Mass spectrometry-based targeted proteomics for analysis of protein mutations

Abstract

Cancers are caused by accumulated DNA mutations. This recognition of the central role of mutations in cancer and recent advances in next-generation sequencing, has initiated the massive screening of clinical samples and the identification of 1000s of cancer-associated gene mutations. However, proteomic analysis of the expressed mutation products lags far behind genomic (transcriptomic) analysis. With comprehensive global proteomics analysis, only a small percentage of single nucleotide variants detected by DNA and RNA sequencing have been observed as single amino acid variants due to current technical limitations. Proteomic analysis of mutations is important with the potential to advance cancer biomarker development and the discovery of new therapeutic targets for more effective disease treatment. Targeted proteomics using selected reaction monitoring (also known as multiple reaction monitoring) and parallel reaction monitoring, has emerged as a powerful tool with significant advantages over global proteomics for analysis of protein mutations in terms of detection sensitivity, quantitation accuracy and overall practicality (e.g., reliable identification and the scale of quantification). Herein we review recent advances in the targeted proteomics technology for enhancing detection sensitivity and multiplexing capability and highlight its broad biomedical applications for analysis of protein mutations in human bodily fluids, tissues, and cell lines. Furthermore, we review recent applications of top-down proteomics for analysis of protein mutations. Unlike the commonly used bottom-up proteomics which requires digestion of proteins into peptides, top-down proteomics directly analyzes intact proteins for more precise characterization of mutation isoforms. Finally, general perspectives on the potential of achieving both high sensitivity and high sample throughput for large-scale targeted detection and quantification of important protein mutations are discussed.

1. INTRODUCTION

Cancers are caused by accumulated DNA mutations (Bignell et al., 2010; Cairns, 1975; Hanahan and Weinberg, 2000; Hanahan and Weinberg, 2011), which can contribute to cancer by activating protein function (oncogenes) or inactivating protein function (tumor suppressors). The recent advances in next-generation sequencing applied to large numbers of clinical samples has already led to the identification of 1000s of mutations (Shu et al., 2017). They can be divided into neutral (passenger) and disease-associated (driver) mutations. The passenger mutations occur spontaneously in somatic cells and have little or no apparent functional consequence on tumor progression, while the driver mutations play a key role in the malignant transformation and tumor progression (Kumar et al., 2019). However, it is challenging to sort out the cancer-associated mutations with functional significance (Manolio et al., 2009).

Functions encoded in the genome are generally executed by proteins; >2/3 of FDA-approved drugs directly target proteins (Santos et al., 2017). Mutation can affect protein folding and stability, protein function, subcellular localization, and protein-protein interactions (Reva et al., 2011; Sheynkman et al., 2016; Sheynkman et al., 2014; Smith et al., 2013). Protein mutations are not simply associated with tumor cells but actually responsible for altered cellular function and aberrant signaling pathways. Thus, they play a major role in cancer development for tumorigenesis, progression and therapy resistance (Reva et al., 2011; Smith et al., 2013; Tan et al., 2020). Furthermore, mutations at the genome level may not be readily reflected at the protein level. Many studies have revealed that mRNA abundance is poorly correlated with protein abundance across different types of tumor tissues (Lorentzian et al., 2019; Rodland et al., 2018; Sinha et al., 2019; Zhang et al., 2019). This strongly suggests that protein translation and degradation are tightly regulated and play a critical role in determining gene function, and mRNA may lack a direct link to protein activity and gene function. In addition, protein mutations produced only by tumor cells can also provide a unique opportunity for cancer biomarker development (Jensen et al., 1998; Nadal et al., 2014; Nishikawa et al., 2009; Wang et al., 2011). Therefore, direct analysis of protein mutations is critically important for discovery of functionally important mutations to improve our understanding of tumorigenesis, progression and therapy resistance (Reva et al., 2011) and to provide new therapeutic targets (Lima et al., 2019).

With recent technological advances in sample preparation and liquid chromatography–mass spectrometry (LC-MS) instrumentation, MS-based proteomics has emerged as a powerful tool for quantitative genome-scale profiling of proteome and precise comprehensive characterization of proteoforms with posttranslational modifications (PTMs) (e.g., phosphorylation, acetylation, and glycosylation) (Mertins et al., 2018; Rodland et al., 2018; Rodriguez and Pennington, 2018; Sheynkman et al., 2016; Zhang et al., 2019). The deep proteomic analysis is complementary to available genomic (and transcriptomic) analysis by providing additional biological insights that would have been difficult or impossible to obtain solely through genomics approaches (e.g., protein PTM analysis) (Kiseleva et al., 2018; Latysheva and Babu, 2016; Nadal et al., 2014; Smith et al., 2013). With continuous contributions by the NCI CPTAC (Clinical Proteomic Tumor Analysis Consortium) program, integrated analysis of proteomic and genomic data (i.e., proteogenomic analysis) has recently become an emerging trend with many high-impact papers for characterization of various cancer types (Chen et al., 2020; Dou et al., 2020; Gao et al., 2019; Gillette et al., 2020; Jiang et al., 2019; Mertins et al., 2016; Vasaikar et al., 2019; Wang et al., 2021; Zhang et al., 2014; Zhang et al., 2016). Proteogenomic analysis has been demonstrated to provide a more comprehensive understanding of disease biology and new insights into the functional regulation, which underscores the importance of comprehensive proteome and PTM characterization (Dou et al., 2020; Mertins et al., 2016; Vasaikar et al., 2019; Wang et al., 2021; Wisniewski et al., 2009; Zhang et al., 2014; Zhang et al., 2016). Nevertheless, the use of unbiased “discovery” MS-based proteomics platforms for large-scale analysis of mutation proteoforms has not been very effective thus far.

There are three common types of protein mutations encoded by mutated genes (Kiseleva et al., 2018; Latysheva and Babu, 2016; Reva et al., 2011; Vegvari, 2016; Wang et al., 2011): single nucleotide variants (SNVs), alternative splicing variants (ASVs), and gene fusion products. Coding SNVs at the protein level can be classified into three types: synonymous (sense) with normal function, which does not change the corresponding amino acid; nonsense with truncated form, which introduces a premature stop codon; nonsynonymous (missense) with mutation, which changes the corresponding amino acid (i.e., single amino acid variant (SAAV) or single amino acid polymorphism (SAP) (Vegvari, 2016). Mass spectrometry (MS)-based global proteomics has been attempted for deep profiling of protein mutations. However, only ~4–10% of single nucleotide variants detected by both DNA and RNA sequencing were observed as SAAVs (Rodland et al., 2018; Ruggles et al., 2016; Zhang et al., 2019). Very recently, Lubman’s group has applied three different global proteomics strategies for deep profiling of mutant peptides in PANC-1 cell line (Tan et al., 2020). Only ~540 SAAVs have been identified and which SAAVs are functionally important is largely unknown. Nevertheless, this study represents the most comprehensive discovery of SAAVs in single cell line. In addition, the same group has performed global proteomics analysis of subpopulations of breast cancer stem cells to study SAAVs and their relation to breast cancer (Tan et al., 2017). 374 unique SAAVs were identified in total, where 27 SAAVs are cancer-related SAAVs and 135 SAAVs were only found in the cancer stem cell population when compared to mature luminal cells. The low success rate can be attributed to lacking generic databases for sequence search, higher false discovery rates from an enlarged sequence search space, and mutant peptides often present at a low abundance resulting in escaping MS/MS sequencing (Rodland et al., 2018; Sheynkman et al., 2016; Zhang et al., 2019). Targeted proteomics has been broadly used for validation of mutations from global discovery (Lichti et al., 2015; Mostovenko et al., 2018; Tan et al., 2020; Vegvari, 2016) and for accurate quantification of mutations which are cancer-specific markers (Lesur et al., 2015; Wang et al., 2011) because it has higher sensitivity than global proteomics (Shi et al., 2012) and avoids complex database search (Figure 1).

FIGURE 1.

Schematic diagram for large-scale discovery of protein mutations by bottom-up MS-based proteomics (black arrow: global proteomics for analysis of protein mutations; blue arrow: targeted proteomics for analysis of protein mutations; dash black arrow: validation of global discovery with targeted proteomics).

Two types of MS-based global proteomics are used for global discovery: the commonly used data-dependent acquisition (DDA) and the more recently implemented data-independent acquisition (DIA) (Aebersold and Mann, 2016). However, it remains a challenge for reproducible detection and quantification of surrogate peptides from protein mutations because they often have low expression levels and happen at low frequency rates. For example, in DDA the MS randomly samples peptides for fragmentation and is biased to frequently pick the peptides with higher abundance, then leading to a bias against low abundance mutant peptides. Finally, the enzymatic digestion for shotgun proteomics also presents an issue in terms of sequence coverage since mutation containing peptides based on enzymatic cleavage patterns are sometimes simply too short or too long to be identifiable.

In contrast, MS-based targeted proteomics has the potential for precise broad targeted discovery of functionally important protein mutations because it selects the predefined targeted peptides without bias selection, and it has higher detection sensitivity, quantitation accuracy (accurate or absolute quantification), and reproducibility (≤10% CV) (Shi et al., 2016; Shi et al., 2012) with the same level of detection specificity (easy to distinguish SAAVs (Lesur et al., 2015; Lichti et al., 2015; Tan et al., 2020; Wang et al., 2011) as global proteomics (Figure 1). Furthermore, the shortcoming of targeted proteomics, lower multiplexing for target analytes, has recently been addressed effectively with enabling simultaneous quantification of ~1,000 analytes in a single analysis (Stopfer et al., 2021).

In this review we provide an overview of recent advances in bottom-up targeted proteomics and highlight its broad applications for identification and quantification of protein mutations in human bodily fluids, tissues, and cell lines. Furthermore, we briefly review recent applications of top-down targeted proteomics for analysis of protein mutations due to its unique ability for precise characterization of intact protein mutation isoforms. Finally, future perspectives in targeted proteomics for large-scale targeted discovery of functionally important protein mutations as biomarkers and therapeutic targets are discussed.

2. RECENT TECHNOLOGICAL ADVANCES IN TARGETED PROTEOMICS

Targeted proteomics can be divided into two groups: selected reaction monitoring (SRM), also known as multiple reaction monitoring (MRM), and parallel reaction monitoring (PRM). SRM is a classic targeted proteomics approach performed on a triple quadrupole mass spectrometer (QqQ MS). It exploits the unique features of QqQ with two levels of mass selection (i.e., Q1 and Q3 isolation of precursor ions and their product ions, respectively) and a relatively long dwell time (in general 10 ms per pair of precursor/product ions) over a narrow m/z window (± 0.02 m/z), which result in significantly improved selectivity and sensitivity, at least one to two orders of magnitude higher than full scan global proteomics analysis (Lange et al., 2008; Shi et al., 2016; Shi et al., 2012). The high specificity, quantitation accuracy (accurate or absolute quantification), and reproducibility make it ideally suitable for studying protein mutations.

PRM is conceptually similar to SRM (Shi et al., 2016). The major difference between PRM and SRM is that in PRM full MS/MS spectra are acquired for each precursor in the high-resolution accurate-mass Orbitrap mass analyzer, while for SRM the predefined product ions are monitored by the low-resolution quadrupole mass analyzer. The principle and performance of SRM and PRM have been well documented with systematic performance comparison (Kockmann et al., 2016; Ronsein et al., 2015; Shi et al., 2016). Compared to antibody-based approaches, SRM/PRM has higher specificity for site-specific detection and quantification of protein mutations. Recent advances in targeted proteomics detection sensitivity and analyte multiplexing capability (e.g., simultaneous quantification of 100s of proteins (Lee et al., 2020; Shi et al., 2012) and phosphorylation sites (Stopfer et al., 2021) enable researchers to perform discovery and validation of 100s-1000s of low-abundance protein mutations.

2.1. Detection sensitivity

Direct LC-S/PRM can provide detection sensitivity for reproducible quantification of target proteins in the concentration range of 4–5 orders of magnitude. However, such sensitivity is insufficient to cover the entire proteome (e.g., protein mutations at the low expression level) with up to 7 and 11 orders of magnitude for mammalian cell lines and human plasma/serum, respectively. Therefore, different front-end sample processing strategies have been explored to greatly enhance detection sensitivity: target analyte enrichment with affinity reagents and sample fractionation to enrich target analytes and reduce complexity (Figures 2 and 3).

FIGURE 2.

Antibody-based affinity enrichment for improving targeted proteomics detection sensitivity: (A) Protein immunoaffinity capture (MSIA-SRM); (B) Peptide immunoaffinity capture (SISCAPA-SRM); (C) Sequential immunoaffinity capture (Immuno-SRM).

FIGURE 3.

High-resolution LC-based enrichment for improving targeted proteomics detection sensitivity: (A) PRISM-SRM; (B) DD-SRM; (C) T-μLC-SRM. Reprinted with permission from Zhang et al. Anal Chem 2018, 90(3):1870–1880 for Figure 3C. Copyright 2018 American Chemical Society

2.1.1. Antibody-based affinity enrichment

A. Protein immunoaffinity capture

With antibody-based immunoaffinity (IA) enrichment of target proteins or peptides, immuno-SRM can be used for sensitive reproducible quantification of target proteins from complex biological samples. There are three types of immuno-SRM assays: protein IA capture, peptide IA capture, and sequential IA capture (Figure 2). Among them, the protein IA capture is the most commonly used method for identification and quantification of protein mutations because the common shared epitopes recognized by commercially available antibodies that are originally raised against the wild type proteins can be used for enrichment of their corresponding mutant proteins. For example, an anti-SPINK1 antibody was used for effective enrichment of SPINK1 mutant proteins (N34S and P55S), followed by LC-SRM analysis. Immuno-SRM enabled to detect SPINK1 mutations at the limit of detection (LOD) of 0.5 ng/mL in patient sera at different disease conditions (Ravela et al., 2018) (Figure 2A).

B. Peptide immunoaffinity capture

The peptide IA capture is commonly referred as stable isotope standards and capture by antipeptide antibodies (SISCAPA) (Anderson et al., 2004; Anderson et al., 2009; Whiteaker et al., 2010; Whiteaker et al., 2012) (Figure 2B). When compared to the protein IA capture, SISCAPA has an advantage to reduce background interference more significantly with selective enrichment of target peptides of interest rather than all the peptides from target proteins. It can provide an average of ~100-fold enhancement in detection sensitivity of targeted proteomics (Anderson et al., 2004). Potential clinical applications of SISCAPA-SRM have been illustrated by implementation of SISCAPA in a clinical laboratory environment for quantification of low-abundance plasma biomarker proteins (Whiteaker and Paulovich, 2011). Another advantage over the protein IA capture is that harsh sample processing conditions can be used for SISCAPA without the need to maintain native protein epitopes for protein enrichment. SISCAPA-SRM was applied for targeted quantification of membrane proteins in tissue homogenates (Fan and Neubert, 2016) and plasma biomarkers in the presence of autoantibodies (Hoofnagle et al., 2008; Kushnir et al., 2013). In addition, Katafuchi et al., developed a highly sensitive SISCAPA-SRM assay for enabling detection of FGF15 at LOD of 0.1 ng/mL in mouse plasma (Katafuchi et al., 2015).

C. Sequential immunoaffinity capture

To further increase immuno-SRM detection sensitivity, sequential IA enrichment, first targeting protein and then surrogate peptide for enrichment, can be developed for detection and quantification of extremely low-abundance proteins (Figure 2C). One example is to measure β-nerve growth factor (β-NGF) in human serum. β-NGF protein was first isolated from patient serum using an anti-protein antibody. Subsequently, the peptide IA capture was performed to further reduce background interference using an antibody against the surrogate peptide for the β-NGF protein. As a result, reliable SRM quantification of β-NGF at the levels of ~7–450 pg/mL in human serum can be achieved with the inter-assay CV of <15% (Neubert et al., 2013). The sequential enrichment was demonstrated to have additional >10-fold improvement in detection sensitivity when compared to single-step immuno-SRM analysis.

2.1.2. High-resolution LC-based enrichment

A. PRISM-SRM

Antibody-based immuno-SRM is attractive for enhanced detection sensitivity, however it shares similar limitations as other antibody-based approaches (e.g., unavailability of antibodies for new proteins and the long lead time and high cost for the development of high-quality antibodies). To address this issue, we have developed an antibody-independent high-resolution LC-based approach termed PRISM (high-pressure, high-resolution separations coupled with intelligent selection and multiplexing) for highly sensitive SRM quantification of target proteins in complex biological samples (Figure 3A) (Shi et al., 2012). PRISM-SRM capitalizes on high-resolution reversed-phase liquid chromatographic separations for analyte enrichment, intelligent selection of target fractions via on-line SRM monitoring of heavy internal standards, and the partial orthogonality between high pH and low pH RPLC separations. The accurate elution profiles of the internal standards allow precise determination of the locations of target peptides in the 96-well plate, thus allowing the selection of the most informative target fractions for downstream LC-SRM analysis. It is also practically feasible to use other alternative strategies to pinpoint the target fractions, such as off-line determination of the locations of target fractions and relying on the reproducibility of HPLC to locate target fractions from subsequent experiments. Furthermore, a limited number of target fractions eluted at different times during the first-dimension separation can be multiplexed to enhance the overall sample throughput (Figure 3A). Fractions eluted at early, middle, and late retention times have little overlap in their elution profiles of the second-dimension LC separation, and thus can be effectively combined before LC-SRM analysis.

In contrast to standard LC-SRM, PRISM-SRM can provide ~200-fold improvement in detection sensitivity (Shi et al., 2012; Shi and Qian, 2013). PRISM-SRM allows for accurate and reproducible quantification of plasma proteins at the 50–100 pg/mL range in human blood with IgY14 immunoaffinity depletion (Shi et al., 2012) and the sub-ng/mL to low ng/mL range without immunoaffinity depletion (Shi et al., 2013). When applied to human cell lines it can reliably quantify proteins at 10–100 copies per cell (Shi et al., 2016). Such sensitivity is comparable to or even better than most analytically validated ELISAs (Shi et al., 2012; Shi et al., 2014). PRISM-SRM enabled the detection and quantification of all the selected EGFR pathway proteins including extremely low-abundant negative feedback regulators (e.g., DUSP4,6 and SPRY4) and the low expression EGFR in MCF7 that are rarely detected by other sensitive MS platforms or ELISAs (Catenacci et al., 2012). Very recently, we have applied PRISM-SRM to evaluate 52 candidate biomarkers in prostate cancer primary tumors (Gao et al., 2020). A 5-protein classifier demonstrated significant improvement over the standard of care base models using only the clinical and pathological variables in predicting distant metastasis and patient stratification.

Besides sensitive quantification of low-abundance proteins, we have also applied PRISM-SRM for site-specific quantification of stoichiometric protein phosphorylation (Shi et al., 2015), SAAVs (Vegvari, 2016; Wang et al., 2017), and gene fusion products (He et al., 2014). PRISM-SRM was demonstrated to enable direct quantification of ERK phosphorylation isoforms (pT, pY, pTpY) in human mammary epithelial cells (HMECs) and their phosphorylation dynamics in HMEC treated by different doses of EGF using as little as 25 μg tryptic peptides from whole cell lysates (Shi et al., 2015). When compared to immobilized metal affinity chromatography (IMAC), PRISM provided ~10-fold higher signal intensities, presumably due to the better peptide recovery of PRISM. The description of PRISM-SRM quantification of SAAVs and gene fusion products can be found in the following Section 3.

B. DD-SRM

To further improve targeted proteomics detection sensitivity, we have developed another “deep-dive” SRM (DD-SRM) approach that capitalizes on multidimensional high-resolution RPLC separation for target peptide separation and enrichment combined with precise selection of target peptide fractions (Nie et al., 2017) (Figure 3B). The concept of DD-SRM was built on three features of multidimensional RPLC separations: 1) the partial orthogonality between low- and high-pH RPLC separations that provides high peak capacity, and thus the combination of low-high-low pH RPLC separations would significantly reduce matrix background interference, 2) the consecutive implementation of three-dimensional RPLC separations (high-flow, microflow, and nanoflow LC) that allows for exceptionally high sample loading (e.g., ~4 mg for DD-SRM, ~25–50 μg for PRISM-SRM, and ~0.2–0.5 μg for LC-SRM), and 3) precise selection of target peptide fractions of interest by online SRM monitoring due to the compatibility of RPLC separation buffers with electrospray ionization. RPLC separation provides higher resolution and reproducibility and does not need additional sample cleanup to minimize sample handling for improved peptide recovery. All these features contribute to the ability of DD-SRM to detect extremely low abundance proteins in complex biological samples.

DD-SRM was demonstrated to enable detection and quantification of target proteins at ≤10 pg/mL levels in nondepleted human blood or <10 copies per cell in human tissue. Such levels of sensitivity are better than those of many analytically validated immunoassays. In contrast to conventional LC-SRM at detection sensitivity of ~100 ng/mL in human blood, DD-SRM improves SRM detection sensitivity by >4 orders of magnitude due to its greatly increased sample loading and reduced background interference. When compared to PRISM-SRM, DD-SRM provides ~100-fold improvement in detection sensitivity due to significantly enhanced sample loading as well as ultrahigh resolving power from serial orthogonal low-high-low pH RPLC separations. Furthermore, unlike immuno-SRM methods which require approximately 1 mL of human blood to achieve detection sensitivity of ~100 pg/mL, DD-SRM has ~10-fold higher sensitivity with ~5-fold less starting material (i.e., only ~200 μL of human blood).

C. T-μLC-SRM

To improve robustness and sample throughput, Zhang et al., has recently developed a novel trapping-micro-LC-SRM (T-μLC-SRM) method for sensitive quantification of target proteins (Zhang et al., 2018). It employs a dual-flow system for high-capacity loading and sensitive μLC-SRM analysis (Figure 3C). T-μLC-SRM consists of two synchronized LC units where high-flow LC at high pH is used for online trapping of high amounts of biological samples on a large-capacity trapping column, and low-flow μLC-MS at low pH is used for separation and detection. It utilizes the partial orthogonality between high and low pH RPLC for sample loading and analysis and selective-trapping and delivery as well as narrow-window-isolation SRM, which significantly reduce background noise and thus improve SRM detection sensitivity and selectivity. In addition, unique column configuration and peak compression were introduced to enable large-capacity, rapid, and quantitative loading while maintaining excellent chromatographic resolution. When compared to conventional LC-SRM, T-μLC-SRM improves detection sensitivity by up to 25-fold with exceptional robustness. No appreciable peak deterioration or loss of sensitivity was observed after >1,500 injections of tissue or plasma samples. It can be readily implemented for multiplexed quantification of many low-abundance proteins and coupled to other affinity-based enrichment methods to further improve detection sensitivity.

2.1.3. Other enrichment strategies

Besides the above methods, the use of nanoparticles (NPs) has recently emerged as an alternative method for highly sensitive, specific enrichment of proteoforms. NPs have similar size and diffusion kinetics as proteins, allowing effective penetration through complex biological mixtures, and high surface-to-volume ratios to enhance protein interaction. Furthermore, they are versatile scaffolds with the ability to couple diverse affinity ligands for effective protein binding and capture (Kelley et al., 2014; You et al., 2007). Tiambeng et al. have recently designed an organosilane surface functionalization molecule bearing a cysteine (Cys)-thiol reactive handle for synthesizing peptide-functionalized superparamagnetic NPs to capture and enrich low-abundance proteins from human serum (Tiambeng et al., 2020). The newly synthesized NPs enabled highly specific reproducible enrichment of cTnI (<1 ng/mL) by efficient depletion of highly abundant proteins (e.g., serum albumin at >10¹⁰ higher concentration than cTnI). Following the enrichment, top-down targeted proteomics was used to reveal different proteoforms of cTnI from human blood (Tiambeng et al., 2020).

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) is another convenient method for enrichment of target proteins based on their molecular weight. When combined with LC-SRM (termed GeLC-SRM), SDS-PAGE has been used for detection and quantification of protein mutations. Halvey et al., applied GeLC-SRM for reliable quantification of wide-type (WT) and mutant KRAS in relatively low protein inputs (5–50 μg) (Halvey et al., 2012). It allows one to detect KRAS mutant variants (G12D, G13D, G12V, G12S) in a panel of cancer cell lines. They also measured KRAS mutant peptides in fluid from benign pancreatic cysts and pancreatic cancers at concentrations from 0.08–1.1 fmol/μg of protein. This result demonstrated that GeLC-SRM provides a robust, sensitive approach for targeted quantification of mutant proteins in complex biological samples.

2.2. Analyte multiplexing capability

Recent advances in QqQ MS allow for simultaneous quantification of up to 30,000 transitions per analysis in the retention time (RT) scheduling mode. This can be translated into ~5,000 target analytes assuming 6 transitions for each pair of heavy and light analytes. Thus, with the well-established conditions the most advanced QqQ MS (e.g., Thermo TSQ Altis) has the potential for multiplexed quantification of 1,000s of protein mutations in a single LC-SRM analysis. We have recently applied TSQ Altis for reliable measurements of ~150 HNSCC protein markers with 2–3 surrogate peptides per protein (a total of ~300 target peptides) across ~1,000 serum samples (Lee et al., 2020). SureQuant operated in Thermo Orbitrap Exploris, Eclipse, or Lumos MS instrument, is another high-multiplexing method for enabling highly sensitive, multiplexed, reliable, and accurate quantification of ~100s-1000s of target analytes. This method has been built upon traditional internal standard (IS)-PRM (Gallien et al., 2015) with leveraging heavy isotope-labeled IS trigger peptides to efficiently guide MS acquisition in real-time analysis. The use of trigger peptides eliminates the need for RT scheduling to greatly expand multiplexing capability for highly multiplexed analysis of target proteins or peptides. Very recently, SureQuant has been used for reliable quantification of 385 commonly dysregulated pTyr targets with high quantitative accuracy, enhances target detection success rates, and improves the robustness and usability of targeted acquisition (Stopfer et al., 2021). It has also been used for multiplexed quantification of 804 surrogate peptides for 582 plasma proteins with extending the dynamic range by >1 order of magnitude in contrast to conventional SRM.

3. TARGETED PROTEOMICS FOR ANALYSIS OF PROTEIN MUTATIONS

In recent years S/PRM-based targeted proteomics has been demonstrated to have significant advantages over traditional antibody-based assays for reliable quantification of protein mutations in different diseases (Table 1). The protein mutation candidates are generated from either genomic studies or proteogenomic analysis (Figure 1). Furthermore, it has been recognized that novel mutant peptides from global proteomics discovery should be reported with extreme caution because of frequent false positive identification (Li et al., 2011; Nesvizhskii, 2014). Validation of these mutant peptides with targeted proteomics is often required to avoid artificial assignment (Bunger et al., 2007; Chernobrovkin et al., 2015). For example, proteogenomic analysis has recently been used for identification of protein mutations, p53^R273C for breast cancer (Dimitrakopoulos et al., 2017), CAPN2^D22E for prostate cancer (Kwon et al., 2019), and HSPA12A^E365G, AHNAK2^P1072S, and GAA^V220L for neurodegenerative disease for neurodegenerative disease (Wingo et al., 2017). These protein mutations have been further validated by LC-S/PRM to confirm their existence. In this section, we will review recent studies for detection and quantification of protein mutations using targeted proteomics, including SAAVs, alternative splicing variants, and gene fusion products.

TABLE 1.

Discovery and validation of protein mutations using targeted proteomics

Bottom-up targeted proteomics
Mutation type	Disease type	Protein	Mutation site	Specimen	Assay type	Clinical application	Ref.
SAAVs	Pancreatic cancer	Serotransferrin	I448V	Serum samples	LC-SRM	Diagnosis biomarker	Nie et al., 2017
	Pancreatic and colorectal cancer	KRAS	G12V	Cell lines, tissue and cyst fluids	Immuno-SRM	Diagnosis biomarker	Wang et al., 2011
			G12D
	Pancreatic and colorectal cancer	KRAS	G12D	Tissues	Immuno-SRM	Diagnosis biomarker	Ruppen-Cañás et al., 2012
	Pancreatic cancer	KRAS	G12D	Cell Lines	PRISM-SRM	Therapeutic target	Tan et al., 2020
		p53	R273H
		SLC37A4	G88D
	Pancreatic disease	KRAS	G12D	Cell lines and tissues	SDS-PAGE-based MRM (GeLC-MRM)	Diagnosis biomarker	Halvey et al., 2012
			G13D
			G12V
			G12S
	Pancreatitis	SPINK1	N34S	Serum samples	Immuno-pseudo-MRM	Prediction of chronic pancreatitis	Ravela et al., 2018
			P55S
	Prostate cancer	PSA	L132I	Plasma and seminal samples	Immuno-SRM (plasma depletion)	Diagnosis biomarker	Vegvari et al., 2013
	Prostate Cancer	CAPN2	D22E	Cell lines and tissues	PRM	Prognosis and prediction of advanced prostate cancer	Kwon et al., 2019
	Prostate cancer	SPOP	Y87N	Cell lines	PRISM-SRM	Diagnosis biomarker	Wang et al., 2017
			F102C
			F133V
	Neurodegenerative disease	Heat shock 70 kDa protein 12A (HSPA12A)	E365G	Brain tissues	PRM	Diagnosis biomarker	Wingo et al., 2017
		AHNAK nucleoprotein 2 (AHNAK2)	P1072S
		lysosomal alpha-glucosidase (GAA)	V220L
	Obesity and diabetes	Complement component C7	P587T	Plasma samples	LC-SRM	Diagnosis biomarker	Su et al., 2011
		Complement factor H	V62I
		Complement component C5	D966Y
	Age-related macular degeneration	Complement factor H (CFH)	Y402H	Plasma samples	LC-SRM	Diagnosis biomarker	Zhang et al., 2017
			I62V
	Brain disease	ALDH5A1	H180Y	Brain tissues	LC-SRM	Diagnosis biomarker	Su et al., 2014
		HADH	P215T
		RBP1	M50V
		GRIA1	N768S
		CDC42	K163R
	Glioblastoma	Cytoplasmic C-1-tetrahydrofolate synthase (MTHFD1)	R653Q	Stem cells	LC-PRM	Therapeutic target	Mostovenko et al., 2018
	Neurological and developmental disorder	HSP10	L73F	Primary cells	LC-SRM	Diagnosis biomarker	Bie et al., 2016
	Breast cancer	p53	R273C	Cell lines and tissues	LC-SRM and PRM	Diagnosis biomarker	Dimitrakopoulos et al., 2017
	Dileucineopathies	GLUT1	P485L	Cell lines and primary cells	LC-PRM	Diagnosis biomarker	Meyer et al., 2018
		ITPR1	P1059L
		CACNA1H	P648L
	Infertility	TEX101	G99V	Spermatozoa	Immuno-PRM (anti-TEX101)	Diagnosis biomarker	Schiza et al., 2019
	Colon Cancer	KRAS	G13D	Cell lines	LC-SRM	Monitor of disease progress	Demory Beckler et al., 2013
	Colorectal carcinoma	BRAF	V600E	Tissues	Immuno-SRM	Therapeutic target	Chen et al., 2016
	Colorectal adenocarcinoma			Cell lines	Immuno-SRM	Diagnosis biomarker	Lin et al., 2019
	Lung disease	Pulmonary surfactant protein A (SP-A)	Q223K	Bronchoalveolar lavage	SDS-PAGE-based MRM (GeLC-MRM)	Genotyping validation	Foster et al., 2014
	Congenital adrenal hyperplasia	CYP21A2	L388R	Plasma samples	SDS-PAGE-based MRM (GeLC-MRM)	Diagnosis biomarker	Brønstad et al., 2014
			E140K
			P45L
			V211M
			V281L
	Alexander disease	GFAP	R79C	Tissues	SDS-PAGE-based MRM (GeLC-MRM)	Diagnosis biomarker	Heaven rt al., 2019
			R239H
			R416W
ASVs		PKM1/2		iPSC lines and tissues	LC-PRM	Splicing isoform function	Lau et al., 2019
		MYOM1
		NDUA5
		TENV
		SVIL
		RTR2
		MYBPC3
	Neuronal disease	NRX1	Alternative splicing 3, 4 and 6	Brain tissues	LC-SRM	Cell recognition processes	Schreiner et al., 2014
		NRX2	Alternative splicing 3 and 6
		NRX3	Alternative splicing 3, 4 and 6
	Alzheimer’s disease	Tau	0N3R	Cerebrospinal fluids	Combination of IP and GeLC-MRM	Diagnosis biomarkers and therapeutic targets	Xu et al., 2021
			1N3R
			2N3R
			0N4R
			1N4R
			2N4R
	Premature aging disorders	Lamin A/C	78–89	Cell lines	LC-SRM	Diagnosis biomarker	Al-Qahtani et al., 2019
			547–572
			529–553
			529–615
	Heart disease	TPM1	189–212	Tissues	LC-PRM	Discovering functionally relevant isoforms in the heart.	Han et al., 2021
	Histone-deregulated disease	H2A	22 H2A variants	Mouse testes	LC-MRM/PRM	Histone variant and isoform identification	El Kennani et al., 2018
		H2B	3 H2B variants
	Non-small cell lung cancer (NSCLC)	Osteopontin (OPN)	16–81	Plasma samples	Immuno-SRM	Diagnosis biomarker	Wu et al., 2012
			16–58_72–81
			16–30_58–81
Gene fusion product	Prostate cancer	TMPRSS2-ERG		Cell line and tissues	PRISM-SRM	Diagnosis biomarker	He et al., 2014
				Cell lines	Immuno-SRM		Fu et al., 2021
Top-down targeted proteomics
Mutation type	Disease type	Protein	Mutation site	Specimen	Assay type	Clinical application	Ref.
SAAVs	Hemoglobinopathies	Hemoglobin A	E6V	whole blood samples	ETD and pseudo-SRM	Diagnosis biomarker	Coelho Graça et al., 2012
			E6K
	Muscle-related diseases	Tropomyosin (Tpm)	R38Q	Skeletal muscles	Offline multi-step purification	Diagnosis biomarker	Jin et al., 2016
			P64L
	Colorectal cancer	KRAS4b	WT	Cell line and tissues	Immunoaffinity enrichment	Therapeutic targets	Ntai et al., 2018
			G13D

3.1. SAAVs

SAAV, also known as SAP, is the most common protein mutation and often occurs on tumor-related proteins. SAAVs may lead to significant alteration in protein function and are directly associated with human diseases, and thus reliable quantification of SAAV mutations is crucial to understand molecular mechanism in disease (or cancer) initiation and progression with the potential of developing highly specific protein mutation markers for diagnosis and prognosis and revealing therapeutic targets for better treatment.

3.1.1. Pancreatic cancer

Pancreatic cancer is the 4th most common cancer for both men and women and 11th most common cause of death globally (Ilic and Ilic, 2016). CA19–9 glycoprotein is used for clinical diagnosis, but it lacks sufficient sensitivity and specificity. To identify alternative biomarkers for this lethal cancer, Lubman’s group has applied global proteomics for deep profiling of SAAV peptides in PANC-1 cell line and human serum, and then validated the SAAV candidate biomarkers of interest by targeted proteomics (Nie et al., 2014; Tan et al., 2019; Tan et al., 2020). One mutant peptide from serotransferrin (I448V) was measured in a cohort of serum samples from pancreatic cancer patients. When combined with α−1-antichymotrypsin (AACT) and thrombospondin-1 (THBS1), the mutant peptide has shown an excellent diagnostic performance for significant differentiation of pancreatic cancer patient from healthy controls and pancreatitis (Nie et al., 2014).

Oncogenic KRAS gene mutations are known to be one of characteristics of pancreatic cancer (Waddell et al., 2015). For sensitive detection of KRAS protein mutations, a pan-KRAS antibody which recognizes the common epitope on both the WT and mutant KRAS (G12V and G12D) was used for efficient enrichment of KRAS protein mutations. The enriched samples were analyzed by LC-SRM with detection of both KRAS mutations (Wang et al., 2011). With the use of a more sensitive targeted proteomics platform, immuno-SRM enabled for quantification of both WT KRAS and KRAS^G12D as low as 12 amol (0.25 pg) in tumor tissue samples from pancreatic and colorectal cancer patients (Ruppen-Canas et al., 2012). Besides antibody-based enrichment, Halvey et al. employed SDS-PAGE for selectively enriching KRAS proteins. Four KRAS variants (G12D, G13D, G12V and G12S) were identified by GeLC-SRM from a panel of cancer cell lines and fluid from benign pancreatic cysts and pancreatic cancer patients (Halvey et al., 2012). GeLC-SRM was used for reliable quantification of their corresponding mutant peptides at concentrations of ~0.08–1.1 fmol/μg, which makes it possible to use KRAS protein mutations as specific diagnosis biomarkers for pancreatic cancer.

In addition, SPINK1 mutation (N34S) was reported to be associated with earlier onset and pancreatic insufficiencies (Witt et al., 2000). Both wild type and mutant (N34S and P55S) SPINK1 can be effectively enriched using an anti-SPINK1 antibody followed by LC-SRM analysis. Immuno-SRM assays have been demonstrated to enable detection of SPINK1 mutations at the LOD of 0.5 ng/mL in patient sera at different disease conditions (Ravela et al., 2018).

3.1.2. Prostate cancer

Prostate cancer (PCa) is the second most common cancer among men worldwide (Jemal et al., 2011). Prostate-specific antigen (PSA) is routinely used for clinical diagnosis and frequently chosen as a model protein for the S/PRM method development (Fortin et al., 2009; Li et al., 2011; Liu et al., 2012). With immuno-SRM assays, the PSA^L132I mutation was identified in the PCa patient plasma, and PSA^L132I was estimated to be present in 10% of the human population (Vegvari et al., 2013). However, its value for PCa screening needs further evaluation in large cohorts of clinical samples. We have recently used targeted proteomics for analysis of SPOP mutations (up to 15% of PCa patients) which represent one of the most specific biomarkers that define distinct molecular subtypes of PCa. Based on genomic data we selected 11 surrogate peptides for SPOP SAAVs (Y87C/N, F102C, K129C, D130H, W131C/G, F133L/V/S and K134N) (Wang et al., 2017). We applied a tiered approach (first LC-SRM then PRISM-SRM) for quantifying them in PCa cell lines. PRISM-SRM enable confident detection of three most frequent SPOP mutations (Y87N, F102C, and F133V) in the mutation positive cell lines but not in the negative cell lines (Wang et al., 2017). Lower expression of the F133V mutation and WT SPOP was observed when compared to that of other two F102C and Y87N mutations. This result suggests that SRM-based targeted proteomics is a promising tool for multiplexed targeted discovery of protein mutations predicted from genomic data.

3.1.3. Aging- or neurodegeneration-related diseases

Point mutations of single amino acids were proved to cause functional change of proteins in age-related diseases (e.g., β-amyloid precursor protein in Alzheimer’s disease (Di Fede et al., 2009)). SRM-based targeted proteomics was used to detect and quantify SAAVs (e.g., complement component C7^P587T, ALDH5A1^H180Y, HADH^P215T, RBP1^M50V, GRIA1^N768S, CDC42^K163Rand MTHFD1^R653Q) which were reported to be significantly associated with obesity, diabetes (Su et al., 2011), age-related macular degeneration (Zhang et al., 2017), brain disease (Su et al., 2014), and glioblastoma (Mostovenko et al., 2018). Targeted proteomics analysis of plasma SAAVs provides a new way for assessing physiological or pathological traits. For functional study, Bie et al., applied SRM assays to measure the steady state levels of the WT and L73F mutant HSP10 in fibroblasts of a patient with a complex disease history (e.g., infantile spasms, hypotonia, developmental delay, a slightly enlarged liver, macrocephaly, and mild non-specific dysmorphic features) (Bie et al., 2016). Based on SRM quantification, they discovered that mutant HSP10^L73F not only decreased the formation of HSP60/HSP10 chaperonin complex but also reduced the expression of SOD2. This result suggests that mutant HSP10^L73F acts as a strong contributing factor for the disorder in the affected patient.

3.1.4. Other diseases

Targeted proteomics has also been used for quantification of mutant proteins in other diseases, such as GLUT1^P485L, ITPR1^P1059L, and CACNA1H^P648L for in dileucineopathies (Meyer et al., 2018), TEX101^G99V in infertility (Schiza et al., 2019), and KRAS^G13D and BRAF^V600E in colon cancer (Chen et al., 2016; Demory Beckler et al., 2013; Lin et al., 2019). With gel electrophoresis fractionation, SRM-based targeted proteomics was applied for detection and quantification of SP-A variants (Gln223 and Lys223) from bronchoalveolar lavage (BAL)(Foster et al., 2014), GFAP variants (R79C, R239H and R416W) for Alexander disease(Heaven et al., 2019), and CYP21A2 variants (L388R, E140K, P45L, V211M and V281L) from congenital adrenal hyperplasia (CAH) patient plasma(Bronstad et al., 2014). The mutations of L388R and E140K for CYP21A2 were demonstrated to affect the enzyme activity for steroid synthesis in the adrenal cortex(Bronstad et al., 2014).

3.2. ASVs

Large-scale RNA-seq sequencing has shown that there are over 100,000 alternative splicing transcripts in the human genome (Pan et al., 2008; Wang et al., 2008). However, whether their corresponding protein isoforms are produced remains largely unknown due to the lack of suitable proteomics tools. Recent technological advances in bottom-up MS proteomics have greatly increased the ASV identification, but current bottom-up proteomics still lacks the depth to cover the majority of the ASV proteome and the sufficient resolution to discern 10s-100s of closely related protein isoforms that differ by only a single peptide (Nesvizhskii and Aebersold, 2005). One promising tool to tackle these challenges is targeted proteomics, as demonstrated by SRM-based measurement of alternative splicing forms of neuronal neurexin receptors (Schreiner et al., 2015).

3.2.1. Neuronal disease

Neurexins are a class of synaptic adhesion molecules that play an important role in the synapse formation and function (Dean et al., 2003). In mammals, three neurexin genes (Nrxn1–3) are transcribed into long alpha-neurexin and short beta-neurexin transcripts with >1,000 unique isoforms in the adult brain due to the presence of six alternatively spliced segments (AS1–6) and extensive alternative splicing (Baudouin and Scheiffele, 2010; Schreiner et al., 2014). The diversity of neurexin molecules may serve as synaptic recognition events which control neuronal wiring and function (Reissner et al., 2013). To detect endogenous neurexin splicing protein variants, Schreiner et al. developed SRM assays for quantification of surrogate peptides which are unique to segments AS3, 4, and 6 (Schreiner et al., 2015). With these SRM assays, 17 unique protein isoforms can be reliably detected and quantified. For the panel of SRM assays, pan-neurexin peptides and peptides unique to either the alpha or beta form of neurexin were also included to measure the total neurexin and neurexin-alpha/beta, respectively. Such SRM assays not only allow them to answer the fundamental question about which splicing variants are translated to proteins, but also provide a proteomics tool for highly specific quantification of protein splicing variants.

Another example is targeted quantification of ASVs from Alzheimer’s disease (AD), one of the most common causes of dementia (Kukull and Bowen, 2002). Tau protein was demonstrated to promote microtubule assembly, stabilize microtubules, and maintain normal morphology of neurons (Weingarten et al., 1975). It has at least six isoforms and each isoform has differential physiological roles (Andreadis, 2005). Quantification of individual Tau isoform may be helpful to better understand the role of Tau in AD and to improve AD diagnostic and treatment accuracy. Very recently, Xu et al. developed multidimensional separations (i.e., the combined immunoprecipitation and SDS-PAGE) coupled with LC-SRM for detection and quantification of all the six Tau isoforms with ~250-fold improvement in detection sensitivity (Xu et al., 2021). The multidimensional targeted proteomics platform was then applied for simultaneous quantification of these isoforms in CSF of AD patients.

3.2.2. Other diseases

Lamin A/C mutations have been linked to numerous heritable disorders including premature aging syndrome (Worman, 2018). SRM assays have recently been developed for highly specific quantification of four splicing variants (lamin A, lamin C, lamin^AΔ10, and lamin^AΔ50) for the Lamin A/C proteins in MCF7 and U937 cancer cell lines(Al-Qahtani et al., 2019). Except for lamin^AΔ50, three out of four Lamin A/C variants have been detected with direct LC-SRM. The performance of splicing variant SRM assays (e.g., sensitivity, reproducibility, and reliability) has further been evaluated and compared with that of qRT-PCR for quantification of splicing variants at the mRNA level.

To improve the number and the confidence of ASVs identification, Lau et al. developed an isoform-inferred approach in which RNA-seq data are used to guide the discovery of splicing transcripts harboring junction pairs (Lau et al., 2019). They identified >1,000 isoform sequences which were not documented previously. They further validated selected junction sequences by performing PRM-based targeted quantification. Prior to LC-PRM analysis, normal human heart tissue lysates were digested and separated into 10 fractions with high-pH RPLC to improve PRM sensitivity. Targeted quantification not only validated of known junction peptides, but also identified six protein isoforms. For easy method adoption, the same group has recently published the detailed workflow for the computation-assisted targeted proteomics method (Han et al., 2021). Targeted proteomics analysis of histone variants is another example because of the existence of many variants for histone proteins (El Kennani et al., 2017). Kennani et al. demonstrated the power of the targeted proteomics for precise quantification of histone variants and isoforms with high sequence similarity (El Kennani et al., 2018). These assays can be easily applied to study human histone variants whose abundance may be dysregulated in various diseases.

In addition, Wu et al. developed immuno-SRM assays for quantification of three osteopontin splicing variants (OPNa, b, and c) in human plasma (Wu et al., 2012). All OPN isoforms were immunocaptured to reduce sample complexity using the pan-OPN antibody and SRM assays were used for targeted quantification of isoform-specific signature peptides. For the non-small cell lung cancer (NSCLC) patients, the concentrations of OPNa were found to be substantially elevated when compared to normal controls.

3.3. Protein fusion products

Gene fusions are generated through structural rearrangements (e.g., inversions, interstitial deletion, or translocations of two independent genes). They can lead to the dysregulation of gene expression, the formation of novel fusion proteins, and the truncation of protein products. Recent studies have revealed that fusion genes can act as driver mutations across diverse cancer types (Watson et al., 2013). Fusion between the androgen-responsive gene TMPRSS2 and the transcription factor ERG (TMPRSS2-ERG) is a common driver mutation in nearly 50% of PCa (Nam et al., 2007) and has been used as drug target (Shao et al., 2020) as well as diagnostic (Yang et al., 2016) and prognostic biomarkers (Hagglof et al., 2014). Due to the lack of high-quality antibodies suitable for quantitative analysis, studies of TMPRSS2-ERG gene fusions have seldom been performed at the protein level.

With the development of the antibody-independent, highly sensitive and specific PRISM-SRM platform, we have applied PRISM-SRM assays for reliable quantification of the TMPRSS2-ERG gene fusion protein products in PCa cell lines and prostate tumor tissues (He et al., 2014). At least two distinct ERG protein isoforms were observed to be simultaneously expressed in TMPRSS2-ERG gene fusion positive samples as evidenced by the concomitant detection of two mutually exclusive peptides in two patient tumors and the VCaP cell line. Three peptides, shared across almost all fusion protein products, were determined to be the most abundant peptides, providing “signature” peptide markers for detection of ERG over-expression resulting from TMPRSS2-ERG gene fusion. With its demonstrated analytical performance, PRISM-SRM is a valuable tool for studying gene fusion protein products without the need of specific affinity reagents.

Very recently, Fu et al. developed immuno-SRM assays for quantification of a low-abundance T1E4 TMPRSS2-ERG fusion protein and its isoforms in VCaP cells (Fu et al., 2021). With accurate quantification of the total ERG and its four unique isoforms, immuno-SRM revealed that the T1E4-ERG isoform accounted for ~50% of the total ERG protein in VCaP cells and formalin-fixed paraffin-embedded PCa tissues. With immuno-SRM assay, the amount of total ERG was estimated as 2.2 fg per VCaP cell, consistent with the level of 1.8 fg per cell from our PRISM-SRM assay (He et al., 2014). This suggests that both targeted proteomics methods are reliable for quantification of the TMPRSS2-ERG gene fusion products. Immuno-SRM assays present an alternative tool for quantifying ERG and its isoforms in clinical samples, thus paving the way for development of more accurate PCa diagnostics.

4. TOP-DOWN PROTEOMICS FOR ANALYSIS OF PROTEIN MUTATIONS

Unlike bottom-up proteomics that requires the digestion of proteins into peptides, top-down proteomics directly analyzes intact proteins, allowing precise and comprehensive characterization of various proteoforms such as protein mutations from genetic polymorphisms and RNA splice variants as well as PTMs (Aebersold et al., 2018; Schaffer et al., 2019; Smith and Kelleher, 2018). Since it directly analyzes intact proteins, top-down proteomics can potentially provide the richest data for both reliable protein identification and precise localization of mutation and PTM sites for a complete view of proteoforms (Smith and Kelleher, 2018). Similar to bottom-up proteomics, top-down proteomics can also be performed in either a shotgun fashion (i.e., global profiling) or a targeted manner (Chen et al., 2018). However, there are many technical challenges for current top-down proteomics technologies (e.g., difficulties in characterization of high molecular weight and low-abundance proteins, and bioinformatics tools for analysis of complex proteoform MS data) (Schaffer et al., 2019), which severely hinder its broad applications with technical limitations in depth and throughput.

4.1. Global top-down proteomics

In top-down proteomics, intact proteins extracted from cells are typically fractionated by LC, electrophoresis, or serial size exclusion chromatography (sSEC) followed by RPLC-MS/MS analysis. Current state-of-the-art global top-down proteomics workflows enable to identify ~3,000–6,000 proteoforms corresponding to ~800–1,200 proteins. Using a multiple dimensional separation system (electrophoresis, solution isoelectric focusing (sIEF), and gel-eluted liquid fraction entrapment electrophoresis (GELFrEE)), ~1,000 proteins were identified from HeLa cells with >3,000 proteoforms (Tran et al., 2011) and ~1,200 proteins from H1299 cells with >5,000 proteoforms (Catherman et al., 2013). Cai et al. developed a sSEC strategy for high-resolution size-based fractionation of 10–223 kDa intact proteins from human heart tissues. The cSEC fractions were further separated by RPLC followed by MS analysis. In contrast to one dimensional RPLC separation, >4,000 unique proteoforms were identified with a 15-fold increase in the detection of proteins above 60kDa (Cai et al., 2017). Using an orthogonal multidimensional separation platform coupling SEC and RPLC based protein prefractionation to capillary zone electrophoresis (CZE)-MS/MS with ~4,000 of peak capacity, ~5,700 proteoforms were identified from the E. coli proteome (McCool et al., 2018), representing the largest bacterial top-down proteomics data set. A 10-fold improvement in the number of identified proteoforms was achieved when compared with previous CZE-MS/MS studies. Toby et al. applied the GELFrEE fractionation and RPLC-MS/MS for global analysis of human blood and ~2,900 proteoforms were identified from human peripheral blood mononuclear cells (Toby et al., 2019). Using quantitative top-down proteomics analysis of specific regions of the female mouse brain, several proteoform changes in abundance were observed after mice were trained for cocaine-conditioned place preference. These observations provide insight into estrogen signaling in the brain with potential new approaches to treat women with cocaine use disorder (Park et al., 2019). However, most proteoforms were caused by the events of various PTMs (e.g., methylation, acetylation, and phosphorylation) and truncations at the protein termini. Few protein mutations were identified by global profiling presumably due to their low frequency rate and/or the low- or moderate-abundance of their corresponding proteins.

4.2. Targeted top-down proteomics

To improve detection sensitivity for analysis of protein mutations, targeted top-down proteomics were used for detection and quantification of proteoforms from specific proteins of interest. With the combination of electron transfer dissociation and pseudo-SRM, targeted top-down proteomics was used for analysis of hemoglobin A and hemoglobin variants where glutamic acid at position 6 of the β globin chain is replaced by either valine (hemoglobin S) or lysine (hemoglobin C) in clinical samples (Coelho Graca et al., 2012). Subtle difference in the β globin chain can be unambiguously identified from the whole blood by direct injection. With the use of a top-down LC/MS strategy for antibody-independent purification and offline top-down MS analysis, tropomyosin (Tpm) isoforms associated with muscle diseases were reliably identified and quantified across different skeletal muscles from multiple species, including swine, rat, and human (Jin et al., 2016). Among these isoforms Tpm1.1 was confirmed to have two amino acid polymorphisms R38Q and P64L and N-terminal acetylation. This study demonstrates the utility of targeted top-down MS for precise characterization of proteoforms from single protein.

Another example for targeted top-down MS is to study how genetically encoded mutations affect PTMs on the same protein molecule by detection and quantification of mutation-specific consequences (Ntai et al., 2018). Immunoaffinity enrichment combined with top-down MS was used to discover and quantify KRAS proteoforms caused by the KRAS^G13D mutation. Analysis of isogenic CRC cell lines has shown a direct link between the knockout of the mutant G13D allele and the complete nitrosylation of cysteine 118 of the remaining WT KRAS4b. Application of the top-down MS workflow to other cancer cell lines with three mutations at Gly12 along with primary colorectal tumor samples enabled to quantify mutant versus WT KRAS4b expression ratios to reveal the major difference in the levels of C-terminal carboxymethylation, a modification critical for membrane association. This study highlights the importance of targeted top-down proteomics for discovery of dynamic PTMs underlying key regulatory mechanisms with the potential link to disease stage and chance of survival.

5. FUTURE PERSPECTIVES

The central dogma of biology is the flow of genetic information from DNA to RNA to protein. Recent technological advances allow for cost-effective whole genome (transcriptome) analysis of genetic mutations, resulting in the identification of 1000s of cancer-associated mutations (Reva et al., 2011). However, proteomics technologies for comprehensive analysis of protein mutations are lagging far behind genomics technologies, which severely prevents a more comprehensive view of cancer-associated mutations through integrated protegenomic analysis and the identification of functionally important mutations to connect disease genotype to phenotype. More importantly, protein mutations identified in tumor cells not only confirm genetic mutations but actually are responsible for tumorigenesis and altered cellular functions (Wang et al., 2011). Therefore, large-scale analysis of protein mutations promises a unique opportunity for cancer biomarker development and discovery of new therapeutic targets for more effective disease treatment (Figure 4).

FIGURE 4.

Future developments for large-scale discovery of protein mutations and their applications to biomedical research.

5.1. Detection sensitivity and sample throughput

Future developments will focus on significant improvements in detection sensitivity and sample throughput for rapid deep profiling of protein mutations across 100s of clinical samples. Enhancing detection sensitivity could be achieved by effective integration of ultralow-flow LC (Shen et al., 2003; Shen et al., 2004; Shen et al., 2002) or capillary electrophoresis (CE) (Sun et al., 2013) and a high-efficiency ion source/ion transmission interface (Cox et al., 2014; Marginean et al., 2010; Page et al., 2008; Tang et al., 2011) with the most advanced MS platform for sensitive detection of low-abundance protein mutations. Sample throughput could be increased by using ultrafast high-resolution ion mobility-based gas-phase separation (e.g., timsTOF (Meier et al., 2018), FAIMS (Hebert et al., 2018) and SLIM IMS (Ibrahim et al., 2017)) to replace current slow liquid-phase (LC or CE) separations, and effective integration of liquid- and gas-phase separations for greatly reducing separation time (i.e., improved sample throughput) but without trading off separation resolving power. For example, the combined low-resolution FAIMS and high-resolution LC separations enabled analysis of up to 2,000 peptides per minute and more than 5,000 protein groups in 20 mins (Bekker-Jensen et al., 2020). All these advancements could significantly improve the performance of both bottom-up and top-down MS platforms for in-depth profiling of protein mutations in a reasonable time frame. They will also help close the gap between proteomics and transcriptomics or genomics technologies for more effective proteogenomic analysis of mutations. For confident detection of protein mutations by global proteomics profiling, an accurate and more complete mutation database is a prerequisite as the identification relies on the quality of the database (Schaffer et al., 2019; Sheynkman et al., 2016). In addition, new bioinformatics tools are also needed to better control the FDR (Schaffer et al., 2019; Sheynkman et al., 2016).

5.2. Large-scale targeted discovery of protein mutations

MS-based bottom-up global proteomics has been used for deep profiling of protein mutations. However, only low percentage of SNVs detected by either DNA or RNA sequencing had been validated as SAAVs at the protein level (Rodland et al., 2018; Ruggles et al., 2016; Zhang et al., 2019). Unlike global proteomics that need database searching for mutant protein identification and may fail to detect important mutations at low expression levels, targeted proteomics can be used to exclusively analyze 1000s of presumably important mutations obtained from curated literature, mutation database, and prediction tools. For example, targeted proteomics has been used to validate mutations from global discovery (Lichti et al., 2015; Mostovenko et al., 2018; Tan et al., 2020; Vegvari, 2016) and to accurately quantify cancer-specific mutations as markers (Lesur et al., 2015; Wang et al., 2011). With recently demonstrated exceptionally high multiplexing capability (e.g., ~1000s of target analytes in single LC-PRM analysis operated in the SureQuant mode (Stopfer et al., 2021)) and constant improvement in MS detection sensitivity, targeted proteomics is well suited for precise and comprehensive targeted discovery (identification) and quantification of functionally important protein mutations.

To further improve the success rate of targeted discovery, the use of multiple enzymes in a complementary fashion needs to be considered to improve sequence coverage (Sheynkman et al., 2016; Trevisiol et al., 2016). This is critically important for large-scale profiling of protein mutations because of an uneven distribution of their cleavage sites. If only trypsin (lysine and arginine as the cleavage sites) is used, some tryptic peptides may be too long or too short to be effectively detected and sequenced by MS. In the case of no suitable tryptic peptide available to cover the mutation site, a different proteolytic enzyme may be used to generate peptides covering the targeted mutation site for S/PRM. This was evidenced in our study for quantification of SPOP mutations using the enzyme of Arg-C rather than trypsin for digestion because there are many lysine residues in the SPOP mutation region (Wang et al., 2017).

5.3. Discovery of disease biomarkers and therapeutic targets

Protein mutation markers are supposed to provide complementary diagnostic and prognostic values to genetic mutation markers because the correlation between protein and mRNA abundance is typically low to moderate (Rodland et al., 2018; Sinha et al., 2019; Zhang et al., 2019). Proteogenomic analysis of the same mutation markers across the same cohort of clinical samples is expected to fully reveal disease biology for accurate diagnosis, prognosis, and better monitoring of the response to therapy, contributing to better opportunities toward precision medicine. It has also been recognized that comprehensive proteogenomic analysis is powerful for differentiating functionally important mutations (e.g., driver mutations) from passenger mutations and for revealing the connection between genotype and phenotype (Dou et al., 2020; Mertins et al., 2016; Rodland et al., 2018; Ruggles et al., 2016; Sheynkman et al., 2016; Vasaikar et al., 2019; Wang et al., 2021; Zhang et al., 2016). With constant improvement in MS-based proteomics technologies, more functionally important protein mutations are expected to be detected and quantified in clinical samples. Since these protein mutations are expressed only in tumor cells and responsible for tumorigenesis and progression, the expression level of these mutations can more accurately reflect cellular phenotype and regulatory processes than genetic mutations (Vegvari, 2016; Wang et al., 2011). Thus, they have the potential to be novel disease markers with high specificity for early diagnosis and prognosis and monitoring disease progression.

Cancers are caused by the accumulation of mutations (Hanahan and Weinberg, 2000). An individual mutation may slightly increase the risk of cancer development, but the combination of multiple mutations can result in a significant level of risk. Therefore, multiplexed targeted proteomics technologies are important for quantification of 100s of functionally important mutations across 100–1000s of clinical samples to prioritize the most promising disease biomarkers. Furthermore, the stoichiometries of protein mutations (i.e., the abundance ratio of mutant over the total mutant and WT proteins) can be accurately determined by multiplexed targeted proteomics. When compared to the abundance of protein mutations alone, the threshold of their stoichiometries may be more critical to define the onset of tumorigenesis and tumor progression, which could facilitate biomarker discovery and improving our understanding of the biology of complex diseases.

Besides the above, functionally important protein mutations may also alter protein structures, which could affect protein association and dissociation, leading to aberrant protein-protein interactions and signaling transduction pathways (e.g., different dynamic signaling patterns of PTMs) (Reva et al., 2011). Therefore, high-resolution structural analysis of disease-associated functional mutations allows for discovery of new therapeutic targets for targeted therapy. In addition, aberrant signaling pathways caused by disease-associated mutations are another attractive therapeutic target because activation of signaling pathways contributes to tumor growth and therapeutic resistance, and constitutively active signaling is commonly observed in cancer (Vegvari, 2016). Another important yet difficult to access research area is the gain or loss of PTMs caused by SAAVs. This is particularly important for protein glycosylation because the large glycan group will alter protein folding, resulting in changing protein function. Limited studies on glycosylation were conducted by Lubman’s and Goldman’s research groups (Fan et al., 2018; Mazumder et al., 2012; Tan et al., 2019). Future studies for simultaneous targeted quantification of PTMs and SAAVs are needed to help better understand the functional impact of SAAVs on human diseases.

6. CONCLUSIONS

Proteomic analysis of mutations is still at the infancy stage. With the recent advances in LC and MS instrumentation, MS-based global proteomics allows genome-scale proteome profiling and proteomic data are being increasingly integrated with genomic data for proteogenomic analysis, which is expected to significantly improve our understanding of the biology of the diseases and the potential of moving toward precision medicine. However, from the genome-scale proteome profiling data, less than 10% single nucleotide variants were detected as SAAVs at the protein level. The low success rate can be attributed to lacking generic databases for sequence search and low abundance mutant peptides as well as high false discovery. Targeted proteomics is a powerful tool for targeted discovery and validation of protein mutations because there is no need for database search, and it has high specificity, multiplexing capability, and reproducibility. When combined with the front-end enrichment of target analytes to enhance detection sensitivity, targeted proteomics has been broadly used for multiplexed detection and quantification of low abundance protein mutations and identification of functionally important mutations. These mutations are responsible for tumorigenesis and tumor progression, and thus they can be used as disease markers for diagnosis and monitoring disease progression due to their high specificity to tumor cells. Moreover, functional studies of these mutations have the potential for discovery of new therapeutic targets. Targeted top-down proteomics has also been used for analysis of protein mutations with precise localization of mutation sites in the intact proteins. Still, there are many technical challenges for current top-down proteomics, which greatly limits its broad applications at present. Further improvement in detection sensitivity and sample throughput for targeted proteomics is needed for cost effective, rapid, sensitive quantification of many important protein mutations in large clinical cohorts for translation of mutation markers into clinical use.

Abbreviations:

AACT

α-1-antichymotrypsin

AD

Alzheimer’s disease

ASV

alternative splicing variant

AS

alternative splicing

BAL

bronchoalveolar lavage

β-NGF

β-nerve growth factor

CAH

congenital adrenal hyperplasia

CE

capillary electrophoresis

cTnI

cardiac troponin I

Cys

cysteine

CZE

capillary zone electrophoresis

DD-SRM

deep-dive selected reaction monitoring

DDA

data-dependent acquisition

DIA

data-independent acquisition

FDR

false-discovery rate

GeLC-SRM

gel electrophoresis-LC-SRM

GELFrEE

gel-eluted liquid fraction entrapment electrophoresis

HMEC

human mammary epithelial cell

HNSCC

head and neck squamous cell carcinoma

IA

immunoaffinity

IMAC

immobilized metal affinity chromatography

IS

internal standard

LC-MS

liquid chromatography–mass spectrometry

LOD

limit of detection

MRM

multiple reaction monitoring

MS

mass spectrometry

NCI CPTAC

National Cancer Institute Clinical Proteomic Tumor Analysis Consortium

NPs

nanoparticles

Nrxn

neurexin gene

NSCLC

non-small cell lung cancer

OPN

osteopontin

PCa

prostate cancer

PRISM

high-pressure, high-resolution separations coupled with intelligent selection and multiplexing

PRM

parallel reaction monitoring

PSA

prostate-specific antigen

PTM

posttranslational modification

QqQ

triple quadrupole mass spectrometer

RPLC

reversed-phase LC

RT

retention time

SAAV

single amino acid variant

SAP

single amino acid polymorphism

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

sIEF

solution isoelectric focusing

SISCAPA

stable isotope standards and capture by antipeptide antibodies

SNV

single nucleotide variant

SOD2

superoxide dismutase 2

SPINK1

pancreatic secretory trypsin inhibitor Kazal type 1

SPOP

Speckle-type POZ protein

SRM

selected reaction monitoring

sSEC

serial size exclusion chromatogram

THBS1

thrombospondin-1

Tpm

tropomyosin

T-μLC-MS

trapping-micro-LC-MS

WT

wide type

Biography

Tai-Tu Lin obtained his PhD degree form National Taiwan university in 2017, studying membrane proteomics and glycobiology under the guidance of Professors Yu-Ju Chen and Chi-Huey Wong. He is currently working as a postdoctoral fellow in the Integrative Omics Group, Biological Sciences Division, Pacific Northwest National Lab under Dr. Wei-Jun Qian, focusing on developing mass spectrometric assays for precisely monitoring the pathophysiology of disease and applying the state-of-the-art mass spectrometry method for biomarker discovery.

Tong Zhang obtained his PhD degree from the University of Florida in 2016, studying biochemistry and proteomics under Professors Sixue Chen and Alice C. Harmon. He worked as a postdoctoral fellow in the Integrative Omics Group, Biological Sciences Division, Pacific Northwest National Laboratory under Dr. Wei-Jun Qian, focusing on developing proteomic strategies to characterize biological systems. He is currently working as a research scientist in the Seattle Children’s Research Institute. His research interest includes highly sensitive proteomic detection of mutations that cause human diseases, large-scale identification of post-translational modifications in the proteomes, and mechanistic studies for understanding how redox modifications relay cell signaling.

Reta B. Kitata obtained his PhD degree from National Tsing Hua University in 2017, studying membrane proteomics and cancer biology under the guidance of Professor Yu-Ju Chen. Later he worked as postdoctoral fellow in Academia Sinica, Taiwan. He is currently working as research associate in Integrative Omics Group, Biological Sciences Division, Pacific Northwest National Laboratory under Dr. Tujin Shi focusing on spatial phosphoproteome mapping of human tissues.

Tao Liu obtained his PhD in Biochemistry and Molecular Biology in 2001 from Shanghai Institute of Biochemistry, Chinese Academy of Sciences. He is a Senior Staff Scientist and Team Leader in the Integrative Omics group in the Biological Sciences Division at the Pacific Northwest National Laboratory. His research includes development and application of advanced MS technologies for protein post-translational modifications, single-cell, and multiomic characterization of cancers, leading to improved understanding of aberrant regulatory and signal transduction networks underlying cancers, as well as cancer biomarker discovery and verification.

Richard D. Smith received his PhD degree in physical chemistry from the University of Utah. He is currently Director of Proteomics Research, a Battelle Fellow, as well as Chief Scientist for the Biological Sciences Division at Pacific Northwest National Laboratory. He is the author or co-author of more than 1100 peer reviewed publications, holds 75 US patents, and his recognition has included the 2003 American Chemical Society Award for Analytical Chemistry, the 2009 HUPO Discovery Award in Proteomics Sciences and the 2013 ASMS Distinguished Contribution in Mass Spectrometry Award. His early research included the development of the combinations of both capillary supercritical fluid chromatography and capillary electrophoresis separations with mass spectrometry. He also led in the early development and use of high-resolution capillary LC combined with high resolution mass spectrometry for applications in proteomics. More recently his research has included contributions to ion mobility spectrometry (IMS) development, including Structures for Lossless Ion Manipulations (SLIM) IMS for achieving much higher resolution ion mobility separations with mass spectrometry, for increasing the speed, sensitivity and utility of measurements.

Wei-Jun Qian obtained his Ph.D. in Bioanalytical Chemistry from the Chemistry Department at the University of Florida in 2002. He is currently a Laboratory Fellow at the Pacific Northwest National Laboratory and a member of the Integrative Omics group within the Biological Sciences Division. His research involves the development of chemical proteomic approaches for posttranslational modifications, particularly cysteine thiol-based redox modifications, and targeted quantification with applications in metabolic and oxidative stress-related disease areas.

Tujin Shi obtained his PhD degree from York University in 2007, working under the guidance of Professors K. W. Michael Siu and Alan C. Hopkinson. He is a senior scientist in the Integrative Omics Group, Biological Sciences Division, Pacific Northwest National Laboratory. His research interest encompasses targeted proteomics, single-cell proteomics, liquid biopsy proteomics, and spatial tissue mapping as well as gas-phase ion chemistry. His team members develop mass spectrometry-based proteomics approaches and then apply them to address current biomedical need in combination with other advanced technologies.