Technical note: proteomic approaches to fundamental questions about neutrophil biology

Review of technologic advances that examine proteomics and its significant contributions to the understanding of neutrophil biology.

Abstract

Proteomics is one of a group of technologies that generates high-throughput, large-scale datasets that can be used to understand cell or organ functions at a systems level. This review will focus on the application of proteomics to the understanding of neutrophil biology. The strengths and weaknesses of common proteomic methods and their application to neutrophils are reviewed, with the goal of evaluating whether the technology is ready to advance our understanding of neutrophil biology.

Introduction

In the 1990s, signal transduction and cell-response pathways were considered to be linear arrays in which a single “upstream” event controlled a single “downstream” response. This understanding implied that biologic functions could be explained in terms of a direct link between genes and their translation products. The subsequent development of “omic” technologies, which allowed large-scale evaluation of genes, transcription events, and proteins, established a different reality, in which signaling pathways existed in networks that allowed for considerable interactive relationships [1, 2].

An understanding that cell phenotypes are controlled through complex signaling pathways led to the realization that networks can produce biologic outcomes not predicted by traditional reductionist approaches. Those outcomes, referred to as emergent properties [3], led to the development of an independent field of study called systems biology. The goal of systems biology as it relates to innate immunity is to generate models that predict the molecular networks responsible for normal innate-immune responses and to define those networks that produce phenotypic changes resulting in disease [4, 5]. Those predictive models are then tested in a hypothesis-driven manner using traditional molecular and cellular approaches. The iterative application of systems biology and traditional hypothesis-driven research promises to provide a more complete understanding of innate immunity.

Predictive models are generated by integrating data from technologies that generate high-throughput, large-scale datasets. Those technologies come under names, such as genomics, transcriptomics, proteomics, lipidomics, and metabolomics. Although systems biology is in its infancy and not without limitations, rapid technological advancement has validated the concept. This review will focus on the application of one technology—proteomics—to the understanding neutrophil biology. We will examine the strengths and weaknesses of proteomic approaches, and by reviewing the application of proteomics to neutrophils, we will evaluate whether that technology has advanced our understanding of neutrophil biology.

WHAT IS PROTEOMICS?

Proteomics is formally defined as the large-scale, systematic analysis of all proteins in an organ, cell, or system for their identity, quantity, and function. Development of proteomics as a field has primarily depended on improvements in scale, efficiency, and sensitivity of protein identification, largely through advances in MS and computational approaches to handling the large datasets that are generated.

Mass spectrometers contain three elements: an energy source that converts molecules into gas-phase ions; an analyzer that uses electromagnetic fields to separate ions based on their m/z ratio; and a detector that records the number of ions at each m/z ratio. The development of soft ionization that converts molecules into a cloud of gas-phase ions without excessive fragmentation was the key step that allowed for mass spectrometric analysis of peptides and proteins. In general, the m/z value is scaled to a charge (z) = +1, thereby providing the mass (m) of the ion in Daltons. This mass value, and the accuracy with which it is determined, is important in deducing the molecule's identity. For purposes of reproducibility, most MS-based experiments are conducted at an acidic pH to aid in the positive charging of the protein or peptide by the transfer of a proton (H⁺) to or from a molecule present in the sample solvent or sample matrix.

Figure 1 illustrates the basic features of the mass spectrometers described below. ESI and MALDI are the two principal methods for introducing biological samples into a mass spectrometer [6, 7]. ESI, but not MALDI, allows for direct and real-time analysis of samples through combination with LC. This advantage has made ESI the principal method of sample introduction into mass spectrometers for proteomics experiments. In experiments focused on direct analysis of biological samples that do not require separation of complex mixtures with LC, MALDI has the advantage of allowing rapid data collection from multiple samples.

Figure 1. (A–E) Acquisition of proteomic data.

The approach to acquiring MS datasets is influenced significantly by the complexity of the sample. (A) Samples are digested and data collected, using a MALDI or an ESI approach; former for higher-abundance/lower-complexity and later for lower-abundance/higher-complexity samples. The first data to be collected is MS¹-level data that provide information on the peptides present in the sample. The subsequent data are MS² (fragmentation)- or higher-level data that provide the information to reconstruct an amino acid sequence for the peptide or on PTMs. (B) MALDI-TOF instruments derive sample m/z values from the time the ions take to travel from the target plate to the detector and often use an ion reflector to improve the mass accuracies of the measured m/z values through refocusing ions into tighter packets prior to reaching the detector. (C) MS/MS, such as the MALDI-TOF-TOF instrument, can perform an initial MS¹ scan and select a peptide or set of peptides for sequential experiments, wherein the ions are isolated and fragmented and the fragments mass-measured. The selection of ions is most often conducted in a data-dependent fashion with the most intense signal selected first and then proceeding in rank order to less-intense peptides. For proteomic experiments, ESI instruments, such as the QqQ instruments, are coupled on-line to HPLC instruments. The sample is introduced into the MS via a fine aerosol (spray), and the ions that are created by evaporation of the solvent are steered into the instrument using guiding Qs (bundles of metal rods through which a current is passed and controlled to create tunable radiofrequency fields). For QqQ instruments (D), the ions can be allowed to pass straight through to the detector, and based on the tuning of the Qs, the m/z can be calculated. For MS/MS experiments, the second Q (q2) is filled with a gas, such as helium, and m/z values tuned in by the first Q (Q1) can be fragmented and then measured by the third Q (Q3). For ion trap instruments (E), the ions are guided into an electromagnetic trap using Qs and filled with ions. These ions can be pushed out of the trap using radiofrequency fields in a controlled fashion and m/z values measured (MS¹ scan). If desired, the trap can be refilled, a targeted population of ions (common m/z values) retained, then fragmented by collision-induced dissociation activation, and subsequently, mass-measured.

The measurement of separate m/z values requires the instrument to control the order in which ions reach the detector. To achieve this control, most instruments have relied on the interactions of ions with electromagnetic fields using TOF- or Q-based mass analyzers (scanning or ion-beam analyzers). More recent methods for ordering complex mixtures of peptides into discrete packets of ions use trapping analyzers equipped for low (LITs)- and high (FTICR and Orbitrap analyzers)-resolution mass measurements, thus achieving low- or high-mass accuracies, respectively [8–10]. Mass spectrometers with trapping analyzers have evolved to become the workhorse instruments for bottom-up or shotgun proteomics. Mass spectrometers with TOF or QqQ mass analyzers have evolved to become the instruments of choice for the qualitative or quantitative analysis of targeted protein samples [11]. MS/MS results in fragmentation of peptides at the C–N bond by a process of collision-induced dissociation. Mass spectrometers with this capability slowly heat gas-phase peptides by multiple collisions with rare gas atoms. In addition to allowing sequence information, this technique is used to identify specific PTMs and the amino acid that is modified [12]. Table 1 lists the characteristics and capabilities of the commonly used mass spectrometers.

Table 1. Comparison of Characteristics and Performances of Commonly Used Mass Spectrometers.

Instrument	Mass resolution	Mass accuracy	Sensitivity	Applications
QqQ (Thermo-TSQ)	2000	100 ppm	Attomole	Peptide quantification, high throughput, high complexity
Q-TOF (Waters SYNAPT G2–5)	20–40,000	<2 ppm	Femtomole attomole	Bottom-up peptide quantification, structural elicidation of isobaric species, qualitative and quantitative identification of protein complex components
Ultrahigh resolution Q-TOF (Bruker MaXis UHR-TOF)	40,000	<2 ppm	Femtomole attomole	Bottom-up peptide identification, top-down qualitative and quantification protein studies, qualitative and quantitative identification of protein complex components
Triple TOF (TOF/TOF/TOF) (AB Sciex 5600)	25–40,000	<2 ppm	Femtomole attomole	Peptide quantification
LIT (Thermo LTQ)	2000	100 ppm	Femtomole	Qualitative and semiquantitative (label-free) peptide-based proteomics, qualitative and quantitative identification of protein complex components
Q-LIT (AB Sciex QTRAP)	2000	100 ppm	Attomole	Qualitative and quantitative bottom-up proteomic analysis of medium-complexity samples, peptide quantification, qualitative and quantitative identification of protein components of moderate complexity
Q-Orbitrap (Thermo Q-Exactive)	140,000	<2 ppm	Femtomole attomole	Qualitative and quantitative (label-free, chemical labeling, metabolic labeling) peptide-based proteomics, high throughput, high complexity, top-down protein identification of smaller proteins, qualitative and quantitative identification of protein components of moderate-to-high complexity using label-free or labeling approaches
LIT-Orbitrap (Thermo LTQ-Orbitrap, XL, ELITE)	>450,000	2 ppm	Femtomole attomole	Qualitative and quantitative (label-free, chemical labeling, metabolic labeling) peptide-based proteomics, high throughput, high complexity, recently used with data acquisition in a data-independent fashion; high resolution, high mass accuracy measurements of low-to-high molecular-weight proteins analyzed in a top-down fashion; when coupled with electron-transfer dissociation, this platform represents the current state-of-the-art for top-down protein analysis and for structural analysis of glycan/proteoglycan structures (glycomics); qualitative and quantitative identification of protein components of moderate-to-high complexity using label-free or labeling approaches
LIT-FTICR (Thermo LTQ-FT-Ultra)	1,000,000	<2 ppm	Femtomole	Protein identification, quantification, PTM identification, top-down protein identification; high resolution, high mass accuracy measurements of low-to-high molecular-weight proteins analyzed in a top-down fashion; qualitative and quantitative identification of protein components of moderate-to-high complexity using label-free or labeling approaches

Identification of proteins can be top-down, where whole proteins are analyzed, or bottom-up, where enzymatic or chemically produced peptides are analyzed [11–13]. Bottom-up proteomic studies are applied to large-scale or high-throughput analysis of complex samples. Although bottom-up approaches have been applied widely and are effective, they have the disadvantages of information loss during conversion of proteins to peptides, inability to identify all peptides generated, and limitations of dynamic range as a result of the large number of peptides generated from highly abundant proteins. Top-down proteomics has the theoretical advantage of having the entire protein sequence available for analysis. Limitations of the top-down approach include difficulty with fragmentation of larger proteins, requirement for very high-resolution mass measurements, and low throughput and efficiency.

TYPES OF PROTEOMIC STUDIES

Expression proteomics

Proteomic studies fall into three broad categories: expression, structural, and functional. Expression proteomics are studies that define and quantify the protein components present in an organ, tissue, cell, organelle, or body fluid. To date, the majority of expression proteomic studies has provided qualitative, descriptive lists of proteins identified under static or resting conditions. The ability to answer meaningful biologic questions depends on advances in techniques to quantify proteins under dynamic conditions, and despite some advances, determining differences in proteomes related to specific cell phenotypes remains challenging. Although the enhanced sensitivity and accuracy of mass spectrometers has resulted in more complete identification of proteomes, identification of pathways and networks within those proteomes continues to occur primarily through post hoc pathway analysis using curated databases.

Structural proteomics

The goal of structural proteomics is to define PINs. The goals and potential problems with approaches to defining PINs have been reviewed recently [1, 14]. Experiments typically involve precipitation of a bait protein to capture and identify interacting proteins. In cell lines, the preferred method of isolating bait proteins is affinity purification of tagged proteins introduced into cells, typically by transfection. The use of tandem affinity-tagged proteins allows a significantly higher level of selectivity of protein purification. The use of cell lines and genetic manipulation of protein expression introduce potential problems in application of results to endogenous cells and tissues, particularly neutrophils that have a short lifespan. An alternative approach is immunoprecipitation of proteins from cell lines or endogenous cells, whenever an antibody is available. However, the lack of specificity of many immunoprecipitating antibodies creates a new set of problems with this approach. Confirmation of the interaction between two proteins can be accomplished by performing reciprocal purification. As copurification does not guarantee a direct interaction between two proteins, applying cross-linking techniques allows confirmation of direct interactions. Recently, an approach to global protein complex identification, using chromatographic separation of human-cultured cell extracts, followed by protein identification by LC-MS/MS, was reported [15]. A total of 622 putative protein complexes containing 2634 unique proteins were defined. Thus, large-scale identification of PIN may be possible in the near future. Even when direct interactions are demonstrated, PIN studies are still limited by the inability to define the directionality of that interaction.

Functional proteomics

Functional proteomic studies examine the mechanisms by which proteins communicate with each other through establishment of PSNs. MS approaches to defining PSNs have been reviewed previously [1, 2]. MS is now capable of identification of most PTMs, including phosphorylation, methylation, acetylation, ubiquitination, glycosylation, and proteolysis, and the amino acid undergoing the modification can be determined. Because of the importance of kinases in cellular functional responses and technological advances in phosphopeptide enrichment, most PTM studies have focused on protein phosphorylation. An advantage to those studies is that the direction of protein interactions can be determined. However, the definition of signaling networks is still constrained by limitations in the ability to determine the kinases and phosphatases controlling phosphorylation and the ability to identify other proteins that regulate or bind to phosphorylation sites.

DESIGN AND INTERPRETATION OF PROTEOMIC STUDIES

The approach to designing and interpreting proteomic studies is similar to that used for other experiments. Thus, identification of the experimental goals as well as an appreciation for the strengths and limitations of proteomic techniques are necessary prerequisites for good study design. In this section, we will review the common steps used in proteomic studies and the caveats that should be considered in designing such experiments. As proteomic studies require compromises among a number of factors, including the sensitivity and selectivity of protein identification, the reproducibility and accuracy of identification, and the dynamic range of the proteome being examined, it is useful to design experiments after discussing their goals with a mass spectrometrist or protein chemist.

Isolation of protein-containing structures

The increasing sensitivity and accuracy of mass spectrometers have enhanced identification of low-abundance proteins. As proteins that are transient components of signaling complexes or that are phosphorylated are typically expressed in low abundance, these improvements in MS make it more likely that proteomics studies will identify physiologically relevant proteins. However, it is also more likely that identified proteins will arise from low-level contaminants rather than the proteome of interest. Thus, investigators need to identify the level of sensitivity, i.e., the relative abundance of the proteins of interest, needed for successful protein identification. The higher the level of sensitivity demanded by the experiment, the greater attention must be paid to enrichment strategies for the organelle, complex, or other proteome to be analyzed.

Protein extraction

Once the desired proteome has been enriched, the appropriate technology must be applied to extract proteins from that proteome. Optimal extraction techniques applicable to all cell compartments do not exist, and studies of a global nature require multiple approaches tailored to the compartments being examined to achieve maximal proteomic coverage. Extraction from some tissues, cells, and cellular compartments can be particularly difficult. For example, in neutrophils, the ease with which proteins can be extracted from the different granule subsets varies greatly. Protein extraction from azurophil granules is challenging, possibly a result of packaging of proteins into a dense mucopolysaccharide core [16]. Protein extraction from cell membranes has also proved difficult. Whereas optimal extraction from membranes can be achieved using SDS, subsequent separation and fragmentation are markedly impaired unless gel-based techniques are used. An alternative approach using methanol-based protein extraction has been described and applied successfully to protein extraction from neutrophil plasma and secretory vesicle membranes [17]. Degradation of proteins by proteases released by cell disruption is a potential problem in all cell types, and it is a particular problem whenever neutrophils are disrupted. A number of protease inhibitors are available that can be added to the lysis buffer or introduced into cells prior to lysis. PMSF, 4-(2-aminoethyl)-benzenesulfonylfluoride hydrochloride, and diisopropylfluorophosphate are irreversible serine protease inhibitors, whereas benzamidine competitively inhibits serine proteases. EDTA binds the metal ions needed for metalloproteinases to function, and EGTA prevents activiation of calcium-dependent proteinases. Leupeptin and pepstatin are commonly used inhibitors of cysteine and aspartic proteases, respectively.

Protein separation

Optimal identification of proteins by MS requires prefractionation of proteins and/or peptides prior to MS. For many years, the protein fractionation technique of choice was 2-DE. The advantages of 2-DE were the ability to perform this separation in almost any laboratory, the availability of quantitative information from the gels, and the ability to combine samples from several experiments using differential labeling with specific fluorescent dyes. The disadvantages of 2-DE include its time-consuming nature, its bias against proteins of low or high molecular size or pI, and the impaired separation of membrane proteins. For those reasons, most investigators are transitioning to gel-free separation of proteins or peptides by LC, HPLC, or UPLC. Time savings have been achieved by performing peptide separation in line, resulting in direct injection of peptide fractions into the mass spectrometer. With the use of real-time sample introduction using HPLC or UPLC, investigators are able to take advantage of improvements in chromatographic separation of peptides directed at the selective isolation of specific groups of peptides, such as phosphorylated peptides using metal oxide affinity chelate- or immobilized metal affinity chelate-based resins [18, 19].

Protein digestion

Bottom-up protein analysis requires conversion of proteins to peptides that optimally contain 7–35 residues that are protonated and have a low charge state and high m/z [14]. Peptide generation has been performed almost exclusively by trypsin digestion as a result of the availability of inexpensive recombinant proteomics-grade trypsin. However, trypsin cleavage is relatively time-intensive and temperature-dependent, and 100% sequence coverage is difficult, as the arginine and lysine cleavage sites are not uniformly distributed in proteins, particularly membrane proteins. Although 100% sequence coverage is not necessary for protein identification, it is necessary to define sites of protein–protein interaction and to identify the set of PTMs present. A strategy to overcome the limitations imposed by trypsin includes performing protein digestion with a combination of trypsin and another enzyme (e.g., chymotrypsin), an endoprotease (e.g., Lys-C), or a chemical cleavage agent (e.g., cyanogen bromide).

In contrast to bottom-up proteomics, top-down analysis occurs by introducing intact proteins into the mass spectrometer, where they are subjected to gas-phase fragmentation. That approach eliminates the time-consuming step of protein digestion, provides access to the entire protein sequence, and enables identification of PTMs. However, this approach requires mass spectrometers with very high resolution to achieve sufficient mass accuracy. Fourier-transformed ion cyclotron resonance mass spectrometers have been the choice for this approach, although recent advances in Orbitrap analyzers have resulted in high-resolution observations and will likely result in increased use of the top-down approach.

Protein quantification

Large-scale comparison of protein expression associated with different cell phenotypes requires quantitative proteomic analysis. An advantage of gel-based protein separation techniques is that they can be modified to obtain quantitative data. This is the basis of DIGE, where differential fluorescent labeling of proteins obtained from cells expressing different phenotypes is subjected to 2-DE in the same gel. Comparison of fluorescent intensity with an internal standard allows comparison of relative expression levels.

Quantitative studies of protein expression by gel-free proteomics can be performed by in vivo metabolic labeling, chemical labeling, or unlabeled/label-free approaches. Whereas large-scale approaches for the quantification of whole proteins using a mass spectrometer are under development, all of these three approaches currently use peptide-based methods that are consistent with bottom-up proteomic studies [20]. Metabolic labeling requires incorporation of specific, stable, isotope-labeled amino acids into proteins [21]. The most common metabolic labeling experiment is the method known as SILAC [22]. These experiments are initiated using a defined cell-culture media that are depleted of a specific amino acid. In one condition, the media are supplemented using the naturally occurring amino acid and in a second condition, using a synthetic amino acid, wherein carbon, nitrogen, and sometimes, oxygen or hydrogen are heavy stable isotopes (e.g., carbon-13, nitrogen-15, oxygen-18, or deuterium). In most cases, the control condition is grown in the heavy media (to conserve costs), and a variety of treatment conditions are grown in normal or light media. After the defined treatment time, the cells are lifted and counted, and proportional numbers of heavy and light cells are mixed and lysed and the proteomic study conducted. It is important to note here that in shotgun experiments, the quantification of the protein is based on the integrated intensity of the MS¹ data for the peptide and not the fragmentation data. A developing direction for SILAC, which may have promise for proteomic studies of terminally differentiated cells, such as neutrophils, is the labeling of whole animals using feed, supplemented with heavy amino acids [23].

Chemical labeling of samples is another approach to multiplexing samples for quantification. Chemical labeling involves derivatization of proteolytic peptides derived after protein extraction from individual cell lysates or tissue samples, usually by amine-reactive chemistry. In most cases, this approach has the advantage of admixing multiple samples (multiplexing). In some cases, labeling is achieved using inexpensive reagents, such as reductive methylation of lysine side-chains or ¹⁸O-labeling of C-terminal carboxylate groups [24, 25]. As with stable, isotope-labeled amino acids, quantification of peptides using chemical derivatization is based on MS¹ data (with ¹⁸O-labeling) or MS² data (with isobaric-tagging reagents) [24, 26]. Other approaches can allow for multiplexing of up to six samples (TMT reagent; Thermo Fisher Scientific, San Jose, CA, USA) or eight samples (iTRAQ reagent; AB Sciex, Framingham, MA, USA) using isobaric-tagging reagents [27]. These approaches can be limited by incomplete fragmentation of labeled parent ions or under estimation of peptide-specific report ion abundances.

Label-free quantitation has the advantage of requiring no manipulation of proteins and can be performed by spectral counting or peak integration. Spectral counting is straightforward and facile in nature. However, it can only be applied to comparison MS/MS datasets after assignment of spectra by informatic programs, such as Sequest, Mascot, or X! Tandem.

Quantitative studies can also be conducted using a targeted proteomic approach, wherein prior knowledge of specific peptides has been gained. In these cases, the mass spectrometric experiment is designed to select for and base the quantification on the peptide-specific fragmentation data. This approach is termed SRM or MRM, and it requires the inclusion of a stable, isotope-labeled peptide for absolute quantification [28]. As compared with a shotgun proteomics experiment, the SRM/MRM approach achieves a high degree of combined specificity and sensitivity for quantification of the target analyte. SRM and MRM typically use the selectivity of a QqQ instrument (e.g., TSQ Quantum; Thermo Fisher Scientific, San Jose, CA, USA) to allow only ions of a specific mass to be transmitted from the first Q to the second Q. The second Q acts as a collision cell to generate fragments of the parent ions. In a SRM experiment, daughter ions are selected by the third Q and used for quantification of the parent peptide. Comparison of endogenous “light” peptide abundance with internal standard “heavy” peptides, defined as a peptide, where one amino acid is enriched in stable, heavy isotopes of carbon (¹³C) and nitrogen (¹⁵N), allows quantification based on external calibration curves composed of combinations of the heavy and light peptides.

Bioinformatic analysis

The final step in proteomic studies is an analysis of what may be an extensive list of proteins. A number of public and proprietary bioinformatics tools for analysis and modeling of proteomic data have been reviewed [4, 5, 29, 30]. One consequence of the development of more-sensitive mass spectrometers is a corresponding increase in the size of the MS/MS datasets to be analyzed. Given the statistical nature of peptide assignment to the MS/MS spectra, the possibility of more frequently assigning incorrect MS/MS spectra by random chance has resulted in a focus on developing methods, such as the use of decoy databases, to control false discovery rates and ensure accurate results. A second and more recent development in methods to analyze MS/MS datasets is the analysis of cross-linked proteins. Reagents, such as disuccinimidyl suberate (Pierce Biotechnology, Thermo Fisher Scientific, Rockford, IL, USA), are used to cross-link proteins found in protein complexes. Proteolysis of these complexes yields two peptides linked by the cross-linking reagent. The MS/MS spectra that result from the analysis of these cross-linked peptides cannot be assigned by routine approaches. This area of bioinformatics is still developing, but with maturation, should improve data analysis [31]. Statistical analysis of datasets is possible with appropriate sample replicates, as well as replicates of each individual MS run.

DEFINITION AND QUANTIFICATION OF NEUTROPHIL PROTEINS

The initial application of proteomics to a cell or tissue typically involves determining the proteins expressed without quantitation, and the application of proteomics to neutrophils has been no exception. Early studies using gel-based protein separation techniques and low-sensitivity mass spectrometers identified only highly abundant neutrophil proteins. Improvements in technology and methodology subsequently allowed more than 1000 proteins to be identified in single studies of human neutrophils or their organelles. Some of the major contributions to the neutrophil proteome have come from studies by Tomazella et al. [32], Trusch et al. [33], Uriarte et al. [17], and Jethwaney et al. [34].

Tomazella et al. [32] analyzed detergent-insoluble proteins and cytosolic proteins from human neutrophil lysates by gel electrophoresis, followed by LC-MS/MS. A total of 1249 proteins were identified, of which, 30% were from metabolic pathways. Trusch et al. [33] used 2D LC coupled to MS/MS to identify 253 unique proteins from human granulocyte cytosol. Uriarte et al. [17] used 2D-LC-MS/MS to analyze proteins extracted from plasma membrane- and secretory vesicle membrane-enriched fractions and identified 1116 proteins, of which, 266 were also identified by Tomazella et al. [32]. Analysis of those proteins using Ingenuity Pathways Knowledge Base analysis demonstrated that proteins belonging to unique canonical pathways distinguished the two membrane fractions. Feuk-Lagerstedt et al. [35] used gel-based protein separation and HPLC-MS/MS to identify 43 and 37 unique proteins from secretory vesicle membranes and plasma membranes, respectively, 20 of which represented new proteins in the neutrophil proteome.

Other potentially more instructive proteomic studies have examined specific structures within the neutrophil. Lominadze et al. [16] determined the distribution of 286 proteins in gelatinase, specific, and azurophil granules, of which, 174 were new to neutrophils. Feuk-Lagerstedt et al. [35] examined the proteins contained in detergent-resistant domains of the membranes from azurophil granules. A total of 106 proteins were identified, including a number of cytoskeletal proteins, many of which were also identified in one or more granule subset by Lominadze et al. [16].

Two other studies examined the composition of the cytoskeleton in neutrophils. Xu et al. [36] examined the composition of the neutrophil cytoskeleton as the detergent-insoluble fraction obtained from cytoplasm, phagosome, and plasma membrane. Proteins in those fractions were subjected to 2-DE in-gel trypsin digestion and analyzed by MALDI-TOF-MS. That analysis identified 138 unique proteins, of which, 72 were associated with the cytoskeleton from cytosol, 39 from the plasma membrane, and 27 from the phagosome. Metabolic enzymes were enriched in the cytoplasmic cytoskeleton, whereas signaling proteins were enriched in the plasma membrane cytoskeleton. Very few metabolic enzymes or signaling proteins were associated with the phagosomal cytoskeleton. Nebl et al. [37] separated proteins from detergent-resistant fractions from neutrophil plasma membranes by SDS-PAGE and analyzed peptides from in-gel digestion by MALDI-MS/MS. Twenty-seven proteins were identified and their expression confirmed by immunoblot analysis, seven of which were also identified in the study by Xu et al. [36].

Burlak et al. [38] used a proteomic approach to define the components associated with phagosome maturation. Proteins from phagosomes isolated after 30 min of receptor-mediated phagocytosis were separated by 1-DE or 2-DE, and peptides generated by in-gel digestion were identified by MS. A total of 198 unique proteins were identified, including 30 cytoskeletal proteins, 21 proteins associated with the ER, 28 signal transduction proteins, 29 host defense proteins, and 49 proteins associated with metabolic processes. Comparison of those proteins with the proteins identified from macrophage phagosomes suggested common mechanisms of phagocytosis, in addition to microbicidal activity [39].

We used the studies outlined above to compile a catalog of proteins identified in human neutrophils. Taken together, those studies have identified over 2300 proteins. However, compiling a list of proteins in the neutrophil proteome omits critical data, such as the primary intracellular location and the changes in protein expression or location that accompany neutrophil activation. The article by Rørvig and colleagues [unpublished results] in this issue illustrates the abililty to use proteomics to define the location of proteins in neutrophils. They used a more sensitive instrument, the LTQ Orbitrap, to perform a quantitative comparison of protein expression among the various neutrophil granule subsets. Their data defined at least eight patterns of protein expression in granules, secretory vesicles, and cell membrane from resting neutrophils.

The effect of advances in instrumentation and proteomic techniques on the ability to determine changes in protein expression following neutrophil activation was illustrated by two studies published eight years apart. In 2002, Fessler et al. [40] used 2-DE to separate proteins from resting and LPS-stimulated neutrophils, which were then identified by MALDI-TOF-MS. Analyzing changes in intensity of ∼1200 protein spots in gels from each of six different experiments identified ∼125 protein spots with increased expression, 4 h after LPS stimulation, whereas ∼110 spots showed reduced intensity. However, only 125 spots were concordant among six gels, of which, only 25 proteins with statistically significant changes in expression were identified. We performed a PubMed search, in which we found that only five of those 25 proteins had been previously linked to LPS stimulation of any cell. The discrepancy between the number of protein spots with altered expression on 2D gels and the number of proteins identified illustrates the limitations of gel-based protein identification. Typically, proteins contained in a number of spots cannot be identified by MS, as a result of technical limitations and low protein content. Additionally, changes in density of spots can occur, as PTMs shift the pI of the proteins rather than a true change in protein expression. In 2010, Kotz et al. [41] used a microfluidic approach to isolate human neutrophils from whole blood. The cells were subsequently analyzed for changes in gene and protein expression following stimulation with LPS or a combination of IFN-γ and GM-CSF. Proteins were identified following disruption of neutrophils by urea or hypotonic lysis, with or without 2,2,2-trifluoroethanol, and analyzed by reversed-phase LC-MS/MS. With the use of those three disruption techniques, a total of 364, 207, and 560 proteins, respectively, were identified, including 70 proteins that demonstrated a change in expression following stimulation by LPS, a threefold increase in sensitivity over that obtained by Fessler et al. [40]. Of those 70 proteins, the authors identified 15 that were previously shown to be related to LPS stimulation. The number of proteins with altered expression after LPS stimulation, not previously associated with a LPS, suggests that the effect of LPS on neutrophils is extensive.

Each of the above studies evaluating protein expression after LPS stimulation also examined changes in mRNA expression. Fessler et al. [40] identified 923 genes common to three individual donors, and they reported increased expression of 100 genes and decreased expression of 56 genes following LPS stimulation. Those numbers correlate well with the number of spots with altered protein expression on their gels. However, protein and gene expression correlated poorly, as gene expression and protein expression moved in the same direction in only five of 13 proteins analyzed. Kotz et al. [41] screened 21,000 genes and found that 2452 showed a greater than twofold change in expression. They evaluated concordance of gene and protein expression for the 15 proteins known to interact with LPS and found that gene and protein expression moved in the same direction for 12 genes. Although protein and gene expression moved in the opposite direction for a number of proteins in the group as a whole, the authors did not provide information on the total number of genes that demonstrated discordance. These studies indicate that current technology has the ability to define changes in protein expression and location in neutrophils stimulated in vitro or obtained from patients with inflammatory diseases. The studies also indicate that changes in gene and protein expression can measure different events, leading to discordant results. Gene and protein expression changes have also been compared during myeloid development using a murine cell line [42, 43]. Those studies showed a general concordance in the timing of changes in protein and gene expression during differentiation, and several transcription factors not associated previously with myeloid development were identified.

The ability of neutrophils to change expression of surface molecules and release microbicidal agents, such as granule contents, neutrophil extracellular traps, and microparticles, has recently been examined using proteomic approaches. Galkina et al. [44] reported that neutrophils attach to bacteria and other cells through elongated membrane tethers or TVEs. To characterize the physiologic relevance of those tethers, a proteomic analysis of secreted proteins and proteins in isolated TVE following neutrophil adhesion to fibronectin was performed by gel-based protein separation followed by MALDI-MS/MS of peptides generated by in-gel protein digestion [44]. A total of 14 secreted proteins were identified, including granule proteins, metabolic enzymes, and S100 proteins. A total of 23 proteins were identified from TVE, including the same granule proteins, cytoskeletal components, metabolic enzymes, and the same S100 proteins. That the relatively small number of proteins identified in each group resulted from the limitations in the technical approach is suggested by a study by Zhang et al. [45], who used NanoLC-MS/MS to define the proteome secreted by neutrophils, stimulated by two peptides, derived from chromogranin A. Those authors identified 73 unique proteins, including the same granule proteins, cytoskeletal proteins, and S100 proteins identified by Galkina et al. [44].

Timár et al. [46] recently reported that microvesicles released by neutrophils, incubated with opsonized bacteria, possessed antimicrobial activity, whereas microvesicles released after other stimuli, including fMLF, TNF-α, PMA, and LPS, did not alter bacterial growth. To define the basis of that microbistatic activity, a quantitative proteomic analysis was performed by methanol protein extraction and 2D-LC-MS/MS on microvesicles obtained from unstimulated, PMA-stimulated, and bacteria-stimulated neutrophils. A total of 282 proteins were identified from the three groups, and a quantitative comparison of the 100 most highly expressed proteins was performed by spectral counting. Of the 29 proteins demonstrating a >40% increase in expression in bacteristatic microvesicles, 15 were known to possess antibacterial activity, and 26 were expressed in neutrophil granules. Those results suggest that granules can contribute antibacterial proteins to microvesicles following specific stimuli.

DEFINITION OF NEUTROPHIL PINs

Whereas the studies described in the previous section have used advances in MS to establish extensive proteomes of neutrophils and their substructures, identification of pathways and networks within the proteome has occurred, primarily through post hoc analysis of curated databases. A few studies suggest that a strength of proteomics is the ability to identify PINs in neutrophils using direct experimental approaches, as illustrated by the following examples.

The protein kinase Akt was previously reported to regulate oxidase activity, chemotaxis, actin polymerization, and apoptosis in neutrophils [47–49]. To define the molecular mechanisms of that regulation, Rane et al. [50] used immunoprecipitation of Akt to identify Akt-binding partners. Immunoprecipitates were subjected to 2-DE, and MALDI-TOF-MS was used to identify proteins following in-gel digestion. A C-terminal portion of the GABA_BR2 was identified. Subsequent studies determined that the GABA_BR2 was present in neutrophils, interacted with Akt, and mediated neutrophil chemotaxis and tubulin reorganization. Florentinus et al. [51] isolated FcR complexes from live neutrophils and performed a quantitative analysis of proteins identified by 2D-LC-ESI-MS/MS. A total of 245 proteins were identified from the FcR complex, of which, 117 were cytoskeletal proteins. The authors generated a protein-interaction network, describing reorganization of the actin cytoskeleton by FcRs using the mathematical modeling packages, Cytoscape, Osprey, and String.

Finally, Richmond and colleagues [52, 53] immunoprecipitated CXCR2 from differentiated HL-60 granulocytes, following incubation with or without stimulation with CXCL8. The immunoprecipitated proteins were separated by gel electrophoresis and in-gel trypsin digests analyzed by LC-MS/MS. They reported identification of 23 proteins associated with an unstimulated receptor (11 proteins), with a CXCL8-stimulated receptor (six proteins), or with both conditions (six proteins). Characterization of two proteins, vasodilator-stimulated phosphoprotein and IQGAP1, confirmed direct association with the receptor and defines roles for both proteins in receptor-mediated cell functions.

DEFINITION OF NEUTROPHIL PSNs

Although genomic systems biology approaches have been applied to neutrophils with some success [41, 54–56], large-scale proteomic studies of PTMs to define signaling networks in neutrophils remain scarce. MS can be used to define or confirm PTMs of amino acids of specific proteins. In the case of neutrophils, examples include the definition of N-linked glycosylation of gp91^phox and p22^phox [57], phosphorylation of S100A8/9 [58], and S-nitrosylation of S100A8 [59]. More-generalized approaches aimed at discovering proteins involved in signaling pathways have also been reported. Boldt et al. [60] used differential 2-DE-DIGE to identify differences in protein expression after fMLF stimulation of neutrophils. Subsequently, increased phosphorylation of L-plastin and reduced phosphorylation of moesin, cofilin, and stathmin were identified by phosphospecific gel staining with Pro-Q Diamond and protein analysis of in-gel tryptic digests by LC-MS/MS. Luerman et al. [61] used phosphopeptide enrichment with titanium dioxide or gallium affinity chromatography of proteins digested from the surface of specific granules derived from unstimulated and stimulated neutrophils to identify phosphoproteins by LC-MS/MS. A total of 79 phosphoproteins were identified from resting neutrophil granules, 81 following 1 min of fMLF stimulation, and 118 following 2 min of stimulation. Despite that large pool of phosphoproteins on granules, a model of the signaling networks that control specific granule exocytosis was not elucidated.

Studies identifying protease substrates and defining the peptides generated by those proteases have expanded the role of neutrophil degranulation in regulating innate and adaptive immunity. For example, Starr et al. [62] performed a proteomic analysis on the degradation products of leukolysin (MMP25). In addition to determining that MMP25 processes CXC and CC chemokines into active forms, a cleavage product of vimentin, which stimulated phagocytosis of neutrophils, was identified. This study demonstrates the ability of proteomics to assist in defining downstream targets of enzymes released during neutrophil degranulation.

CONCLUSIONS

Neutrophil proteomic studies have not yet fulfilled the goal of identifying the emergent properties created by complex molecular networks, as defined by the systems biology paradigm. This review has summarized the technologic and experimental progress, indicating that the tools are rapidly becoming available, which will allow a systems biology approach to neutrophil biology and innate immunity. The application of proteomics to neutrophils, to date, offers evidence that an iterative interaction between proteomic approaches and traditional hypothesis-driven studies will provide new insights into the role of neutrophils in innate- and adaptive-immune responses. Extension of those insights into disease-related networks promises to lead to improved understanding of complex immune and inflammatory diseases, to identify biomarkers of those diseases, and to define new therapeutic targets.

DISCLOSURES

The authors declare no conflict of interest.