Still with significant technological developments, the use of MRM–MS for protein quantification is not yet routine. This can be attributed to the tremendous time commitment to method development, optimization and validation for quantification of protein markers. While information regarding in silico
design is improving, MRM transitions for the peptides yielding the best precursor ion intensities and the most dominant MS/MS fragment ions are most effective if derived from empirical data, when available. The reason for this is because the accurate prediction of fragmentation patterns for any given instrument and peptide sequence is still difficult, and remains a research question which is discussed briefly below. Furthermore, selection of a fragment ion, the Q3 target, for peptide and protein quantification is even more challenging than small molecule quantification because one is obliged to select a fragment ion that contains the isotopic label to exploit the resolving power and mass difference in Q3. For example, it is common to isotopically label lysine and arginine residues in a sample preparation scheme that includes trypsin as a proteolytic enzyme. Unfortunately, the labeled lysine and arginine residue at the C-terminus will result in the entire b-ion series which will not contain the label and therefore should not be used as Q3 masses as they will not discriminate in Q3 from the native peptide. Thus, careful consideration must be taken in the selection of an MRM transition, as the most abundant fragment ion cannot always be selected. Furthermore, a key requirement of the MRM methodology is the ability to distinguish between correct identifications and false positives. Peptides are much larger and more complex than small molecules, share considerable homology and are measured in very complex matrices. Since each fragmentation product of a peptide only provides information about one position in that peptide, it is highly likely to have interference, i.e. other peptides resulting in the same transition as the peptide analyte in question [14
]. A schematic representation of peptide fragmentation is shown in highlighting some possible fragment ions after CID and outlines a comparison of SRM to MRM. By requiring multiple fragmentation product ions in the Q3 which together result in the same chromatographic elution time, the method can be highly specific providing confidence of the peptide identification in the absence of a full-scan tandem mass spectrum.
Different tryptic peptides from the same protein can produce ion currents differing by factors of 103
in LC–MS/MS experiments, excluding peptides that are not detected at all [15
]. This variation is due to multiple factors, including propensity to ionize in the electrospray source, coincidence in elution time with other easily ionizing peptides, efficiency of release during tryptic digestion and the presence of unrecognized post translational modifications (PTM) arising due to biology or artificially during sample preparation [16
]. A consistent response is required across the dynamic range of the analyte measured to ensure proper quantification of both native and isotope-labeled peptide pairs [17
]. A potential source of interference is in-source fragmentation of abundant peptides where the fragment ions rather than the precursors are the source of interference. This is caused by coincidence of the primary or secondary fragment of the precursor that has the same or nearly the same mass as the analyte transition of interest [18
]. This can be a significant issue for quantification and is dependent on the sensitivity level one is attempting to achieve. It is critical to select transition ions that maximize specificity and potentially minimize interferences from co-eluting species that fall within the mass windows and tolerances of the detector. Decreased mass resolution (1000–3000 FWHH) in the triple quadrupole instruments requires careful consideration of peptide separation so as to ensure accurate measurement of the peptide of choice because of the decreased separation of analytes at the same nominal masses [19
]. Therefore, in practice, initial LC–MS/MS experiments are performed on the biological sample to obtain preliminary information on the peptide characteristics, both ionization and fragmentation. Most commonly a full scan MS/MS spectrum in a linear ion-trap or TOF/TOF instrument is gathered. Once the precursor ion and fragmentation characteristics are noted, the precursor ion of interest and multiple fragment ions are chosen for MRM transitions and the sample is transferred to a triple quadrupole instrument. These transitions, based on the experimentally determined fragmentation pattern and CID parameters are further optimized on the triple quadrupole instrument. The sample is then re-analyzed with identical LC conditions to maintain the retention time with the goal of creating a set of MRM transitions with the best qualitative and quantitative properties per peptide of interest.
To address the tremendous time commitment and multiple rounds of analyses, sometimes on different instruments, required for the development of MRM–MS methodologies, a novel hybrid instrument (4000 QTRAP, Applied Biosystems) is now available which is both a triple quadrupole and a linear trap and offers a unique opportunity to design MRM–MS methods [20
]. Termed the MIDAS (MRM-initiated detection and sequencing) workflow [21
], when significant signal in a specific MRM transition is detected (both the parent mass in Q1 and fragment mass in Q3 are isolated and detected), the instrument switches the third quadrapole to ion trap mode and collects a full scan tandem mass spectrum. This full scan ion trap MS/MS obtained from the targeted detection can be submitted to a database search to confirm that the detected peptide is the peptide of interest and be utilized to further optimize the target for Q3. In this way the MIDAS workflow can develop MRM–MS methods in one step, eliminating the need to do preliminary method development confirming selectivity of peptide identification. Like traditional MRM–MS methodology development, the MIDAS workflow is still dependent on two things: (i) that the chosen MRM peptide ionizes efficiently enough to be measured by MS satisfying the first ion, Q1, of the MRM; and (ii) that the peptide fragments well enough and with sufficient intensity to generate the second ion, Q3, of the MRM. If using an in silico
methodology to predict ionizable peptides such as; a module in MIDAS, P3 (Proteotypic Peptide Predictor) [22
], TIQAM (Targeted Identification for Quantitative Analysis by MRM) [23
], as discussed below, or other in silico
prediction methods based on amino-acid sequence such as MDIP (Minimum Acceptable Detectability for Identified Peptides) [24
], APEX (Absolute Protein Expression) [25
] and those predicted are determined empirically to not ionize, another peptide that does ionize from the same protein can be tested and subsequently used instead. TIQAM is another software-derived workflow similar to MIDAS aimed at reducing the time commitment required for the development and validation of MRM assays [23
]. In this workflow, the proteome is extensively mapped utilizing standard global discovery proteomic work-flows. These discover data are then gathered into the PeptideAtlas database or similar user-derived database where those proteotypic peptides uniquely identifying proteins of interest are assembled to establish MRM transitions based on the validated MS2 spectrum. The user can then decide on how many rounds of MRM-triggered quantification experiments for this validated listing of transitions in order to gleam biologically useful information. TIQAM also offers a protein/peptide-centered-view to predict proteotypic peptides, perform in silico
digestions, and view the MRM trace further optimizing the method development.
Whether or not empirical data has been collected, it is recommended that MRM transitions are designed for chosen peptides approximately conforming to the following rules [1
- With electrospray ionization, doubly charged parent ions are preferred over singly charged for Q1, however a triply charged parent ion mass is assumed if the peptide sequence contains a histidine residue and can be used if the peptide has a mass greater than 1500 Da.
- y-ions are most often preferred over b-ions to ensure the transition contains a labeling site. Case in point, with the AQUA labeling strategy, it is common to isotopically label the C-terminal lysine and/or arginine due to trypsin proteolysis thus resulting in all y-ions isotopically-labeled. In contrast, the mTRAQ and iTRAQ reagents label the N-terminal and lysine side chains. Therefore, all b-ions and all y-ions of a C-terminal lysine peptide will be labeled.
- The Q3 fragment ion should have a greater m/z than the selected Q1 parent ion, should contain at least four amino acids and not be complimentary unless providing verification on the labeling site. This enhances selectivity and eliminates low mass ion overlap for multiply charged parent ions.
- If the sequence contains a proline residue, the high abundance y-ion created from fragmentation N-terminal to the proline is selected.
After choosing the most sensitive transition for quantification, it is recommended that two additional transitions are chosen to provide the most selectivity and combined reduced interference from other peptides. The aim being is to assemble in the method at least three transitions per peptide and it is recommended to have three peptides per protein with at least one peptide having a truly unique sequence upon a BLAST search. However, in practice, biomedical considerations and particulars regarding peptide length and sequence homology may not necessitate these recommendations. Additionally, since this methodology is truly measuring a peptide as a stochiometric representation of the protein in question, one must consider which peptides to use if there are known biological modifications to the protein such as truncation. If using these recommendations, one transition can be used for quantification and two transitions for qualitative identification of the peptide, as shown in , thereby collectively identifying the peptide with high confidence and allowing optimal cycle time for sensitivity and multiplexing multiple analyses. Alternatively, it is possible to use one charge state for quantitative calculation and the other for verification of identification or use both charge states and average them for quantification of all analyses, however this has yet to be shown in the literature.
Figure 4: Extracted ion chromatogram of AQUA MRM spectra for prostate specific antigen (PSA) peptide LSEPAELTDAVK. The heavy-labeled peptide (C-terminal K is labeled with 13C6 and 15N2) is extracted in the top panel. The bottom panel shows the light, i.e. endogenous (more ...)
Finally, because the MRM methodology is based on a specific peptide or peptides that are used as stochiometric representation of a protein, and proteins are quantified by comparing the peptide peak area with an internal standard, whether it is a relative or absolute standard, it is imperative that peptides utilized are selective to the particular protein isoform of interest. In the past, measurement of small nuclear polymorphisms (SNPs) and quantitative mRNA expression by PCR technologies have been well established which can easily differentiate between highly homologous isoforms [26–34
]. However, there still remains a need to develop similar methodologies for proteins. This is an area where MRM–MS is really gaining interest and momentum. Not only can a targeted approach such as MRM–MS lessen the ever challenging dynamic range issues most commonly encountered during global proteomics experiments thereby digging deeper into the low abundance proteome and low stochiometric PTM, this methodology is showing tremendous utility in the verification of both global proteomics identification and quantification data [12
]. Unfortunately, global proteomics data can be inconclusive. Often these data include one-peptide identifications or multiple peptide identifications which do not discriminate between highly homologous isoforms. Therefore, it is often a requirement to verify global quantitative proteomics results with orthogonal methodologies such as Western blotting. Unfortunately, there may not be an available antibody with or without PTM specificity needed or discriminatory antibodies which distinguish closely related isoforms. Additionally, quantification is less than absolute and these current immunological methodologies to verify proteomics experiments are low-throughput. Therefore, isotope-labeled peptide and protein standards are particularly useful for focusing instrument detection on low abundant proteins and PTMs (most commonly one-peptide identifications) and the measurement of highly homologous isoforms where multiple signature discriminatory peptides can be examined by MRM in a single experiment [40
]. Specificity can be achieved by the selection of sequences with no homology in BLAST, as well chromatographic retention standardization. MRM methodologies are particularly useful for measurement of highly homologous protein isoforms because of the high level of specificity achieved [11
]. Single amino-acid changes in a peptide sequence can often be distinguished by MRM enabling resolution and quantification of closely related isoforms. Furthermore, targeted detection at higher sensitivity often leads to information on lower abundance isoforms not detected by the intensity driven precursor selection of data dependent global strategies. Therefore, these aspects of MRM methodologies enable precise quantification of protein isoforms and overcome some of the limitations of classical shotgun quantification proteomic strategies thereby obtaining more sensitive, specific and selective information. For instance, Rosalind Jenkins et al
., utilized the ICAT labeling methodology to not only discriminate between highly homologous cytochrome P450 enzymes in liver microsomes but also extended their methodology to relatively quantify each of the isoenzymes with MRM–MS [11
]. Then utilizing synthesized stable isotopically-labeled peptides, they extended their methodology even further to determine absolute quantification of the isoenzymes.
Utility of MRM analysis
The confident identification of PTM have always been and still is an enormous challenge to proteomic specialists. The ability to detect phosphorylation, acetylation, glycosylation, and even ubiquination has become a specialized skill set because each requires knowledge in not only specialized sample preparation but also in MS method development. Because the control of reversible protein phosphorylation is central to most intra- and inter-cellular signal transduction, it is a most popular topic of interest to molecular biologists and pharamcologists. Unfortunately, development of good antibodies that recognize distinct phosphorylation states is a cumbersome, time-consuming and expensive process that requires a priori
knowledge of the proteins and its phosphorylation sites [17
]. Additionally, due to the high combinatory possibilities of multiple phosphorylation sites, many different antibodies need to be developed to fully understand the dynamic process [9
]. There is also a possibility of antibody cross-reactivity with different versions of multi-phosphorylated peptides leading to misleading quantification data and stochiometric analysis [5
]. Despite these issues, antibody-based approaches have been more favored for the determination of phosphorylation in signal transduction studies. However, by taking advantage of the characteristic fragmentation pattern of phosphopeptides, i.e. the NL of the phosphate group, it is been possible to identify phosphopeptides in the presence of non-phosphorylated peptides in the mass spectrometer. Unfortunately, this method often results in many non-phosphopeptides surreptitiously generating a similar NL, causing false positives that must be reconciled by advanced bioinformatic strategies, and most importantly, causes the mass spectrometer to waste scan time detecting non-phosphorylated peptides. However, by using a priori
biological knowledge, one can design a targeted MRM–MS-based method to achieve a higher level of selectivity and sensitivity. These MRM–MS-based targeted analyses of phosphorylation and quantification have been conducted [5–7
]. This approach is much different than other MS-based techniques for identifying phosphopeptides, like NL scan or triple play experimentation as it is highly selective and can generate absolute quantification of the dynamic stoichiometric phosphorylation events.
Acetylation is another PTM of great biological interest and is arguably easier to detect in a mass spectrometer than phosphorylation as the modification is not as liable chemically. As such, it is often searched for indiscriminately as a variable modification during database algorithm searching. This however greatly increases the search time and cost. While MRM–MS has not been commonly used to focus on acetylation as a PTM like phosphorylation, Griffiths et al
., demonstrated an interesting approach [8
]. Since acetylation can be detected by the characteristic reporter ion of 126.1 m/z
after CID, corresponding to the immonium ion of the acetyl–lysine, the authors utilized a MRM–MS approach to detect acetylation in a test case proteins. Using the MIDAS workflow, the authors generated a listing of the proteins’ amino-acid sequence and the respective sequences that are purported to contain an acetylated lysine. For the methodology, Q1 precursor scanning was set to dynamically scan over a mass range of 450–1200 m/z
. Q3 was held to only transmit the acetyl–lysine immonium ion of 126.1 m/z
. If and when a transition was detected, Q3 switched to ion trap and captured the full scan tandem mass spectra thereby allowing database searching and sequencing of the peptide. These authors later compared this MRM–MS-based approach to a typical data-dependent MS approach and found the MRM approach to be approximately 10-fold more sensitive.
Probably one of the most chemically complex PTM of great biological interest, especially in the area of biomarker discovery, is glycosylation. Utilizing a combination of N
-glycosite isolation and stable-isotope-labeled standard MRM–MS, Stahl-Zeng et al
., analyzed and quantified N
-glycosylated peptides in plasma detecting sub-ng/ml concentrations [45
]. These authors also prove that this targeted analysis can accurately quantify over a dynamic range of five orders of magnitude. This is an additional order of magnitude over what can even be detected in a global proteomics experiment, let alone quantified, demonstrating a great power of MRM–MS-based proteomics approaches. Also, utilizing MRM–MS to monitor the asparagine-linked glycosylation, Hulsmeier et al
] detected several serum glycoproteins to diagnose deficiencies of N
-glycosylation biosynthetic pathways in congenital disorders of glycosylation. These deficiencies, associated with endocrine and coagulation disorders, manifest themselves with different N
-glycosylation site occupancies and are hypothesized to correlate to severity of disease. Prior to this publication, the exact degree of protein under-glycosylation had not been determined due to lack of a precise measurement. Utilizing MRM methodologies, Hulsmeier et al
. was able for the first time, measure precisely N
-glycosylation and correlate the degree of N
-glycosylation and severity of disease. Like other PTM measured and analyzed with the targeted MRM–MS approach, Sahana Mollah et al
. utilized the MIDAS workflow to identify conclusively four ubiquitinated sites on recombinant receptor interacting protein [43
Additional targeted proteomics methods
Other directed MS-based analyses that are used for targeted approaches are inclusion list scanning, NL, immonium ion detection and PIM. Inclusion list scanning, listing precursor masses of interest for detection rather than scanning entire mass ranges [47
], NL scanning and immonium ion detection can be used individually or combined with MRM–MS approaches as shown and referenced above and therefore will not be discussed in detail here. PIM, theoretically, is a compilation of different MS methods. In its purest form, PIM is most akin to selected ion monitoring (SIM) which is an older technique that was used to increase sensitivity and dynamic range of detection because instruments did not have the faster scan times found today. In a SIM methodology, an instrument is focused on measuring only one ion at any given time. It is now mostly used to computationally extract a particular ion from complex spectra for ease of visualization. Like SIM, PIM is only measuring one ion at any given time, but it is the product ion after CID that is monitored as opposed to the precursor ion in a SIM method. A newer version of PIM was recently pioneered by Vathany Kulasingam et al
]. These authors utilized PIM as a technique specifically for ion trap instruments in lieu of triple quadrupoles. Because precursor mass selection, CID and detection of the product ions of interest are all occurring in the same compartment, i.e. space of the instrument it is possible to monitor all product ions from fragmentation of one precursor ion. This will provide additional structural information without increasing the duty cycle of the instrument. Showing proof of principle, in combination with antibody capture, the most sensitive sample purification methodologies, the authors were able to achieve a limit of detection of 10 amol with a coefficient variation of <20% with a linear range of over four orders of magnitude. An interesting compartment of the literature, albeit scant, is the comparisons of these targeted methodologies. All work done in this area concludes that MRM or other MS-based targeted methodologies enhance sensitivity and selectivity over global proteomics methodologies. For example, Hasmik Keshishian et al
., noted that the sensitivity of targeted MS analysis enhances the lower detection limit for peptides by up to 100-fold when compared to unbiased MS analysis [18
]. Additionally, Charanjit Sandhu et al
., also compared a data-dependent acquisition, where a precursor ion is selected for fragmentation-based on intensities with a targeted peptide monitoring, where predefined m/z
ratios are selected for fragmentation to MRM. They conclude that the targeted peptide monitoring far exceeds the data-dependent acquisition methodology for sensitivity.
These targeted MS-based methodologies are often combined or compared against affinity-based methodologies. While most of these methodologies are well seasoned, some novel approaches have been developed and combined with MRM–MS. It had been shown that two different antibodies used in a sandwich immunoassay can provide extreme sensitivity. However, the limited ability to multiplex immunoassays make it necessary to use an alternative method. SISCAPA (Stable Isotope Standards and Capture by Anti-Peptide Antibodies) was designed and pioneered by Anderson et al
., to address this issue. In SISCAPA, specific tryptic peptides are selected as stochiometric representatives of a protein. Anti-peptide antibodies are generated and immobilized on nanoaffinity columns to enrich these specific peptides along with spiked stable-isotope-labeled internal standards. Both the heavy (internal standards) and light peptides are enriched from a complex matrix such as human serum with the immobilized antibodies. The enriched sample is then introduced into a mass spectrometer for analysis. Both relative and absolute quantification information about the protein of interest is then gathered [49
]. This method was further optimized for greater precision and accuracy by Whiteaker et al
., reporting a physiologically relevant sensitivity in the ng/mL range [50
]. Whiteaker et al
. also utilized the a combination of antibody and MS approaches to confirm global proteomics results from mouse model of breast cancer while characterizing and transferring a couple of the proteins identified previously to a plasma biomarker [38
]. Finally, Kornilayev et al
., extended the SISCAPA approach developed by Anderson et al
. by utilizing polyclonal monospecific anti-peptide antibodies raised against unique cytochrome P450 isoenzymes specific tryptic peptides for qualitative and quantitative proteomic analysis. The point of this approach is that every isoenzyme contains a unique tryptic peptide(s) that can be used for differential analysis. Chemical proteomics is also a novel affinity-based methodology that is gaining popularity. Pioneered by Bantscheff et al
., small molecule non-selective kinase inhibitors were immobilized onto beads (kinobeads) and incubated with whole cell lysate in an effort to bind both known and novel protein targets of the inhibitors. Coupled with iTRAQ labeling, the authors were able to quantitatively profile both known protein targets, asses the binding affinity and map the phosphorylation status of the captured proteome [51