Identification of Cripto-1 as a Novel Serologic Marker for Breast and Colon Cancer

Human Cripto-1 (CR-1) is the founding member of the epidermal growth factor (EGF)-CFC family (Cripto in humans, FRL1 in Xenopus, and Cryptic in mice) of structurally related proteins that have been identified in several vertebrate species (1–4). The EGF-CFC protein family is characterized by the presence of two important functional domains: a modified EGF-like domain and a unique cysteine-rich CFC motif. The variant EGF-like motif in the EGF-CFC proteins is divergent from the canonical EGF-domain that is present in EGF-related peptides such as EGF, transforming growth factor α, and heregulins, and explains the observation that CR-1 does not directly interact with type I receptor tyrosine kinases (i.e., EGF-receptor, erbB-2, erbB-3, or erbB-4; refs. 5, 6). The COOH terminus contains, in some cases, consensus sequences for a glycosylphosphatidylinositol anchorage site that serves to attach the protein to the cell membrane. However, CR-1 can also be found in the conditioned medium of several cell lines, probably by cleavage of the glycosylphosphatidylinositol linkage by phosphatidyl inositol-phospholipase C or glycosylphosphatidylinositol-phospholipase D, suggesting that EGF-CFC proteins can function as secreted or cell membrane–associated proteins (7–9).

EGF-CFC proteins perform regulatory functions related to cell and tissue patterning during embryogenesis (1, 3, 4). CR-1 has been show to maintain the pluripotential capacity and self-renewal of human and mouse embryonic stem cells and is therefore considered a stem cell marker (10, 11). In the adult, CR-1 is reexpressed in an inappropriate fashion in a wide range of epithelial cancers, suggesting a link between stem cells and tumor biology (12). In this regard, CR-1 is expressed at elevated levels, as detected by immunohistochemistry, in 50% to 80% of different types of primary human carcinomas, including breast, colon, stomach, pancreas, lung, ovary, cervix, and testis (2, 13). The high frequency of expression of CR-1 in different types of human carcinomas, and its low expression in normal tissues, suggests that CR-1 may be a suitable target for therapy in human malignancies. In fact, anti-CR-1 antisense oligonucleotides or neutralizing blocking monoclonal antibodies have been shown to strongly inhibit the in vitro and in vivo growth of human breast, colon, ovarian, and testicular carcinoma cells (14–18). Experiments in vitro have shown that CR-1 can increase migration, invasion, epithelial-to-mesenchymal transition, and branching morphogenesis of several mouse and human mammary epithelial cell lines and can promote tumor angiogenesis (2, 19–23). Finally, transgenic mice studies have shown that overexpression of a human CR-1 transgene in the mammary gland under the control of the mouse mammary tumor virus (MMTV) or whey acidic protein promoter results in mammary hyperplasias and adenocarcinomas (19, 24, 25).

In the present study, we investigated whether CR-1, such as other glycosylphosphatidylinositol-linked tumor markers including carcinoembryonic antigen, can be released by tumor cells and reach the bloodstream, where it could be specifically detected for diagnostic and prognostic purposes (26). We therefore developed a sandwich-type ELISA to analyze CR-1 levels in plasma samples from healthy volunteers and from patients with colon and breast cancer. Immunohistochemistry and real-time reverse transcription-PCR (RT-PCR) were used to detect CR-1 expression in colon and/or breast tumor tissues.

Plasma from healthy donors and cancer patients. After obtaining informed consent, plasma samples were collected in EDTA-containing vacutubes from a total of 108 patients at ITN-Fondazione Pascale in Naples, Italy. Of these patients, 87 were diagnosed with colon and breast cancer and 21 patients were diagnosed with benign breast lesions (fibrocystic disease, adenosis, and fibrosis). Plasma samples from 21 healthy volunteers were also collected and analyzed as controls. All plasma samples were stored at −80°C until further analysis.

Plasma of MMTV-1 CR-1 transgenic mice. Plasma samples from MMTV-CR-1 transgenic mice and FVB/N control mice were centrifuged at 4°C for 15 minutes at 2,000 rpm and were stored at −80°C until further analysis.

Recombinant proteins and antibodies. Recombinant human CR-1, EGF, amphiregulin, heparin-binding EGF, and heregulin β1 were purchased from R&D Systems (Minneapolis, MN). The mouse monoclonal anti-CR-1 antibody (A10.B2.18) was kindly provided by Biogen-Idec (Cambridge, MA). Rabbit polyclonal anti-CR-1 antibody was developed by Biocon (Frederick, MD).

ELISA for CR-1 in plasma samples. A sandwich-type ELISA was developed for the quantitative measurement of soluble CR-1 in plasma samples. The anti-CR-1 monoclonal antibody A10.B2.18 (1 μg/well) was adsorbed to 96-well microtiter plates for 1 hour at room temperature. The plates were blocked with 2% milk, washed thrice with washing buffer (Kirkegaard & Perry Laboratories, Gaithersburg, MD) and incubated with 100 μL of undiluted plasma samples, CR-1 recombinant protein (diluted in 1% milk) at concentrations ranging from 3.7 pg/well (100 μL) to 2 ng/well (100 μL), and negative controls (2% milk blocking buffer) overnight at 4°C. Unbound CR-1 was removed by washing the plates five times with washing buffer (Kirkegaard & Perry Laboratories) and a rabbit polyclonal anti-CR-1 antibody (1:3,000; Biocon) was added to the plates for 1 hour at room temperature. The plates were then washed five times with washing buffer and incubated with goat anti-rabbit IgG conjugated to horseradish peroxidase (1:3,000; Amersham Pharmacia Biotech, Piscataway, NJ) for 1 hour at room temperature. The plates were developed with 3,3′,5,5′-tetramethyl-benzidine peroxidase substrate (Kirkegaard & Perry Laboratories) and absorbance was read at 450 nm. All the samples were assessed in duplicate.

Determination of sensitivity and specificity. Recombinant human CR-1 protein was used to generate a standard curve. Cross-reactivity with other homologous EGF-like proteins was investigated using purified recombinant EGF, amphiregulin, heparin binding-EGF, and heregulin β1 at a concentration of 20 μg/mL. The sensitivity of the CR-1 sandwich ELISA was defined as the ratio between true positives (TP) and the sum of true positives and false-negatives (FN) [TP / (TP + FN)]; specificity was defined as the ratio between true negatives (TN) and the sum of true negatives and false-positives (FP) [TN / (TN + FP)]. Accuracy was defined as the ratio between the sum of true positives and true negatives and the total number of subjects (n) [(TP + TN) / n].

Immunohistochemistry. After surgery, matching tumor tissue samples were obtained from 15 patients with breast carcinoma and 7 patients with colon carcinoma, and were used for immunohistochemistry. Each 5-μm paraffin-embedded section was dewaxed in xylene and rehydrated in a series of graded ethanols. Endogenous peroxidase activity was blocked with 3% H₂O₂ in methanol for 10 minutes. Antigen retrieval was done by incubating the sections with ready-to-use proteinase K solution (DakoCytomation, Carpinteria, CA) for 6 minutes at room temperature and washed several times in PBS. Immunostaining with anti-CR-1 monoclonal antibody (MAB2771, diluted to 1 μg/mL; R&D Systems) was done using the Vectastain ABC immunoperoxidase kit and Vector 3,3′-diaminobenzidine peroxidase substrate kit (both from Vector Labs, Burlingame, CA) following the manufacturer's instructions. Sections were counterstained with hematoxylin and observed with a light microscope. The intensity of CR-1 expression was evaluated by scoring the percentage of positive staining cells per field observed. The final score per section analyzed was calculated from the mean percentage of staining from at least four different fields observed at 400× magnification.

Real-time RT-PCR. For real-time RT-PCR, normal human breast from reduction mammoplasty, and breast cancer tissue samples were obtained from the Fox Chase Cancer Center. RNA was isolated from these samples as previously described (27). Quantification of CR-1 mRNA was done using TaqMan technology. The following primers and probe for the analysis of CR-1 mRNA expression were selected by using Primer Express software (PE Biosystems, Foster City, CA): forward primer 5′-CACGATGTGCGCAAAGAGA-3′; reverse primer 5′-TGACCGTGCCAGCATTTACA-3′; TaqMan probe [labeled with carboxyfluorescein (FAM)] 5′-CCATGACACCTGGCTGCCCAAGAA-3′. All RT-PCR reactions were done on the ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA) using the fluorescent TaqMan methodology (TaqMan One-Step RT-PCR Master Mix Reagents, Applied Biosystems). The PCR cycle at which the fluorescence intensity increases above the background signal is called the cycle threshold (Ct). In total, 100 ng of total RNA was used for each RT-PCR reaction in a total volume of 50 μL. Primers and probe concentration for the target gene were optimized according to the manufacturer's procedure. The thermal cycling conditions comprised 30 minutes at 48°C, 10 minutes at 95°C, and 40 cycles of 15-second denaturalization at 95°C and 60-second annealing at 60°C. Each sample was analyzed in triplicate, normalized against 18S as a control gene and was expressed in relation to a calibrator sample. Relative quantification of CR-1 mRNA within the samples examined was done by using the ΔΔCt method (ΔCt_sample − ΔCt_calibrator = ΔΔCt; relative quantity = 2^-ΔΔCt), as suggested by the manufacturer (PE Biosystems).

Statistical analysis. The nonparametric Mann-Whitney U test was used to assess the statistical significances of the differences between the various groups. Pearson's regression analysis and Kruskal-Wallis rank sum test were used to assess the correlation between CR-1 values in cancer patients and clinicopathologic variables. Statistical calculations were done with the use of Statistical Package for Social Sciences software package, version 11.0 (SPSS, Chicago, IL). All statistical tests were two-sided, and data were considered statistically significant when P < 0.05.

Human CR-1 sandwich ELISA. We developed an antibody sandwich ELISA to detect CR-1 in human plasma samples. Antibody sandwich ELISA has been shown to be more sensitive than ELISA, in which the antigen is directly bound to the plate (28, 29). An anti-CR-1 monoclonal antibody (A10.B2.18) was used as the capture antibody and an anti-CR-1 rabbit polyclonal antibody was used as the detection antibody. A secondary goat anti-rabbit polyclonal antibody, labeled with horseradish peroxidase, was then used and absorbance was measured at 450 nm. We used purified human CR-1 recombinant protein to generate a calibration curve. CR-1 was diluted in 1% milk to a concentration ranging from 0.037 to 20 ng/mL. The assay showed a linear relationship within this range (Fig. 1A). We next assessed cross-reactivities of mouse monoclonal and rabbit polyclonal anti-CR-1 antibodies. Because the EGF-like domain of CR-1 shares amino acid sequence similarity with members of the EGF family of peptides (3), we examined potential cross-reactivity with EGF, amphiregulin, heparin binding-EGF, or heregulin β1. All of the recombinant proteins produced readings comparable with the background signal, even at concentrations 1,000-fold higher than CR-1 (Fig. 1B). These data confirm that this sandwich ELISA measures CR-1 with high specificity because it efficiently discriminates CR-1 from other similar proteins.

Detection of human CR-1 in the plasma of MMTV-CR-1 transgenic mice. Because human CR-1 is overexpressed in the mammary gland of MMTV-CR-1 transgenic animals and because these animals, after several pregnancies, develop mammary hyperplasias and papillary adenocarcinomas, we evaluated whether soluble CR-1 could be detected in the plasma of MMTV-CR-1 transgenic animals (25). A total of four MMTV-CR-1 transgenic mice for each group were analyzed. We were not able to detect any CR-1 in the plasma of FVB/N control mice because the anti-CR-1 monoclonal antibody (A10.B2.18) used in the sandwich ELISA could only react with human CR-1 transgene protein and not with endogenous mouse Cr-1 protein (data not shown), as previously described (17). As shown in Fig. 1C, human CR-1 was detected in the plasma of MMTV-CR-1 mice during pregnancy (0.047 ± 0.0035 ng/mL; mean ± SD) and in mice with hyperplastic lesions of the mammary gland (0.041 ± 0.0056 ng/mL) or with mammary adenocarcinomas (0.052 ± 0.0021 ng/mL). Because the MMTV promoter induces the constitutive expression of high levels of CR-1 in the mammary gland of MMTV-CR-1 transgenic mice, no statistically significant differences in CR-1 levels were observed among these three groups of animals. Therefore, we conclude from these results that a sandwich ELISA for CR-1 can specifically detect human CR-1 in biological fluids, such as plasma.

Identification of CR-1 in plasma of patients with cancer. CR-1 levels were analyzed in 87 plasma samples derived from cancer patients with either colon carcinoma (n = 33) or breast carcinoma (n = 54) at different clinical stages (Table 1A). We also examined 21 plasma samples from healthy subjects. As shown in Table 1B and Fig. 2A, very low levels of CR-1 (mean ± SD) were detected in the plasma of healthy volunteers (0.32 ± 0.19 ng/mL). CR-1 levels in the plasma of patients with colon and breast carcinomas were significantly higher when compared with the control group (P < 0.001). The highest plasma levels of CR-1 were detected in colorectal cancer (4.68 ± 3.5 ng/mL; Table 1B). As a cutoff level, we chose the average of CR-1 values in the control group +2 SD (0.7 ng/mL). At this cutoff level, the test was positive in the plasma of all patients with cancer (100% sensitivity), and in 1 of 21 healthy donors (95% specificity). The accuracy of the test was 99%. No statistically significant differences for CR-1 plasma levels were found among patients with different stages of disease (Fig. 2B). Similarly, using two different univariate analysis tests, no statistically significant correlations were observed between CR-1 concentration in the plasma and various clinicopathologic variables in patients with breast and colon cancer, including tumor size, lymph node involvement, proliferative index as assessed by Ki-67 staining, estrogen and progesterone receptor status, or erbB-2 status where appropriate (data not shown).

Detection of CR-1 in the plasma of patients with benign breast lesions. We next assessed the levels of CR-1 in the plasma of 21 patients with benign breast lesions. These patients had clinical and radiological findings that were highly suggestive for breast carcinoma. Cytologic examination of fine-needle aspirate biopsy was not informative, or revealed the presence of atypical cells suggestive of hyperplasia. Therefore, these patients underwent surgical excision of the breast lesions. The plasma levels of CR-1 in patients carrying benign breast lesions were 1.7 ± 0.99 ng/mL (Table 2). The difference between the mean plasma CR-1 levels detected in patients with benign breast lesions and mean plasma CR-1 levels detected in patients with breast carcinoma (2.97 ± 1.48 ng/mL) was statistically significant (P < 0.001; Fig. 3). Within the group of patients with benign breast lesions, a trend towards higher levels of CR-1 were found in a subset of these patients (n = 7) with additional histologic signs of sclerosis, although this difference was not statistically significant (Table 2). Interestingly, high levels of plasma CR-1 were also found in a small group of patients (n = 6) with fibroadenoma, a pathologic breast benign lesion that is characterized by the presence of epithelial hyperplasia and fibrosis (Table 2). The difference in the mean CR-1 plasma levels between the fibroadenoma group and the nonsclerotic lesions is statistically significant (P < 0.05).

When a level of 2 ng/mL was arbitrarily chosen to separate benign breast lesions from breast carcinomas, the test was positive in 34 of the 51 patients with stage I or stage II breast cancer (67%), and in only 5 of 21 patients with benign breast lesions (24%; P < 0.001). Interestingly, four out of five benign breast lesions that were positive for CR-1, as detected by sandwich ELISA, were also characterized by the presence of sclerosis.

Immunohistochemical localization of CR-1 in tumor tissues from cancer patients. We next evaluated whether soluble CR-1, which was detected in the plasma of patients with cancer, originated from the primary tumor. Using immunohistochemistry, we analyzed CR-1 expression in colon and breast tumor tissue sections derived from patients that had high plasma levels of soluble circulating CR-1 protein. Specific staining for CR-1 was detected in all the colon cancer (n = 7) and breast cancer (n = 15) tissue sections that were examined (Fig. 4A). Weak to negative staining for CR-1 was detected in normal breast or colon tissue adjacent to the malignant lesions (data not shown). However, the levels of CR-1 in the plasma of patients with colorectal and breast cancer were not found to be associated with the degree of positivity for CR-1 in tumor sections (data not shown).

Detection of CR-1 mRNA in normal and tumor breast tissues. Finally, we assessed the levels of expression of CR-1 mRNA in another set of normal and malignant human breast tissues obtained by the Fox Chase Cancer Center by using real-time RT-PCR. Total RNA was obtained from normal and breast cancer tissues. As shown in Fig. 4B, CR-1 mRNA could be detected in normal breast tissue samples. More importantly, a large increase in the levels of CR-1 mRNA expression was observed in six human primary breast carcinomas as compared with normal breast tissue. These data are in agreement with the results from the plasma and immunohistochemical analysis for CR-1 expression in patients with breast cancer.

Despite the significant improvements in cancer therapy which have been achieved in the last 20 years, the prognosis for patients with cancer is mainly affected by the stage of the disease at diagnosis. The majority of human cancers are curable if diagnosed at an early stage, whereas death is common when the disease is discovered at a late stage. Therefore, the discovery of biomarkers for early diagnosis represents a major effort of cancer research.

Different biological markers have been proposed for the early detection of cancer. For example, early diagnosis of colorectal neoplasia can be accomplished by detection in the feces of mutations in ras or p53, or of changes in microsatellites of DNA (30). However, these assays are costly, and they can only be done in specialized molecular biology laboratories. Urokinase plasminogen activator and its receptor in nipple aspirate fluid samples are independent predictors of the presence of breast cancer (31). The feasibility of this procedure is significantly affected by the availability to obtain nipple aspirates, and by compliance of the patients with this approach. Serum markers for early diagnosis of cancer offer several advantages including the low cost and the possibility of detection with rapid and simple tests. However, the levels of a serum marker generally reflect tumor burden, and for this reason, they are not sensitive enough to be used for screening and early diagnosis of cancer (32). Growth factors, including transforming growth factor α and insulin-like growth factors, have also been proposed as potential tumor markers (33). Although epidemiologic evidence suggests a role for insulin-like growth factors and their binding proteins in the pathogenesis of different cancers, the value of serum insulin-like growth factors and insulin-like growth factor binding proteins for early diagnosis has not yet been established (34). In addition, some studies have also suggested that transforming growth factor α serum levels are elevated in patients with cancer compared with healthy controls; however, the diagnostic and/or prognostic value of serum transforming growth factor α in patients with carcinoma has not yet been proven (33, 35).

This article provides, for the first time, evidence that CR-1 might represent a more reliable marker for the detection of cancer. In fact, a significant increase in the plasma levels of CR-1 was observed in patients affected by two of the most frequent types of human carcinomas, breast, and colon, as compared with healthy volunteers. The sensitivity of the ELISA assay to detect patients with cancer was 100% with a specificity of 95%. Interestingly, higher CR-1 plasma levels were found in patients with breast cancer at an early stage of disease as compared with levels detected in patients with benign breast lesions and in healthy volunteers, suggesting that CR-1 might indeed be a useful marker in this setting. No correlation was found between tumor size and plasma levels of CR-1 in patients with colon and breast cancer. However, the absolute number of patients with advanced disease in our cohort was low compared with patients with early stage disease, and this might have affected our analysis. Furthermore, CR-1 exists in two biologically active forms as a cell membrane–associated, glycosylphosphatidylinositol-linked protein and as a soluble protein (7–9). No information is available on the mechanism(s) that regulate the processing of these two forms in human carcinomas as influenced by the expression of glycosylphosphatidylinositol-specific phospholipases by carcinoma cells. Finally, the amount of plasma CR-1 might be affected by the extent of necrosis within the tumor mass. In fact, no correlation was found between the intensity of staining in the primary tumor and the levels of CR-1 in the plasma of patients with cancer in whom this was assessed.

Our findings are in agreement with previous studies demonstrating that CR-1 is usually expressed in normal tissues at low levels, and that an increase in CR-1 expression is an early event in carcinogenesis (1, 4). In fact, no expression of CR-1 has been previously reported to occur in normal breast and colon tissue. In this regard, we found that CR-1 mRNA is expressed at low levels in normal human breast as compared with the levels of CR-1 mRNA in breast tumors. In contrast, a high frequency of CR-1 expression has been observed in colon tissue surrounding primary tumors and in the colon mucosa of individuals at high-risk, in whom CR-1 expression is associated with increased proliferation (36). The expression of CR-1 has also been observed in colon cancer premalignant lesions such as colon adenomas and in gastric metaplasia, suggesting that the expression of CR-1 may be a common event in human tumorigenesis (37, 38). In agreement with these findings, moderate levels of plasma CR-1 were found in women with non–cancer lesions of the breast including hyperplasia or atypical hyperplasia. In this subgroup of patients at high-risk, with clinical and radiological findings suggestive of breast carcinoma, preoperative levels of CR-1 might be a useful tool to help orient diagnosis and therapeutic intervention.

Within the nonmalignant breast lesions, the levels of plasma CR-1 were higher in fibroadenomas and in lesions characterized by the presence of sclerosis as compared with lesions without sclerosis. Recent evidence from studies of fibrotic disorders, including renal, liver, and lung fibrosis, supports the view that a substantial number of fibroblasts which are found in these conditions are derived from epithelial cells through a phenomenon termed epithelial-to-mesenchymal transition (39–41). Interestingly, our group has previously shown that overexpression of CR-1 in mouse and human mammary epithelial cells leads to morphologic and functional modifications that are typical of epithelial-to-mesenchymal transition (19, 22). Therefore, we hypothesize that in nonmalignant breast lesions with sclerosis, CR-1 might have a role in inducing epithelial-to-mesenchymal transition in mammary epithelial cells (19). The implications of this phenomenon in breast tumorigenesis warrant further exploration.

In conclusion, for the first time, this study provides evidence that plasma levels of CR-1 are elevated in patients with colon and breast carcinomas. The observation that CR-1 levels are elevated in patients with early breast cancer suggests that this marker might be useful for the early detection of cancer. Further prospective studies are needed to substantiate this hypothesis.