Identification of new accessible tumor antigens in human colon cancer by ex vivo protein biotinylation and comparative mass spectrometry analysis


  • Paolo Conrotto,

    1. Department of Chemistry and Applied Biosciences, ETH Zurich, Wolfang-Pauli-Strasse 10, 8093 Zurich, Switzerland
    Search for more papers by this author
  • Christoph Roesli,

    1. Department of Chemistry and Applied Biosciences, ETH Zurich, Wolfang-Pauli-Strasse 10, 8093 Zurich, Switzerland
    Search for more papers by this author
  • Jascha Rybak,

    1. Department of Chemistry and Applied Biosciences, ETH Zurich, Wolfang-Pauli-Strasse 10, 8093 Zurich, Switzerland
    Search for more papers by this author
  • Philippe Kischel,

    1. Metastasis Research Laboratory, Center of Experimental Cancer Research, University of Liège, Tour de Pathologie-1, 4000 Liege, Belgium
    Search for more papers by this author
  • David Waltregny,

    1. Metastasis Research Laboratory, Center of Experimental Cancer Research, University of Liège, Tour de Pathologie-1, 4000 Liege, Belgium
    Search for more papers by this author
  • Dario Neri,

    Corresponding author
    1. Department of Chemistry and Applied Biosciences, ETH Zurich, Wolfang-Pauli-Strasse 10, 8093 Zurich, Switzerland
    • Department of Chemistry and Applied Biosciences, ETHZ, Wolfang-Pauli-Strasse 10, 8093 Zurich, Switzerland
    Search for more papers by this author
    • Fax: + 41-446-331-358.

  • Vincent Castronovo

    1. Metastasis Research Laboratory, Center of Experimental Cancer Research, University of Liège, Tour de Pathologie-1, 4000 Liege, Belgium
    Search for more papers by this author

  • Conflict of Interest: All the authors declare that they have not any conflict of interest regarding the results and experimental approaches described in the present manuscript.


One of the most promising new strategies for the development of efficacious cancer therapies relies on the targeted delivery of biopharmaceutical to the tumor environment by the use of selective and specific antibodies. The identification of accessible perivascular proteins selectively overexpressed in cancer tissue may facilitate the development of antibody-based biopharmaceutical administration. This approach is potentially highly selective and specific, combining the presence of tumor biomarkers readily accessible from the blood vessels and the high rate of angiogenesis characteristic of cancer tissues. We performed ex vivo perfusions of surgically resected human colon cancer using a reactive ester derivative of biotin, thus achieving a selective covalent modification of accessible proteins in vascular structures and stroma. After extraction and purification, biotinylated proteins were digested and the resulting peptides submitted to a comparative mass spectrometry-based proteomic analysis, revealing quantitative differences between normal and cancer colon. Sixty-seven of the total 367 proteins identified were found to be preferentially expressed at the tumor site. We generated human monoclonal antibodies against 2 potential tumor targets, NGAL and GW112, and we proved their selective expression in cancer colon and not or barely in healthy tissues. This article presents the first proteomic analysis of human colorectal cancer structures readily accessible from the tumor vasculature, revealing the overexpression of novel tumor antigens which may serve as selective targets for antibody-based imaging and therapeutic biomolecular strategies. © 2008 Wiley-Liss, Inc.

Colorectal cancer (CRC) is the third most common cause of cancer death in the Western world.1 CRC formation follows a well-defined stage progression model already described in the 30s (the adenoma-carcinoma sequence),2 in which a progressive series of changes in cellular proliferation and differentiation leads to the onset and development of the malignancy.3 The sequence of molecular events and epigenetic contributions,4 which leads to the tumor establishment, growth, invasion and metastasis is becoming increasingly understood,5, 6 thanks to advances in genetics, molecular biology and in vivo models. However, most current mechanistic models focus almost exclusively on the localized lesions rather than on the contribution of the tumor stroma and of the tumor neo-vasculature.7

When detected at early stages, the majority of CRC can be cured with high success rates, and there is a strong correlation between tumor progression at the moment of the diagnosis and survival rate.5 For this purpose a number of screening tests have been developed, ranging from fecal occult blood test to colonoscopy.5 However the ideal screening test must be specific, sensitive and noninvasive, and further improvements are needed to achieve these goals. Immunophotodetection procedures have been proposed as an avenue to improve the specificity as well as the sensitivity of diagnostic techniques. This strategy crucially relies on the availability of suitable antibodies, specific to good-quality tumor-associated antigens.8, 9 CRC pharmacological treatment invariably features the use of chemotherapic agents (5-fluorouracil, leucovorin, oxaliplatin and irinotecan),10, 11 but biopharmaceutical agents are increasingly used in combination protocols. Two monoclonal antibodies, bevacizumab (Avastin™), a humanized antibody against the Vascular Endothelial Growth Factor A, and cetuximab (Erbitux™), a chimeric antibody targeting the Epidermal Growth Factor Receptor,12 have recently been approved. Yet, even the most recent combination protocols are rarely curative,13 and a number of side effects have been registered in association with bevacizumab treatment.14, 15

The antibody-based delivery of therapeutic agents (e.g., drugs, cytokines, radionuclides, photosensitizers) to the tumor neo-vasculature is gaining increasing recognition as a promising modality for the treatment of cancer.16–19 This pharmaceutical strategy crucially relies on the availability of good quality tumor-associated antigens, which are easily accessible for intravenously-administered antibodies. Target discovery efforts have largely relied on transcriptomic methodologies. These approaches, however, require extensive antibody-based validation procedures (e.g., by immunohistochemistry (IHC)) and have not led so far to the development of novel clinical candidates. By contrast, proteomic studies of CRC have until now mainly been restricted to the analysis of tumor cell lines20 and of tumor lysates.1, 7, 21

Aiming at the identification of tumor markers suitable for antibody-based therapeutic intervention, our group has recently developed a method for the detection of antigens accessible from the blood stream, based on the terminal perfusion of tumor-bearing rodents with a reactive ester of biotin. This in vivo labeling procedure is followed by the recovery of the biotinylated proteins from normal organs and tumors, which are then proteolytically digested and submitted to a comparative mass spectrometry analysis.22 This approach has recently been extended to the ex vivo perfusion of surgically resected human kidney with cancer, thus directly revealing the overexpression of markers in tumor vascular structures.23 Furthermore, an alternative technique for biotinylation of blood-accessible antigens has been recently developed to enable the exploration of non ex vivo perfusable tumors such as breast cancer.24

In this article, we report for the first time the ex vivo perfusion of surgically resected human colon with cancer using a reactive biotin ester derivative, followed by the proteomic identification of labeled proteins and by the validation of their overexpression in CRC. This study provides a repertoire of proteins which are expressed in the CRC neo-vasculature and in the surrounding stroma. Furthermore, we have generated human monoclonal antibodies specific to 2 of the most promising targets (Neutrophil Gelatinase-Associated Lipocalin [NGAL] and GW112), which react in immunohistochemical procedures and which may serve as tools for the in vivo targeting of CRC.

Material and methods


This study was started upon approval by the ethical committee of the University Hospital of Liège (Belgium). Criteria adopted for patient selection were as follows: (i) diagnosis of caecum adenocarcinoma as determined by standard imaging and subsequent endoscopically guided biopsies, (ii) a therapeutic indication for ileocolectomy, and (iii) a tumor size and localization that allowed healthy portions of the resected colon segment to be clearly distinguished and used as normal controls. Immunohistochemical procedures compatible with the detection of specific proteins without biotin interference were adopted for the diagnostic histopathological analysis. Informed consent was obtained from all patients.

Ex vivo vascular perfusion

Surgery was performed according to a standard procedure, which includes the ligation and section of colic vessels and subsequent ileocolectomy. The colic artery carried a longer suture for immediate identification in the perfusion step. Within 2 min after ileocolectomy, the colic artery was cannulated, the colic vein was opened (by removing the suture) to allow outflow of the perfusate, and perfusion via the colic artery was started. Colons were first perfused for 7–9 min with 500 mL of a 1 mg/mL solution of sulfo-N-hydroxysuccinimide-LC-biotin in PBS, washing away blood components and labeling accessible primary amine-containing structures with biotin. Immediately afterward, a second perfusion step with 450 mL of PBS containing a 50 mM solution of the primary amine Tris was performed for 8–9 min, in order to quench unreacted biotinylation reagent. All perfusion solutions contained 10% dextran-40 (GE Healthcare, Uppsala, Sweden) as a plasma expander and were pre-warmed at 40°C. Both perfusion steps were performed with a pressure of 100–150 mm Hg. Successful procedure was indicated by the wash out of blood during the first minutes of perfusion and subsequent flow of clear perfusate out of the colic vein. After perfusion, the ileocolectomy specimens were washed with 50 mM Tris in PBS, dried, rubbed with black ink to allow the later pathologic investigation of surgical margins, and then opened along the antimesocolic axis. Successful experiment resulted in a whitish color of the tissues. Specimens from the tumor and from the normal colon tissue (unaffected by the tumor) were excised (from well perfused, whitish parts) and immediately snap-frozen for proteomics and histochemical analyses or paraformaldehyde-fixed and paraffin-embedded for histochemical analyses. As negative controls, unperfused colectomy specimens were used, and specimens were taken as described above from the tumor and from normal colon tissue. For specific information about the examined specimens see Supplementary Table 1a.

Protein enrichment, nanocapillary HPLC fractionation and mass spectrometry analysis

Thawed colon specimens were immediately homogenized and sonicated, and biotin-labeled proteins were enriched on straptividin resin followed by trypsin digestion, as described previously.23 This procedure has been tested to be optimal for freshly frozen human material, in contrast to what was shown for cell culture material.25 Tryptic peptides were separated by reverse phase (RP) nano-HPLC controlled by Chromeleon software.23 Peptides were eluted at a flow rate of 300 nL/min, with a buffer B:buffer A gradient of 0–30% for 7 min, 30–80% for 67 min, 80–100% for 3 min and 100% for 5 min (buffer A: 2% acetonitrile 0,1% TFA; buffer B: 80% acetonitrile 0,1% TFA). Eluted fractions were mixed with a solution of 3 mg/mL α-cyano-4-hydroxycinnamic acid, 277 pmol/mL neurotensin as internal standard, 0,1% TFA and 70% acetonitrile and spotted on a target plate using the on-line Probot system (Dionex, Sunnyvale, CA).23 Mass spectrometry analysis was performed as previously described.23 Briefly, the 4,700 Proteomics Analyzer (Applied Biosystem, Foster City, CA) MALDI TOF/TOF was used, and MS spectra were summed from 2,000 laser shots. An automated external calibration was performed using a mixture of five peptide standards in the range 900–2,400 m/z, Gvc spotted in 6 different wells. All fractions were first measured in MS mode and automatically scanned for significant peaks. Precursors with a signal-to-noise ratio >100 (maximum of 15 ions/spot) were selected for fragmentation by Collision Induced Dissociation. The MS/MS spectra were summed up from 2,500 to 5,000 laser shots. Spectra were processed with the Global Protein Service Analyzer (Applied Biosystem, Foster City, CA) using MASCOT software (Matrix science, Boston, MA). Only peptide sequences with C.I. score higher than 95% were taken into account.

PCR analysis

PCR was performed on cDNA libraries of human liver, brain, kidney, heart, placenta, normal colon and cancer colon (Biochain Institute Inc., Hayward, CA). A PCR mix was prepared with 0.5 μM primers, 250 μM dNTPs, 0.5 unit Taq DNA Polimerase (Qiagen, Hilden, Germany), buffer 10× (Qiagen, Hilden, Germany), cDNA library (0,7 μL/50 μL PCR mix). The following primers, annealing temperatures and extension times were used: Cathepsin G FW: 5′-ATGCAGCCACTCCTGCTTCTG-3′; Rev: 5′-TCACAGGGGGGTCTCCATCTGAT-3′; 53°C, 75 sec; Emilin-1 FW: 5′-CCAGCCAGCCGCCACAGGAA-3′; Rev: 5′-CATTCCTGAGGGGGCCTGCGG-3′; 54°C, 80 sec; GW112 FW: 5′-GGACAAGATGAGGCCCGGCC-3′; Rev: 5′-TTACTGGGGCTTCTGCAAGACAG-3′; 55°C, 100 sec; NGAL FW: 5′-ATGCCCCTAGGTCTCCTGTG-3′; Rev: 5′-TCAGCCGTCGATACACTGGTC-3′; 53°C, 75 sec; S100A9 FW: 5′-ATGACTTGCAAAATGTCGCAGC-3′; Rev: 5′-TTAGGGGGTGCCCTCCCC-3′; 50°C, 45 sec; Actin Control primers from Biochain Institute Inc.; 55°C, 30 sec.

Cells, reagents and DNA constructs

Human Embryonic Kidney (HEK) 293 EBNA cells were obtained from ATCC and cultured in Dulbecco Modified Eagle Medium (Gibco), in the presence of 10% fetal calf serum (Gibco, Carlsbad, USA), 100 U/mL penicillin G, 100 μg/mL streptomycin and 250 μg/mL amphotericin B (Gibco, Carlsbad, CA). Cells were transfected with the calcium-phosphate method, and selected with Hygromycin (200 μg/mL, Invitrogen, Carlsbad, CA).

NGAL, Cathepsin G and Emilin-1 antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA); anti-polyhistidine peroxidase conjugated antibody was from Sigma-Aldrich (St. Louis, MO).

GW112 and NGAL were amplified from a human placenta cDNA library (Biochain Institute Inc., Hayward, CA), and subcloned into the pCEP4 vector (Invitrogen, Carlsbad, CA), using the Kpn-BamH I and Hind III-Xho restriction sites, respectively, and inserting a histidine tag at the C-terminus.

Phage display selection and antibody purification

Purification of histidine-tagged antigens was carried out essentially according to the QIAexpressionist protocol (Qiagen, fifth edition), collecting medium from stably transfected HEK 293 EBNA cells. The ETH Gold phage library, set up in our laboratory, was challenged against antigens immobilized on immunotube, as previously described.26 Bound phages carrying the single chain variable fragment (scFv) of an antibody were used to infect exponentially growing E. coli TG1 strain. Phage production was performed as described.26, 27 After two subsequent rounds of selection, monoclonal antibodies were tested in ELISA.27 Monoclonal scFvs with high affinity were purified as previously described.26


IHC was performed on acetone-fixed 10 μm sections of freshly frozen human specimens. Anti-NGAL, -Emilin-1 and -Cathepsin G commercial antibodies were diluted 1:500, and detected by the horse radish peroxidase (HRP)-conjugated secondary antibodies and AEC substrate chromogen (DakoCytomation, Glostrup, Denmark). The anti-NGAL monoclonal antibody produced in our laboratory was used at a concentration of 10 μg/mL and detected using alkaline phosphatase and Fast Red solution, as previously described.28 Biotinylation was detected using alkaline phosphatase-conjugated streptavidin (Biospa, Milan, Italy). Immunofluorescence was performed using anti-NGAL and anti-GW112 antibodies in scFv format detected with Alexa 546 immunoglobulin-conjugate (Eugene, OR), in combination with the anti-von Willebrand Factor (vWF) antibody (DakoCytomation, Glostrup, Denmark) detected with Alexa 488 immunoglobulin-conjugate (Eugene, OR). GW112 IHC on a large number of specimens was performed using formalin-fixed, paraffin-embedded tissue sections of human colon, breast and prostate cancer specimens. Anti-GW112 monoclonal antibody was used at a concentration of 8.5 μg/mL in combination with the anti-c-myc antibody (Santa Cruz Biotechnology, 1/200), and detected using avidin-peroxidase conjugates (Vectastain ABC kit from Vector Laboratories, Burlingame, CA) according to the manufacturer's instructions. Images were taken with Zeiss Axiovert S100TV microscope and the Zeiss AxioCam, and visualized with the AxioVision Release 4.5 software (Zeiss, Chester, VA).


Ex vivo biotinylation of surgically resected human colon specimens

Colon tissues were surgically resected from 3 patients with CRC. Immediately after resection, the specimens were perfused ex vivo using a solution containing a reactive ester derivative of biotin (sulfo-NHS-LC-biotin) with an impaired diffusion through biological membranes,29 followed by a quenching step with a primary amine.22 This procedure leads to the covalent modification of primary amines within the proteins which are accessible to the biotin reagent during the perfusion step (Fig. 1a). The efficiency of the biotinylation reaction was confirmed by histochemistry, using streptavidin-alkaline phosphatase conjugate as detection reagent (Fig. 1b). Sample from an unperfused colon specimen was used as negative control for the subsequent proteomic analysis.

Figure 1.

Ex vivo perfusion and biotinylation process. (a) Schematic representation of the ex vivo perfusion with a derivative ester of biotin. Proteins readily accessible from blood vessels were biotin-labeled, enriched and identified with mass spectrometry. (b) Histochemistry with alkaline phosphatase-conjugated streptavidin with tumor (I, II) or normal (III, IV) colon sections. Perfused tissues (I, III) show a specific staining, absent in the not perfused counterpart (II, IV).

Proteomic analysis of biotin-perfused tissues

Following perfusion, colon tissues were homogenized and the resulting complex protein mixture was enriched for biotinylated proteins by chromatography on streptavidin-coated beads, performed in the presence of 1% SDS. After on-resin tryptic digestion and elution, the resulting peptide mixture was further separated by nano-HPLC and analyzed by MALDI TOF/TOF as described.23 This analysis led to the identification of 367 different proteins, of which 67 were found only in tumor specimens, 121 only in normal colon, and 179 in both sample types. Table I presents a list of some of the most striking differences in protein expression between normal and cancer colon, as well as abundant proteins found in both samples, such as serum components (albumin) and abundant structural proteins (fibronectin, tenascins, collagens). The complete list of all proteins identified is shown in the Supplementary Table 2.

Table I. Proteins Identified in Normal and Cancer Colon
Acc. No.Protein nameNormal colonCancer colonBiological function
  1. Lists of some proteins identified during the proteomic analysis by means of corresponding tryptic peptides. The Accession Numbers and the Protein Name refer to UniProtKB/TrEMBL data bank. The number of specimens corresponds to different, individual surgical samples processed by ex vivo biotinylation and derived from the patients described in Supplementary Table 1. In total, 5 cancer specimens (2 from patient 1, 1 from patient 2 and 2 from patient 3) and 10 normal colon specimens (3 from patient 1, 3 from patient 2 and 4 from patient 3) were used for the proteomic analysis. The last column describes the putative function of the protein, as deduced from literature data.

P07996Thrombospondin-1 precursor04Cell-to-cell and cell-to-matrix interactions
Q99715Collagen alpha-1(XII) chain precursor04Structural function
Q5SYV9NGAL02Lipid metabolism
P08311Cathepsin G02Proteolysis.
P06702Protein S100-A902Inflammatory responses, cell-cell signal
O75815Breast cancer anti-estrogen resistance protein 302Intercellualr signal; regulation of cell proliferation
Q5T5R0Protein tyrosine phosphatase receptor02Tyrosine phosphatase activity
Q6UX06GW11201Unknown; overexpressed in gastric and colon cancers
O75145Liprin-alpha-301Regulation of the disassembly of focal adhesions and cell-ECM adhesion
P13611Versican precursor01Regulation of cell motility, growth and differentiation
Q3LX97Serine protease inhibitor Kazal-type 501Peptidase activity
Q9HCB6Vascular smooth muscle cell growth-promoting factor01Cell adhesion protein
Q6F3F7Developmentally regulated G-protein-coupled receptor α-201Growth and development regulation
Q59F63Transforming growth factor beta-induced protein90Cell adhesion
P01019Angiotensinogen precursor70Regulation of volume and mineral balance of body fluids
P35613Basigin precursor60MMPS regulation
O15230Laminin alpha-530Regulation of cell attachment, migration and organization
Q99795A33 antigen precursor30Cell-cell recognition and signaling
P56199Integrin alpha-120Laminin and collagen receptor
Q13162Peroxiredoxin-420Regulation of redox metabolism.
P48960CD97 antigen precursor20Cell adhesion and signaling processes
Q86YG0Growth-inhibiting protein2020Transport carrier
Q5D0D7Albumin105Carrier activity
P51888Prolargin precursor103Anchoring protein for basement membranes
P98160Basement membrane- specific heparan sulfate proteoglycan105Structural component in ECM
P07585Decorin93Support and regulation of fibrils formation.
Q15063Periostin precursor55Cell attachment and spreading
P24821Tenascin precursor74Regulation of cell adhesion and migration
P21810Biglycan precursor74Involved in collagen fibers assembly
Q14112Nidogen-2 precursor44Cell adhesion glycoprotein
Q9Y6C2Emilin-122Cell adhesion and vessel assembly

Validation of protein overexpression in CRC

Cathepsin G, NGAL and GW112 represent some of the most striking differences observed between CRC and normal colon specimens (Table I). Emilin-1 was found in both normal colon and cancer specimens, but with higher signal intensity and peptide coverage in the tumor counterpart.

To verify the expression patterns of these genes, a PCR analysis was performed on cDNA libraries derived from different human organs. In all cases, a preferential expression in CRC was observed. By contrast, actin (chosen as positive control) was found to be expressed in all of the organs analyzed (Fig. 2). Consistent with the mass spectrometric data, S100A9 was found to be overexpressed in the colon cancer sample, but was also observed in many normal tissues. For Cathepsin G, NGAL and Emilin-1, expression was also observed in human placenta, suggesting a possible association with active angiogenesis.

Figure 2.

PCR validation of MS data. 5 potential targets were screened for their expression in normal versus cancer colon, together with 5 other organs. The same amount of cDNA was used in all samples, as proved by actin amplification. MW: molecular weights.

We then investigated the expression of these proteins by IHC experiments, using the commercially available antibodies against Cathepsin G, Emilin-1 and NGAL. Figure 3 reveals a specific staining of Emilin-1 in tumor sections (Figs. 3a and 3b), whereas normal colon sections (Fig. 3d) and tumor sections incubated with anti-Emilin-1 antibody plus the corresponding inhibitory peptide (Fig. 3c) did not display a staining reaction. Emilin-1 appears to be mainly localized to the extracellular spaces among colon villi, consistent with its role as secreted glycoprotein. Similar results were obtained with anti-Cathepsin G antibody, yielding a selective staining of tumor sections (Figs. 3e and 3f), compared to the normal colon counterpart (Figs. 3g and 3h). Cathepsin G is a protease produced by neutrophils and released mainly at inflammation sites; the immunohistochemical analysis reveals a strong staining of specific cells, likely infiltrated macrophages. Similarly, commercial anti-NGAL antibody reveals a preferential staining of colon cancer (Fig. 3i) compared to normal colon sections (Fig. 3m). A strong antigen expression was also observed in human MCF7 breast tumors and Lovo colon tumors grafted in mice, but not in melanoma, medulloblastoma, head and neck squamous cell carcinoma or in the negative control (data not shown). The staining pattern was mainly confined to the tumor stroma, compatible with the fact that NGAL is a secreted protein.

Figure 3.

Immunohistochemistry validation of MS data. (ad) Staining with anti-Emilin-1 antibody; a specific staining is detectable in cancer colon specimens (a, b) compared to normal colon (d). Incubation of anti-Emilin-1 antibody with its epitope inhibitor peptide suppresses the staining reaction in cancer tissue (c). (eh) Immunohistochemistry with anti-Cathepsin G antibody; colon cancer tissues (e, f) reveal a specific signal compared to the normal colon counterpart (g, h). (i, m) Staining with anti-NGAL commercial antibody. Cancer (i) and normal (m) colon sections stained with commercial anti-NGAL antibody. (l, nr) Immunohistochemistry with monoclonal anti-NGAL antibody produced in our laboratory in scFv format. Colon cancer sections from 3 different patients (panels l,o,p) incubated with anti-NGAL antibodies exhibit a specific staining, which was not detectable in portions of normal colon (n), or when the primary antibody was omitted (cancer colon panel q; normal colon panel r). The experiments were performed in triplicate with freshly frozen cancer and normal specimens.

Anti-GW112 and anti-NGAL antibody selection and production

In addition to the immunohistochemical studies described in the previous section, we focused our attention on GW112, a protein suspected to be over-expressed in gastric, colorectal, lung and breast cancer based on transcriptomic studies,30–32 and recently described in gastric cancer tissues.33 Since monoclonal antibodies for GW112 were not available, we generated human monoclonal antibodies against this protein using antibody phage technology34 with a large synthetic antibody library recently produced in our laboratory.26 Furthermore, in view of the promising immunohistochemical results obtained with polyclonal antibodies specific to NGAL, we used antibody phage technology in order to produce monoclonal antibodies also against human NGAL. Both GW112 and NGAL were recombinantly expressed as secreted proteins in stably-transfected HEK293-EBNA cells, in order to ensure proper folding and correct disulfide bond formation. After purification on nickel resin, the purity of the corresponding recombinant proteins (which were used for phage antibody biopanning) was checked by SDS-PAGE analysis (Supplementary Fig. 1). After two rounds of panning on immobilized antigen, several antibody fragments in scFv format were found that specifically recognized the cognate antigen. The amino acid sequences of variable regions of the two monoclonal antibodies against GW112 and NGAL used for immunohistochemical studies are showed in Supplementary Table 3.

In the IHC experiments performed, 3 out of 4 CRC tissues tested with the anti-NGAL scFv antibody show strong immunoreactivity, as illustrated in Figure 3 (panels l, o, p); no significant staining was observed in normal colon (Fig. 3n), or in tissues incubated in the absence of the antibody (Fig. 3q,r). While most of the NGAL staining was detectable in tumor stroma, expression on tumor blood vessels was also observed, as evidenced by a two-colors immunofluorescence analysis with the anti-vWF antibody (Figs. 5a, 5d and 5g).

Sections from paraffin-embedded, formalin-fixed human colon (n = 29), breast (n = 5), and prostate (n = 5) carcinomas were subjected to immunohistochemical detection of GW112 using our specific human monoclonal antibody in scFv format. The clinico-pathological characteristics of the malignant lesions are detailed in Supplementary Table 1. Nearly all sections also contained non-neoplastic adjacent glands. Negative control experiments included omission of primary scFv antibody and omission of both primary scFv and secondary anti-myc antibodies. In such experiments no specific staining was observed (Figs. 4c, 4f and 4i). In a few cases however, when only the anti-scFv antibody was omitted from the immunohistochemical procedure, focal, mild, nuclear reactivity could be observed, presumably as a result of the detection of nuclear myc in the cells. Strong anti-GW112 immunoreactivity was observed in the tumor tissue of a subset of colon, breast, and prostate cancers (panels 4a, 4d and 4g). Overall, 21 out of the 33 colon cancers tested (29 paraffin-embedded + 4 freshly frozen tissues) showed strong reactivity (staining intensity = 2–3, including 14 cases with more than 80% of the tumor cells labeled). Beside a GW112 immunoreactivity spread in the entire tumor environment, a co-localization with the endothelial marker vWF was observed (Fig. 5). In all cases, corresponding normal colonic glands harbored no (staining intensity = 0) or a low level (staining intensity = 1) of anti-GW112 reactivity (Fig. 4b). Similar findings were observed in the breast cancer lesions analyzed. Out of the 5 breast cancers included in this study, 3 of them showed strong anti-GW112 immunoreactivity (staining intensity = 2–3 with most of the tumor cells (>80%) being positive), while matched normal glands exhibited no or a low level of anti-GW112 staining (0–1; Figs. 4d and 4e). By contrast, among the 5 prostate adenocarcinomas tested, only 1 case showed strong (staining intensity = 3), focal (±15% of the tumor cells) anti-GW112 reactivity (Fig. 4g). As in breast and colon normal glands, non-neoplastic glands stained negative or weakly (staining intensity = 1) for the anti-GW112 antibody (Fig. 4h).

Figure 4.

Immunohistochemical detection of GW112 in tumoral and matched normal tissue sections from colon (a, b, c), breast (d, e, f) and prostate (g, h, i) cancers with the home-made scFv antibody. Experiments in which the anti-GW112 scFv antibody was omitted from the immunohistochemical procedure served as negative controls (panels c, f, i). Sections were counterstained with hematoxylin. Original magnification: a, b, d and g: ×100 (scale bars: 100 μm); c, e, f, h, i and inserts in a, c, d and g: ×400 (scale bars: 25 μm).

Figure 5.

Two-colors immunofluorescence of colon cancer tissues. Sections were stained with anti-NGAL (a), or anti-GW112 (b,c) antibodies, together with the anti-von Willebrand Factor antibody (df). Merged panels (gi) shows the presence of both antigens in proximity of the blood vessels. A large number but not all the blood vessels express the antigens, as indicated by the asterisks in panel h. Magnification: a, b, d, e, g, h: scale bars 100 μm; c, f, i: scale bars 25 μm.


The antibody-based targeted delivery of bioactive molecules to sites of CRC may facilitate cancer detection using scintigraphic or fluorescence-based imaging techniques,8, 35 and may yield biopharmaceuticals with improved therapeutic index.36–38 Such tumor targeting strategies crucially rely on the availability of good target antigens. Tumor-associated antigens which are easily reachable from the tumor neo-vasculature are particularly attractive, considering their accessibility for intravenously-administered antibody-based therapeutic agents. The development of such biopharmaceutical strategies is particularly important for CRC, in view of its high vascularization rate, the need for curative therapeutic agents and the high incidence of colon malignancies.

This work presents for the first time a comprehensive proteomic study of antigens expressed in CRC structures, which are readily accessible from the tumor vasculature. We have used a combination of ex vivo perfusion with biotinylation reagents of surgically resected human colon with cancer and mass spectrometry for the discovery of novel tumor associated antigens. Furthermore, we have generated human monoclonal antibodies against two of the most promising tumor targets (GW112 and NGAL), which may facilitate the development of antibody-based anticancer strategies.

In total, we identified 367 proteins, 67 of which were found to be overexpressed in colon cancer, whereas 121 were detected only in normal colon. Among other, we identified only in normal colon (i.e. downregulated in CRC) anti-oncogenic molecules such as the Growth-inhibiting protein 20, Thy-1 antigen (a negative regulator of cell migration), FRAS-related extracellular matrix protein1 (having a role in epidermal differentiation and adhesion), PEDF (a potent inhibitor of angiogenesis). By contrast, we found only in CRC the breast cancer resistance protein-3 (which regulates the proliferation of breast cancer cells), tyrosine phosphatase receptors and G-protein coupled receptors. We focused our interest on some of the most promising candidate tumor markers. Both transcriptomic and immunohistochemical analysis confirmed the overexpression of Cathepsin G, Emilin-1, NGAL and GW112 in CRC compared to normal colon specimens. Cathepsin G is a serine protease produced by neutrophils and secreted in the extracellular matrix at sites of inflammation. Serine proteases are active in host pathogen degradation, regulation of chemokines and cytokines, and activation of receptors such as integrins, PARs, CAMs, thus affecting cell adhesion and spreading.39 Moreover, the correlation between inflammation and inflammatory agents with malignancy initiation and promotion is well accepted.40

Emilin-1 is a glycoprotein which plays an important role in the organization of extracellular matrix of elastic tissues, and it is highly expressed during embryo development in vascular system.41 Emilin-1 knock-out mice have an altered vascular tree, showing a decrease in both peripheral vessel diameter and wall thickness; this results in unbalanced blood pressure homeostasis of surrounding tissues.42

NGAL, also known as Lipocalin 2, is a 25 kDa protein involved in iron transport. Although its exact role in tumor environment is not yet completely understood, different studies showed high level of NGAL expression in adenocarcinoma of lung and pancreas,43 primary breast cancer,44 ovarian cancer45 and oesophageal carcinoma.46 Particularly, a high protein expression has been detected in CRC but not in normal colon.43 Despite these promising data of NGAL as a broad spectrum tumor biomarker, no further in vivo studies in animal models or clinical translation methodologies have been so far reported, to our knowledge.

GW112 is a recently identified protein30 which, based on trascriptomic studies, has been postulated to be overexpressed in gastric,31, 47 pancreas,32 lung32 and colon31, 32 tumors, with a barely detectable level of transcripts in other tumors as well as in healthy tissues. A positive correlation between GW112 mRNA amount and gastric cancer progression,47 together with a proliferative activity in a pancreatic cell line indicated by small interfering RNA experiments, suggested the involvement of this protein in the establishment and/or development of some types of malignancies, mainly concerning the gastroenteric tract. Recently, the GW112 expression in gastric cancer tissues has been shown.33 Nevertheless, the absence of monoclonal antibodies has so far prevented the characterization of this protein, and its possible use as tumor biomarker.

We generated human monoclonal antibodies against GW112 and NGAL, which were used to study the tissue distribution of these antigens. Overall, we showed a strong GW112 expression in 64% of the colon cancers and 60% of the breast cancers analyzed, while no widespread protein expression was found in prostate cancer (20% of tissues tested).

The availability of these human monoclonal antibodies will allow the study of GW112 and NGAL function by selective in vivo inhibition of their function, as well as their characterization as suitable targets for antibody-based delivery strategies, using quantitative biodistribution analysis, in full analogy to other human antibodies previously developed by our lab which are currently being investigated in clinical trials.48 Human monoclonal antibodies, rather than murine of chimeric antibodies, are nowadays considered to be necessary for the development of low-immunogenicity antibody-based therapeutic strategies.18, 49


We thank the Functional genomics Center Zurich for access to instrumentation and technical support.