Proteomics Clin. Appl. 2014, 8, 543–553
543
DOI 10.1002/prca.201400034
REVIEW
Targeted proteomics of myofilament phosphorylation and other protein posttranslational modifications Genaro A. Ramirez-Correa1 , Maria Isabel Martinez-Ferrando2 , Pingbo Zhang3 and Anne M. Murphy1 1
Department of Pediatrics/Division of Cardiology, Johns Hopkins University School of Medicine, Baltimore, MD, USA 2 Department of Comparative Biology and Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA 3 The Hopkins Bayview Proteomics Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA Global cardiac myofilament protein phosphorylation levels, and their site-specific stoichiometry, are physiologically and clinically relevant for heart function. Unlike myofilament phosphorylation, other PTMs such as O-GlcNAcylation are just beginning to gain attention due to their potential physiological and clinical implications. This review will focus on what is currently known about cardiac troponin I phosphorylation, and on the potential physiological and clinical impact of targeted proteomics including new findings on cardiac troponin I sites and stoichiometry. We will then discuss the increasing recognition of other myofilament PTMs functional relevance and the potential of targeted MS approaches, particularly MRM, for accelerating their systematic characterization. In addition, we will broadly discuss the development and application of MRM to quantitatively assess site-specific PTMs. Finally, we will give an overview of expert’s consensus on MRM methods design/validation and best practices to develop MRM assays intended to reach clinical application. The unique ability of MRM and similar methods to identify and quantify cardiac myofilament PTMs is likely to become central in answering important biological questions in the field of cardiac integrative physiology.
Received: March 28, 2014 Revised: May 29, 2014 Accepted: June 24, 2014
Keywords: Cardiac myofilaments / Heart failure
1
Introduction
The myofilament proteins are the molecular motors of cardiomyocytes. Satisfactory heart pumping function depends ultimately on the force generated by the sarcomere [1], which serves as both, a contractile apparatus and a metabolic signaling hub [2]. Activation of the myofilament during systole initiates with the release of activator Ca2+ from sarcoplasmic reticulum and results in cardiac muscle force generation and shortening through highly coordinated Ca2+ -dependent interactions between the thick and thin myofilament proteins.
Correspondence: Professor Anne M. Murphy, Department of Pediatrics, Division of Cardiology, 720 Rutland Avenue/Ross Bldg. 1144, Baltimore, MD 21205, USA E-mail: [email protected] Fax: +1-410-614-0699 Abbreviations: cTnI, cardiac troponin I; HF, heart failure; PKA, cAMP-dependent protein kinase; PKC, protein kinase C; WB, Western blot C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
The thin filaments, actin–tropomyosin–troponin (ATmTn), three states (blocked-closed-open) are defined in relation to their three distinct myosin-binding properties, the equilibrium between which responds to changes in intracellular Ca2+ [3–7]. A variety of PTMs have been reported to occur on thick and thin filament proteins, including phosphorylation, proteolysis, O-GlcNAcylation, acetylation, methylation, and oxidation, among others. Changes in PTMs modulate the response of myofilaments to calcium on a “beat to beat” and on a long-term basis [8–11]. In this review, we will discuss the new insights on site-specific phosphorylation offered by targeted proteomics techniques, such as MRM, and how this technique provided new insight on functional regulation by cardiac troponin I [12]. We will also discuss the discovery and clinical application potential of this powerful technique, if applied to other physiologically relevant myofilament proteins and other PTMs.
Colour Online: See the article online to view Figs. 1–4 in colour. www.clinical.proteomics-journal.com
544
2
G. A. Ramirez-Correa et al.
Roles of myofilament PTMs in regulating heart function in health and diseases
2.1 Regulation of cardiac function by dynamic myofilament phosphorylation and new cTnI functional sites by MRM Modulation of myofilament function by rapid phosphorylation in response to a physiological neurohumoral stimulus was first demonstrated 38 years ago for cardiac troponin I (cTnI) [13]. Since then a whole field of research was born, which focused on learning about each relevant myofilament site-specific phosphorylation site, their effects in physiological settings, and the kinases/phosphatases in charge of controlling their dynamic changes. However, assigning a physiological role to each myofilament or site-specific phosphorylation within the same protein, either in normal hearts or in the development of heart failure (HF), has been more daunting than expected [14]. Fortunately, the technological advances of MS-based targeted proteomics and its recent application in myofilament proteins has helped to provide integrative physiologic insight. Phosphorylation of myofilament proteins, such as troponin T, myosin-binding protein C, myosin light chain 2, alpha-tropomyosin, and titin, among others, is also critical for proper regulation of myocardial contractility, and there are excellent reviews covering this topic in depth [2, 8, 10, 15–23]. Here we will discuss the classical view of how cTnI phosphorylation regulates contractility. We will then consider the new insights on site-specific phosphorylation elucidated by advances in proteomics approaches. A classical example is the modulation of cTnI phosphorylation through -adrenergic stimulation as a critical factor for proper contractile function and for rapid adaptations to enhanced workload [9]. Perhaps even more important is the delicate balance of site-specific phosphorylation stoichiometry of cTnI. For example, cTnI is phosphorylated by cAMP-dependent protein kinase (PKA), cGMPdependent protein kinase (PKG), and/or by protein kinase D (PKD) at Ser-23 and Ser-24, and by protein kinase C (PKC) at three sites (Ser-43, Ser-45, and Thr-144) [24, 25]. All these sites have been found to be important for intrinsic regulation of heart contractility. Experimental evidence in Tg mice harboring single or combined phosphomimics or phosphosite ablation in PKA, PKC, or both have shown that alteration of phosphorylation stoichiometry alone may dramatically change force–frequency response [26–28] and the Frank–Starling law [29]. The changes in myocardial contractility arise from alterations in myofilament calcium–force dependence and relaxation kinetics [27,29–31]. In fact, there is strong evidence that cTnI phosphorylation influences three main rate-limiting steps of myocardial contractility, which includes the kinetics of cross-bridge cycling with actin, the kinetics of Ca2+ release from cTnC, and the co-
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Proteomics Clin. Appl. 2014, 8, 543–553
operativity of the myofilament apparatus [2]. For a summary, see the illustrations of cTnI showing the canonical phosphorylation sites of PKA (Ser-23, Ser-24) and PKC (Ser-43, Ser-45, and Thr-142) (Fig. 1A), with the described physiological consequences when studied in skinned muscles force–calcium (pCa2+ ) relationships (Fig. 1B). A summary of our previous study using two transgenic mouse models shows the effect of mimicking constitutive pseudophosphorylation in intact muscle twitch and Ca2+ transient dynamics (Fig. 1C). Note that the cTnI sequence nomenclature uses the rat sequence and excludes the methionine at position 1. The transgenic mice Tg cTnIDD22, 23 mimic constitutive pseudophosphorylation of PKA and display improved force–frequency response by means of generating the same or more force per unit of Ca2+ . In contrast, the transgenic mice Tg cTnIAD22,23 DD42, 44 mimic changes at PKA and PKC sites, believed to model what is present in human HF and display a lack of response to force–frequency challenge [27], which is a hallmark of human HF. Indeed, pioneering studies on human tissues and animal models suggested that HF may be associated with a decreased phosphorylation of cTnI at PKA sites [32], which later on was consistently corroborated by several groups [33,34]. See a representative Western blot (WB) of reduced cTnI Ser-23, Ser-24 phosphorylation in human HF (Fig. 1D), adapted from Frazier and colleagues [35]. However, it was not until recently that the phosphorylation status of cTnI at canonical PKC site (Thr142) in human myocardium was known, mainly due to the lack of phosphospecific antibodies. A recent study, employing a site-specific cTnI MRM-based assay in human heart, unraveled surprising results. Among these, a PKC site (Thr-142) turned out to be the most prominently cTnI-phosphorylated site, and to be, indeed, increased in ischemic HF (by 58%) and in idiopathic dilated cardiomyopathy (by 82%) [12]. This is followed in abundance by other PKC sites, Ser-41 and Ser43, respectively. This snapshot, taken by MRM, might help to clarify why the physiological effects of phosphorylation at Ser43, Ser-45, and Thr-142 dominate those of phosphorylation at Ser-23, Ser-24 [26–29, 36, 37]. The X-ray crystal structure of the human cardiac troponin complex was not available until 2003 [38], however, highresolution structural information on the N-terminus cTnI (1–34) containing Ser-23 and Ser-24 was prominently lacking. Later on, Howarth and colleagues [39] presented an NMR solution structure of N-terminus (1–32). Their structural information, in combination with molecular dynamic modeling, offered an explanation for the structural consequences of cTnI Ser-23, Ser-24 phosphorylation on inter- and intramolecular interactions of the troponin complex. To review the structure–function consequences of cTnI Ser-23, Ser-24 phosphorylation, we display the original structure of human cardiac troponin complex from Takeda and colleagues structure (1J1E.pdb) (Fig. 2A and B), and overimposed the NMR structure model (2JPW.pdb) of Howarth and colleagues in a front view (Fig. 1A) and in a 90⬚ axis rotation (Fig. 2B). A
www.clinical.proteomics-journal.com
Proteomics Clin. Appl. 2014, 8, 543–553
545
Figure 1. Physiological consequences of canonical cTnI PKA and PKC phosphorylation. (A) Schematic representation of cTnI and canonical PKA and PKC phosphorylation sites. (B) Schematic representation of skinned cardiac muscle force–calcium steady-state relationships, at baseline (solid line) and after PKA and PKC phosphorylation (dotted line). Red arrow indicates a decrease in calcium sensitivity and black arrow a decrease in maximal calcium activated force (Fmax ). (C) Force [Ca2+ ]i hysteresis loops in intact cardiac muscle from Ntg and two transgenic mouse models showing the effect of mimicking constitutive pseudophosphorylation of PKA and PKC sites on force–frequency response. In the loops, A is the point of minimum force and intracellular calcium, B represents maximal calcium levels before force development occurs, C is the point of maximal force before relaxation starts, black is 2 Hz and red is 4 Hz of stimulation frequency. Adapted from Ramirez-Correa and colleagues [27]. (D) Representative Western blot showing reduced cTnI Ser-23, Ser-24 phosphorylation in human heart failure, adapted from Frazier and colleagues [35].
molecular model of cTnI alone is also illustrated as a composite transparent surface and cartoon (Fig. 2C and D). Currently, it is generally accepted that phosphorylation of cTnI Ser-23, Ser-24 allows for a more structured state. This process not only stabilizes the intermolecular interactions of N-terminal and the inhibitory region of cTnI, but also weakens the Nterm helix (21–30) interactions with the N-term cTnC, where its physiological calcium affinity site is located. We first illustrate this interaction when phosphorylated at Ser-23, Ser-24 the cTnI N-term goes (2JPW.pdb) wrapped around the top of the N-term lobe of TnC (Fig. 2A and B). However, in a molecular model of unmodified cTnI (alone), we highlight the N-term in light pink, the known phosphorylatable sites as green spheres, and the newly found phosphorylated sites as red spheres, with the purpose of showing the complexity that phosphorylation at these residues brings to the understanding of cardiac troponin complex structure and function. See Table 1 for a summary of cTnI-phosphorylation sites, their potential physiological role, and clinical applications. Thus, although there are multiple levels of regulation, ample evidence indicates that even subtle changes in my C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ofilament proteins PTM stoichiometry, particularly in the regulatory subunits, such as cTnI, may profoundly impact myocardial contractility.
2.2 Regulation of myofilament function by other PTMs (O-linked GlcNAcylation, acetylation, and Ox-PTMs) and potential characterization by MRM In recent years, new PTMs have been identified on sarcomeric proteins that are substrates of classical PTMs, and equally important new PTMs that also modulate cardiac contractility have been identified. Newer PTMs range from O-GlcNAcylation [40–43], acetylation [44, 45], to a variety of PTMs that could be grouped under oxidative PTMs or Ox-PTMs [46, 47], such as S-nitrosylation, S-glutathionylation, sulfhydration, disulfide bonds formation, sulfenylation, sulfinic and sulfonic acid modifications. Although only a limited number of Ox-PTMs have been documented in myofilament proteins, the current vast landscape www.clinical.proteomics-journal.com
546
G. A. Ramirez-Correa et al.
Proteomics Clin. Appl. 2014, 8, 543–553
Figure 2. Troponin complex structure and molecular models of cTnI Ser-23, Ser-24 phosphorylation effects. (A) PyMol model of human cardiac troponin complex structure (1J1E.pdb) from Takeda and colleagues [38], and cTnI N-term model from NMR structure (2JPW.pdb) from Howarth and colleagues [39], overimposed in a front view. (B) Models from (A) in a 90⬚ axis rotation. (C) A PyMol molecular model of cTnI alone performed in Phyre2 (protein structure prediction web interface [90]) is illustrated as a composite transparent surface and cartoon, known phosphorylation sites are labeled in green and new sites in red. N-term residues 1–34 from the model are labeled in light pink. (D) Models from (C) in a 90⬚ axis rotation.
of myofilament PTMs leaves much to be discovered about the precise biological role of PTMs other than phosphorylation. Our group has studied O-linked glycosylation of the myofilament and found 32 O-GlcNAcylation sites on cardiac myofilaments to the array of newer PTMs that can modulate
cardiac contractility [43]. There is rapidly accumulating evidence for the functional importance of this PTM in diabetic [48] and hypertrophic or failing hearts [49]. O-GlcNAc modification cycles rapidly, like phosphorylation, occurs at serine and threonine residues of proteins, is highly abundant, and
Table 1. Summary of cTnI-phosphorylation sites, potential functional impact, and applications
Protein name
Site
Functional impact
cTnI S4, S5 (phosphorylation)
Interacts with inhibitory region when S22, S23 are phosphorylated Recessive mutations cause dilated cardiomyopathy S22, S23 ↓ Ca2+ sensitivity, ↑ Off rate of Ca2+ from TnC, force–frequency response, cross-bridge kinetics Y25 Similar to S22, S23? S41, S43 ↓ Ca2+ sensitivity, force–frequency response, maximal force T50
Does not change in HF
S76, T77 IT-arm region potential modulation with cTnT T142 ↓ Ca2+ sensitivity, force–frequency response, maximal force S150 Potential metabolic sensor, cAMPK target S165 Switch peptide between actin sites, next to histidine button (sensor of acidosis) T180 ? S198 ?
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Potential clinical application
Method
Refs.
HF serum marker?
MRM and LTQ-Orbitrap Velos
[12, 91]
MRM
[12] [12]
Reference for cTnI phosphorylation? HF serum marker?
MRM and LTQ-Orbitrap Velos MRM and LTQ-Orbitrap Velos
HF serum marker?
MRM
[12]
?
In vitro phosphorylation MRM
Not detected
Stage or prognosis marker?
HF or HCM serum marker? Stage or prognosis marker?
CRT responder?
MRM, LTQ-Orbitrap
[12, 92]
[12]
[12, 91]
www.clinical.proteomics-journal.com
547
Proteomics Clin. Appl. 2014, 8, 543–553
is involved in many cellular processes [50]. However, the progress in protein O-GlcNAcylation discovery has been limited, mainly due to the technical challenges associated to site-specific labeling and enrichment strategies, which in turn leads to no consistent site-specific identification. There are some proof of principle studies using MRM to estimate stoichiometry of O-GlcNAcylation in noncomplex samples [51, 52]. Currently, there is no standard method to be used in complex biological samples. Such a method would be desirable to study myofilament subproteome derived from normal subjects and disease conditions, that is, HF, diabetes, or cardiac hypertrophy. Classically, chemoenzymatic tagging techniques, such as Click iT or Bioortogonal chemistry, in combination with different enrichment strategies are used to site map O-GlcNAc modifications. In principle, these techniques are compatible with developing MRM assays to quantify OGlcNAc modifications in myofilament proteins. In addition to classic MRM approaches, an emerging MS approach is termed as SWATH-MS (sequential window acquisition for all theoretical fragment-ion spectra MS). SWATH-MS, which has been recently implemented by Dr. Aebersold’s group [53], is gaining popularity rapidly among the MS community [54] as a next-generation targeted proteomics technique [55]. Technically, as long as a mass shift (either known or hypothetical) is expected and internal standard peptides can be synthesized (with PTMs or chemical modifications), an MRM assay can be designed to identify and quantify PTMs in myofilament proteins. In addition, relative or absolute quantification has been successfully achieved when MRM has been
combined with other MS quantitative techniques. Some examples are SILAC, stable-isotope labeling by amino acids in mammals (SILAC Mice) [56], protein standard absolute quantification (PSAQ), ICAT, iTRAQ, nonisobaric tags for relative and absolute quantification (mTRAQ), [18 O] chemical incorporation during trypsinization, and AQUA synthetic reference peptides [57]. Thus, MRM holds a tremendous potential to characterize old and new myofilament PTMs and help in understanding important biological questions.
3
MRM and hyper-reaction monitoring: New applications for an old technique
3.1 History of MRM Currently, MRM is often used as synonymous to SRM. Since the development of what was referred to as SRM 3 decades ago, triple quadrupole mass spectrometers have been successfully used to identify and quantify small molecules and their metabolites, in urine, blood, and other body fluids [54, 55]. In one of the earliest studies, Kondrat and colleagues compared the sensitivity and specificity of MRM and complete massanalyzed ion kinetic energy spectra for analysis of organics compounds, such as cocaine [58]. In clinical chemistry, measurement of specific peptides by field desorption MS was first performed in the early 80s by using stable isotope labeled standards for neural peptides, as reviewed by Marx [54]. Triple quadrupole mass spectrometers instrumentation can, in principle, be operated also in discovery mode (Fig. 3A)
Figure 3. Schematic representation of triple quadrupole mass spectrometer used in discovery or targeted mode. (A) Schematic representation of a 4000 QTRAP hybrid triple quadrupole linear IT mass spectrometer (AB SCIEX) operated in discovery mode. (B) Same instrument operated in targeted mode, Q1 serves as first filter for precursor ions, Q2 is the fragmentation chamber, and Q3 is the second mass filter for preselected ions, the time of co-elution and m/z ratio intensity are used for relative or absolute quantification. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.clinical.proteomics-journal.com
548
G. A. Ramirez-Correa et al.
to identify as many peptides/proteins as possible. However these instruments are almost never used in this mode. On the other hand, they are specially designed for targeted mode (Fig. 3B), which provide the highest sensitivity to detect the low abundant proteins, and the peptide precursor and product ions (transitions) [59, 60]. High sensitivity is achieved by preselecting their unique mass, and multiple filters (Quads) achieve the greatest reproducibility and quantitative accuracy in samples. Although triple quadrupole mass spectrometers monitor one peptide precursor and one product ion at a given time (i.e., 40 ms), a method built with a list of peptides and their transitions allows the instrument for MRMs. For most of laboratories around the world, the assessment of protein expression and their PTMs depends on WB analysis. Although WB has been invaluable in protein-based research, its use as a widespread clinical application presents important limitations. The obstacles of WB include (i) restricted availability of specific antibodies (just a small portion of the proteome), (ii) antibody production is costly and lengthy, (iii) limited multiplex capabilities [61], (iv) inability to distinguish single amino acid mutants, and (v) inability to detect site-specific PTMs [62]. The inability to detect specific residues PTMs is problematic especially between adjacent phosphorylated sites, such as cTnI Ser-23, and Ser-24. Although site-specific PTMs antibody production is possible, it remains technically challenging and expensive. A technique widely used in the clinical settings is ELISA. Although ELISA is able to quantify and multiplex, however because it is dependent on antibodies it has similar limitations to WB analysis [61]. In order to translate proteomic and biomarker discoveries rapidly into clinical application, the biomedical community should ideally have access to a methodology that does not rely on antibodies. This methodology should be highly accurate and reproducible when quantifying a biomarker in complex samples. In addition, such methodology should be high-throughput and be readily multiplexed. Once validated in pilot studies and ultimately larger clinical populations, commercial quality assays that would yield clinically actionable results within 45–60 min from sample collection would be developed [61]. Recent technical developments in MRM assays position this technique as a leading methodology for translational applications.
3.2 Recent developments in MRM In the last decade, the scientific community has steadily extended the applications of MRM from metabolite or small molecules to peptide/protein quantification. Unlike small molecules or metabolites, proteins are large, thus they are typically studied by MS techniques after digestion into small peptides, usually by trypsin or other proteases. Distinct from small molecules, tryptic peptides yield more complex patterns of ion fragmentation, thus complicating selection of transitions to monitor [57]. In order to overcome the challenge of C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Proteomics Clin. Appl. 2014, 8, 543–553
Table 2. Summary of recent advances in MRM
Process
Improvement
Selection of target proteotypic peptides
Skyline and associated tools, access to Peptide Libraries, Peptide Atlas, Global Proteome Machine, NIST, and others. Skyline, SRM Atlas, MRM pilot, and others. Collision dissociation and ion fragmentation patterns; different yor b-ion intensities between instruments, that is, ion trap vs. triple quadrupole (QQQ). Empirical data from QQQ, hybrid quadrupole TOF and ion trap are preferred. However, the formula to estimate collision energy correlation to m/z ratios of b- and y-ions has been recently improved [93]. Crude synthetic peptide libraries and their QQ-type information is optimal for SRM development on low abundance proteins for high-throughput and high confidence assays [67]. Monitoring smaller windows centered on peptide predicted retention time circumvents the limits associated to the number of transitions and the LOD of SRM experiments. Enables reproducible and precise quantification of thousands of proteins in a single instrument run based on data-independent acquisition [82].
Transitions or ion-pair selection Collision energy protocols
Surrogate Peptide Libraries
Scheduled MRM data acquisition
Hyper-reaction monitoring
identifying the best peptides, and associated transitions to monitor for a given protein, one method described in detail by Grote et al. is to first do in silico analysis of digestion products but then to test and select the best peptides derived from purified protein for the biomarker of interest [63]. Other workflows and approaches, particularly for high-throughput analysis employ sophisticated bioinformatics tools and data set repositories [57]. The basics of targeted peptide selection and validation are briefly discussed in the section below, and more details are available in excellent methodological papers [59, 60]. Some of the recent developments in MRM [57,64–67] are summarized in Table 2. The versatility of proteomic applications for MRM continues to grow. A search of the term “Multiple-ReactionMonitoring” in the last 20 years (1993–2013) in PubMed leads to approximately 3300 entries and shows a steady increase in the number of applications of the methodology. Therefore in 2012, not surprisingly, MRM-targeted proteomics was recognized as the method of the year by the journal Nature Methods [54, 55]. An important MRM application to highlight is the www.clinical.proteomics-journal.com
549
Proteomics Clin. Appl. 2014, 8, 543–553
proteomic analysis of various PTMs, such as phosphorylation [68–73], tyrosine nitration [74], cysteine oxidation [75], acetylation [76], methylation [76], ubiquitinylation [76–79], and more recently, O-GlcNAcylation [51, 52]. Most, if not all, of these PTMs are potentially important on cardiac myofilament function. MRM is an application that will likely serve to answer the many biological questions that remain in the growing puzzle of site-specific PTMs and myofilament function. A promising and ongoing application of MRM is their use in translation medicine for biomarker validation, due to the high sensibility and reproducibility of the assay [61]. The advancements of LC-MS/MS instrumentation and software analysis will help to develop targeted peptide quantification with accuracy and sensitivity needed for use in clinical translation medicine [80, 81], one of the fastest growing areas in MS proteomics. Finally, other methods, which like SWATH are based on data-independent acquisition and hybrid instrument platforms, are referred as hyper-reaction monitoring because they enable reproducible and precise quantification of thousands of proteins in a single instrument run [82].
3.3 General MRM method design and validation With the help of commercial or open-source software, MRM methods can be designed to quantify a selected set of proteins (tens to a few hundreds). Selection of the proteins list to monitor is hypothesis-driven and case-dependent. The general principle relies upon the assumption that after uniform protein digestion, the many resulting peptides are present in equal amounts, although they display a wide range of m/z ratios in the first filter (Q1). A selected list of peptide precursor, in particular the double and triple charged, are subjected to fragmentation in the collision cell (Q2), several ion products (or transitions) from each precursor ion are generated. The signal abundance of product ions and their retention time is indicative of the target protein abundance and is recorded (Q3) and expressed as an AUC (Fig. 3B). The key for the success of an MRM method is picking measurable peptides and their specific transitions, preferentially based upon preliminary experimental data, then the method can be further refined. However, this step can also be based on previous studies or web data repositories/spectral libraries such as Peptide Atlas (www.peptideatlas.org/), National Institute of Standards and Technology (NIST) (http://peptide.nist.gov/), or Global Proteome Machine (GPM) (ftp://ftp.thegpm.org) [83], among others. Ideally, two to three peptides per a given protein should be picked and at least one of them should unequivocally correspond to the proteins of study (“proteotypic” peptide) [57]. In addition, software tools are available to predict proteotypic peptides and their likely ion fragmentation pattern during a real MS experiment, for example, SVM technique
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
for evaluating proteotypic peptides [84]. These software tools are especially useful when working on organisms, or a particular cell type or subproteome, for which previous data are not abundant or available. Finally, choosing appropriate transitions (product ion pairs) from each selected proteotypic peptide is the key step to increase selectivity. To calculate and optimize transitions in silico there are numerous available MRM method design software, for example, MIDASTM (MRM-initiated detection and sequencing) [72] and MRM pilot (AB SCIEX), however the gold standard seems to be the open-source Skyline [85] from MacCoss Lab, which has been installed more than 13 000 times [54]. Nonetheless, it is crucial that the ions selected by in silico have to be optimized prior to final assay for the best signals on different MS/MS platforms. Thus, validation is a necessary step to select the final transitions for the newly designed MRM method. In order to achieve absolute quantification, the sample is spiked with various concentrations of specific labeled synthetic peptides to generate a reference standard curve [86]. However, this approach restricts the number of peptides, to be quantified precisely, to only a few dozen [60]. The concepts summarized above are the general agreement and are reviewed more in depth on excellent methodological chapters, reviews, and tutorials [57, 59, 60, 62–64, 87, 88].
3.4 Application of MRM to quantitatively assess site-specific cTnI phosphorylation The innovative aspect of the work of Zhang and colleagues relies on their comprehensive coverage of all the potentially phosphorylated sites. In Fig. 4 we display a simplified workflow of human cTnI MRM, which is a first of its class study [12]. First, the authors evaluated appropriate transitions from the peptides in their mono-, bi-, or unphosphorylated status in silico. Second, unmodified and modified peptides were confirmed by LC-MS/MS on an LTQ Orbitrap platform. Based on these data and with the aid of MRM Pilot Software (AB SCIEX), three to five transitions (ion pairs) per peptide were chosen for monitoring. An internal standard peptide sequence (without PTMs) was designed to quantify total cTnI and was synthesized to harbor heavy nitrogen (N15 heavy). On the other hand, targeted peptides, both the mono- and biphosphorylated peptides were synthesized as the light versions (N13 light). Finally, only the optimal transitions were selected and instrument parameters were optimized manually for each of those 30 assays that consisted on a total of 145 transitions. Absolute quantification was achieved by calibration curves for each targeted transition, where the LODs, LOQs, and the CV were determined to then analyze the real experimental human heart samples, while spiking in known amounts of standards.
www.clinical.proteomics-journal.com
550
G. A. Ramirez-Correa et al.
Proteomics Clin. Appl. 2014, 8, 543–553
Figure 4. Troponin I workflow of MRM experiment in a triple quadrupole mass spectrometer. (A) Schematic representation of MRM assay development and validation by Zhang and colleagues [12], using an LTQ Orbitrap (Thermo Scientific) and a 4000 QTRAP hybrid triple quadrupole linear IT mass spectrometer (AB SCIEX). (B) Small human heart explanted samples (100 mg) were used for the study, only the bands corresponding to cTnI (corroborated by MALDI-TOF) were in-gel trypsin digested, potential phosphosites were corroborated in LTQ Orbitrap, synthetic standard were validated, stable isotope labeled standards were the nonphosphorylated version of the peptides, adapted from Zhang and colleagues [12].
4
Translational potential of targeted proteomics assays
As growing popularity of MRM and other targeted proteomic approaches in translational medicine, a group of experts in the scientific community recently described the use of targeted proteomic assays in different tiers [81]. In this publication, Tier 1 refers to development of targeted MS approach (such as MRM assay), which is sufficiently accurate for clinical or diagnostic use. This level of assay must have high-quality control, including the use of labeled reference standards for every analyte and must demonstrate a high degree of precision (CV
543
DOI 10.1002/prca.201400034
REVIEW
Targeted proteomics of myofilament phosphorylation and other protein posttranslational modifications Genaro A. Ramirez-Correa1 , Maria Isabel Martinez-Ferrando2 , Pingbo Zhang3 and Anne M. Murphy1 1
Department of Pediatrics/Division of Cardiology, Johns Hopkins University School of Medicine, Baltimore, MD, USA 2 Department of Comparative Biology and Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA 3 The Hopkins Bayview Proteomics Center, Johns Hopkins University School of Medicine, Baltimore, MD, USA Global cardiac myofilament protein phosphorylation levels, and their site-specific stoichiometry, are physiologically and clinically relevant for heart function. Unlike myofilament phosphorylation, other PTMs such as O-GlcNAcylation are just beginning to gain attention due to their potential physiological and clinical implications. This review will focus on what is currently known about cardiac troponin I phosphorylation, and on the potential physiological and clinical impact of targeted proteomics including new findings on cardiac troponin I sites and stoichiometry. We will then discuss the increasing recognition of other myofilament PTMs functional relevance and the potential of targeted MS approaches, particularly MRM, for accelerating their systematic characterization. In addition, we will broadly discuss the development and application of MRM to quantitatively assess site-specific PTMs. Finally, we will give an overview of expert’s consensus on MRM methods design/validation and best practices to develop MRM assays intended to reach clinical application. The unique ability of MRM and similar methods to identify and quantify cardiac myofilament PTMs is likely to become central in answering important biological questions in the field of cardiac integrative physiology.
Received: March 28, 2014 Revised: May 29, 2014 Accepted: June 24, 2014
Keywords: Cardiac myofilaments / Heart failure
1
Introduction
The myofilament proteins are the molecular motors of cardiomyocytes. Satisfactory heart pumping function depends ultimately on the force generated by the sarcomere [1], which serves as both, a contractile apparatus and a metabolic signaling hub [2]. Activation of the myofilament during systole initiates with the release of activator Ca2+ from sarcoplasmic reticulum and results in cardiac muscle force generation and shortening through highly coordinated Ca2+ -dependent interactions between the thick and thin myofilament proteins.
Correspondence: Professor Anne M. Murphy, Department of Pediatrics, Division of Cardiology, 720 Rutland Avenue/Ross Bldg. 1144, Baltimore, MD 21205, USA E-mail: [email protected] Fax: +1-410-614-0699 Abbreviations: cTnI, cardiac troponin I; HF, heart failure; PKA, cAMP-dependent protein kinase; PKC, protein kinase C; WB, Western blot C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
The thin filaments, actin–tropomyosin–troponin (ATmTn), three states (blocked-closed-open) are defined in relation to their three distinct myosin-binding properties, the equilibrium between which responds to changes in intracellular Ca2+ [3–7]. A variety of PTMs have been reported to occur on thick and thin filament proteins, including phosphorylation, proteolysis, O-GlcNAcylation, acetylation, methylation, and oxidation, among others. Changes in PTMs modulate the response of myofilaments to calcium on a “beat to beat” and on a long-term basis [8–11]. In this review, we will discuss the new insights on site-specific phosphorylation offered by targeted proteomics techniques, such as MRM, and how this technique provided new insight on functional regulation by cardiac troponin I [12]. We will also discuss the discovery and clinical application potential of this powerful technique, if applied to other physiologically relevant myofilament proteins and other PTMs.
Colour Online: See the article online to view Figs. 1–4 in colour. www.clinical.proteomics-journal.com
544
2
G. A. Ramirez-Correa et al.
Roles of myofilament PTMs in regulating heart function in health and diseases
2.1 Regulation of cardiac function by dynamic myofilament phosphorylation and new cTnI functional sites by MRM Modulation of myofilament function by rapid phosphorylation in response to a physiological neurohumoral stimulus was first demonstrated 38 years ago for cardiac troponin I (cTnI) [13]. Since then a whole field of research was born, which focused on learning about each relevant myofilament site-specific phosphorylation site, their effects in physiological settings, and the kinases/phosphatases in charge of controlling their dynamic changes. However, assigning a physiological role to each myofilament or site-specific phosphorylation within the same protein, either in normal hearts or in the development of heart failure (HF), has been more daunting than expected [14]. Fortunately, the technological advances of MS-based targeted proteomics and its recent application in myofilament proteins has helped to provide integrative physiologic insight. Phosphorylation of myofilament proteins, such as troponin T, myosin-binding protein C, myosin light chain 2, alpha-tropomyosin, and titin, among others, is also critical for proper regulation of myocardial contractility, and there are excellent reviews covering this topic in depth [2, 8, 10, 15–23]. Here we will discuss the classical view of how cTnI phosphorylation regulates contractility. We will then consider the new insights on site-specific phosphorylation elucidated by advances in proteomics approaches. A classical example is the modulation of cTnI phosphorylation through -adrenergic stimulation as a critical factor for proper contractile function and for rapid adaptations to enhanced workload [9]. Perhaps even more important is the delicate balance of site-specific phosphorylation stoichiometry of cTnI. For example, cTnI is phosphorylated by cAMP-dependent protein kinase (PKA), cGMPdependent protein kinase (PKG), and/or by protein kinase D (PKD) at Ser-23 and Ser-24, and by protein kinase C (PKC) at three sites (Ser-43, Ser-45, and Thr-144) [24, 25]. All these sites have been found to be important for intrinsic regulation of heart contractility. Experimental evidence in Tg mice harboring single or combined phosphomimics or phosphosite ablation in PKA, PKC, or both have shown that alteration of phosphorylation stoichiometry alone may dramatically change force–frequency response [26–28] and the Frank–Starling law [29]. The changes in myocardial contractility arise from alterations in myofilament calcium–force dependence and relaxation kinetics [27,29–31]. In fact, there is strong evidence that cTnI phosphorylation influences three main rate-limiting steps of myocardial contractility, which includes the kinetics of cross-bridge cycling with actin, the kinetics of Ca2+ release from cTnC, and the co-
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Proteomics Clin. Appl. 2014, 8, 543–553
operativity of the myofilament apparatus [2]. For a summary, see the illustrations of cTnI showing the canonical phosphorylation sites of PKA (Ser-23, Ser-24) and PKC (Ser-43, Ser-45, and Thr-142) (Fig. 1A), with the described physiological consequences when studied in skinned muscles force–calcium (pCa2+ ) relationships (Fig. 1B). A summary of our previous study using two transgenic mouse models shows the effect of mimicking constitutive pseudophosphorylation in intact muscle twitch and Ca2+ transient dynamics (Fig. 1C). Note that the cTnI sequence nomenclature uses the rat sequence and excludes the methionine at position 1. The transgenic mice Tg cTnIDD22, 23 mimic constitutive pseudophosphorylation of PKA and display improved force–frequency response by means of generating the same or more force per unit of Ca2+ . In contrast, the transgenic mice Tg cTnIAD22,23 DD42, 44 mimic changes at PKA and PKC sites, believed to model what is present in human HF and display a lack of response to force–frequency challenge [27], which is a hallmark of human HF. Indeed, pioneering studies on human tissues and animal models suggested that HF may be associated with a decreased phosphorylation of cTnI at PKA sites [32], which later on was consistently corroborated by several groups [33,34]. See a representative Western blot (WB) of reduced cTnI Ser-23, Ser-24 phosphorylation in human HF (Fig. 1D), adapted from Frazier and colleagues [35]. However, it was not until recently that the phosphorylation status of cTnI at canonical PKC site (Thr142) in human myocardium was known, mainly due to the lack of phosphospecific antibodies. A recent study, employing a site-specific cTnI MRM-based assay in human heart, unraveled surprising results. Among these, a PKC site (Thr-142) turned out to be the most prominently cTnI-phosphorylated site, and to be, indeed, increased in ischemic HF (by 58%) and in idiopathic dilated cardiomyopathy (by 82%) [12]. This is followed in abundance by other PKC sites, Ser-41 and Ser43, respectively. This snapshot, taken by MRM, might help to clarify why the physiological effects of phosphorylation at Ser43, Ser-45, and Thr-142 dominate those of phosphorylation at Ser-23, Ser-24 [26–29, 36, 37]. The X-ray crystal structure of the human cardiac troponin complex was not available until 2003 [38], however, highresolution structural information on the N-terminus cTnI (1–34) containing Ser-23 and Ser-24 was prominently lacking. Later on, Howarth and colleagues [39] presented an NMR solution structure of N-terminus (1–32). Their structural information, in combination with molecular dynamic modeling, offered an explanation for the structural consequences of cTnI Ser-23, Ser-24 phosphorylation on inter- and intramolecular interactions of the troponin complex. To review the structure–function consequences of cTnI Ser-23, Ser-24 phosphorylation, we display the original structure of human cardiac troponin complex from Takeda and colleagues structure (1J1E.pdb) (Fig. 2A and B), and overimposed the NMR structure model (2JPW.pdb) of Howarth and colleagues in a front view (Fig. 1A) and in a 90⬚ axis rotation (Fig. 2B). A
www.clinical.proteomics-journal.com
Proteomics Clin. Appl. 2014, 8, 543–553
545
Figure 1. Physiological consequences of canonical cTnI PKA and PKC phosphorylation. (A) Schematic representation of cTnI and canonical PKA and PKC phosphorylation sites. (B) Schematic representation of skinned cardiac muscle force–calcium steady-state relationships, at baseline (solid line) and after PKA and PKC phosphorylation (dotted line). Red arrow indicates a decrease in calcium sensitivity and black arrow a decrease in maximal calcium activated force (Fmax ). (C) Force [Ca2+ ]i hysteresis loops in intact cardiac muscle from Ntg and two transgenic mouse models showing the effect of mimicking constitutive pseudophosphorylation of PKA and PKC sites on force–frequency response. In the loops, A is the point of minimum force and intracellular calcium, B represents maximal calcium levels before force development occurs, C is the point of maximal force before relaxation starts, black is 2 Hz and red is 4 Hz of stimulation frequency. Adapted from Ramirez-Correa and colleagues [27]. (D) Representative Western blot showing reduced cTnI Ser-23, Ser-24 phosphorylation in human heart failure, adapted from Frazier and colleagues [35].
molecular model of cTnI alone is also illustrated as a composite transparent surface and cartoon (Fig. 2C and D). Currently, it is generally accepted that phosphorylation of cTnI Ser-23, Ser-24 allows for a more structured state. This process not only stabilizes the intermolecular interactions of N-terminal and the inhibitory region of cTnI, but also weakens the Nterm helix (21–30) interactions with the N-term cTnC, where its physiological calcium affinity site is located. We first illustrate this interaction when phosphorylated at Ser-23, Ser-24 the cTnI N-term goes (2JPW.pdb) wrapped around the top of the N-term lobe of TnC (Fig. 2A and B). However, in a molecular model of unmodified cTnI (alone), we highlight the N-term in light pink, the known phosphorylatable sites as green spheres, and the newly found phosphorylated sites as red spheres, with the purpose of showing the complexity that phosphorylation at these residues brings to the understanding of cardiac troponin complex structure and function. See Table 1 for a summary of cTnI-phosphorylation sites, their potential physiological role, and clinical applications. Thus, although there are multiple levels of regulation, ample evidence indicates that even subtle changes in my C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
ofilament proteins PTM stoichiometry, particularly in the regulatory subunits, such as cTnI, may profoundly impact myocardial contractility.
2.2 Regulation of myofilament function by other PTMs (O-linked GlcNAcylation, acetylation, and Ox-PTMs) and potential characterization by MRM In recent years, new PTMs have been identified on sarcomeric proteins that are substrates of classical PTMs, and equally important new PTMs that also modulate cardiac contractility have been identified. Newer PTMs range from O-GlcNAcylation [40–43], acetylation [44, 45], to a variety of PTMs that could be grouped under oxidative PTMs or Ox-PTMs [46, 47], such as S-nitrosylation, S-glutathionylation, sulfhydration, disulfide bonds formation, sulfenylation, sulfinic and sulfonic acid modifications. Although only a limited number of Ox-PTMs have been documented in myofilament proteins, the current vast landscape www.clinical.proteomics-journal.com
546
G. A. Ramirez-Correa et al.
Proteomics Clin. Appl. 2014, 8, 543–553
Figure 2. Troponin complex structure and molecular models of cTnI Ser-23, Ser-24 phosphorylation effects. (A) PyMol model of human cardiac troponin complex structure (1J1E.pdb) from Takeda and colleagues [38], and cTnI N-term model from NMR structure (2JPW.pdb) from Howarth and colleagues [39], overimposed in a front view. (B) Models from (A) in a 90⬚ axis rotation. (C) A PyMol molecular model of cTnI alone performed in Phyre2 (protein structure prediction web interface [90]) is illustrated as a composite transparent surface and cartoon, known phosphorylation sites are labeled in green and new sites in red. N-term residues 1–34 from the model are labeled in light pink. (D) Models from (C) in a 90⬚ axis rotation.
of myofilament PTMs leaves much to be discovered about the precise biological role of PTMs other than phosphorylation. Our group has studied O-linked glycosylation of the myofilament and found 32 O-GlcNAcylation sites on cardiac myofilaments to the array of newer PTMs that can modulate
cardiac contractility [43]. There is rapidly accumulating evidence for the functional importance of this PTM in diabetic [48] and hypertrophic or failing hearts [49]. O-GlcNAc modification cycles rapidly, like phosphorylation, occurs at serine and threonine residues of proteins, is highly abundant, and
Table 1. Summary of cTnI-phosphorylation sites, potential functional impact, and applications
Protein name
Site
Functional impact
cTnI S4, S5 (phosphorylation)
Interacts with inhibitory region when S22, S23 are phosphorylated Recessive mutations cause dilated cardiomyopathy S22, S23 ↓ Ca2+ sensitivity, ↑ Off rate of Ca2+ from TnC, force–frequency response, cross-bridge kinetics Y25 Similar to S22, S23? S41, S43 ↓ Ca2+ sensitivity, force–frequency response, maximal force T50
Does not change in HF
S76, T77 IT-arm region potential modulation with cTnT T142 ↓ Ca2+ sensitivity, force–frequency response, maximal force S150 Potential metabolic sensor, cAMPK target S165 Switch peptide between actin sites, next to histidine button (sensor of acidosis) T180 ? S198 ?
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Potential clinical application
Method
Refs.
HF serum marker?
MRM and LTQ-Orbitrap Velos
[12, 91]
MRM
[12] [12]
Reference for cTnI phosphorylation? HF serum marker?
MRM and LTQ-Orbitrap Velos MRM and LTQ-Orbitrap Velos
HF serum marker?
MRM
[12]
?
In vitro phosphorylation MRM
Not detected
Stage or prognosis marker?
HF or HCM serum marker? Stage or prognosis marker?
CRT responder?
MRM, LTQ-Orbitrap
[12, 92]
[12]
[12, 91]
www.clinical.proteomics-journal.com
547
Proteomics Clin. Appl. 2014, 8, 543–553
is involved in many cellular processes [50]. However, the progress in protein O-GlcNAcylation discovery has been limited, mainly due to the technical challenges associated to site-specific labeling and enrichment strategies, which in turn leads to no consistent site-specific identification. There are some proof of principle studies using MRM to estimate stoichiometry of O-GlcNAcylation in noncomplex samples [51, 52]. Currently, there is no standard method to be used in complex biological samples. Such a method would be desirable to study myofilament subproteome derived from normal subjects and disease conditions, that is, HF, diabetes, or cardiac hypertrophy. Classically, chemoenzymatic tagging techniques, such as Click iT or Bioortogonal chemistry, in combination with different enrichment strategies are used to site map O-GlcNAc modifications. In principle, these techniques are compatible with developing MRM assays to quantify OGlcNAc modifications in myofilament proteins. In addition to classic MRM approaches, an emerging MS approach is termed as SWATH-MS (sequential window acquisition for all theoretical fragment-ion spectra MS). SWATH-MS, which has been recently implemented by Dr. Aebersold’s group [53], is gaining popularity rapidly among the MS community [54] as a next-generation targeted proteomics technique [55]. Technically, as long as a mass shift (either known or hypothetical) is expected and internal standard peptides can be synthesized (with PTMs or chemical modifications), an MRM assay can be designed to identify and quantify PTMs in myofilament proteins. In addition, relative or absolute quantification has been successfully achieved when MRM has been
combined with other MS quantitative techniques. Some examples are SILAC, stable-isotope labeling by amino acids in mammals (SILAC Mice) [56], protein standard absolute quantification (PSAQ), ICAT, iTRAQ, nonisobaric tags for relative and absolute quantification (mTRAQ), [18 O] chemical incorporation during trypsinization, and AQUA synthetic reference peptides [57]. Thus, MRM holds a tremendous potential to characterize old and new myofilament PTMs and help in understanding important biological questions.
3
MRM and hyper-reaction monitoring: New applications for an old technique
3.1 History of MRM Currently, MRM is often used as synonymous to SRM. Since the development of what was referred to as SRM 3 decades ago, triple quadrupole mass spectrometers have been successfully used to identify and quantify small molecules and their metabolites, in urine, blood, and other body fluids [54, 55]. In one of the earliest studies, Kondrat and colleagues compared the sensitivity and specificity of MRM and complete massanalyzed ion kinetic energy spectra for analysis of organics compounds, such as cocaine [58]. In clinical chemistry, measurement of specific peptides by field desorption MS was first performed in the early 80s by using stable isotope labeled standards for neural peptides, as reviewed by Marx [54]. Triple quadrupole mass spectrometers instrumentation can, in principle, be operated also in discovery mode (Fig. 3A)
Figure 3. Schematic representation of triple quadrupole mass spectrometer used in discovery or targeted mode. (A) Schematic representation of a 4000 QTRAP hybrid triple quadrupole linear IT mass spectrometer (AB SCIEX) operated in discovery mode. (B) Same instrument operated in targeted mode, Q1 serves as first filter for precursor ions, Q2 is the fragmentation chamber, and Q3 is the second mass filter for preselected ions, the time of co-elution and m/z ratio intensity are used for relative or absolute quantification. C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.clinical.proteomics-journal.com
548
G. A. Ramirez-Correa et al.
to identify as many peptides/proteins as possible. However these instruments are almost never used in this mode. On the other hand, they are specially designed for targeted mode (Fig. 3B), which provide the highest sensitivity to detect the low abundant proteins, and the peptide precursor and product ions (transitions) [59, 60]. High sensitivity is achieved by preselecting their unique mass, and multiple filters (Quads) achieve the greatest reproducibility and quantitative accuracy in samples. Although triple quadrupole mass spectrometers monitor one peptide precursor and one product ion at a given time (i.e., 40 ms), a method built with a list of peptides and their transitions allows the instrument for MRMs. For most of laboratories around the world, the assessment of protein expression and their PTMs depends on WB analysis. Although WB has been invaluable in protein-based research, its use as a widespread clinical application presents important limitations. The obstacles of WB include (i) restricted availability of specific antibodies (just a small portion of the proteome), (ii) antibody production is costly and lengthy, (iii) limited multiplex capabilities [61], (iv) inability to distinguish single amino acid mutants, and (v) inability to detect site-specific PTMs [62]. The inability to detect specific residues PTMs is problematic especially between adjacent phosphorylated sites, such as cTnI Ser-23, and Ser-24. Although site-specific PTMs antibody production is possible, it remains technically challenging and expensive. A technique widely used in the clinical settings is ELISA. Although ELISA is able to quantify and multiplex, however because it is dependent on antibodies it has similar limitations to WB analysis [61]. In order to translate proteomic and biomarker discoveries rapidly into clinical application, the biomedical community should ideally have access to a methodology that does not rely on antibodies. This methodology should be highly accurate and reproducible when quantifying a biomarker in complex samples. In addition, such methodology should be high-throughput and be readily multiplexed. Once validated in pilot studies and ultimately larger clinical populations, commercial quality assays that would yield clinically actionable results within 45–60 min from sample collection would be developed [61]. Recent technical developments in MRM assays position this technique as a leading methodology for translational applications.
3.2 Recent developments in MRM In the last decade, the scientific community has steadily extended the applications of MRM from metabolite or small molecules to peptide/protein quantification. Unlike small molecules or metabolites, proteins are large, thus they are typically studied by MS techniques after digestion into small peptides, usually by trypsin or other proteases. Distinct from small molecules, tryptic peptides yield more complex patterns of ion fragmentation, thus complicating selection of transitions to monitor [57]. In order to overcome the challenge of C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Proteomics Clin. Appl. 2014, 8, 543–553
Table 2. Summary of recent advances in MRM
Process
Improvement
Selection of target proteotypic peptides
Skyline and associated tools, access to Peptide Libraries, Peptide Atlas, Global Proteome Machine, NIST, and others. Skyline, SRM Atlas, MRM pilot, and others. Collision dissociation and ion fragmentation patterns; different yor b-ion intensities between instruments, that is, ion trap vs. triple quadrupole (QQQ). Empirical data from QQQ, hybrid quadrupole TOF and ion trap are preferred. However, the formula to estimate collision energy correlation to m/z ratios of b- and y-ions has been recently improved [93]. Crude synthetic peptide libraries and their QQ-type information is optimal for SRM development on low abundance proteins for high-throughput and high confidence assays [67]. Monitoring smaller windows centered on peptide predicted retention time circumvents the limits associated to the number of transitions and the LOD of SRM experiments. Enables reproducible and precise quantification of thousands of proteins in a single instrument run based on data-independent acquisition [82].
Transitions or ion-pair selection Collision energy protocols
Surrogate Peptide Libraries
Scheduled MRM data acquisition
Hyper-reaction monitoring
identifying the best peptides, and associated transitions to monitor for a given protein, one method described in detail by Grote et al. is to first do in silico analysis of digestion products but then to test and select the best peptides derived from purified protein for the biomarker of interest [63]. Other workflows and approaches, particularly for high-throughput analysis employ sophisticated bioinformatics tools and data set repositories [57]. The basics of targeted peptide selection and validation are briefly discussed in the section below, and more details are available in excellent methodological papers [59, 60]. Some of the recent developments in MRM [57,64–67] are summarized in Table 2. The versatility of proteomic applications for MRM continues to grow. A search of the term “Multiple-ReactionMonitoring” in the last 20 years (1993–2013) in PubMed leads to approximately 3300 entries and shows a steady increase in the number of applications of the methodology. Therefore in 2012, not surprisingly, MRM-targeted proteomics was recognized as the method of the year by the journal Nature Methods [54, 55]. An important MRM application to highlight is the www.clinical.proteomics-journal.com
549
Proteomics Clin. Appl. 2014, 8, 543–553
proteomic analysis of various PTMs, such as phosphorylation [68–73], tyrosine nitration [74], cysteine oxidation [75], acetylation [76], methylation [76], ubiquitinylation [76–79], and more recently, O-GlcNAcylation [51, 52]. Most, if not all, of these PTMs are potentially important on cardiac myofilament function. MRM is an application that will likely serve to answer the many biological questions that remain in the growing puzzle of site-specific PTMs and myofilament function. A promising and ongoing application of MRM is their use in translation medicine for biomarker validation, due to the high sensibility and reproducibility of the assay [61]. The advancements of LC-MS/MS instrumentation and software analysis will help to develop targeted peptide quantification with accuracy and sensitivity needed for use in clinical translation medicine [80, 81], one of the fastest growing areas in MS proteomics. Finally, other methods, which like SWATH are based on data-independent acquisition and hybrid instrument platforms, are referred as hyper-reaction monitoring because they enable reproducible and precise quantification of thousands of proteins in a single instrument run [82].
3.3 General MRM method design and validation With the help of commercial or open-source software, MRM methods can be designed to quantify a selected set of proteins (tens to a few hundreds). Selection of the proteins list to monitor is hypothesis-driven and case-dependent. The general principle relies upon the assumption that after uniform protein digestion, the many resulting peptides are present in equal amounts, although they display a wide range of m/z ratios in the first filter (Q1). A selected list of peptide precursor, in particular the double and triple charged, are subjected to fragmentation in the collision cell (Q2), several ion products (or transitions) from each precursor ion are generated. The signal abundance of product ions and their retention time is indicative of the target protein abundance and is recorded (Q3) and expressed as an AUC (Fig. 3B). The key for the success of an MRM method is picking measurable peptides and their specific transitions, preferentially based upon preliminary experimental data, then the method can be further refined. However, this step can also be based on previous studies or web data repositories/spectral libraries such as Peptide Atlas (www.peptideatlas.org/), National Institute of Standards and Technology (NIST) (http://peptide.nist.gov/), or Global Proteome Machine (GPM) (ftp://ftp.thegpm.org) [83], among others. Ideally, two to three peptides per a given protein should be picked and at least one of them should unequivocally correspond to the proteins of study (“proteotypic” peptide) [57]. In addition, software tools are available to predict proteotypic peptides and their likely ion fragmentation pattern during a real MS experiment, for example, SVM technique
C 2014 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
for evaluating proteotypic peptides [84]. These software tools are especially useful when working on organisms, or a particular cell type or subproteome, for which previous data are not abundant or available. Finally, choosing appropriate transitions (product ion pairs) from each selected proteotypic peptide is the key step to increase selectivity. To calculate and optimize transitions in silico there are numerous available MRM method design software, for example, MIDASTM (MRM-initiated detection and sequencing) [72] and MRM pilot (AB SCIEX), however the gold standard seems to be the open-source Skyline [85] from MacCoss Lab, which has been installed more than 13 000 times [54]. Nonetheless, it is crucial that the ions selected by in silico have to be optimized prior to final assay for the best signals on different MS/MS platforms. Thus, validation is a necessary step to select the final transitions for the newly designed MRM method. In order to achieve absolute quantification, the sample is spiked with various concentrations of specific labeled synthetic peptides to generate a reference standard curve [86]. However, this approach restricts the number of peptides, to be quantified precisely, to only a few dozen [60]. The concepts summarized above are the general agreement and are reviewed more in depth on excellent methodological chapters, reviews, and tutorials [57, 59, 60, 62–64, 87, 88].
3.4 Application of MRM to quantitatively assess site-specific cTnI phosphorylation The innovative aspect of the work of Zhang and colleagues relies on their comprehensive coverage of all the potentially phosphorylated sites. In Fig. 4 we display a simplified workflow of human cTnI MRM, which is a first of its class study [12]. First, the authors evaluated appropriate transitions from the peptides in their mono-, bi-, or unphosphorylated status in silico. Second, unmodified and modified peptides were confirmed by LC-MS/MS on an LTQ Orbitrap platform. Based on these data and with the aid of MRM Pilot Software (AB SCIEX), three to five transitions (ion pairs) per peptide were chosen for monitoring. An internal standard peptide sequence (without PTMs) was designed to quantify total cTnI and was synthesized to harbor heavy nitrogen (N15 heavy). On the other hand, targeted peptides, both the mono- and biphosphorylated peptides were synthesized as the light versions (N13 light). Finally, only the optimal transitions were selected and instrument parameters were optimized manually for each of those 30 assays that consisted on a total of 145 transitions. Absolute quantification was achieved by calibration curves for each targeted transition, where the LODs, LOQs, and the CV were determined to then analyze the real experimental human heart samples, while spiking in known amounts of standards.
www.clinical.proteomics-journal.com
550
G. A. Ramirez-Correa et al.
Proteomics Clin. Appl. 2014, 8, 543–553
Figure 4. Troponin I workflow of MRM experiment in a triple quadrupole mass spectrometer. (A) Schematic representation of MRM assay development and validation by Zhang and colleagues [12], using an LTQ Orbitrap (Thermo Scientific) and a 4000 QTRAP hybrid triple quadrupole linear IT mass spectrometer (AB SCIEX). (B) Small human heart explanted samples (100 mg) were used for the study, only the bands corresponding to cTnI (corroborated by MALDI-TOF) were in-gel trypsin digested, potential phosphosites were corroborated in LTQ Orbitrap, synthetic standard were validated, stable isotope labeled standards were the nonphosphorylated version of the peptides, adapted from Zhang and colleagues [12].
4
Translational potential of targeted proteomics assays
As growing popularity of MRM and other targeted proteomic approaches in translational medicine, a group of experts in the scientific community recently described the use of targeted proteomic assays in different tiers [81]. In this publication, Tier 1 refers to development of targeted MS approach (such as MRM assay), which is sufficiently accurate for clinical or diagnostic use. This level of assay must have high-quality control, including the use of labeled reference standards for every analyte and must demonstrate a high degree of precision (CV
