The Epithelial Glycoprotein 2 (EGP-2) Promoter-driven Epithelial-specific Expression of EGP-2 in Transgenic Mice: A New Model to Study Carcinoma-directed Immunotherapy¹

Since its discovery in 1979 (1), the human pancarcinoma associated EGP-2,³ also referred to as 17–1A or Ep-CAM, has become one of the major targets for carcinoma-directed immunotherapy and, more recently, also for gene therapy (2). This is not surprising because EGP-2 is intensively and uniformly expressed on a variety of epithelial tissue-derived cancers, such as those of the breast, pancreas, gonads, and gastrointestinal, respiratory, and urinary tracts. In pathology, EGP-2 serves as a suitable marker for differential diagnosis and prognosis of several types of carcinomas (3, 4). Antibodies to this M_r 38,000 transmembrane antigen have been successfully used in patients for imaging of small cell lung cancer (5) and for adjuvant treatment of minimal residual disease of colon carcinoma, leading to an increased survival of 7 years for this otherwise poorly prognosed disease (6). Various groups have tried to enhance the potential of anti-EGP-2-mediated immunotherapy by creating single-chain variable-region fragments to reduce the size of antibodies or by humanizing constant regions of the mouse antibodies to lower immunogenicity. Effector functions have been improved by designing bispecific constructs to retarget immune effector cells to tumor cells, by fusing antibodies to cytokines, drugs, or gene therapy vehicles, or by developing vaccines to EGP-2 (2, 7, 8, 9, 10). Still, the overall therapeutic efficacy of these immunotherapeutic devices has yet to be established.

EGP-2 is a transmembrane glycoprotein that has been characterized as a homotypic adhesion protein. Its main function, however, is still poorly understood (11). In normal tissues, EGP-2 is expressed on the basolateral cell surface of the simple, transitional, and pseudostratified epithelium of the respiratory, gastrointestinal, and urinary tracts, and of the pancreas, gonads, uterus, and cervix. Other tissues, such as heart, spleen, muscle, brain, and connective tissues lack the EGP-2 antigen (12). When using EGP-2 as a target for immunotherapy, there is a risk that side effects will be induced by the binding of the antigen to normal tissue. Indeed, toxicity has been observed after treatment of patients with high-affinity anti-EGP-2 Mabs (13) and with a high affinity anti-EGP-2 Mab-derived bispecific antibody (8).

Animal models have a proven validity for testing and validating the applicability of a therapeutic concept. The relevance of an animal model for studying toxicity during therapeutic targeting of the EGP-2 antigen greatly depends on the specific expression of EGP-2 on normal epithelial tissues. Endogenous EGP-2 expressed by the mouse itself has been used to study anti-EGP-2 immunotherapy strategies (14). However, the distribution pattern of mouse EGP-2 differs from that in humans in that in the mouse, EGP-2 is not only expressed in epithelia but also in lymphoid organs, such as spleen and thymus, and in T- and B cells, as well as in dendritic cells (15, 16, 17). Therefore, results obtained in wild-type mice using mouse EGP-2 as a target may not be of direct relevance for humans.

Transgenic mice have extensively been used to determine the function of proteins, both in the development of disease and for the evaluation of anti-disease therapies. Although the EGP-2 cDNA was already cloned in 1990 (18), no suitable transgenic animal model in which the EGP-2 protein is accurately expressed has been generated thus far. This is probably attributable to the fact that appropriate regulatory sequences were not available. Here, we describe the isolation of the EGP-2 regulatory sequences and the application of these sequences to direct epithelial-specific EGP-2 expression in mice in a way similar to the human situation. We evaluated these EGP-2 transgenic mice for their use as a model to study the effect of Mab-based EGP-2-directed immunotherapy on established syngeneic EGP-2-positive tumors and on EGP-2-expressing normal tissues.

By SalI/SacII digestion, a DNA fragment containing part of exon 1 of the EGP-2 gene with an additional 200 bp of EGP-2 upstream sequences was derived from plasmid GA21726–22RS, kindly provided by Dr. A. J. Linnenbach (Wistar Institute, Philadelphia, PA). This fragment was used as a probe to identify an EGP-2-containing BAC clone (Genome systems Inc, St. Louis, MO) from which a 55-kb XhoI fragment containing the human EGP-2 gene could be isolated. DNA from this single positive BAC clone was purified according to standard methods for BAC DNA isolation (19). The presence of EGP-2 genomic sequences was determined by PCR of exons 2–3 (sense strand primer, 5′-ATAATAATCGTCAATGCCAGTGTA; and antisense strand primer, 5′-ATCATAAAGCCCATCATTGTTCT) and exon 9 (sense strand primer, 5′-TCAGATAAAGGAGATGGGTGAGA; and antisense strand primer, 5′-GGCAGCTTTCAATCACAAATCAG). By restriction analysis and subsequent Southern blotting using the SalI/SacII fragment of the GA21726–22RS plasmid or the 1.5-kb EGP-2 cDNA as probe, it could be determined that at least 10 kb of upstream sequences and 31 kb of downstream sequences were present (Fig. 1 A–C). The 55-kb XhoI fragment was introduced into oocytes of FVB/N mice (Harlan, Leiden, the Netherlands) according to standard methods (20). By PCR and Southern blot analyses, three mice were found positive for the EGP-2 transgene. Of these founders, two lines transmitted the transgene to their progeny. These two founder mice were named FVB/N-TgN(EGP2BAC)2 pm and FVB/N-TgN(EGP2BAC)3 pm, but they will be referred to throughout the article as BACF2 and BACF3. Both the BACF2 and BACF3 mice lines were healthy and fertile and expressed the EGP-2 mRNA as determined by RT-PCR using intron spanning exon 3–7 primers (21). FISH was carried out essentially as described previously (22) using the BAC vector containing the 55-kb XhoI EGP-2 genomic sequence as a probe. The line expressing the highest copy number was selected for additional studies. For the investigation of tumor growth, the EGP-2 transgenic FVB/N mice were crossed with C57/BL6 (Harlan) wild-type mice.

Tissue-culture supernatant of the hybridomas MOC31 and Bly-1 (anti-EGP-2, IgG1 and antihuman CD20, IgG1) was purified by protein A column chromatography (Amersham Pharmacia Biotech AB, Uppsala, Sweden), biotinylated, and filled out in a stock volume of 1 mg/ml. The recombinant fully humanized anti-EGP-2 Mab UBS 54 (7) was kindly provided by Dr. T. Lochtenberg (University Hospital Utrecht, Utrecht, the Netherlands). Horseradish peroxidase-conjugated streptavidin (SA-PO) diluted 1:100 or goat antihuman immunoglobulin diluted 1:50 (both from Dakopatts, Glostrup, Denmark) was used to detect the bound Mabs. Immunoperoxidase stainings were performed on 5-μm-thick, air-dried cryosections made from snap-frozen biopsies. After acetone fixation and rehydration, diluted antibody 1:100 or 1:10 and conjugate in the presence of 2% normal mouse serum were applied to the sections and incubated at room temperature. As substrate for specific staining, 0.01% H₂O₂ was used in combination with 3-amino-9-ethyl-carbazole (AEC; Sigma Chemical Co., Bornhem, Belgium). Counterstaining was performed using a Mayers hematoxylin solution (Merck). In vivo localized MOC31^bio or Bly-1^bio was detected by applying only SA-PO, diluted 1:25, to the cryosections, which were subsequently stained as described above. By blocking the cryosections with avidin for 15 min and with biotin for 15 min (both from Dakopatts) prior to staining with SA-PO, it could be demonstrated that biotin was responsible for the staining observed.

To analyze EGP-2 protein expression by SDS-PAGE and Western blotting, tissues were homogenized in SDS-PAGE sample buffer without 2-mercaptoethanol (19). Total protein concentration was determined by SDS-PAGE and subsequent Coomassie Brilliant Blue staining (19). Approximately 10 μg of protein were fractionated by SDS-PAGE and electroblotted onto nitrocellulose filters (Amersham Pharmacia Biotech, Little Chalfont, United Kingdom) using the Bio-Rad miniprotean II system according to the manufacturer’s protocol (Bio-Rad Laboratories, Hercules, CA). After blocking, the filters were incubated with 1:100 diluted MOC31^bio Mab, and specific binding of the antibody was detected by 1:5000-diluted SA-PO and chemoluminescence (Pierce, Rockford, IL).

The murine B16.F10 melanoma cell line was obtained from the American Type Culture Collection. The EGP-2-transfected B16 melanoma cell line, B16.C215, was kindly provided by Dr. M. Dohlsten (University of Lund, Lund, Sweden) and cultured in the presence of 500 μg/ml G418 (Life Technologies, Inc., NY) as described previously (23). The cell lines were maintained in DMEM (Life Technologies, Inc.) supplemented with 50 μg/ml gentamicin sulfate (Biowhittaker, Vervier, Belgium), 2 mm l-glutamine (Life Technologies, Inc.), and 10% FCS (Bodinco, Alkmaar, the Netherlands), at 37°C in humidified 5% CO₂ atmosphere.

Flow cytometric analysis was performed on 100 μl of peripheral EDTA blood of EGP-2 transgenic and nontransgenic mice. To analyze the blood cells for the presence of EGP-2, they were incubated with PBS or 1:100 times diluted MOC31^bio or Bly-1^bio at room temperature for 30 min. After one wash with 2 ml of PBS, 100 μl of SA-FITC (Becton Dickinson, Mountain View, CA) were added to the cell pellet, which was incubated in the dark at room temperature for 30 min. Cells were resuspended in 1 ml of Facs-lysing solution (Becton Dickinson) incubated in the dark at room temperature for 10 min, washed once with 2 ml of PBS, and resuspended in a final volume of 200 μl of PBS for analysis. The samples were analyzed on a Coulter Elite cytometer (Coulter Electronics, Hialeah, FL) at 488 nm. Immunofluorescence emission was measured using a 525-nm bandpassfilter.

In vivo studies were performed with 8- to 12-week-old male and female EGP-2 transgenic and nontransgenic FVB/N/C57/BL6 F1 hybrid mice. Tumors were induced by injecting 1 × 10⁴ B16.F10 or B16.C215 cells s.c., and subsequent tumor growth was monitored every 2nd day. In case of a palpable tumor, tumor length (L) and diameter (D) were measured, and tumor volume was calculated according to the formula L × (D)² × 0.4. All of the animals were killed on day 25, irrespective of tumor size. On the basis of the tumor growth curve, biotinylated Mab MOC31 (anti-EGP-2; IgG1) or biotinylated isotype-matched control Mab Bly-1 (antihuman CD20; IgG1) were injected i.v. at a dose of 100 μg/100 g body weight in four transgenic mice each. This concentration of antibody was based on our previous experiences and on that of others in animal models and patients (6, 24, 25). As additional controls, four transgenic animals were given injections i.v. with an equal volume of PBS and four nontransgenic animals, with the biotinylated MOC31. Essential organs as listed below were isolated 24 h after the injection, carefully frozen in isopentane (Merck, Darmstadt, Germany), and stored at −80°C until further analysis.

To achieve the desired human-identical EGP-2 expression pattern in mice, a BAC clone was isolated that contained the 14-kb EGP-2 genomic sequences including 10-kb-upstream and ∼31-kb-downstream sequences, as determined by Southern blot analysis and PCR (Fig. 1, A–C). EGP-2 transgenic mice were generated by injecting the 55-kb XhoI EGP-2-spanning genomic DNA fragment that was isolated from this BAC clone into FVB/N mice oocytes. Three mice were found positive for the EGP-2 transgene by PCR and Southern blot analysis. Of these founders, two lines transmitted the transgene to their progeny. Comparison of the band intensity of HindIII-digested DNA in a Southern analysis of these two founder mice, designated BACF2 and BACF3, with those of the original BAC clone and plasmid standards indicated that BACF2 and BACF3 carried approximately three and two copies, respectively, of the EGP-2 gene (data not shown). Both of the lines were healthy and fertile, and after 20 months of observation, no abnormalities have been detected to date.

Analysis of BACF2 and BACF3-derived mRNA by RT-PCR for the presence of EGP-2 and endogenous β2-microglobulin transcripts revealed expression of EGP-2 in the lung, kidney, thymus, stomach, pancreas, and ovary plus uterus, small intestine, and colon, whereas no EGP-2 mRNA was detected in brain, heart, muscle, spleen, liver, or skin tissue (Fig. 2,A). This EGP-2 expression pattern was confirmed by SDS-PAGE, Western blotting, and subsequent analysis using the biotinylated EGP-2 specific Mab MOC31 (Fig. 2 B). The intensity of the band does not imply the expression level, because the number of cells expressing EGP-2 differs between the organs analyzed. Expression level was established by immunohistochemical analysis. In nontransgenic littermates, EGP-2-specific mRNA as well as EGP-2 protein could not be detected. Furthermore, it was established that the observed transgene expression pattern was integration site-independent as determined by FISH analysis of the two established founder lines (data not shown).

Immunohistochemical analysis using the biotinylated Mab MOC31 as well as the humanized anti-EGP-2 antibody UBS 54 revealed that the EGP-2 regulatory sequences directed the EGP-2 expression to the membrane of epithelial cells demonstrating a level of EGP-2 expression and a distribution pattern similar to the one seen in the human situation (Fig. 3 and Table 1). The kidney showed strong EGP-2 expression in the epithelial cells of Henle’s loop, whereas Bowman’s capsule and the proximal and distal tubuli stained weakly positive. Tracheal epithelium, stratified bronchial epithelium of the major airways, aveolum epithelium, and epithelial tissue found in the mucous glands of the bronchial mucosa also stained positive (Fig. 3,A). Of the gastrointestinal tract, epithelium of the villi and of crypts of the small intestine (Fig. 3,D) and colon, as well as the surface epithelium of the stomach, showed EGP-2 expression, whereas the gastric glands appeared negative. EGP-2 expression was also observed in the glandular epithelium of the endometrium, the tubuli seminiferi of the testis, and in the epithelium of the ovarium fallopian ducts. Furthermore, the exocrine and ductal epithelia of the pancreatic tissue were EGP-2 positive, and the endocrine epithelial tissue stained only weakly positive (Fig. 3,G). In the liver, EGP-2 expression was observed in the bile duct epithelium, whereas the hepatocytes were negative. In the thymus, Hassall’s corpuscles stained positive for EGP-2. No EGP-2 expression was observed in T- and B cells nor in dendritic cells as determined by flow cytometrical analysis (Table 1).

Because EGP-2 is one of the best-studied tumor-associated antigens and has been frequently used as a target for experimental and clinical cancer immunotherapy, we induced human EGP-2-positive tumors in the EGP-2 transgenic mice. To this end, we made use of the B16.F10 murine melanoma cell line stably transfected with the EGP-2 cDNA. The resulting cell line has been designated B16.C215 (23). To adapt these C57/BL6-derived tumor cell lines to the EGP-2 transgenic FVB/N mice, transgenic FVB/N/C57/BL6 F1 hybrid mice were generated, and tumor growth was monitored after s.c. inoculation of EGP-2-positive B16.C215 or EGP-2-negative B16.F10 tumors. The EGP-2 expression level and pattern in the hybrid genetic background remained similar to the observed expression level and pattern in the EGP-2 transgenic FVB/N mice, as was established by immunohistochemical analysis (Table 2). Tumor outgrowth and EGP-2 expression were evaluated 1, 2, and 3 weeks after s.c. tumor inoculation by Western blot analysis. Despite the absence of G418, EGP-2 expression by the B16.C215 cells remained present for at least 3 weeks after tumor inoculation (Fig. 4,A). The differences in EGP-2 expression are accounted for by the amount of normal tissue present when isolating a tumor. As the tumor grows, the total amount of protein is more and more accounted for by the tumor tissue. To monitor tumor growth more accurately, B16.F10 or B16.C215 tumors were induced s.c. in eight transgenic and eight nontransgenic FVB/N/C57/BL6 hybrid mice, and tumor volumes were calculated according to the formula L × (D)² × 0.4 every second day. This experiment was repeated three times, and tumors were isolated on day 25, irrespective of tumor size. No statistical significant difference in growth of the s.c.-induced B16.F10 or B16.C215 tumors were observed in the EGP-2 transgenic mice as compared with nontransgenic littermates (Fig. 4 B).

To study the effect of the presence of EGP-2 on normal tissue on the localization of anti-EGP-2 Mabs, biotinylated MOC31 (MOC31^bio) was injected i.v. in four transgenic and four nontransgenic mice bearing both the B16.F10 and B16.C215 tumors. As controls, PBS and biotinylated Bly-1 (Bly-1^bio), an isotype-matched sham Mab, were injected each into four transgenic tumor-bearing mice. Injection into the tail vein took place 3 weeks after tumor inoculation, irrespective of tumor size. Mice were killed 24 h after the injection. Tumors and relevant organs were isolated and analyzed immunohistochemically, using a high concentration of SA-PO. i.v. administered MOC31^bio localized specifically at the EGP-2-positive B16.C215 tumors in both transgenic and nontransgenic animals (Fig. 5,A, parts A and G), whereas no such localization was observed at the B16.F10 tumors analyzed (Fig. 5,A, parts B and H). In addition, no staining of the B16.F10 or B16.C215 tumors could be observed when PBS or Bly-1^bio were injected (Fig. 5 A, parts C–F).

Table 2 presents the results of the immunological stainings performed on frozen tissue sections of the different organs. i.v.-applied MOC31^bio could not be detected in EGP-2-negative brain and spleen tissue nor in the EGP-2-expressing tissues such as thymus, colon, small intestine, lung and pancreas (Fig. 5,B, parts I, J, and K). In other tissues, such as those of stomach, liver, skin glands, and kidney, the localization was inconclusive, because the presence of endogenous biotin caused an aspecific background staining as a result of the binding of endogenous biotin to SA-PO, which was used in a high concentration. Because the staining pattern observed in nontransgenic mice that were given i.v. injections of MOC31^bio and in transgenic mice that were given i.v. injections of Bly-1^bioor PBS was similar to the pattern observed in transgenic mice that were given i.v. injections of MOC31^bio (Table 2), this pattern could not be attributable to either Mab clearance or specific EGP-2-driven localization. Lack of any positive staining on incubation of the tissue slides with avidin and biotin prior to incubation with SA-PO ruled out the possibility that endogenous peroxidase was responsible for the observed staining pattern.

The human EGP-2 (Ep-CAM or pancarcinoma-associated protein 17–1A), encoded by the GA733–2 gene (26, 27), is expressed in a majority of human epithelial neoplasias. It has, therefore, been a target for immunotherapy and gene therapy strategies (2, 28). However, EGP-2 is an epithelial-differentiation antigen and not a tumor-specific antigen (12). Therefore, effective anti-EGP-2 therapy may cause severe side effects. Animals transgenic for EGP-2 with a distribution pattern of EGP-2 similar to that in humans may provide an appropriate experimental model from which we can learn about the potential efficacy, limitations, and safety of anti-EGP-2-directed immunotherapeutic strategies before initiating clinical trials. Here we present a transgenic mouse model expressing the EGP-2 antigen accurately on both normal and tumor tissue.

EGP-2 transgenic animal models have been generated before. Balzar et al. (12) reported on a transgenic mouse expressing EGP-2 cDNA under control of the mouse mammary tumor virus (MMTV) promoter. A transgenic rat expressing EGP-2 cDNA under control of the keratin 18 (K18) regulatory sequences was generated at our own laboratory (29). However, these EGP-2 transgenic animal models failed to express EGP-2 in a pattern similar to that observed in humans. A major disadvantage of these previously reported models is that EGP-2 is not expressed in the gastrointestinal tract nor the lung, and, in the mouse model, also not in the pancreas, whereas anti-EGP-2 immunotherapy is most frequently applied in patients with carcinomas derived from the epithelial cells of these tissues. To generate EGP-2 transgenic mice with a tissue-specific distribution pattern similar to the one observed in humans, we isolated the EGP-2 genomic sequences including its 5′ and 3′ sequences and used these to generate transgenic mice. Similar distribution patterns were observed in two independently generated EGP-transgenic mouse lines. FISH analysis demonstrated that this expression pattern was integration site-independent and, therefore, must be directed by 5′, 3′, and intron-specific regulatory sequences. This indicates that these sequences can indeed be used to direct epithelial cell-specific expression in vivo. To determine whether the EGP-2 regulatory sequences were capable of appropriate transmembrane expression of EGP-2, we used the high-affinity Mab MOC31, which recognizes an epitope in the first EGF-like repeat of the extracellular domain of the EGP-2 molecule (12, 30). Specific MOC31 Mab-binding to EGP-2 expressed on the membrane of normal epithelial tissues of EGP-2 transgenic mice gave a staining pattern that was very similar to the MOC31 staining pattern of human EGP-2-expressing normal epithelial tissues, suggesting accurate transmembrane expression of the transgene. The MOC31 Mab turns out to be very specific for the human EGP-2 antigen because no cross-reactivity was observed with the mouse homologue of EGP-2. This correlates with the previously described findings by Zaloudik et al. (14) who did not see any cross-reactivity of their Mab against human EGP-2 with various animal tissue, which did stain positive with the polyclonal antibody against human EGP-2.

The human EGP-2 protein consists of an extracellular domain with two EGF-like motifs, a transmembrane region of 23 hydrophobic amino acid residues, and a relatively short 26-residue highly charged cytoplasmic domain with an internalization motif (12). Several studies have pointed to a role for EGP-2 as a signaling molecule leading to regulation of proliferation and differentiation of epithelial cells and also to a morphoregulatory role (4, 31, 32). On transfection with EGP-2, cells incapable of intercellular adhesion formed aggregates, which suggested a homotypic adhesion function for EGP-2 (33). The exact role of EGP-2 in epithelial cells still remains to be elucidated, however. No evidence of additional or aberrant adhesion was found in the EGP-2 transgenic mice. Survival of the EGP-2 transgenic mice did not differ from nontransgenic littermates over a 20-month period. Expression of EGP-2 on the ovary duct and sertoli cells apparently did not affect the fertility of the transgenic animals, because female transgenic animals gave birth to viable transgenic offspring. These observations question the function of EGP-2 as a homotypic adhesion molecule. Neither neoplastic lesions nor morphological aberrations could be observed in the EGP-2 transgenic mice, in contrast to observations in the mammary gland of the MMTV-EGP-2 transgenic mice in which ductal hyperplasia was observed and differentiation of lobular and ductal cells was affected by the ectopically expressed human EGP-2 (12). The difference between these two EGP-2 transgenic mice models might be explained by differences in expression level, in localization of the EGP-2 protein, and in interactions with other proteins expressed in the tissues involved.

Our interest in producing an EGP-2 transgenic mouse lies in the desire to have an in vivo tumor model to study anti-EGP-2-directed immunotherapeutic modalities. A model frequently used to study anti-EGP-2 immunotherapy is the EGP-2-transfected B16.F10 tumor model (23). To adapt this C57/BL6 tumor model to the EGP-2 transgenic FVB/N mice, transgenic FVB/N/C57/BL6 F1 hybrids were generated. In these mice, both the EGP-2-negative parental B16.F10 tumor and the EGP-2-positive B16.C215 tumor could be induced, the latter without loss of EGP-2 expression. Successful Mab-based anti-EGP-2 immunotherapy depends on the ability of the Mab to discriminate in vivo between the antigen expressed by the tumor and the antigen expressed by normal tissue (28). To determine whether this essential property of an immunotherapeutic strategy could be studied in our EGP-2 transgenic B16.C215 tumor model, the high-affinity anti-EGP-2 antibody MOC31^bio was injected i.v. Immunohistochemical analyses revealed that i.v. applied MOC31^bio localized specifically to the EGP-2-positive tumor and not to the EGP-2 negative tumor. Furthermore, no binding was observed of the i.v. applied MOC31^bio to the EGP-2 expressed on the membrane of colon, lung, small intestine, and pancreas. In contrast, direct application of anti-EGP-2 on frozen-tissue slides of these organs led to a ready detection of EGP-2 in these tissues. These results demonstrate that, in this EGP-2 transgenic mouse tumor model, one can indeed discriminate between EGP-2 expressed on the tumor and EGP-2 expressed on normal tissue. This indicates that the EGP-2 expressed on normal tissue is not readily accessible for antibodies. Presumably, as has been suggested to be the case in normal tissue in humans, as assessed by radioimmunolocalization of MOC31 on lung cancer patients (34), the basal lamina shields the circulation. The presence of endogenous biotin in several other organs prevented the study of MOC31^bio and Bly-1^bio localization in these organs. Future studies using, e.g., humanized anti-EGP-2 antibodies and anti-EGP-2 single-chain variable region fragments, may provide more insight on the in vivo behavior of anti-EGP-2 antibodies in these organs.

Using both human tissue culture cells and animal models, it has been established that expression or overexpression of EGP-2 correlates with both benign and malignant proliferation of epithelial cells (4, 32, 35, 36, 37). The EGP-2-transgenic-mouse tumor model presented here provides an excellent tool to study this dualistic role of EGP-2 in tumor development and the additional signaling responsible for either phenotype, particularly when the EGP-2 transgenic mice are cross-bred with mice that are genetically predisposed to develop different types of tumors (38, 39, 40). In addition, because the endogenous EGP-2 regulatory sequences have been used to direct EGP-2 expression in these transgenic mice, they can also be used to evaluate the importance of EGP-2 during embryonic development or morphogenesis. Although several studies suggest an important role for EGP-2 during embryogenesis (12, 41), actually only limited information is available concerning the expression of the EGP-2 gene during human embryonic development. Anticancer therapies targeting the EGP-2 antigen require an appropriate preclinical model for investigators to study their efficacy and possible toxicity. Because our EGP-2-transgenic-mice tumor model expresses EGP-2 in a pattern similar to that of humans, in normal and tumor tissue, it may serve as a means to better understand the function of EGP-2 in development, maintenance, and tumorigenesis of epithelial tissues and as a useful tool for the development and evaluation of therapeutic strategies that use EGP-2 as a target.

We thank Hans Bartels for the i.v. injections of the mice, Mathieu Platteel, Hendrik E. Moorlag, Marianne L. C. van der Horst, and Anita E. Niemarkt for the tissue sectioning and the immunohistochemical stainings, and Jelleke Dokter for culturing the B16.F10 and B16.C215 cell lines.