Serum protein signature may improve detection of ductal carcinoma in situ of the breast

Abstract

Ductal carcinoma in situ (DCIS) of the breast is part of a spectrum of preinvasive lesions that originate within normal breast tissue and progress to invasive breast cancer. The detection of DCIS is important for the reduction of mortality from breast cancer, but the diagnosis of preinvasive breast tumors is hampered by the lack of an adequate detection method. To identify the changes in protein expression during the initial stage of tumorigenesis and to identify the presence of new DCIS markers, we analysed serum from 60 patients with breast cancer and 60 normal controls using mass spectrometry. A 23-protein index was generated that correctly distinguishes the DCIS and control groups with sensitivities and specificities in excess of 80% in two independent cohorts. Two candidate peptides were purified and identified as platelet factor 4 (PF-4) and complement C3adesArg anaphylatoxin (C3adesArg) using liquid chromatography–tandem mass spectrometry (LC–MS/MS). In an independent serum set of 165 patients, PF-4 and C3adesArg were significantly upregulated in DCIS compared with non-cancerous controls, as validated using western blot and enzyme-linked immunosorbent assay. We conclude that our serum protein-based test, used in conjunction with image-based screening practices, could improve the sensitivity and specificity of breast cancer detection.

Introduction

Since the widespread introduction of screening mammography, ductal carcinoma in situ (DCIS) stands out among breast cancer diagnoses. The incidence of DCIS has increased 10-fold in the past two decades, especially among women older than 50 years. Approximately 64 000 cases of DCIS are diagnosed annually in the United States, and it accounts for 20% of all incidences of breast cancer (Jemal et al., 2008). Recently, however, the value of mammography screening has been questioned, as the incidence of stage 2 and stage 3 breast cancer has not significantly declined. This suggests a bias in the detection of early-stage cancers, especially in women with dense breast tissue and in premenopausal women (Elmore et al., 2005). A recent European Organization for Research and Treatment of Cancer study involving more than 1000 patients showed that 66% of young women, compared with 25% of women over 40 years, were diagnosed with DCIS after discovering symptomatic lesions. These young women are not routinely tested by mammography screening as the mammography screening age range is 40–70 years in Europe (Bijker et al., 2006). Alternative imaging methods, such as magnetic resonance imaging, have been proposed. Although it is a very sensitive detection tool and has become the standard for women at high risk of developing breast cancer, it lacks sufficient specificity and cost effectiveness for use as a general screening tool (Esserman et al., 2007). Therefore, there is an urgent need to develop complementary approaches to improve the sensitivity and specificity of DCIS screening.

As whole blood is considered to provide a dynamic representation of an individual's physiological and pathological status, human serum/plasma represents the most extensively studied biological matrix in the quest for cancer biomarkers (Hanash et al., 2008). However, most of the biomarkers currently used in clinical settings are frequently elevated in patients with metastatic disease. Therefore, these markers have insufficient predictive value as a screening blood assay and, in conjunction with diagnostic imaging, history and physical examination, are only recommended for monitoring patients with metastatic disease during active therapy (Harris et al., 2007). The search for specific disease-associated biomarker signatures is, consequently, driven by the need to use multiple biomarkers to increase diagnostic power with respect to preinvasive breast cancer (Levenson, 2007).

Recent progress in proteomics has unlocked novel avenues for the discovery of cancer-related biomarkers (Hanash, 2003; Cho, 2007). In recent years, liquid chromatography–tandem mass spectrometry (LC–MS/MS) has been used for the detection and identification of the panels of proteins and is likely to have an increasing role in oncoproteomics (Van der Merwe et al., 2007). A related technology, surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI–TOF), combined with computational methodologies, has also been used to determine protein profiles (Solassol et al., 2006). Numerous studies have already shown that this methodology can be used to uncover proteomic expression patterns linked with cancer, and some expression patterns have shown high promises in the detection of early-stage cancers (Petricoin and Liotta, 2004; Kang et al., 2005; Kozak et al., 2005; Jacot et al., 2008).

The aim of this study was to establish a serum protein signature for DCIS and to evaluate the sensitivity and specificity of this protein profile for diagnosing preinvasive breast cancer. To do this, protein profiles were generated in a learning set from the serum of patients diagnosed with DCIS and noncancerous controls using SELDI–TOF. Several statistical algorithms were developed to compare the spectra from the two groups of samples. The robustness of the protein panel was then tested using an independent validation set. Finally, we aimed to identify putative biomarkers using LC–MS/MS to provide insight into the pathological processes involved in DCIS and to analyse possible clinical diagnostic assays.

Results

Significant differences exist between the serum of ductal carcinoma in situ and normal patients

We first tested our experimental reproducibility using a pool of normal serum samples spotted randomly onto CM10 chips together with experimental serum samples. The coefficients of variation of 15 reliable peaks were calculated by average peak intensity values derived from 20 different runs. Acceptable average coefficients of variation of 10 and 17% were obtained for intra- and inter-assay variability, respectively. This is consistent with previously reported studies using SELDI–TOF (Semmes et al., 2005; Albrethsen, 2007).

A total of 60 serum samples corresponding to the discovery serum set 1 (30 DCIS and 30 normal samples) were included in a subsequent analysis using SELDI–TOF. Fractionated serum, rather than whole serum, was used in this analysis because fractionation markedly increases the resolution and sensitivity of protein peak detection without any loss of minor proteins within the range of molecular weights detected using SELDI–TOF (Solassol et al., 2005; Mange et al., 2008). A total of 289 peaks were picked and clustered by the Biomarker Wizard software (Ciphergen, Fremount, CA, USA; see Excel files of absolute linear and normalized log-transformed intensity values of all serum in Supplementary Material Table S1). This set of discriminating peaks was then further refined to a 23-protein peak signature with the best prediction performance derived from the discovery set (P<0.05 and intensity 2). Among these peaks, 12 were overexpressed and 11 were underexpressed in serum from patients with breast cancer when compared with the control samples. The mean amplitude of these peaks for the two groups of patients is given in Table 1. These peaks have areas under the receiver operating characteristic curve ranging from 0.67 to 0.80, indicating possible diagnostic utility, especially when several of these peaks could be used to build a classifier.

Table 1 Twenty-three statistically significant protein peaks are differentially expressed in DCIS and control samples

Multivariate analyses

To identify a multiprotein signature that can differentiate serum from individuals with DCIS from healthy controls, a heat cluster map, indicating the up- and downregulation of possible biomarker candidates, was generated for a multivariate analysis of the samples in the discovery set. Protein peaks between 2.5 and 50.0 kDa revealed four normal and two DCIS samples that were misclustered (Figure 1). We then used the 23-protein peak panel to determine the diagnostic performance using a support vector machine (SVM) classification algorithm, taking into account the biomarker intercorrelations. SVM achieved an overall classification accuracy of 96.7%, a sensitivity of 96.5% and a specificity of 96.7% in distinguishing normal samples from DCIS samples. An area under the curve (AUC) linear multiprotein index was used as a second classification method and confirmed the performance of our protein signature with an accuracy, sensitivity and specificity very similar to those above and an AUC value for the composite index (0.992) higher than those of the individual protein peaks (0.67–0.80). Table 2 indicates the mean performance of both classification methods. On the basis of the results of the test set, we validated our classifiers with a second set of samples (serum set 2), which included 60 independent samples (30 DCIS and 30 normal samples). The results indicate that overall classification accuracies were maintained with 83.0% (sensitivity 85.2% and specificity 81.2%) and 81.4% (sensitivity 82.1% and specificity 80.7%) for SVM classification and AUC linear multiprotein index classification indexes, respectively.

Figure 1
figure1

Protein expression profiling of serum from breast cancer and control patients. An overview of a hierarchical clustering analysis obtained from 23 protein profiles of 30 breast cancer and 30 control serum samples is shown. Each row label represents a protein with a given m/z, and each column label represents an individual case. The intensity of the red or green color indicates that the relative protein concentration is higher or lower than the mean value, respectively.

Table 2 Diagnostic performance of classifiers

Peak identification

As a first step to identify the 23 biomarkers from our protein signature, we selected the 5905 and 8926 Da peaks for further analysis. These peaks were selected because of their clear separation from other peaks and strikingly different expressions in DCIS versus control patients (Figures 2a and 3a). The proteins were first enriched on Q HyperD F anion and S HyperD F cation columns before size exclusion fractionation. The eluted fractions were then subjected to tricine–sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE). After silver staining, the bands with an apparent molecular weight close to those expected were excised from the gel, destained and prepared for both direct in-gel tryptic digestion and passive elution. The proteins were identified by tryptic digestion followed by nano-LC–MS/MS. The 5905 and 8926 Da markers were identified as platelet factor 4 (PF-4) and complement C3adesArg anaphylatoxin (C3adesArg) with sequence coverages of 43 and 29%, respectively (Figures 2b and 3b). The identification of PF-4 and C3adesArg was validated using a mass spectrometric immunoassay with polyclonal anti-human PF-4 and anti-human C3adesArg antibodies. We incubated a DCIS serum sample with immobilized antibodies to bind and deplete these proteins from the serum and then analysed the immune-depleted serum using SELDI–TOF. The expected peaks of PF-4 (5905 Da, Figure 2c) and C3adesArg (8926 Da, Figure 3c) were completely depleted with the corresponding specific antibody, confirming the identity of the 5905 and 8926 Da peaks.

Figure 2
figure2

Identification of the candidate 5905 Da marker. (a) Representative spectra from two cancer-free patients (upper panel) and two patients with DCIS (lower panel). The frame indicates the position of the 5905 Da peak overexpressed in fraction 2 of DCIS patients. An overlay of the 5905 Da peak obtained from the serum of patients with DCIS (red) and healthy individuals (blue) is shown in the upper box. (b) Mass spectrometric identification of PF-4 using nano-liquid chromatography–tandem mass spectrometry (LC–MS/MS) sequencing. The amino acid sequences obtained using MS/MS are in bold. The characterized peptides are listed in the table with their position in the PF-4 sequence. (c) Immunodepletion of PF-4 using a polyclonal antibody. Fraction 2 of a DCIS serum sample (upper panel), the PF-4 immunodepleted sample (middle panel) and the eluted fraction (lower panel) were analysed using CM10 ProteinChip array and surface-enhanced laser desorption/ionization time-of-flight mass spectrometry.

Figure 3
figure3

Identification of the candidate 8926 Da marker. (a) Representative spectra from two cancer-free patients (upper panel) and two patients with DCIS (lower panel). The frame indicates the position of the 8926 Da peak overexpressed in fraction 1 of DCIS patients. An overlay of 8926 Da peak obtained from the serum of patients with DCIS (red) and healthy individuals (blue) is shown in the upper box. (b) Mass spectrometric identification of complement C3adesArg anaphylatoxin using nano-liquid chromatography–tandem mass spectrometry (LC–MS/MS) sequencing. The amino acid sequences obtained using MS/MS are in bold. The characterized peptides are listed in the table with their position in the complement C3adesArg anaphylatoxin (C3adesArg) sequence. (c) Immunodepletion of complement C3adesArg using a monoclonal antibody. Fraction 1 of a DCIS serum sample (upper panel), the complement C3adesArg immunodepleted sample (middle panel) and the eluted fraction (lower panel) were analysed using CM10 ProteinChip array and surface-enhanced laser desorption/ionization time-of-flight mass spectrometry.

From their m/z ratio, the fraction from which they were derived, the ProteinChip surface of capture, as well as data from previously performed serum profiling studies (Rai et al., 2002; Kozak et al., 2005; Goncalves et al., 2006; Ward et al., 2006; Roesch-Ely et al., 2007; Timms et al., 2007), it was possible to identify several other potential protein biomarkers present in the multiprotein index. Thus, the 15 120, 15 870 and 79235 Da proteins that were upregulated in the serum of DCIS patients were identified as α- and β-hemoglobin and transferrin, respectively. These identities were similarly confirmed using MS/MS analysis (data not shown).

Validation of PF-4 and C3adesArg expression in serum

As PF-4 and C3adesArg seemed to be upregulated in the serum of patients with DCIS, the level of expression of each protein was analysed in serum from 30 additional newly diagnosed DCIS patients and 95 non-cancerous control patients (healthy donors, patients with autoimmune diseases (AID) and patients with benign breast lesions (BBL)). PF-4 expression was significantly higher in the serum of patients with DCIS as compared with that of non-cancerous patients (Figure 4), confirming our SELDI–TOF results. C3adesArg expression was also found to be significantly higher in the DCIS serum samples (19.05 ng/ml, 95% confidence interval (CI) 18.06–20.05) as compared with healthy donor serum samples (12.88 ng/ml, 95% CI 12. 62–13.15), AID samples (14.30 ng/ml, 95% CI 14.12–14.49) and benign breast lesion samples (15.70 ng/ml, 95% CI 14.13–17.28), as well as compared with all non-cancerous control samples collectively (13.91 ng/ml, 95% CI 13.57–14.26; Figure 5). Finally, we evaluated PF-4 and C3adesArg expression levels in early breast disease progression. In all, 20 newly diagnosed T1N0 and 20 T2N0 breast cancers were analysed. PF-4 and C3adesArg expression levels were significantly higher in early invasive breast cancer than in non-cancerous samples. However, no significant difference was observed between DCIS and invasive breast cancer samples for either protein (Figures 4 and 5).

Figure 4
figure4

Expression of PF-4 in the serum of breast cancer and control patients. (a) A western blot for PF-4 was performed on 30 DCIS, 40 healthy donors, 40 AID, 15 BBL and 40 early-stage breast cancer samples. (b) The PF-4 band intensity was quantified in DCIS and non-cancerous samples and normalized to a single sample. The median is indicated as a horizontal line.

Figure 5
figure5

Expression of C3adesArg anaphylatoxin (C3adesArg) in the serum from breast cancer and control patients. The concentration of complement C3adesArg in the serum was determined using enzyme-linked immunosorbent assay on the same samples used for the PF-4 western blot. The median is indicated as a horizontal line.

Discussion

The potential benefit from a detection methodology designed to identify early-stage breast cancer is clear. Mammography has been shown to be the most effective screening tool for detecting breast cancer early and for saving lives. However, mammography has intrinsic limitations that may be difficult to overcome. The sensitivity of mammography ranges between 63 and 87%, depending on age, breast density and tumor characteristics (Elmore et al., 2005). Therefore, diagnostic biomarkers may be helpful in increasing the positive predictive value of mammographic lesions, thereby decreasing the number of women who undergo unnecessary biopsies (Orel et al., 1999). In addition, biomarkers may also be used to select cases for more sensitive diagnostic techniques, such as magnetic resonance imaging. Finally, diagnostic biomarkers may be important for younger women (40 years) who are not recommended for regular mammography.

In this study, we determined whether protein profiling using SELDI–TOF analysis may be complementary to the conventional screening methods for smaller breast tumors (<50 mm). Proteins that were differentially expressed in the serum of breast cancer patients and cancer-free controls were selected and subjected to biostatistical analysis combining monoparametric (Biomarker Wizard) and two multiparametric indices (SVM classification and AUC linear multiprotein index classification). A 23-protein index was generated that correctly distinguishes the DCIS and control groups with sensitivities and specificities in excess of 80% in two independent cohorts. Importantly, the components identified by this multiprotein index include PF-4, C3adesArg, transferrin and α- and β-hemoglobin. Finally, PF-4 and C3adesArg were validated to be significantly upregulated in the serum of 30 newly diagnosed DCIS patients as compared with 135 non-cancerous subjects, showing their potential usefulness as markers of preinvasive breast tumors.

A similar use of SELDI–TOF has been published previously, further suggesting that protein profiling can be used to diagnose breast cancer (Li et al., 2002, 2005; Vlahou et al., 2003; Becker et al., 2004; Hu et al., 2005; Gast et al., 2006, 2008; Mathelin et al., 2006; Song et al., 2006; Belluco et al., 2007; Brozkova et al., 2008). However, most of these studies include samples from women with late-stage and advanced metastatic disease, and the use of this technique for early diagnosis and/or cancer screening has been limited. Recently, Belluco et al. (2007) reported excellent capability of their seven-peak classifier to discriminate serum from stage 1 breast cancer patients from that of controls in an independent cohort. However, and importantly, none of the seven protein peaks were structurally identified. As SELDI–TOF is likely to identify false-positive biomarkers, the identification and validation of one or several protein peaks from the protein index is essential.

In our study, we identified two proteins using LC–MS/MS that were found to be significantly upregulated in DCIS serum samples, a truncated form of PF-4 and C3adesArg anaphylatoxin. Human C3 is the most abundant complement protein in serum (about 1.2 mg/ml) and is made up of a and b fragments (Sahu and Lambris, 2001). The proteolytic activation of native C3 occurs by either the classical (C4b, C2a) or alternative (C3b, Bb) pathway. C3 convertase cleaves C3 between residues 726 and 727 (Arg–Ser), generating C3b and an N-terminal fragment, C3a (8.9 kDa) (Sahu and Lambris, 2001). C3a has high biological activity and is able to trigger the degranulation of mast cells and basophils, which produces a local inflammatory response (the desArg form represents a stable inactivated form of complement C3a). It is very difficult to ascertain whether biomarker proteins are generated by the cancer itself or by secondary effects, such as an inflammatory reaction. The mechanism by which chronic inflammation is related to cancer prognosis remains unclear. Inflammatory proteins have been shown to have relative relevance as serum prognostic biomarkers. For example, Pierce et al. (2009) has very recently proposed several inflammatory circulating proteins as important prognostic markers for long-term survival in breast cancer patients, independent of race, tumor stage and body mass index. It is likely that concomitant inflammation in breast cancer is causing the increase in the level of C3adesArg in our study. Indeed, in patients suffering from AID, C3adesArg was also significantly increased as compared with controls (Figure 5). Interestingly, Li et al. (2002) observed increased serum peaks of 8.9 and 8.1 kDa, which allowed discrimination between breast cancer patients and healthy individuals. These peaks were further identified as C3adesArg (8.9 kDa) and a C-terminal truncated form of this protein (C3adesArg8, 8.1 kDa) using structural analysis (Li et al., 2005). Subsequent analysis of an independent sample set confirmed an increase of the 8.1 and 8.9 kDa C3a fragments in breast cancer (Li et al., 2005), which was confirmed by a third study (Mathelin et al., 2006). Finally, C3adesArg is not unique to breast cancer and has been shown to be elevated in other cancers, such as ovarian cancer (Bjorge et al., 2005), chronic lymphoid malignancies (Miguet et al., 2006), hepatocellular carcinoma (Lee et al., 2006) and colon cancer (Ward et al., 2006). We also identify, for the first time, PF-4 as a potential biomarker in early-stage breast cancer. PF-4 is a tetrameric, lysine-rich member of the CXC chemokine family and is produced almost exclusively by megakaryocytes. PF-4 is an important modulator of endothelial cell proliferation (Maione et al., 1990; Gupta and Singh, 1994) and angiogenesis (Sharpe et al., 1990; Hagedorn et al., 2001). PF-4 may influence tumor growth by modifying the effects of vascular endothelial growth factor or by binding and neutralizing heparin and other sulfated glycosaminoglycans that are required for the binding of proangiogenic factors (Gengrinovitch et al., 1995; Perollet et al., 1998). PF-4 has been recently identified by SELDI–TOF as a biomarker of early tumor growth in mice (Cervi et al., 2008). In our work, we found that a cleaved form of PF-4 was significantly increased in DCIS as compared with the controls. The cleavage site is located downstream from the N-terminal CXC motif and corresponds to a loss of the first large loop of the PF-4 molecule. Cleavage has been previously shown to greatly potentiate the endothelial cell growth inhibitory activity of PF-4, showing that the N-terminal processing of PF-4 is important for endothelial cell regulation (Gupta et al., 1995). This proteolytic regulation of PF-4 may be an important physiological mechanism that controls the response to neoplasia. The increase in platelet-associated PF-4 in dormant (nonangiogenic) tumors may, therefore, reflect the functional inhibition of angiogenesis in preinvasive breast cancer, particularly in intermediate- and high-grade DCIS lesions that secrete large amounts of vascular endothelial growth factor (Pavlakis et al., 2008). Finally, although increases in C3adesArg and PF-4 do not seem to be breast cancer specific, as currently observed for individually used biomarkers, their identification contributes important information to our 23-protein index. Our serum protein-based test, used in conjunction with image-based screening practices, could improve the sensitivity and specificity of breast cancer detection. In addition, the highly potent derivative PF-4, described in this study, may have direct therapeutic potential for early treatment of angiogenic DCIS with high risk of invasion.

There has been strong criticism that SELDI–TOF-based analyses identify only highly abundant and acute phase proteins. Of the five markers we identified, three were highly abundant and acute-phase proteins. Transferrin has been identified by SELDI–TOF as a putative ovarian (Rai et al., 2002; Kozak et al., 2005), colon (Ward et al., 2006) and breast (Goncalves et al., 2006) cancer biomarker. α- and β-Hemoglobin have also been identified as putative markers for ovarian and head and neck cancer (Roesch-Ely et al., 2007). With this in mind, neither transferrin nor hemoglobin can serve as robust breast cancer biomarkers unless they are used in combination with additional validation studies. Several reports show that these potential biomarkers need to be further analysed. Indeed, transferrin has been shown to regulate the metastatic capacity of various solid tumors, including breast cancer (Nicolson et al., 1992; Cavanaugh and Nicolson, 1998; Cavanaugh et al., 1999). In addition, transferrin was also shown to promote an angiogenic phenotype (Carlevaro et al., 1997). We were initially concerned with the potential use of hemoglobin, as mechanical handling and/or sample preparation can result in red blood cell lysis and hemoglobin release. However, as we omitted samples that were obviously red from our analyses, we considerably reduced the possibility that the hemoglobin identified in our screen was an artifact. In addition, Alexander et al. (2004) found the level of hemoglobin β-chain isoforms to be associated with breast cancer in nipple aspirate fluid, which is normally free of blood contamination. As biochemical modifications of the membrane of erythrocytes in women with ovarian cancer may increase susceptibility to hemolysis of red blood cells, α- and β-hemoglobin have been considered as potential markers for ovarian cancer (Kopczynski et al., 1998).

Early detection remains one of the most urgent issues in breast cancer research (Kuerer et al., 2009). Our study shows a potential pattern of biomarkers in the serum of DCIS patients that can clearly classify tumors with good sensitivity and specificity. Importantly, the classification methods used were shown to be robust in independent validation set. However, this study presents several limitations. Several patients were misclassified: two DCIS samples corresponding to low-grade lesions (<20 mm) were incorrectly classified as normal. The small size likely explains this false-negative result and proves that validation with large numbers of patients is needed before more widely applicable immunoassays can be developed. In future studies, the specificity of these biomarkers for the detection of breast cancer needs to be addressed by analysing specimens from other types of cancer. After our results, a prospective study is now ongoing at our institution. Finally, no significant correlation was found between the concentration of the markers and any histological (such as size or inflammatory status) or biological parameter. Therefore, the discriminatory power of these markers most likely reflects the malignant nature of the tumor rather than its histological or molecular characteristics.

We are aware that further analyses in cohorts of at-risk patients are needed to validate a final diagnostic protein profile for preinvasive breast cancer. However, our findings may serve as a ‘proof-of-concept’, supporting important implications for prospective assessment of proteomic profiling as a screening tool. We provide additional data on the potential advantage of the SELDI–TOF technique to detect small tumors. Furthermore, we identified several potential biomarkers in DCIS that also belong to the protein profile of more advanced stages of breast cancer and identified a truncated form of PF-4 as a potential biomarker. We conclude that SELDI–TOF should be applied for prospective assessment for screening purpose.

Materials and methods

Patient selection

Female breast cancer patients treated at the Val d'aurelle Cancer Center (Montpellier, France) between September 2004 and February 2008 were enrolled. The study protocol was approved by the Comités-Consultatifs-de-Protection-des-Personnes (Montpellier) and the RBM03-63-INSERM review board. All patients provided informed consent to participate in the study.

The discovery study population consisted of 120 subjects: 60 women who underwent surgery and had a histopathological diagnosis of breast cancer and 60 age-matched unaffected women with negative mammograms, negative physical breast exams for at least 4 years and no history of previous malignancy, immunodeficiency, autoimmune disorder, hepatitis or HIV infection. The patients were staged according to the American Joint Committee on Cancer Staging Manual, sixth edition (Greene et al., 2002). The discovery study population was subsequently subdivided into serum set 1 and serum set 2 for the SELDI analysis. The validation study population consisted of 30 additional newly diagnosed DCIS patients (serum set 3), 40 healthy donors, 40 patients with AID accompanied by inflammation (rheumatoid arthritis, n=20; systemic lupus erythematosus, n=20), 15 patients with benign breast lesions (ductal micropapillomatosis, n=2; fibroadenoma, n=2; mastitis, n=1; papilloma, n=1; fibroepithelial lesions, n=3; atypical ductal hyperplasia, n=6) and 40 patients with early-stage primary breast cancers (T1N0, n=20 and T2N0, n=20). The characteristics of the DCIS population are reported in Table 3. All of the samples were collected by the same procedure and were immediately stored at −80 °C as previously described (Jacot et al., 2008).

Table 3 Clinicopathological characteristics of ductal carcinomas in situ

Proteomic analysis

To optimize the peak resolution and the number of protein peaks detected, an anion-exchange fractionation procedure was performed as described previously (Solassol et al., 2005). Each fraction was applied to a weak cation-exchange ProteinChip array surface (CM10). For ProteinChip array binding, CM10 arrays were equilibrated with binding buffer (100 mm sodium acetate, pH 4), and 10 μl of the fractionated eluate and 90 μl of binding buffer were added to each spot. After 30 min, the arrays were washed thrice with binding buffer for 5 min and rinsed twice with water. A saturated solution of sinapinic acid in 50% acetonitrile and 0.5% (v/v) trifluoroacetic acid was applied to the chip array surface twice on each spot as an energy absorbing matrix. The arrays were read on a Protein Biological System II ProteinChip reader according to the manufacturer's instructions (Ciphergen). Peak detection was performed using the ProteinChip Biomarker software (version 3.0, Ciphergen).

Reproducibility

The accuracy of the mass detection of the ProteinChip Biomarker System, version IIc (PBSIIc) was calibrated externally using the All-in-1 protein II molecular mass standards (Ciphergen). The protein data were collected by averaging a total of 192 laser shots at a laser intensity of 185 and sensitivity of 8 in a positive mode.

The reproducibility was estimated using a control sample that was randomly spotted on each chip to measure the variability of fractionation, on-chip spotting and data acquisition. The intensity values for the selected peaks were used to calculate the intra- and inter-array coefficients of variation.

Peak detection

The analysis of the peaks was performed using the ProteinChip and Biomarker Wizard software (Ciphergen). The background was subtracted from the spectra, and the peak intensities were normalized to the total ion current of m/z between 2.5 and 50 kDa. Automatic peak detection was performed with the following settings: (1) signal-to-noise ratio of four for the first pass and two for the second pass, (2) minimal peak threshold at 15% of all spectra and (3) cluster mass window at 0.5% of mass.

Protein identification

For protein identification, peaks of interest were enriched from the serum samples using Q HyperD F anion and S HyperD F cation columns, followed by size exclusion fractionation. The eluted fractions were then subjected to tricine–sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and in-gel tryptic digestion followed by nanoLC–MS/MS (for details, see Supplementary Material).

Platelet factor 4 western blot

To further confirm the protein chip results, 100 μg of each serum sample was denatured, separated in a 16% tricine–sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) gel (Schagger, 2006), and transferred onto a polyvinylidene fluoride membrane (Millipore, Molsheim, France), which was then blocked for 1 h at room temperature with 5% skim milk in phosphate-buffered saline. PF-4 was detected by incubation overnight at 4 °C with 1:1000 of an anti-human PF-4 antibody (PreProtech, Rocky Hill, NJ, USA). The membranes were then incubated with horseradish peroxidase–conjugated goat anti-rabbit immunoglobulin G antibody (Jackson ImmunoResearch Laboratories, West Grove, PA, USA) at a 1:3000 dilution for 1 h at room temperature. The protein signals were detected using a Super Signal West Pico Chemiluminescent Substrate kit (Pierce, Rockford, IL, USA) followed by autoradiography on Hyperfilm ECL (GE Healthcare, Orsay, France). The densities of bands were quantified using ImageJ (version 1.41o, National Institutes of Health, Bethesda, MA, USA). The sample densities were normalized to the density of a single control sample that was run on each blot.

Immunoassays

The human complement C3adesArg enzyme immunoassay was carried out in duplicate according to the manufacturer's instructions (Assay Designs, Stressgen, Ann Arbor, MI, USA). For this assay, the serum samples were diluted 1:200.

Statistical data analysis

The univariate statistical analysis of SELDI–TOF peak masses and relative intensity value determinations were conducted using the nonparametric Mann–Whitney U-test. The differentially expressed proteins were defined with a P-value<0.05. The data were analysed using various statistical algorithms including AUC linear multiprotein index classification and SVM analysis. SVM was chosen because this learning scheme is known to perform well when the number of samples in the data set is small relative to the number of attributes. In our case, the attributes were the peaks. SVM identified the maximum margin hyperplane, which was the hyperplane separating the two classes of samples in an n-dimensional space while maximizing the distance between the hyperplane and the closest training point. SVM was implemented using the MultiExperiment Viewer software (Mev, version 4, Boston, MA, USA) (Saeed et al., 2003). The AUC linear multiprotein index classification characterizes the discrimination between two well-defined populations and was implemented on mROC software (Montpellier, France) (Kramar et al., 2001).

For the validation set, statistical analyses were performed using GraphPad InStat (version 3.06, La Jolla, CA, USA). Student's t-test and Mann–Whitney U-test were used for the comparison of two groups. One-way analysis of variance with Bonferroni multiple comparison post-test and nonparametric analysis of variance with Dunn's multiple comparison post-test were used for comparisons among more than two groups. A probability level of P<0.05 was chosen for statistical significance.

Conflict of interest

The authors declare no conflict of interest.

References

  1. Albrethsen J . (2007). Reproducibility in protein profiling by MALDI-TOF mass spectrometry. Clin Chem 53: 852–858.

  2. Alexander H, Stegner AL, Wagner-Mann C, Du Bois GC, Alexander S, Sauter ER . (2004). Proteomic analysis to identify breast cancer biomarkers in nipple aspirate fluid. Clin Cancer Res 10: 7500–7510.

  3. Becker S, Cazares LH, Watson P, Lynch H, Semmes OJ, Drake RR et al. (2004). Surfaced-enhanced laser desorption/ionization time-of-flight (SELDI-TOF) differentiation of serum protein profiles of BRCA-1 and sporadic breast cancer. Ann Surg Oncol 11: 907–914.

  4. Belluco C, Petricoin EF, Mammano E, Facchiano F, Ross-Rucker S, Nitti D et al. (2007). Serum proteomic analysis identifies a highly sensitive and specific discriminatory pattern in stage 1 breast cancer. Ann Surg Oncol 14: 2470–2476.

  5. Bijker N, Meijnen P, Peterse JL, Bogaerts J, Van Hoorebeeck I, Julien JP et al. (2006). Breast-conserving treatment with or without radiotherapy in ductal carcinoma-in-situ: ten-year results of European Organisation for Research and Treatment of Cancer randomized phase III trial 10853--a study by the EORTC Breast Cancer Cooperative Group and EORTC Radiotherapy Group. J Clin Oncol 24: 3381–3387.

  6. Bjorge L, Hakulinen J, Vintermyr OK, Jarva H, Jensen TS, Iversen OE et al. (2005). Ascitic complement system in ovarian cancer. Br J Cancer 92: 895–905.

  7. Brozkova K, Budinska E, Bouchal P, Hernychova L, Knoflickova D, Valik D et al. (2008). Surface-enhanced laser desorption/ionization time-of-flight proteomic profiling of breast carcinomas identifies clinicopathologically relevant groups of patients similar to previously defined clusters from cDNA expression. Breast Cancer Res 10: R48.

  8. Carlevaro MF, Albini A, Ribatti D, Gentili C, Benelli R, Cermelli S et al. (1997). Transferrin promotes endothelial cell migration and invasion: implication in cartilage neovascularization. J Cell Biol 136: 1375–1384.

  9. Cavanaugh PG, Jia L, Zou Y, Nicolson GL . (1999). Transferrin receptor overexpression enhances transferrin responsiveness and the metastatic growth of a rat mammary adenocarcinoma cell line. Breast Cancer Res Treat 56: 203–217.

  10. Cavanaugh PG, Nicolson GL . (1998). Selection of highly metastatic rat MTLn2 mammary adenocarcinoma cell variants using in vitro growth response to transferrin. J Cell Physiol 174: 48–57.

  11. Cervi D, Yip TT, Bhattacharya N, Podust VN, Peterson J, Abou-Slaybi A et al. (2008). Platelet-associated PF-4 as a biomarker of early tumor growth. Blood 111: 1201–1207.

  12. Cho WC . (2007). Contribution of oncoproteomics to cancer biomarker discovery. Mol Cancer 6: 25.

  13. Elmore JG, Armstrong K, Lehman CD, Fletcher SW . (2005). Screening for breast cancer. JAMA 293: 1245–1256.

  14. Esserman LJ, Shieh Y, Park JW, Ozanne EM . (2007). A role for biomarkers in the screening and diagnosis of breast cancer in younger women. Expert Rev Mol Diagn 7: 533–544.

  15. Gast MC, Bonfrer JM, van Dulken EJ, de Kock L, Rutgers EJ, Schellens JH et al. (2006). SELDI-TOF MS serum protein profiles in breast cancer: assessment of robustness and validity. Cancer Biomark 2: 235–248.

  16. Gast MC, van Tinteren H, Bontenbal M, van Hoesel RQ, Nooij MA, Rodenhuis S et al. (2008). Haptoglobin phenotype is not a predictor of recurrence free survival in high-risk primary breast cancer patients. BMC Cancer 8: 389.

  17. Gengrinovitch S, Greenberg SM, Cohen T, Gitay-Goren H, Rockwell P, Maione TE et al. (1995). Platelet factor-4 inhibits the mitogenic activity of VEGF121 and VEGF165 using several concurrent mechanisms. J Biol Chem 270: 15059–15065.

  18. Goncalves A, Esterni B, Bertucci F, Sauvan R, Chabannon C, Cubizolles M et al. (2006). Postoperative serum proteomic profiles may predict metastatic relapse in high-risk primary breast cancer patients receiving adjuvant chemotherapy. Oncogene 25: 981–989.

  19. Greene FL, Page DL, Fleming ID et al. (2002). Cancer Staging Manual, 6th edn. Springer: New York.

  20. Gupta SK, Hassel T, Singh JP . (1995). A potent inhibitor of endothelial cell proliferation is generated by proteolytic cleavage of the chemokine platelet factor 4. Proc Natl Acad Sci USA 92: 7799–7803.

  21. Gupta SK, Singh JP . (1994). Inhibition of endothelial cell proliferation by platelet factor-4 involves a unique action on S phase progression. J Cell Biol 127: 1121–1127.

  22. Hagedorn M, Zilberberg L, Lozano RM, Cuevas P, Canron X, Redondo-Horcajo M et al. (2001). A short peptide domain of platelet factor 4 blocks angiogenic key events induced by FGF-2. FASEB J 15: 550–552.

  23. Hanash S . (2003). Disease proteomics. Nature 422: 226–232.

  24. Hanash SM, Pitteri SJ, Faca VM . (2008). Mining the plasma proteome for cancer biomarkers. Nature 452: 571–579.

  25. Harris L, Fritsche H, Mennel R, Norton L, Ravdin P, Taube S et al. (2007). American Society of Clinical Oncology 2007 update of recommendations for the use of tumor markers in breast cancer. J Clin Oncol 25: 5287–5312.

  26. Hu Y, Zhang S, Yu J, Liu J, Zheng S . (2005). SELDI-TOF-MS: the proteomics and bioinformatics approaches in the diagnosis of breast cancer. Breast 14: 250–255.

  27. Jacot W, Lhermitte L, Dossat N, Pujol JL, Molinari N, Daures JP et al. (2008). Serum proteomic profiling of lung cancer in high-risk groups and determination of clinical outcomes. J Thorac Oncol 3: 840–850.

  28. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T et al. (2008). Cancer statistics, 2008. CA Cancer J Clin 58: 71–96.

  29. Kang X, Xu Y, Wu X, Liang Y, Wang C, Guo J et al. (2005). Proteomic fingerprints for potential application to early diagnosis of severe acute respiratory syndrome. Clin Chem 51: 56–64.

  30. Kopczynski Z, Kuzniak J, Thielemann A, Kaczmarek J, Rybczynska M . (1998). The biochemical modification of the erythrocyte membranes from women with ovarian cancer. Br J Cancer 78: 466–471.

  31. Kozak KR, Su F, Whitelegge JP, Faull K, Reddy S, Farias-Eisner R . (2005). Characterization of serum biomarkers for detection of early stage ovarian cancer. Proteomics 5: 4589–4596.

  32. Kramar A, Faraggi D, Fortune A, Reiser B . (2001). mROC: a computer program for combining tumour markers in predicting disease states. Comput Methods Programs Biomed 66: 199–207.

  33. Kuerer HM, Albarracin CT, Yang WT, Cardiff RD, Brewster AM, Symmans WF et al. (2009). Ductal carcinoma in situ: state of the science and roadmap to advance the field. J Clin Oncol 27: 279–288.

  34. Lee IN, Chen CH, Sheu JC, Lee HS, Huang GT, Chen DS et al. (2006). Identification of complement C3a as a candidate biomarker in human chronic hepatitis C and HCV-related hepatocellular carcinoma using a proteomics approach. Proteomics 6: 2865–2873.

  35. Levenson VV . (2007). Biomarkers for early detection of breast cancer: what, when, and where? Biochim Biophys Acta 1770: 847–856.

  36. Li J, Orlandi R, White CN, Rosenzweig J, Zhao J, Seregni E et al. (2005). Independent validation of candidate breast cancer serum biomarkers identified by mass spectrometry. Clin Chem 51: 2229–2235.

  37. Li J, Zhang Z, Rosenzweig J, Wang YY, Chan DW . (2002). Proteomics and bioinformatics approaches for identification of serum biomarkers to detect breast cancer. Clin Chem 48: 1296–1304.

  38. Maione TE, Gray GS, Petro J, Hunt AJ, Donner AL, Bauer SI et al. (1990). Inhibition of angiogenesis by recombinant human platelet factor-4 and related peptides. Science 247: 77–79.

  39. Mange A, Bellet V, Tuaillon E, Van de Perre P, Solassol J . (2008). Comprehensive proteomic analysis of the human milk proteome: contribution of protein fractionation. J Chromatogr B Analyt Technol Biomed Life Sci 876: 252–256.

  40. Mathelin C, Cromer A, Wendling C, Tomasetto C, Rio MC . (2006). Serum biomarkers for detection of breast cancers: A prospective study. Breast Cancer Res Treat 96: 83–90.

  41. Miguet L, Bogumil R, Decloquement P, Herbrecht R, Potier N, Mauvieux L et al. (2006). Discovery and identification of potential biomarkers in a prospective study of chronic lymphoid malignancies using SELDI-TOF-MS. J Proteome Res 5: 2258–2269.

  42. Nicolson GL, Cavanaugh PG, Inoue T . (1992). Differential stimulation of the growth of lung-metastasizing tumor cells by lung (paracrine) growth factors: identification of transferrin-like mitogens in lung tissue-conditioned medium. J Natl Cancer Inst Monogr 13: 153–161.

  43. Orel SG, Kay N, Reynolds C, Sullivan DC . (1999). BI-RADS categorization as a predictor of malignancy. Radiology 211: 845–850.

  44. Pavlakis K, Messini I, Vrekoussis T, Yiannou P, Keramopoullos D, Louvrou N et al. (2008). The assessment of angiogenesis and fibroblastic stromagenesis in hyperplastic and pre-invasive breast lesions. BMC Cancer 8: 88.

  45. Perollet C, Han ZC, Savona C, Caen JP, Bikfalvi A . (1998). Platelet factor 4 modulates fibroblast growth factor 2 (FGF-2) activity and inhibits FGF-2 dimerization. Blood 91: 3289–3299.

  46. Petricoin EF, Liotta LA . (2004). SELDI-TOF-based serum proteomic pattern diagnostics for early detection of cancer. Curr Opin Biotechnol 15: 24–30.

  47. Pierce BL, Ballard-Barbash R, Bernstein L, Baumgartner RN, Neuhouser ML, Wener MH et al. (2009). Elevated biomarkers of inflammation are associated with reduced survival among breast cancer patients. J Clin Oncol 27: 3437–3444.

  48. Rai AJ, Zhang Z, Rosenzweig J, Shih Ie M, Pham T, Fung ET et al. (2002). Proteomic approaches to tumor marker discovery. Arch Pathol Lab Med 126: 1518–1526.

  49. Roesch-Ely M, Nees M, Karsai S, Ruess A, Bogumil R, Warnken U et al. (2007). Proteomic analysis reveals successive aberrations in protein expression from healthy mucosa to invasive head and neck cancer. Oncogene 26: 54–64.

  50. Saeed AI, Sharov V, White J, Li J, Liang W, Bhagabati N et al. (2003). TM4: a free, open-source system for microarray data management and analysis. Biotechniques 34: 374–378.

  51. Sahu A, Lambris JD . (2001). Structure and biology of complement protein C3, a connecting link between innate and acquired immunity. Immunol Rev 180: 35–48.

  52. Schagger H . (2006). Tricine-SDS-PAGE. Nat Protoc 1: 16–22.

  53. Semmes OJ, Feng Z, Adam BL, Banez LL, Bigbee WL, Campos D et al. (2005). Evaluation of serum protein profiling by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry for the detection of prostate cancer: I. Assessment of platform reproducibility. Clin Chem 51: 102–112.

  54. Sharpe RJ, Byers HR, Scott CF, Bauer SI, Maione TE . (1990). Growth inhibition of murine melanoma and human colon carcinoma by recombinant human platelet factor 4. J Natl Cancer Inst 82: 848–853.

  55. Solassol J, Jacot W, Lhermitte L, Boulle N, Maudelonde T, Mange A . (2006). Clinical proteomics and mass spectrometry profiling for cancer detection. Expert Rev Proteomics 3: 311–320.

  56. Solassol J, Marin P, Demettre E, Rouanet P, Bockaert J, Maudelonde T et al. (2005). Proteomic detection of prostate-specific antigen using a serum fractionation procedure: potential implication for new low-abundance cancer biomarkers detection. Anal Biochem 338: 26–31.

  57. Song J, Patel M, Rosenzweig CN, Chan-Li Y, Sokoll LJ, Fung ET et al. (2006). Quantification of fragments of human serum inter-alpha-trypsin inhibitor heavy chain 4 by a surface-enhanced laser desorption/ionization-based immunoassay. Clin Chem 52: 1045–1053.

  58. Timms JF, Arslan-Low E, Gentry-Maharaj A, Luo Z, T'Jampens D, Podust VN et al. (2007). Preanalytic influence of sample handling on SELDI-TOF serum protein profiles. Clin Chem 53: 645–656.

  59. Van der Merwe DE, Oikonomopoulou K, Marshall J, Diamandis EP . (2007). Mass spectrometry: uncovering the cancer proteome for diagnostics. Adv Cancer Res 96: 23–50.

  60. Vlahou A, Laronga C, Wilson L, Gregory B, Fournier K, McGaughey D et al. (2003). A novel approach toward development of a rapid blood test for breast cancer. Clin Breast Cancer 4: 203–209.

  61. Ward DG, Suggett N, Cheng Y, Wei W, Johnson H, Billingham LJ et al. (2006). Identification of serum biomarkers for colon cancer by proteomic analysis. Br J Cancer 94: 1898–1905.

Download references

Acknowledgements

This work was supported in part by grants from the Comité de l'Hérault de la Ligue Contre le Cancer. Serum collection was supported by INSERM (RBM 03-63).

Author information

Correspondence to A Mangé.

Additional information

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Solassol, J., Rouanet, P., Lamy, P. et al. Serum protein signature may improve detection of ductal carcinoma in situ of the breast. Oncogene 29, 550–560 (2010). https://doi.org/10.1038/onc.2009.341

Download citation

Keywords

  • breast
  • diagnosis
  • serum
  • profiling

Further reading