We used the quantitative proteomic technique iTRAQ® to analyze pooled, immunodepleted serum from ovarian cancer patients and non-cancer controls for the discovery of proteins that are present at increased levels in ovarian cancer serum and may be candidates for biomarkers. For our biomarker discovery experiments, we selected patients with advanced disease under the assumption that these patients may have higher levels of ovarian cancer proteins in their sera than patients with early stage disease, thus increasing the likelihood of detecting ovarian cancer specific biomarkers. The serum samples were pooled in order to reduce variation between individual samples and to increase the overall number of samples tested. The technique of pooling samples prior to proteomic analysis has been performed by others in similar studies with body fluids [12–14].
Relative expression levels of 14 proteins in serum from ovarian cancer patients were 1.6-fold or more higher than the corresponding serum protein levels of healthy women in at least one experiment. Nine proteins were identified in all five iTRAQ® experiments, two proteins were identified in four experiments, two proteins were identified in three experiments, and one protein was identified in two experiments.
Several of the proteins identified in this study have previously been identified as potential ovarian cancer biomarkers [15–18], demonstrating that iTRAQ® labeling of serum can be used for the identification of differentially expressed serum protein biomarkers. For example, haptoglobin was first shown to be elevated in the sera of women with ovarian cancer almost 40 years ago . Others have confirmed this finding numerous times, most recently by 2DE , and also with SELDI and ELISA . In addition to the iTRAQ® data presented here, we have also shown by DIGE analysis that several haptoglobin subunits are present at elevated levels in ovarian cancer sera . Serum amyloid A1 [20, 21], and C-reactive protein [22, 23] have also previously been shown to be elevated in ovarian cancer; in our analysis of iTRAQ® labeled sera these proteins had average C/N ratios of 8.9 and 6.5 respectively, although the N/N peak height was below the threshold for quantitation and not reported by the ProteinPilot™ software. Consequently, these proteins did not meet our criteria for inclusion in Figure 1; however the relative expression, statistical parameters for quantitation, and the peptide information for these proteins are in Additional file 1, Table S1. C-reactive protein was identified by a single peptide in sera depleted by the MARS column; the spectrum for this peptide is shown in Additional file 3, Figure S1.
Notably, we identified several candidate biomarker proteins which have not been extensively studied. For example, apolipoprotein-E (ApoE) was identified by multiple peptides in all five iTRAQ® experiments, and was ~1.6-fold elevated in the ovarian cancer serum pools. Although commonly known for its role in lipid transport, ApoE has previously been shown to be overexpressed in ovarian cancer cells [24, 25] where it is required for proliferation and survival .
The protein LRG1 was determined to have robustly increased levels of expression in ovarian cancer sera by iTRAQ®. We have previously shown an increase in several LRG1 isoforms in ovarian cancer serum by DIGE . LRG1 expression in ovarian cancer sera was validated by Western blot in the immunodepleted serum pools, and by ELISA in individual, undepleted serum samples (data not shown). Although LRG1 has been classified as an "acute phase" protein, produced by the liver in response to injury or infection [26, 27], we have evidence that LRG1 is also being produced by ovarian cancer cells and may contribute to the increased levels of LRG1 found in patient sera (manuscript in preparation). LRG1 levels have also been shown to be elevated in the medium and low abundance serum proteins of lung and pancreatic cancers [28–30].
ECM1 was identified in two iTRAQ® experiments, with an ovarian cancer-to-control ratio of ~1.6. The increase in expression of ECM1 was validated by Western blot in all of the ovarian cancer serum pools. ECM1 is a secreted protein, which has been shown to be overexpressed in a number of different epithelial cell tumors . Interestingly, ECM1 was identified in two proteomic analyses of ovarian cancer ascites cells [32, 33], and was also highly enriched in the conditioned media from ovarian cancer cell lines  and in chemoresistant ovarian tumor tissue .
PRG4 was identified by iTRAQ®, with a relative protein abundance in ovarian cancer-to-control of ~2.1. PRG4, also known as megakaryocyte stimulating factor and hemangiopoietin, is a multifunctional proteoglycan with growth promoting properties in hemangioblasts . It is synthesized by chondrocytes and plays a role in joint lubrication , however it is also synthesized by the liver, heart, lung, and bone .
LBP1 was identified with an iTRAQ® ratio of 1.95 in ovarian cancer vs. control sera, which was validated by Western blot. LBP1 is proposed to play a role in immune activation , however, the identification of LBP1 in ovarian cancer ascites cells using MS [32, 33] also supports a potential role in ovarian cancer.
We have previously reported the identification of candidate serum biomarkers for ovarian cancer by DIGE analysis of pooled, depleted serum . Four proteins were shown to have increased levels in ovarian cancer serum by both proteomic methods: haptoglobin, alpha-1 antichymotrypsin, serum amyloid P-component, and LRG1. Five other proteins were found at increased levels in ovarian cancer serum by DIGE, while 12 additional proteins were discovered by iTRAQ®, demonstrating that these two techniques for identification of candidate biomarkers are complementary.
Our study is the first to report immunoaffinity depletion in combination with iTRAQ® labeling as a means to quantitate serum proteins in ovarian cancer. Gagne et al.  used the quantitative proteomic techniques of both iTRAQ® and 2DE to compare two ovarian cancer cell lines, one highly proliferative and the other with low malignant potential, to identify proteins associated with invasion and metastasis. Similarly, proteins associated with prostate cancer progression were identified using iTRAQ® labeling and comparison of highly and poorly metastatic prostate cancer cell lines . Other iTRAQ® studies have compared tumor to normal tissue for biomarker discovery (e.g. endometrial, head and neck cancers [40–42], and hepatocellular carcinoma ). iTRAQ® analysis of biofluids such as saliva  and cerebral spinal fluid  has also been reported. Interestingly, a recent quantitative proteomic analysis of ovarian tumor tissue using ICAT labeling examined the relative expression of tumor proteins in chemoresistant vs. chemosensitive tumors, and found that five of the proteins identified in this study [ECM1, LRG1, orosomucoid 1 (alpha-1-acid glycoprotein) and alpha-1-antitrypsin] were all overexpressed in chemoresistant tumors , strengthening their potential use as biomarkers.
In an iTRAQ® analysis of serum proteins in patients following brain injury, Hergenroeder et al.  identified 160 proteins (≥ 95% confidence) in sera immunodepleted by the IgY12 column. Fifteen of the proteins identified were increased in abundance compared to controls. The relatively small number of serum proteins that they found to be upregulated by iTRAQ® parallels the numbers seen in our study. The proteins that were elevated in serum from patients with traumatic brain injury included many of the same serum proteins we found elevated in ovarian cancer (e.g. haptoglobin, orosomucoids-1 and -2, alpha-1 antitrypsin, serum amyloid P-component, and alpha-1 antichymotrypsin), which suggests that these proteins are not specific to ovarian cancer . However, several of these "acute phase" proteins have recently been shown to discriminate between malignant and benign gynecological disease, and thus may have a role in a "multi-analyte" panel of ovarian cancer biomarkers . In general, the results indicate that the continued presence of highly abundant serum proteins post-depletion, as well as the presence of moderately abundant serum proteins, may interfere with the detection of low abundance proteins that are more specific to the disease.
Recently, Lin et al.  reported the identification of serum biomarkers for ovarian cancer using depletion of both high and moderately abundant proteins in combination with LC-MS/MS and ICAT® labeling. Using their "deep depletion" method, they were able to identify 222 proteins, about three times as many as they found with depletion of only the most abundant proteins. Similar to our study, they found LRG1 levels were increased in ovarian cancer serum. In contrast, they reported elevated levels of a number of proteins (e.g., hemopexin, histidine-rich glycoprotein, and vitronectin) that were identified in our study but did not meet our criteria for overexpression in ovarian cancer serum. This is potentially due to the small number of samples (two pools of ten individuals in each group) used in their study.