Renee Ruhaak1; Carol Stroble1, 2; Qiuting Hong1; Suzanne Miyamoto2; Kyoungmi Kim1; Gary Leiserowitz2; Carlito Lebrilla1
1UC Davis, Davis, CA; 2UC Davis Medical Center, Sacramento, CA
A QQQ-MS approach is presented for the discovery and validation of serum protein and glycopeptide levels as discriminators of disease.
Protein glycosylation has been proposed as a new source of potential biomarkers for diseases as diverse as cancer and infection. We have previously reported on the use of serum glycans for the diagnosis of ovarian cancer (OC). However, these studies have focused on released glycans thereby eliminating protein-specific information. Shotgun glycoproteomics analyses, which would provide protein-specific information, have thus far not been widely successful due to the large heterogeneity in protein glycosylation. Instead, we have focused on protein-specific glycosylation using a targeted approach. Multiple reaction monitoring on QQQ-MS is an excellent tool for protein- and site- specific quantification of protein glycosylation. We developed transitions for many abundant serum proteins, and apply the method towards glycopeptide biomarker discovery for ovarian cancer.
Protein standards were used to develop MRM transitions. Standards of the 9 proteins immunogobulin G (IgG), immunoglobulin A (IgA), immunoglobulin M (IgM), haptoglobin (HP), transferrin (TF), alpha-1-acid glycoprotein (AGP), alpha-1- antitrypsin (A1AT), alpha-2-macroglobulin (A2MG) and complement C3 (C3) were treated with DTT and IAA prior to digestion using trypsin as well as trypsin/chymotrypsin. First, their site-specific glycosylation patterns were determined using nLC-chip-Q-TOF mass spectrometry. Subsequently, quantitative MRM transitions were developed for peptides and glycopeptides using UPLC-QQQ-MS. The absolute protein concentration was determined using the peptide signals. Glycopeptide signals were normalized to the absolute protein amount to determine the protein- and site- specific glycosylation pattern. The method was applied to serum samples of ovarian cancer patients and their controls.
For each of the proteins, site-specific glycosylation patterns were determined using literature search and Q-TOF fragmentation data. Based on these data, transitions were developed corresponding to glycopeptides as well as nonglycosylated peptides, where the latter allows for protein quantification. The proteins IgG, IgA, IgM, HP, TF, AGP, A1AT, A2MG and C3 were targeted thus far. For each of the glycopeptides, transitions were developed to fragments with m/z 204 (HexNAc) or m/z 366 (Hex-HexNAc), except for high-mannose glycopeptides where transitions were developed to the peptide+HexNAc fragment. The method is highly accurate with RSD’s reported for the glycopeptides of less than 10%.
The method was applied to a sample set of 40 ovarian cancer cases and their age-matched controls for biomarker discovery. Levels of 18 glycopeptides of IgG, 13 glycopeptides of IgA and 10 glycopeptides of IgM were shown to alter significantly (either over- or under-expressed) with OC at a false discovery rate of <0.05. Multiplex classifiers combining multiple glycopeptides together were developed for each of the proteins with the highest accuracy for determining ovarian cancer of 91.1% when combining five glycopeptides of IgG. For IgA, multiplexing resulted in a highest accuracy of 88%, while the highest accuracy was 80% for IgM.
The method was then applied to an independent test set for validation. When testing the multiplex classifiers developed using the training set in samples of the independent test set, classification accuracies of cancer state of 90%, 80% and 74% were achieved for IgG, IgA and IgM, respectively. We are currently developing a model combining the glycopeptides from the different proteins that could allow the definition of a targeted QQQ-based OC diagnostic test with greater accuracy. This is the first attempt at quantitation of the glycosylation of several proteins simultaneously and shows the potential of targeted glycopeptide analysis as a diagnostic tool.