Proteomics to study the diversity and dynamics of proteasome complexes: from fundamentals to the clinic

Abstract

This article covers the latest contributions of proteomics to the structural and functional characterization of proteasomes and their associated proteins, but also to the detection of proteasomes as clinical biomarkers in diseases. Proteasomes are highly heterogenous supramolecular complexes and constitute important cellular proteases controlling the pool of proteins involved in key cellular functions. The comprehension of the structure/function relationship of proteasomes is therefore of major interest in biology. Numerous biochemical methods have been employed to purify proteasomes, and have led to the identification of complexes of various compositions – depending on the experimental conditions and the type of strategy used. In association with protein separation and enrichment techniques, modern mass spectrometry instruments and mass spectrometry-based quantitative methods, they have led to unprecedented breakthroughs in the in-depth analysis of the diversity and dynamics of proteasome composition and localization under various stimuli or pathological contexts. Proteasome inhibitors are now used in clinics for the treatment of cancer, and recent studies propose that the proteasome should be considered as a predictive biomarker for various pathologies.

Keywords::

• 2D electrophoresis
• affinity purification–mass spectrometry
• biomarker
• in vivo crosslinking
• multiple reaction monitoring
• post-translational modifications
• proteasome heterogeneity
• proteasome interactome
• quantitative proteomics
• subcellular localization

Figure 1. Diversity of proteasome complexes.

Mammalian proteasome complexes result from the assembly of a 20S core complex, the 20S proteasome, with one or two regulatory particles (RPs) of identical or different protein composition. The 20S core complex is a stable entity assembled in four stacked ring-shape heptamers, α7β7β7α7. Three subunits, β1, β2 and β5, of each β ring exhibit three distinct catalytic activities. These standard β subunits can be completely or partially replaced by three immunosubunits (β1i, β2i and β5i, respectively), which leads to various forms of 20S proteasomes (immunoproteasome and intermediate-type proteasomes). The two α-rings are implicated in the recruitment of RPs at each side of the 20S core complex. In mammals, four activators have been identified: the 19S RPs, PA28αβ, PA28γ and PA200.

Figure 2. Affinity purification methods of proteasome complexes.

(1) Purification using a tagged 19S regulator subunit, (2) purification using a tagged domain recognizing a subunit of the 19S regulator, (3) purification using a tagged 20S core particle subunit and (4) purification using an antibody specifically directed against a conserved 20S proteasome subunit.

-: Not possible; +: Good; ++: Very good.

PIP: Proteasome-interacting proteins; UBL: Ubiquitin-like; UIM: Ubiquitin-interacting motif; Xlink: Proteins have been in vivo crosslinked prior to purification.

Figure 3. Targeted detection of peptides by mass spectrometry using selected reaction monitoring acquisition mode.

m/z: Mass-to-charge ratio; SRM: Selected reaction monitoring.

It is now clear that proteins mainly act in association with other proteins in a dynamic way to reach a tight control of cellular processes [1]. In order to address these issues, proteomics has turned into functional proteomics, which has become a driving force to develop powerful tools for the study of protein machines, and ultimately, protein networks. Among the eukaryotic cellular protein machineries, the ubiquitin–proteasome system (UPS) is of particular importance since it is involved in the selective degradation of most short-lived intracellular proteins [2,3]. In most cases, the degradation process is initialized through the polyubiquitination of the substrate. The tagged protein is then efficiently recognized by the proteasome, unfolded and degraded. Major biological processes, such as cell cycle progression, apoptosis, DNA repair, epitope generation and cell quality control, are tightly regulated by this system [4]. Many studies have demonstrated that a dysregulation of this machinery is related to various pathologies such as neurodegenerative diseases [5] and cancer [6]; this contributes to the identification of proteasomes as therapeutic targets, especially for some cancers [7].

Proteasome complexes consist of a 20S catalytic core particle (CP) comprising 14–17 different proteins, either alone or associated with one or two regulatory particles (RPs) that can be of identical or different protein composition (Figure 1). In mammals, four activators have been identified: the 19S RP, PA28αβ, PA28γ and PA200. The 26S proteasome is a particular form where the 20S CP is capped by two 19S RPs, forming a 2.4 MDa complex; it is involved in the ubiquitin-dependent degradation of proteins through specialized functions of specific subunits of the 19S, such as polyubiquitinated substrate recognition, ATP-dependent substrate unfolding and ubiquitin recycling. Hybrid proteasomes correspond to forms where the 20S CP is capped with two different regulators, mainly 19S RP and PA28. The eukaryotic 20S proteasome is assembled in four stacked ring-shape heptamers, with seven unique α-subunits in the two outer rings and seven unique β-subunits in the two inner rings. The two β-rings each contain three catalytic subunits – β1, β2 and β5 – which are totally or partially replaced by the so-called immunosubunits – β1i, β2i and β5i – in the immunoproteasome or intermediate-type proteasomes, respectively. The catalytic subunits are responsible for three main proteasome proteolytic activities (trypsin-like, chymotrypsin-like and caspase-like), which can be modulated by the replacement of standard subunits with immunosubunits [8]. The immunoproteasome is induced during the immune response in mammals but, together with intermediate-type proteasomes, also exists as constitutive proteasome complexes – depending on tissues or cell type – and can have distinct proteolytic activities [9–14]. Theoretically, taking into account the rules of cooperative assembly of inducible catalytic subunits, 13 different configurations of intermediate-type proteasomes can be formed, in addition to the standard proteasome and the immunoproteasome (Figure 1)[11]. The functional roles of these different proteasome subtypes are largely unresolved, but a specialized function might be associated with each of them. The thymoproteasome is another recently discovered form of 20S proteasome exclusively expressed in cortical thymic epithelial cells of vertebrates and containing the β5t subunit, a novel catalytic site with unusual enzymatic activity, probably involved in the positive selection of developing thymocytes [15]. The association of the 20S CP with its different regulators is ensured through the two α-rings, which also regulate the entry of substrates into the CP.

Proteasome complexes therefore exhibit a high degree of heterogeneity in their overall subunit composition and can, in addition, recruit other so-called proteasome-interacting proteins (PIPs), the identification and role of which are one of the challenges of today’s proteasome research. Proteasomes constitute dynamic structures with respect to the cellular environment, such as inflammation [16], the cell type [9,17,18], its subcellular localization [19–21], the tissue [22–24], the stage of development [25,26] or the pathophysiologic context [10,27–31]. This diversity is probably the result of specialized functions of each individual proteasome form [32], providing varied means to adapt protein degradation pathways to changing conditions in the cell [33]. Therefore, a precise structural characterization, as well as a functional understanding of each proteasome subtype, are needed and still constitute a challenge in the field of proteasomes. Unfortunately, no precise structural determination of the whole 26S proteasome complex is currently available, probably owing to the tremendous complexity, diversity and lability of proteasomes. So far, only subcomplexes have been crystallized for atomic structure determination [34], but recent progress has been made in determining the 26S proteasome structure at subnanometer resolution [35]. Proteomic strategies constitute complementary methods to structure determination approaches for the study of protein machines [36]. The association of modern mass spectrometry (MS) strategies and efficient biochemical approaches, in particular affinity purification (AP) methods, has been critical for the initial success of functional proteomics [37–39]. Today’s challenges for proteomics in the analysis of protein complexes are to determine the composition and stoichiometry of protein complexes and to measure their variations, either in different cell types or after inducing a protein expression change, or over time after applying a stimulus, in order to distinguish stable from dynamic partners and to identify post-translational modifications (PTMs) on interactants.

Detailed proteomic approaches to study proteasomes have recently been reviewed [40,41]. Therefore, this article will mainly cover the developments over the last 4 years in proteasome proteomics, and will envisage future trends for improving our knowledge of this complex machinery and for the consideration of proteasomes as clinical biomarkers in cancer or other pathologies.

Latest developments for the analysis of proteasome complexes by proteomic approaches

Purification of proteasomes

The detailed structural characterization of a given protein complex begins with its purification. When the complex is an enzyme like the proteasome, the challenges are both to preserve its enzymatic activity for further functional studies and to obtain a high purification yield. Given the high heterogeneity of proteasomes and the differences in stability and dynamics of the PIPs, the experimental conditions and the type of strategy used will highly influence the purity, the composition of complexes obtained (20S CP, 26S proteasome or the whole diversity of proteasome complexes), and the nature of the interactants purified (specific, stable and dynamic).

The 20S proteasome is a salt-resistant complex, but its association with the different regulators (Figure 1) and partners is much more labile [42,43]. This explains why the number of surveys dealing with 26S proteasomes and the characterization of other existing proteasomes by proteomic strategies is much more limited than those for studying the 20S core complex. Selective salt precipitations, differential or gradient centrifugations and multidimensional liquid chromatographies combined with glycerol gradient ultracentrifugation have been used to purify the 20S CP (see [41] for a review) and the 26S proteasome [44–46] from various species to apparent homogeneity. Recent in-depth characterization of 20S proteasomes subtypes were reported in studies relying on high-resolution anion-exchange chromatography [12] and native isoelectric focusing–free-flow electrophoresis (IEF–FFE) [13] as the final purification step. The different fractions contained reasonably well-separated standard proteasome and immunoproteasomes but also intermediate-type proteasomes. Interestingly, IEF–FFE of intact proteasome complexes can distinguish between differentially phosphorylated 20S proteasome complexes [13]. Intermediate-type proteasomes could be further discriminated into several subtypes, which differed mainly in the composition of their catalytic β-subunits and their enzymatic properties [12,13], but also in their subcellular compartment repartition [11], their tissue/cell distribution [10,12,13], their susceptibility to proteasome inhibitors or with respect to pathologies [10,12]. These classical biochemical methods combined with further subunit identification have therefore improved our knowledge of the 20S proteasome structure/function relationship. However, they remain better suited for the study of 20S than 26S proteasomes, especially when high salt concentrations are used during elution steps [43,47].

The study of native and fully functional forms of proteasomes (20S proteasome in association with its different RPs: single 20S, singly and doubly PA28-capped, singly and doubly 19S-capped, 19S and PA28 hybrid) can be performed using 2D blue native (BN)/sodium dodecyl sulfate polyacrylamide gel electrophoresis (PAGE), which allows for their separation. When combined with MS, the dynamic variations of the different complexes can be analyzed under various stimuli [48,49]. In association with antibody-based gel shift assay, BN PAGE can also constitute an alternative strategy for identifying specific partners from the same multi-protein complex [50], as already demonstrated for proteasome complexes [48,51].

Alternatively, AP methods such as immunoprecipitation, immunochromatography and epitope tagging strategies, although costly and technically challenging, can generate a high yield of pure and functional proteasome complexes from a limited amount of starting material. These methods are also faster and simpler because they can be performed in 1 day and in a one-step procedure. In the past decade, several AP approaches have been developed to purify proteasome complexes. Associated with technological developments in MS-based proteomics, they have clearly enlarged our knowledge concerning protein complex networks [37,39,52]. These approaches can differ in several ways:

• • The bait protein employed to catch the complexes;

• • The use of an epitope tagging strategy;

• • The composition of the buffers used along the purification scheme;

• • The sample preparation strategy, for example, the possibility to use an in vivo chemical treatment to strengthen protein–protein interactions.

As will be discussed, these key features influence the diversity of proteasome complexes purified, the efficiency of PIPs recovery, the probability of catching endogenous interactions and the possibility of handling clinical samples (Figure 2).

Purification using a tagged 19S regulator subunit

A fruitful AP–MS approach to analyze protein complexes relies on overexpression of tagged proteins in relevant cell lines. It is particularly suitable for proteins present at low endogenous levels and for labile complexes like 26S proteasomes (Figure 2, purification method 1). Such a strategy using a tagged 19S subunit and MS analyses of coimmunoprecipitated proteins identified Adrm1, the human orthologue of the yeast Rpn13 subunit [53,54], and confirmed arsenite-inducible RNA-associated protein as an arsenite-inducible subunit of the 19S proteasomal cap [55]. Two-step methods, such as tandem affinity purification (TAP), can also be applied by introducing two different tags, which allows the purification of the complexes with high purity and lowers the number of contaminant proteins [56]. However, TAP strategies are typically multistep procedures involving long incubation times and multiple elutions, which may destabilize labile protein–protein interactions This problem has been successfully overcome by combining TAP and formaldehyde in vivo cross-linking to efficiently capture labile 26S proteasome partners in yeast [57]. In this strategy, a histidine biotin tag, consisting of an in vivo biotinylation signal flanked by two hexahistidine tags, was introduced at the C-terminus of several 19S proteasome subunits. Quantitative MS analysis then allowed the identification of a global map of the yeast 26S proteasome network consisting of 471 PIPs [57]. However, this approach has been performed in denaturing conditions, which prevents further in vitro proteolytic activity measurement of the purified complexes. More generally, the fact that a subunit of the 19S regulator is used as prey might constitute a limitation for two reasons. First, the 19S regulator was recently shown to exert several proteolysis-unrelated biological functions independently of the 20S CP. These include transcription, DNA repair and chromatin remodeling [58,59], and might therefore lead to a specific 19S regulator interactome, which cannot be distinguished from the proteolytic-related proteasome interactome. Second, only proteasome complexes containing at least one 19S particle can be isolated. Free 20S particles and proteasomes involving solely PA28 or PA200 regulators are not purified, even if they are abundant in the entire pool of proteasome complexes [60,61], and might be associated with specific partners. Interestingly, recent surveys describe possible implications of these proteasomes, in particular PA28-associated proteasomes, in various pathologies, such as Parkinson’s disease [30], cancer [31,62,63], viral infection [64,65] and alcoholic liver injury [66].

Purification using a tagged domain recognizing a subunit of the 19S regulator

Targeting wild-type endogenous complexes is an appropriate way of studying protein complexes, since any sample type can be analyzed and no cellular perturbation is expected, as might happen when using overexpression or tagging strategies. Strategies using a tagged domain recognizing the endogenous protein complex of interest as bait have recently been developed for 26S proteasomes purification. The high affinity of the ubiquitin-like (UBL) domain of either hHR23A [67] or hHR23B [68] proteins for the ubiquitin-interacting motif (UIM)-2 domain of S5a (Rpn10), a subunit of the 19S regulator, was used (Figure 2, purification method 2). These two domains were cloned and produced as recombinant glutathione-S-transferase (GST) fusion proteins. In both approaches, the purification workflow consisted of:

• • Binding of proteasomes and other UIM-containing proteins to the GST-UBL;

• • Elution with an excess of His-UIM-2;

• • Removal of His-UIM-2 using nickel-nitrilotriacetic acid resin.

Further MS experiments allowed the identification of proteasomes and PIPs obtained from rat skeletal muscle [68], rat cortex [60] and human embryonic kidney (HEK-293) cells [67]. Interestingly, together with 20S and 19S proteasome subunits, as well as known PIPs such as proteasome activators, chaperones or ubiquitin conjugates, several proteins belonging to the UPS but not previously known to associate with the 26S proteasome were identified. Among these, some are deubiquitinating enzymes, ubiquitin-conjugating enzymes and ubiquitin ligases. The UBL AP–MS approach is of great interest, since it permits endogenous and functional proteasome complexes from tissues or biological fluids of whatever mammalian origin to be obtained. The UBL domain of hHR23B might be more suited to efficient proteasome complex purification as it presents a higher affinity than the one of hHR23A for S5a [69]. However, the specificity of the proteasome interactome obtained needs further validation, as several of the proteins purified using this approach, in particular all the cellular proteins exhibiting an UIM domain, might be captured through direct binding to the UBL domain. Indeed, the UIM domain occurs in a wide variety of proteins and might constitute a general ubiquitin-binding motif [70]. Moreover, this approach suffers from the same limitations as the purification strategy using a tagged 19S subunit as previously mentioned.

Purification using a tagged or a native 20S CP subunit

Affinity methods using a subunit of the 20S core complex, associated with all of its interactors, are more prone to catch the whole diversity of proteasome complexes (Figure 2, purification methods 3 and 4). ATP and glycerol preserve the interactions between the 20S CP and the 19S regulators [42,43,49], and are therefore required for a successful purification of 26S proteasomes. On the other hand, ATP is not required to maintain PA200- and PA28-associated proteasome complexes, but these interactions are salt sensitive [49,71]. Strategies using the genetic replacement of β4 [42,43,55,72] in yeast or, more recently, of α7 in plants [73] with an affinity-tagged variant, have been described and led to the efficient and rapid purification of a heterogeneous collection of proteasomes (Figure 2, purification method 3). These approaches necessitate careful optimization to find the best tag, its best location on the gene and, most importantly, to check the proper integration of the modified subunit in the tightly associated 20S core complex so that the biological and/or catalytic activities are preserved. Simpler strategies using an antibody specifically directed against a conserved 20S proteasome subunit are probably more biologically relevant and also should allow clinical samples to be dealt with (Figure 2, purification method 4). However, efficient and specific antibodies against one of the protein complex components and resistance to quite harsh washing conditions (to increase specificity) are required. Any constitutive subunit of the 20S CP might be used as prey, as long as the epitope is not hindered when the physiological complex is assembled. The α-subunits of 20S proteasome are excellent candidates since they are theoretically present in all existing subtypes of 20S proteasomes. Strategies using monoclonal antibodies directed against the 20S core α2 subunit [74], α3 subunit [75] and α6 subunit [76] were recently successful to purify proteasome complexes. The mouse IgG₁ monoclonal antibody MCP21 [77] is of particular interest, not only because it efficiently and specifically recognizes the 20S core α2 subunit but also because it can easily be produced by a commercially available hybridoma cell line. Our group has reported an efficient workflow for the immunopurification of highly pure 20S proteasomes in a single-step procedure [9,78]. This protocol was then miniaturized to deal with a small number of epithelial colorectal cells (30 × 10⁶ cells) available from patient tissue samples [79]. It allowed the determination and comparison of 20S proteasome complex subunit composition from tumor and corresponding healthy human colorectal cells from clinical samples using 2DE.

This purification strategy was then further optimized to purify the whole pool of proteasome complexes in erythrocytes. ATP and glycerol were added in all buffers to enable the weak interactions between the CP and its regulatory complexes to be maintained [74]. A negative control experiment associated with a label-free differential quantitative proteomic method was conducted to distinguish nonspecific interactors from putative PIPs. In this approach, wild-type endogenous complexes are targeted and can be purified from any human sample as starting material, such as tissues, biopsies, cultured cells and biological fluids. This approach is thus compatible with clinical studies. Successive antibody-based immunoprecipitation directed against two different 20S subunits has also been used to characterize new forms of intermediate-type 20S proteasomes containing either one (β_5i) or two (β_1i and β_5i) immunosubunits, and to show that these proteasomes were able to uniquely process some tumor antigens of clinical interest [14].

To conclude proteasome complexe purification, the previously described approaches represent complementary tools and might be employed depending on which proteasome complex is targeted: a specific 20S subtype, 20S proteasome associated with a specific regulator (like the 19S regulator) or the whole diversity of proteasome complexes. As far as the exhaustive identification of PIPs is concerned, targeting one subunit of the 19S or other 20S proteasome regulators (PA28s or PA200) for the AP might probably be more relevant to efficiently catch labile and dynamic interactors. Indeed, the major known PIPs, such as deubiquitinating enzymes, ubiquitin ligases and multiubiquitin chain-binding proteins, associate transiently or weakly with RPs and not with the 20S core. However, in this case, each RP will have to be targeted independently to identify the exhaustive interactome of all proteasome forms. Moreover, this strategy is not likely to catch some specific 20S proteasome partners such as assembly chaperones or other PIPs implicated in the ubiquitin-independent proteasome degradation (for reviews, see [80]). An alternative approach to purify all PIPs in a single purification scheme would be to target one subunit of the 20S CP while maintaining the labile bonds within entire complexes, so that long-distance linkages could be preserved. As demonstrated recently, in vivo cross-linking of protein–protein interactions can represent a good strategy to strengthen protein interactions [81] and has been used for the identification of proteasome interaction networks [57,74], while chemical cross-linking after proteasome purification has been used for structural studies [35,82,83]. Among several cross-linkers used in such approaches, formaldehyde fulfills several criteria for efficient stabilization of endogenous linkages between nucleic acids or proteins [81,84]. Formaldehyde is also a reversible cross-linker that facilitates MS identification of proteins. Formaldehyde is introduced very early in the proteomic workflow, so that real biological events can be frozen. However, we experienced a detrimental effects of formaldehyde for the recovery of high-molecular-weight PIPs, possibly owing to their precipitation/insolubility during the purification steps [74]. Therefore, both native and cross-linking approaches might be necessary to increase the number of identified proteins. Moreover, careful experimental design with optimization of formaldehyde concentration and inclusion of proper controls is needed to discriminate true interactors from contaminants. Associated with the purification procedure targeting all existing proteasome complexes through the α2 subunit of 20S proteasomes described by our group, in vivo formaldehyde cross-linking has proven to significantly increase the number of PIPs purified compared with a strategy without any cross-linking, while carrying out a rapid and single-step AP experiment and preserving the proteolytic activity of proteasomes [74]. Although the proteasome forms obtained at the end of the purification process are not separated from each other, this strategy could constitute an interesting tool in clinical studies intending to find a possible relationship between the global proteasome profile and a particular cellular state or pathology. It could constitute an interesting approach to improve proteasome complex recovery from tissues, but careful experimental design with optimization of formaldehyde concentration and inclusion of proper controls should be added in order to discriminate true interactors from contaminants.

Proteomic approaches for the study of proteasomes

As previously mentioned in the introduction, proteomic strategies are particularly suited for the study of protein complexes (see reviews [37,38,85,86]). Beyond the identification of the complex interactants, they enable the characterization of PTMs on complex subunits and possibly on partners, PTMs being recognized to tightly regulate protein–protein interactions, and their dynamics upon cellular stimulation [87–89]. More recently, quantitative proteomics has become a valued tool for the analysis of further protein complexes, allowing the identification of transient and labile interactants (for reviews, see [39,52,90]), the measurement of protein complex dynamics in terms of composition and subcellular localization and the determination of protein complex stoichiometry.

Proteomics for the characterization of proteasome subunits

The high-resolving separation power of 2DE associated with MS detection is particularly suited to the analysis of 20S proteasomes because they are easily soluble in common urea buffers, usually employed for the first-dimension separation. In the first reference map of 20S proteasome from human erythrocytes [78], more than 30 spots could be unambiguously detected on the 2DE gel and identified using MS, although only 14–17 distinct subunits were expected. MS results confirmed the proteolytic processing at the N-terminus of β1, β2 and β5 catalytic subunits, thus liberating the threonine residue that is essential for the three catalytic activities of the CP [91]. Numerous other studies have been undertaken and similar 2DE patterns of purified 20S proteasomes from various species, tissues or cell types could be obtained (reviewed in [13,40]), confirming that the 20S proteasome presents a high structural heterogeneity, likely in all types of cells.

19S proteasome subunits exhibit broader isoelectric point (pI) and molecular weight ranges (∼4.7–8.5 and 30,000–106,000 Da, respectively) and, as experienced recently, 2DE does not usually permit the detection of all 19S subunits, possibly owing to stability problems [76,92–95]. Although several studies using 2DE have revealed new PIPs, such as Adrm1 (an orthologue of Rpn13 in yeast) [94] and thioredoxin-related protein 32 (TRP32), a PIP bound to the 19S regulator through Rpn11 [96], shotgun approaches (1D electrophoresis [1DE]–liquid chromatography [LC]–MS/MS or 2D–LC–MS/MS) are more prone to identify proteasome subunits and PIPs [42,57,66–68,73,74,97].

Proteasome subunits exhibit numerous PTMs that might affect the activity of the proteasome but also its stability/assembly, plasticity, subcellular localization and interaction with other molecules [40]. Major limitations for the identification of PTMs on proteins reside in the fact that they are often substoichiometric and dynamic. Recently developed sophisticated strategies, in addition to new MS-based technologies, have enabled progress on the characterization of PTMs on proteasomes. Table 1 summarizes up-to-date N-terminal processing and endogenous phosphorylation sites identified and localized by MS/MS on human, mouse and yeast proteasomes.

Most 20S proteasome subunits exist as two to four different forms, mostly of different pIs, but also some of different molecular weights, which are separated by 2DE. Standard MALDI-TOF MS/MS or LC–electrospray ionization–MS/MS strategies have permitted explanation of such heterogeneity in some cases only. For example, the phosphorylation of α7 at Ser250 can explain the observed shift in pI for two spots corresponding to this subunit [78]. However, although similar strategies have revealed N-acetylations on several 20S proteasome subunits [17,78,98–100], they have failed to resolve the structural differences implied by the multiple spots observed for most subunits.

A recently described combination of top-down and bottom-up proteomic approaches permitted a deep characterization of the 20S proteasome subunits and their different forms from human erythrocytes [100]. The top-down analysis of each subunit after separation by 2DE and passive elution from the gel could give the experimental mass of each form with a mass accuracy of approximately 0.01%. Comparison with the theoretical mass could then give an indication of a possible PTM occurring on the protein. For example, an increment of mass of 42 and 80 Da is likely to be correlated to an N-acetylation and a phosphorylation, respectively, as obtained in this study for many subunits. This type of MS-based study is particularly powerful since it permits the choice of the most adapted bottom-up strategy to further check the suspected PTM. For instance, if a phosphorylation is suspected, one can decide to use an analysis workflow specifically dedicated to phosphopeptides [101,102], such as immobilized metal ion chromatography (IMAC) or TiO₂ enrichment, to enhance the detection sensitivity by the bottom-up experiment. The suspicion of a glycosylation will direct toward other types of enrichment strategies [103,104], such as purification with lectins.

Over the past decade, several sensitive post-electrophoretic stains have been developed as other interesting tools for the specific detection of PTMs, such as phosphorylation, glycosylation and oxidation [105–108]. Importantly, the newly developed dyes are compatible with further MS and can be used before and as a complementary approach to the use of general protein stains dedicated to the complete detection of all proteins in the gel [109]. Workflows associating high-resolution 2DE separation to specific enrichment strategies and multiplex detection systems, for example fluorescent PTM-sensitive dyes, specific chemical-derivation reagents, antibodies and MS, were implemented for the characterization of PTMs including N-acetylation, glycosylation, phosphorylation, oxidation, and nitrosylation on purified proteasomes from murine heart [110] and yeast [93]. Many proteasome 20S and 19S subunits responded positively to at least one of these different detection systems. MS experiments conducted could not confirm all of these. This may be explained by an incomplete protein sequence coverage hampering the detection of some modifications, but possibly also owing to false-positive detections. An additional difficulty of confident PTM confirmation by MS may arise from the use of collision-induced dissociation (CID), which remains the most common ion activation technique employed in laboratories devoted to proteomics today. In CID, activation of the selected ions occurs by collision(s) with neutral gas molecules in a collision cell. For the analysis of PTMs such as phosphorylations and glycosylations, this process typically results in poor informative phosphopeptides and glycopeptides MS/MS spectra, dominated by a neutral loss of the phosphate and glycoside group, respectively. Fragmentation techniques developed more recently, namely electron capture dissociation (ECD) and electron transfer dissociation (ETD), result in more confident identification of peptides modified with a labile moiety (for more details on these techniques, refer to the recent reviews [111–113]). Alternating both fragmentation techniques, ETD and CID, increases the amount of information derived from peptide fragmentation, and the new generation of high-resolution, sensitive and fast-sequencing LTQ Orbitrap mass spectrometers are equipped with both fragmentation modes [114,115]. Therefore, CID/ETD fragmentations, in association with protein fractionation, 2DE separation and/or phosphopeptides enrichment techniques, enabled the identification and assignment of new phosphorylation sites on proteasome complexes from yeast [116] or mammalian tissues (Table 1)[23]. However, none of the five 19S subunits and nine 20S subunits that had been previously determined to be potentially O-Glc-NAc modified in Drosophila melanogaster using immunological and lectin-binding methods [117] could be confirmed by the latest dedicated enrichment and CID/ETD fragmentation methods [118], showing how challenging the MS determination of glycosylation sites still remains. Future developments to improve dedicated tools for the enrichment of specific glycosylated proteins therefore seem needed.

Ubiquitin attachment sites are probably easier to identify because trypsin digestion results in the cleavage at the C-terminal arginine residue in ubiquitin, leaving a characteristic and stable GG or LRGG tag on the lysine residue in the peptide that was covalently linked to ubiquitin [119,120]. Using this signature tryptic footprint, several 20S and 19S proteasome subunits were detected as ubiquitinated in various eukaryotes species after immunopurification of proteasomes [73,121] or after large-scale purification of all cellular ubiquitinated proteins by immunoaffinity chromatographic procedures [122,123]. These results are in agreement with a previous report [124] suggesting that there is an autoregulation of active proteasome levels in cells by autodegradation of nonfunctional 20S proteasome [73,122]. Alternatively, ubiquitination could impart new functions to specific subunits, as demonstrated very recently for Rpn10, one of the ubiquitin receptors of the 26S proteasome [121]. Monoubiquitination of Rpn10 strongly inhibits its UIM-mediated interaction with polyubiquitinated proteins, thereby downregulating substrate recruitment to the proteasome. This event is notably decreased in stress conditions, presumably for more efficient protein degradation.

The recent improvements in the sensitivity of detection using modern mass spectrometers associated with sophisticated enrichment techniques offer an unprecedented breakthrough for the identification of PTMs on proteins. The forthcoming challenges will be to assign a biological function to each of these modifications. As far as proteasomes are concerned, results from the literature highlight a high variability in the PTM pattern between tissues, species and pathophysiological status, probably linked to a regulated proteasome activity, but this has to be confirmed in most cases.

The most relevant approaches will probably consist of linking an identified PTM with a particular biological context. In most cases, the detection system, such as ³²P radiolabeling, PTM-directed antibodies, PTM-sensitive dyes or pI/PM shifts on 2DE gels, will not give any information about the precise site of modification. Therefore, the knowledge of potential modification sites reported on a proteasome subunit by global proteomics could help to design targeted experiments, such as targeted MS/MS approaches or directed mutagenesis, to confirm the biological role of a given site of modification. Recently, an observed pI shift of several 20S proteasome subunits from Alzheimer’s disease (AD) brain samples using 2DE pointed out possible variations in PTMs on proteasomes [125]. The previously reported phosphorylation and N-acetylation sites on the α7 subunit were then further investigated using MS/MS as this subunit was one for which a significant pI shift had been observed. Finally moderate N-terminal acetylation and dephosphorylation of subunit α7 were demonstrated in the brain of AD patients. Similarly, 2DE separation of 20S proteasome forms in Jurkat T cells showed a decrease in the phosphorylation status of α7 at the Ser²⁵⁰ site, as a result of 5-fluorouracil-induced apoptosis [27], and this is in accordance with previous work showing a change in phosphorylation of 20S proteasome subunits after induction of apoptosis by doxorubicin in proerythroleukemic K562 cells [126]. The biological incidence of the dephosphorylation of α7, even moderate, might well have an impact on 26S proteasome stability. In order to characterize variations of PTM on proteasome subunits, future developments will require the association of PTM-specific workflows with quantitative proteomic strategies specifically developed for the quantitative analysis of protein complexes, including concurrent identification of interactors and their state and stoichiometry of modification [88]. A recent survey associating proteasome purification, TiO₂ enrichment and isobaric tags for relative and absolute quantitation (iTRAQ) labeling demonstrated that the already described phosphorylation of Rpn2 subunit on Thr²⁷³(Table 1) induces the inhibition of proteasome activity in the context of osmotic stress [127]. Sensitive multiple reaction monitoring (MRM) methods will also constitute promising targeted quantitative approaches for the determination of the stoichiometry of PTM-modified sites, as recently described for phosphorylation [128] and ubiquitination [129–131].

Functional analysis of proteasomes by quantitative proteomics

Quantitative proteomics has evolved as an valuable tool for characterization of the biological function of proteins (for a review, see [132]). As reviewed in the next section, recently developed strategies have permitted significant progression in our knowledge of proteasome biology.

Quantitative analysis of the variations in 20S proteasomes subunit composition

2DE associated with image analysis using dedicated software is an attractive technique for the quantification of 20S proteasomes because it enables the separation of all α and β subunits including their numerous forms and their quantification in a single experiment. As already reviewed [41] and more recently published [28,133], it has been used in numerous surveys for the biochemical characterization of 20S proteasomes in different contexts. However, numerous challenges such as spot matching, 2DE gel normalization, and lack of reproducibility of separation or staining have to be overcome for successful quantification. The recently developed DIGE technology has become the method of choice for the detection of differences by image analysis after protein separation by 2DE (see recent reviews [134,135]). In DIGE-based proteomics, the experimental and control samples are derivatized with different fluorophores and the compared sets of proteomes comigrate in a single run, minimizing gel-to-gel variations. The strategy also includes an internal standard, facilitating the normalization step. Recently, this technique enabled a similar 20S proteasome high-resolution 2D pattern as the one previously obtained using commassie blue staining [78] to be obtained, with multiple forms for most proteasomal subunits [125]. pI shifts and changes in relative abundance between healthy and AD brain samples could be observed for several proteasome subunits, possibly due to PTM variations on these subunits. Recently, a study by DIGE of global proteome variation in HL-60 cells, treated with methotrexate, an inhibitor of nucleotide biosynthesis used for the treatment of cancers, pointed out several subunits of the 26S proteasome. Their downregulation might explain the accumulation of ubiquitinated proteins and promotes apoptosis [136]. The catalytic subunit β2 was also identified as one protein associated with multidrug resistance in LoVo human colon cancer cells by DIGE, and this regulation was further validated by immunofluorescence and Western blotting [137]. Numerous other recent global proteome differential analyses using 2DE quantification and MS identification report variations in proteasome subunit abundance. However, measurement of the variations observed using complementary approaches remains necessary to validate the results.

Recently, MS-based quantification techniques have been successfully developed, in association or not with 2DE separation, for the identification of variations in 20S proteasome composition and abundance either in different cell types or in a pathophysiological context. Chemical labeling at the protein level by isotope-coded affinity tag (ICAT) [18,99] or by stable isotope labeling with amino acids in cell culture (SILAC) [27], and at the peptide level after protein digestion [24,138], or even label-free MS quantitation [22,138], have emerged as alternative techniques advantageously replacing the classical post-electrophoretic quantification by image analysis.

When several subunit forms are separated by 2DE prior to quantitation, the variations of each form can be distinguished [18,27,99]. Schmidt et al. recently described a SILAC-based quantitative study of 20S proteasomes separated by 2DE to better understand the observed decreased proteasome-mediated proteolysis of short-lived proteins in apoptotic cells [27]. Significant quantitative changes were observed on the 2DE gels for two forms of the α7 subunit, highlighting a decrease of phosphorylation status, further confirmed by quantitative MS analysis.

Liquid chromatography–MS analysis of isotope-labeled peptides directly after proteasome purification can only detect global variations of each subunit in most cases. However, this can be a sound and straightforward approach for the comparison of the relative abundance of catalytic subunits or PIPs. Raijmakers et al. successfully pointed out the differences in the composition between chromatographically purified bovine liver and spleen 20S core proteasome complexes [24]. In this study, online and on-column sequential derivatization of peptides, by dimethylation using cyanoborohydride and either regular or deuterated formaldehyde, was performed. The results obtained demonstrated that all three catalytic immunosubunits – β1i, β2i and β5i – as well as the PA28 α/β regulator of the 20S proteasome were more abundantly present in the spleen-derived proteasome preparation, as expected from previous observations [139]. In another survey, Gomes et al. compared purified 20S proteasomes from heart and liver using three different quantitative approaches, namely label-free quantitation, chemical labeling using ¹⁶O/¹⁸O post-digestion labeling and immunoblotting [138]. Again, a clear difference in inducible catalytic subunits, mainly β1i and β5i, was observed in the liver when compared with the heart, although the concentrations of proteasomes per cell/tissue are largely different. Recently, a broader study on heart, kidney, liver, lung, thymus and spleen tissues compared the relative abundance of all α and β subunits from purified 20S proteasomes [22]. As expected, no major change was observed on α1 to α7, β3, β4, β6 and β7 subunits (which are shared by the constitutive and the immunoproteasome); however, differences in the relative abundance of the immunosubunits and their constitutive counterparts could be precisely determined and confirmed the tissue-specific distribution of the immunoproteasome (more abundant in spleen and thymus lymphoid tissues). It also pointed out the existence of immunoproteasome hybrids wherein the subunits β1, β2 and β5 have only been partially replaced by their β1i, β2i or β5i counterparts, and the specific presence in the thymus of the β5t catalytic subunit [15,140].

Identification of specific PIPs & determination of their association stability by quantitative proteomics

Proteomic analysis can be very efficient in identifying interacting proteins. Improvements in MS sensitivity and acquisition speed allow the production of ever larger lists of identified proteins. These, however, may contain contaminating proteins that needs to be distinguished from true partners [141]. These nonspecific co-purified proteins can represent up to 95% of the total pool of proteins identified [142]. More stringent purification methods can be used to decrease the risk of false-positive interactants, but they also restrict the identification of the strongest partners. Modern quantitative proteomics can circumvent this problem by direct quantitative comparison of each identified protein in specific and control pull-downs. These strategies have been recently extensively reviewed [90,141] and essentially rely on isotopic labeling at the protein [143–145] or at the peptide level [88,146], or on label-free quantitative approaches [147,148]. Labeling methods at the protein level usually give a more accurate relative quantification because the label is introduced early in the workflow and labeled samples (usually two or three) are then mixed and analyzed in the same analytical run. By contrast, a label-free approach necessitates parallel sample treatment and analysis. Accurate quantification thus requires a high analytical reproducibility of the workflow and replicate analyses to be performed and handled using dedicated bioinformatics tools. The main advantage of label-free approaches is that they can be used to compare any type and number of samples. High confidence protein–protein interactions are revealed thanks to their high ratios and can be thereby discriminated from a great number of background proteins. These quantitative strategies have successfully been applied to greatly enlarge our current knowledge of the proteasome network. SILAC [57,97] and label-free [74] quantitative approaches have been described in association with AP strategies to characterize proteasome complexes and discriminate true PIPs from nonspecific co-purified proteins. In the classical SILAC strategy, the samples to be compared are mixed very early, at the cellular level, to minimize quantitation errors arising from separate sample handlings. Therefore, after cell lysis, the differentially labeled protein extracts containing either the tagged bait or the control are in contact for a certain time during the AP step. This permits dynamic exchange of the most transient interactors. The more dynamic the interaction is, the faster the exchange is. Therefore, only stable interactions can be detected with high ratios using this approach, termed purification after mixing-SILAC (PAM-SILAC). Lower ratios can be indicative of transient or labile linkages within the complexes and need further investigation using time-controlled PAM-SILAC [97]. An alternative strategy to detect both stable and the most dynamic PIPs was to mix the two protein extracts after purification (mixing after purification [MAP] strategy). Finally, a combination of both strategies is of great interest to obtain a wide interaction network, while gaining information about the nature of the interaction (stable or dynamic). Some well-known PIPs of mammalian proteasomes, such as Usp14 deubiquitinating enzyme or hHR23B multi-ubiqutin chain-binding protein, and even the Rpn13 subunit of 19S regulator, were classified as dynamic, while others were found to be stable PIPs, based on their SILAC ratios [97]. These approaches have been reviewed very recently [39].

Interestingly, a quantitative comparison between a native and a cross-linked protein extract can also give indication about the strength of the interaction of a protein within a complex, as demonstrated recently for PIPs using formaldehyde reticulation [74]. The strategy was to stabilize the most labile and/or dynamic interactions so that the in vitro immunoaffinity-isolated proteasome complex network could be enlarged. The benefit of formaldehyde treatment was evaluated by estimating the relative abundance of PIPs in purified samples from formaldehyde-treated and nontreated erythrocytes using label-free quantification. Interestingly, results indicated that the most significant protein linkage stabilization was observed for the most remote subunits from the α2 subunit of 20S proteasome, used as prey protein. Indeed, subunits from the 19S lid (Rpn3 and Rpn5-12) exhibited high formaldehyde-enrichment ratios compared with subunits from the 20S CP or subunits from the 19S base (Rpts and Rpn1-2). Other PIPs were also only recovered after formaldehyde treatment, which suggested loose interactions within the complex.

Quantitative analysis of the variation in PIPs to elucidate the roles of proteasomes in biological processes & in pathologies

Altered proteasome-mediated degradation is believed to be related to diverse pathologies such as cancer, neurodegenerative disorders, aging or muscle atrophy [7,149–152] through dysfunctions of cellular biological processes such as cell-cycle progression, apoptosis or DNA damage repair, or as a consequence of environmental stresses, such as oxidative stress [153]. Modulation of protein–protein interactions constitutes a new area of research for the development of new drugs and this strategy might be well adapted to proteasomes as increasing evidence demonstrates that some of these disorders might correlate with alterations of PIPs and proteasome regulatory complexes binding to the proteasome [31,66,76,154]. PIPs associate with the proteasome and have an important impact on its structure, assembly, stability and subcellular localization, as well as its substrate recognition and degradation efficiencies. Therefore, the specific modulation of the interaction between the proteasome and some PIPs could constitute new therapeutic strategies. As reviewed in the following section, comparative expression proteomics using modern MS-based quantitative approaches has proven to be of great interest to study dynamic variations of PIPs under normal or pathological contexts.

The proteasome tightly regulates major cellular processes by controlling the cellular pool of key regulatory short-lived proteins. In an effort to better understand how 26S proteasomes regulate cell cycle transitions, the dynamic variations of yeast PIPs along the cellular cycle progress were studied thanks to their SILAC ratio profiles by the QTAX strategy [155]. This enabled categorization of the PIPs and to evaluate whether these groups of proteins were enriched in specific cellular functions, biological pathways or complexes. Interestingly, this analysis showed that a group of PIPs only captured at G1 and S phases was enriched in proteins involved in the yeast mating pheromone-response pathway that induces cell-cycle arrest.

Another very important function of proteasomes is to ensure the quality control of the cell by degrading abnormal and nonfunctional proteins generated under normal and stress conditions [4,16]. Oxidative stress is associated with important pathophysiological events in a variety of diseases, such as neurodegenerative diseases [5,156] or in aging [157,158]. Proteasomes seem to play an important role in preventing the increase of intracellular oxidative stress through the degradation of defective ribosomal products [16] and oxidized proteins, both in the cytosol and the nucleus [159], but also through the regulation of pro-apoptotic proteins level [160]. Inhibition of the proteasome may contribute to the toxicity associated with oxidative stress [161–163]. The mechanism responsible for the decrease in proteasome activity caused by excessive oxidative stress is still not clear. The formation of stable 4-hydroxynonenal adducts on proteins, thereby forming crosslinks that are difficult to be degraded, but also on proteasomes themselves, was reported to inhibit proteasome activity [159,164,165]. Phosphorylation and S-glutathionylation have also been reported on several 20S or 19S proteasome subunits upon exposition to diverse oxidative stresses [165–169]. Another reason for proteasome impairment associated with oxidative stress might be an alteration of the association of PIPs and regulatory complexes with the proteasome, as suggested by the following surveys. An impaired assembly of the 26S proteasome and a decreased expression of Rpn11, the major DUB 19S subunit, were found to be associated with aging [157]. H₂O₂-induced S-glutathionylation of Rpn2 might induce aberrant interaction of the Rpn1/Rpn2 complex with other proteasomal subunits [168]. Cells lacking arsenite-inducible RNA-associated protein, an arsenic-inducible PIP, accumulate polyubiquitinated substrates when exposed to oxidative stress caused by arsenite [55]. Recently, arsenite-induced changes in proteasome composition were investigated by a SILAC-based quantitative analysis in mammalian cells [76]. TRP32, a PIP with thiol reductase activity primarily detected in 2DE-separated 26S proteasomes [96], was found to be an arsenite-regulated protein by SILAC. The 26S-associated TRP32 level was shown to decrease after oxidative stress exposure of mammalian cells, but the functional significance of this variation could not be elucidated. In another recent survey, alterations in proteasome composition were investigated in the context of alcoholic liver injury [66], where significant inhibition of proteasome activity has been reported. Changes in the amount of PIPs upon chronic ethanol feeding were determined by ICAT and label-free quantitative approaches. PA28α/β proteasome activator and Ecm29, a known PIP also found to be responsible for 26S proteasome disassembly under oxidative stress [170], and three major proteasome-associated deubiquitinases, were significantly decreased in the ethanol-treated purified proteasomes. The lack of these important PIPs within proteasome complexes might be responsible for the observed dramatic decrease of proteasome activity and accumulation of polyubiquitinated proteins.

Quantitative proteomics to elucidate spatial distribution of proteasomes

Proteasomes have been detected in numerous cellular compartments, including the cytosol, nucleus, ER, Golgi apparatus [41] and, to a lesser extent, associated with membranes [171]. Most studies relied on subcellular fractionation followed by immunoblotting or, on in situ observation using immunohistochemistry, immunofluorescence or confocal microscopies [41]. These targeted approaches necessitate the availability of specific antibodies against the proteins of interest.

Quantitative proteomics also constitute a precious tool to assign proteins to their specific organelle or subcellular compartments [172–175] or to profile dynamic changes in the subcellular localization of proteins in cells under different physiological states [19,21]. They permit a global view of the cellular distribution of proteins to be obtained because they are essentially not targeted approaches and can therefore localize previously uncharacterized proteins. Moreover, they can permit the localization of specific protein forms, such as splicing isoforms or forms with PTM, to be assigned, which is of great help for more in-depth biological interpretation. A prerequisite is to optimize an efficient fractionation workflow so that highly pure separated subcellular compartments can be obtained. In a recent study, Boisvert et al. described a successful SILAC-based proteomic strategy, termed ‘spatial proteomics’, to establish a quantitative map of the relative subcellular localization of each protein, and to characterize their dynamic changes induced by DNA damage [19]. Interestingly, while at steady state, the 26S proteasome subunits clustered predominantly in the cytoplasmic fraction, DNA damage was found to induce the relocation of the 20S complex and of all its regulators (PA28s, 19S) in the nucleus. The mechanisms explaining how proteasomes are recruited to the nucleus are not yet known but confirm previous reports linking the function of proteasomes to the DNA repair pathway [176]. Quantitative proteomics on purified proteasomes rather than on whole cellular lysates might permit a more precise distribution of these complexes throughout the cell to be obtained. Indeed, each proteasomal subunit might be found in mature proteasomes but also in premature proteasome forms [177,178] and as free subunits in the cell, where they could have an additional extraproteasomal function, as recently proposed for Rpn10 [179,180].

Quantitative analysis of ubiquitin chain structures and of polyubiquitinated substrates to better understand proteasome-related ubiquitination & its function

Although an increasing number of examples show that structured proteins can be processed by proteasomes independent of ubiquitination (see review [80]), it is widely acknowledged that ubiquitin conjugation constitutes the major signal for 26S proteasome-mediated degradation. Protein ubiquitination is accomplished through a cascade of three consecutive enzymatic reactions involving activating enzymes (E1), conjugating enzymes (E2) and ligases (E3) [2]. It results in the formation of an isopeptide bond between the ε-amino group of a lysine residue within the substrate and the carboxyl group of the C-terminal glycine residue of ubiquitin. Because ubiquitin displays seven accessible lysine residues (K6, K11, K27, K29, K33, K48 and K63) and an N-terminal amino group, which all constitute a possible anchor for polyubiquitination, the diversity of polyubiquitin (polyUb) chains is high [119,181]. Protein ubiquitination is central to many cellular processes [182,183] and the fate of a modified protein depends mainly on the number of ubiquitin moieties attached to it and on the chain topology. Mechanisms of recognition and processing of ubiquitin–protein conjugates by the proteasome have been reviewed recently [184]. PolyUb chains on substrates allow both addressing and binding to the proteasome, the main requirements believed for efficient proteolysis being the particular topology of K48-linked polyUb chains [185] and a minimal chain length of four ubiquitin moieties [186]. Monoubiquitination or polyubiquitination through K63 are generally thought to act in a range of other proteasome-independent cellular processes (for a review, see [187]). However, recent surveys performed, for many, with the aid of quantitative proteomics, raised the question of more complicated mechanisms for substrate recognition and degradation efficiencies by proteasomes complexes. MS is a powerful tool for the study of ubiquitin chain topology as each type of ubiquitin linkage gives a specific tryptic digest: monoubiquitination produces characteristic unbranched ubiquitin tryptic peptides, whereas polyubiquitination generates modified tryptic peptides of ubiquitin where a tag of two G residues is attached to the K which is involved in the chain linkage.

A subset of proteins with K11 linkages have been reported to be degraded by the proteasome [130,188,189]. Moreover, the K11 GG-tagged tryptic peptide of ubiquitin, detected in tryptic digests of whole cell lysates but also of proteasome-bound proteins, is increased, together with its K48 counterpart, upon MG132 inhibition of the proteasome [131] or in yeast strains with mutation in proteasome components [129]. Similarly, all non-K63 polyUb linkages accumulate upon proteasome inhibition [129,190]. Whether K63 polyUb linkages have the ability to target substrates for proteasomal degradation seems to be more a matter of debate. MS-based quantitative analyses of K63 linkages levels upon proteasome inhibition in Saccharomyces cerevisiae or mammalian cells show either a significant increase, suggesting the involvement of K63 as a targeting signal for the 26S proteasome [190–192], no significant changes [129] or a delayed accumulation for long time inhibition [131], possibly in response to stress caused by proteasome inhibition rather than on account of a proteolytic function of K63 linkages. In vitro, however, K63 polyUb chains bind the 26S proteasome as efficiently as K48 chains [131], and are also able to target the degradation of several proteins [130,191,193,194]. Quantitative MS analysis of the polyUb chains K48/K63 ratio showed a higher abundance of K48 chains bound to the proteasome compared with the ratio observed in the whole lysate [131]. Proteasome-bound deubiquitinating activities, which more efficiently degrade long K63 chains than their K48 counterparts, might account for such discrepancies and enable the escape of K63 polyUb substrates from proteasomal degradation [131].

Therefore, even if some specific polyUb linkages are required for proteolysis of particular substrates, most seem competent in addressing proteins to proteasomes, although with different efficiencies. This capacity might constitute a cellular means to more efficiently respond to diverse stresses by an increased and massive proteasome-mediated degradation of proteins. The major limitation of the quantitative proteomic strategies developed to date resides in the difficulty to discriminate a mixture of different homogeneous ubiquitin–ubiquitin linkages from branched mixed-linkage chain topologies. A tryptic peptide of ubiquitin containing two Ub footprints at K29 and K33 could, however, be MS-detected in vivo in yeast [119,195] and in plant [196], suggesting highly complex chain topologies, independently of the species.

Moreover, even if MS is capable of determining global ubiquitin chain topology, it cannot discriminate a free polyUb chain from a protein-linked one. The only way to identify an ubiquitinated protein in a whole cellular lysate is the detection of the mass shift of 114.04 Da, induced by the GG signature on the tryptic peptide triggering the ubiquitin moiety, which is plausible but quite unlikely. Therefore, ubiquitinated proteomes are better studied using affinity-based enrichment methodologies [119,123,190,196–198] and, in association with quantitative proteomics strategies, changes in ubiquitination profiles can be used to better understand ubiquitination and its function. These approaches permitted the identification of hundreds of putative ubiquitinated proteins, but the Ub attachment sites were resolved for only a small subset of these proteins [190,196]. Little sequence consensus was observed, at least in plant [196], and the discovery of new ubiquination sites might therefore rely on the development of ever-sensitive and highly dynamic range MS technologies.

Quantitative analyses comparing the ubiquitinated proteomes of untreated cells and cells treated with proteasome inhibitors constitute interesting tools to identify putative proteasome substrates [190]. As described recently, quantitative proteomics can also help to identify a subset of proteasome-targeted proteins that are substrates of a specific E3 ubiquitin ligase [199].

Role of MS & proteomics for the study of proteasome complexes as therapeutic targets

Proteasome inhibitors are now used in clinics for the treatment of some types of cancer but the cellular pathways that are affected at the molecular level are still not completely defined. Global quantitative proteomic approaches can play an important role in elucidating these pathways. Here, we review such studies that have been undertaken so far, and will also highlight some recent studies that propose consideration of the proteasome as a predictive biomarker for cancer therapies and other diseases, and discuss the growing role of MS in the detection of biomarkers.

Global quantitative proteomic approaches represent an attractive way of deciphering the cellular pathways affected by proteasome inhibition. They can also be valuable to understand resistance mechanisms associated with drug treatment and to help define new therapeutic targets to be used in combination with proteasome inhibition. However, several challenges have to be faced using this approach. A first difficulty comes from the fact that proteasome inhibition will affect many processes and lead to the accumulation of many proteins. Thus, the inhibition time point chosen for the study may be an important parameter. A second difficulty is that the pathways already known to be affected and to play a role in the apoptotic cascade may involve low-abundance proteins, such as, for example, the NF-κB protein [200]. This may explain why few studies have attempted such experiments up to now using various cellular models and several proteasome inhibitors. These studies were based on 2DE separation and differential analysis followed by MALDI–MS or LC–MS/MS protein identification to determine the proteins affected by proteasome inhibition in rat cardiac myocytes [201], in human umbilical vein endothelial cells [202], in colorectal cell lines [203], and in mantle cell lymphoma cell lines [204]. These studies were able to separate more than 1000 protein spots on average and to identify three up to 75 varying proteins. They demonstrate that proteins from the heat-shock protein (HSP) family (HSP70 in most cases) were increased for all cell lines and regardless of the proteasome inhibitor used. Other proteins were then found to be more specific from the cell line studied. Interestingly, sensitive mantle cell lymphoma cell lines differed from resistant ones after inhibition with 25 nM bortezomib and exhibited additional alterations in protein levels of energy metabolism, RNA and transcriptional regulation, and cell division [204]. This opens new perspectives to explore combination therapies targeting complementary pathways for the treatment of lymphoma.

In a very recent study, a global quantitative proteomic approach was used to characterize aggregating proteins in human neuroblastoma cells after proteasome inhibition [205]. The authors performed metabolic labeling using SILAC and identified the proteins affected by proteasome inhibition using MG132 treatment, which is a widely used proteasome inhibitor even though it may inhibit other proteases in the cell. This quantitative strategy was combined with protein fractionation using a sucrose gradient to focus on proteins accumulating in aggregates, which are known to be formed after proteasome inhibition. Moreover, the use of a recent highly sensitive and high acquisition-speed mass spectrometer such as the LTQ Orbitrap instrument provided an efficient tool to identify less abundant proteins. This approach allowed more than 500 proteins to be identified as enriched in aggregates after MG132 treatment, corresponding to 74% of the total number of identified proteins. The identified proteins suggest that aggregates formed upon proteasome inhibition are in part similar to other protein aggregates found in neurodegenerative diseases, but also contain additional proteins associated with the ubiquitin system and also demonstrate an increased level of HSP70.

Another aspect of the analysis of proteasome as a therapeutic target lies in its detection as a diagnostic and/or prognostic biomarker in body fluids. This is supported by the fact that an elevated proteasome amount has been found in several tumor cells [206]. It was then hypothesized that this elevated concentration could also be found in fluids in contact with the tumor cells. Even though the origin of the extracellular proteasome detected in body fluids is still enigmatic [171], it was proposed that it could reflect a pathological state of patients and that it could represent a reliable biomarker for several diseases. Extracellular proteasome concentration has mainly been measured in bronchoalveolar lavage fluid and in blood plasma or serum, where it is referred to as circulating proteasome, using ELISA (for a review, see [171]). In recent years an increasing number of clinical studies have been conducted. They all demonstrate an elevated concentration of 20S proteasome in body fluids of patients with a wide range of pathologies, such as burn and inhalation injury [207,208], acute respiratory distress syndrome [209], melanoma [210,211], malignant transformation of liver cirrhosis [212], sepsis [213] and epithelial ovarian cancer [214]. It can thus be concluded that extracellular and circulating 20S proteasomes, the concentration of which can easily be measured from body fluids, constitute a new class of biomarkers of several pathologies. In some cases, it could efficiently be used to assess efficacy of treatment.

However, the precise composition of circulating 20S complexes is not known when using an ELISA, and various proteasome subtype profiles could exist for an equivalent 20S proteasome amount. In a pathological context one could then imagine using these profiles of a tissue or plasma sample from patients as a more precise clinical tool for diagnosis or prognosis, which could lead to more targeted therapies. Indeed, development of inhibitors targeting a specific proteasome subtype could constitute a sound strategy to reduce side effects of nonspecific inhibitors [215,216]. Immunoproteasome-specific inhibitors(IPSI)-001, the lead compound of a series of recently developed IPSIs, demonstrated a more than 100-fold increased selectivity towards the immunoproteasome. Therefore, tissues expressing predominantly standard proteasomes should not be targeted by this inhibitor – unlike the current generation of proteasome inhibitors, such as bortezomib, which is proteasome specific but not 20S proteasome subtype specific. IPSIs were shown to constitute as valuable alternative drugs for the treatment of multiple myeloma patients who were exhibiting resistance to conventional drugs and nonspecific proteasome inhibitors [217]. Other pathologies where elevated expression of immunoproteasome has been reported in various tissues – such as autoimmune diseases [218], inflammatory bowel disease [29,219,220], major hepatic diseases [221] and degenerative diseases [222] – could also benefit from this new class of proteasome-targeted inhibitors. In the case of inflammatory bowel disease patients, it has recently been demonstrated that the subunit pattern of 20S proteasome from surgical tissue specimens was suitable to differentiate Crohn’s disease and ulcerative colitis: two forms of this disease [28]. It is therefore important to study further proteasome profiles and to develop robust analytical tools to accurately and reproducibly measure proteasome subunit compositions.

Targeted MS techniques can play a pivotal role in the measurement of all existing subunits of 20S proteasome complexes, including immunosubunits and post-translationaly modified subunits, from patient body fluids. In the studies performed so far, antibodies have mainly been used to measure a few subunits of the standard and the immunoproteasome complexes. An initial attempt to characterize all subunits of circulating proteasome has been reported using 2DE of immunopurified 20S complexes [211], but the low amount of 20S proteasome available and the protein composition of plasma samples were important limiting factors to the success of this approach. Thus, only the composition of proteasome from plasma of patients with stage IV melanoma could be determined and showed a mixture of standard proteasome and immunoproteasome complexes. Targeted MS has been shown to be a sensitive and specific way to measure biomarkers in body fluids [223]. It has limits of detection comparable with those of ELISA for some proteins, as demonstrated for prostate-specific antigen [224,225]. It is based on the definition of proteotypic peptides generated after trypsin digestion, which are specific of the protein to be measured. Although targeted MS analysis does not measure intact proteins, peptides can be carefully chosen when necessary to cover the N- and C-termini of a protein and to detect modified peptides. These peptides are then analyzed by MS using a selected reaction monitoring acquisition mode, which consists of selecting a defined peptide, fragmenting it and specifically analyzing the most abundant fragments formed (Figure 3; for a review, see [226]). The promising potential of this MS-based approach lies in its sensitivity and high selectivity. Even more importantly, it is fast and it can be multiplexed to measure several tens of proteins in a single LC–MS/MS run [227]. These multiple single reaction monitoring transitions correspond to the MRM mode. Very recent applications have been performed using biological fluids, mainly serum [225,228] and feces [229] from patients, for the detection of several biomarkers in cancer. It is thus expected that targeted proteomics using MRM is going to play a growing role in biomarker validation in the near future.

Expert commentary

Newly developed AP approaches now permit the entire proteasome network to be obtained in a simple procedure, and some of these strategies are compatible with clinical studies, allowing the study of proteasomes in real pathological contexts. Meanwhile, the latest improvements in protein/peptide separation, enrichment and labeling techniques in MS and in bioinformatics enabled proteomics to tackle functional studies. These advances contributed to significant progress in proteasome biology knowledge. In particular, the understanding of the variations of proteasome quantities, subunit composition, subunit PTM modification or subcellular localization in response to various stimuli or in disease states could be improved. Further studies are needed to fully characterize the dynamics of proteasome diversity, and the major challenge will probably be to understand the interplay between the structure of a specific proteasome form and its cellular function in a given physiological or pathological context. Future development in targeted quantitative proteomics could then help to validate such a relationship in a large cohort of patient samples.

Five-year view

Proteomics has evolved as a valuable tool for the study of proteasomes for both fundamental and clinical purposes. Indeed, while 2DE separation of purified proteasome complexes and global proteomic approaches can play an important role in understanding the structure/function relationship of proteasome subtypes and in elucidating the cellular pathways affected during proteasome inhibition, targeted strategies such as MRM are on the way to providing promising therapeutic approaches for the quantitative analysis of proteasomes in biological fluids or for the detection of pathology-related UPS components. Indeed, sensitive and specific analytical mass spectrometric strategies will be required to access the quantification of each low-abundance proteasome form in complex patient samples. Nevertheless, recent progress in the field of structural biology has been made using MS for the analysis of intact protein complexes. Applied to proteasome complexes, these approaches have contributed, for example, to the determination of the stoichiometry and the architecture of the 19S regulatory complex. One can foresee that this field, which brings complementary information, will combine efficiently with proteomics data in the near future to provide a more comprehensive view of proteasome complexes in the cell.

Table 1. Summary of the N-terminal processing and endogenous phosphorylation sites identified and localized by MS/MS on human, mouse and yeast proteasomes.

Name	Gene name (human)	Protein name (human)	Protein name (yeast)	PTM (amino acid, ^species)^†‡
20S proteasome subunits
α-type 1	PSMA1	α6 [230]	α6_sc [231]	Ac(M1^h) [98]; p(S14^y,T16^y) [232,233]
α-type 2	PSMA2	α2 [230]	α2_sc [231]	Ac(A2^h,m) [17,100]; p(S198^m) [23]; p(S13,53^y) [232]
α-type 3	PSMA3	α7 [230]	α7_sc [231]	Ac(S2^h,m) [17,74,98,100]^§; Fr(S2^h) [74]; p(S249^h) [98,100]; p(S250^m) [23,234]; p(S13^m,T186^m) [23]; p(S258, 263, 264^y) [235]; p(T278, 279^y) [232]
α-type 4	PSMA4	α3 [230]	α3_sc [231]	Ac(S2^h) [100]; p(S75^h) [98]; p(S33,75,173^m) [23]
α-type 5	PSMA5	α5 [230]	α5_sc [231]	Ac(M1^h,m) [17,74,98,100]; p(S134^m,T230^m) [23] p(S35,55,56,87,251^y) [93,232,233,236]
α-type 6	PSMA6	α1 [230]	α1_sc [231]	Ac(S2^h,y) [74,93,100]; p(S237^y) [232]
α-type 7	PSMA7	α4 [230]	α4_sc [231]	Ac(S2^h,m) [17,98,100]; p(T97^m) [23]; p(T60^y) [93,116]; p(S69^y) [93,232]
β-type 1	PSMB1	β6 [230]	β6_sc [231]	Tr(R29^h) [98,100]; p(S167^m) [23]; p(S53,209^y) [116,232]
β-type 2	PSMB2	β4 [230]	β4_sc [231]	Ac(M1^h,m) [17,74,98,100]; p(S39^m) [23]; p(S76,82^y) [232]
β-type 3	PSMB3	β3 [230]	β3_sc [231]	Ac(S2^h,m) [17,74,98,100]; p(S31,33^y) [232,236]
β-type 4	PSMB4	β7 [230]	β7_sc [231]	Tr(T46^h,m) [17,100]; p(S93^m) [23]
β-type 5	PSMB5	β5 [230]	β5_sc [231]	Tr(T60^h) [98,100]; p(T48^m) [23]; p(S51^y) [232]
β-type 6	PSMB6	β1 [230]	β1_sc [231]	Tr(T35^h,m) [17,98,100]; p(S157^m) [23]; p(T91^y) [232]
β-type 7	PSMB7	β2 [230]	β2_sc [231]	Tr(T44^h,m) [17,98,100]; p(T273^m) [23]
β-type 8	PSMB8	β5i [230]
β-type 9	PSMB9	β1i [230]		Tr(T21^m) [17]
β-type 10	PSMB10	β2i [230]
19S proteasome subunits
Regulatory subunit 4	PSMC1	S4 [237]	Rpt2 [238]	Myr(G2^h) [98]; p(S176,180,182,356^y;T178,357^y; Y181) [232,233,236]
Regulatory subunit 7	PSMC2	S7 [237]	Rpt1 [238]	Fr(P2^h,m) [74,93,98]; p(S164,231^y) [93,232,236]
Regulatory subunit 6A	PSMC3	S6a [237]	Rpt5 [238]	Ac(M1^h) [74,98]; p(S9^h) [98]; p(S181^y;T11,27,369,370^y) [93,232,236]
Regulatory subunit 6B	PSMC4	S6b [237]	Rpt3 [238]	Ac(M1^h) [74,98]; p(T8^y) [93,232]
Regulatory subunit 8	PSMC5	S8 [237]	Rpt6 [238]	Ac(A2^h) [74,98]; p(S120^h) [239]
Regulatory subunit 10B	PSMC6	S10b [237]	Rpt4 [238]	Ac(A2^h) [98]; p(S292^y) [233,236]
Non-ATPase regulatory subunit 1	PSMD1	S1 [237]	Rpn2 [238]	Ac(M1^h) [74,98]^§; Fr(M1^h) [74,98]; p(T273,311,315^h) [98,240]; p(S56,883,932^y;T801^y) [93,232]
Non-ATPase regulatory subunit 2	PSMD2	S2 [237]	Rpn1 [238]	Ac(M1^h) [98]; p(S16,361^h) [98]; p(S11,16,19,178,179,182,187^y; T24,190,640^y) [93,116,232,236]
Non-ATPase regulatory subunit 3	PSMD3	S3 [237]	Rpn3 [238]	p(S454^y) [232]
Non-ATPase regulatory subunit 4	PSMD4	S5a [237]	Rpn10 [238]	Fr(V2^h) [74,98]; p(S358,361^h) [98]; p(S259^y) [236]
Non-ATPase regulatory subunit 5	PSMD5	S5b [237]
Non-ATPase regulatory subunit 6	PSMD6	S10a [237]	Rpn7 [238]	Fr(P2^h) [74,98]; p(S8,77^y) [232]
Non-ATPase regulatory subunit 7	PSMD7	S12 [237]	Rpn8 [238]	Fr(P2^h) [74,98]; p(T186^h) [98]; p(S311,314,317,319,321^y; T327^y) [93,232,236,233]
Non-ATPase regulatory subunit 8	PSMD8	S14 [237]	Rpn12 [238]
non-ATPase regulatory subunit 9	PSMD9	S15 [237]	Rpn15 [238]
Non-ATPase regulatory subunit 10	PSMD10	Gankyrin		Ac(M1^h) [98]
Non-ATPase regulatory subunit 11	PSMD11	S9 [237]	Rpn6 [238]	Ac(A2^h) [74,98]; p(S13,78^h) [98]; p(S38^y) [93]; (T287^y) [236]
Non-ATPase regulatory subunit 12	PSMD12		Rpn5 [238]	p(S26^y) [93]
Non-ATPase regulatory subunit 13	PSMD13	S11 [237]	Rpn9 [238]	Fr(M1^h) [74,98]; p(S106,266^h) [98]
Non-ATPase regulatory subunit 14	PSMD14	S13 [237]	Rpn11 [238]	p(S224^h) [98]; p(S243^y) [93]; p(T262^y) [232]
Proteasomal ubiquitin receptor ADRM1	ADRM1	hRpn13	Rpn13 [238]	Ac(T2^h) [74,98]; Ac(S2^y) [93]; p(S132,133,135,140^y) [93,232,233,236]
Proteasomal ATPase-associated factor 1	PAAF1		Rpn14 [238]
26S proteasome complex subunit DSS1	SHFM1 DSS1	DSS1	Rpn15 [238]	Ac(S2^y) [93]; p(S12^y) [93]
Other proteasome activators
Proteasome activator complex subunit 1	PSME1	PA28α – 11Sα – REGα
Proteasome activator complex subunit 2	PSME2	PA28β – 11Sβ – REGβ
Proteasome activator complex subunit 3	PSME3	PA28γ – 11Sγ – REGγ
Proteasome activator complex subunit 4	PSME4	PA200	Blm10	p(S11,29,34,35,56,62^y;T64,66^y) [93,116,232,233,236]

^†Superscript letters in this column denote species.

^‡Determined by CID unless otherwise stated (underlined: determined by ETD and CID; underlined and bold: determined by ETD only).

^§Partial N-acetylation. The free N-terminus was also observed.

Ac: N-acetylation; CID: Collision-induced dissociation; ETD: Electron transfer dissociation; Fr: Free N-terminus; h: Human; m: Mouse; Myr: Myristoylation; P: Phosphorylation; PTM: Post-translational modification; Tr: N-terminal truncated; y: Yeast.

Key issues

• • Proteasomes constitute highly heterogenous complexes and are regulated in a dynamic way.

• • The different 20S proteasome subtypes act as platforms recruiting regulatory complexes and proteasome-interacting proteins of diverse structures and of unknown functions.

• • The latest affinity purification strategies can target the whole diversity of endogenous proteasomes and proteasome-interacting proteins, and are compatible with clinical studies.

• • 2D electrophoresis and shotgun approaches, associated with mass spectrometry, are complementary tools for the characterization of the structural diversities and networks of proteasomes.

• • The new generation of high-resolution, sensitive and fast-sequencing LTQ Orbitrap mass spectrometers equipped with complementary collision-induced dissociation and electron transfer dissociation fragmentation modes results in a more confident identification of post-translational modifications on proteasome subunits. In particular, in association with protein fractionation and specific enrichment techniques, they enabled significant increases in our knowledge on the proteasome phosphorylation status.

• • Quantitative proteomics has evolved as a valuable tool for the analysis of the biological function of proteasomes through characterization of the proteasome network and its dynamics, and of proteasome subcellular localization, in response to various stimuli.

• • K48-linked polyubiquitin chains are probably the most efficient proteasome-targeting signals. Most of the other polyubiquitin linkages probably also address, although less efficiently, proteins to proteasomes, and this might help to modulate protein degradation in response to the cellular demand.

• • Targeted proteomics using multiple reaction monitoring will probably play a growing role in biomarker validation in the near future. Multiple reaction monitoring could constitute an efficient complementary technique to ELISA for the detection of proteasome as a diagnostic and/or prognostic biomarker in body fluids.

Financial & competing interests disclosure

The authors have no 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. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

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