Megakaryocyte Potentiation Factor Cleaved from Mesothelin Precursor Is a Useful Tumor Marker in the Serum of Patients with Mesothelioma

Megakaryocyte potentiation factor (MPF) is a 33-kDa soluble protein that is produced by proteolytic cleavage of the 69-kDa mesothelin precursor protein (1, 2). The other cleavage product is mesothelin, a cell membrane–associated 40-kDa glycosylphosphatidylinositol-linked glycoprotein (1). Mesothelin is present on normal mesothelium but is highly expressed on mesotheliomas and other tumors, including ovarian and pancreatic cancers (3–8). The biological function of mesothelin and MPF is not clear. MPF was originally identified as a polypeptide from the culture supernatant of a pancreatic cancer cell line and was shown to stimulate the megakaryocyte colony-forming activity of murine interleukin-3 in mouse bone marrow cell cultures (9). However, mesothelin knockout mice, which do not produce the mesothelin precursor protein encoding both MPF and membrane-bound mesothelin, showed no discernible phenotype and had normal platelet counts (10). Recently, mesothelin has been shown to bind to CA125/MUC16 that is present on many ovarian cancer cells, and it was suggested that mesothelin might have a role in the metastatic spread of ovarian cancer in the peritoneal cavity (11).

Malignant mesothelioma is a highly aggressive tumor of serosal cavities that arises from the mesothelial cells of the pleura, peritoneum, or pericardium (12). Although the incidence of mesothelioma in the United States seems to be leveling off, its incidence is still increasing in other countries (13–15). Although some advances have been made in the treatment of this disease, the prognosis of patients is still poor with a median overall survival of about 9 to 12 months (12). However, surgery can lead to improved survival in the small percentage of patients who are diagnosed at an earlier stage (16). A sensitive tumor marker that allows early diagnosis of mesothelioma in people at risk for this disease would, therefore, be clinically useful.

Because virtually all epithelial mesotheliomas overexpress mesothelin, we wanted to determine if MPF, a proteolytic fragment of the mesothelin precursor protein, is shed into the blood stream of patients with mesothelioma. To measure MPF that may be released into the circulation, we generated monoclonal antibodies (mAb) specifically directed to MPF. Using anti-MPF mAbs that react with different epitopes on the MPF protein, we developed a sandwich ELISA to measure the MPF level in human serum. Our results show elevated MPF levels in the sera of patients with mesothelioma compared with healthy donors. Our findings also suggest that measurements of MPF in the blood could be potentially useful to monitor treatment response.

Human serum samples. Seventy serum samples from healthy blood donors obtained from Bioreclamation, Inc. (Hicksville, NY) were used as controls. The median age of healthy volunteers was 39 years (range, 18-65 years; 36 males and 34 females). Sera from 56 patients with mesothelioma (all were of the epithelial type) were obtained from patients being evaluated for participation in phase I studies of the anti-mesothelin immunotoxin SS1P (17). These studies were approved by the Institutional Review Board of the National Cancer Institute, and all patients gave informed consent. All of the sera came from patients with advanced unresectable disease. The median age of patients with mesothelioma was 57 years (range, 26-79 years; 42 males and 14 females). Serum samples were kept at −80°C until the assay. For the study on the effect of tumor debulking on the level of serum MPF, serum samples before and after surgery were obtained from patients with mesothelioma who were treated on Surgery Branch protocols at the National Cancer Institute. These patients underwent tumor reduction with intraoperative continuous hyperthermic peritoneal perfusion with cisplatin (18). The patients were treated on an Institutional Review Board–approved protocol.

Generation of MPF-rFc and MPF proteins. MPF was expressed as a fusion protein with rabbit IgG Fc (rFc) in HEK293T cells as previously described (19). The DNA fragment encoding MPF protein (amino acids 1-291) was PCR amplified from the plasmid pMH107 (Genbank accession no. AY743922; ref. 20) and was inserted into the rFc expression plasmid based on pcDNA3 (Invitrogen, Carlsbad, CA). Mesothelin (amino acids 296-599) was also expressed as a fusion protein with rFc. MPF protein was released from the MPF-rFc by Prescission protease treatment (Amersham Biosciences, Piscataway, NJ) that cut the recognition site between MPF and rFc. After cleavage, the Prescission protease that is a fusion protein with glutathione S-transferase was removed by glutathione-Sepharose (Amersham Biosciences) column. The purity of the prepared MPF protein was shown by SDS-PAGE analysis (Fig. 1).

Generation of mAbs against MPF protein. Ten female mesothelin-deficient mice (6-8 weeks old) were immunized thrice with the MPF-rFc fusion protein (50 μg in 100 μL PBS) i.p. at intervals of 2 weeks. All animals were maintained in accordance with institutional guidelines. Three days before fusion, MPF protein (20 μg) was injected i.p. Cell fusions of the splenocytes and the SP2/0 cells, selection of the hybridomas, and screening for mAb production by an indirect ELISA were carried out as previously described (21). In the ELISA, each well of Nunc-Immuno plates (Nalge Nunc International, Rochester, NY) were coated with a solution containing 50 ng/50 μL of purified MPF protein and incubated with 50 μL of 10-day supernatants of hybridomas. The bound mAbs were detected by an incubation with horseradish peroxidase–conjugated goat anti-mouse IgG antibody (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) diluted 1:2,000. As the substrate, immunopure tetramethylbenzidine substrate solution (Pierce, Rockford, IL) and a 2 N solution of sulfuric acid were used. Plates coated with mesothelin-rFc protein were used as negative control for this assay. The selected hybridomas were grown in a CELLINE flask (INTEGRA Biosciences, Chur, Switzerland). The mAbs were harvested from supernatants, and isotypes were determined by a mouse immunoglobulin isotyping kit (Roche Applied Science, Indianapolis, IN) and purified on a protein A column (Amersham Biosciences).

Whole-cell extract preparation and Western blot analysis. Mesothelin-expressing A431/K5 cells were harvested, and whole-cell lysates were prepared. The cells were washed thrice with 10 mL of ice-cold PBS each time. Radioimmunoprecipitation assay buffer [50 mmol/L Tris-HCl (pH 7.5), 150 mmol/L NaCl, 1 mmol/L EDTA (pH 8), 0.1% Triton X-100, 1 mmol/L phenylmethylsulfonyl fluoride, 1 μg/mL aprotinin, 1 μg/mL leupeptin, and 1 μg/mL pepstatin] was used to solubilize cells on the dish. The cell solution was cleared by centrifugation at 20,000 × g at 4°C. The supernatant (whole-cell lysate) was frozen in small aliquots on dry ice and stored at −80°C. Protein concentration was determined by bicinchoninic acid assay (Pierce) according to the manufacturer's protocol. The whole-cell lysates (40 μg per lane) and the recombinant mesothelin proteins (1 μg per lane) purified from Escherichia coli (15) were separated by a 4% to 20% Tris-glycine SDS-PAGE and transferred onto nitrocellulose filters. The filters were incubated with primary antibodies for 2 hours at the following concentrations: 2 μg/mL anti-MPF mAb (MPF44), 1 μg/mL anti-mesothelin mAb (MN), and 1 μg/mL anti-tubulin (Santa Cruz Biotechnology, Santa Cruz, CA). Primary antibodies were detected by a horseradish peroxidase–conjugated secondary antibody and goat anti-mouse immunoglobulin G (Biosource, Camarillo CA) at 1:20,000 for 1 hour. MPF and mesothelin proteins were visualized by the enhanced chemiluminescence kit (Amersham Biosciences).

PNGase F digestion. Whole-cell lysates (40 μg) or purified recombinant proteins (10 μg) were digested with PNGase F (New England Biolabs, Ipswich, MA) for 2 hours at 37°C as instructed by the manufacturer. N-linked deglycosylation of cell lysates was then examined by Western blot analysis.

Western blot using recombinant MPF protein. The reactivity of the anti-MPF mAbs to SDS-denatured antigen was tested in a Western blot analysis. MPF-rFc, bovine serum albumin, and MPF (50 ng of each) were separated on 4% to 20% SDS polyacrylamide gels (Bio-Rad, Hercules, CA) under reducing conditions. Proteins were transferred to a 0.2-mm Immunoblot polyvinylidene difluoride membrane (Bio-Rad) in transfer buffer [25 mmol/L Tris-HCl, 192 mmol/L glycine, 30% (v/v) methanol (pH 8.3)] at 4°C for 1 hour at 240 mA. Western blotting procedures were the same as the Western blot using cell lysates described above.

Topographical epitope mapping. Topographical epitopes of anti-MPF mAbs were determined by the assessment of the mutual competition of all possible pairs of mAbs in label-free competitive ELISA (22). Briefly, an immune complex formed by incubation of MPF-rFc and a mAb (competitor) is tested for binding to another mAb (indicator) that had been captured by goat anti-mouse IgG (Jackson ImmunoResearch Laboratories) coated on the plate. If the topographical epitopes of these two mAbs are the same, MPF-rFc will not be able to bind to the indicator mAb because the competitor mAb has already occupied the epitope for the indicator mAb. If the epitopes are different, MPF-rFc binds to the indicator mAb because the epitope for the indicator mAb is free. The MPF-rFc bound to the plates was detected by horseradish peroxidase–labeled anti-rabbit IgG (Jackson ImmunoResearch Laboratories). The epitopes are classified into groups based on the mutual competition of mAbs as described previously (22).

Preparation of biotinylated mAb MPF25. Four milligrams of purified mAb MPF25 in 1 mL of PBS was combined with 1:20 volume of 1 mol/L NaHCO₃ (final pH 8.3). A N-hydroxysuccinimide ester of EZ-Link sulfo-N-hydroxysuccinimide-LC-biotin (Pierce) dissolved in water at 18 mg/mL (10 mmol/L) and 50 mL of the solution was immediately mixed with the mAb solution (ratio of 20 biotins per mAb molecule). After incubation for 2 hours at room temperature with end-over-end rotation, the solution was dialyzed (cutoff value: 12,000-14,000; Invitrogen) twice against 5 liters of PBS for 2 days at 4°C. The solution was diluted to 1 mg/mL with PBS and combined with 2 mg/mL bovine serum albumin and 0.1% sodium azide. Aliquots were stored at −80°C.

Sandwich ELISA. Purified mAb MPF49 (4 μg/mL in PBS) was absorbed onto ELISA plates (50 μL/well) at 4°C overnight. The plates were then blocked with blocking buffer (25% DMEM, 5% FCS, 25 mmol/L HEPES, 0.5% bovine serum albumin, 0.1% azide = 100 μL/well) for 30 minutes at room temperature. After washing, diluted samples or standards were added to the washed plates (50 μL/well). After overnight incubation at 4°C, the plates were washed twice and incubated with biotinylated mAb MPF25 (0.5 μg/mL in blocking buffer). Plates were then washed twice and incubated with 1:2,000 dilution of streptavidin-horseradish peroxidase with 0.05% Tween 20 in blocking buffer for 1 hour at room temperature (50 μL/well). After washing the plates four times, tetramethylbenzidine substrate solution was added to each well (100 μL/well). The enzyme reaction was allowed to proceed at room temperature for 15 minutes; 2 N sulfuric acid (50 μL/well) was added, and the absorbance at 450 nm was measured with the reference at 600 nm. Color development stopping solution was done until the absorbance signal of an 8 ng/mL MPF standard attained an absorbance of 0.8 at 450 nm. For measurements of MPF in culture supernatants, a standard curve using purified recombinant MPF was constructed. A_{450 nm} values were used to represent the MPF values in serum.

Preparation of culture supernatant from various cancer cell lines. The human ovarian cancer cell lines A1847 and IGROV-1 (19), the human cervical cancer cell line HeLa, the human epidermoid carcinoma cell line A431, its stably expressing mesothelin-transfectant A431/K5 (23), and the human osteosarcoma cell line OHS-M1 (24) were obtained from laboratory stocks. Cells were cultured in DMEM or RPMI-based medium supplemented with 10% to 20% fetal bovine serum. For time course studies of MPF release into the culture supernatant, we used A431/K5 cells which are stable transfectants of mesothelin and A431 cells. The cells were seeded on day 0 (6 × 10⁴ per well) in 96-well culture plates. At each time point, medium from three wells was harvested. Each supernatant was centrifuged and kept at −80°C until the assay. The concentration of MPF was determined using sandwich ELISA (recombinant MPF was used as standard curve).

Fluorescence-activated cell sorting analysis to detect mesothelin expression on cells. Cultured cells (2 × 10⁵) were dissociated with dissociation buffer (Sigma-Aldrich, St. Louis, MO). Each sample was washed twice in fluorescence-activated cell sorting (FACS) buffer (PBS with 5% fetal bovine serum and 0.1% sodium azide); 1 μg/mL of anti-mesothelin mAb MB (18) was added to the cells and incubated for 1 hour on ice. Cells were washed twice with FACS buffer, resuspended in 100 μL of secondary antibody (R-PE conjugated goat anti-mouse IgG diluted 1:100; Biosource), and incubated for another hour on ice. Finally, cells were washed twice and analyzed on a FACSCalibur machine (Becton Dickinson, Franklin Drive, NJ) using the CellQuest software. The reactivities by FACS are shown with geometric means of FACS signals.

Statistical analysis. For the data comparison, nonparametric Mann-Whitney test was used. P < 0.05 was considered statistically significant.

Generation of a recombinant MPF-rFc fusion protein and generation of anti-MPF mAbs. The mesothelin precursor protein contains 622 amino acids and is composed of MPF (amino acids 1-295) and mesothelin (amino acids 296-622). To obtain mAbs that recognize MPF, we prepared a MPF-rFc fusion protein composed of the full-length MPF (amino acids 1-291) connected by a peptide linker containing a Prescission protease site to an Fc fragment of rabbit IgG1. Purified MPF was prepared by treating the fusion protein with Prescission protease (Fig. 1).

Mesothelin-deficient mice were used for the immunization because it seemed possible they would respond to MPF immunization better than the wild-type mice because their immune system had not been educated to the mouse homologue of human mesothelin. To try and obtain as many different mAbs as possible, 10 mesothelin-deficient mice were immunized with MPF-rFc. All mice showed a high titer against MPF (>1:10,000). From four mice, 56 hybridomas secreting mAbs specifically reacting with MPF were established (32 IgG1, 10 IgG2a, 10 IgG2b, 2 IgG3, and 2 IgM). Seven mAbs showing the highest ELISA signal were further characterized (Table 1). All seven mAbs bound to MPF (amino acids 1-291) and did not bind to a recombinant control fusion protein consisting of mesothelin (amino acids 296-599) and rFc.

The reactivities of the anti-MPF mAbs 25, 44, and 49 to SDS-denatured antigen were tested in a Western blot analysis, in which MPF-rFc, bovine serum albumin, and MPF were loaded on the same gel (Fig. 2A). Both mAbs MPF25 and MPF44 reacted with recombinant MPF-rFc protein and with MPF protein and did not react with bovine serum albumin. The reactivity of mAb MPF44 was stronger than that of mAb MPF25. In the MPF-rFc lane, mAb MPF44 reacted strongly with the fusion protein and also detected a small amount of free MPF (Fig. 2A, lane 4). mAb MPF49 did not react with either MPF-Fc or MPF and probably reacts with a conformational epitope (data not shown).

To determine the nature of MPF produced by a mesothelin-expressing cell line, mAb MPF44 was used to analyze MPF expression in A431/K5 cells that stably express the full-length mesothelin precursor protein (72 kDa). In samples not treated with PNGase F, the anti-MPF mAb detects a 72-kDa protein that also reacts with an antibody to mesothelin, indicating this is the mesothelin precursor protein (Fig. 2B). There is also a smaller band (40 kDa) that reacts with only the antibody to mesothelin. A protein the size of MPF was not detected within the cell, indicating that MPF was released after its precursor reached the cell surface. In cell extracts treated with PNGase F, the size of the band reacting with both antibodies was reduced to 50 kDa, consistent with deglycosylation. The size of the band reacting only with the antibody to mesothelin was also reduced. Mesothelin contains three potential N-linked glycosylation sites, whereas MPF has only one site (1). This experiment shows that anti-MPF antibody is useful for detecting the full-length mesothelin precursor protein in cells. There is no detectable free MPF in cells, indicating it is released after the precursor reaches the cell surface.

Topographical epitope mapping of anti-MPF mAbs. To construct a sandwich ELISA that can measure MPF in tissue culture supernatants and in the blood of patients, we need to use mAbs that react with different epitopes on MPF. To clarify the topographical epitopes of the mAbs, a competitive ELISA was carried out as described in Materials and Methods (label-free competitive ELISA; ref. 22). As shown in Table 2, the competition percentages of competitors varied from 4 (no competition) to 119 (complete competition), depending on the contribution of indicator and competition of the mAbs. When antibody binding is reduced by >40% for one pair of mAbs in a mutual manner, the two mAbs are judged to be located in the same topographical epitope group. This method revealed that the mAbs can be divided into three epitope groups (epitopes 1-3; Table 2). This result indicates that mAbs from different groups can bind to MPF at the same time. To develop a sandwich ELISA, we chose three mAbs (MPF25, MPF44, and MPF49) with the highest titers in each epitope group (Table 1). Six different ELISAs were constructed using different combination of mAbs, and we evaluated the linearity and sensitivity of each of them. We chose the pair of mAbs MPF25 and MPF49 to construct a sandwich ELISA because this combination showed the best performance (data not shown).

Measurement of MPF by ELISA. A sandwich ELISA was developed using the anti-MPF mAbs MPF25 and MPF49. The standard curve using purified recombinant MPF is shown in Fig. 3. Reproducibility of this sandwich ELISA is in an acceptable range over the range of 1 to 8 ng/mL of MPF (<10% of CV, n = 9). The assay was then used to determine if stably expressing mesothelin-transfectant A431/K5 cells that express the full-length mesothelin gene and are known to contain mesothelin protein on their surface release MPF into the medium. The data in Fig. 4 show that A431/K5 cells secrete MPF in a time-dependent manner over 6 days. As expected, no MPF was detected in the supernatants of the epidermoid carcinoma cell line A431 that does not express the mesothelin gene (Fig. 4). Finally, the assay was used for the detection of the MPF from the native cancer cells. The production of MPF by various cancer cell lines derived from different cancer types was measured using cells growing in 96-well culture plates. As shown in Table 3, three mesothelin-positive cell lines (HeLa, A1847, and IGROV1) secrete MPF. No MPF was detected in the media from mesothelin-negative OHS-M1 osteosarcoma and A431 epidermoid carcinoma cells.

MPF in serum of patients with mesothelioma. A total of 56 sera were obtained from patients with a histopathologic confirmation of mesothelioma. Sera from healthy volunteers (n = 70) with a median age 39 years were used to determine the cutoff for positivity in the sandwich ELISA. The average absorbance ±SD from healthy volunteers using a sandwich ELISA was 0.019 ± 0.005. Absorbance values of a 1:10 diluted serum that exceeded the mean absorbance value of sera from healthy donors by 3 SDs were considered positive. This cutoff based on absorbance values is similar to what has been described for other ELISA systems (19, 25). Thus, serum samples with absorbance values of ≥0.034 were considered to have elevated MPF levels. As shown in Fig. 5, MPF was not detected in healthy donors (n = 70). However, MPF was detected in the sera from 91% of patients with mesothelioma (51 of 56 patients). The average absorbance ±SD of the mesothelioma sera was 0.76 ± 0.63. These two groups were compared by nonparametric Mann-Whitney test. The P revealed that the difference between these groups was statistically very significant (P < 0.0001).

Effect of surgical treatment on serum MPF levels in patients with peritoneal mesotheliomas. To determine if serum MPF levels can be used to measure treatment response, we measured MPF levels in four patients with peritoneal mesothelioma before and after surgical tumor debulking. The clinical characteristics of the patients are summarized in Table 4. Three patients (A, B, and C) with mesothelin-expressing mesotheliomas had elevated serum MPF levels before therapy (Fig. 6). In patients A and B, the serum MPF levels decreased immediately after surgery. Patient C was found to have extensive tumor spread, and only limited tumor removal (∼10%) could be done. In this patient, post-surgical serum MPF values did not change significantly from the baseline. Patient D had a mesothelin-negative tumor, and no MPF was detected before or after surgery. These results indicate that MPF may be useful to monitor response to therapy.

Using mesothelin-deficient mice and immunization with recombinant MPF-Fc fusion protein, we generated several mAbs that specifically recognize MPF. Next, we did topographical epitope mapping of these anti-MPF antibodies and determined that they recognize three different epitope groups of the MPF protein. Using mAbs MPF25 and MPF49 that bind to two different epitopes on MPF, we developed an ELISA to measure MPF in cell culture supernatants and human serum.

Using our sandwich ELISA, we evaluated MPF levels in culture supernatant of several cell lines. Elevated MPF levels were noted in ovarian and cervical cancer cell lines that were mesothelin positive by FACS analysis. In contrast MPF levels in the culture supernatant were absent in mesothelin-negative epidermoid carcinoma and osteosarcoma cell lines. These results show that MPF is indeed shed into the media of cell lines that express mesothelin (Table 3).

To determine whether MPF is shed into blood, we analyzed serum samples from 70 normal healthy donors and 56 patients with advanced malignant mesothelioma. We choose to use serum samples from patients with mesothelioma because virtually all epithelial mesotheliomas overexpress mesothelin (26). Our results showed that MPF levels were elevated in 91% of these patients. In addition, using serum samples from patients with peritoneal mesothelioma who were undergoing debulking surgery, MPF levels decreased in patients who had optimal tumor debulking. In the current study, we used samples from patients with late-stage disease who came to the National Cancer Institute to be evaluated for a phase 1 clinical trial. In the future, we plan to test serum from patients with early-stage mesothelioma and other mesothelin-expressing cancers. We also plan to measure MPF levels in other groups of individuals, including humans exposed to asbestos and individuals of different ages.

The cDNA encoding mesothelin and MPF was identified by two different groups using different approaches. Our group identified mesothelin as the antigen recognized by mAbK1, which we obtained by immunizing mice with the ovarian cancer cell line OVCAR-3 (1, 3). On the other hand, MPF was identified by Yamaguchi et al. from the culture supernatant of a pancreatic cancer cell line and had some megarkaryocyte potentiating activity in mouse bone marrow cultures (2, 9). Both the MPF and mesothelin are encoded by the same gene that produces a precursor protein of 69 kDa. This precursor protein has a furin cleavage site (RPRFRR) between amino acids 288 and 293. Cleavage at this furin site results in the release of MPF from the cell surface while the mesothelin is attached to the cell membrane by a glycosylphosphatidylinositol anchor (17). We have recently shown that cell-bound mesothelin can also be released into the blood of patients with mesothelioma and ovarian cancer and may be a useful marker for mesothelin-expressing cancers (27). The release of mesothelin from the cell membrane is most likely mediated by phospholipase C or D or other proteases because glycosylphosphatidylinositol-linked proteins are frequently shed from the cell surface by such proteases (28, 29). In addition, soluble mesothelin-related proteins have been described in serum of patients with ovarian cancer and mesothelioma (25, 30).

Although MPF is presumed to be released by proteolytic cleavage of the 69-kDa precursor protein by furin, its detection in human blood has been hampered by the lack of antibodies that specifically recognize it. Using anti-MPF mAbs that we generated, we developed an ELISA to measure it in human serum. Elevated MPF levels were noted in 91% of serum samples obtained from patients with advanced mesothelioma. Because mesothelin is expressed by several other cancers, especially ovarian and pancreatic cancer, it will be important to study whether elevated MPF levels are also present in serum from patients with these cancers. MPF may also be a useful marker for these cancers. Additional studies are also needed to determine if serum MPF levels can be used as a surrogate marker for response to therapy. Furthermore, it will be important to correlate serum MPF levels with serum mesothelin levels in patients with mesothelin-expressing cancers.

We thank Dr. Mark Willingham (Department of Pathology, Wake Forest University, NC) for pathologic diagnoses and members of the Laboratory of Molecular Biology for useful discussions.