Quantitative proteomics and biomarker discovery in human cancer

Early diagnosis and prevention is crucial for reducing the mortality and morbidity of cancer. Current clinical examination, such as mammography and invasive needle or surgical evaluation for breast cancer, and chest x-ray for lung cancer, for example, is often not sufficiently sensitive for early diagnosis. In order to detect cancer in its early stages, it is necessary to identify biomarkers for asymptomatic patients who may have cancer. Biomarkers could be proteins, metabolites or electrolytes whose differential expression indicates the presence of disease [1]. Biomarkers of protein origin are powerful, both in monitoring the development of cancer and in determining the efficacy and safety of drugs. Novel biomarkers are urgently needed for the early diagnosis and treatment of cancer. Nowadays, the remarkable progress in proteomics technology has offered unprecedented opportunity for biomarker discovery.

Traditionally, 2DE has been the standard method for comparing protein expression between normal and disease-perturbed states. However, it is now known that 2DE lacks resolution and sensitivity. Only abundant proteins can be resolved on a 2D gel. Biomarkers, which are usually low in abundance, are rarely found with the 2DE method. Gel-free or mass spectrometry (MS)-based proteomics techniques are emerging as the choice for quantitatively measuring protein levels, due to better sensitivity and reproducibility over 2DE-based methods [2,3]. However, quantitative proteomic profiling of complex biological samples for the purposes of biomarker discovery remains a challenge in proteomics. Multiple innovative profiling techniques have been introduced with the aims of comprehensively identifying and quantifying proteins. These can be approximately divided into two categories: isotope-labeled and label-free MS.

Isotope-labeled mass spectrometry

Isotope-labeling methods have been developed that introduce stable isotope tags to proteins via chemical reactions using isotope-coded affinity tags (ICAT) [4,5] and isobaric tag for relative and absolute quantitation (iTRAQ) [6], enzymatic labeling, for example using ¹⁸O water for trypsin digestion [7], or via metabolic labeling (stable isotope labeling of amino acids in cell culture [SILAC]) [8]. The pioneering ICAT technology selectively targets peptides containing a specific amino acid (cysteine) with a stable isotope-coded internal reference or standard [9]. The extracted proteins from treatment and control samples are labeled with either light or heavy ICAT reagents by reacting with cysteinyl thiols on the proteins. Peptides containing the labeled and unlabeled ICAT tags are recovered by avidin affinity chromatography and are then analyzed by liquid chromatography (LC)-MS/MS. Differential protein expression is determined by the isotope peak ratio of the peptide. Enrichment of low-abundance proteins can be performed through cell lysate fractionation [10]. ICAT technology has been widely used for protein identification and quantification in mammalian, liver and breast tumor cells [11]. However, disadvantages of ICAT analyses are obvious: it is only applicable to proteins containing cysteine, it can only identify 300–400 proteins, far fewer than the 2DE method, and the peptides contain large labels, which makes database searching more difficult, especially for short peptides [10].

Isobaric tag for relative and absolute quantitation is another labeling technique first developed by Ross et al., which use special isobaric tags to label proteins extracted from samples for comparison [12]. The proteins from control and treatment samples have the same mass after reacting with the iTRAQ reagents. These peptides give four ion species of different masses upon collision-induced dissociation. These ions allow samples to be quantified in MS/MS mode [12]. The amine specificity of the labeling reagents makes most peptides amenable to this labeling strategy with no loss of information. This is especially important for proteins with post-translational modifications, such as phosphorylation and glycosylation. In addition, the multiplexing capacity of these reagents allows for comparison among multiple cellular states. However, as a chemical labeling method, iTRAQ may generate side products during labeling and cause some loss of analytic sensitivity because chemical strategies involve a derivatization step that might not be complete.

Absolute quantification (AQUA), uses synthesized isotopically labeled peptides that mimic native peptides as internal standards [13,14]. The method has the potential for high-throughput and multiplexed sample analysis [15]. SILAC, which takes a similar strategy but utilizes different media containing light or heavy forms of particular amino acids, has emerged as a popular label-based quantification technique [16]. It was first developed by Mann et al., based on metabolic incorporation of ‘light-’ or ‘heavy-’ form amino acids into the proteins in living cultured cells [8]. Usually, heavily labeled (¹³C or ¹⁵N) arginine, lysine or both are used in culture medium to ensure complete labeling of every trypsinized peptide fragment. In experiments, one cell population is fed with regular amino acids, while the other is fed with ¹³C- or ¹⁵N-labeled amino acids. After several rounds of cell division, heavy amino acids will be incorporated into newly synthesized proteins. In the MS spectrum, the light and heavy peptides will show up as two distinct peaks separated by the incremental mass of the labeled amino acids. By comparing the signal intensity, relative quantification can be achieved. Owing to its simplicity in principle, SILAC is widely used for biomarker discovery [17], cell signaling dynamics [18], identification of posttranslational modification sites [19,20], protein–protein interaction [21–23] and subcellular proteomics [24].

The dynamics of protein turnover is another key feature to the understanding of regulation of protein expression in cells [25,26]. Recently, we developed a general method for determination of protein synthesis rate using labeling of amino acids with deuterium or ¹⁵N at low enrichment [27,28]. This method, ‘modified SILAC’ (mSILAC), can measure protein synthesis rate quantitatively, based on analysis of mass isotopmer distribution (MIDA) of the newly synthesized protein [28]. Once precursor enrichment is known, protein synthesis is determined from isotopomer distribution. In experiments with 30–50% enriched ¹⁵N amino acids, incorporation of ¹⁵N amino acids result in sufficient mass shift in the new protein, the ¹⁵N enrichment can be estimated from the mass shift by curve fitting, and the expected isotopomer distribution of the new peptide can be generated by the concatenation function. Synthesis rate is then calculated by multiple linear regression analysis of the observed peptide spectrum on the expected new and the old (unlabeled) spectra [28]. This method obviates the need for the use of a 100% newly synthesized protein as a reference as in Vogt’s methods [29,30]. The concatenation function provides an ideal 100% labeled spectrum and multiple regression analysis uses all the information from the mass spectrum. Our mathematical algorithm represents a major improvement in the calculation of protein synthesis rate, permitting the use of isotope labeling of protein through the pathways of amino acid metabolism with low cost isotopes [27,28].

For small peptides, another method, multiple-reaction monitoring (MRM), attempts to monitor prespecified ions (and their daughter product ions after MS/MS) as specific signatures for a particular protein instead of aiming to measure everything in a sample. In this method, initial selective scanning for a particular precursor ion in the first MS is followed by scans of a particular transition. Only one of the ions produced during collision-induced dissociation is selected with the aid of prior knowledge of reliable precursors and transition state [31]. Due to limitations in mass spectrometer dynamic range and resolution of chromatography separation, MRM is frequently combined with immunodepletion, size-exclusion chromatography [32] or enrichment of stable isotope standards with capture by antipeptide antibody [33]. These hybrid methods are particularly attractive because they can precisely measure the quantity of low-abundance proteins, which are otherwise bypassed in other conventional methods. MRM is comparable to ELISA with regard to sensitivity but obviate the necessity of antibody, which is costly to obtain and sometimes unavailable. MRM is normally more accurate than ELISA because it does not have propensity of cross-reaction as antibodies have.

Label-free mass spectrometry

With the advances of new instrumentation, computing power and advanced bioinformatics, generic label-free LC-MS shotgun screening methods, such as MudPIT, have been alternatives for relative and absolute protein quantitation in biological samples. Without requiring modification of the analytes, the label-free approaches were based on the observation that MS intensity is linearly proportional to ion concentration, even in complex samples such as serum [34]. However, it was later observed that not all peptides are equally detectable because of competition between ions, dynamic-range limitation and sensitivity of MS instrumentation [35]. A new technique, known as spectral counting, circumvented these pitfalls by not looking at differences in ion peak areas or intensities, as used in isotope-based methods [35]. Nevertheless, the main bottleneck for label-free LC-MS still lies in computer softwares and sample preparation. Although many software tools are available, there is a large space for improvement, such as user interface and calculation efficiency of workflow. Moreover, the accuracy of label-free quantitation depends largely on experimental set-up, such as protein extraction and sample stability, which is illustrated by the fact that the quantitative reproducibility of technical replicates is much better than that of experimental replicates [36,37]. With improvements in the efficiency of data analysis workflow, label-free MS will potentially be widely used for biomarker discovery and validation.

Conclusion

The recent development of MS-based strategies for absolute protein quantification offers great opportunity for biomarker discovery. While the prospects of this technology are exciting and promising, the current methodology is far from perfect. First, most current methods require complicated sample preparation, such as immunosubtraction, multidimensional LC separation, immunoaffinity and solid-phase extraction, in order to enhance the analytical dynamic range and detection sensitivity. To establish high-throughput pipelines, we should ideally have a one-step preparation. Second, useful and validated biomarkers are still rare based on these developed methods because low-abundance biomarkers are always immersed in large quantities of routine proteins, especially in plasma samples. There is a large space for improvement of sensitivity.

Financial & competing interests disclosure

This work is jointly funded by the Bone Biology Program of the Cancer and Smoking Related Disease Research Program and the Nebraska Tobacco Settlement Biomedical Research Program (289104–845610 to GX), and partially supported by a grant awarded to Wai-Nang Paul Lee from the UCLA Center of Excellence in Pancreatic disease (P01 AT003960–01) and Harbor-UCLA GCRC Mass Spectrometry Core (M01 RR00425–33). The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.