In vitro and in vivo activity of MT201, a fully human monoclonal antibody for pancarcinoma treatment

In our study, a novel, fully human, recombinant monoclonal antibody of the IgG1 isotype, called MT201, was characterized for its binding properties, complement‐dependent (CDC) and antibody‐dependent cellular cytotoxicity (ADCC), as well as for its in vivo antitumor activity in a nude mouse model. MT201 was found to bind its target, the epithelial cell adhesion molecule (Ep‐CAM; also called 17‐1A antigen, KSA, EGP‐2, GA733‐2), with low affinity in a range similar to that of the clinically validated, murine monoclonal IgG2a antibody edrecolomab (Panorex®). MT201 exhibited Ep‐CAM‐specific CDC with a potency similar to that of edrecolomab. However, the efficacy of ADCC of MT201, as mediated by human immune effector cells, was by 2 orders of magnitude higher than that of edrecolomab. Addition of human serum reduced the ADCC of MT201 while it essentially abolished ADCC of edrecolomab within the concentration range tested. In a nude mouse xenograft model, growth of tumors derived from the human colon carcinoma line HT‐29 was significantly and comparably suppressed by MT201 and edrecolomab. The fully human nature and the improved ADCC of MT201 with human effector cells will make MT201 a promising candidate for the clinical development of a novel pan‐carcinoma antibody that is superior to edrecolomab. © 2002 Wiley‐Liss, Inc.

The use of humanized or murine/human chimeric monoclonal antibodies of the IgG1 subtype is now well established for the treatment of human cancers. Such antibodies are effective as monotherapy and in combination with chemotherapy. Treatment of advanced breast cancer by the HER-2-specific trastuzumab (Herceptin) and of follicular non-Hodgkin B-cell lymphoma by the CD20-specific rituximab (Mabthera, Rituxan) was shown to lead to increased overall survival. 1,2 A number of other promising IgG1 antibodies targeting EGFR for carcinoma and CD52, CD33 and CD22 for lymphoma and leukemia treatment are in late stages of clinical development. 3 Human IgG1 is thought to eliminate tumor cells by CDC, ADCC and, depending on the target, additionally by direct pro-apoptotic signaling or growth factor receptor antagonism. A recent study by Clynes et al. 4 suggested that ADCC is a major in vivo mechanism of IgG1 action.
A murine monoclonal IgG2a antibody called edrecolomab (17-1A; Panorex) was among the first monoclonal antibodies administered to humans for treatment of cancer. 5,6 Murine IgG2a is the functional equivalent of human IgG1, which to some extent can also exhibit ADCC and CDC with human effector cells and human complement, respectively. Edrecolomab was obtained by immunization of mice with human colon cancer cells and recognizes the pan-epithelial differentiation antigen Ep-CAM, 7-10 which is widely involved in homotypic cell adhesion of epithelial cells. Since then, numerous clinical trials have been undertaken with edrecolomab and other Ep-CAM-specific murine, chimeric and humanized monoclonal antibodies of various affinities. Ep-CAMspecific antibodies were also tested preclinically and clinically in the form of conjugates with toxins, radioisotopes and the cytokines IL-2 and GM-CSF. This vast experience with monoclonal antibodies and derivatives against the Ep-CAM target has however not yet been translated into an established antibody-based therapy in the clinic. The only exception is edrecolomab, which obtained market approval for a limited number of years in Germany. In an academically sponsored phase III trial, monotherapy with edrecolomab was shown to increase the overall survival after 10 years by 32% in patients with resected colorectal carcinoma at the Dukes C stage. 11,12 A recent, much larger phase III trial has shown that edrecolomab monotherapy was inferior after a 3-year observation period to the meanwhile introduced chemotherapy with 5-fluorouracil/leucovorin. 13 This led to the withdrawal of edrecolomab's market approval. Nevertheless, the edrecolomab monotherapy arm in the large phase III trial confirmed the overall survival observed in the Riethmüller study.
A persistent issue with the pan-carcinoma target Ep-CAM is its presence on normal epithelia as a tissue-specific cell adhesion molecule. 10,14 However, a recent study with transgenic mice expressing human Ep-CAM suggested that Ep-CAM is a valid tumor target based on its differential accessibility to parenterally administered Ep-CAM antibodies. 15 I.v.-injected anti-human Ep-CAM antibody did homogenously stain human Ep-CAM-expressing syngenic tumors in transgenic mice while no significant binding of the antibody to the Ep-CAM present on normal tissue such as colon, pancreas or lung was detected. These normal tissues were however strongly stained with the human Ep-CAM-specific antibody on sections prepared for immunohistochemical analysis. Recent immunohistochemistry studies of tumor samples from prostate, gastric and head and neck cancers have shown that Ep-CAM expression can increase with disease progression. 16 -19 This apparent overexpression may further widen the therapeutic window of the Ep-CAM target in certain indications.
The low-affinity anti-Ep-CAM antibody edrecolomab is fairly well tolerated in humans where doses of several grams were occasionally given without causing serious side effects. By contrast, the high-affinity antibody 3622/W94 had a low maximum tolerated dose of 30 mg and caused pancreatitis at higher concentrations. 20 -22 Conjugates of other Ep-CAM antibodies with toxins also showed significant side effects as is expected for antibodies attacking epithelia. 23 Edrecolomab and 3622W94 were reported to bind the same subdomain of Ep-CAM, 24 suggesting that affinity and not epitope recognition was responsible for the difference in tolerability. A drawback of edrecolomab for human therapy is its murine nature, resulting in a neutralizing immune response, short serum half-life, no option for chronic treatment and reduced compatibility with human immune effector mechanisms.
We therefore developed a fully human monoclonal antibody against Ep-CAM with the goal to preserve the safety profile of the low-affinity edrecolomab but to improve its antitumor activity. In our study, we characterized this novel antibody, called MT201 (HD69) 25 and compared its in vitro and in vivo properties with those of edrecolomab. We anticipate that the human monoclonal antibody MT201 has the potential to significantly improve upon the clinically validated murine monoclonal antibody edrecolomab not solely with respect to better pharmacokinetic properties resulting from its fully human nature, but also with respect to its higher efficacy of tumor cell elimination via human effector cells.

Cell lines and PBMC
CHO cells were purchased from the American Type Cell Culture Collection (ATCC, Rockville, MD). HT-29 cells, SW480, 22RV1 and LnCAP lines were obtained from the Deutsche Sammlung von Mikroorganismen und Zelllinien (DSMZ, Braunschweig, Germany) and KATO III from the European Collection of Cell Cultures (ECACC, Salisbury UK). Cells were cultured as recommended by the suppliers. A CHO cell clone transfected with human Ep-CAM cDNA was produced as described. 25 Peripheral blood mononuclear cells (PBMC) were prepared by Ficoll density centrifugation from enriched lymphocyte preparations (buffy coats) obtained from local blood banks. PBMC were prepared on the same day of buffy coat receipt. Erythrocytes were removed from PBMC by erythrocyte lysis buffer (155 mM NH 4 Cl, 10 mM KHC 3 and 100 M ethylenediamine tetraacetic acid; EDTA) and thrombocytes removed via the supernatant obtained after centrifugation of PBMC at 100g for 10 min PBMC were typically used between days 2 and 4 after preparation without additional stimulation.

Antibodies
MT201 was produced at Micromet AG (Martinstied, Germany) by a CHO cell clone and purified to homogeneity by protein-A affinity and anion exchange chromatography. 25 The murine monoclonal antibodies M79, 323/A3, 7C1 and 425 were kind gifts of Drs. P. Kufer and J. Johnson (Institute of Immunology, Munich University, Munich, Germany). Edrecolomab (Panorex) was purchased from GlaxoSmithKline, Munich, Germany. Chimeric 17-1A called C46 was a kind gift of Dr. Hakan Mellstedt of Karolinska University, Stockholm, Sweden. FITC-labeled goatanti-mouse IgG was from Pharmingen, Heidelberg, Germany (no. 554001) and used at a 1:20 dilution. FITC-labeled mouse antihuman IgG was purchased from ICN, Eschwege, Germany (no. 672171) and used at a dilution of 1:60.
For FITC-labeling of M79, MT201, 323/A3 and edrecolomab, the monoclonal antibodies were dialyzed against 2 L borate buffer (50 mM sodium borate, pH 8.3, 100 mM NaCl). One mg of antibody in 1 ml was reacted with 4 l of fluorescein-NHS (Sigma, Taufkirchen, Germany; 20 mg/ml stored in dimethyl sulfoxide [DMSO] at Ϫ20°C) and 26 l DMSO for 1 hr at room temperature. The mixture was then dialyzed overnight against 2 L of PBS, 0.1% sodium azide and stored in 100 l aliquots at 4°C.

Binding studies
Antibody binding to cells was studied by flow cytometry using a FACSCalibur instrument (Beckton Dickinson, Heidelberg, Germany) equipped with a 488 nm argon laser. Data were analyzed by Cellquest Software (Becton Dickinson). Cells were washed by FACS buffer containing phosphate-buffered saline, 1% fetal calf serum (FCS) and 0.05% sodium azide. Flow and cleaning solutions were purchased from Becton Dickinson and used according to the manufacturer's instructions. FACS data were quantitated as histograms by determining the mean fluorescence intensity (MFI) as proposed by Diamond and Demaggio. 26 Binding competition studies were done in 2 series. In 1 series, FITC-labeled antibodies (10 g/ml) and varying concentrations of unlabeled competitor antibodies were premixed before addition to cells. In a second series, fluorescently labeled antibodies were allowed to first bind to cells at 4°C for 30 min before addition of unlabeled antibodies. Antibody binding to 120,000 -200,000 human Ep-CAM cDNA-transfected CHO cells was performed in 30 -50 l FACS buffer at room temperature for 30 -60 min. Ten thousand events were recorded by FACS and MFI determined.

Determination of binding constants
For Scatchard analysis, the binding of MT201 to human gastric carcinoma line KATO III was titrated using FACS analysis essentially as described by Krause et al. 27 MT201 was detected by a sandwich of unlabeled mouse-anti-human IgG (Pharmingen; no. 34161D) and FITC-labeled goat-anti-mouse IgG (Pharmingen; each at a concentration of 25 g/ml. Binding reactions with 10 different concentrations of MT201 were performed in 50 l with 200,000 cells for 30 min at 4°C followed by 2 washes in FACS buffer. Secondary and tertiary antibodies were subsequently incubated for 30 min at 4°C followed by washes and flow cytometric analysis. The sandwich detection of MT201 was necessary to quantitate the number of bound MT201 by using the QIFIKIT kit (DAKO, Glostrup, Denmark; no. K0078). This kit allows one to quantitate by FACS the number of bound murine antibody to calibrated beads with 3,200 -689,000 bound IgG molecules. A strictly linear relationship of MFI and bound IgG was obtained. Data were analysed by Scatchard plotting and direct KD determination from binding curves using the PRISM software program version 3.02 (GraphPad Software, San Diego, CA).
Biacore analysis was performed by Inventus BioTec GmbH (Muenster, Germany) using a Biacore 2000 reader (Applied Biosystems, Uppsala, Sweden). Soluble, recombinant extracellular domain of human Ep-CAM was produced and purified from the supernatant of stably transfected CHO cells. 25 Ep-CAM protein was coated to CM5 flow cells (Becton Dickinson) using the Amine Coupling Kit as described by the manufacturer. Binding studies of MT201 and edrecolomab were performed in a running buffer containing 10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA and 0.005% surfactant (P-20). K D and rate constants were determined from sensorgrams collected with 12 different antibody concentrations.

CDC assay
The FACS-based assay used 50,000 KATO III cells per reaction. One hundred microliters of trypsinized, medium-washed KATO III cells in RPMI/10% FCS were reacted with 20 l antibody solution. The CDC reaction was started by addition of 80 l of a solution containing 25% human serum in RPMI/10% FCS, resulting in a final concentration of 10% human serum with active complement. Human serum was collected from healthy donors. An aliquot was treated at 56°C and served as a control containing inactivated complement. Incubation was for 2 hr at 37°C in an atmosphere of 5% CO 2 . Cells were collected by centrifugation and resuspended in FACS buffer containing 1 g/ml propidium iodide (PI). Ten thousend events were collected by flow cytometry. CDC was determined from the number of surviving PI-negative cells in gate R1.

ADCC assay
Immune effector cells (PBMC) were prepared as described above. One million target cells were labeled with 0.2 M calcein AM (Molecular Probes, Göttingen, Germany; no. C-1430) for 30 min at 37°C in cell culture medium. After 2 washes in PBS, a cell density of 5 ϫ 10 5 cells/ml was adjusted in RPMI/10% FCS and 100 l aliquots of 50,000 cells used per assay reaction. Cultured PBMC were washed in PBS followed by RPMI/10% FCS and adjusted by dilution with RPMI/10% FCS to the desired E:T ratio. Antibodies were diluted in RPMI/10% FCS to the required concentration. If not otherwise indicated, a standard reaction at 37°C/5% CO 2 was for 4 hr and used 50,000 calcein-AM-labeled target cells, 1 million PBMC (E:T ratio of 1:20) and 20 l antibody in a total volume of 100 l. After the reaction, cells were collected by centrifugation and resuspended in FACS buffer containing 1 g/ml PI. One hundred thousand events were collected by flow cytometry.
Quantitation of cytotoxicity was based on the number of alive and dead target cells in the control reaction without antibody. This was necessary because effector cells tend to take up calcein AM released from dead target cells, in particular, during longer incubation periods. Where indicated, either specific or overall ADCC was determined. Sigmoidal dose response curves typically had r 2 values Ͼ0.95 as determined by Prism Software.

Animal experiments
In-house bred, male athymic NMRI:nu/nu mice with a body weight of 30 Ϯ 1 g and an age of 6 -8 weeks were used for testing the effect of MT201, edrecolomab and C46 on tumor growth. The mice were held under sterile standardized conditions. These were 22°C, 50 Ϯ 5% relative humidity, a 12 hr light/dark rhythm, autoclaved food and bedding (by SSniff, Soest, Germany) and ad libitum acidified tap water. All animal experiments were performed according to the German Animal Protection Law with permission from the responsible local authority. HT-29 colon carcinoma cells were produced by cell culture. One million cells in a volume of 0.1 ml PBS were injected into the left flank of nude mice on day 0. Treatment with antibodies at 30 g in 0.1 ml PBS started on day 1 after tumor-cell inoculation and was repeated at days 4 and 7. Antibodies were administered by the tail vein. As controls, PBS and rituximab (30 g/dose) were used. Statistical analysis of tumor growth was performed by the Mann-Whitney test. The significance level was p Յ 0.05.

Binding specificity of MT201
The cell-binding properties of MT201 were investigated by flow cytometry. A cell line not expressing human Ep-CAM (chinese hamster ovary, CHO), CHO cells stably transfected with human Ep-CAM cDNA (Ep-CAM-CHO) 25 and a human gastric carcinoma line naturally expressing Ep-CAM (KATO III) were tested for MT201 binding. Figure 1a shows FACS histograms of the 3 cell lines in the presence of 1.6 and 100 g/ml MT201, as detected by a secondary FITC-labeled mouse-anti-human IgG. Essentially no binding of MT201 to rodent CHO cells was detected (Fig. 1a, upper panel). However, strong and dose-dependent MT201 binding was seen with the Ep-CAM cDNA-transfected CHO line (middle panel). KATO III also showed binding of MT201 (lower panel). A fine titration of antibody binding suggests that CHO-Ep-CAM cells expressed much higher levels of Ep-CAM than KATO III cells, while no binding to untransfected CHO cells was detected up to a concentration of 400 g/ml MT201 (Fig. 1b).
A panel of additional human carcinoma-derived cell lines was tested for antibody binding by FACS. MT201 binding was detected with the colon cancer lines HT-29 and SW480 and the prostate cancer lines 22RV1 and LnCAP (data not shown). Likewise, we observed binding of MT201 to human monocytic cells and a subpopulation of lymphocytes (most likely NK cells) that was presumably mediated by Fc␥ receptor binding of MT201 to those immune cells (data not shown).

Low-affinity Ep-CAM binding of MT201 is similar to that of edrecolomab
Binding constants for MT201 were determined by both a FACSbased binding assay using KATO III cells and a plasmon resonance-based assay with the Biacore analyzer. Scatchard analysis of cell binding data yielded a mean dissociation constant for MT201 of 1.38 ϫ 10 Ϫ7 M (r 2 ϭ 0.9423). Biacore analysis using sensor chips coated with the recombinant extracellular domain of human Ep-CAM yielded a dissociation constant of 1.75 ϫ 10 Ϫ7 M, a k on of 2.63 ϫ 10 5 M Ϫ1 s Ϫ1 and a k off of 4.61 ϫ 10 Ϫ2 s Ϫ1 . With these affinity and rate constants, MT201 exhibited similar binding characteristics as the clinically validated murine monoclonal antibody edrecolomab. Under identical Biacore assay conditions, a dissoci-

Epitope recognition of MT201 relative to other anti-Ep-CAM MAbs
To investigate the relatedness of epitopes, the binding of MT201 to Ep-CAM-transfected CHO cells was studied in the presence of increasing amounts of 3 other Ep-CAM-specific monoclonal antibodies. The 3 other mAbs were M79, 28 the murine monoclonal antibody used to isolate variable domains of MT201 by chain shuffling, 25 edrecolomab and 323/A3, a high-affinity murine monoclonal antibody 29 used to develop the clinically tested, humanized monoclonal antibody 3622/W94. Each antibody was labeled by activated fluorescein and binding competition studied by FACS in the presence of increasing concentrations of respective unlabeled antibodies either after prebinding of the FITC-labeled antibody to cells at 4°C (Fig. 2, right panels) or by simultaneous addition of FITC-labeled and unlabeled antibodies (left panels). Murine antibodies against HER-2 (7C1) or EGFR (mAb 425) served as nonspecific controls and did not show significant binding competition throughout the experiment (Fig. 2).
Binding of FITC-labeled MT201 was best competed by 323/A3 followed by M79 and self (Fig. 2a,b). Edrecolomab exhibited the weakest inhibition of MT201 binding. Cell binding of FITClabeled M79 was very similarly competed by MT201, edrecolomab and self and best by 323/A3 (Fig. 2c,d). FITC-labeled edrecolomab was well competed by all antibodies except for MT201 (Fig. 2e,f). Binding of FITC-labeled 323/A3 was very weakly competed by the low-affinity antibodies MT201, edrecolomab and M79, but was efficiently competed by self (Fig. 2g,h).
In conclusion, the mutual competition of all 4 tested antibodies suggests their binding to a common subdomain of the extracellular domain of human Ep-CAM. The efficiency and dose-dependence of binding competition was consistent with the relative affinities of mAbs for Ep-CAM. For 323/A3 and edrecolomab, the subdomain of Ep-CAM was previously determined to be the immunodominant, outer EGF-like domain I. 24 Because both M79 and 323/A3 did compete efficiently with all other antibodies, it appears they recognized epitopes on the same Ep-CAM subdomain that are overlapping with those recognized by MT201 and edrecolomab. Within this Ep-CAM subdomain, MT201 and edrecolomab apparently recognized the most distant epitopes, as was evident from their weakest cross-competition among the 4 anti-Ep-CAM mAbs tested.

MT201 and edrecolomab exhibit comparable CDC
CDC of MT201 was investigated in a FACS-based assay (Fig.  3). This assay monitors the lysis of KATO III cells by the uptake of propidium iodide (PI). At a final human serum concentration of 10%, a significant and dose-dependent tumor cell lysis was observed in the presence of 6.3 or 100 g/ml MT201 (Fig. 3a). This was evident by the appearance of a PI-positive cell population and the concomitant disappearance of the PI-negative alive cell population in the shown gate. Heat-inactivated human serum did not lead to significant cell lysis in the presence of 100 g/ml MT201. The human IgG1 MT201 and the murine IgG2a edrecolomab exhibited a comparable dose response of CDC with human serum. Both antibodies reached half-maximal cell lysis at approximately 10 g/ml (Fig. 3b). Another antibody with a human IgG1 Fc domain that is specific for CD20 (rituximab) was not cytotoxic against KATO III cells in the presence of human serum.
CDC of MT201 was specific for Ep-CAM. The viability of CHO cells was not affected up to 100 g/ml of MT201 in the presence of human serum (Fig. 3c). However, CHO cells expressing human Ep-CAM were very efficiently eliminated by MT201. KATO III cells required an approximately 10-fold higher concentration of MT201 to be lysed by CDC to the same extent as CHO-Ep-CAM cells. This correlates with the lower level of MT201 binding to KATO III cells (Fig. 1b).
A number of additional Ep-CAM-positive human tumor cell lines were tested and exhibited a much reduced susceptibility to CDC by MT201 compared to KATO III cells (data not shown). This may be related to the expression of complement resistance factors on tumor cells such as CD46, CD55 and CD59, of which CD55 and CD59 are absent from KATO III cells increasing their CDC susceptibility. 30

ADCC of MT201
We established a FACS-based ADCC assay monitoring the fate of target cells over the high background of human PBMC used as effector cells. To this end, KATO III cells were loaded with the fluorescent dye calcein AM, which is trapped inside live cells through esterase-mediated cleavage. 31 Labeled, live target cells could be efficiently separated in FL-1 from unlabeled PBMC (Fig.  4a, right gate). PI allowed for separation and quantitation of calcein AM-labeled dead target cells in the upper gate (Fig. 4a). Upon incubation with MT201 at an effector:target cell (E:T) ratio of 20:1, a dose-dependent decrease of live cells in the right gate and a concomitant increase of PI-positive cells in the upper gate was observed. A short 4 hr assay period was chosen to also observe distinct levels of overall tumor cell lysis in the plateau phase of dose-response plots.
ADCC by MT201 was specific. Figure 4b shows that no ADCC was observed with CHO cells unless this rodent cell line was stably transfected with human Ep-CAM cDNA. In contrast to CDC (Fig. 3), KATO III cells were significantly more sensitive to ADCC than Ep-CAM-transfected CHO cells (Fig. 4b). The halfmaximal doses (ED 50 values) for ADCC were below 1 g/ml.
ADCC of MT201 was dependent on the E:T ratio (Fig. 4c). Essentially no cell lysis by MT201 was observed in the absence of effector cells. With increasing ratios of human PBMC, the extent of cell lysis, as seen in the plateau phase, increased. Although it doubled with an E:T ratio going from 10:1 to 20:1, not much of an increase was seen upon a further doubling to 40:1. The ED 50 values of ADCC were not strongly affected by the various E:T ratios.

PBMC donor variation strongly affects ADCC
PBMC were prepared from various healthy donors and all tested for ADCC in the presence of MT201 at the same E:T ratio of 20:1. Donor PBMCs exhibited a significant variation of ADCC with respect to the extent of nonspecific and specific tumor cell lysis, as well as ED 50 values. Figure 5a compares the dose-response behaviour of ADCC by 6 distinct donor PBMCs. More than 10-fold variations among donors were observed. A statistical analysis (Fig. 5b) compares ED 50 values obtained from dose-response analyses of a total of 23 healthy donors. Two-thirds of donor PBMCs gave ED 50 values between 100 and 500 ng/ml MT201 and only 4 donor PBMC gave ED 50 values Ͼ1 g/ml.

MT201 exhibits much higher ADCC than edrecolomab
ADCC of MT201 and edrecolomab was compared using KATO III as target cells (E:T ratio of 10:1 in a 20 hr assay). A highly active donor PBMC was selected to see an optimal performance of edrecolomab. These PBMC gave an ED 50 value of ADCC for MT201 below 1 ng/ml (Fig. 6a). Under identical conditions, edrecolomab exhibited an ED 50 value Ͼ100 ng/ml. A similarly large difference in activity between MT201 and edrecolomab was seen with less-active PBMC effectors (data not shown).
Upon addition of 50% human serum, a source of extra human IgG, the ADCC of MT201 was reduced by 2 orders of magnitude from an ED 50 value of 1 ng/ml to approximately 100 ng/ml (Fig.  6b). Under identical conditions, human serum essentially abolished the ADCC of edrecolomab (Fig. 6c). No significant cell lysis was seen up to 100 g/ml edrecolomab, a concentration of the antibody that is unlikely to be exceeded much in vivo.

Inhibition of tumor growth by MT201 in a nude mouse model
The in vivo efficacy of MT201 against tumors was investigated in a nude mouse model using the Ep-CAM-positive human colon carcinoma line HT-29. KATO III cells, which worked well in vitro as a target cell line, did not develop tumors in nude mice. MT201 or controls were administered via the tail vein on days 1, 4 and 7 following the subcutaneous (s.c.) injection of 1 ϫ 10 6 HT-29 tumor cells. Solid-tumor growth to a mean size exceeding 0.35 cm 3 was observed with both the PBS control and an IgG1 isotype control not recognizing HT-29 cells (i.e., CD20-specific rituximab) (Fig. 7a). Administration of 3 times 30 g MT201 led to a statistically significant inhibition of tumor growth (p ϭ 0.05) that lasted for the entire observation period of 37 days. Three times 3 g MT201 also caused a reduction in tumor growth but the effect did not reach statistical significance. The 30 g dose of MT201 was diluted for treatment of mice either in PBS or in a buffer containing 1% of the excipient polyvinylpyrrolidone (PVP-17) that was used for formulation of clinical test material. PVP did not significantly affect the in vivo efficacy of MT201.
The in vivo efficacy of MT201 was compared to that of murine edrecolomab (IgG2a) and C46, a chimerized version of edrecolomab with a human IgG1 backbone. 32 Thirty-microgram doses of all 3 antibodies given on days 1, 4 and 7 after s.c. inoculation of 10 6 tumor cells did inhibit tumor growth in the HT-29 nude mouse model (Fig. 7b). No statistically significant difference was obtained when the activities of 3 antibodies were compared among each other. There was, however, a statistically significant difference when the activity of the antibodies was compared to the PBS and IgG1 isotype (rituximab) controls. Edrecolomab and MT201 gave a number of significant readings (p Յ 0.05; marked by asterisks), whereas C46 exhibited only a significant reading on day 21. The difference between edrecolomab and C46 may reflect the different efficiency of murine IgG2a and human IgG1 Fc effector domains in a murine background. MT201, which performed in the mouse background similarly well as edrecolomab, may thus be in a human background superior to edrecolomab and chimerized edrecolomab.

DISCUSSION
In our study, we characterized in detail the human monoclonal antibody MT201 for Ep-CAM-specific cell binding, epitope recognition in relationship to other well-characterized anti-Ep-CAM monoclonal antibodies, CDC and ADCC and inhibition of tumor take in a nude mouse model. The in vitro and in vivo data suggest that the human monoclonal antibody MT201 has characteristics desirable for the development of a next-generation anti-Ep-CAM antibody: low affinity, potent ADCC with human effector cells, CDC activity and inhibition of tumor growth in vivo. Moreover, the overall sequence identity of 95% of VH and VL domains of MT201 to human germline sequences would predict a very low immunogenicity in man.
Despite many disappointing efforts, Ep-CAM still appears as an attractive pan-carcinoma target for antibody-based therapies. Unlike the clinically validated antibody targets HER-2 and EGFR, Ep-CAM is not overexpressed on tumor cells as a consequence of gene amplification but, as an epithelial differentiation antigen, is constitutively expressed on almost all carcinomas. HER-2-positive breast tumors only occur in 15-30% of patients. 33 The frequency of EGFR overexpression in breast cancer is ca 40% 34 and somewhat higher in other tumors. 35 By contrast, Ep-CAM expression on certain tumors such as prostate and gastric carcinoma reaches 100% as examined by immunohistochemistry. 16,36 Moreover, Ep-CAM was found to be significantly upregulated with disease progression in several cancers including breast cancer, [17][18][19]37 suggesting that Ep-CAM expression on tumor cells confers a growth, invasion and/or survival benefit. A recent study showed that Ep-CAM can interact with the negative regulatory receptor LAIR-1 present on most immune cells. 38 This indicates that Ep-CAM expression may confer a general suppression of immune cell activity in the microenvironment of the tumor. The widespread expression of Ep-CAM with respect to carcinoma type, frequency in patient population and on cells of a given tumor make Ep-CAM a unique and promising pan-carcinoma target. Additional efforts are thus warranted to optimally exploit the therapeutic window offered by the differential accessibility of Ep-CAM to antibodies, 15 its widespread expression and its upregulation on certain tumors.  The vast clinical experience with the murine monoclonal antibody edrecolomab and its unique properties are pointing the way to an optimized Ep-CAM-specific antibody therapy. Ep-CAMspecific human IgG1 appears particularly useful for treatment of minimal residual disease (micrometastasis) where low tumor load and small tumor size do not overburden the limited efficacy and penetration capacity of "naked" IgG. With more advanced stages of cancer, a combination of IgG with chemotherapy may help reduce tumor load and increase the tumor's accessibility for antibodies. Currently, MT201 is the only anti-Ep-CAM monoclonal antibody of human origin under clinical development. Its VH domain was isolated from a human B-cell repertoire that has not yet undergone somatic mutations and deletion of autoreactive specificities. 25 We determined by Scatchard and Biacore analysis an affinity of MT201 for Ep-CAM in the 100 nM range. This affinity for Ep-CAM was only slightly higher than that of edrecolomab. We thus reached our goal of obtaining a fully human IgG1 with an affinity close to that of the clinically validated edrecolomab.
Competition analysis of MT201 with the Ep-CAM-specific monoclonals edrecolomab, M79 and 323/A3 suggests that all 4 monoclonal antibodies recognize human Ep-CAM in a very related fashion. We observed that edrecolomab binding was efficiently competed by 323/A3 and vice versa, suggesting that their epitopes on the EGF-like domain I are essentially overlapping. The distinct toxicity of the 2 antibodies in man may thus be related to their Ͼ100-fold difference in affinity rather than the recognition of distinct epitopes. High-affinity anti-Ep-CAM antibodies may interfere as antagonistic autoantibodies with the adhesion function of Ep-CAM in normal epithelium, resulting in perturbed epithelial integrity. The increased levels of pancreatic enzymes in blood observed with the humanized version of 323/A2 would be consistent with this hypothesis. Epithelial damage by the humanized version of 323/A3 could also be a consequence of CDC-or ADCC-mediated damage to epithelial cells. However, this should then also be observed for edrecolomab, which does not significantly increase amylase levels in blood.
Although edrecolomab and MT201 only very weakly competed each other for their binding to Ep-CAM, both were well competed by 323/A3 and M79. This indicated that the epitopes recognized by MT201 and edrecolomab were closely adjacent though not strictly overlapping. Binding studies with truncated Ep-CAM showed that both 323/A3 and edrecolomab do bind the immunodominant EGFlike domain I of Ep-CAM. 10,24 We infer from the competition data that M79 and MT201 also recognized this subdomain, although this conclusion may require further support from binding studies using truncated Ep-CAM versions.
Although the cytotoxicity of MT201 mediated by ADCC and CDC was substantial, MT201 alone did not show any detectable effect on the viability of Ep-CAM-positive human tumor cell lines in the absence of complement (human serum) or human immune effector cells. The monoclonal antibody does therefore not belong to the class of antitumor antibodies that directly elicit pro-apoptotic signaling or growth inhibition by recognition of a growth factor receptor. We do not know whether MT201 binding to Ep-CAM produces any kind of intracellular signal and it is currently not known whether any of the other anti-Ep-CAM antibodies does. It is interesting to note that in vitro studies suggested that both trastuzumab and rituximab render tumor cells apoptotic by inhibition of growth and survival signals. 39,40 However, in Fc␥ KO mice rendered incapable of ADCC, both antibodies lost most of their antitumor activity against tumor cell lines that are otherwise susceptible to the pro-apoptotic activity of the antibodies in vitro. 4 We noted a substantially higher ADCC of MT201 compared to edrecolomab. This may not be related to the slightly higher binding affinity of MT201 because the CDC activity of the 2 antibodies was very similar. We rather attribute the substantial difference in ADCC to an improved binding of MT201 to the Fc␥ receptors of human effector cells that may be less well activated by the murine IgG2a edrecolomab. When human serum was added to ADCC reactions, ADCC was strongly reduced. A likely interpretation of this inhibition is that the extra human IgG in serum was competing for MT201 binding to Fc␥ receptors. In the case of edrecolomab, this virtually abolished ADCC with human effector cells, suggesting that under physiologic conditions edrecolomab cannot elicit substantial ADCC but may largely act against tumor cells via CDC. The in vitro data indicate that in human serum MT201 and edrecolomab may require concentrations Ͼ1 g/ml to efficiently eliminate tumor cells via ADCC or CDC.
We expect that edrecolomab is more efficient with murine than human effector cells, whereas the opposite may be the case for MT201. The in vivo activity of MT201 in the nude mouse model suggests that the human antibody is also effective with murine effector cells and/or complement. Further studies are necessary to explore the compatibility of human and murine IgG with human and murine cytotoxic effector mechanisms in order to better understand the validity of in vivo and in vitro model systems that cross species barriers. Pharmacokinetic behaviour adds even more differences to the 2 systems. Compared to human IgG1, the serum half-life of murine IgG2a within humans is markedly reduced apparently because of its low affinity to FcRn. 41 Likewise, the half-life of human IgG isotypes is much reduced in mice. 42 On these notions, a direct comparison between edrecolomab and MT201 in a mouse background seems to provide limited information with respect to the situation in humans. The animal data would predict that MT201, which in mice was similarly potent as edrecolomab, may perform far better in humans because of a longer half-life and improved ADCC with human effector cells. Our in vitro ADCC experiment in the presence of human serum and testing human effector cells from various donors may have a better predictive potential about the in vivo efficacy of MT201 (and edrecolomab) in humans than the nude mouse model.
A chimeric version of edrecolomab with a human IgG1 Fc part called C46 32 was clinically tested in combination with the immune-stimulatory cytokine GM-CSF. 43,44 As expected, the inci-dence of an immune response against the chimeric form was reduced compared to the murine version even in the presence of GM-CSF. Of note, dosing and side effects associated with the chimeric version were not reported to be strikingly different from those of murine edrecolomab, although the chimeric edrecolomab is expected to better interact with human immune cells and therefore has improved ADCC activity. This supports the notion that side effects of high-affinity Ep-CAM monoclonal antibodies are associated with an antagonistic effect on Ep-CAM function rather than increased ADCC or CDC activity against epithelium. An ongoing phase I trial with MT201 will reveal the side effect profile of MT201, its immunogenicity, serum half-life and perhaps a maximally tolerated dose.