Automated sample preparation platform for mass spectrometry-based plasma proteomics and biomarker discovery

Abstract

The identification of novel biomarkers from human plasma remains a critical need in order to develop and monitor drug therapies for nearly all disease areas. The discovery of novel plasma biomarkers is, however, significantly hampered by the complexity and dynamic range of proteins within plasma, as well as the inherent variability in composition from patient to patient. In addition, it is widely accepted that most soluble plasma biomarkers for diseases such as cancer will be represented by tissue leakage products, circulating in plasma at low levels. It is therefore necessary to find approaches with the prerequisite level of sensitivity in such a complex biological matrix. Strategies for fractionating the plasma proteome have been suggested, but improvements in sensitivity are often negated by the resultant process variability. Here we describe an approach using multidimensional chromatography and on-line protein derivatization, which allows for higher sensitivity, whilst minimizing the process variability. In order to evaluate this automated process fully, we demonstrate three levels of processing and compare sensitivity, throughput and reproducibility. We demonstrate that high sensitivity analysis of the human plasma proteome is possible down to the low ng/mL or even high pg/mL level with a high degree of technical reproducibility.

Figure 1

An automated setup for the multidimensional LC system, allowing an on-line integration of depletion, protein derivatization and fractionation. Connection of depletion and RP columns into a single separation unit, through common LC pumps (A, B, C, D), capillary linkages and two 6-port valves (X, Y). Pumps A and B are delivering buffers for loading the proteins on-trap, and, if deployed, for a depletion procedure. The pumps C and D provide for acetonitrile gradient in RP protein fractionation. The reagents for derivatization (dithiothreitol, iodoacetamide) and desalting (water) are injected through the valve Y.

Figure 2

Distribution of CVs by sample group. Distribution of CV% for peak areas and their medians from peptide maps. Automated workflows: B (no depletion), C (95% depletion) and D (95% depletion and fractionation, fraction where the peptide is most abundant is shown). A: Benchmarking with an arbitrary plasma sample representing precision of the LC-MS measurement itself.

Figure 3

The identification power of three workflows on human plasma illustrated as a Venn diagram to demonstrate the overlap between the methods. Each Venn diagram shows the number of unique proteins detected for undepleted whole plasma, plasma depleted by immunoaffinity and plasma depleted by immunoaffinity and subsequent reverse phase separation. In diagram (A) (on the left) 10 µL whole plasma was analysed. Where immunodepletion was performed the most abundant 14 proteins were removed and where RP HPLC was performed 10 fractions were collected. In diagram (B) (on the right) the starting material was increased to 250 µL plasma. This then enabled higher fractionation. Where immunodepletion was performed the most abundant 14 proteins were removed and a further immunodepletion (Supermix) was used to remove up to 99% of the total protein mass. Where RP HPLC was used 20 fractions were collected. The left hand panel (A) therefore represents the options for the current automation, whereas (B) represents the potential for the extreme of fractionation.