Assessment of a Novel VEGF Targeted Agent Using Patient-Derived Tumor Tissue Xenograft Models of Colon Carcinoma with Lymphatic and Hepatic Metastases

The lack of appropriate tumor models of primary tumors and corresponding metastases that can reliably predict for response to anticancer agents remains a major deficiency in the clinical practice of cancer therapy. It was the aim of our study to establish patient-derived tumor tissue (PDTT) xenograft models of colon carcinoma with lymphatic and hepatic metastases useful for testing of novel molecularly targeted agents. PDTT of primary colon carcinoma, lymphatic and hepatic metastases were used to create xenograft models. Hematoxylin and eosin staining, immunohistochemical staining, genome-wide gene expression analysis, pyrosequencing, qRT-PCR, and western blotting were used to determine the biological stability of the xenografts during serial transplantation compared with the original tumor tissues. Early passages of the PDTT xenograft models of primary colon carcinoma, lymphatic and hepatic metastases revealed a high degree of similarity with the original clinical tumor samples with regard to histology, immunohistochemistry, genes expression, and mutation status as well as mRNA expression. After we have ascertained that these xenografts models retained similar histopathological features and molecular signatures as the original tumors, drug sensitivities of the xenografts to a novel VEGF targeted agent, FP3 was evaluated. In this study, PDTT xenograft models of colon carcinoma with lymphatic and hepatic metastasis have been successfully established. They provide appropriate models for testing of novel molecularly targeted agents.


Introduction
Animal models have been used in front-line preclinical studies for predicting efficacy and possible toxicities of anticancer drugs in cancer patients [1]. Advancing a laboratory candidate drug from preclinical testing into testing in phase II clinical trials is based on the assumption that cancer models used in the laboratory are clinically predictive [2]. One of the most serious obstacles confronting investigators involved in the development and assessment of new anticancer drugs is the failure of rodent tumor models to predict reliably as to whether a given drug will have prospective anticancer activity with acceptable toxicity when applied to humans. Current tumor models used for drug evaluation generally consist of implantation into immunodeficient mice of xenografts generated from well-established human cancer cell lines that have already adapted to in vitro growth. These models have been used extensively for decades for rapid screening of the anticancer drug efficacy [3,4]. Such models have proven useful for identifying cellular and molecular mechanisms underlying metastasis and for developing new therapeutics. However, limited effectiveness exists, which severely restrains the predictive power of such models assessing the responses of patients' tumors to anticancer drugs in the clinic. The highly anaplastic cancer cells cultivated in vitro represent the extreme derivates from highly advanced cancers and are not associated with original tumor stroma, which now has been recognized as a crucial factor in the pathogenesis of cancer metastasis. In recent years, various groups have initiated the development of more relevant models based on xenografting of primary human tumor tissue in immunodeficient mice. Such patient-derived tumor tissue (PDTT) xenograft models are mainly constructed by introducing advanced tumor cells into the subcutaneous graft site. These xenografts models retain similar morphology, architecture and molecular signatures as the original cancers and thus should be used for rapid screening of potential therapeutics.
In recent years, many studies have focused on the heterogeneity found in primary tumors and corresponding metastases with the consideration that evaluation of metastatic rather than primary sites could be of clinical relevance [5]. Numerous reports have evaluated the heterogeneity in primary tumors and corresponding metastases in a range of solid tumors such as breast cancer [6,7,8,9,10,11,12,13], colorectal cancer [14,15,16,17] and nonsmall cell lung cancer (NSCLC) [18,19]. The main purpose of investigating the heterogeneity found in primary tumors and corresponding metastases is to evaluate the effect of such heterogeneity on the efficacy of anticancer therapy and cancer patients' prognosis. The primary tumor and its corresponding metastases are different at the molecular marker expression or gene status levels and that these differences may affect the clinical outcome of anticancer therapy [20]. Monaco et al. suggested that the EGFR and KRAS status of primary lung carcinomas might not predict the status in the corresponding metastases. Their observation may have important implications for molecular testing for EGFR-targeted therapies [21]. A retrospective study investigated the role of PTEN loss, Akt phosphorylation and KRAS mutations in primary colorectal tumors and their corresponding metastases on the activity of cetuximab plus irinotecan [22]. This study gave us direct evidence to reveal that the genetic heterogeneity in primary colorectal tumors and their corresponding metastases have different responses to EGFR-targeted therapy. On these considerations mentioned above, to establish the PDTT xenograft models of primary tumor and corresponding metastases for assessment of the response of both the primary tumor and the metastases to some novel drugs is extremely useful.
FP3 (also named as KH902 or KH903) is an engineered protein which contains the extracellular domain 2 of VEGF receptor 1 (Flt-1) and extracellular domain 3 and 4 of VEGF receptor 2 (Flk-1, KDR) fused to the Fc portion of human immunoglobulin G1 [23,24]. Previous studies indicated that FP3 had promise as a local antiangiogenic treatment of human CNV (choroidal neovascularization) -related AMD (age-related macular degeneration) [23,25,26,27]. In subsequent studies, it was demonstrated that FP3 has an inhibitory efficacy in VEGF-mediated proliferation and migration of human umbilical vein endothelial cells, and in VEGF-mediated vessel sprouting of rat aortic ring in vitro [24]. It was also demonstrated that FP3 has an antitumor effect in a NSCLC cell line (A549) xenograft model [24] and a PDTT xenograft model of gastric carcinoma [28] in nude mice.
It was the aim of our study to establish PDTT xenograft models of colon carcinoma with lymphatic and hepatic metastasis useful for testing of a novel molecularly targeted agent, FP3.

Patient and tissue samples
Tumor specimens were obtained at initial surgery from a 40year-old female colon carcinoma patient with lymphatic and hepatic metastases. Prior written informed consent was obtained from the patient and the study received ethics board approval at First Affiliated Hospital, College of Medicine, Zhejiang University. The patient had not received chemotherapy or radiation therapy before surgery. The histological type was determined according to WHO criteria. The tumor was diagnosed as mucinous adenocarcinoma (T3N2M1). The tumor samples of colon carcinoma with lymphatic and hepatic metastases were put into medium immediately after surgical resection under sterile conditions and transported without delay to the animal facility.

Establishment of xenografts
Four-to six-week-old female BALB/c nude mice purchased from Slaccas (Slaccas Laboratory Animal, Shanghai, China) were housed in a barrier facility and acclimated to 12-h light-dark cycles for at least three days before use. The use of experimental animals adhered to the ''Principles of Laboratory Animal Care'' (NIH publication #85-23, revised in 1985). All experiments were approved by the Institutional Animal Care and Use Committee of Zhejiang University (approval ID: SYXK(ZHE)2005-0072). The method to establish the PDTT xenograft models of human colon carcinoma with lymphatic and hepatic metastases were followed as described previously [5,28,29,30]. Briefly, tumors were placed in RPMI 1640 medium supplemented with 20% FBS and 0.05% penicillin/streptomycin solution in an ice bath in the surgical site. Tumors were then transferred to a sterile Petri dish containing RPMI 1640 medium with supplements mentioned above. Thin slices of tumor were diced into 26263 mm 3 pieces and washed thrice with RPMI 1640 with supplements mentioned above. Under anesthesia with isofluorane, tumors were implanted into 4-to 6-week-old female athymic nude mice by a small incision and subcutaneous pocket made in one side of the lower back in which one tumor piece is deposited in the pocket. While the pocket was still open, one drop of 1006 penicillin/streptomycin solution was placed into the opening. Half of the rest of the tumors was cryopreserved in liquid nitrogen and the other half was immediately snap-frozen and stored at 280uC and processed for biological studies, such as genetic, genomic, mRNA expression and protein expression analyses. For each tumor, twenty mice were used. Growth of established tumor xenografts was monitored at least twice weekly by vernier caliper measurement of the length (a) and width (b) of tumor. Tumor volumes was calculated as (a6b 2 )/2. For the first week following implantation, a small bump would be visible where the tumor was inserted. At 14 to 22 weeks following implantation, a tumor began to appear at the site of implantation with 1000 to 1500 mm 3 in volume. At a size of about 1500 mm 3 , tumors were removed for serial transplantation. Tumor-bearing animals were anesthetized with diethyl ether and sacrificed by cervical dislocation. Animals were placed immedi-ately in an ice water bath for 2 minutes. The mice were then placed in 75% ethanol for 2 minutes, and transferred to a laminar flow hood for dissection. Tumors were minced under sterile conditions and implanted in successive nude mice as described above. Tumors were passaged no more than 10 times. Following transplantation, tumors were allowed to grow to 200 to 500 mm 3 before initiation of treatment for drug evaluation. Numerous samples from early passages were stored in the tissue bank and cryopreserved in liquid nitrogen, and used for further experiments.

Treatment protocol
Xenografts from this second mouse-to-mouse passage (the third generation, G3) were allowed to grow to a size of 200 mm 3 , at which time mice were randomized in the following three groups of treatment, with 10 mice in each group: (a) control (100 ml saline); (b) FP3, 15 mg/kg, i.v., twice per week; (c) Avastin (bevacizumab), 10 mg/kg, i.v., twice per week. Mice were treated during 21 days, monitored twice per week for signs of toxicity, and were weighed once a week. Tumor size was evaluated twice a week by caliper measurements using the following formula: tumor volume = (length6width 2 )/2. Relative tumor growth inhibition (TGI) was calculated by relative tumor growth of treated mice divided by relative tumor growth of control mice (T/C). Experiments were terminated on day 30.

Histology and immunohistochemistry
Selected tumor specimen were fixed in 10% neutral-buffered formalin and embedded in paraffin. Five-micrometer sections were cut, dewaxed, and then rehydrated and stained with hematoxylin and eosin (H&E) as described previously [31,32].
For immunohistochemical staining, five micromolar sections were cut, dewaxed, rehydrated, and subjected to antigen retrieval. After blocking endogenous peroxidase activity, the sections were incubated with the primary antibodies against CK-20 (1:100), CDX-2 (1:100), EGFR (1:100), VEGF (1:100) and PCNA (1:100) (overnight at 4uC). Immunohistochemistry was performed using the streptavidin-biotin peroxidase complex method (Lab Vision, Fremont, CA). The slides were examined and pictures were taken using an Olympus BX60 (Olympus, Japan). Sections known to stain positively were incubated in each batch and negative controls were also prepared by replacing the primary antibody with preimmune sera.

Fluorescent immunohistochemistry
Selected mice with similar tumor size were anesthetized with ketamine (87 mg/kg) plus xylazine (13 mg/kg) injected intramuscularly. The chest was opened rapidly, and the vasculature was perfused for 3 minutes at a pressure of 120 mmHg with fixative [4% paraformaldehyde in 0.1 mol/L phosphate-buffered saline (PBS), pH 7.4] from an 18-gauge cannula inserted into the aorta via an incision in the left ventricle. Blood and fixative exited through an opening in the right atrium. After the perfusion, the implanted tumor was removed and placed into fixative for 2 hours at 4uC. Specimens were then rinsed several times with PBS, infiltrated overnight with 30% sucrose, embedded in OCT medium and frozen for cryostat sectioning [33]. Cryostat sections 8 to 10 mm in thickness were brought to room temperature, air dried overnight, then fixed in acetone for 10 min. Slides were allowed to air dry for 30 min and were washed three times for 5 min each in PBS. Samples were then incubated in 5% BSA in PBS for 30 min at room temperature to block nonspecific antibody binding. Next, the sections were incubated with two primary antibodies (CD31, 1:100; and a-SMA, 1:200) or VEGFR-2 (1:150) overnight at room temper-ature in humidified chambers diluted in PBS. After several rinses with PBS, specimens were incubated for 1 hour at room temperature with fluorescent (Cy3-or FITC-conjuncted) secondary antibodies (goat anti-rat or goat anti-rabbit) diluted (1:200) in PBS. Specimens were rinsed again with PBS, and mounted in Vectashield (Vector Laboratories, Burlingame, CA) [34,35]. Tissue sections were examined and digitally photographed using a Zeiss Axiophot fluorescence microscope (Carl Zeiss, Thornwood, NY) equipped with single, dual, and triple fluorescence filters and a low-light, externally cooled, three-chip charge-coupled device (CCD) camera (4806640 pixel RGBcolor images, CoolCam; SciMeasure Analytical Systems, Atlanta, GA) and saved as TIFF files.
Genome-wide gene expression analysis RNA was extracted from tumor specimens and from 26262 mm 3 tumor xenograft samples (derived from different passages ranging from 1 to 3), which were taken from sacrificed animals. Samples were snap-frozen and stored in liquid nitrogen until use. Total RNA of homogenized tumor samples was prepared with Trizol RNA extraction reagent (Invitrogen) followed by purification using the RNeasy Mini Kit (Qiagen) according to the manufacturer's recommendations. A DNase I (Qiagen) digestion step was included to eliminate genomic DNA. The quality of the total RNA was checked for integrity using RNA LabChips on the Agilent Bioanalyzer 2100 (Agilent Technologies) and the concentration was measured on the Peqlab NanoDrop. Only RNA with a RNA integrity number larger than 6.5 was used for cDNA synthesis. The one-cycle eukaryotic target labeling assay from Affymetrix was used according to manufacturer's instructions and as previously described [36]. Data analysis was performed as previously described [36].

DNA extraction and mutation analysis
DNA was extracted from paraffin-embedded samples of colon carcinoma with lymphatic and hepatic metastases. For every tumor tissue, 10-mm sections were prepared, and an additional representative 2-mm section was deparaffinised, stained with haematoxylin and eosin, and analysed for detailed morphology. Regions of tumor tissue were marked, and this tissue was extracted with 0.2 M sodium hydroxide in 1 mM edetic acid and neutralised with 100 mM TRIS-TE (pH 6.5). After extraction, DNA was purified with Qiagen PCR purification kit (Qiagen, Hilden, Germany). KRAS gene in exon 1 was analysed at codons 12 and 13 with pyrosequencing using a previously described assay which has been shown to be of greater sensitivity [37]. BRAF gene in exon 15 at codon 600 and PIK3CA gene in exon 9 at codons 539, 542, 545 and 546 and exon 20 at codons 1043, 1044, 1047 and 1049 were analysed with pyrosequencing as previously described [14].

qRT-PCR
Total RNA was extracted from frozen tissues using TRIzol reagent according to the protocol provided by the manufacturer (Invitrogen, Carlsbad, CA). Total RNA was reverse-transcribed into single-strand complementary DNA (cDNA) using moloneymurine leukemia virus (M-MLV) reverse transcriptase (Promega, Madison, WI). Briefly, the RNA was denatured by heating for 5 min at 70uC, followed by rapid cooling on ice, and then used for reverse transcription (2 mg of total RNA, 25 U of RNase inhibitor, 0.5 mmol/L of each dNTP, 1.5 mmol/L reverse primer and 200 U of M-MLV reverse transcriptase in a total volume of 25 ml). For reverse transcription, tubes were incubated at 42uC for 60 min. The expression of a randomly selected set of genes, including dehydropyrimidine deshydrogenase (DPD), nucleotide excision repair-1 (ERCC-1), and thymidylate synthase (TS) in tissues in original tumors (G0) and tumors from the third generation (G3) was analyzed using a fluorescence-based realtime detection method (ABI PRISM 7700 Sequence Detection System [TaqMan]; Perkin-Elmer Applied Biosystems) as previously described [38,39]. Specific primer pairs and probes are listed in Table 1. The b-actin gene was used as an endogenous control for normalization. The qRT-PCR reaction was carried out in triplicate for each sample. The 25 ml PCR mixture was consisted of 1 ml of cDNA template, 1 ml each of sense and antisense primers, 0.75 ml of 59 FAM-and 39 TAMARA-labeled oligonucleotide probe, 2 ml of dNTP mixture, 5 ml of 56 reaction buffer, and 0.125 ml of Taq DNA polymerase. Cycling conditions were 50uC for 2 minutes, and 95uC for 10 minutes, followed by 46 cycles at 95uC for 15 seconds and 60uC for 1 minute.

Western blotting
Protein expression profiles were analyzed by western blotting as previously described [31,40,41]. Briefly, lysates for immunoblotting were prepared by adding lysis buffer [50 mM Tris-HCl (pH 7.4), 1% Nonidet P-40, 0.5% sodium deoxycholate, 150 mM NaCl, 0.02% sodium azide, and 0.1% SDS] containing protease and phosphatase inhibitors (Sigma, St. Louis, MO) to the tumor tissue homogenized in fluid nitrogen. After centrifugation at 15,000 rpm at 4uC for 10 min, the supernatants were collected, and the protein concentration was determined using Bio-Rad Protein Assay Kit (Bio-Rad, Hercules, CA). Protein extracts of tumor lysates (30 mg) were added to a loading buffer [10 mmol/L Tris-HCl (pH 6.8), 1% SDS, 25% glycerol, 0.1 mmol/L mercaptoethanol, and 0.03% bromophenol blue], boiled, and separated on 8% to 12% (w/v) polyacrylamide gels in the presence of SDS. Molecular weights of the immunoreactive proteins were estimated based on the relative migration with colored molecular weight protein markers (Amersham Pharmacia Biotech, Piscataway, NJ). After electrophoresis, the protein blots were electro-transferred to PVDF membranes (Millipore, Billerica, MA). Then, the membranes were blocked at room temperature with 5% nonfat milk in TBS [10 mmol/L Tris-HCl (pH 7.5), 0.5 mol/L NaCl, and 0.05% (v/v) Tween 20] buffer for 1 h. The primary antibodies were diluted at 1:1,000 and the membranes were incubated with primary antibodies overnight at 4uC. The antibodies tested were anti-Akt, anti-ERK, anti-MAPK, anti-mTOR antibodies, anti-EGFR, anti-VEGF, anti-cleaved-caspase-3, anti-PCNA, and phosphorylation-specific antibodies against Akt (Ser 308 and Ser 473 ), ERK (Thr 202 /Tyr 204 ), MAPK (Thr 180 /Tyr 182 ) and mTOR (Ser 2448 ). The next day, the membranes were washed and incubated for 1 h at room temperature with rabbit immunoglobulin G-horseradish peroxidase-conjugated secondary antibodies (Santa Cruz Biotechnology), at a final dilution of 1:5,000. After washing thrice with TBS, antibody binding was visualized using enhanced chemiluminescence detection system (SuperSignal West Pico, Pierce) as described by the manufacturer and autoradiography. To show equal protein loading, the blots were stripped and reprobed for GAPDH.

Statistical analysis
Hierarchical clustering of all microarray experiments was done based on all probe sets (54,675 probe sets) represented on the   HGU133Plus2.0 array (Affymetrix) with a quality p,0.04 using positive correlation and complete linkage. Gene expression of primary tumors was compared with the median arrays of replicate tumors from each xenograft model in a paired t test. Drug sensitivity data are presented as mean 6 SEM and analyzed by SPSS 16.0 software. Difference among mean of the groups is determined with one-way ANOVA. Comparison is considered to be statistically significant if p,0.05.

Patient-derived tumor tissues of colon carcinoma and its corresponding lymphatic and hepatic metastases can be implanted efficiently into nude mice
To test whether patient-derived tumor tissues of primary colon carcinoma and its corresponding lymphatic and hepatic metastases can be engrafted in nude mice, we implanted small pieces of freshly tumor tissue into female athymic nude mice subcutaneously. After two to five months, tumors began to appear at the site of implantation with 1000 to 1500 mm 3 in volume, and xenografts were harvested for serial transplantation. The tumor-bearing rate and tumor growth rate of different generations of PDTT xenografts of primary colon carcinoma and its corresponding lymphatic and hepatic metastases were illustrated in Table 2.

Histological and molecular characterization of xenografts and comparison with original tumors
Using PDTT xenografts as models for preclinical anticancer drug development is based on the assumption that the xenografts would closely resemble the corresponding original tumors. For this purpose, H&E staining, immunohistochemical staining, genomewide gene expression analysis, pyrosequencing, qRT-PCR and western blotting were used to determine the biological stability of the xenograft during serial transplantation compared with the original tumor tissues. In this study, CK-20 and CDX-2 were used as markers for determining the lymphatic and hepatic metastases with a colon carcinoma primary. Immunohistochemical staining showed a positive expression of CK-20 and CDX-2 in the metastases (Fig. 1), which ascertained lymphatic and hepatic metastases with an epithelial origin. Histological examination of the H&E sections showed that the PDTT xenografts were adenocarcinoma with features similar to the original surgical specimens. Fig. 2 shows the morphology of the original tumors (G0) of primary colon carcinoma with lymphatic and hepatic metastases and their third generation (G3) implants in nude mice. There were no significant morphological differences between the tumor resected from the patient and the initial successful implants. The explanted tumor pieces also showed similar VEGF and EGFR expressions compared to the corresponding original tumors (Fig. 3,4).
When using the PDTT xenograft models for testing of the anticancer agents, especially molecularly targeted agents, detailed characterization of the molecules is indispensable. For this purpose, GeneChip HGU133Plus2.0 expression arrays (Affymetrix) was used to determine the status of the genome-wide gene expression of the xenograft models, pyrosequencing was used to determine the mutation status of some selected genes of the  xenograft models, qRT-PCR was used to determine the status of some randomized selected genes of the xenograft models, western blotting was used to determine the expression of some randomized selected signaling pathway proteins of the xenograft models during serial transplantation compared with the original tumor tissues. Our results showed that early passages of the PDTT xenograft models of primary colon carcinoma with lymphatic and hepatic metastases revealed a high degree of similarity with the original clinical tumor samples with regard to KRAS, BRAF, EGFR, and PIK3CA mutation status (see Fig. S1, S2, S3, S4) as well as mRNA expression (data not shown). With regards to the western blotting analysis of 'randomized selected signaling proteins', early passages of the PDTT xenograft models of colon carcinoma lymphatic and hepatic metastases revealed a high degree of similarity with the original clinical tumor samples though these phenomena did not exist between original clinical tumor sample of primary colon cancer and its xenografts (Fig. 5). Based on the results of genome-wide gene expression analysis, we calculated the correlation coefficient between primary cancer and the xenograft derived thereof. The correlation coefficient ranged between 0.988 and 0.991 (0.991 between primary colon carcinoma and its xenograft, 0.989 between lymphatic metastasis and its xenograft, and 0.988 between hepatic metastasis and its xenograft) indicating a high degree of similarity between the primary cancer and the corresponding xenograft model. A paired t test between the primary tumors and xenografts revealed limited differentially expressed probe sets. Clustering based on these probe sets showed a limited distinction between primary tumors and xenografts (Table S1, S2, S3, Fig. 6).
PDTT xenograft models of colon carcinoma and its corresponding lymphatic and hepatic metastases for assessment of a novel VEGF targeted agent Our results indicated that during sequential passage, the PDTT xenograft models retained their similarity to the corresponding  original tumors in morphology, architecture and molecular signatures thus could be used for rapid screening of potential therapeutics. To improve confidence of a novel VEGF blocker during its early preclinical studies allowing for a faster decision is one of purposes of using these xenograft models. To explore the value of FP3 as an antitumor therapeutic and to compare it to other effective agents targeting the VEGF pathway, we evaluated its ability to block the growth of these xenograft models.
After implantation, tumors were allowed to grow for 10 days, forming large retroperitoneal tumors .100 mm 3 . Injections of FP3 (15 mg/kg body weight), Avastin (10 mg/kg body weight) or saline were then given i.v. twice per week for 21 days, after which the animals were killed and tumors excised and measured. No body weight-related toxicity was found in each group. FP3 significantly inhibited the growth of xenografts of primary colon carcinoma and its lymphatic and hepatic metastases in nude mice ( Fig. 7), and resembled the well-defined and generally accepted antitumor activity of Avastin [42].
To evaluate the effects of FP3 on tumor-associated angiogenesis, selected tumors from the above studies were sectioned and immunostained with antibodies to CD31 and a-SMA, so that the vasculature could be visualized (Fig. 8, 9, 10). This analysis revealed that vasculature was nearly absent in FP3-treated xenografts. FP3 (treatment for 21 days) almost completely blocked tumor-associated angiogenesis, with the stunted tumors being largely avascular (Fig. 8: B, E, and H; Fig. 9: B, E, and H; Fig. 10: B, E, and H). In contrast to the FP3-treated tumors, control tumors in saline-treated mice not only were much larger but also had a very high vascular density (Fig. 8: A, D, and G; Fig. 9: A, D, and G; Fig. 10: A, D, and G). These results indicate that FP3 administration reduces xenograft size and concurrently causes decreased microvessel growth.  We next measured cell proliferation in the treated tumors. By immunohistochemical staining, we found that VEGF expression (Fig. 11) and PCNA expression (Fig. 12) in FP3-and Avastintreated tumors were significantly suppressed. However, EGFR expression was not significantly different between any of the treatment groups and saline-treated controls (Fig. 13), suggesting that levels of EGFR are unlikely to be altered by the treatments.
Because inhibition of VEGF signaling can decrease VEGFR-2 expression in certain types of blood vessels [43,44], we asked whether VEGF/VEGFR blockade decreased receptor expression in our tumor models. Treatment of the xenograft models with FP3 for 21 days, decreased expression of VEGFR-2, a marker for growing vasculature (Fig. 14, 15). These results were consistent with the disappearance of endothelial cells expressing this receptor (Fig. 8, 9, 10).

Discussion
Rodent tumor models currently being used and which include transgenic tumor models, and those generated by planting human tumor cell lines subcutaneously in immunodeficient mice, do not sufficiently represent clinical cancer characteristics, especially with regard to metastasis and drug sensitivity [29]. The increasingly used PDTT xenografts models implanted subcutaneously or in subrenal capsule in immunodeficient mice provide a more accurate reflection of human tumor biological characteristics than tumor cell lines. The ability to passage patients' fresh tumor tissues  into large numbers of immunodeficient mice provides possibilities for better preclinical testing of new therapies for the treatment and better outcome for cancer.
In our study, we successfully established the PDTT xenograft models of primary colon carcinoma and its corresponding lymphatic and hepatic metastases. One of vital questions relating  to PDTT xenografts is whether tumor passage in the experimental mice alters the phenotype of primary cancer cells. When developing the anticancer agents, especially molecularly targeted agents, detailed characterisation of the molecules is indispensable.
If there are fundamental changes in tumors from before to after their engrafting, the model may not do well in reflecting the features of human cancers. Therefore, the practicability of this model as a screening platform for new drug development for this  disease will be limited. Similarly, the value of this model as a tool to individualise patient treatment would be limited [40]. Using PDTT xenografts as models for preclinical anticancer drug development is based on the assumption that the xenografts would closely resemble the corresponding original tumors. In our study, we firstly demonstrated that early passages of the PDTT xenograft models of primary colon carcinoma, lymphatic and hepatic metastases revealed a high degree of similarity with the  original clinical tumor sample with regard to histology (Fig. 2), immunohistochemistry (Fig. 3, 4), mutation status (see Fig. S1, S2, S3, S4), mRNA expression (data not shown) as well as genes expression (Fig. 6). As these xenografts models retain similar histopathological features and molecular signatures as the original tumors, they could be used for rapid screening of potential therapeutics not only on the primary tumor, but also on its corresponding metastases.
One of the general applications of the PDTT xenograft models of primary colon carcinoma and its corresponding lymphatic and hepatic metastases is that they can be used as an in vivo screening tool to test the response of both the primary tumor and the metastases to some certain novel drugs. Furthermore, these models may help us to clear whether the primary tumors and corresponding metastases have different response to the same anticancer drugs. In this study, the PDTT xenograft models of primary colon carcinoma and its corresponding lymphatic and hepatic metastases were used for assessment of a novel VEGF targeted agent, FP3. As shown in results, FP3 significantly inhibited the growth of xenografts of primary colon carcinoma and its corresponding lymphatic and hepatic metastases and concurrently caused inhibition of tumor vessels growth dramatically [ Fig. 7, 8, 9, 10]. These results coincide with our previous works [24,28].
In conclusion, our work successfully established the PDTT xenograft models of colon carcinoma with lymphatic and hepatic metastasis useful for testing of a novel molecularly anti-VEGF targeted agent, FP3. Early passages of the PDTT xenograft models of primary colon carcinoma, lymphatic and hepatic metastases revealed a high degree of similarity with the original clinical tumor samples. They provide appropriate models for testing of novel molecularly targeted agents. These models may also help us to clear whether the primary tumors and corresponding metastases have different response to the same anticancer drugs. Figure S1 Mutation analysis by pyrosequencing. Sequence data of KRAS gene in exon 1 at codons 12 and 13 of primary tumor tissues and their early-generation xenograft models tumor tissues. A, standard wild-type sample. Tumor tissues from surgical specimens of primary colon carcinoma (B) and its corresponding lymphatic metastasis (D) and hepatic metastasis (F). Tumor tissues from the third generation of xenograft models of primary colon carcinoma (C) and its corresponding lymphatic metastasis (E) and hepatic metastasis (G). (TIF) Figure S2 Mutation analysis by pyrosequencing. Sequence data of BRAF gene in exon 15 at codon 600 of primary tumor tissues and their early-generation xenograft models tumor tissues. A, standard wild-type sample. Tumor tissues from surgical specimens of primary colon carcinoma (B) and its corresponding lymphatic metastasis (D) and hepatic metastasis (F). Tumor tissues from the third generation of xenograft models of primary colon carcinoma (C) and its corresponding lymphatic metastasis (E) and hepatic metastasis (G). (TIF) Figure S3 Mutation analysis by pyrosequencing. Sequence data of PIK3CA gene in exon 9 at codons 539, 542, 545 and 546 of primary tumor tissues and their early-generation xenograft models tumor tissues. A, standard wild-type sample. Tumor tissues from surgical specimens of primary colon carcinoma (B) and its corresponding lymphatic metastasis (D) and hepatic metastasis (F). Tumor tissues from the third generation of xenograft models of primary colon carcinoma (C) and its corresponding lymphatic metastasis (E) and hepatic metastasis (G). (TIF) Tumor tissues from surgical specimens of primary colon carcinoma (B) and its corresponding lymphatic metastasis (D) and hepatic metastasis (F). Tumor tissues from the third generation of xenograft models of primary colon carcinoma (C) and its corresponding lymphatic metastasis (E) and hepatic metastasis (G).

(TIF)
Table S1 Genes differentially expressed in patient primary colon cancer specimens and its xenograft.