Protein separation on gels and columns

A major analytical challenge of proteomics is the dynamic range of protein concentrations (e.g. estimated 10¹² in human blood)⁽Reference Jacobs, Adkins and Qian¹⁴⁾. Current MS-based proteomic platforms can deliver a dynamic range of 10⁴. This means that the low-abundant proteome has to be addressed by depletion of the most abundant proteins⁽Reference Gong, Li and Yang¹⁵⁾ or by selective enrichment of low-abundant proteins⁽Reference Mechref, Madera and Novotny^16–Reference Wollscheid, Bausch-Fluck and Henderson¹⁸⁾. After protein depletion and/or enrichment, further separation is performed at protein and/or peptide level, based on two-dimensional (2D) gels or on liquid chromatography (LC) or on hybrid approaches (Gel-LC). Fig. 1 summarises the classical proteomic workflows.

Fig. 1. (Colour online) Classical workflows in MS-based discovery proteomics. Both gel-based and liquid chromatography (LC) based approaches are represented, with the latter increasingly taking over the proteomics business. ESI, electrospray ionisation; MALDI, matrix-assisted laser desorption/ionisation.

The gel strategy for protein separation offers the advantage of visualisation of proteins, and, to some extent, of their modifications and, therefore, it preserves the protein context. After protein separation on gel, the gel spots are then excised, digested with trypsin and further processed in LC-tandem MS (MS/MS) or MALDI-MS instruments for protein identification. However, major drawbacks of this method are that (i) hydrophobic proteins such as membrane proteins, very basic proteins and low-abundant proteins are not well represented on the gels and (ii) the procedure is low throughput and difficult to standardise and automate.

The ‘gel-free’ discovery approach for protein identification is referred to as ‘shotgun’ analysis: here, the sample or protein mixture is directly digested in solution. The resulting peptide mixture is then separated on an HPLC usually coupled online to a mass spectrometer for peptide mass identification. The development of ultra-HPLC delivers improved peptide separation⁽Reference Eschelbach and Jorgenson¹⁹⁾ and column efficiency by means of higher mass sensitivity, analytical resolution and speed. The resolving capability of the HPLC separation can be enhanced with 2D HPLC techniques⁽Reference Washburn, Wolters and Yates²⁰⁾ that combine ion-exchange chromatography followed by reverse-phase HPLC.

Peptides can also be separated using gas-phase fractionation, which is defined as iterative mass spectrometric interrogations of a sample over multiple smaller mass-to-charge (m/z) ranges. By doing so, a higher number of unique peptides compared to the ions selected from the wide mass range scan in standard LC-MS/MS analysis can be addressed. Gas-phase fractionation is described as a means to achieve higher proteome coverage than classical LC-MS/MS analyses of a complex peptide mixture⁽Reference Blonder, Rodriguez-Galan and Lucas^21–Reference Yi, Marelli and Lee²³⁾.

Protein identification by MS

Mass spectrometers identify proteins and peptides by determination of their exact masses and by generating information on their amino acid sequences. Today the main ionisation methods deployed are electrospray ionisation⁽Reference Fenn²⁴⁾ and MALDI⁽Reference Tanaka²⁵⁾. These ion sources are combined with various mass analysers that separate the ions by m/z. The most popular analysers in proteomics are ion traps, triple-quadrupoles, time-of-flight tubes, orbitrap and Fourier-transform ion cyclotron resonance, with their specific advantages: high sensitivity and multiple-stage fragmentation for ion traps; high selectivity for triple-quadrupoles; high sensitivity and speed for time-of-flight. Current top-end proteomic machines are orbitraps⁽Reference Makarov, Denisov and Kholomeev²⁶⁾ and Fourier-transform ion cyclotron resonance instruments⁽Reference Nielsen, Savitski and Zubarev²⁷⁾, which provide very high mass accuracy and resolution compared to the other analysers. MS/MS consists of either the fragmentation of a selected precursor peptide ion to generate specific fragment ions for sequence elucidation (data-dependent acquisition); or uncoupled acquisitions of intact and fragment masses with retrospective reconstitution of the parent–daughter ion context (data-independent acquisition; see Emerging Technologies)⁽Reference Panchaud, Scherl and Shaffer²⁸⁾.

Considering the amount of data generated in a single shotgun run, search algorithms have been developed to process the raw data automatically and to compare the measured MS/MS spectra against theoretical fragment ion spectra generated by in silico digestion of protein databases. Comparison of experimental with theoretical spectra results in a list of possible peptide matches, each with an associated score that quantifies the quality of the match. The peptide(s) with the highest score(s) is(are) generally considered for protein identification. Several of these search engines have been commercialised (e.g. Sequest, Mascot or Phenyx), but others are freely available (open source programs, e.g. XTandem or OMSSA). They usually agree on 80% of the identifications⁽Reference Deutsch, Lam and Aebersold²⁹⁾. The two most established applications are Sequest⁽Reference Eng, McCormack and Yates³⁰⁾ and Mascot⁽Reference Perkins, Pappin and Creasy³¹⁾.

Identified peptides and proteins can be further validated by software that applies statistical algorithms to calculate additional scores and probabilities, thereby distinguishing correct from incorrect assignments. PeptideProphet and ProteinProphet from the TransProteomic Pipeline (ISB, Seattle)⁽Reference Keller, Nesvizhskii and Kolker^32, Reference Nesvizhskii, Keller and Kolker³³⁾ determine respectively at peptide and protein level the probabilities of correctness associated with false discovery rate. These tools provide the researcher with means to assess the quality of the data in a dataset-dependent manner and to control the trade-off between false positives (specificity) and false negatives (sensitivity)⁽Reference Urfer, Grzegorczyk and Jung³⁴⁾. The second strategy to elucidate the false-positive/false-negative relationship relies on a database search using a target-decoy database⁽Reference Elias and Gygi³⁵⁾, which is typically generated by reversing the sequences of target protein database. The search is done against the target and the decoy database and allows the estimation of false positives.

Protein quantification

Once qualitative analysis of a sample is achieved (first discovery mode, protein catalogue), very often quantitative information is required to obtain more insight into proteome differences between conditions or over time.

Relative quantification enables the comparison of two or more biological samples/conditions and the identification of (candidate) biomarkers, i.e. proteins that are more or less abundant (up- or down-regulated) in a certain condition compared to another.

The classical method for relative quantification evolved from 2D gels: 2D difference gel electrophoresis is characterised by differential labelling of proteins with fluorescent dyes prior to separation according to isoelectric point and molecular weight⁽Reference Unlu, Morgan and Minden³⁶⁾. An internal standard is used to match the protein patterns across gels thereby facilitating gel alignment, spot matching and quantification. Dedicated software (e.g. Progenesis SameSpots and DeCyder) allow for gel comparison and spot quantification.

Other procedures for relative quantification of proteins rely on metabolic or chemical labelling of proteins by incorporation of stable isotopes (usually ¹³C or ¹⁵N) in the samples to be compared and quantified at MS level. A summary of popular chemical labels used for relative protein quantification is shown on Fig. 2 that depicts a generic peptide with its side chains and modification options. Labelling of proteins and peptides is performed by in vitro chemical or enzymatic derivatisation and include for example: isotope-coded affinity tag⁽Reference Gygi, Rist and Gerber³⁷⁾, isotope tags for relative and absolute quantification⁽Reference Ross, Ambrose and Cutler³⁸⁾, isotope-coded protein label⁽Reference Schmidt, Kellermann and Lottspeich³⁹⁾ or aniline and benzoic acid labelling ⁽Reference Panchaud, Hansson and Affolter⁴⁰⁾. The differentially labelled samples are mixed and infused into an LC-MS/MS instrument. While these techniques involving labelling are accurate, they need specific sample preparation and involve added costs due to stable-isotope labelled reagents. In contrast to chemical labelling, stable isotope labelling by amino acids in cell culture⁽Reference Ong, Blagoev and Kratchmarova⁴¹⁾ consists in labelling of proteins already during cell growth and division by incorporation of labelled amino acids.

Fig. 2. (Colour online) Protein quantification based on chemical stable-isotope labelling. Labels targeting N- and C-termini are reported in light print, whereas labels targeting side chains of amino acids within a polypeptide are reported in dark print. Derivatisations occurring at the N-terminus also affect the ε-amino group of lysine. For each type of label, a reference is indicated⁽Reference Gygi, Rist and Gerber^37, Reference Panchaud, Hansson and Affolter^40, Reference Zhang, Li and Martin^42–Reference Zhang, Sioma and Thompson⁶³⁾. Adapted from Julka and Regnier⁽Reference Julka and Regnier⁴³⁾, authorisation no. 2597551000090.

Recently, methods for label-free relative quantification have been developed. In this case, each sample is separately analysed by LC-MS/MS and resulting data are processed using software designed for LC-MS/MS run alignment, extraction of peptide intensities and peptide counts (e.g. Progenesis LCMS, DeCyder MS, SuperHirn, SpecArray or MSQuant). Currently, two label-free quantification strategies can be used: (a) measuring and comparing the mass spectrometric signal intensity of peptide precursor ions belonging to a particular protein⁽Reference Chelius and Bondarenko^64, Reference Mueller, Rinner and Schmidt⁶⁵⁾; and (b) counting and comparing the number of fragment spectra identifying peptides of a given protein⁽Reference Liu, Sadygov and Yates⁶⁶⁾. In comparison with stable isotopes, label-free protein quantification is simpler to perform; there is technically no limit to the number of sample/conditions to be compared (except the analytical capacity of software/computers/servers); and it can yield an improved proteome coverage with a broader dynamic range⁽Reference Bantscheff, Schirle and Sweetman⁶⁷⁾. However, label-free proteome comparisons are limited to not too complex proteomes, otherwise the peptide-to-peptide alignment becomes too difficult due to LC elution and m/z overlaps.

Absolute quantification of proteins relies on the addition of an internal standard at a known concentration. The Absolute QUAntification method⁽Reference Gerber, Rush and Stemman⁶⁸⁾ uses synthetic stable-isotope labelled proteotypic peptides as internal standards that are otherwise identical to the peptides to be quantified. This approach requires preliminary analyses to select the peptides to be used to quantify the protein(s) of interest (proteotypic peptides). Recently, the QconCAT technology has been developed for parallel production of labelled proteotypic peptides which are then used in multiplexed quantification assays⁽Reference Rivers, Simpson and Robertson⁶⁹⁾. QconCAT consists of an artificial gene, inserted into a vector for expression in Escherischia coli, which is designed to express artificial proteins comprising a concatenation of proteotypic peptides. This latter technology is highly useful for repetitive, multiplexed analysis, i.e. a protein assay-like situation.

The systematic large-scale approach for absolute protein quantification has been referred to as selected-reaction monitoring or multiple-reaction monitoring⁽Reference Lange, Picotti and Domon⁷⁰⁾. This type of protein quantification is based on the quantification of relevant proteotypic peptides and is exclusively performed with triple-quadrupole mass spectrometers. The masses of the peptides and of their most abundant fragments are defined in the method and the mass spectrometer only scans for these as well as for the corresponding stable-isotope labelled proteotypic peptides. A chromatographic peak proportional to the peptide amount appears only if both parent and fragment masses are present (referred to as one transition). These peaks can then be integrated and the peptide and protein concentrations can be calculated by comparison with the internal peptide standard. This method enables the targeted, multiplexed, high-throughput quantification of low-abundance proteins in highly complex mixtures.

Analysis of post-translational modifications: protein functionality

PTM of amino acids give functionality to the proteins through the attachment of functional groups such as phosphate, carbohydrates, acetate or lipids. PTM play crucial roles in regulating the biology of the cell since they can change a protein's physical or chemical property, activity, localisation or stability. Some PTM can be added and removed dynamically as a mechanism for reversibly controlling protein function and cell signalling. Several proteomic techniques have been developed to identify and quantify PTM and allow the study of modifications such as phosphorylation, acetylation, glycosylation or lipid modifications.

Phosphorylation of proteins controls many cellular processes such as growth, differentiation, metabolism, signalling and cell death, and is itself regulated by enzymatic activity (i.e. by kinases and phosphatases). The challenges of phosphoproteomics lie, as for proteomics, in the complexity, dynamic range and temporal dynamics of protein isoforms⁽Reference Nita-Lazar, Saito-Benz and White⁷¹⁾. Several methods have been developed to enrich phosphoproteins and include anti-phosphotyrosine antibodies⁽Reference Pandey, Andersen and Mann⁷²⁾; immobilised metal affinity chromatography⁽Reference Andersson and Porath⁷³⁾; and chemical modification and strong exchange chromatography⁽Reference Beausoleil, Jedrychowski and Schwartz⁷⁴⁾. The analysis of phosphorylated proteins has been facilitated by the development of new fragmentation techniques such as electron capture dissociation⁽Reference Zubarev, Horn and Fridriksson⁷⁵⁾ and electron transfer dissociation⁽Reference Syka, Coon and Schroeder^76, Reference Wiesner, Premsler and Sickmann⁷⁷⁾ that help identify and determine the location of phosphorylations that cannot be as efficiently characterised by standard collision-induced dissociation.

Protein glycosylation is prevalent in proteins that are involved in mechanisms like cell–cell interactions, immune system (e.g. antibodies, MHC) or transport (e.g. transferrin). Also, glycoproteins account for a major proportion of milk and human blood proteomes for which it has been estimated that 70 and 50% of all proteins are glycosylated, respectively. There are two types of glycoproteins: (a) N-glycosylated proteins with carbohydrates linked to the side chain of the asparagine; and (b) O-glycosylated proteins, in which carbohydrates are bound to the side chain of serine, threonine, hydroxylysine or hydroxyproline. The functions of glycoproteins are still incompletely understood⁽Reference Funakoshi and Suzuki⁷⁸⁾. MS can provide information on molecular mass, composition, sequence and sometimes branching of a glycan chain⁽Reference Dell and Morris^79, Reference Tissot, North and Ceroni⁸⁰⁾. As glycoprotein forms are often minor constituents compared to the non-glycosylated proteins, enrichment methods such as cell surface-capture technology⁽Reference Zhang, Li and Martin⁴²⁾ or affinity capture with lectins⁽Reference Yang and Hancock⁸¹⁾ have been set-up.

Acetylation is a PTM that has been associated with several biological processes, especially gene expression regulation by histones, which pack the chromosomes so that they fit into the cell nucleus⁽Reference Allfrey, Faulkner and Mirsky⁸²⁾. Acetylation usually occurs at lysines or on N-terminal groups of peptides or proteins and is involved in the destabilisation of chromatin and recruitment of effector proteins⁽Reference Grant^83–Reference Strahl and Allis⁸⁶⁾. Especially the mass spectral deciphering of the so-termed and above discussed histone codes, a key epigenetic mechanism, is expected to shed light on the phenomenon of metabolic programming: there is compelling evidence that the human body retains a memory of environmental, such as nutritional, impacts and may thereby be ‘metabolically (re)wired’. Such events have been associated with epigenetics, of which histone (de)acetylation is a central mechanism⁽Reference Bonenfant, Coulot and Towbin⁸⁷⁾. Recently O-aceylated serine and threonine residues have been identified⁽Reference Mukherjee, Hao and Orth⁸⁸⁾. Acetylation is stable to peptide fragmentation and can be detected by its characteristic mass shift from unmodified form. Trypsin cleavage at acetyllysine residues is usually blocked, so the acetylated peptides are detected as ‘missed cleavage’ product when performing database searches⁽Reference Witze, Old and Resing⁸⁹⁾. Enrichment of acetylated peptides is difficult and therefore studies have generally characterised protein acetylation on partially purified mixtures (e.g. histones). Immunoaffinity techniques have been developed to purify acetylated peptides: acetyllysine sites were mapped by enriching acetylated peptides using resin-coupled antibodies to acetyllysine⁽Reference Kim, Sprung and Chen⁹⁰⁾.

Lipoproteins are lipid–protein complexes whose primary function is thought to be transport of cholesterol and other lipids and that include five protein classes: chylomicrons, VLDL, intermediate density lipoproteins, LDL and HDL. Many lines of evidence strongly link them to the immune system and macrophage biology⁽Reference Barter, Nicholls and Rye^91–Reference Shiflett, Bishop and Pahwa⁹³⁾. Several approaches have been developed for lipoprotein isolation⁽Reference Havel, Eder and Bragdon^94–Reference Zechner, Moser and Kostner⁹⁷⁾; delipidation⁽Reference Folch, Lees and Sloane Stanley^98, Reference Karlsson, Leanderson and Tagesson⁹⁹⁾; sample preparation (solubilisation and digestion)⁽Reference Farwig, Campbell and Macfarlane^100–Reference Stahlman, Davidsson and Kanmert¹⁰³⁾; and the characterisation of protein component of lipoproteins using MS⁽Reference Mancone, Amicone and Fimia^104, Reference Vaisar, Pennathur and Green¹⁰⁵⁾.

Emerging proteomic technologies

The full characterisation of a given proteome remains a challenge today. This is due to factors such as the large dynamic range of protein expression, complexity of the mixture in terms of numbers of proteins, as well as lack of methods to amplify proteins. However, proteomics is a rapidly expanding field and new analytical approaches are emerging.

Usually, proteomic profiling is done in a data-dependent acquisition mode in which the most abundant ionised peptides from each MS scan are selected for subsequent MS/MS analysis. A data-independent acquisition method, referred to as Precursor Acquisition Independent From Ion Count, consists of the acquisition of MS/MS spectra at every m/z value regardless of whether a precursor ion is observed or not: precursor ion scans (MS scan) are no longer conducted⁽Reference Panchaud, Scherl and Shaffer²⁸⁾. This strategy yields better proteome coverage, higher numbers of identified proteins and an extended dynamic range compared to the classical data-dependent method⁽Reference Bern, Finney and Hoopmann¹⁰⁶⁾.

Imaging MS is a technology enabling the direct examination of the distribution of biomolecules (e.g. proteins, peptides, etc) in cells or tissues⁽Reference Chaurand and Caprioli^107–Reference Stoeckli, Chaurand and Hallahan¹⁰⁹⁾. Imaging MS is principally used for clinical applications⁽Reference Franck, Arafah and Elayed^110, Reference Samsi, Krishnamurthy and Groseclose¹¹¹⁾ and biomarker discovery in diseased tissue⁽Reference Schwamborn, Krieg and Reska^112, Reference Wong, Chan and Ma¹¹³⁾. In imaging MS, frozen tissue sections are mounted on a target plate, covered with a suitable matrix, dried and inserted into a MALDI-MS for spectra acquisition. The mass spectrometer records the spatial distribution of peptides and proteins by scanning the tissue surface with consecutive laser shots. Alternatively, in situ tryptic digestion of a spot on a tissue followed by peptide sequencing of a predicted fragment by MALDI-MS/MS can be done⁽Reference Groseclose, Andersson and Hardesty¹¹⁴⁾. Specific methods have been developed for the analysis of formalin-fixed paraffin-embedded sections⁽Reference Lemaire, Desmons and Tabet¹¹⁵⁾. Often histological staining, either on the same section⁽Reference Chaurand, Schwartz and Capriolo¹¹⁶⁾ or on a serial section⁽Reference Chaurand, Schwartz and Reyzer¹¹⁷⁾, is used to guide the placement of matrix and provides the capability of focusing on areas having a high content of a cell type of interest.

The trend towards biological analysis at decreasing scale, ultimately down to an individual cell, continues, and MS with sensitivity of detecting a few to single molecules will be necessary. Recently, a prototype for a mass spectrometer with single-molecule sensitivity for single-cell proteomics has been designed⁽Reference Naik, Hanay and Hiebert¹¹⁸⁾. Another method for the analysis of protein complexes in single cells, so-called visual proteomics, has been developed⁽Reference Beck, Malmstrom and Lange¹¹⁹⁾ and consists of the combination of quantitative MS with cryo-electron tomography for the detection, counting and localisation of protein complexes.

Proteomics today: a paradigm shift

A recent publication reviews proteome coverage and reports on the detection of protein abundance over seven orders of magnitude with today's high-end platforms⁽Reference Zhang, Faca and Hanash¹²⁰⁾. This impressive power is a combined result of highly improved mass spectrometric instrumentation and data acquisition/processing as well as of highly sophisticated fractionation, enrichment and depletion techniques.

However, given the complexity and dynamics of proteomes, proteomics experiences nowadays a paradigm shift. Strategically speaking, the original hypothesis-free discovery workflow is being increasingly complemented or followed up by either hypothesis-driven analysis or even by candidate-based targeted analysis and validation: a recent review puts the discovery, directed and targeted proteomics approaches into perspective⁽Reference Domon and Aebersold¹²¹⁾. Proteomics has thereby developed from a pure discovery to a screening and validation tool. The discovery workflow (or shotgun approach) aims at identifying large protein sets in a sample, and can include protein quantification (with or without prior protein labelling). The directed proteomics workflow consists of two successive analyses of the same sample. The first analysis is a survey scan aiming at the definition of a list of target peptides used for a second analysis surveying exclusively the peptides of the target list. This approach allows the quantification of less abundant proteins. Finally, targeted proteomics is a hypothesis-driven approach focusing on the detection and quantification of specific peptides associated with the proteins of interest (selected-reaction monitoring or multiple-reaction monitoring).

The other change of proteomic ‘philosophy’ roots in the increasing appreciation of peptides as bioactive, health beneficial food components⁽Reference Minkiewicz, Dziuba and Darewicz¹²²⁾. The analysis of such peptides requires a different analytical approach because these entities vary much more in their chemical nature than classical tryptic peptides generated in shotgun proteomics workflows for protein biomarker identification: multiple-processing parameters and digestive enzymes come into play and these generate not only a large variety in peptide length, sequence and terminal residues but also a number of peptide modifications. Moreover, there is only a single possibility to identify and quantify the native peptide of interest as such molecule is not one of several representatives of a parent protein, as it is typically the case in biomarker research. In view of this food peptidome complexity, it becomes evident that proteomic tools must be further developed and adapted from biomarker to bioactive research⁽Reference Kussmann, Panchaud and Affolter¹²³⁾.

Protein and peptides as food ingredients

Allergens: protein and peptides as food-derived causes of hazard

Food allergies arise from the intake of allergenic food components, which can induce a response from the immune system and lead to clinical symptoms ranging from mild to life threatening⁽Reference Bush and Hefle¹⁶⁸⁾. The prevalence of food allergy is rising; indeed 2% of adults and 5–8% of children in industrialised countries are affected⁽Reference Burks, Kulis and Pons^169–Reference Ortolani, Ispano and Scibilia¹⁷¹⁾. Over 180 protein allergens have been identified so far, the major ones occurring in common foods such as cow's milk, egg, peanut, soyabean, wheat, fish and tree nut⁽Reference Poms, Klein and Anklam¹⁷²⁾. Sensitive consumers have to be protected from undesirable allergic reactions and, therefore, proteomic methods have been developed for accurate allergen identification and quantification⁽Reference Kirsch, Fourdrilis and Dobson^130, Reference Poms, Klein and Anklam^172, Reference Monaci and Visconti¹⁷³⁾.

The classical proteomic strategy to identify food allergens consists of separating food proteins on 2D-PAGE, followed by electro-transfer onto a nitrocellulose membrane and subsequent IgE reactive protein detection by IgE immunoblotting using sera from allergic patients. This method was used to study allergens in wheat⁽Reference Akagawa, Handoyo and Ishii^174, Reference Sotkovsky, Hubalek and Hernychova¹⁷⁵⁾, apple⁽Reference Guarino, Arena and De Simone¹⁷⁶⁾, maize⁽Reference Fasoli, Pastorello and Farioli¹⁷⁷⁾ or sesame seeds⁽Reference Beyer, Bardina and Grishina¹⁷⁸⁾. A systematic proteomic analysis of rice (Oryza sativa) leaf, root and seed using 2D gels followed by MS/MS allowed for the detection and identification of more than 2500 proteins⁽Reference Koller, Washburn and Lange¹⁷⁹⁾ including several previously characterised allergenic proteins. The 2D difference gel electrophoresis method was also used to study several peanut varieties in order to show their low content of major allergens⁽Reference Schmidt, Gelhaus and Latendorf¹⁸⁰⁾. Recently, a method based on spectral counting was developed and successfully applied to the analysis of transgenic peanut lines containing reduced levels of certain major allergens⁽Reference Stevenson, Chu and Ozias-Akins¹⁸¹⁾.

Biomarkers: proteins and peptides as indicators of health and disease

Biomarkers are measurable indicators of different stages in a biological process, ranging from healthy functioning via deviation from such healthy equilibrium to disease onset and development⁽Reference De Roos¹⁸²⁾. Proteins and peptides, which are the main effectors in the body, can be used as such biomarkers. In practice, biomarkers are used for diagnostics, for prognostics, and to measure bioefficacy of nutrients in an intervention study.