The Ubiquitin Proteoform Problem

Abstract

The diversity of ubiquitin modifications is immense. A protein can be monoubiquitylated, multi-monoubiquitylated, and polyubiquitylated with chains varying in size and shape. Ubiquitin itself can be adorned with other ubiquitin-like proteins and smaller functional groups. Considering different combinations of post-translational modifications can give rise to distinct biological outcomes, characterizing ubiquitylated proteoforms of a given protein is paramount. In this Opinion, we review recent advances in detecting and quantifying various ubiquitin proteoforms using mass spectrometry.

Introduction

Posttranslational modifications (PTMs) dramatically expand the functional capacity of proteins [1]. By modifying amino acid side chains of nascent or folded proteins, PTMs license new chemistry, create new recognition motifs, regulate enzymatic activity, and control protein stability and localization. In many instances, proteins exist in several modified forms referred to as proteoforms [2]. Individual proteoforms are capable of eliciting distinct biological responses. Thus, the combinatorial nature of PTMs enables exquisite regulation of biochemical events and equips cells with the ability to rapidly respond to developmental or physiological cues [3].

One of the more intricate PTMs is protein ubiquitylation. Ubiquitin is a highly conserved 76-residue protein with an exceptionally stable β-grasp fold [4]. The C-terminal glycine of ubiquitin (G76) is covalently attached to the ε-amino group of substrate lysines through the action of three enzymes: E1 (2 human enzymes), E2 (35 human enzymes), and E3 (>600 human enzymes). E1 enzymes activate the ubiquitin C-terminus and transfer the ubiquityl moiety to an active site cysteine of an E2 shuttling enzyme [5,6]. The ubiquitin-charged E2 then interacts with a substrate-bound E3 delivering ubiquitin to a single lysine (monoubiquitylation) or multiple lysines (multi-monoubiquitylation) (Figure 1A) [7-9]. The mechanisms by which ubiquitin is transferred to the substrate depend on the type of E3 [10]. Really Interesting New Gene (RING) E3s act as scaffolds, catalyzing transfer largely via a proximity-induced effect. Homologous to E6AP C-Terminus (HECT) and RING-Between-RING (RBR) E3s, on the other hand, utilize a covalent mechanism involving the intermediacy of a ubiquityl~E3 acyl-enzyme species (“~” refers to a reactive thioester).

Figure 1. Ubiquitin proteoforms.

A. Schematic showing the ubiquitylation cascade leading to the formation of different proteoforms. B. Structure of ubiquitin showing the seven lysines and amino terminus that are used to form ubiquitin chains.

The complexity of ubiquitylation stems from the ability to form ubiquitin oligomers (Figure 1A) [11,12]. Ubiquitin has seven lysines (K6, K11, K27, K29, K33, K48, and K63) and an amino-terminus (M1) that provide eight different attachment sites for the C-terminus of another ubiquitin molecule (Figure 1B). The resulting ubiquitin chains vary in length (number of ubiquitin molecules), linkage, and overall architecture (unbranched versus branched). The number of possibilities is astonishing. In addition, ubiquitin subunits can be adorned with smaller functional groups, e.g., phosphoryl [13,14], acetyl [15], and phosphoribosyl [16,17]. Such diversity, which we refer to as ubiquitin proteoforms, makes it difficult to assign biological function. Considering the flux through most signaling pathways is tightly regulated by ubiquitylation it is imperative to characterize the exact nature of the chain(s) attached to a substrate protein [18,19].

The presence of numerous proteoforms of the modified protein itself further complicates functional analysis. According to the Uniprot database there are many human proteins with annotated PTMs on more than one site and there is substantial evidence of crosstalk between PTMs [20,21]. Understanding the information embedded in a particular ubiquitin proteoform therefore requires each modification to be considered in the context of the other PTMs on the same protein, e.g., other ubiquitin modifications, phosphorylation, glycosylation, acetylation, etc. The problem, however, is that the pattern of PTMs also varies, resulting in a combinatorial explosion of proteoforms. Take for example a protein with four sites that can be monoubiquitylated. There are 16 (2⁴) possible proteoforms of this multi-monoubiquitylated protein alone. So how do we characterize all proteoforms to develop a holistic view of the ubiquitin landscape? In this review we discuss different mass spectrometry-based strategies that are used to characterize ubiquitin proteoforms along with their limitations.

Bottom-Up Approach to Proteoform Analysis

The conventional approach to analyzing ubiquitin modifications involves bottom-up proteomics (BUP) [22,23]. As part of the BUP workflow, proteins are broken down into peptide fragments using a protease, typically trypsin or Lys-C [24,25]. Because the C-terminus of ubiquitin ends in Arg-Gly-Gly, trypsinolysis of ubiquitin conjugates leaves a diGly motif on the side chain of a substrate lysine residue (KGG). LC-MS/MS characterization of diGly-modified peptides then informs on the exact site of ubiquitylation. Due to the low abundance of diGly-modified peptides relative to their unmodified counterparts, an additional enrichment step using KGG-specific antibodies is required [26-29]. With the KGG enrichment approach, over 90,000 unique ubiquitylation sites have now been identified and the landscape of ubiquitin modifications has been analyzed in different cellular states [30-36]. In many of these studies, stimulus-dependent changes in the abundance ubiquitin chain linkages have bolstered the idea that distinct chains have distinct biological functions, which is the basis of the ‘ubiquitin code’ hypothesis. Whole cell KGG analyses, however, do not inform on the type of chain attached to a protein. This can be achieved using immunoprecipitation followed by MS, but a more common approach is to use linkage-specific antibodies [37-41], affimers [42,43], or binding domains [44-46] to enrich for a particular chain type and then use BUP to identify the modified proteins. We refer the reader to a series of excellent reviews discussing the utility of linkage-specific chain enrichment strategies [47,48].

The challenge with BUP is that the precise nature of the chain cannot be discerned nor can the relationship between ubiquitylation and other PTMs. What BUP lacks in the ability to characterize proteoforms, is offset by the quantitative information that can be gleaned. Coupling BUP with targeted methods based on heavy reference peptides and selected ion monitoring techniques, e.g., parallel reaction monitoring (PRM) or selected reaction monitoring (SRM)-MS, can provide information on site-specificity, abundance, stoichiometry, and kinetics (Figure 2) [49]. Two recent studies nicely illustrate the power of targeted approaches in the analysis of proteoforms.

Figure 2. Ubiquitin proteoform analysis with BUP.

Scheme for selected ion monitoring using PRM or SRM-MS. Samples are digested with trypsin and Lys-C and mixed with heavy reference peptides corresponding to unmodified or modified forms of the protein prior to analysis by proteomics. MS¹ and MS² intensities are used to determine relative and absolute abundance of individual sites. Information on site-specificity, abundance, stoichiometry, and kinetics can be obtained using PRM/SRM-MS.

Damaged mitochondria are removed through a form of selective autophagy termed mitophagy [50]. The mitochondrially localized kinase PINK1 accumulates on the mitochondrial outer membrane (MOM) upon damage [51-53], where it phosphorylates ubiquitin to recruit and activate the PARKIN RING-Between-RING (RBR) ubiquitin ligase [54-63]. Ubiquitylation of numerous MOM-associated proteins along with the formation of ubiquitin chains then serves to mobilize autophagy receptors for proper disposal of damaged mitochondria [64-67]. To better understand how the coordination between phosphorylation and ubiquitylation control the rate of mitophagy, both label-free and PRM-MS-based quantitation were performed on MOM proteins [68]. Phosphorylated ubiquitin S65 (pS65-Ub) was found to comprise 20% of the ubiquitin pool purified with mitochondria. Increasing the fractional occupancy of pS65-Ub to ~60% reduces the association with mitophagy receptors, suggesting the stoichiometry of ubiquitin phosphorylation has been fine tuned to facilitate the recruitment of both PARKIN and mitophagy receptors. Establishing a temporal order of ubiquitylation by PRM-MS led to the identification of privileged PARKIN targets and the preferred sites of modification. Several lysines within the same target protein were modified with similar kinetics, suggesting multi-ubiquitylated proteoforms could be important for proper mitophagic flux. PRM-MS also offers a highly sensitive method for quantifying ubiquitin chain linkages [69], and in neurons K63-linked chains appear to be the preferred signal for PARKIN-dependent mitophagy [70]. Thus, the combination of heavy reference peptides and PRM-MS provides a powerful means to determine the absolute amounts of different proteoforms even though the precise nature can only be inferred.

Selected ion monitoring has also enabled the characterization of tau proteoforms. Tau is a microtubule-associated protein predominately expressed in neurons [71,72]. Misregulation leads to the formation of aggregates, which are the hallmark of a class of neurodegenerative diseases referred to as tauopathies [73]. Hyperphosphorylation of tau is considered an early event in the aggregation process [74]. A number of other PTMs, including ubiquitylation, have been implicated in disease progression. To define the tau PTMs relevant to tauopathies, specifically Alzheimer’s disease (AD), heavy full-length tau protein was spiked into samples containing postmortem brain tissue prior to proteolytic digestion [75]. The heavy reference peptides facilitate quantitation of the unmodified peptides as opposed to the modified variants [76]. In principle, the abundance of the unmodified peptide will decrease by an amount proportional to the modified form(s). MS-based detection of the unmodified peptide must therefore be highly reproducible, which is why a targeted ion monitoring approach must be employed. Coupling this strategy with statistical analyses uncovered common features in AD patients and identified combinations of PTMs that reflect disease progression. Phosphorylation of tau within the proline-rich region occurs at the earliest stages of disease. Ubiquitylation and acetylation of sites in the microtubule-binding domain occur later and are strongly associated with formation of high molecular weight, detergent-insoluble forms of tau that have prion-like behavior. These results together with cryo-EM data of tau suggest that distinct combinations of ubiquitylation and acetylation contribute to the stability of different filament structures [77]. Whether ubiquitin chains or other ubiquitin modifications play a role in filament formation and thus disease progression remains unclear.

Insights into Chain Architecture Using Middle-Down MS

Although the quantitative information afforded by BUP is unparalleled, the loss of connectivity between modified peptides means the structure of different proteoforms can only be inferred. The inference problem is particularly acute for ubiquitin chains, as insight into chain length and the extent of branching is completely lost.

Branched chains have been implicated in a number of pathways ranging from the cell cycle to immune signaling [41,78-84]. Branched chains are composed of at least two or more linkages and have individual subunits that are modified with two or more ubiquitin molecules. When multiple modifications occur on non-adjacent lysines, BUP is unable to distinguish between branched and unbranched chains. Middle-down proteomics (MDP), however, facilitates the detection of branched chains by exploiting the recalcitrance of ubiquitin to tryptic cleavage. Under native conditions, the peptide bond between R74 and G75 of ubiquitin is the most susceptible to trypsinolysis [85]. With ubiquitin largely intact, distinct regions of a ubiquitin chain can be detected by MS: the caps (Ub_1-74), the unbranched portion (diGly-Ub_1-74), and the branchpoints (2xdiGly-Ub_1-74) (Figure 3). Early studies showed that the ratio of Ub_1-74 to diGly-Ub_1-74 could be used to deduce the length of isolated chains [86]. By coupling MDP with electron-transfer dissociation (ETD) MS² analysis, subsequent studies demonstrated that the extent to which different E3s assemble branched chains could be assessed and information on the linkages present in branchpoints could be obtained [87]. MDP can also be used to measure the abundance of branchpoints in cell extracts after enrichment of ubiquitin chains by immobilized ubiquitin-binding domains (UBDs) [88]. By availing a K11 linkage-specific antibody, for example, K11/K48 branched chains were found to represent ~4% of the total population of the enriched ubiquitin species in G2/M synchronized cells [89].

Figure 3. Middle-down MS analysis of ubiquitin chain architecture.

Ubiquitylated proteins are digested with trypsin under native conditions or treated with LbPro* to generate three distinct ubiquitin variants: Ub_1-74, diGly-Ub_1-74, and 2xdiGly-Ub_1-74. The resulting variants are analyzed by MS¹ and MS² to determine the extent of chain branching and the modification sites.

MDP has also played an instrumental role in identifying UCHL5/UCH37 as a branched chain-selective deubiquitinase. UCH37 was first discovered as a deubiquitinase transiently associated with the 19S regulatory particle of the 26S proteasome [90]. For a long time, the premise was that the proteasomal subunit RPN13 stimulates the ability of UCH37 to trim the distal end of K48 chains, thereby rescuing proteins prematurely sent to the proteasome for degradation. However, it has been hard to reconcile this model with data showing that UCH37 cleaves homotypic K48 chains at very slow rates even in the presence of RPN13 and the proteasome [91-94]. A clue that UCH37 might be targeting K48 linkages in heterotypic chains was provided by quantitative analysis of chain linkages using PRM-MS [95]. It was MDP, however, that allowed for direct visualization of UCH37’s ability to remove K48 branchpoints and regulate proteasomal degradation through this activity.

A major limitation with trypsin-mediated MDP, however, is the need to carefully optimize the experimental conditions to avoid unwanted digestion. Another issue is that minimal trypsinolysis results in cleavage of the ubiquitin-modified protein. Thus, information is forfeited on other PTMs that may promote or be a consequence of ubiquitylation. In a groundbreaking study, the leader protease (LbPro*) of foot-and-mouth disease virus was engineered to specifically recognize and cleave ubiquitin after R74 [96]. With this protease, ubiquitin is ‘clipped’ at its C-terminus leaving behind the signature Kgg motif on the modified protein. The same ubiquitin variants—Ub_1-74, phospho-Ub_1-74, diGly-Ub_1-74, phospho-diGly-Ub_1-74, and 2xdiGly-Ub_1-74—that can be detected by trypsin-mediated MDP are identified with ‘Ub-clipping’, but the protein attached to the chain or mono-ubiquitin is left intact (Figure 3). Ub-clipping has recently been used to investigate the phospho-ubiquitin proteoforms produced during PINK1/PARKIN-dependent mitophagic signaling [70]. The majority of ubiquitin phosphorylation was found to occur on monomeric ubiquitin or the caps of chains.

The Challenge of Chain Length

While advances in BUP and MDP enable the characterization and quantitation of ubiquitin chain linkages, ubiquitin PTMs, and the degree of branching in complex biological samples, the analysis of chain length has largely relied on gel mobility. As BUP experiments have shown, many substrates have multiple ubiquitylation sites making it difficult to discern whether the ‘ubiquitin smear’ on a gel or immunoblot is due to multi-monoubiquitylation or heterogeneity in chain length. Chains with the same number of ubiquitin molecules can also have different gel mobilities due to linkage type and branching. Thus, we have largely been left in the dark regarding the functional relevance of chain length.

The need for characterization methods is further underscored by mounting evidence suggesting that chain length controls the dynamics of ubiquitin-dependent pathways. A classic example is the in vitro study from the Pickart lab showing a K48 tetraubiquitin chain provides the minimal proteasome-targeting signal for the globular protein dihydrofolate reductase [97]. More recently, chain length has emerged as an important factor for substrate selection by the ubiquitin-dependent unfoldase Cdc48/p97/VCP [98-100], the deubiquitinase MINDY [101], and the ubiquitin-directed endoprotease DDI2/Ddi1 [102,103]. In each of these cases, longer chains bearing more than five subunits are preferentially recognized, suggesting there is certain threshold that must be achieved to trigger downstream events. Methods for characterizing chain length could help us understand how often this threshold is met, the degree to which it depends on the nature of the chain and substrate, and the cellular processes that rely on it.

For the purposes of characterization, access to an enzyme that could remove chains en bloc from a substrate would be ideal, as chains would be left intact. If the same enzyme also possessed cleavage specificity for R74 of ubiquitin, then ubiquitylation sites along with other PTMs could be mapped on a substrate protein. In the absence of such a reagent, other approaches have been devised. In a method referred to as Ub-ProT (ubiquitin chain protection from trypsinization), trypsin is combined with a trypsin-resistant, multivalent display of ubiquitin-binding domains to release intact chains from substrates [104] (Figure 4). The lengths are then analyzed using a gel-based assay and the absolute abundance of different linkage type are assessed using BUP. In yeast, substrate-anchored ubiquitin chains were mainly found in the range of monomer to heptamer, with Cdc48 regulating chain length. The challenge is that not all ubiquitin chain linkages are fully protected from trypsinolysis and branched chains seem to be more prone to digestion than homotypic chains.

Figure 4. Ubiquitin chain Protection from Trypsinization (Ub-ProT).

Trypsin-Resistant Tandem Ubiquitin Binding Entities (TR-TUBEs) enable the isolation of polyubiquitinated proteins from cells. Following trypsinolysis, the modified substrate is digested but the chains are left intact. Western blotting and BUP can then be used for characterization.

Top-Down Proteomic Analyses

As the only MS-based platform that is not limited by the protein inference problem, top-down proteomics (TDP) should be well-suited for the analysis of proteoforms bearing ubiquitin chains of different lengths, linkages, and architectures. Proteoforms are left completely intact prior to mass analysis, and identification and localization of chemical modifications are determined by MS² analysis of fragment ions [105]. The major challenge is that with increasing molecular weight the signal-to-noise (S/N) ratio decreases exponentially [106]. The low abundance of ubiquitylated proteins further compounds the problem. Thus, while TDP has been used to characterize well-defined ubiquitylated proteins [107,108], the success of TDP in the analysis of complex, heterogenous mixtures has not been realized. With advances in sample preparation, separation, and instrumentation [109], it will eventually be possible to capture the entire landscape of ubiquitylation events on a protein of interest and investigate the interplay with other PTMs. Indeed, by combining LbPro* with the enrichment of a protein of interest, the characterization of modification crosstalk using TDP is already within reach, and it will be exciting to see future applications of this strategy. The advent of new proteolytic tools that differentiate between ubiquitin-ubiquitin and ubiquitin-target protein linkages would also accelerate TDP analyses. In parallel with modification crosstalk, chain length distribution could be assessed while also avoiding complications associated with the low S/N ratios of polyubiquitylated proteins.

Conclusions and Future Directions

Complex cellular activities are made possible by the diversity of proteoforms that arise from splice variants and PTMs. Even within the same family of proteoforms, distinct combinations of PTMs influence both structure and function, enabling highly sophisticated cellular information processing. The complexity of PTMs like ubiquitylation, however, make it particularly challenging to identify and quantify proteoforms to understand function. Despite these obstacles, mass spectrometry has proven to be an invaluable tool in the study of ubiquitin-containing proteoforms. When combined with heavy isotope-labeled peptides and selective ion monitoring techniques such as PRM or SRM, BUP informs on the site-specificity, abundance, stoichiometry, and kinetics, but correlations between the target protein and the precise nature of the ubiquitin modification are completely lost. MDP offers insights into the architecture of chains, but chain length must be inferred. Since the ultimate goal is to characterize all ubiquitylated proteoforms within a given family and determine which proteoform(s) is/are responsible for a particular biological response, information on all sources of variability must be retained. Thus, TDP will play an increasingly important role in the characterization of ubiquitin proteoforms.