Discovery and verification of gelsolin as a potential biomarker of colorectal adenocarcinoma in a Chinese population: Examining differential protein expression using an iTRAQ labelling-based proteomics approach

1Anal-Colorectal Surgery Institute; 2Department of Pathology; 3Department of Clinical Laboratory, 150th Hospital of PLA, LuoYang, China Correspondence and reprints: Dr Chun-Fang Gao, 150th Hospital of PLA, No.1, HuaXia Road, Luoyang, Henan 47 1000, China. Telephone 86-379-64169327, fax 86-379-64169112, e-mail chunfanggao@126.com Received for publication March 23, 2011. Accepted June 8, 2011 Approximately 1,529,560 cancer-related deaths are estimated to occur annually in the United States; of these, 142,570 are attributed to colorectal cancer (CRC) (1). Due to early detection, clinical prognoses have improved significantly; however, robust screening tests are needed for all types of gastrointestinal cancer. Although a wide variety of screeening modalities for CRC exist, only 40% of cases are detected at early stages (2). The identification of novel protein biomarkers or therapeutic targets may ultimately improve patient care and survival. Proteomics combined with mass spectrometry (MS) has become a powerful method for the global examination of proteins in the postgenomics era, enabling the discovery of cancer biomarkers and drug targets. Although transcriptomics provides the tools for unravelling gene expression networks, proteomics links these networks to protein products and provides additional insight into post-translational modifications that regulate cellular functions, thereby complementing genomic analyses. The identification of differentially expressed proteins in CRC using proteomics suggests that expression patterns of proteins may have some utility in the discovery of cancer biomarkers. Differential tagging with isotopic reagents, such as isotope-coded affinity tags (3), or the more recent variation that uses isobaric tags for relative and absolute quantitation (iTRAQ) reagents followed by multidimensional liquid chromatography (LC) and tandem MS/MS analysis, is emerging as a powerful methodology in the search for disease biomarkers. Many recent studies (4-11) using iTRAQ reagents resulted in the identification and relative quantification of proteins that led to the discovery of potential cancer markers for prostate, radiotherapy-resistant, renal cell, lung, oral, head and neck, breast and oRiginAl ARTiClE

In the present study, we used iTRAQ labelling in combination with two-dimensional (2D) liquid chromatography (LC) and MS to identify potential biomarkers of CRC by comparing protein profiles of tumour tissue with paired normal mucosa, which in turn may facilitate the early diagnosis of CRC. Some of the proteins identified were confirmed by Western blot analysis and immunohistochemistry (IHC). These approaches ensure that the selected proteins demonstrated a consistent pattern of differential expression in CRC and significantly increased the confidence of the observations obtained from the iTRAQ analysis. In addition to their potential utility as biomarkers for CRC, these proteins also provide valuable insight into the still unknown molecular networks and mechanisms that govern the normal-to-malignant conversion of the epithelium.

Patients
Data from all CRC patients who underwent surgical resection at the 150th Hospital, PLA, Luoyang, China, between January and June 2010, were retrieved from the files of the department of pathology. The criteria for inclusion in the study were as follows: presence of sporadic tumours without evidence of polyposis or familial predisposition syndrome; availability of clinical information and tumour stage; availability of tissue material; and a diagnosis of colorectal adenocarcinoma. The criteria for exclusion from the study were the presence of palindromic or metastatic tumour(s), and patients receiving antitumour therapy including chemotherapy, radiotherapy or biotherapy. Of the patients fulfilling these criteria, 66 cases were randomly selected. The patient consent forms and tissue-banking procedures were approved by the Research Ethics Board of the 150th Hospital. Histological diagnosis of each sample was confirmed using microscopic examination of hematoxylin and eosin-stained sections obtained during radical correction surgery.

Tissue material
Paired tissue samples of primary tumours (obtained from surgical resections) and their normal mucosa (not less than 5 cm from the cancerous margin) were flash frozen in liquid nitrogen within 20 min of devitalization after radical correction surgery (n=5, used for iTRAQ and Western blot analysis) or fixed in buffered formalin and embedded in paraffin (n=61, used for IHC).
The frozen samples were sectioned and stored at −80°C. The tissue from the opposite face of the histological section was then washed three times in approximately 1 mL of phosphate buffered saline (PBS) containing a mixture of protease inhibitors as described previously (12) (1 mM 4-[2-aminoethyl] benzenesulfonyl fluoride, 10 μM leupeptin, 1 μg/mL aprotinin and 1 μM pepstatin). The washed tissues were then homogenized in 0.5 mL of PBS containing protease inhibitors using a handheld homogenizer. These homogenates were then flash frozen in liquid nitrogen and stored at −80°C until use.
Neoplastic lesions were classified histologically according to the Vienna classification of digestive epithelial neoplasia (13). The tumour, node, metastsis (TNM) stage of the adenocarcinomas was determined according to the American Joint Committee on Cancer classification (14). The following data were recorded: localization, histological classification and TNM stage for adenocarcinomas (Table 1).

Protein digest, iTRAQ labelling and strong cation-exchange fractionation
Cancerous and paired normal flash-frozen samples from five patients were thawed and clarified by centrifugation, and protein concentrations were determined with a Bradford-type assay using a protein quantification reagent (Bio-Rad, USA). Cancerous samples and paired normal samples were pooled in equal parts and precipitated.
The samples (100 mg each) were reduced, alkylated and digested with trypsin overnight at a protein:enzyme ratio of 20:1 at 37°C. Each digest was concentrated to a volume of 15 μL in a speed vacuum followed by the addition of 15 μL of 1 M TEAB. The iTRAQ reagent was dissolved in 70 μL of ethanol and added to the digest, and the mixture was incubated at room temperature for 1 h. Cancer and normal samples labelled with different iTRAQ reagents were mixed and dried to a volume of 50 μL. The combined peptide mixture was fractionated by strong cation-exchange (SCX) chromatography on an UltiMate highperformance liquid chromatography (HPLC) system (LC Packings, The Netherlands) using a polysulfoethyl A column (2.1100 mm, 5 m, 300 Å [PolyLC, USA]). The sample was subsequently dissolved in 1 mL of SCX loading buffer (25% v/v acetonitrile, 10 mM KH 2 PO 4 [pH 2.8]), and pH was adjusted to 2.8 by adding 1 M phosphoric acid. The entire sample was loaded onto the column and washed isocratically for 30 min at a flow rate of 200 μL/min. Peptides were eluted with a linear gradient of 0 mM to 500 mM KCl (25% v/v acetonitrile, 10 mM KH 2 PO4, pH 2.8) over 30 min at a flow rate of 200 μL/min. The absorbance at 214 nm was monitored, and 15 fractions were collected along the gradient.

lC/MS analysis
Each SCX fraction was dried down, dissolved in 0.1% formic acid and analyzed by MS (Qstar Pulsar [Applied Biosystems-MDS Sciex, USA) interfaced with an Agilent 1100 HPLC system. Peptides were separated on a reverse-phase column packed with 10 cm of C18 beads (36,075 m, 5 m, 120 Å, YMC ODS-AQ [Waters Associates, USA]) with an emitter tip (New Objective, USA) attached. The peptides retained were washed using a gradient ranging from 5% to 40% solvent B (98% acetonitrile, 0.1% formic acid) in solvent A (0.1% formic acid) for 60 min at a flow rate of 300 nL/min. Survey scans were acquired from a spectra of 400 m/z to 1200 m/z, with up to three precursors selected for MS/MS using a dynamic exclusion of 45 s. Rolling collision energy was used to promote fragmentation, with the collision energy range 20% higher than that used for unlabelled peptides because of the presence of iTRAQ tags.

Data analysis
The MS/MS spectra data were extracted and searched for in the Uniprot-Sprot database (version 4, total number of entries: 230,093, entries for Homo sapiens: 14,515) using ProteinPilot version 3.0 software (revision 114732 [Applied Biosystems, USA]) with the Paragon method using the following search parameters: H sapiens as species, trypsin as enzyme (one missed cleavage allowed), cysteine static modification with methylmethanethiosulfate and iTRAQ (peptide labelled at the N terminus and lysine) as sample type. Mass tolerance was set to 0.15 atomic mass units for the precursor, and 0.1 atomic mass units for the fragment ions. The raw peptide identification results from the Paragon algorithm (Applied Biosystems, USA) searches were further processed by the Pro Group algorithm (Applied Biosystems, USA) within the ProteinPilot software before final display. The Pro Group algorithm uses the peptide identification results to determine the minimal set of proteins that can be reported for a given protein confidence threshold. For each protein, Pro Group algorithm reports two types of scores: unused ProtScore and total ProtScore. The total ProtScore is a measurement of all the peptide evidence for a protein, and is analogous to protein scores reported by other protein identification software packages. The unused ProtScore, however, is a measurement of all peptide evidence for a protein that is not better explained by a higher ranking protein. In other words, the unused ProtScore is calculated by using the unique peptides (peptides that are not used by the higher ranking protein) and is a true indicator of the presence of protein. This is how a single-protein member of a multiprotein family was isolated. The protein confidence threshold cut-off for the present study was ProtScore 2.0 (unused) with at least one peptide with 99% confidence. The mean, SD and P values to estimate statistical significance of the protein changes were calculated using Pro Group software. Differentially expressed proteins (ratio >1.2 or <0.8) that were consistent between two independent biological experiments were manually validated and quantified. Peak areas for each of the signature ions (113 and 114) were obtained and corrected according to the manufacturer's instructions to account for isotopic overlap. Only signature ions with intensities <1500 counts were used for quantitation. When intesities of the signature ions were >1500 counts, 1:1 ratios were generated due to detector saturation.

Western blot analysis
To verify the iTRAQ data, samples (1 μg) were separated by 12% sodium dodecyl polyacryamide gel electrophoresis and transferred to polyvinylidene fluoride membranes. To evaluate the percentage of gelsolin-positive cells, the entire section was first scanned to determine the overall distribution of gelsolin expression within the tumour and its degree of heterogeneity; then, the percentage of gelsolin-positive cells was determined in representative areas by examining 10 randomly selected microscopic fields (magnification ×400) of each tissue section. Observations were made by two different observers who were blinded to clinical outcome (ie, the slides were coded, and the pathologists did not have previous knowledge of the local tumour burden, lymphonodular spread and grading of the tissue samples while scoring the immunoreactivity). Sections were scored positive if epithelial cells showed immunopositivity in the cytoplasm, plasma membrane and/or nucleus when judged independently by two scorers. A grading system was used to express the proportion of positive cells in each case as follows: grade (G) 0, negative; G1, <25% positive cells; G2, 25% to 50% positive cells; G3, 50% to 75% positive cells; and G4, >75% positive cells per lesion. In each type of lesion, results were expressed as the percentage of cases of each grade.

Statistical analysis
ProteinPilot software using the Paragon method was used to analyze MS/MS spectra. Results of IHC were analyzed using the c 2 test (continuity correction) or t test. Differences between two means with P<0.05 were considered to be statistically significant.

Quantitative proteomics analysis of CRC and normal tissue
In the present study, iTRAQ-based stable isotope labelling of colorectal tissues was performed to identify dysregulated proteins in CRC. The experimental strategy is shown in Figure 1.
After labelling with iTRAQ reagents (113 for cancer samples and 114 for normal samples), the two samples were mixed and analyzed by LC-MS/MS. The MS/MS fragmentation of the iTRAQ-labelled peptides resulted in signature peaks (at 113.1 and 114.1) for quantitation, while the fragmentation along the peptide backbone resulted in band y-type fragments, which were used to identify the peptide sequence. Only proteins with at least two peptides per protein and a high annotation confidence (≥95% ) were considered, resulting in a total of 802 nonredundant proteins, which included structural proteins, signalling components, enzymes, receptors, transcription factors and chaperones. Eighty-two altered proteins outside the expression range of 0.8 to 1.2 were considered to be potential CRCspecific proteins (Appendix). Among these 82 proteins, gelsolin was specifically identified as a potential biomarker given the following: gelsolin, one of the major actin-binding proteins, is involved in the regulation of actin cytoskeleton organization via its severing and capping activity of actin filaments (15); and loss of gelsolin, a tumour suppressor,

Figure 1) Mass spectrometry (MS)-based approach (isobaric tags for relative and absolute quantitation [iTRAQ]
) to quantitative analysis of colorectal cancer (CRC). Proteins were extracted from CRC tissue and paired normal mucosa, digested with trypsin, labelled with iTRAQ reagent and combined. The resulting peptides were fractioned using strong cation exchange chromatography and analyzed by liquid chromatography (LC)-tandem MS-MS is one of the most frequently occurring molecular defects in ovarian, breast, colon and pancreatic cancers (16)(17)(18)(19). Gelsolin peptide sequences, peptide confidence, peptide ratios, protein ratios and P values calculated by ProteinPilot are presented in Table 2.
According to the protein sequence coverage map, protein sequences with matching peptides could be divided into groups with the highest confidence, moderate confidence, lowest confidence and the portions of the protein sequence for which no peptides were identified. The MS/MS spectra of two matching peptides (AGALNSNDAFVLK and EPAHLMSLFGGKPMIIYK) with the highest confidence are shown in Figure 2, with the peaks for the signature ions shown in the insets.

Western blot analysis
Western blot experiments were performed to validate the iTRAQ assay results. Based on a literature search, gelsolin was chosen. Gelsolin was detected by Western blot from the pooled sample used in a 2D LC-MS/MS experiment. Gelsolin is considered to be downregulated in CRC based on gray scale analysis of the protein strap, which was consistent with the discovery experiment ( Figure 3).  Figure 4C). Gelsolin achieved a sensitivity of 86.89% and specificity of 100% in distinguishing colorectal adenocarcinoma from normal tissue.
In cancer samples (reduced expression in eight cases) with clusters of gelsolin-positive cells, the mean (±SD) percentage of gelsolin-positive tumour cells was 21.18±7.73%. The percentage of gelsolin-positive cells was reduced in cancer compared with paired normal mucosa (P=0.034 [paired sample t test]) ( Figure 4C and 4D). No adenocarcinoma was scored as G4 (>75% of positive cells) or G3 (50% to 75% of positive cells), whereas 37.5% of cases (three of eight) were classified as G2 (25% to 50% of positive cells) and 62.5% of the cases (five of eight) were classified as G1 (<25% of positive cells). Gelsolin-positive cells showed cytoplasmic expression in adenocarcinoma (three of eight samples [37.5%]); mucinous adenocarcinoma (three of eight samples [37.5%]); and signet-ring cell carcinoma (two of eight samples [25%]). Compared with tumours devoid of expression, gelsolin expression tended to be reduced in mucinous adenocarcinoma and signet-ring cell carcinoma (P<0.0001). This decrease in gelsolin expression in CRC was not correlated with age, sex, histological differentiation, TNM stage or tumour localization (Table 1).

DiSCuSSioN
Compared with traditional proteomic methods, such as 2D gel analysis followed by MS/MS identification, the iTRAQ-coupled 2D-LC-MS/MS

Figure 4) Validation of gelsolin using immunohistochemical staining. A Cancerous sample showing a complete absence of detection. b Paired normal mucosa of sample A showing an expression of grade 4. C Cancerous sample showing decreased expression from grade 3 to grade 1 compared with sample D (paired normal mucosa of sample C). All images original magnification ×100
approach provides higher detection sensitivity and holds the promise of effectively depicting cellular protein profiles in tissue (9,11,20). In recent years, this approach has been used in prostate, radiotherapyresistant, renal cell, lung, oral, head and neck, breast and hepatocellular cancers (4)(5)(6)(7)(8)(9)(10)(11).
In the present study, we applied isobaric peptide tags (ie, iTRAQ) and multidimensional LC-MS/MS to identify proteins that are differentially expressed in CRC and paired normal mucosa samples. Differential gelsolin expression was validated by Western blot analysis and IHC. Because adenocarcinoma comprises more than 50% of CRCs, only colorectal adenocarcinoma was included in the present study.
Using a combination of subfractionation, iTRAQ-based labelling, 2D-LC peptide separation, MS and database searches for human proteins, 82 proteins altered outside the range of 0.8 to 1.2 were considered as potential CRC-specific proteins. Downregulation of gelsolin was confirmed by Western blot analysis and IHC. These serial studies suggest that combined proteomics and bioinformatics analysis can generate valid candidates that may be further evaluated for their role in the progression from normal tissue to adenocarcinoma. Although the validation of a protein cannot be extrapolated to the full protein list, it did suggest that the iTRAQ results were reliable.
Abnormal expression of gelsolin has been reported in many types of tumours. The expression of gelsolin has been reported to be frequently silenced in various cancers (16,18,(21)(22)(23)(24)(25)(26)(27). Conversely, overexpression of gelsolin has been found to be a negative prognostic predictor in a subpopulation of patients with nonsmall cell lung cancer, urothelial cancer, and EGFR + /erB-2 + breast cancer (25,28,29). Gelsolin regulates the architecture and dynamics of cells by capping, severing and nucleating actin filaments. However, the precise molecular mechanism behind the reduction of gelsolin expression has not yet been clarified (22,30). Despite the high incidence of CRC, only a few contradictory studies regarding the expression of gelsolin in this type of malignancy and its precursor lesions are available. One IHC study (31) reported weak gelsolin staining in 30 normal human colon mucosa and 22 adenocarcinoma samples. In contrast, another report based on 69 primary human colon adenocarcinomas and their paired normal mucosa (18) showed decreased gelsolin expression in all tumours analyzed, resulting in either a complete absence of detection in 34.8% of cases, or in focal staining of clusters of tumour cells in 65.2% of cases. This latter result was supported by DNA microarray studies that reported reduced expression of gelsolin complementary DNA in colon cancer compared with normal colon tissue (32,33). Furthermore, gelsolin expression and its subcellular distribution seem to correlate with metastatic potential in human colon adenocarcinoma cells (15). Using 2D-LC-MS/MS, gelsolin expression was found to be decreased in CRC in the present study. This result was further validated by Western blot analysis and IHC, with gelsolin expression tested using IHC in 61 primary human CRC and paired normal mucosa samples. Gelsolin expression was detected in all normal colon and rectum samples examined. Compared with normal mucosa, gelsolin expression was decreased in all tumours analyzed (61 of 61 [100%]), resulting either in a complete absence of detection in 86.89% (53 of 61) of samples or in focal staining of clusters of tumour cells in 13.11% (eight of 61) of samples. The rate of complete absence of detection surpasses the rate reported by Gay et al (18). In cases of reduced gelsolin expression, 62.5% showed mucinous adenocarcinoma and signet-ring cell carcinoma, while 37.5% of cases showed adenocarcinoma. This difference was statistically significant (P<0.0001), which suggests that the role of gelsolin in CRC may correlate with histological classification.

CoNCluSioN
The results presented reinforce the idea that iTRAQ and 2D-LC-MS/ MS represent an effective and reliable proteomic process in cancer biomarker discovery; the downregulation of gelsolin expression is a general mechanism during the transition from normal tissue to cancer; and that gelsolin may be a potential biomarker for colorectal adenocarcinoma in the Chinese population.