Immunotoxin targeting EpCAM effectively inhibits peritoneal tumor growth in experimental models of mucinous peritoneal surface malignancies

Authors

  • Kjersti Flatmark,

    Corresponding author
    1. Department of Gastroenterological Surgery, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    • Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Ingrid J. Guldvik,

    1. Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Hege Svensson,

    1. Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Karianne G. Fleten,

    1. Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Vivi Ann Flørenes,

    1. Department of Pathology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Wenche Reed,

    1. Research Innovation and Education, Oslo University Hospital, Oslo, Norway
    Search for more papers by this author
  • Karl-Erik Giercksky,

    1. Department of Gastroenterological Surgery, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    2. University of Oslo, Institute of Clinical Medicine, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Øystein Fodstad,

    1. Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    2. University of Oslo, Institute of Clinical Medicine, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author
  • Yvonne Andersson

    1. Department of Tumor Biology, Norwegian Radium Hospital, Oslo University Hospital, Montebello, Oslo, Norway
    Search for more papers by this author

  • Conflict of interest: None

Correspondence to: Kjersti Flatmark, Department of Tumor Biology, The Norwegian Radium Hospital, Oslo University Hospital, Montebello, 0310 Oslo, Norway, Tel.: +47–22934000, Fax: +47–22781895, E-mail: kjersti.flatmark@rr-research.no

Abstract

Cytoreductive surgery and intraperitoneal (i.p.) chemotherapy constitute a curative treatment option in mucinous peritoneal surface malignancies of intestinal origin, but treatment outcome is highly variable and the search for novel therapies is warranted. Immunotoxins are attractive candidates for targeted therapy in the peritoneal cavity because of direct cytotoxicity, distinct mechanisms of action and tumor cell selectivity. The MOC31PE immunotoxin targets the tumor-associated adhesion protein EpCAM (Epithelial Cell Adhesion Molecule), and has been administered safely in early clinical trials. In our work, the efficacy of i.p. administration of MOC31PE alone and together with mitomycin C (MMC) was investigated in unique animal models of human mucinous peritoneal surface malignancies. In initial model validation experiments, clear differences in efficacy were demonstrated between MMC and oxaliplatin, favoring MMC in five investigated tumor models. Subsequently, MOC31PE and MMC were given as single i.p. injections alone and in combination. In the PMCA-2 model, moderate growth inhibition was obtained with both drugs, while the combination resulted in at least additive effects; whereas the PMP-2 model was highly sensitive to both drugs separately and in combination and intermediate sensitivity was found for the PMCA-3 model. Furthermore, results from ex vivo experiments on freshly obtained mucinous tumor tissue from animals and patients suggested that classic mechanisms of immunotoxin activity were involved, i.e., inhibition of protein synthesis and induction of apoptosis. The present results suggest that adding MOC31PE to MMC-based i.p. chemotherapy should be further explored for EpCAM-expressing peritoneal surface malignancies, and a phase I trial is in preparation.

Mucinous peritoneal surface malignancies of intestinal origin encompass a range of clinical presentations, from the clinically benign manifestation pseudomyxoma peritonei (PMP) to aggressive mucinous peritoneal carcinomatosis from colorectal cancer. Histopathological presentation varies, with increasing cellular atypia, high epithelium to mucin ratio and signet ring cell differentiation being associated with an aggressive phenotype. The treatment is, regardless of histology, aimed at complete surgical removal of all visible tumor deposits followed by chemotherapy, typically some type of perioperative intraperitoneal (i.p.) chemotherapy, often supplemented with postoperative systemic chemotherapy. Prognosis is highly dependent on histopathological subtype, and particularly in peritoneal mucinous carcinomatosis (PMCA), the search for more efficacious chemotherapy regimens is warranted.[1-3] Although the benefit of cytoreductive surgery is well documented in mucinous peritoneal surface malignancies, the optimal chemotherapeutic strategy has not been proven.[2-4] Several drug combinations and delivery strategies are currently being pursued, mostly in early clinical trials, while preclinical assessment of drug efficacy has been impeded by lack of appropriate experimental models.

The MOC31PE immunotoxin is composed of the MOC31 monoclonal antibody targeting the tumor-associated antigen EpCAM (Epithelial Cell Adhesion Molecule), covalently linked to pseudomonas exotoxin (PE) A. The antibody targets the immunotoxin to EpCAM-expressing cells, and when internalized, the toxin effector moiety triggers cell death by catalytic inactivation of vital processes, such as protein synthesis, and by directly inducing apoptosis.[5-7] Different treatment strategies to target EpCAM for killing cancer cells have been tested preclinically and in clinical trials.[8, 9] The efficacy of MOC31PE, in particular, has been extensively studied in preclinical human tumor models in vitro and in vivo,[10] and in a recently completed phase I clinical trial the clinical preparation was well tolerated upon repeated intravenous administrations (ClinicalTrials.gov #NCT01061645, manuscript in preparation). The potential for selective targeting and killing of tumor cells combined with distinct mechanisms of cytotoxicity suggests that immunotoxins may be an interesting class of agents for i.p. treatment of peritoneal surface malignancies in combination with conventional chemotherapeutic drugs. MOC31PE is particularly interesting in the current setting, both because of its clinical tolerability and because EpCAM is highly expressed in gastrointestinal malignancies.[11]

We previously reported the establishment, propagation and validation of five orthotopic animal models of human mucinous peritoneal surface malignancies.[12, 13] In our work, the utility of the models for studying experimental treatment was demonstrated by comparing the efficacy of two drugs commonly used for i.p. cancer treatment in humans, mitomycin C (MMC) and oxaliplatin (OXA). Furthermore, the efficacy of single i.p. injections of MOC31PE immunotoxin on tumor growth was investigated in two of the models alone and in combination with MMC. Finally, potential mechanisms for the observed immunotoxin effects were studied in ex vivo experiments.

Material and Methods

Clinical samples

The clinical samples used for establishment of the animal models from human tumor tissues were previously described.[12, 13] In the present work, mucinous tumor tissue harvested at the time of surgery from two additional patients with clinical PMP was used in short-term cultures to study ex vivo immunotoxin efficacy and cell-death mechanisms. The study was approved by the regional ethics board of south-east Norway and written informed consent was obtained from the patients. The PMP-3 sample was obtained from a 40-year-old woman who previously had been subjected to appendectomy (for a ruptured mucinous cystadenoma) and bilateral salpingo-oophorectomy (for ovarian recurrence), 7 and 2 years previously. A widespread intra-abdominal recurrence was detected and histologically classified as PMCA of intermediate features (according to the Ronnett classification[1]). The PMP-6 sample, also harvested at the time of cytoreductive surgery, was from a 41-year-old male who 1 year previously had an appendectomy for a mucinous cystadenoma of the appendix that ruptured during surgery. He presented with an extensive peritoneal recurrence, histologically classified as disseminated peritoneal adenomucinosis. Both patients were subjected to extensive cytoreductive surgery, accomplishing complete cytoreduction, and they received hyperthermic i.p. chemotherapy with MMC.

Drugs

Stock solutions of MMC (Medac, Wedel, Germany) and OXA (Sanofi Aventis, Paris, France) were prepared in 5% glucose (Fresenius Kabi AB, Uppsala, Sweden) and stored as recommended by the drug manufacturers. The MOC31 antibody (MCA Development, Groningen, The Netherlands) detects the epithelial glycoprotein EpCAM. The immunotoxin was constructed by conjugating the MOC31 antibody to PE (obtained from Dr Darrel Galloway, University of Ohio, Columbus, OH) by a thioether bond formed with the reagent sulfo-SMCC [sulfo-succinimidyl-4-(N-maleimidomethyl) cyclohexane-1-carboxylate] (Pierce, Rockford, IL), as described earlier.[14] Clinical-grade MOC31PE was produced at Fred Hutchinson Cancer Centre (Seattle, CA) and was dissolved and diluted (when applicable) in phosphate-buffered saline (PBS) with 0.1% human serum albumin. For animal experiments, stock solutions of MMC and OXA were diluted in 5% glucose (Fresenius Kabi AB, Uppsala, Sweden).

Animal experiments

The establishment and propagation of the five human orthotopic models of PMP and mucinous peritoneal carcinomatosis were previously described.[12, 13] Briefly, the tumor models were established by implanting human mucinous tumor tissue from patients undergoing cytoreductive surgery of the peritoneal cavity. The macroscopic growth pattern of the models closely resembled their human counterparts, with strict i.p. growth without metastasis development. The tumor tissue, microscopically dominated by large amounts of extracellular mucin with clusters of tumor epithelium, could be harvested as liquiform, highly viscous mucinous tumor tissue and injected i.p. for passage to new generations of mice or used for ex vivo experiments. The presence of tumor tissue was asymptomatic in the mice, to be detected only by visible abdominal distension, which was used as humane end point. In-house bred BALB/c (nu/nu) female mice, aged 4–6 weeks, were used. Animals were kept under specific pathogen-free conditions, and food and water were supplied ad libitum. Housing and all procedures involving animals were performed according to protocols approved by the animal care and use committee.

For the ex vivo studies, mucinous tumor tissue was harvested at autopsy and immediately brought to the laboratory for further processing as described in the respective sections. For the in vivo studies, 200–250 µl mucinous ascites was injected i.p. in each study animal. Animals were randomly assigned to treatment groups (in most cases, six animals per group), and the next day, or on day 6, the study drugs (MMC, OXA or MOC31PE) or vehicle (5% glucose) were administered in the form of a single i.p. injection. Additional control experiments were performed by i.p. administration of MOC31 alone (without the toxin moiety), and another isotype-like (IgG1) antibody (BM7; performed in the PMCA-3 model) as well as human gamma globulin, which contains a mixture of immunoglobulin species (#G 4386, Sigma Aldrich, Saint Louis, MO; PMCA-2, PMP-2 and PMCA-3 models), and no growth inhibition was observed compared to vehicle treatment. The BM7 antibody, which recognizes a glycosylated epitope on the MUC1, was kindly provided by Dr. S. Kaul (Frauenklinik, Heidelberg, Germany). Animals were monitored for symptoms of drug toxicity and for the presence of tumor growth. As tumor growth does not give rise to symptoms of disease in these models, animals were sacrificed by cervical dislocation when and if i.p. tumor growth was detected in the form of abdominal distention or at the end of the experiment (which was defined to be at least two times the observation period of untreated animals). At autopsy, tumor was collected and weighed, and if possible, tumor samples from at least two animals in each treatment group were collected. Samples were formalin fixed, paraffin-embedded, sectioned and hematoxylin–eosin stained, and the presence of typical tumor tissue for each model was verified. Very few animals had to be sacrificed for other reasons than tumor growth and they were excluded from analysis.

Calculation of growth index

Because tumor growth was asymptomatic and had to be evaluated by visual inspection, the observation time and amount of harvested tumor at autopsy individually were insufficiently accurate as outcome parameters, and an alternate approach was worked out to express tumor growth in each animal. A growth index was devised to calculate an end point for each animal, combining these two key parameters, final tumor load (tumor weight) and the time for each tumor to reach its final volume relative to the total duration of the experiment:

display math

where TTotal is the entire duration of the experiment in days and TA is the time from start of the experiment until sacrifice of the specific animal. The fraction was multiplied by 10 to balance the relative contribution of the time factor with that of tumor load at autopsy.

Example of how the growth index is calculated in an imaginary experiment using the equation:

Mouse 1 (vehicle): Killed on day 50 (TA), and 4 g of tumor tissue was harvested

Mouse 2 (very efficacious drug): Killed on day 100 (TA), and 0.3 g of tumor tissue was harvested

Mouse 3 (not so efficacious drug): Killed on day 90 (TA), and 4 g of tumor tissue was harvested

The experiment lasted for 100 days, which gives the TTotal (defined as at least two times the observation period of the vehicle group). Respective growth indices would then be:

Mouse 1: math formula

Mouse 2: math formula

Mouse 3: math formula

Immunohistochemistry

Frozen tissue sections were air dried and stored at −80°C. Buffers, streptavidin-horseradish peroxidase (HRP) and 3′,3′-diaminobenzidine (DAB) substrate used for immunochemistry were all from DAKO (Glostrup, Denmark). After thawing, the slides were fixed for 10 min in ice-cold acetone, washed with buffer and blocked for 10 min at room temperature (RT) with biotin blocking system part one. After washing, the sections were incubated for 10 min at RT with biotin blocking system part two. After washing, sections were incubated with biotinylated human EpCAM antibody (kind gift from Affitech AS, Oslo, Norway) (16 μg/ml in ChemMate antibody diluents) overnight at 4°C. The human anti-EpCAM antibody was biotinylated as recently described.[15] Next day the sections were washed, incubated with streptavidin-HRP complex for 1 hr at RT and subsequently washed before the sections were stained with DAB + chromatin substrate, counterstained with hematoxylin and mounted. Negative control sections were treated as described above except for the incubation with primary antibody.

Cell viability assay

For the ex vivo experiments (cell viability, protein synthesis and Western blot analysis), mucinous tumor tissue from animals or patients was diluted 1:4 and passaged through decreasing dimensions of syringes until a homogenous cell suspension was obtained. Because of the high mucin content, the cell suspension was highly viscous, and determining the number of cells seeded was not possible; standardization was achieved by using the same volume of cell suspension through each experimental series. Using the PMCA-2 model as an example, analyzing ex vivo cell viability of vehicle-treated cells, mean absorbance was 0.986 ± 0.067, representing a SD of 6.8% (in three separate biological experiments with three to six parallels), and in vivo, the variability was on a similar scale (SD: 7.8%). The observed differences between treatment groups were thus deemed to clearly transcend intra-experimental variability, suggesting that using tumor volume as a means of assay standardization was acceptable. For the cell viability assay, the suspension was seeded in 96-well plates (100 microliters per well), MOC31PE (1,000 ng/ml) or vehicle was added and the cells were incubated in a standard tissue culture incubator at 37°C for 24 hr. Solution cell proliferation assay (Promega, Madison, WI) was then added to the wells to quantify the presence of viable cells, and absorbance was measured 1–2 hr later at a wavelength of 490 nm in a Wallac Victor 2 plate reader (Perkin Elmer, Waltham, MA). After correcting for background absorbance, values generated for treated cells and untreated controls were compared and percentage cell viability was calculated. Experiments were set up with three to six parallel samples, and experiments with xenograft tissue were repeated at least three times, whereas analyses of live cells from patient samples could be performed only once.

Protein synthesis assay

Protein synthesis was measured by using the [3H]-leucine incorporation assay.[16] Mucinous cell suspension was seeded in 48-well plates (250 microliters per well) and MOC31PE (1–1,000 ng/ml) or vehicle was added immediately and the suspension was incubated at 37°C for 24 hr. Afterward, cells were washed twice with cold PBS with 0.1% human serum albumin and incubated with [3H]-leucine (2 μCi/ml) in leucine-free medium for 60 min at 37°C. The cells were then washed twice with 5% trichloroacetic acid for 10 and 5 min, respectively, and dissolved in 0.1 M KOH. The resultant solution was transferred to the liquid scintillator Aquasafe 300 Plus (Zinsser Analytic, Frankfurt/Main, Germany). Sample counts were determined in a liquid scintillation counter (LKB Wallac, Perkin Elmer, Boston, MA). Assays were performed in triplicate and repeated three times for PMP-2 and once for PMCA-2.

Western blot analysis

The mucinous cell suspension was seeded in 48-well plates (500 microliters per well). MOC31PE (1,000 ng/ml) or vehicle was immediately added and the cells were incubated at 37°C for 24 hr. Thereafter, the cells were lysed by using an sodium dodecyl sulfate (SDS) boiling method as described previously.10 In brief, cell pellets were resuspended in lysis buffer [2% SDS, 1 mM Na3VO4 and 10 mM Tris–Cl (pH 7.6)], which was held at 100°C when added, and the lysates were boiled for 5 min. After six passages through a 20-G syringe on ice, the lysates were cleared by centrifugation. Protein concentrations were then determined using the BCA (bicinchoninic acid) protein assay (Pierce, Rockford, IL). The lysates were snap frozen in liquid N2 and kept at −70°C. A portion (15 μg) of each protein lysate was fractionated by 4–12% NuPAGE Bis-Tris gel electrophoresis (Invitrogen, Carlsbad, CA) and subsequently transferred by electroblotting to Immobilon membranes (Millipore, Bedford, MA). The filters were probed with an anti-PARP antibody (Roche Diagnostics, Mannheim, Germany), anti-EpCAM (MOC31, IQ Corporation, Groningen, The Netherlands) or with anti-α-tubulin (Calbiochem, La Jolla, CA). Immune complexes were detected with an appropriate HRP-coupled secondary antibody. Peroxidase activity was visualized with enzyme-linked chemiluminescence (Amersham Pharmacia Biotech, Buckinghamshire, UK). The Western blots were probed with anti-α-tubulin antibody to confirm equal loading and transfer of samples.

Statistical analyses

Student's t-tests were performed to compare treatment groups in the various assays, using SPSS (Statistical Package for Social Sciences) version 16.0 (SPSS, Chicago, IL), and p-values <0.05 were considered statistically significant.

Results

In vivo tumor growth inhibition by MMC and OXA

Studies of tumor growth inhibition by MMC and OXA were undertaken to examine how the model systems would perform, comparing two drugs with expected efficacy as they both are in clinical use for treatment of mucinous peritoneal surface malignancies. Tumor take was in all models 100% in control animals. As shown in Table 1, the mean time period for control animals to develop visible abdominal distension ranged between 38 days for the PMP-2 model and 46 days for the PMCA-1 and PMP-1 models. The mean amount of mucinous tumor tissue harvested per tumor-bearing mouse in untreated animals ranged between 4.0 and 6.4 g, the variability probably reflecting difficulty of reproducibly evaluating the degree of abdominal distention in otherwise asymptomatic animals. Several series of experiments were performed to establish relevant dose levels and toxicities using the PMCA-2 model. For MMC, doses of 5 mg/kg were well tolerated, whereas toxicity (failure to thrive or weight loss in the absence of tumor development) was observed at 10 mg/kg, while the highest tolerated dose of OXA was 10 mg/kg (data not shown). Within the tolerated dose levels of MMC, 5 mg/kg almost completely inhibited tumor growth, as only one of 11 animals developed tumor during the observation period, and only a minimal amount of tumor was collected, giving a mean growth index of 0.02. A single dose of 2.5 mg/kg was slightly less efficacious, giving a mean growth index of 0.3 (Fig. 1). In contrast, the highest tolerable dose of OXA (10 mg/kg) only slightly delayed tumor growth in the PMCA-2 model, giving a mean growth index of 7.0, and at 5 mg/kg the mean growth index was equal to that of untreated animals. Experiments were also performed to assess whether the effects were confined to the immediate period after tumor implantation, and interestingly, delaying treatment start until day 6, using the highest tolerable doses for each drug, gave very similar results compared to treatment given on day 1.

Table 1. In vivo tumor growth in five tumor models upon treatment with vehicle, mitomycin C and oxaliplatin
   Mitomycin COxaliplatin
     Day 6
  Vehicle2.5 mg/kg5 mg/kg5 mg/kg5 mg/kg10 mg/kg10 mg/kg
PMCA-2Tumor growth/total number of animals14/141/31/111/56/610/104/4
 Days until sacrifice (mean)42106103106447674
 Mucin (g), mean (SD)4.4 (1.3)1 (0.6)0.02 (0.1)0.1 (0.3)4.9 (0.8)4.5 (1.4)5.8 (1.6)
PMP-1Tumor growth/total number of animals6/60/61/6  6/6 
 Days until sacrifice (mean)46110100  68 
 Mucin (g), mean (SD)5.0 (0.9)01.0 (2.3)  4.7 (0.7) 
PMP-2Tumor growth/total number of animals6/60/60/6  5/6 
 Days until sacrifice (mean)38111111  74 
 Mucin (g), mean (SD)5.1 (0.9)0.7 (0.4)0.5 (0.7)  4.3 (2.3) 
PMCA-1Tumor growth/total number of animals6/63/61/6  6/6 
 Days until sacrifice (mean)469498  68 
 Mucin (g), mean (SD)4.0 (1.3)1.5 (2.3)1 (2.6)  4.7 (2.4) 
PMCA-3Tumor growth/total number of animals6/64/60/5  6/6 
 Days until sacrifice (mean)45102112  63 
 Mucin (g), mean (SD)6.4 (0.7)1.2 (2.4)0  5.3 (1.2) 
Figure 1.

Efficacy of intraperitoneal (i.p.) mitomycin C (MMC) and oxaliplatin (OXA) in five orthotopic models of mucinous peritoneal surface malignancies. To create a single end-point parameter for each animal, the “growth index” was devised (see the Material and Methods section), incorporating into one number the observation time from tumor implantation until the animal was sacrificed and the weight of tumor tissue harvested at autopsy. Error bars indicate standard deviation. (a) Mean growth indices calculated for the PMCA-2 model showing the effect of administrating single i.p. doses of MMC, OXA and vehicle (on day 1 or day 6 after tumor implantation). (b) Mean growth indices calculated for the PMP-1, PMP-2, PMCA-1 and PMCA-3 models after single i.p. doses of vehicle, MMC 2.5 mg/kg, MMC 5 mg/kg and OXA 10 mg/kg (on day 1 after tumor implantation); V, vehicle. Significant differences compared to vehicle: *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Based on these results, in the remaining four models, 2.5 and 5 mg/kg of MMC and 10 mg/kg of OXA were used. In spite of variability between the models, some apparent trends were observed. In all models, growth inhibition induced by OXA was clearly inferior to that of MMC, as OXA gave rise to only slight reduction of growth indices compared to vehicle-treated animals (Fig. 1). In the PMP-1 and PMP-2 models, both tested doses of MMC almost completely inhibited tumor growth within the observation period, and for PMCA-1 and PMCA-3, 5 mg/kg was more efficacious than 2.5 mg/kg MMC.

Expression of EpCAM in xenografts and patient tissues

A prerequisite for optimal immunotoxin activity is the presence of the target antigen on the surface of target cells. Expression of EpCAM was detected by immunohistochemistry in tumor cells in samples from all the animal models. Figures 2a2c depict the results for the PMCA-2 and PMP-2 models, indicating the presence of EpCAM in the cytoplasm and on the cell membrane of tumor cells. Additionally, Western immunoblot analysis was performed on tissue lysates from xenografts using the mouse monoclonal MOC31 antibody, confirming the presence of EpCAM in all examined samples (Fig. 2d).

Figure 2.

Expression of EpCAM as assessed by immunohistochemistry; PMP-2 (a), PMCA-2 (b) models, negative control (c) and Western immunoblot analysis of lysates from all five xenografts (d).

In vivo tumor growth inhibition by MOC31PE and MMC

In a new set of animal experiments, the efficacy of administering single i.p. injections of MOC31PE and MMC alone or in combination was assessed in two of the animal models. In the PMCA-2 model, MMC 1 mg/kg inhibited tumor growth, but clearly less effectively than the higher doses (2.5 and 5 mg/kg) used in the model validation series. MOC31PE (150 µg/kg, previously determined as maximum tolerable i.p. dose in this strain of mice, data not shown) similarly brought about a clear reduction in growth index, while combination of the two drugs in the same doses had at least additive effect, almost completely preventing tumor growth in the animals (Fig. 3a). The PMP-2 model was clearly more sensitive to both MMC and MOC31PE, as all tested doses resulted in substantial inhibition of tumor growth in this model (Fig. 3b), and even the lowest immunotoxin dose (6.4 µg/kg) resulted in more than 50% growth inhibition. Similar to the validation studies, delaying treatment start until day 6 gave almost equal results compared to day 1 administration. No advantage of combining the drugs could be detected for the tested drug combinations (MMC 1 mg/kg + MOC31PE 150 µg/kg; MMC 1 mg/kg + MOC31PE 6.4 µg/kg and MMC 0.2 mg/kg + MOC31PE 6.4 µg/kg). Interestingly, intravenous injection of a low dose of MOC31PE (32 µg/kg) did not have any effect on i.p. tumor growth, suggesting that the effect was caused by a local interaction in the peritoneal cavity. The PMCA-3 model exhibited intermediate sensitivity toward both drugs (Fig. 3c). Combining the two lowest drug doses given (MMC 1 mg/kg + MOC31PE 6.4 µg/kg) completely prevented tumor growth in the PMCA-3 model.

Figure 3.

Mean growth indices were calculated for the PMCA-2, PMP-2 and PMCA-3 models to demonstrate the effects of administrating single intraperitoneal doses of vehicle, mitomycin C (MMC), MOC31PE and combinations of the drugs. Error bars indicate standard deviation. (a) PMCA-2. All treatment groups were significantly different compared to vehicle (p = 0.005 for MMC, p < 0.001 for MOC31PE and the combination), and the combination treatment was more efficacious than either of the drugs as monotherapy (p < 0.001). (b) PMP-2. All i.p. treatment groups were significantly different compared to vehicle (p < 0.001 for all groups with exception of MOC31PE 6.4 µg/kg; p = 0.04). No significant differences were detected between i.p. treatment groups or between intravenous (i.v.) treatment with MOC31PE 32 µg/kg and vehicle. No additional growth inhibition was observed when combining MMC and MOC31PE (MMC 1 mg/kg + MOC31PE 150 µg/kg; MMC 1 mg/kg + MOC31PE 6.4 µg/kg and MMC 0.2 mg/kg + MOC31PE 6.4 µg/kg). Delaying administration of MOC31PE until day 6 was equally effective as day 1 treatment (150 µg/kg). (c) PMCA-3. All i.p. treatment groups were significantly different compared to vehicle (p ≤ 0.001 for all groups). No tumor growth was observed in the combination group receiving low doses of MMC and MOC31PE.

Ex vivo activity of MOC31PE

To investigate possible mechanisms of immunotoxin activity, short-term (24 hr) ex vivo experiments were performed using tumor tissue harvested either from tumor-bearing animals (PMCA-2 and PMP-2) or taken directly from the operating theatre from patients undergoing surgery (PMP-3 and PMP-6). Ex vivo treatment of mucinous tumor tissue with MOC31PE (1,000 ng/ml) for 24 hr resulted in ∼25% reduction of cell viability in both tumor models compared to vehicle-treated cells (Fig. 4a). Ex vivo activity of the immunotoxin was also observed in the patient samples, with 44 and 12% reduction of cell viability for the PMP-3 and PMP-6 models, respectively.

Figure 4.

Ex vivo short-term cultures of mucinous tumor tissue from the PMCA-2 and PMP-2 animal models and of two samples of mucinous tumor tissue taken directly from the operating theatre. Error bars indicate standard deviation. (a) Cell viability was assessed 24 hr after addition of MOC31PE (1,000 ng/ml) using the MTS assay and is expressed as percentage of the value obtained in vehicle-treated cells; V, vehicle treatment. *Significant differences compared to vehicle, p = 0.005. #Significant differences compared to vehicle, p = 0.03. (b and c) Protein synthesis inhibition was determined by measuring 3H-leucine incorporation after treatment with MOC31PE or vehicle in PMCA-2 (b) and PMP-2 (c) and is expressed as a percentage of the value obtained in vehicle-treated cells. *Significant differences compared to vehicle, p < 0.001. #Significant differences compared to vehicle, p = 0.006. (d) Western blot analysis was performed on protein lysates after exposure to MOC31PE or vehicle for 24 hr. The blots were incubated with anti-poly(ADP-ribose) polymerase (PARP) antibody or anti-α-tubulin antibody as a loading control. The 118- and 85-kDa bands represent the uncleaved and cleaved versions of PARP, respectively. Alpha-tubulin detection was unsuccessful for the PMP-6 sample.

Mechanisms of immunotoxin activity

Inhibition of protein synthesis has been shown to be one of the main mechanisms involved in immunotoxin-related cell death, and in the ex vivo setting, a dose-response relationship was present when mucinous tumor tissue was treated with MOC31PE in both models (Figs. 4b and 4c). Protein synthesis was decreased by ∼20 and 40% after 24-hr treatment with MOC31PE 100 or 1,000 ng/ml, respectively.

Another typical feature of immunotoxin-mediated cell death is the induction of apoptosis, which can be visualized by the detection of poly(ADP-ribose) polymerase (PARP) cleavage by Western immunoblot analysis. Upon ex vivo treatment of mucinous tumor tissue with MOC31-PE immunotoxin, PARP cleavage was detected in PMP-2 xenograft tissue as well as in both samples taken directly from patients. In the PMCA-2 model, the 85-kDa cleavage product was present in the untreated sample at a relatively substantial level, and no increase was detected after immunotoxin treatment, although depletion of the uncleaved form was detected, suggesting that apoptosis-related cleavage still had taken place (Fig. 4d). No PARP inactivation was evident in cells treated with the MOC31 antibody alone or with the PE toxin alone, showing that the effect was solely dependent on the specific cytotoxic effect of MOC31PE (data not shown).

Discussion

The rationale for applying local therapy (cytoreductive surgery and i.p. chemotherapy) in peritoneal surface malignancies rests on the conception of the peritoneal cavity as a unique compartment, with the peritoneum as a barrier against local invasion and metastatic tumor spread.[17] Thus, in principle, even if cancerous dissemination has occurred widely in the peritoneal cavity, complete tumor eradication may be accomplished provided the peritoneal barrier has not been compromised. The possibility of adding immunotoxin treatment to standard therapy is particularly appealing, because it has the potential to provide direct cytotoxicity and has a unique mechanism of action compared to other cytotoxic drugs. Our study reports our first experience with assessment of experimental treatment in orthotopic models of mucinous peritoneal surface malignancies, and specifically the use of MOC31PE immunotoxin in these models.

Differential sensitivity toward MMC and OXA was observed in all five models, in all cases favoring MMC over OXA, using the highest doses tolerated by the mice. Only one patient (from whom tissue was used to establish the PMCA-2 model) had previously received chemotherapy, in the form of adjuvant 5-fluorouracil after the primary procedure and neoadjuvant 5-fluoruracil in combination with OXA upon diagnosis of peritoneal carcinomatosis. MMC and OXA are the drugs most extensively used in i.p. chemotherapy for mucinous peritoneal surface malignancies, MMC being the preferred drug in PMP, whereas OXA is being studied as an alternative to MMC in the treatment of peritoneal carcinomatosis from colorectal cancer.[18] The low efficacy observed for OXA in these models was somewhat surprising, but in accordance with findings recently reported from a model of peritoneal carcinomatosis based on the nonmucinous human colorectal carcinoma cell line HT29.[19] In the human treatment setting, i.p. therapy is directed toward eradication of low-volume disease, and particularly against exfoliated tumor cells in the peritoneal cavity. Delivering treatment on day 1 in the models was aimed to mimic this situation, but interestingly, the same differences were observed when the drugs were administered on day 6 (Table 1), indicating that the effects were not confined to the immediate period after tumor implantation. Thus, the results suggest that the current orthotopic models of mucinous peritoneal surface malignancies can be effectively used to study treatment efficacy of i.p. administered drugs.

Although all three tested models were sensitive to MMC and MOC31PE, they exhibited very different sensitivity toward each drug given alone and in combination. In the PMCA-2 model, moderate effects were observed with MMC and MOC31PE alone, whereas the combination was significantly more efficacious than either drug given as monotherapy. In contrast, in the PMP-2 model, all tested drug doses were efficacious, and no additional benefit was gained from combining the drugs, suggesting that further dose reduction might be appropriate in this very sensitive model. The PMCA-3 model exhibited intermediate sensitivity toward both drugs. It would be tempting to speculate that the observed differential sensitivity might reflect the aggressiveness of the tumor models, because the PMCA-2 and PMCA-3 models were derived from patients with aggressive histology, whereas the PMP-2 model represents an intermediate histological subtype, originating from an appendiceal cystadenoma.[12] However, and perhaps more likely, molecular characteristics independent of histological subtype might determine the sensitive phenotype, but presently no specific determinant of MMC sensitivity has been defined. For MOC31PE, the best indicative factor is cell surface antigen expression, which was essentially similar in all the models as assessed by immunohistochemistry. Thus, in addition to demonstrating the efficacy of administering MOC31PE in this setting, our results suggest the presence of differential drug sensitivity in peritoneal surface malignancies that currently has not been explored in human disease, and these models might represent potential tools for generating a rationale to improve preclinical drug selection for subsequent testing in humans.

Ex vivo, MOC31PE inhibited cell viability in the PMCA-2 and PMP-2 models, and similar findings were made in the samples of mucinous tumor tissue cultured directly from two patients (PMP-3 and PMP-6). The ex vivo effects of MOC31PE were relatively modest compared to the in vivo efficacy, probably reflecting differences in microenvironment between the ex vivo culturing conditions and the peritoneal cavity in nude mice. However, the consistent effects on cell viability suggested that the ex vivo situation could still be appropriate for studying mechanisms for MOC31PE-mediated cell death. Importantly, both the ex vivo and the in vivo conditions aimed to simulate the therapeutic challenge encountered after cytoreductive surgery when free tumor cells without vasculature are left in the peritoneal cavity. In subsequent experiments, mechanistic evidence of immunotoxin activity was identified in the examined samples; specifically, inhibition of protein synthesis as assessed by 3H-leucine incorporation (in PMCA-2 and PMP-2) and induction of apoptosis by demonstrating PARP cleavage (additionally in PMP-3 and PMP-6) (Fig. 4). The identification of some of the classic features of immunotoxin activity suggests that MOC31PE exerts its antitumor activity in mucinous peritoneal surface malignancies through mechanisms similar to what has been previously described by us and others in other model systems.[5-7, 20] Assuming that these results are representative of the in vivo situation, the potential for successfully combining MOC31PE with other chemotherapeutic agents based on the unique mechanisms of action of the immunotoxin is also likely to be present in the setting of i.p. treatment.

I.p. administration after cytoreductive surgery for mucinous peritoneal malignancies of intestinal origin would in our opinion for several reasons be an ideal setting for exploring clinical efficacy of treatment with MOC31PE. The tumor burden after cytoreductive surgery is very low and restricted to the peritoneal cavity and its surfaces, which would allow optimal interaction between MOC31PE and remaining tumor cells. EpCAM was highly expressed in all the models, and the observed ex vivo effects of MOC31PE in two samples taken directly from patients suggested a similarly sensitive phenotype. Although it should be verified in a broader cohort of mucinous peritoneal surface malignancies, it is known from several immunohistochemical studies that most colorectal tumors express EpCAM, and MOC31PE might thus be an ideal cytotoxic molecule for targeting this disease.[11] Clinical implementation of immunotoxins for cancer treatment has been hampered by neutralizing anti-immunotoxin antibody response, making repeated administrations a challenge, and in some cases by liver toxicity, neurotoxicity and vascular leak syndrome. In our phase I clinical trial with MOC31PE, anti-immunotoxin antibody response was observed in addition to mild, reversible liver toxicity. In i.p. treatment of peritoneal surface malignancies, the most likely schedule for MOC31PE would involve one single administration of the drug after cytoreductive surgery, alleviating the concern for neutralizing antibodies. Finally, no antagonistic interaction was observed with MMC in our work, suggesting that MOC31PE could probably be added to the standard treatment setting.

Conclusion

Our orthotopic animal models of mucinous peritoneal surface malignancies were shown to effectively display differences in efficacy upon single i.p. injections of MMC, OXA and MOC31PE. Importantly, MOC31PE efficaciously inhibited tumor growth alone and in combination with MMC, most likely through classic mechanisms of immunotoxin activity, as shown in ex vivo experiments. Interestingly, substantial differences in sensitivity toward MMC and MOC31PE were observed between the PMCA-2, PMP-2 and PMCA-3 models, suggesting further studies to clarify the molecular basis and clinical relevance of this finding. The present results support the consideration of MOC31PE for early-phase clinical trials in EpCAM-expressing mucinous peritoneal surface malignancies of intestinal origin, and a phase I clinical trial is currently being designed.

Ancillary