Novel Strategies to Address the Challenges in Top-Down Proteomics

Abstract

Top-down mass spectrometry (MS)-based proteomics is a powerful technology for comprehensively characterizing proteoforms to decipher post-translational modifications (PTMs) together with genetic variations and alternative splicing isoforms toward a proteome-wide understanding of protein functions. In the past decade, top-down proteomics has experienced rapid growth benefiting from groundbreaking technological advances, which have begun to reveal the potential of top-down proteomics for understanding basic biological functions, unraveling disease mechanisms, and discovering new biomarkers. However, many challenges remain to be comprehensively addressed. In this Account & Perspective, we discuss the major challenges currently facing the top-down proteomics field, particularly in protein solubility, proteome dynamic range, proteome complexity, data analysis, proteoform–function relationship, and analytical throughput for precision medicine. We specifically review the major technology developments addressing these challenges with an emphasis on our research group’s efforts, including the development of top-down MS-compatible surfactants for protein solubilization, functionalized nanoparticles for the enrichment of low-abundance proteoforms, strategies for multidimensional chromatography separation of proteins, and a new comprehensive user-friendly software package for top-down proteomics. We have also made efforts to connect proteoforms with biological functions and provide our visions on what the future holds for top-down proteomics.

Graphical Abstract

1. INTRODUCTION

In the post-genomics era, a comprehensive analysis of “proteoforms”^1,2—all protein products that arise from a single gene due to genetic variations, alternative splicing, and post-translational modifications (PTMs)—is essential for understanding biological systems at a functional level and for dissecting complex molecular systems with consideration of individual variability toward precision medicine.^3–7 The well-established “bottom-up” proteomics⁸ analyzes peptides from protein digests and has become invaluable for the high-throughput identification and quantification of proteins as well as exploration of proteome structure and function,^9,10 but it is suboptimal for characterizing PTMs and sequence variants due to the “peptide-to-protein” inference problem.^2,11 In contrast, top-down mass spectrometry (MS)-based proteomics analyzes intact proteins, thus making it the premier technology for comprehensive analysis of proteoforms to decipher PTM codes together with genetic mutations and alternative splicing.^1–7

Top-down MS was first coined by McLafferty and co-workers in 1999,¹² and the early top-down MS studies generally took a targeted approach focusing on offline purified single proteins or simple protein mixtures due to the inherent limitations of intact protein analysis.^12–17 Top-down proteomics, pioneered by Kelleher and co-workers, includes the front-end fractionation of intact proteins, high-resolution MS,¹⁸ and back-end informatics,¹⁹ making it possible to identify proteins from complex mixtures.^6,20,21 In the past decade, top-down proteomics has experienced rapid growth due to many technological advances,^22–32 thereby enabling the applications of top-down proteomics to complex protein mixtures and for understanding basic biological functions, unraveling disease mechanisms, and discovering new biomarkers.^33–45 However, many challenges still exist for top-down MS-based proteomics, especially protein solubility, dynamic range, proteome complexity, data analysis, proteoform–function relationships, and analytical throughput of top-down proteomics for precision medicine (Figure 1). Previous reviews on top-down proteomics have already described various aspects of top-down proteomics such as technological advancements and biomedical applications, among others.^{4–6,17,21,46–53} In this Account & Perspective, we will focus on the most pressing challenges currently facing the top-down proteomics field. We illustrate how recent and future technological innovation could help address these challenges to advance top-down proteomics toward the mainstream. Given the focus of this special issue, we specifically detail the major strides our research group has made to address these current challenges, and we will provide our vision of what the future holds for top-down proteomics.

Figure 1.

Addressing the challenges in top-down proteomics through novel strategies. Illustration of the current major challenges in top-down proteomics. These challenges include protein solubility, proteome dynamic range, proteome complexity, intact protein data analysis, proteoform–function relationships, and analytical throughput.

2. ADDRESSING THE CHALLENGE OF PROTEIN SOLUBILITY

Protein solubility represents a major challenge in top-down MS-based proteomics, especially for membrane proteins^54,55 and extracellular matrix (ECM) proteins,^56,57 which are difficult to extract and solubilize from tissues and cells but play significant biological roles.^5,48,58–60 Despite the success of several important top-down proteomics studies of membrane proteins,^{25,54,61–65} the time-consuming preparations and labor-intensive workups prior to MS analysis leave significant room for technological improvement in this area. Moreover, early work generally focused on targeting a specific protein of interest or the mitochondrial membrane subproteome.^{25,54,61–65} Surfactants (also known as detergents) are typically used in protein extraction buffers to effectively solubilize proteins from cells and tissues.^66,67 However, conventional surfactants that are capable of solubilizing and denaturing proteins with high efficiency, such as sodium dodecyl sulfate (SDS), are not compatible with downstream MS analysis due to the signal suppression of proteins and thus require extensive cleanup that can result in protein loss and irreproducibility.^66,68 Although non-ionic surfactants, such as dodecyl β-d-maltoside (DDM) and octyl β-d-glucopyranoside (OG), are compatible with MS when used at lower concentrations,^68,69 they are relatively mild surfactants with limited solubilization ability.⁶⁷ Acid-cleavable surfactants have been previously developed for bottom-up proteomics, such as RapiGest (RG)⁷⁰ (also known as ALS),²⁰ ProteaseMax,⁷¹ and MS-compatible slowly degradable surfactant (MaSDeS),⁷² to enable protein solubilization and improve in-gel or in-solution digestion efficiency. However, none of these surfactants are directly compatible with top-down MS-based proteomics.⁵⁹

To address this problem, it is essential to develop novel top-down MS-compatible surfactants that can quickly degrade into innocuous non-surfactant byproducts, facilitating easy surfactant removal before MS analysis. We recently identified a novel photocleavable surfactant, 4-hexylphenylazosulfonate (“Azo”), which effectively solubilizes proteins with similar performance to SDS but can rapidly degrade upon exposure to UV irradiation, thus enabling top-down proteomics (Figure 2A–D).⁵⁹ A systematic comparison of Azo with commonly used surfactants (i.e., SDS, RapiGest, ProteaseMax, MaSDeS, DDM, OG, digitonin) was performed to evaluate their ability to solubilize proteins from the insoluble cardiac tissue pellets, which confirmed that Azo is the only surfactant that can effectively solubilize proteins while maintaining top-down MS compatibility (Figure 2E–G). Moreover, Azo can greatly enhance access to proteins from many cellular components including the nucleus, mitochondria, plasma membrane, endoplasmic reticulum, cytoplasm, cytoskeleton, and extracellular region using top-down proteomics.⁵⁹ Importantly, the use of Azo enables the facile solubilization of membrane proteins from human cardiac tissue for comprehensive characterization and localization of PTMs.⁵⁹ For example, the PTMs of phospholamban, a well-known cardiac regulatory protein, were characterized and a palmitoylation modification to cysteine 36 within the transmembrane region was localized.⁵⁹ Moreover, 46 subunits of the electron transport chain and 51 proteins with transmembrane domains, including all of the subunits of the ATP synthase complex, were confidently identified with high mass accuracy directly from cardiac tissue. Since the synthesis of Azo is very simple and Azo can be used as a general SDS replacement in SDS-PAGE, this surfactant has a broader impact beyond MS-based proteomics.⁵⁹

Figure 2.

A novel photocleavable surfactant to address protein solubility challenge. (A) Scheme illustrating the use of Azo in solubilizing proteins, followed by rapid degradation with UV irradiation, and MS analysis of the intact proteins. Note that the molecules are not drawn to scale. (B) Degradation of Azo into 4-hexylphenol, 4-hexylbenzene, nitrogen, and hydrogen sulfate under UV irradiation. (C) Synthetic scheme for Azo. (D) UV–Vis spectra of Azo (0.1%) degradation as a function of time, showing that Azo can be rapidly degraded by UV irradiation at ambient temperature. SDS–PAGE analysis (E) and protein assay (F) for the evaluation of effectiveness of surfactant-aided protein extractions (E3) following the initial HEPES buffer extractions (E1 and E2) to deplete the cytosolic proteins from the cardiac tissue. NS, no surfactant (serving as a control). Error bars represent the standard error of measurement for protein assay experiments (n = 3). (G) ESI–MS analysis of ubiquitin with 0.1% surfactant to investigate MS compatibility. The mass spectra were normalized to an intensity of 1.7 × 10⁶. Data are representative of three independent experiments. (H) Scheme for Azo-enabled high-throughput top-down and bottom-up proteomics. Figure adapted from refs 59 and 73. Copyright 2019 Springer Nature (A–G). Copyright 2020 John Wiley & Sons, Inc. (H).

We further showed that Azo is a versatile proteomics surfactant compatible with both top-down and bottom-up proteomics workflows.^59,73 Rapid enzymatic digestion by trypsin (~1 h) was observed due to the highly denaturing nature of Azo facilitating the opportunity for high-throughput proteomics. Additionally, an Azo-aided bottom-up method was developed to study integral membrane proteins, providing over a 6-fold increase in the number of identified integral membrane proteins when compared to conventional sample preparation strategies.⁷³ The unique ability of Azo to facilitate both top-down and bottom-up proteomics, including the analysis of membrane proteins, makes it a valuable tool for proteomics sample preparation (Figure 2E). Recently, we expanded the use of Azo to extracellular matrix (ECM) proteomics.⁶⁰ The ECM provides an architectural meshwork for surrounding cells, and the dysregulation of ECM proteins directly contributes to various pathologies and progression in human diseases such as cancers and cardiovascular diseases.^56,57,74,75 However, the biochemical/proteomic characterization of ECM proteins is extremely challenging due to the nearly insoluble nature of the ECM.^35,36 To address these challenges, we established a new decellularization/extraction method enabled by Azo^59,73 for the efficient enrichment of ECM proteins, minimizing the sample cleanup before LC-MS/MS analysis.⁶⁰ Overall, we have shown that Azo could enhance proteomic analysis through simple protein sample preparation and streamlined proteomic workflows, so we anticipate Azo will be extended to other proteomics applications. Nonetheless, it should be noted that certain protein classes are currently more amenable to acid/base extractions, such as acid-soluble histones.^76,77

MS-based proteomics experiments are commonly performed under denaturing conditions to obtain the primary structural analysis of proteins.⁷⁸ In contrast, native MS performed under non-denaturing conditions seeks to preserve whole protein structures and their complexes, which enables the study of protein–protein interactions and ligand/drug binding.^79–81 The recent integration of native MS with top-down proteomics (“native top-down proteomics”) allowed the structural characterization of macromolecular complexes together with comprehensive sequence mapping and proteoform characterization.^26,82–86 In particular, the rapidly developing native membrane proteomics has huge potential for unlocking structural and functional information on these important proteins.^26,69 Solubilizing membrane proteins for native MS requires keeping proteins stable and as “native” as possible. Toward this end, modular surfactants were designed and employed by Urner et al. for improved purification and native MS of G-protein-coupled receptors.⁸⁷ In an effort to further increase the “native” environment, Chorev et al. developed a method to analyze protein complexes ejected directly from native E. coli outer membranes.⁸⁸ Finally, one of the most significant advancements was the development of “nativeomics”, wherein native MS was used to precisely identify ligands bound to membrane protein assemblies, providing insight into their functional role.⁸² Conceivably, the continued developments in the area of native membrane proteomics, including new membrane mimics and purification strategies, will be needed for analyzing endogenous membrane proteins/protein complexes (particularly as it shifts toward studying eukaryotic membrane proteins).⁸⁴

3. ADDRESSING THE CHALLENGE IN THE HIGH DYNAMIC RANGE OF THE HUMAN PROTEOME

The high dynamic range of the proteome (10⁸–10¹²) makes detection of low-abundance proteoforms extremely difficult.⁸⁹ Front-end protein enrichment strategies are therefore required to specifically capture and enrich low-abundance proteoforms from the complex biological milieu before MS analysis.^90,91 Antibodies have historically been the dominant affinity reagents for protein capture and quantification in biomedical research.^91–93 Antibody-based affinity purification (AP) has been a preferred approach in the targeted analysis of intact proteins and protein complexes.^4,94,95 For example, AP has been utilized to couple with top-down mass spectrometry (MS) for the purification of cardiac troponin from both human and animal heart tissues to identify disease-associated proteoform alterations.^34,35 Nonetheless, antibody-based technologies have significant limitations, including the high cost of antibody production, difficulty in the generation of antibodies with high specificity, limited availability of high-quality antibodies, batch-to-batch variability of the antibodies, and relatively low stability.^96–99 Hence, there is an urgent need to develop next-generation affinity reagents.^92,100,101

To address this challenge, we have developed surface-functionalized multivalent superparamagnetic nanoparticles (NPs) as a general affinity platform to capture and enrich low-abundance proteoforms with high specificity, provided the NPs are functionalized with a suitable affinity reagent.^102–104 Functionalized NPs have several unique advantages for protein enrichment: (a) NPs possess comparable size and diffusion kinetics to antibodies, allowing for better penetration through complex protein mixtures and promoting effective capture of low-abundance proteins; (b) NPs can be produced with simple, fast, and scalable synthesis, in contrast to the long time required to produce antibodies; (c) NPs have high surface area to volume ratios due to their nanoscale size (5–20 nm) which translates into high surface ligand density and increased protein interactions; (d) NPs can have multiple flexible binding sites allowing multivalent ligand interactions which can increase the overall binding affinity and ensure the efficient capture of lower-abundance proteins.^102–104

The first generation of functionalized superparamagnetic iron oxide (Fe₃O₄) NPs was developed to universally enrich phosphoproteins regardless of the specific phosphorylation site (Ser, Thr, or Tyr phosphorylation).¹⁰² Subsequently, an integrated top-down proteomics workflow has been established for comprehensive phosphoprotein characterization enabled by coupling specific intact phosphoprotein enrichment by cobalt ferrite (CoFe₂O₄) NPs with online LC-MS/MS analysis.¹⁰⁴ Roberts et al. have further developed a reproducible large-scale synthesis of surface silanized Fe₃O₄ NPs as an enabling nanoproteomics platform and demonstrated its applications for the highly specific enrichment of the human heart phosphoproteome.¹⁰³ In this work, a reproducible and large-scale (200 mg) synthesis of surface-silanized magnetite (Fe₃O₄) NPs for biological applications was developed. As an alternative to antibodies, these NPs were surface functionalized with a dinuclear Zn(II)-dipicolylamine (Zn-DPA) complex as a phosphoprotein affinity ligand to specifically enrich intact phosphoproteins.¹⁰³ By further integrating with a LC-MS/MS workflow suitable for intact protein analysis, this highly scalable platform was used for the global enrichment, identification, and characterization of endogenous phosphoproteins from human cardiac tissues.

Recently, we have exploited this nanoproteomics platform to enrich cardiac troponin I (cTnI), the gold-standard biomarker for acute myocardial infarction (also known as a heart attack), from human serum with high specificity and sensitivity.³⁰ The human blood proteome is enormously complex, possessing a huge dynamic range (10¹²) dominated by the presence of numerous highly abundant proteins, such as human serum albumin (HSA), making the analysis of low-abundance proteins extraordinarily challenging.^89,90 Therefore, a nanoproteomics strategy using peptide-functionalized NPs was developed to specifically capture and enrich low-abundance cTnI (<1 ng/mL) from human serum with high specificity and reproducibility (Figure 3A). By coupling this nanoproteomics strategy with top-down MS, diverse cTnI proteoforms were enriched and characterized with high reproducibility (Figure 3B).³⁰ Importantly, these carefully designed peptide-functionalized NPs preserved all endogenous cTnI proteoform distributions and faithfully retained the endogenous cTnI PTM profiles at every step of the enrichment process without introduction of artifactual modifications (Figure 3C). The ability to accurately quantify multiple proteoforms concurrently marks a unique advantage that the top-down MS approach provides over bottom-up MS and conventional immuno-based techniques.^{4,98,99,105,106} Moreover, these NPs can directly capture and enrich cTnI from human serum with high sensitivity (LOD of ~0.06 ng/mL) and high specificity, showing near-complete resistance to nonspecific binding of the highly abundant HSA. These NPs not only outperformed conventional monoclonal antibody platforms for serum cTnI enrichment but also faithfully and holistically preserve all endogenous cTnI proteoforms (Figure 3C). We are currently applying this nanoproteomics platform to human patient blood samples to establish the relationships between the cTnI proteoforms and underlying disease etiology for precision medicine.

Figure 3.

Nanoproteomics enables proteoform-resolved analysis of low-abundance cardiac troponin I in human serum. (A) Silanization of Fe₃O₄ NPs using an allene carboxamide-based organosilane monomer (BAPTES) for cysteine thiol-specific bioconjugation. The rationally designed NPs are surface functionalized with a 13-mer peptide that has a high affinity for cTnI (NP-Pep) for cTnI enrichment. (B) Top-down MS-based evaluation of cTnI enrichment using three different synthetic batches of NP-Pep showing the reproducible enrichment performance. (C) Evaluation of total cTnI proteoform recovery. The deconvoluted top-down mass spectra corresponding to cTnI proteoforms were used to calculate the relative abundance of each cTnI proteoform when normalizing for total protein amount injected. Proteoform abundance data are representative of n = 6 independent experiments with error bars indicating the standard error of the mean. cTnI relative abundance data. Roman numerals correspond to N-terminally acetylated cTnI proteoforms following Met exclusion: (i) ppcTnI[1–207]; (ii) cTnI; (iii) pcTnI; (iv) ppcTnI; (v) cTnI[1–205]; (vi) pcTnI[1–205]; (vii) cTnI[1–206]; (viii) pcTnI[1–206]. (D) Nanoproteomics assay utilizing NP-Pep for specific enrichment of cTnI from serum and subsequent top-down MS analysis of cTnI proteoforms. cTnI is first spiked into human serum to prepare the loading mixture (L). The NPs are then incubated with the serum loading mixture, the cTnI-bound NPs are magnetically isolated, and the unwanted and nonspecific proteins are removed as flow-through (F). The captured cTnI is then eluted, and the final elution fraction after enrichment is analyzed by top-down LC-MS/MS. The cTnI (~10–20 ng/mL) spiked in the human serum (10 mg) were extracted from various human hearts: (i) and (ii), donor hearts; (iii) and (iv), diseased hearts with dilated cardiomyopathy, (v) and (vi), post-mortem hearts. p, phosphorylation. pp, bisphosphorylation. Figure adapted from ref 30. Copyright 2020 Springer Nature.

This top-down nanoproteomics platform has a unique combination of specificity, scalability, reproducibility, and efficiency for the capture and enrichment of low-abundance proteoforms and can be widely applied to many other biological systems of interest to enable proteoform-resolved analysis of low-abundance proteins in general. Future applications synergizing the advances in aptamer^107,108 developments with nanotechnologies^109,110 may enable the application of alternative affinity reagents to new nanoparticle platforms for robust intact protein analysis by top-down MS. Nonetheless, advances in in silico methods for affinity reagent design and virtual screening can lead to the further development of affinity reagents such as affimers and improvements in phage display can help address these challenges.³⁰ Moreover, it should be noted that enrichment strategies may introduce systemic bias in protein sampling and requires careful evaluation of the enrichment method and data analysis to ensure quantitative reproducibility.^48,111 We anticipate continuous technology advancements in the coming years for the enrichment of the low-abundance proteoform and the subsequent quantitative top-down MS analysis.

4. ADDRESSING THE CHALLENGES IN PROTEOME COMPLEXITY

The complexity of the proteome remains a significant challenge to top-down proteomics, which necessitates the separation of intact proteins prior to the MS analysis.^{22,50,112,113} As MS instruments capable of top-down analysis have become more widely available, the demand for the further development of intact protein separations has grown.⁵ Despite the central roles played by liquid chromatography (LC) in the development of bottom-up proteomics,¹¹⁴ chromatographic separation of proteins has been considered very challenging and underdeveloped.^5,112 However, recent improvements to liquid chromatographs,^115,116 advances in coreshell and monolithic stationary phases,^117,118 and the development of new column chemistry and selectivities¹¹⁹ have greatly improved the separation resolution and efficiency of intact protein mixtures. In this section, we will discuss size-based chromatography separation, non-denaturing separations, ion mobility spectrometry, and the creative integration of several orthogonal separation modes in multidimensional liquid chromatography (MDLC), which have shown great potentials to enhance top-down proteomics (Figure 4A).

Figure 4.

Novel separation strategies to address proteome complexity. (A) Comparison of the various protein separation techniques. Ion-exchange chromatography (IEX or IEC), hydrophobic interaction chromatography (HIC), reverse phase chromatography (RPC), size-exclusion chromatography (SEC), and hydrophilic interaction chromatography (HILIC) are shown as representative separation techniques. (B) Overview of the serial size-exclusion chromatography (sSEC) strategy for complex protein mixture separation. (1) Protein mixtures pass through two sSEC columns (1000–500 Å) to separate the proteins by high (1000 Å) and low (500 Å) molecular weights, (2) specific protein fractions are collected, (3) sample fractions are analyzed using online RPC-MS, (4) individual proteins are characterized. Figure adapted from ref 29. Copyright 2017 American Chemical Society. (C) Online HIC-MS of mAb mixtures on a maXis II Q-TOF mass spectrometer. Mass spectrum of a specific mAb (mAb2) showed the detection of monomers, dimers (30× zoom-in), and trimers (100× zoom-in). A deconvoluted mass spectrum of the mAb2 monomer is shown with annotated glycosylation forms (red triangle, fucose; blue square, GlcNAc; green circle, mannose; yellow circle, galactose); a hollow square represents the loss of one GlcNAc (−203 Da), a hollow triangle represents the preservation of C-terminal Lys on the heavy chain (+128 Da), and an asterisk represents the addition of a hexose (+162 Da). GxF indicates Fc-oligosaccharides terminated by x number of galactoses. Figure adapted from ref 23. Copyright 2018 American Chemical Society. (D) Illustration of the 3D IEC (or IEX) /HIC/RPC-MS/MS separation strategy. HIC was employed as a second dimension of separation (following IEX) prior to top-down RPC-MS/MS analysis. Online RPC-MS analysis of intact proteins from a HEK293 cell lysate following IEC-HIC fractionations (3DLC approach) with representative RPC/MS results of 3DLC (IEC-HIC-RPC) is shown. Figure adapted from ref 148. Copyright 2015 American Chemical Society.

The broad range of protein molecular weights within a proteome can exceed 5 orders of magnitude,¹²⁰ making global top-down coverage of the proteome difficult. This challenge is most acute for large proteins because, as protein size increases, the signal-to-noise ratio decreases due to an increased number of charge states and isotopomers.¹²¹ To address this problem, Tran and Doucette developed gel-eluted liquid fraction entrapment electrophoresis (GELFrEE) as an offline technique to bin proteins into size-selected fractions.¹²² Another method, termed Passively Eluting Proteins from Polyacrylamide gels as Intact species for MS (PEPPI-MS), has been developed for size-based proteome fractionation prior to LC-MS/MS.²⁴ The PEPPI-MS method enables recovery of separated protein bands from SDS-PAGE gels using equipment common to biological and biochemical laboratories. Recently, our group developed serial size exclusion chromatography (sSEC) for size-based proteome fractionation followed by RPLC enabling the detection of proteoforms up to 223 kDa using a Q-TOF mass spectrometer (Figure 4B).²⁹ In contrast to conventional SEC methods which use only one separation column,¹²³ the sSEC method links multiple columns of differing porosity to more effectively separate smaller proteoforms from large proteoforms, facilitating the characterization of high molecular weight proteins.²⁹ Proteins in excess of 100 kDa that were previously only detectable using ultrahigh-resolution MS, such as cardiac myosin binding protein C (140 kDa)¹²⁴ and myosin heavy chain (223 kDa),¹²⁵ could be characterized using online RPLC-MS/MS using a Q-TOF mass spectrometer.²⁹ We further extended this method for characterization of large cardiac proteoforms, such as swine aconitate hydratase (82.5 kDa), using a 12T FTICR-MS without front-end separation.¹²⁶ Unlike GELFrEE or PEPPI-MS, sSEC can fractionate proteins in both offline and online modes due to MS compatibility of the mobile phases used. This flexibility allows much greater sample throughput than other size-based separations.

Mobile phase compatibility with ESI is also essential to the development of new separation selectivities.¹²⁷ Protein separation techniques such as ion-exchange chromatography (IEX) and hydrophobic interaction chromatography (HIC)¹²⁸ rely on high concentrations of buffer salt in the mobile phase to stabilize the tertiary structure of the protein and influence the selectivity of the separation, but the conventional buffers such as phosphate, sulfate, or citrate salts are non-volatile.^129,130 Recent work has allowed online coupling of both IEX and HIC with MS using the volatile buffer ammonium acetate.^130,131 In collaboration with Andy Alpert from PolyLC, we developed a series of more hydrophobic HIC materials that can separate proteins using MS-compatible concentrations of ammonium acetate preserving protein structure rather than promoting retention. We have shown for the first time the MS-compatible HIC method, separating a complex E. coli cell lysate in only 25 min (Figure 4C), demonstrating the potential of online HIC-MS for both qualitative and quantitative top-down proteomics.¹³⁰ The stabilizing effect of the volatile salt and limited time on column resulted in native-like charge state distributions for proteins, aiding the detection of large protein species up to 206 kDa in the cell lysate.¹³⁰ Further development of HIC-MS by Chen et al. separated two intact IgG species, offering great potential for HIC-MS as a biotherapeutic characterization technique.²³

There is also significant interest in the development of capillary electrophoresis-MS (CE-MS) as a non-denaturing separation technique.^132–135 The wide variety of orthogonality of separation selectivity to conventional LC-MS methods and low sample volume requirements have made CE-MS a very attractive technique for native separations.^136,137 As new commercial systems and discrete devices are developed,^138,139 we anticipate that CE-MS will rise to increased prominence in top-down proteomics.

In addition to the growing number of liquid-phase separation techniques, the use of gas-phase ion mobility to separate intact proteins prior to top-down MS has gained significant attention recently.^140–142 Ion mobility spectrometry (IMS) is a gas-phase separation with selectivity based on a molecule’s drift time or mobility, a unique property of each molecule based on how it behaves in the drift tube.¹⁴³ IMS has experienced continual growth throughout the years, yielding novel techniques such as drift tube ion mobility, traveling wave ion mobility, trapped ion mobility, field asymmetric ion mobility, and differential mobility ion mobility among others.¹⁴³ High-resolution ion mobility separation has the potential to rapidly separate proteoforms with high sequence homology with limited front-end separation. Recently, the Ruotolo group has exploited traveling wave ion mobility for the top-down characterization of chemically derivatized native protein molecules.¹⁴⁴ We believe advances in ion mobility technologies will attract enormous attention from the top-down proteomics community in the future as new methods and applications showing detailed molecular characterization are established.

Unlike peptides, proteins have a much more diverse range of physiochemical properties; therefore, MDLC by combining multiple orthogonal separation modalities is necessary to separate various proteins in a proteome.^5,112,145 Although extensively used in bottom-up separations, the use of two-dimensional LC (2DLC) coupled to MS in top-down proteomics was more recent.^29,146,147 Much of the 2DLC-MS work for top-down proteomics has used a comprehensive “offline” coupling strategy, collecting fractions across the first dimension of separation (¹D) and then analyzing all fractions using the second dimension of separation (²D) interfaced directly with the MS. Offline ¹D separation allows samples to be buffer-exchanged or concentrated between dimensions, allowing the coupling of MS-incompatible separation techniques. Our group has developed an offline 2DLC strategy coupling HIC and RPLC (Figure 4D).¹⁴⁶ The orthogonality of HIC and RPLC separations greatly extended the range of separable protein polarities in an E. coli cell lysate.¹⁴⁶ To enable better characterization of large proteoforms, we used a 2DLC approach coupling sSEC to separate proteins by size and then used RPLC-MS to separate by hydrophobicity.²⁹ We further developed three-dimensional liquid chromatography (3DLC), coupling HIC-IEX-RPC with offline ¹D HIC and ²D IEX separation before ³D online RPLC-MS (Figure 4D).¹⁴⁸ We achieved a 14-fold improvement in protein identifications of a human embryonic kidney cell lysate using 3D HIC-IEX-RPLC-MS compared to 2D IEX-RPLC-MS.¹⁴⁸

Though the advantages of MDLC are clear, the process is time-consuming and labor-intensive. Online MDLC strategies can alleviate this problem through automation, using a valve interface to automate transfer of eluent from ¹D to ²D.¹⁴⁵ While many coupling techniques for online 2DLC exist, the comprehensive mode, which transfers all eluent from ¹D to ²D, has the greatest potential to aid global top-down proteomics. Several groups have employed comprehensive online 2DLC for top-down proteomic analyses.^28,149,150 Moore and Jorgenson developed an online comprehensive 3DLC system for peptide separation, which has not yet been extended for use with intact proteins.¹⁵¹ In the coming years, we anticipate the power of MDLC and the attraction of automation will lead to new combinations of separation modalities. We also expect the use of more non-denaturing techniques in separation strategies as the interest in native MS grows within the top-down community.

5. ADDRESSING THE CHALLENGE IN TOP-DOWN PROTEOMICS DATA ANALYSIS

Analysis of top-down proteomics data can be complex and labor-intensive, and represents a major barrier for new users.^5,152 Thus, software solutions capable of deconvoluting spectral data and searching generated results against a database have been developed. Top-down proteomics software, ProSight PC, was first introduced by the Kelleher group in 2001, and several generations of this software have been developed since its inception.^153–155 Moreover, the MS-Align¹⁵⁶ and TopPIC¹⁵⁷ database search algorithms introduced by Liu and colleagues have proven to be valuable tools for top-down proteomics data analysis. Recently, Informed Proteomics, a software developed by Park et al., is an open-source top-down proteomics software that uses a LC-MS feature-finding algorithm (ProMex) and a database search algorithm (MSPathFinder).¹⁵⁸ Other database search algorithms frequently used for top-down data analysis include pTOP,¹⁵⁹ Mascot Top Down,³² and Proteoform Suite.¹⁶⁰ Nonetheless, most data analysis tools for top-down proteomics either focus on protein identification, lack a user-friendly interface, or are proprietary to instrument vendors and thus reduce accessibility to users. To address these challenges, our group developed MASH Suite Pro, an open-source software package for top-down proteomics that combines data analysis tools including protein identification, quantification, and characterization into a simple package.^152,161 However, this software is limited to processing Thermo raw data files and uses one deconvolution and one search algorithm.

To improve flexibility in data processing and expand access to multiple vendor formats, our group created MASH Explorer (Figure 5A), a top-down proteomics software that processes targeted- and discovery-based MS data for unambiguous proteoform characterization and identification (Figure 5A–C).³¹ This software incorporates several deconvolution and database search algorithms in a user-friendly software environment and can process data from multiple vendor formats.³¹ Moreover, it provides visualization tools for simple validation and correction of computational outputs, which is an essential step to ensure the reliability of proteomics data.³¹ MASH Explorer has proven to be an indispensable tool for the top-down proteomics community, with a global user base of 2086 MASH users as of March 1, 2021 (64.5% from North America, 0.6% from South America, 17.6% from Europe, and 7.8% from Asia) (Figure 5D).

Figure 5.

MASH Explorer as a universal software environment for top-down proteomics. (A) Schematic of the various MASH Explorer functions for proteomics data processing. The main functions of MASH Explorer include data import, spectral deconvolution, workflow automation, data validation, protein identification, and graphical output. MASH Explorer utilizes a new data processing module based on the ProteoWizard Library to accept various data input file formats from major instrument vendors (e.g., Thermo, Bruker, and Waters). Raw MS and MS/MS data files are then processed by deconvolution algorithms (i.e., MS-Deconv, TopFD, eTHRASH, pParseTD, Flashdeconv, and UniDec) and database search algorithms (i.e., MS-Align+, TopPIC, pTop, and MSPathFinderT). (B) Illustration of “Discovery mode” for LC-MS/MS data processing. “Discovery mode” can handle batch LC-MS/MS raw data files and includes features such as data import, data processing (deconvolution and database search), and data validation for protein identification. (C) Illustration of the “Targeted Mode” workflow for MASH Explorer. The “Targeted Mode” workflow includes data import, spectral deconvolution to identify and verify isotopic distributions, database search based on identified isotopic distributions, and proteoform characterization by matching identified isotopic distributions to the target proteoform sequence. (D) Cartoon schematic of a “world map” featuring the location distribution of MASH users across the globe. There are currently 2086 active users (March 1, 2021) with ~65% of users from North America, ~18% from Europe, and ~8% from Asia. Figure updated and adapted from ref 31. Copyright 2020 American Chemical Society.

Importantly, MASH Explorer has the flexibility to incorporate new and updated software developed by computational scientists. McIlwain et al. developed an ensemble machine learning technique that can process and combine the resultant peak lists from multiple deconvolution algorithms, thus improving the accuracy and confidence in protein identifications from database searches.¹⁶² Hence, we are currently incorporating this machine learning method in the next generation of MASH Explorer to enhance proteoform identification and characterization. We have also recently integrated the FlashDeconv deconvolution algorithm¹⁶³ to MASH Explorer. FlashDeconv is a rapid deconvolution algorithm, outpacing other deconvolution tools in the top-down proteomics field by applying a logarithmic transformation to the mass-to-charge ratio allowing for spectral pattern matching. Using this approach, Jeong et al. observed more detected masses and fewer deconvolution artifacts compared to other methods used. With the addition of FlashDeconv, the throughput of MASH Explorer has been greatly improved.

Outside of being a universal and user-friendly software environment for top-down proteomics, future generations of MASH Explorer will be made amenable to other proteomics applications such as native top-down proteomics. In particular, the incorporation of UniDec¹⁶⁴ into MASH Explorer will be an exciting feature for the native MS community. UniDec developed by Marty et al. is a universal deconvolution algorithm that applies a Bayesian algorithm to separate the mass and charge dimension of a complex spectrum.¹⁶⁴ This software provides fast and robust deconvolution of both isotopically and non-isotopically resolved mass spectra while also incorporating ion mobility-mass spectra. Adding UniDec into the MASH Explorer software is anticipated to enable native protein identification with simultaneous visualization of protein fragment ions.

As the top-down proteomics community continues to grow, the need for universal, comprehensive, and globally accessible top-down proteomics software will increase tremendously. Future software will also need to address the challenges in the identification and characterization of large proteoforms because of the difficulties in obtaining sufficient MS/MS sequence coverage. Combination of multiple MS/MS techniques including but not limited to activated ion electron capture¹⁶⁵/transfer dissociation,¹⁶⁶ and more recently developed ultraviolet photodissociation,¹⁶⁷ and electron ionization dissociation,^168,169 will likely have great potential for proteoform characterization and therefore will require the necessary software tools for effective data analysis to be available to the top-down proteomics community.

6. ADDRESSING THE CHALLENGE IN LINKING PROTEOFORMS TO FUNCTION

Proteins are the primary effectors of function in biology, and thus, the top-down proteomics field is actively seeking to link proteoform-resolved information with functional characteristics in various biological systems.¹⁷⁰ Chamot-Rooke et al. revealed the functionally critical PilE proteoforms carrying phosphoglycerol groups on serines were tightly associated with crossing of the epithelial barrier and access to the bloodstream in cerebrospinal meningitis (N. meningitidis).³³ Paša-Tolić and co-workers showed a unique protein S-thiolation switch as a result of infection-like conditions in Salmonella typhimurium.³⁶ Carel et al. employed top-down AP-MS to decipher the functional role of O-mycoloylation in targeting the outer membrane proteins to the mycomembrane in bacteria.¹⁷¹ Kelleher and co-workers demonstrated the crosstalk between mutation and PTMs after immunoaffinity purification and top-down MS analysis of the KRAS4b proteoforms, which provides new insights in KRAS biology in cancer.⁴⁰ Agar and co-workers applied top-down MS to ALS-patient immunopurified super-oxide dismutase-1 (SOD1) and revealed that, among the three prevalent SOD1 proteoforms, one had a neurotoxic function and another one had a neuroprotective function.¹⁷²

Our group has previously combined top-down proteomics and functional measurements to study sarcopenia, the deterioration of muscle mass and strength as a result of aging, a major burden on the healthcare system, but the molecular mechanism underlying sarcopenia-associated muscle dysfunction remains largely unknown.^39,173 We uncovered a progressive age-related decrease in the phosphorylation of myosin regulatory light chain (RLC), a critical regulatory protein in muscle contraction, in the skeletal muscle of aging rats using top-down targeted proteomics (Figure 6A).³⁹ The sites of decreasing phosphorylation were localized to Ser14/15, of which Ser14 phosphorylation was a previously unidentified phosphorylation site in RLC from fast-twitch skeletal muscle. The functional measurement of the mechanical properties of single fast-twitch fibers indicated that age-related decreases in the contractile properties of sarcopenic fast-twitch muscles could be accounted for by the decreased RLC phosphorylation. This study provides strong evidence for the functional role of RLC phosphorylation at Ser14/15 in sarcopenia-related muscle dysfunction and the potential therapeutic modulation of RLC phosphorylation as a potential avenue to treat sarcopenia.³⁹ In another study, we developed a top-down proteomics method that allowed for simultaneous quantification of sarcomeric proteoforms and contractile measurements from rat fast- and slow-twitch skeletal muscle tissues to correlate muscle function with sarcomeric proteoform expression.¹⁷³ Importantly, there was a connection with the sarcomeric proteoforms found in the aging rats, which was consistent with the functional phenotype. Age-related changes were observed in the phosphorylation levels of Z-disc proteins, which represented the first report of PTMs in the Z-disc associated with aging. The use of proteoform measurements with muscle contraction properties will continue to prove to be essential in the relationship between proteoforms and function in human muscle diseases.

Figure 6.

Linking protoeoforms with biological function using top-down proteomics. (A) Integrated approach combining top-down targeted proteomics with mechanical measurements to elucidate the molecular mechanism(s) underlying age-related sarcopenia. This approach includes the following: (1) use of a rat model of age-related sarcopenia; (2) isolation of skeletal muscle for proteomic and mechanical analyses; (3) top-down targeted proteomics for RLC proteoform analysis; (4) MS-based proteoform quantification; (5) MS/MS analysis for the comprehensive characterization of RLC proteoform sequences and PTMs; (6) mechanical measurements on single fibers; and (7) correlation of the targeted proteomics data with functional data to explain the sarcopenic phenotype. mo, month. Figure adapted from ref 39. Copyright 2016 American Chemical Society. (B) Schematic of integrated functional assessments and the top-down proteomics workflow for the same hiPSC-ECT. (1) hiPSCs are differentiated into CMs and CFs which are used to generate hiPSC-ECTs. (2) Functional assessments are performed on the hiPSC-ECTs to measure the isometric twitch force. (3) Sarcomeric proteins are extracted via a dual extraction method from the functionally tested hiPSC-ECTs. (4) Top-down proteomics is performed on the functionally tested hiPSC-ECTs. (5) Integrated assessment of hiPSC-ECT constructs. Figure adapted from ref 45. Copyright 2021 American Chemical Society.

Recently, our lab has been developing methods that integrate omics data with functional properties of human cardiomyocytes differentiated from induced pluripotent stem cells (hiPSC-CM).^43,45 We established an unbiased method combining proteomics with functional measurements for the accurate and comprehensive assessment of hiPSC-CM maturation.⁴³ In addition to the detection of known maturation-associated contractile protein alterations, for the first time, contractile protein PTMs such as decreased phosphorylation of α-tropomyosin were identified as potential novel hiPSC-CMs maturation markers. While most hiPSC-CM research is performed using 2D monolayer systems, there has been a significant effort to develop three-dimensional (3D) engineered cardiac tissues (ECTs) because ECTs more faithfully reflect the relevant cardiac physiology of endogenous cardiac tissues.⁴⁵ However, many of the metrics used to assess hiPSC-ECTs are often performed on separate samples due to incompatibilities between the various measurements, which is problematic due to the highly heterogeneous nature of stem-cell-derived CMs. To address this issue, Melby et al. developed a functionally integrated top-down proteomics method for a standardized assessment of hiPSC-ECTs (Figure 6B).⁴⁵ Isometric twitch force measurements were performed on two groups of hiPSC-ECTs, and using the same samples, top-down proteomics was implemented to analyze sarcomeric proteoforms. This integrated approach allowed for the determination of direct correlations between functional properties and sarcomeric proteoforms. We envision that this method will be highly valuable in future studies using hiPSC-ECTs for disease modeling, cardiotoxicity screenings, and regenerative therapies. Although all studies have linked proteoforms with functions in various biological systems and diseases to some extent, there remains a challenge to establish a direct causal relationship between proteoform changes and the corresponding functional properties to further understand the cause and effect relationship.

7. THROUGHPUT OF TOP-DOWN PROTEOMICS FOR PRECISION MEDICINE

Precision medicine, wherein disease treatment and prevention consider individual variability, is increasingly recognized as the future of modern health care.^174,175 High-throughput omics technologies are becoming enabling forces for precision medicine with the promise to characterize patients at an extraordinarily detailed molecular level.^176,177 In the post-genomics era, a comprehensive analysis of proteoforms including PTMs and mutations/alternative splicing is essential to gain a transformative understanding of disease mechanisms and identifying new therapeutic targets.^3,178–180

While top-down proteomics is not yet fully incorporated in the clinic, there are several translational studies that eloquently demonstrate the power that this technology could provide for clinical applications.^{4,30,34,40–42,48} Top-down proteomics has been applied to important clinical samples, such as cardiac,^34,44 liver,¹⁸¹ and tumor⁴⁰ biopsy samples, blood,⁴¹ and serum,³⁰ to understand the role of proteoforms in human disease. For example, Kelleher and co-workers applied top-down proteomics for the analysis of proteoform signatures from acute rejection fractions of circulating tumor cells and determined there was a distinct proteoform signature between patients that received a liver transplant versus the control group.¹⁸¹ Our group has linked altered cardiac proteoforms to disease phenotypes in heart failure using both animal models and human clinical samples.^34,35,38 Phosphorylated proteoforms of cardiac troponin I (cTnI), a gold standard biomarker for detection of cardiac injury, were identified by top-down quantitative proteomics to be candidate biomarkers for chronic heart failure.³⁴ Top-down proteomics has also identified actin isoforms as a potential cardiac disease marker¹⁸² and discovered novel phosphorylation of a critical Z-disc protein, Enigma Homologue isoform 2 in acute myocardial infarction.³⁸

Recently, we applied top-down proteomics to septal myectomy samples collected from hypertrophic cardiomyopathy (HCM) patients with severe outflow tract obstruction and revealed a shared sarcomeric proteoform pattern across HCM tissues when compared to non-failing donor heart tissues.⁴⁴ HCM is a genetically heterogeneous disease with nearly 1400 mutations identified in at least 11 genes that encode proteins within the sarcomere.¹⁸³ However, regardless of various HCM-causing mutations, the detected sarcomeric proteoform profiles of these samples were highly consistent. This finding suggests that common signaling pathways are activated in the myocardium of obstructive HCM patients with similar phenotypes requiring septal myectomy intervention and opens the door for future therapeutic interventions targeting HCM phenotypes, which are more accurately reflected by the proteoform profiles rather than the specific genotype.⁴⁴ These various examples demonstrate the importance of proteoform-level molecular details, which provide critical insights into disease progression, in a variety of clinical research areas.

To fully reach the potential of top-down proteomics in precision medicine and translate to the clinic, several challenges mentioned in this perspective must be addressed. Future technology developments that can enable the routine proteoform-resolved characterization of proteins >50 kDa are of particular interest, as over half of the proteome contains proteins of this size.²⁹ To develop a top-down proteomics assay in the clinic, it is essential to analyze biofluids, such as blood and serum, which still suffer from the immense challenge of finding the low-abundance biomarkers of interest among an ocean of other highly abundant blood proteins in the vast biological milieu.³⁰ Important progress has been made with standardized top-down proteomics protocols for the analysis of peripheral blood mononuclear cells and other common biofluids; however, challenges related to proteome complexity remain.⁴² We anticipate that novel approaches such as nanoproteomics,³⁰ and other enrichment techniques, will prove to be critical to addressing this major challenge. Moreover, rigor and reproducibility need to be established to ensure robust and unbiased experimental design, analysis, interpretation, and reporting of results, which is especially critical for the further development of top-down quantitative proteomics toward its clinical use. Top-down proteomics recently has made major strides in the reproducible and quantitative analysis of proteforms.^{4,43,44,48,106} Arguably, it has great potential to be more rigorous and reproducible than other approaches such as conventional bottom-up MS- and antibody-based techniques. Nonetheless, further method development and technology improvements will be required to extend the application of top-down MS-based quantitation toward complex PTMs or heavily modified proteins.

Another pressing issue as it pertains to the utility of proteoform analysis in clinical applications is the throughput of current top-down proteomics workflows. Great progress has been made in high-throughput clinical applications for bottom-up proteomics pioneered by several groups such as Regnier,¹⁸⁴ Van Eyk,^185,186 and Carr.¹⁸⁷ Collectively, these studies demonstrate both the feasibility and benefits of using proteomics for the analysis of biomolecules from clinical samples. However, except for matrix assisted laser desorption ionization-TOF-based intact protein assays,^7,188 the typical discovery-based top-down proteomics still suffers from low throughput and there is an urgent need to optimize the entire workflow including sample preparation, separation, data acquisition, and analysis. To speed up sample preparation, the use of automation has the potential to revolutionize the application of top-down proteomics in a clinical setting by greatly enhancing accuracy, reproducibility, and time spent in preparing samples. Although online LC methods are common-place in current top-down MS workflows, given the high speed of ion mobility separations,¹⁴³ further technology development with improved front-end ion mobility separations may enable rapid and robust protein separation. Similarly, the Evosep system implements sensitive nanoflow liquid separations using a rapid reverse phase gradient proving to be a valuable clinical tool.^189,190 We anticipate a greater array of new column chemistry will enable this technology to be applied for proteoform-resolved clinical sample analysis. Furthermore, for discovery-based top-down proteomics experiments, the process of deconvoluting the highly complex data and then searching the peak list against a database generally takes several hours to fully complete. Discovery-based searches are far slower than target-based searches but do not require a priori knowledge of the proteoform of interest, which for some clinical applications is not a possibility. Conceivably, top-down targeted proteomics focusing on one or multiple protein targets might have great advantages to be used in the clinic. Currently, the nanoproteomics cTnI-NP assay and the HCM clinical patient analysis by top-down proteomics require ~2.5 h total analysis time, when accounting for sample preparation, workup, and final data acquisition.^30,44 We anticipate future technology developments in both software and hardware will enhance the throughput of top-down proteomics, enabling the use of top-down proteomics for precision medicine and translation to the clinic in the near future.

8. CONCLUSION AND FUTURE DIRECTIONS

In this Account & Perspective, we have provided our view of the major challenges affiliated with top-down proteomics at present and new strategies (including those already developed and to be developed) to address these challenges. We have focused on the most pressing areas including protein solubility, dynamic range of the proteome, proteome complexity, data analysis, linking proteoforms to biological function, and throughput of top-down proteomics for precision medicine. There are also other ongoing challenges associated with top-down proteomics such as the sensitivity, rigor, and reproducibility of top-down proteomics. As an invited contribution for the Special Focus Issue of JASMS that is dedicated to Ying Ge in honor of her 2020 Biemann Medal, this Account & Perspective focused more on the contributions of our laboratory to the rapidly growing top-down proteomics field. We greatly appreciate and value the contributions by other top-down proteomics laboratories all over the world, some of which are also discussed here, but it is not possible to comprehensively review all advances in this single Account & Perspective. We envision that the Consortium of Top-down Proteomics (www.topdownproteomics.org) will continue to develop and grow as a community and foster international collaborations to further advance top-down proteomics. It is also essential to establish interdisciplinary teams with diverse expertise in different fields to continuously address the remaining challenges for top-down proteomics to reach its full potential. We firmly believe that future advancements in the many strategies discussed herein, together with the continued improvement in MS instrumentation, device miniaturization to improve sensitivity, automation for high-throughput sample processing, data acquisition, and analysis, will enable the ubiquitous application of top-down proteomics in biomedical research for precision medicine.