Antibody Library-based Tumor Endothelial Cells Surface Proteomic Functional Screen Reveals Migration-stimulating Factor as an Anti-angiogenic Target^,

Abstract

Angiogenesis is critical for cancer development and metastasis. Here we have employed a functional antibody library-based proteomic screen to identify proteins that participate in and might be used as therapeutic targets for tumor-related angiogenesis. Mice were immunized with human esophageal cancer endothelial cells (HECEC). The antibody library was established with the mouse spleen cells the serum of which had most anti-angiogenic effect. Monoclonal antibodies were subjected to an immunoreactive and functional screen and monoclonal antibodies that reacted strongly with cell surface antigens of HECECs and influenced their behavior were selected. Antigens that recognized by the antibodies were obtained by immunoprecipitation and then identified by mass spectrometry analysis. Migration-stimulating factor (MSF), the antigen of 1D2 antibody was identified using this approach. Further studies demonstrated that the 1D2 antibody suppressed MSF-effected migration and adhesion of HECECs on fibronectin matrix. Biodistribution assay showed that MSF targeting antibody 1D2 could specifically home to the xenograft with humanized blood vessel. Targeting treatment with 1D2 antibody significantly suppressed tumor growth through inhibition of human tumor-related angiogenesis. These results indicate that the functional antibody library-based proteomic screen can successfully identify proteins that involved in tumor-related angiogenesis and MSF may be a target for the anti-angiogenic treatment of the esophageal cancer.

Selective in vivo targeting of a single organ or diseased tissues such as a solid tumor remains a desirable but elusive goal for molecular medicine (1,2). Such targeting would permit more effective imaging and provide new modes of drug and gene therapies for many acquired and genetic diseases. Folkman (3) reported that tumor growth is angiogenesis-dependent, which leads to the development of anti-angiogenic therapy. Studies have shown that the tumor vasculature is highly specialized. A global survey of mRNA expression by the serial analysis of gene expression has revealed many striking differences between endothelial cells isolated from human colon cancers and those from adjacent normal tissues (4). A recent study has also disclosed differential gene expression profile of endothelial cells in malignant breast cancer tissue compared with normal tissue (5). These dysregulated genes may be candidate biomarkers of tumor-related angiogenesis and potential targets for the development of antiangiogenic drugs. Annexin I is such a protein identified by subtractive proteomic mapping, which shows to be specifically expressed in breast tumor endothelium and might enable tumor targeting for human breast malignancy (6).

Although anti-angiogenic therapy is conceptually highly appealing for tumor treatment, few probes directing to the native endothelial cell surface proteins show validated tissue and function-specific pharmacodelivery in vivo (7–9). This may be due in part to difficulties in isolating the endothelium from tumor tissues. Additionally, tumor endothelial cells rapidly lose their tumor-specific properties in vitro because these properties are regulated by signals derived from the local tissue microenvironment that cannot be duplicated in vitro (10,11). Moreover, it is labor-intensive and time-consuming work to identify targets from membrane proteins of tumor endothelium and further validate their clinical application in tumor targeting therapies.

In the present study, we developed and used a functional proteomic screen to identify and validate targets that enable tumor anti-angiogenic therapy. Human esophageal cancer endothelium was isolated from carcinoma tissues, and their tumor-specific properties were maintained by co-culturing with tumor cells. Subsequently, a functional monoclonal antibody library was established by immunizing mice with tumor endothelial cells. Antibodies that specifically recognized surface proteins of tumor endothelium were selected from the library, and their antigens were identified by immunoprecipitation and mass spectrometry. Using this strategy we identified MSF1, an isoform of fibronectin, as a tumor vascular target for anti-angiogenic therapy.

EXPERIMENTAL PROCEDURES

Tissue Specimens—

Fresh tissues of esophageal carcinomas and matched histologically normal tissues were procured from surgical resection specimens collected by the Department of Pathology in Cancer Hospital, Chinese Academy of Medical Sciences, Beijing, China. Primary tumor regions and the corresponding histological normal tissues from the same patients were separated by experienced pathologists and immediately stored at −70 °C until use. Patients did not receive any treatment prior to surgery and signed informed consent forms for sample collection.

Isolation of Endothelial Cells from Human Tissues—

Human esophageal cancer endothelial cells (HECECs) were isolated from esophageal squamous cell carcinoma tissues (12,13). Briefly, to isolate pure endothelial cell population from human esophageal squamous cell carcinoma, cancerous tissues were obtained <30 min after surgical removal. The tissues were washed with bovine serum albumin (Sigma)/Hanks (Invitrogen) and cut into slices. Following incubation with collagenase for 2 h at 37 °C, cells were filtered sequentially through 400 μm, then 100-μm meshes, and centrifuged for 15 min at 800 × g in 25% Percoll (Sigma)/D-MEM (Invitrogen). Cells were harvested, then incubated within D-MEM, 15% FBS (Hyclone Labs, Logan, UT), and 100 μg/ml ECGS (endothelial cells growth supplement; Sigma) in dishes coated with 2% gelatin (Sigma). After 2–3 days, the cells were approximately 50% confluent. Magnetic beads (Miltenyi Biotec, Gmbh, Bergisch Gladbach, Germany) coupled with anti-CD31 (Endogen, Woburn, MA) were added to bind to the endothelial cells. Cells were washed three times with D-MEM to remove excess beads. Following treatment with 10 mm EDTA/0.1% trypsin (Invitrogen), cells bound to magnetic beads were subfractionated by magnetic attraction, then washed with D-MEM, and further subfractionated by magnetic attraction. The subfraction by magnetic attraction was repeated three times to purify the HECECs. Finally, endothelial cells were incubated in dishes coated with 2% gelatin (Sigma) in D-MEM supplemented with 10% FBS and 100 μg/ml ECGS at 37 °C under an atmosphere containing 5% CO₂. Human esophageal normal endothelial cells (HENECs) were similarly isolated from human esophageal tissues using the procedure described above.

Cell Culture—

KYSE150, a human esophageal squamous carcinoma cell line, was maintained in MEM medium (Invitrogen) containing 10% FBS. Human liver sinusoidal endothelial cells (LSECs) were maintained in D-MEM supplemented with 10% FBS (14). HECECs, HENECs, and human umbilical vein endothelial cells (HUVECs; CASCADE C-003–5C) were cultured in dishes coated with a thin layer of 2% gelatin (Sigma) and were maintained in endothelial medium comprising D-MEM supplemented with 10% FBS and 100 μg/ml ECGS. Before commencing experiments, the HENECs, HUVECs, and LSECs were cultured to complete confluence and maintained with D-MEM supplemented with 10% FBS for 3 days for reversion to quiescence.

For co-cultivation of HECECs with KYSE150, 1 × 10⁴ HECEC with 1 × 10⁴ KYSE150s were mixed and co-cultured in 6-well plates for 7 days. 1 × 10⁴ HECECs were cultured separately with and without tumor-conditioned medium of KYSE150s, respectively, as controls. These cells were all cultured in plates coated with a thin layer of 2% gelatin (Sigma) and were maintained in endothelial medium containing D-MEM (Invitrogen) supplemented with 10% FBS and 100 μg/ml ECGS.

Immunocytochemistry—

Endothelial cells cultured in 24-well plates were washed three times with PBS. After fixation with acetone:methanol (50:50 v/v) for 15 min, the cells were washed with PBS as before. Endogenous peroxidase activity was inactivated using 3% H₂O₂ in PBS, then pre-blocked with 10% goat serum for 30 min at 37 °C. Cells were subsequently incubated for 45 min at 4 °C with primary antibodies: anti-VEGFR1 (Sigma) at 1:100 dilution; anti-VEGFR2 (Sigma) at 1:200 dilution; and anti-integrin β3 (Chemicon, Temecula, CA) at 1:200 dilution. Then biotin-labeled second antibody was used following three times washing with PBS. Standard avidin-biotin complex peroxidase/3′,3-diaminobenzidine (Dako, Carpinteria, CA) staining was performed to visualize the biomarkers.

Cell Count and Proliferation Assay—

For co-cultured and monoculture model, cells were digested with trypsin and resuspended in PBS; the total cell number was counted using a microscope at ×100 magnification. The cells in co-culture model were fixed and labeled with anti-FactorVIII antibody (Dako) followed by fluorescein isothiocyanate-labeled secondary antibody. These cells were analyzed by fluorescence-activated cell sorter to calculate the ratio of endothelial cells in co-culture model by three independent methods. Monocultured KYSE150 cells were used as control for fluorescence-activated cell sorter analysis. The endothelial cell number in co-culture model was calculated by multiplying total cell number with endothelial cell ratio. For cell proliferation assays, cells were counted with CCK-8 at the indicated days according to the manufacturer's instructions and measured at 450/655 nm: these experiments were performed in triplicate and repeated twice.

Monoclonal Antibody Library Construction and Immunocytochemical Screen—

A library of monoclonal antibodies was generated from mice immunized with fixed HECEC cells using established procedures (15). The antibody subtypes were identified using the Clonotyping system (SouthernBiotech). Antigen immunoreaction was performed to scan for antibodies capable of reacting with endothelium. Briefly, endothelial cells were fixed with paraformaldehyde, then incubated with hybridoma supernatant. Immunocytochemistry was performed as described above.

Cell Surface Immunofluorescent Analysis—

Cells cultured on slides were incubated with a panel of antibodies, including monoclonal antibodies from the library (10 μg/ml) and rabbit anti-α-tubulin antibody (5 μg/ml; Sigma) and detected with either fluorescein isothiocyanate-goat anti-mouse IgG and Cy3-goat anti-rabbit IgG (Jackson ImmunoResearch), respectively. Finally, cells were mounted with Vectashield fluorescence mounting medium containing 4′,6-diamidino-2-phenylindole and subjected to immunofluorescent analysis (Nikon).

Immunohistochemistry—

Immunohistochemical analysis was performed with esophageal cancer tissues and serial sections from xenograft tumors. Standard avidin-biotin complex peroxidase immunohistochemical staining was performed following primary antibodies reaction, including each clone of monoclonal antibodies (10 μg/ml), anti-CD31 (reacts with both human and mouse CD31; Abcam, Cambridge, UK) at 1:200 dilution, anti-huCD31 (reacts only with human CD31′, Zymed Laboratories Inc.) at 1:100 dilution.

Western blot Analysis—

For Western blot analysis, proteins of HECEC were extracted by radioimmune precipitation assay buffer (50 mm Tris, pH 7.4, 150 mm NaCl, 1% Nonidet P-40, 1% Triton X-100, 0.1% SDS, 0.5% sodium deoxycholate) supplemented with a mixture of protease inhibitors (Sigma). Blots on polyvinylidene difluoride membranes were blocked and then probed with 1D2 antibody (1 mg/ml). The immunoreaction was visualized by super ECL detection reagent (Applygen, Beijing, China) following incubation with horseradish peroxidase-conjugated secondary antibodies.

Immunoprecipitation and Mass Spectrometry—

HECEC cells were lysed by radioimmune precipitation assay buffer with 0.02% (w/v) NaN₃ and a mixture of protease inhibitors. Immunoprecipitation was done essentially as described previously (16) using 1D2 antibody, which was chemically coupled to CNBr-activated-Sepharose 4B (GE Healthcare). Immunoprecipitated proteins were analyzed by SDS-PAGE and silver staining.

Stained bands were excised and subjected to in-gel tryptic digestion. The digested peptides were loaded onto a C18 reversed phase capillary column (50 × 0.2-mm inner diameter; Michrom, Auburn, CA) with buffer A (0.1% formic acid, 2% acetonitrile, 97.9% H₂O, both v/v). The peptides were eluted with 5–40% buffer B (0.1% formic acid, 89.9% acetonitrile, 10% H₂O, both v/v; flow rate, 2 μl/min for 90 min).

Mass spectrometric measurements were performed using an LTQ XL ion trap mass spectrometer (Thermo Fisher, San Jose, CA) equipped with an ADVANCE plug and play nanoelectrospray source (Michrom, Auburn, CA). Operation of the Agilent 1200 capillary pump and the mass spectrometer was fully automated during the entire procedure, using the Xcalibur 2.0.7 control system (Thermo Fisher). Continuous cycles of one full scan (m/z 400 to 2000) were followed by seven data-dependent MS/MS measurements at 35% normalized collision energy. MS/MS measurements were allowed for an enabled exclusion list of 50 m/z values (±1.5 Da) and a maximum time limit of 1 min.

MS/MS spectra were extracted from raw files requiring a minimum of 10 signals with an intensity of at least 1000 units. Extracted MS/MS spectra were automatically assigned to the best-matching peptide sequences using the SEQUEST algorithm and the SEQUEST Browser software package, Bioworks 3.3.1 (Thermo Fisher). SEQUEST searches were performed on a personal computer against an International Protein Index human protein database (v3.51) containing 74,049 entries downloaded as FASTA-formatted sequences from the website of the European Bioinformatics Institute. To increase search speed, the protein database was pre-processed to create a binary database containing all possible tryptic peptides. One missed tryptic cleavage was permitted and a static modification of +57 Da on cysteine residue. The precursor ion and fragment ion mass tolerance was set as 2.0 and 1.0 Da, respectively.

The SEQUEST criteria were set as following: (1) DeltaCn score was at least 0.1; (2) Rsp was 1; and (3) Xcorr ≥ 2.2 for +1 charged peptides; Xcorr ≥ 2.8 for +2-charged peptides; Xcorr ≥ 3.5 for +3-charged peptides.

In Vitro Tube Formation Assay—

96-well plates were coated with 50 μl of ice-cold Matrigel mixed with FN. 2 × 10⁴ HECECs were suspended in 100 μl of culture medium with varying amounts of mAb 1D2 or normal mouse IgG, seeded on the matrix and then cultured at 37 °C for 6 h.

Tumor Model—

Nude mice (BALB/c nu/nu, females, 5-week-old) were purchased from the Jackson Laboratory (Vitalriver, Beijing, China). Tumor model was founded as described previously (12). Briefly, 1 × 10⁶ tumor cells KYSE150s mixed with 4 × 10⁶ HECECs were subcutaneously inoculated on the right back to give ample, well circumscribed, tumors for biodistribution analysis, gamma scintigraphic imaging, and for in vivo tumor therapy by 1D2 antibody.

Biodistribution Analysis—

1D2 antibody or normal mouse antibody was conjugated to ^99mTc (17). Biodistribution analysis was performed as described previously (18). Tumor-bearing mice were anesthetized and injected via the tail vein with ^99mTc-labeled antibody (2 μg IgG, 60 μCi/μg). Mice were anesthetized for thoracotomy, blood sampling by cardiac puncture, and organ removal at different time points for the analysis of biodistribution of the ^99mTc antibody. Tissues were washed with normal saline, freed from adhering tissues, and weighed prior radioactivity measurements by gamma scintillation counting.

Gamma Scintigraphic Imaging—

The tumor-bearing mice were administered 360 μCi (12 μg) of the ^99mTc-labeled antibody through the tail vein. For the control, a tumor-bearing mouse was administered 360 μCi (12 μg) of the ^99mTc-labeled 1D2 antibody and 0.6 mg of cold 1D2 antibody or mIgG (without ^99mTc-labeled). At different time points, the mice were fixed on a board, and imaging was performed using a single photon emission computerized tomography (LC 75-005, Diacam, Siemens) gamma camera.

In Vivo Anti-tumor Therapy—

Female BALB/c nude mice (Weitonglihua Biotechnology, Beijing, China) (n = 30, 5–6-week-old) were inoculated subcutaneously with a mixture of 4 × 10⁶ HECECs and 1.5 × 10⁶ KYSE150s. Tumor was allowed to reach a diameter of 2–5 mm, then randomly divided into five groups of six animals. Four of these groups received intraperitoneal injections of (i) 10 mg/kg of mAb 1D2 for high dosage treatment, (ii) 2.5 mg/kg of mAb 1D2 for low dosage treatment, (iii) 10 mg/kg of normal mIgG, or (iv) PBS, respectively. The respective dosages were administered once daily for 5 days followed by twice weekly for 2 weeks. The remaining group of six mice was treated with 10 mg/kg of mAb 1D2 when the average tumor length reached 8 mm. A separate group of 6 female BALB/c nude mice was inoculated by 1.5 × 10⁶ KYSE150s per mouse as a negative control of the tumor model, and did not receive any treatment. Tumor volumes were measured with a caliper twice weekly and calculated as π/6 × length × width². Tumors were weighed at the conclusion of the experiment.

Analysis of Vessels Density—

Images of immunohistochemically stained sections were collected using a microscope at ×100 magnification. Three ×100 fields (0.5 mm²) for each tumor were analyzed to count the average micro-vessel density and humanized micro-vessel density according to anti-CD31 or anti-huCD31 staining, respectively (12,19). Statistical analysis was performed with Microsoft Excel 2000 to determine whether there were significant differences in vessel densities between pair wise groups.

RESULTS

Co-cultivation of HECECs with KYSE150s Maintains Tumor-specific Properties of HECECs in Vitro—

In the co-culture model, both HECECs and the human esophageal squamous carcinoma cell KYSE150s underwent morphological changes when they contacted. The KYSE150 cells aggregated and formed a nest-like structure, and the HECECs assumed a narrow extended shape, infiltrating into the intervals of the KYSE150 cell nests and aligning themselves with the KYSE150 aggregates. By day 6, the HECECs formed net-like structures resembling a vascular network with the involvement of KYSE150 cells (Fig. 1A). The morphology, on the whole, was similar to natural cancer tissues, where the cancer nests were surrounded by stromal tissue containing small blood vessels. However, monocultured HECECs in the endothelial medium or tumor conditional media could not form the network-like structure. Furthermore, the co-cultured HECECs adopted a more extensive spindle-shaped morphology compared with monocultured HECECs with or without conditional medium. To evaluate their growth stimulation, HECECs were labeled by Cy3-Von Willebrand factor and counted 5 days after co-cultivation. The number of HECECs in co-culture was significantly greater than the number of HECECs in monoculture model without tumor conditional medium (Fig. 1B). The activation of HECECs was further determined by immunocytochemical detection of VEGFR1, VEGFR2, and integrin β3. Co-cultured HECECs showed more intensive expression of these molecules (Fig. 1C).

Fig. 1.

HECECs restore tumor endothelial features in the co-culture model. A, HECECs co-cultured with KYSE150 cells. The morphology of both cell types after 3 days of co-culture shows that HECECs begin to surround tumor cells. After 6 days of co-culture the nests of tumor cells are surrounded by HECECs. The arrow points to the nest of tumor cells, and triangle points to the HECECs. B, the cell number of HECECs. HECECs were maintained in co-culture model or monoculture model with or without conditional medium, respectively. C, immunocytochemical detection of molecules associated with endothelial activation. Upper panels show the results for monocultured HECECs, and the lower panels show the results for co-cultured HECECs. The asterisk denotes statistical significance (p < 0.05).

The Functional Antibody Library Was Established by Immunizing Mouse with Co-cultured HECECs—

Co-cultured HECECs and KYSE150s were separated from the endothelial/tumor cells mixture by selective digestion and fixed immediately for the immunization of five mice. At 12 months post-immunization, the serum from immunized mice was collected for functional identification. The serum of #3 mouse showed the highest anti-angiogenic effects in vitro. As shown in Fig. 2A, the serum of mouse #3 was most effective at suppressing HECEC cell growth. We selected sera of the mice showing the highest anti-proliferative effect for further functional assay, i.e. the sera of mice #3 and #5. Immune serum of 1:200 dilution showed remarkable inhibition rate, whereas normal serum at this dilution did inhibit HECEC proliferation significantly. So we chose dilution of 1:200 for further functional analysis. In accordance with the proliferation assay, the serum of mouse #3 showed a higher inhibitory rate than the serum of mouse #5 in assays including the in vitro tube formation assay (Fig. 2B), the migration assay of HECECs on Matrigel (C), and the adhesion assay of HECECs with KYSE150 cells (D). As a viable cell membrane is not permeable to antibodies, there must be functional antibodies directing to cell surface proteins of HECECs in the serum. Therefore, the antibody library raised from these immunized mice should contain functional antibodies that could suppress angiogenesis of HECECs. Consequently, a library of 2,021 monoclonal antibodies (mAbs) was raised by fusing SP2/0 cells with the spleen cells of mouse #3.

Fig. 2.

HECECs behaviors are disturbed by the serum of immunized mice. A, the proliferation of HECECs is suppressed by the serum of HECECs-immunized mice. Left panel, the proliferation assay was performed with serum of each of five different mice at different dilutions; right panel, proliferation curves of HECECs maintained by 1:200 diluted immune serum of mouse #3, non-immunized mouse serum, and PBS. B, effect of 1:200 diluted immune serum on HECECs tube formation. C, effect of 1:200 diluted immune serum on the migration of HECECs. D, effect of 1:200 diluted immune serum on the adhesion of HECECs to KYSE150 cells. Tumor cells are labeled with green fluorescence. The asterisks denote statistical significance (p < 0.05).

Functional Cell Surface Protein Screen of HECECs by Antibody Library against HECECs—

We immunocytochemically screened the library for monoclonal antibodies that bound specifically to HECECs in comparison with other endothelium. A total of 296 monoclonal antibodies showed reactivity with HECECs, ENECs, LESCs, or HUVECs. 60 of these mAbs specifically reacted with HECECs compared with other endothelia. Fig. 3A shows a typical view of the immunocytochemical reactivity of monoclonal antibody with different endothelial cells. 60 mAbs that strongly reacted with HECECs were subjected to immunohistochemical analysis with carcinoma tissues or normal tissues of human esophagus. Among these mAbs, 47 clones of monoclonal antibodies showed immunostaining in malignant or normal tissue, and 17 of these antibodies showed more intensive staining with tumor blood vessels than normal blood vessels of esophagus. However, five clones of these antibodies that strongly reacted with cancerous vessels also exhibited malignant cell staining. Fig. 3B shows the typical view of immunohistochemical staining of antibodies with carcinoma tissues.

Fig. 3.

HECEC cell surface protein screen using functional library against HECECs. A, a typical view of the reactivity of monoclonal antibodies with HECECs, ENECs, LESCs, and HUVECs. B, immunocytochemistry of mAbs 1D2, 15H10, and 6A2 with human esophageal carcinoma tissues (n = 10). C, double immunofluorescence of HECECs without permeabilization. Left panel, α-tubulin is negatively stained; middle panel, 1D2 staining; right panel, the merged view of α-tubulin staining, 1D2 antibody staining, DAPI staining, and the white light image of the cells. D, the behavior of HECECs is disturbed by the cell surface reactive mAbs. An inhibitory rate, which is lower than zero represents the stimulatory effect of some mAbs.

We subsequently screened the above 17 mAbs of their ability to bind HECEC surface antigens. Double immunofluorescence staining was performed with these monoclonal antibodies and anti-α-tubulin antibody. The anti-α-tubulin immunoreaction was conducted to confirm that these Abs could not access the cytoplasm in cells without permeabilization (20). Nine of the 17 antibodies recognized HECEC cell surface antigens (Fig. 3C) and were selected for further functional analysis. The ability of these mAbs to inhibit the adhesion of HECECs with KYES150 cells, migration, and tube formation of HECEC cells on Matrigel were screened with the control of normal mouse IgG (Fig. 3D). Of the nine mAbs tested, 2F7, 10B8, 15H10, and 6A2 demonstrated inhibition of adhesion, and 2E3 demonstrated inhibition of migration. 15H10 and 2E3 showed a prominent suppression of tube formation (inhibitory rate ≥ 20%).

The Antigen of mAb 1D2 Is Identified as Migration-stimulating Factor—

The molecular weight of the cell surface antigen of mAb 1D2 is approximately 70 kDa as determined by Western blotting and SDS-PAGE following immunoprecipitation (Fig. 4A). Peptides of this antigen were generated by in-gel tryptic digestion, and the mixture of peptides was analyzed by ion-trap electrospray mass spectrometry; parent ions were subsequently selected from the total ion chromatogram for de novo sequencing from fragmentation by MS/MS. The deduced sequences of these peptides (Fig. 4C) confirmed the identity of this antigen as an isoform of fibronectin (IPI: IPI00411462), named migration-stimulating factor (MSF). The primary sequence of MSF (Fig. 4B) indicated a theoretical molecular weight of 72 kDa for MSF, which was in consistence with the mass of this protein determined from Western blotting and SDS-PAGE.

Fig. 4.

MSF is identified as the antigen of mAb 1D2. A, the antigen of 1D2 antibody is identified as ∼70 kDa protein. Left panel, the silver staining of 1D2 antibody immunoprecipitated protein. Right panel, Western blotting of HECEC cellular protein with 1D2 antibody. B, the primary sequence of MSF. Underlines show the peptides that were detected and identified by mass spectrometric analysis following in-gel tryptic digestion. C, MS/MS sequencing of one of the peptides (the green underlined sequence in B); the deduced sequence WRPVSIPPRNLGY corresponds to residues 645–657 of MSF.

MSF Is Involved in HECECs Migration and Adhesion on FN-containing Matrix—

Our functional screen illustrated that mAb 1D2 had no effect on adhesion, migration, and tube formation of HECECs. However, several reports (21,22) suggest that fibronectins always adapted homo- or hetero-polymers to participate in matrix assembling and interaction of endothelium with extracellular matrix (ECM). So we redesigned the in vitro functional analysis to test the role of MSF in HECECs angiogenesis. The migration and adhesion of HECECs on Matrigel or collagen I was not disturbed by 1D2 mAb. However, if HECECs were cultured on fibronectin 1, the 1D2 mAb suppressed migration or adhesion in a dose-dependent manner. Wound-healing assays showed HECECs migrated much more slowly on fibronectin 1-coated plates when 1D2 mAb was added into the culture medium (Fig. 5A). When cell motility was examined using the haptotactic cell migration assay on a fibronectin - coated polycarbonate membrane, 1D2 mAb significantly inhibited HECECs migration to the bottom chamber (Fig. 5B). As migration inhibition could be the result of an increase in the adhesion of HECECs to the substrate, we evaluated the adhesive abilities by measuring the number of HECECs attached to FN1 at different concentration of 1D2 mAb added to the culture medium. As shown in Fig. 5C, 1D2 mAb also inhibited the attachment of HECECs to FN1. However, neutralization of MSF by 1D2 mAb did not affect tube formation of HECECs on FN1-containing matrix. A human hepatic sinusoid endothelial cell line (HSLEC), which barely expressed MSF, was subjected to these analyses in parallel as the control. 1D2 mAb did not disturb the behavior of HSLECs on the FN matrix (supplemental Fig. S1). Therefore, MSF might interact with FN1 to promote angiogenesis of HECECs.

Fig. 5.

MSF participates in esophageal cancer-related angiogenesis. A, wound-healing assay of HECEC cells on matrix of FN. The 1D2 antibody significantly suppresses HECEC cell migration in a dose-dependent manner. B, transwell migration assay of HECEC cells through matrix of FN. The 1D2 antibody significantly suppresses HECEC cell migration by a dose-dependent manner. C, adhesion assay of HECEC cells with matrix of FN. The 1D2 antibody significantly suppresses HECEC cell attaching with FN by a dose-dependent manner. The asterisks denote statistical significance (p < 0.05).

1D2 mAb Specifically Targeted to Human Tumor Vessels in Engrafted Tumor—

As there is no murine counterpart to human MSF, we established an engrafted tumor model with humanized blood vessels by subcutaneously injecting a mixture of human esophageal tumor cell line with HECECs into nude mice (12). In this in vivo model, HECECs formed humanized tumor neovasculature in cooperation with murine blood vessels. This model was used to determine the ability of 1D2 mAb to target to humanized tumor blood vessels in vivo.

We administered ⁹⁹Te-radiolabeled 1D2 antibody intravenously into nude mice bearing esophageal tumor with humanized blood vessels. Mice were sacrificed at 4-h and 24-h post-administration to detect the percentage of injected dose per gram (% ID/g) and tumor:normal tissue ratios. The % ID/g of tumors increased whereas the % ID/g of all normal organs and blood reduced with time. At 24-h post-injection, the % ID/g of most observed organs were clearly lower than that of tumors. The % ID/g of blood was about 26% higher than that of tumor. This finding could be attributable to the long half-life of 1D2 mAb in mice. The % ID/g of kidney was markedly higher than that of tumor, potentially because ⁹⁹Te may be rapidly released from the 1D2 mAb and cleared by the kidney. Tumor:normal tissue ratios also indicated tumor-specific immunotargeting of 1D2 mAb in vivo as these values continued to rise considerably over 24 h (Fig. 6, A and B).

Fig. 6.

1D2 antibody specifically homes to engrafted KESY150 tumors with humanized vessels. A, % ID/g at indicated time post-administration via the tail vein. B, tumor:normal tissue ratios (T/NT) at indicated time post-administration via the tail vein. C and D, gamma scintigraphic imaging of KESY150 tumor-bearing mice. C, the auto-radiographs were recorded at the indicated hours after injecting the ^99mTc-labeled 1D2 mAb via the tail vein. The radio signal increasingly concentrated at the tumor location with time, and the tumor outline was exhibited after 24 h of injection. D, the auto-radiographs were recorded at the indicated hours after injecting the mixture with ⁹⁹mTc-labeled 1D2 mAb with 7.5-fold of cold 1D2 mAb via the tail vein. The tumor-targeting of ^99mTc-radiolabeled 1D2 mAb was inhibited by competition with the 7.5-fold excess of cold 1D2 mAb. The radio signal did not increasingly concentrate in the engraft tumor with time, and the tumor outline was not exhibited. The arrow refers to the bladder, and the triangle directs the tumor.

⁹⁹Te-radiolabeled 1D2 mAb was also injected into mice via the tail vein for gamma scintigraphic imaging. The autoradiographs were recorded at 2h, 4h and 24h post-administration (Fig. 6C). The visible biodistribution revealed the accumulation of 1D2 mAb in tumors. The tumor targeting of ^99mTc-radiolabeled 1D2 mAb was inhibited by competition with 7.5-fold excess of cold 1D2 mAb (Fig. 6D) but not by normal mouse IgG (data not shown).

Neutralization of MSF with mAb 1D2 Targeted to Human Tumor Angiogenesis and Suppressed Tumor Growth in Vivo—

Nude mice bearing esophageal tumor with humanized blood vessels were treated by 1D2 mAb in either low dose or high dose. The mean tumor volumes after treatment with 1D2 mAb were significantly lower than controls treated with normal mouse IgG or PBS (Fig. 7A). The tumor weight was measured 29 days after subcutaneous injection of tumor cells. Treatment with 1D2 mAb effectively reduced the mean tumor weights (Fig. 7B). We also investigated the effect of delayed high dose treatment with ID2 mAb with treatment commenced at 7 days post subcutaneous inoculation, as at this time the tumor had grown to a mean volume of 0.2 cm³. Delayed treatment with 1D2 antibody also induced a similar reduction in esophageal tumor weight.

Fig. 7.

MSF-targeting antibody 1D2 suppresses tumor growth through inhibiting angiogenesis. A, 1D2 antibody suppresses tumor growth. Left panel, the photo of tumor burden on mouse backs. Right panel, the growth curve of indicated group. B, tumor weight of indicated group after mice are sacrificed at 29 days after subcutaneously inoculation. C, the neovasculature in engrafted tumor of mIgG-treated group and high dosage of 1D2-treated group. The total vessel is stained by anti-CD31, and the humanized vessel is stained by anti-human CD31. D, the statistical plots of micro-vessel density in each group. The humanized vessel density is remarkably decreased in 1D2-treated groups, whereas the murine vessel density does not change significantly. The asterisks denote statistical significance (p < 0.05).

The organs, including heart, lung, kidney, spleen, and enteron were subjected to histological analysis. No obvious systemic side effects were observed in the organs of treated mice (data not shown). The average mouse weight, excluding the tumor weight, of treated groups and control groups did not show significant difference when the treatment ended (supplemental Fig. S2).

To estimate if the tumor growth inhibition was because of the decrease of angiogenesis, we performed immunohistochemistry to count micro-vessel density in tumors. Humanized blood vessels were stained by anti-huCD31 and total blood vessels were stained by anti-CD31 (Fig. 7C). The total micro-vessel density of 1D2-treated groups were was about 20% lower than the mIgG treated group. The humanized vessel density decreased about 50% in low dose treated group and delayed treated group, and near 65% in high dose treated group in contrasted with that of mIgG treated group. We counted murine blood vessels by subtracting total micro-vessel density with humanized vessel density (Fig. 7D). We revealed that the murine blood vessel of 1D2-treated groups did not decease comparing with that of mIgG treated group. The reduction of total vessel in the tumor of 1D2-treated groups should resulted from the decrease of humanized vessel in these tumors.

DISCUSSION

Tumor angiogenesis and vascular remodeling are complex processes. These processes are induced and controlled by pro-angiogenic factors presented by the malignant cells, including secreting factors in the ECM as well as molecules expressed on cell membranes that participate in direct endothelium-tumor and/or endothelium-ECM interactions (23). Moreover, the host tissue also confers tissue-specific properties to the endothelium. The vascular endothelium in different tissues is not only morphologically but also functionally different (24,25). The vasculature of individual organ or tissue has been proposed to be sufficiently organ-specific to allow targeting of the vasculature in specific organs via “vascular addressing” (26,27). Therefore, the endothelia of different organs form a contiguous but not homogeneous system that responds differently to tissue-specific angiogenic promoters/inhibitors and pharmaceutical agents (28,29). As an extension of vasculature in individual organs, tumor endothelium also exhibits tissue specificity (30). However, the tumor-specific properties are mostly tumor microenvironment-dependent (10,11, 31). We have previously established an anti-HECECs antibody library by directly immunizing mouse with HECECs. The endothelial cells from esophageal cancer were cultured in vitro to obtain a sufficient population for our studies and did not receive any treatment to restore the tumor-specific characteristic. Few functional and specific antibody direct to HECECs was gained from this library (32). Therefore how to maintain the tumor-specific properties of tumor endothelial cells in vitro is critical for tumor angiogenesis studies.

In the present study, to restore the tumor-specific feature of these endothelial cells, a co-culture model was applied by mixing esophageal tumor cells with the HECECs for in vitro culture. In the co-culture model, we observed the interactions between these two cell types: a cancer nest-like morphology was formed, which was also observed by co-culturing HUVECs with tumor cells (33). In addition, the HECECs grew more quickly in our co-culture model than in monoculture model and expressed higher level of tumor angiogenesis-related molecules, including VEGFR1, VEGFR2, and integrin β3 (34,35). These observations suggested that the co-culture model could mimic the tumor microenvironment and therefore restore the tumor-specific properties of HECECs in both morphological and functional aspects.

For functional proteomic screen of cell surface targets, we employed these endothelial cells to immunize mice for the generation of monoclonal antibodies against HECECs. A similar approach for proteomic screen was also effectively applied to identify novel function of known genes (36,37). We modified this approach by inspecting the immune serum with functional in vitro assay. The serum that could suppress angiogenic behavior of endothelial cells indicated functional monoclonal antibodies existed in the serum. From five immunized mice, we finally chose mouse #3, whose serum could suppress angiogenic behavior of HECECs most, for the generation of the antibody library. This library might contain more functional mAbs than the library derived from other mice. Then a functional monoclonal antibody library was established and used for functional proteomic screen of cell surface antigens targeting tumor-related angiogenesis. These monoclonal antibodies were sequentially subjected to immunocytochemistry and immunohistochemistry analysis; those predominantly reacted with esophageal cancer endothelia were selected for the cell surface immunoreaction and functional analysis. After this screen, monoclonal antibodies specifically directed to functional surface targets of HECECs were selected for antigen identification.

Using immunoprecipitation and mass spectrometry, the antigen of 1D2 mAb was identified as MSF, a genetically truncated onco-fetal fibronectin isoform (38). Fibronectin is a modular glycoprotein consisting of a number of functional domains named on the basis of their binding affinities for other matrix macromolecules and members of the integrin family (21,22). Fibronectin mediates cellular interactions with the ECM and plays important roles in cell adhesion, migration, growth, and differentiation (21,22). Although fibronectin has been studied for over two decades, this remarkably complex protein is still the subject of exciting discoveries, such as the isoform of alternatively spliced extra domains A and B. These isoforms are markers of neovasculature of tumors and can be used as anti-angiogenic targets for tumor treatment (39–41). MSF has been defined to express in fetal and cancer patient fibroblasts and seldom expresses in normal adult tissues (42,43). Although MSF is possibly involved in angiogenesis, there is little direct evidence to support the viewpoint that this molecular play a promotive role in tumor-related angiogenesis. A recent report revealed that MSF is predominantly expressed not only in a wide range of human cancer-associated fibroblasts, but also in tumor vascular endothelial cells, which implies a role of MSF in tumor-related angiogenesis (38,44). In our study, we demonstrated that MSF was intensively expressed in esophageal carcinoma endothelial cells as well as tumor-associated fibroblasts. However, the 1D2 mAb did not disturb the migration and adhesion of HECECs on Matrigel or collagen I. MSF is highly homology to the N-terminal 70-kDa fragment of FN that mediates the interaction of FN to cellular surfaces and initiates the interaction among FN molecules to form disulfide-stabilized multimers for FN-related ECM assembly and therefore influences the behavior of cells on ECM (22). We added fibronectin 1 to the matrix and found that the 1D2 antibody could significantly repress the migration and adhesion of HECECs on the matrix.

Biodistribution analysis and gamma scintigraphic imaging revealed that 1D2 mAb could rapidly and specifically home to the esophageal tumors containing humanized blood vessels, which indicated specific in vivo targeting of 1D2 mAb to esophageal carcinoma vessel endothelial cells. Finally, we tested the therapeutic potential of MSF targeting antibodies in vivo. 1D2 mAb caused a 70% reduction in tumor growth. The delayed treated group also had an average inhibition of tumor growth by 57%, although the immunotargeting therapy began when the tumor volume was about 0.2 mm³, a volume that tumor neovasculature had been founded at that time. We did not observe any obvious systemic side effects in mice, even when the dosage was increased to 10 mg/kg. Therefore, the maximal tolerated dose could be higher than 10 mg/kg. We also found the vessel density of tumors in treated group was significantly lower than that in the control group, and the decreased neovasculature formation was because of the suppression of humanized angiogenesis but not the murine-derived vessel formation. We therefore deduced that the therapeutic effect was apparently because of suppression of humanized tumor neovascularization.

Taken together, we established an antibody library-based proteomics approach to functional screen tumor-related angiogenic targets for tumor treatment. Through this approach, functional mAbs were validated to specifically target to esophageal cancer endothelial cells. MSF was identified as a target for esophageal cancer angiogenesis, and the MSF targeting antibody showed effective anti-angiogenesis activity for the cancer therapy. The antibody library-based tumor endothelial cells surface proteomic functional screen reveals migration-stimulating factor as an anti-angiogenic target.