Comparison of tear protein levels in breast cancer patients and healthy controls using a de novo proteomic approach

Noninvasive biomarkers are urgently needed for early detection of breast cancer since the risk of recurrence, morbidity and mortality are closely related to disease stage at the time of primary surgery. In the past decade, many proteomics-based approaches were developed that utilize the protein profiling of human body fluids or identification of putative biomarkers to obtain more knowledge on the effects of cancer emergence and progression. Herein, we report on an analysis of proteins in the tear fluid from breast carcinoma patients and healthy women using a de novo proteomic approach and 25 mixed samples from each group. This study included 25 patients with primary invasive breast carcinoma and 25 age-matched healthy controls. We performed a MALDI-TOF-TOF-driven semi-quantitative comparison of tear protein levels in cancer (CA) and control (CTRL) using a de novo approach in pooled samples. Over 150 proteins in the tear fluid of CTRL and CA were identified. Using an in-house-developed algorithm we found more than 20 proteins distinctly upregulated or downregulated in the CTRL and CA groups. We identified several proteins that had modified expression in breast cancer patients. These proteins are involved in host immune system pathways (e.g., C1Q1 or S100A8) and different metabolic cascades (ALDH3A or TPI). Further validation of the results in an independent population combined with individual protein profiling of participants is needed to confirm the specificity of our findings and may lead to a better understanding of the pathological mechanism of breast cancer.


Introduction
Breast cancer is still the leading cause of death in women worldwide (1). Although the detection rate of breast carcinoma has improved, many female patients die from metastatic relapse. Mammography is the best available method for detection of breast cancer after the age of 50; although, the detection rate of mammography is not as good in younger women due to their high density breast tissue (2). Early detection is beneficial in the fight against breast cancer. Currently, there are no clinical biomarkers available for early detection of breast cancer. Markers such as CA15. 3 and CEA are useful, in combination with imaging and physical examination, for monitoring ongoing treatment in breast cancer patients with metastatic disease; although, they both lack the clinical specificity and sensitivity to be used routinely as a clinical diagnostic tool (3).
The development of high-throughput techniques in Proteomics expanded the search for new biomarkers and enabled the identification of proteins that may have a crucial role in emerging and progressing breast cancer. Proteome analysis of body fluids, such as sera, tear film, or urine, is a hot topic in Proteomics (4)(5)(6). Li et al found three differently regulated proteins in the sera of breast cancer patients and healthy subjects using surface-enhanced laser desorption/ionization time-offlight based protein profiling in 2002 and Mathelin et al tried to validate these putative biomarkers, determining only two of them could be used for the discrimination of cancer patients (7,8) (reviewed in refs. 9,10). Some studies examined the nipple aspirate fluid of breast cancer patients and healthy patients (11). In 2005, Pawlik et al showed 17 distinctly regulated peptides; whereas, Li et al found different protein distribution patterns in the nipple aspirate fluid and ductal lavage with the use of SELDI-TOF mass spectrometry (12,13). Since then, many other protein profiling studies were published that used matrixassisted laser desorption/ionization time-of-flight/time-of-flight mass spectrometry with differently regulated proteins (14)(15)(16). The advantage of the MALDI-TOF-TOF MS is the subsequent identification of the proteins of interest. In a previous study, we reported data from MALDI-TOF-TOF-based profiling of the sera that could distinguish breast cancer patients from agematched healthy controls, and we could classify cancer patients with a high sensitivity of 89% (17).
Another proteomics-based approach for the exploration of cancer-derived differences is the highly-precise microarray platform. This approach can serve, instead of the common ELISA test, as a validation tool for the biomarkers identified from prior MALDI-TOF-TOF-based explorations of the proteome. Here, the antibodies are fixed on a highly-optimized surface. In this manner, several protein levels can be measured simultaneously due to the small required volume (nl) of the reagents. After fixation of the antibodies, the surfaces can be incubated with body fluids containing the appropriate proteins. This high-throughput technique is also very common for the profiling of carcinoma tissue or body fluids of diseased patients due to its miniaturized size, accuracy, and automated handling (18)(19)(20) (reviewed in ref. 21). Several comparative studies of breast cancer and healthy sera have been published. Our study group reported the regulation of several proteins were significantly different in the sera of breast cancer patients (22). The discovery of different protein patterns in diseased cohorts and control samples and subsequent identification of these biomarkers is a promising method of obtaining knowledge about the effects of several diseases (6,23,24). A well-developed and clinically proven biomarker signature could lead to early detection of cancer, which can have great benefits for patients.
Most proteomic studies of breast carcinoma published so far concentrate on profiling the tissue or body fluids near the emergence spot. Little is known about the proteome changes of distant body fluids. Some research groups examined the protein profiles of alternative body fluids such as urine or saliva and several differently regulated proteins were reported (reviewed in refs. 25,26). Previously, we showed different protein distributions in the tear fluid of breast cancer patients and healthy controls in a SELDI-TOF-based profiling study (17,27). Another comparative MALDI-TOF-TOF-driven analysis of healthy dog's tear fluid and dogs diagnosed with cancer has been published (28). To our knowledge, no other comparative tear fluid proteomic studies for breast cancer have been reported. Tear fluid has unique properties as retrieval is minimally invasive and it does not contain as many highly-abundant proteins as serum.
Herein, we report a MALDI-TOF-TOF-driven semi-quantitative comparison of tear protein levels in cancer (CA) and control (CTRL) using a de novo approach in pooled samples. Using a signature of biomarkers significantly decreased or increased in groups of CA and CTRL could help to discriminate diseased women from the healthy population with high specificity and sensitivity and possibly lead to the establishment of a molecular diagnostic tool for breast cancer.

Materials and methods
Comparison of tear protein levels in pooled samples from CA and CTRL. This de novo study included 50 female subjects, 25 patients were diagnosed with primary invasive breast carcinoma and treated at the University Medical Center Mainz. At the time of diagnosis, none of the patients had developed distant metastases. Patients' characteristics are summarized in Table I. The healthy control subjects were 25 age-matched women without any known malignancies who were treated at the University of Mainz medical center. All study members gave their informed consent for voluntary participation in this study. The protocols were approved by the institutional ethics committee in accordance with the ethical standard of Declaration of Helsinki (1964).
Sample retrieval. Tear fluid was obtained from all participants using a Schirmer Strip. After the samples were drawn, the strips were frozen immediately at -80˚C to prevent protein degradation. Tear proteins were prepared under strict and identical conditions for all patients. Prior to the experiments, the wet strip part was cut into small pieces and incubated with n-Dodecyl-β-D-maltoside overnight at 4˚C with constant shaking. The next day, the eluates were briefly centrifuged and transferred into fresh tubes. All samples were stored prior to analysis at -20˚C.
Sample processing. For the comparison of protein levels in CTRL and CA, each of the 25 tear eluates were pooled together accordingly to the group and precipitated with three times the volume of acetone overnight at -80˚C. The next day, tear proteins were centrifuged at 14000 x g and 4˚C to prevent protein degradation. The supernatant was discarded and the proteins were resuspended in PBS. Protein concentrations were measured with the BCA Protein Assay kit (Thermo Scientific, Rockford IL, USA), according to the manufacturer's protocol.
1D SDS-PAGE and sample purification. Pooled tear proteins (60 µg) from CTRL and CA were separated by molecular weight using 1D SDS-PAGE (gels, buffers, and equipment all purchased from Invitrogen, Darmstadt, Germany). After gel electrophoresis, the lanes were stained overnight and then rinsed with double-distilled water. In the next step, the lanes were subdivided into 32 bands and the proteins were digested with endopeptidase trypsin according to the modified digestion  (29). For the purification and desalting of peptides, automated sample handling was preferred to reduce the fluctuations from measured proteins due to manual processing of the samples. The purification and the stepwise elution of the peptides with 10-50% acetonitrile were performed using C18 ZipTips (Millipore, Billerica, USA) on the Freedom EVO ® , purification station (Tecan Group Ltd., Männedorf, Switzerland). The eluted proteins (3 µl) were directly spotted on the MALDI TOF/TOF polished steel target and coated with 3 µl crystallization matrix (20 mg cinnamic acid/50% acetonitrile/2% trifluoroacetic acid). The matrix included 0.5 µl of a Reserpine solution (1 mg/ml) dissolved in methanol for signal normalization. All samples were measured head-to-head to avoid protein degradation and measurement fluctuations in the MALDI-TOF/ TOF mass spectrometer (UltraflexII, Bruker Daltonik GmbH, Bremen, Germany). The peak detection was performed with internal calibration mix (Peptide calibration standard II, Bruker Daltonik GmbH).
Data processing. After MALDI-TOF/TOF measurements, the spectra were exported into Proteomics Pipeline Mainz (P 2 M) software, developed in-house, and normalized according to the Reserpine peaks. Proteins were identified using the MASCOT protein search tool (30). The Swissprot database was chosen for the identification of proteins (31). The following general parameters were used: carbamidomethyl as a global modification and oxidation (M) as a variable modification with an MS tolerance of 100 ppm and MS/MS tolerance of 0.8. Only one miscleaved site was allowed and the MudPIT scoring system was used. For further analysis of the protein regulation levels, the intensities of the peptides for each protein were summed and the ratio of the intensity between both groups was calculated for each protein. Significant differences in protein expression levels were defined as at least two times higher or lower expression than the other group. STRING and Cytoscape software were used for the analysis of protein-protein interactions (32,33).

Results
In this study we conducted an explorative and comparative analysis of the tear proteome of breast carcinoma patients and age-matched healthy controls. We tried to minimize protein degradation and fluctuations of protein measurements to achieve a precise comparison of protein levels. One person performed the experimental steps for the preparation of tear samples for 1D SDS-PAGE until the transfer of digested fractions onto the sample plate for the robotic purification station. The peptide purification was performed automatically to avoid fluctuations due to the manual handling of samples. Likewise, the experimental steps from the precipitation of the tear eluates were also performed by the same person.
Semiquantitative comparison of protein levels in CA and CTRL. After destaining, a grid made of 32 bands was put under the gel for a more accurate comparison of the proteins. Each of the 32 bands from CTRL and CA were cut out and digested with trypsin overnight. Fig. 1 shows the samples after 1D SDS-PAGE separation and staining with Coomassie dye (Colloidal Blue Staining kit, Invitrogen). After digestion and automated fractionation, the peptides were measured in a MALDI-TOF-TOF mass spectrometer. Representative fractions from both groups are shown in Fig. 2. All spectra were normalized using Protein Pipeline Mainz software, which was developed in-house, and the appropriate tear proteins were identified with the MASCOT search tool.
Protein identification. After extensive comparison of the spectra obtained using the annotated proteins in the SWISSPROT Homo sapiens database under the given conditions and MOWSE score, we were able to identify over 150 proteins in the CTRL and CA. The complete merged list of identified proteins is summarized in Table II. To obtain an overview on the relevance and role of the identified proteins, we clustered the proteins in accordance to their molecular functions using the software Cytoscape 2.7.0, as shown in Fig. 3. The Cytoscape software often shows several overlapping molecular functions and distributions into several biological processes; therefore, we created an overview of one mapping possibility for a large number of the identified tear proteins.
Using the in-house-developed algorithm, we compared the protein levels in both groups. More than 20 proteins were distinctly upregulated or downregulated in the CTRL and CA groups and were involved in many biological processes such as metabolism (ALDH3A or TPI) or immune response (e.g., C1Q1 or S100A8). Table III shows a detailed list of the increased or decreased proteins in the tear fluid of breast cancer patients. Of note, the findings include inflammation proteins or complement factors for pathologic processes such as cancer that have already been described (34)(35)(36). Moreover, several proteins show at least four-fold higher (Extracellular sulfatase Sulf-1, Cystatin SA, cst2; 5-AMP-activated protein kinase subunit gamma-3, prkag3; Triosephosphate isomerase, tpi1; Microtubule-associated tumor suppressor 1, mtus1; Transferrin receptor protein 1, trfc; and Putative lipocalin 1-like protein 1, lcn1l1) or lower levels (DNA damage-binding protein 1, ddb1; Protein S100-A9, s100a9; and GTP-binding protein Di-Ras2, diras2) in CA. An overview of the proteins differently regulated in the CA group was constructed according to their regulation using the STRING tool and is shown in Fig. 4 (32).

Discussion
Data from high-throughput proteomic technologies, such as SELDI-TOF MS, MALDI-TOF-TOF MS, and microarray platforms, have recently increased. These techniques allow simultaneous protein profiling and subsequent identification of proteins and their subunits (5,37,38). A huge number of proteome studies have been published for proteome comparison of cancer patients and controls. Likewise, different proteomic studies reported significant differences in protein levels in the body fluids of breast cancer patients and healthy subjects (38,39). In our study, we concentrated on the tear proteome for several reasons. First, the sample retrieval is minimally invasive for the participants and tear fluid is easy to obtain with a simple Schirmer test. Second, the tear proteome contains no high-abundant proteins, such as albumin and immunoglobulins that are found in serum; therefore, it is not necessary to perform additional depletion steps that may cause distortion of potentially important proteins. In addition, we find it very intriguing to explore the tear proteome for potential biomarkers of breast cancer as it is an uncommon approach.
Some of the differently regulated proteins in our de novo pooled experiment have been reported to be altered in the tear fluid of patients with ophthalmic disease. Zhou et al reported S100A8 and S100A9 are increased in patients with dry eyes and Grus et al reported an increase in protein S100A8 (34,40). Both proteins belong to the family of S100 calcium-binding proteins, whose members seem to be involved in pro-inflammatory pathways as previously reported by Nacken et al (35). Some Figure 2. Mass spectra of the digested and purified peptides from the CTRL and CA groups. The molecular weight of the digested peptides is shown on the x-axis; while, the y-axis represents the intensity of the signals obtained. The spectra show the peptides derived from the first gel fraction in both groups. Spectra 1 and 2 show the peptides eluted with 10% acetonitrile (ACN). Spectra 3 and 4 demonstrate the peptides after elution with 15% ACN. Spectra 5 and 6 represent peptides from the elution step with 30% ACN. All spectra show the reproducible high intensity of signals. The mass spectra show appropriate differences in the peptide patterns according to the CTRL and CA pools.   repair. In addition, many of the differently regulated proteins are involved in metabolic processes, e.g., TPI or MDH1 in glycolysis and the citric acid cycle, which are both increased in the tear fluid of cancer patients. However, higher levels of autoantibodies against TPI1 have been reported in the sera of breast cancer patients (36). In our previous studies, we found several alterations in protein expression in the sera and tear fluid of breast cancer patients (22,41). Further analysis of the SELDI-TOF-based tear proteome profiling identified the protein S100A4 to be increased in the tears of breast cancer patients (data not shown). This result was confirmed in this study. The protein S100A4 was also previously found to be upregulated in patients with dry eye syndrome (40). Noteworthy, we observed several alterations in the level of proteins involved in immune response, such as complement factor C1Q1 or fragments of immunoglobulins (Table II). Also, several complement factors  have been reported to be differentially regulated in the sera of cancer patients (42,43). Although, some of the results were controversial and may have resulted from different storage and handling conditions (44). Thus, members of the complement system may have additional roles. Markiewski et al reported tumor growth was promoted by C5a in their experiments with a cervical cancer mouse model (45,46). To our knowledge, little is known about protein expression in the tear fluid of breast cancer patients. Only a very small number of tear proteome studies concerning proteome changes during breast cancer or cancer in general have been published. Further subsequent analyses and validation of our results in a tear protein study with an independent population and a higher number of participants will follow that also includes individual profiling. The findings from this study are intriguing as they may deepen the understanding of the impact of cancer and several cancer-driven pathways. Our study demonstrates that different biological processes are altered not only in prominent and broadly investigated body fluids such as serum and plasma, but also in discrete fluids such as tears that are located far away from the cancer site. As we already mentioned, several proteins have been reported to be modified in various types of body fluids, such as nipple aspirate fluid or urine. Our pilot study adds to these findings and shows again the complexity and multiple impacts of breast cancer while emerging and developing in the host, affecting biological processes and signal cascades. Moreover, we propose that a biomarker panel consisting of different proteins could accurately discriminate cancer patients from healthy controls. Therefore, it is important to examine the protein levels in an independent study population using individual protein profiling to validate our results. Further de novo approaches and validation of our results could lead to a better understanding of the pathological mechanism of breast cancer.