Abstract:
Serological biomarkers capable of detecting early stage cancers or the presence of micrometastases have great potential to improve early diagnosis and clinical management of cancers after initial detection. However, direct detection of novel, specific cancer biomarkers in patient populations has been frustrated by the great complexity of plasma, high variability of plasma protein levels in human populations, and the difficulty of detecting very low abundance proteins. Due to the small size of early tumors and micrometastases, highly specific biomarkers shed by the tumors are expected to be very low abundance in cancer patient plasma and even lower abundance in subjects without tumors. Use of mouse cancer models is the most effective solution to the dual problems of human plasma proteome variability and need for great depth of analysis. We developed a 4-D plasma proteome profiling method that is very effective in detecting many low abundance plasma proteins. The 4-D profiling method consists of: immunoaffinity depletion of abundant proteins, fractionation of depleted plasma using MicroSol IEF, 1-D SDS gel separation of MicroSol IEF fractions followed by cutting each lane into uniform segments and digesting each slice with trypsin, and analysis of trypsin digests using nanocapillary reverse phase chromatography interfaced to a high sensitivity mass spectrometer for LC-MS/MS analysis. This method can identify many candidate biomarkers in xenograft mouse cancer models by identifying human proteins in a sea of mouse plasma proteins. This strategy removes the need for quantitative comparisons that are often complicated by protein level fluctuations caused by non-tumorigenic events or non-specific acute phase reaction changes. Analyses of these proteomes can identify up to 200 human proteins, and some of these proteins are expected to be specific cancer biomarkers. Two related key challenges are to acquire MS/MS spectra on very low abundance proteins by optimizing the nanoLC separation and data acquisition during LC-MS/MS analysis on hybrid ion trap mass spectrometers, and to use MS/MS data analysis strategies that identify low abundance "1-hit wonder" proteins with high confidence. These strategies utilize a comprehensive species-tagged database of human and mouse proteins with an appended decoy reverse database, initial database search with low stringency constraints, and data filters that yield false positive rates of less than 1% using high precursor mass accuracy as the primary filter constraint. The next step is to verify putative human proteins and to distinguish between protein isoforms and homologs using targeted LC-MS/MS analysis on a hybrid ion trap mass spectrometer.
A second model system and analysis approach is to analyze plasma from transgenic mouse models during tumor progression utilizing a 3-D analysis method coupled with label-free quantitation of ion current signals in LC-MS chromatograms. The label-free quantitation software must be capable of combining related signals from multiple LC-MS runs because a single LC-MS run per proteome does not have sufficient depth of analysis.
The next step is to validate the most promising candidate biomarkers in human plasma samples by establishing appropriate medium throughput quantitative assays that are capable of analyzing at least 100 specimens in a reasonable timeframe. So far, the most promising approach appears to be "Top 20" protein depletion followed by Gel-LC-MS/MS using label-free quantitation of ion currents in the LC-MS chromatograms. As with the second mouse model, the software must be capable of combining signals across multiple LC-MS runs per proteome.