- Email: [email protected]
Progress in Mass Spectrometry Acquisition Approach for Quantitative Proteomics
CHINESE JOURNAL OF ANALYTICAL CHEMISTRY Volume 42, Issue 12, December 2014 Online English edition of the Chinese language journal
Cite this article as: Chin J Anal Chem, 2014, 42(12), 1859–1868.
REVIEW
Progress in Mass Spectrometry Acquisition Approach for Quantitative Proteomics ZHANG Wei* Thermo Fisher Scientific, Shanghai 201206, China
Abstract: As the major approach for quantitative proteomics, classic quantitative mass spectrometry (MS) faces new challenges, such as interferences from complex matrices and limits on analytical throughput. Recent progresses in MS technologies, including development of synchronous precursor selection (SPS), mass defect isobaric labeling, parallel reaction monitoring (PRM), multiplexing acquisition (MSX), and various novel data-independent acquisition (DIA) strategies, provide viable solutions for problems in relative and absolute quantification in proteomics. This review analyzed the current bottlenecks in the field of quantitative proteomics, summarized the most recent advances in quantitative MS acquisition, and highlighted the characteristics and the advantages of these new techniques in quantitative proteomics. Key Words: Quantitative proteomics; Synchronous precursor selection; Parallel reaction monitoring; Data-independent acquisition; Review
1
Introduction
Currently, the focus of proteomics has shifted from qualitative to quantitative studies. Quantitative proteomics requires quantitation of the proteome of a cell, tissue, or even entire organism, and plays a significant role in investigation of mechanism of basic biological processes as well as biomarker identification and validation[1,2]. Proteome quantitation can be divided into relative and absolute quantification[3]. Relative quantitation is often applied in comparative studies, which involves large-scale, high-throughput quantitative MS analyses of samples collected under normal or pathological conditions aiming to identify the precise proteome differences. Two major techniques, stable isotope labeling or label-free strategies, can be employed in such studies[4,5]. Absolute quantitation seeks to determine absolute quantity of specific protein, which is typically achieved by monitoring unique peptides of target protein, and calculating concentrations based on peak area ratios to known amount of peptide standards (external method) or isotope-labeled “heavy” peptide standards (internal method). Selected/multiple
reaction monitoring (SRM/MRM) of unique peptides is the most frequently employed MS approach[6]. Stable isotope labeling is the classic approach for relative quantitative proteomics. The labeled samples are mixed prior to MS analysis, so that relative quantitation can be achieved in one run, avoiding the instability in LC separation and minimizing error. Stable isotopes can be introduced metabolically (SILAC)[7], enzymatically (18O labeling)[8], and chemically (dimethyl labeling)[9], whereas relative quantitation is usually based on the chromatogram peak area ratio of precursor ions. However, MS1 based quantitation suffers from low labeling efficiency, narrow dynamic range, and poor sensitivity, and thus isobaric labeling based MS2 quantitation is gaining more and more attention[10]. Using isobaric labeling, the same peptide originated from different samples displays identical m/z at MS1 level, and thereby improves method sensitivity significantly. Meanwhile the reporter ions with different m/z are generated at MS2 level, providing quantitation information as well as b/y ion identification. In addition, a wider dynamic range can be attained with this approach. iTRAQ[11] and TMT[12] are two major isobaric
Received 10 September 2014; accepted 18 October 2014 * Corresponding author. Email: [email protected] Copyright © 2014, Changchun Institute of Applied Chemistry, Chinese Academy of Sciences. Published by Elsevier Limited. All rights reserved. DOI: 10.1016/S1872-2040(14)60788-X
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
labeling reagents, with labeling capacity of 8-plex and 6-plex, respectively. Despite these improvements, isobaric labeling methods face interference issues caused by co-eluting peptides. Specifically, on the LC side, the matrices of proteomic samples are usually very complex, with a significant number of peptides co-eluted from the column; whereas on the MS side, an isolation window of 2 m/z is typically used when filtering precursor ions for MS2 analysis. In combination, these technical pitfalls result in the lack of discrimination between target peptide and co-eluting near-m/z peptides, leading to erroneous incorporation of the reporter ions of co-eluting peptides, compromising quantitative accuracy[13,14]. Ting et al[15] has shown that the co-eluting peptides in complex samples can dramatically affect reporter ion intensity, distort peptide and protein ratios, leading to underestimation of proteome changes. Such problems have become the bottleneck in isobaric labeling method development. Triple quadrupole-based SRM (or MRM), the gold standard for MS quantitation, is widely used for absolute quantitation of proteins[6]. SRM monitors the precursor and product ion mass of specific peptides. The first quadrupole (Q1) filters for the precursor ions, and Q3 selects for product ions after precursor fragmentation in Q2. By monitoring the signal of unique precursor-product transition, interfering ions can be mostly excluded. Peptide confirmation and quantitation can be achieved either with external standard using calibration curves, or by spiking with known amount of stable isotope-labeled peptide internal standards[6,16]. There is no doubt that SRM method offers high sensitivity and wide dynamic range, and is indispensible for target protein verification and quantitation. However, accompanying the progress of quantitative proteomics, SRM is challenged with more complex sample matrices, lower abundance of target proteins and ease of suppression by high abundance components. In addition, the low resolution SRM fails to effectively exclude interferences from complex matrices, and is prone to false positive results[17,18]. On the other hand, the demand on analysis throughput also increases rapidly, with the need to monitor tens of thousands of transitions in one analytical run, while SRM method can only manage a limited number of transitions simultaneously due to its speed and sensitivity limit[19]. Moreover, optimization of SRM transition and collision energy can be overwhelming for proteomic research, especially for those biomarker and systems biology studies requiring large sample sizes[20,21]. Therefore, absolute proteome quantitation also faces intimidating technical challenges. Recently, technical improvement of high-resolution mass spectrometers (such as Orbitrap), and innovations in data acquisition strategies, including isobaric labeling-based synchronous precursor selection (SPS) and mass defect isobaric labeling, as well as high-resolution parallel reaction monitoring (PRM) and multiplexing acquisition (MSX) for
SRM type analysis, and various novel data-independent acquisition (DIA) strategies, have brought new opportunities to solve these challenges in quantitative proteomics. 2
Progress in isobaric labeling technique: synchronous precursor selection (SPS) and mass defect isobaric labeling
As mentioned earlier, isobaric labeling techniques represented by iTRAQ and TMT face the challenges on two fronts: (1) massive co-eluting near-m/z peptides co-fragment with target peptide, which affect the measurement of reporter ion intensity, and distort quantitative results[13‒15]; (2) isobaric tags have limited labeling capacity. Taking iTRAQ as an example, its reporter group (N-methylpiperazine) consists 6 C and 2 N, with limiting labeling capacity to 8-plex, thus unsuitable for higher sample throughput[22]. Although larger reporter group supports higher labeling capacity, it hurts sensitivity[23]. To solve the problem caused by co-eluting peptides, Ting et al[15] analyzed TMT-labeled samples using MS3 scan in LTQ-Orbitrap (Fig.1). They first employed relatively low collision energy for MS2 CID (35%) fragmentation to generate b/y ions for sequence identification without over-fragmenting the TMT tags; then within the 110%–160% m/z range of parent ion, selected the product ion with highest intensity to undergo thorough reporter group fragmentation by high energy HCD (60%) and to determine reporter ion ratios with MS3 detection. They demonstrated that, through selection on both precursor ion and product ion level, elimination of co-eluting interference can be achieved, yielding a quantitative ratio matching the theoretical one. For example, in a model two-proteome peptide mixture sample from Lys-C protein digests of yeast (mimicking sample) and the human HeLa cell line (functioned as interference), when the theoretical ratio was 10:1, classic MS2 analysis only got about 3, while MS3 reported a ratio of 11.7, which was very close to the theoretical ratio. However, signal intensity drops significantly in MS3 results compared to MS2, leading to a much lower sensitivity of MS3 in quantitation despite its high quantitation accuracy, thus limiting its application in real sample analysis[24]. Although the recently developed TMTc[25] and QuantMode [26] methods are effective in dealing with co-eluting peptides, these methods involve tedious procedures without offering a quantitative accuracy comparable to that of MS3. Synchronous precursor selection (SPS) technique completely solved the signal intensity drawback of MS3. SPS uses MultiNotch technique which allows for synchronous isolation of multiple ions (≤ 15, Fig.2A) in a linear trap[27]. With this technique, SPS method can select multiple b/y fragments from the target peptide in MS2 for MS3 experiment, resulting in accumulation of reporter ions and a substantial improvement
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
Fig.1 Schematic elucidation of MS3 method[15]. (A) Acquisition workflow of MS3, (B) Comparison of MS3 and classic MS2 results
Fig.2 Schematic elucidation of SPS MS3 method[24,28] 3
(A) Acquisition workflow of SPS MS ; (B) Influence of classic MS2, MS3 and SPS MS3 for reporter ion intensity; (C) Comparison of sensitivity of MS3 and SPS MS3; (D) Comparison of accuracy of classic MS2 and SPS MS3
of MS3 signal (Fig.2B)[24,28]. Experimental results have demonstrated that the spectra generated using SPS method for MS3 quantitation (CID fragmentation in MS2, then select multiple fragment ions for MS3 HCD fragmentation and Orbitrap detection) are comparable to MS2 spectra, and the success rate of quantitation is over 70% which is similar to that of classic MS2 (Fig.2C)[24,28]. Moreover, compared to the reporter ion from the single product ion, the accumulation of reporter ion from multiple product ions in MS3 gives rise to more stable ratios, and consequently better reproducibility and accuracy (Fig.2D). Besides, SPS MS3 quantitation is compatible with trypsin digested samples which are more common than Lys-C digestion required in traditional MS3, empowering SPS MS3 with wider applicability in routine
quantitative proteomics[29]. Weekes et al[30] investigated the process of CMV infection of fibroblasts and underlying mechanism using SPS MS3 method. They obtained precise quantitative data on over 8000 proteins (including 1184 cell surface proteins) at 8 time points during infection, conducted real-time dynamic quantitative analysis, and established the field of quantitative temporal viromics. As an attempt to improve the throughput limit of isobaric labeling tags, Dephoure et al[31] combined the triplex metabolic labeling SILAC and 6-plex isobaric tag TMT to quantitate 18 samples in parallel and obtained satisfactory results. But the expansion of labeling capacity was not achieved fundamentally until the introduction of mass defect isobaric labeling in recent years. More specifically, subtle
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
mass defect differences exist between 13C/12C and 15N/14N due to nuclear binding energy, and mass defect isobaric labeling utilizes the 6.32 mDa difference between 13C14N and 12C15N to realize higher labeling capacity by replacing 13C with 15N (Fig.3A). For example, by replacing one 13C atom with 15N in TMT127 and TMT129 reporter group of TMT 6-plex, one will get TMT127L (12C8H1615N1+, 127.1247608 Da)/TMT127H (12C713C1H1614N1+, 127.1310808 Da) and TMT129L (12C613C2H1615N1+, 129.1314705 Da)/TMT129H (12C513C3H1614N1+, 129.1377904 Da), raising labeling capacity to 8-plex[32]. Similarly, TMT 10-plex and 18-plex can also be accomplished without altering reporter group structure, thus providing an effective solution to the multiplexing capacity issue of parallel samples in quantitative proteomics[33,34]. On the other hand, in the range of m/z 100–200, as to the 6.32 mDa difference, a resolution of over 50000 is required to achieve baseline separation (Fig.3B), which is beyond the capacity of common high resolution mass spectrometers[33]. But for the fourier-transformation-based ultra-high resolution Orbitrap and FT-ICR whose resolution is inversely proportional to m/z, discrimination of mass defect labels can be easily realized[34]. Currently, mass defect labeling based-TMT 10-plex has been commercialized, and widely applied in quantitative proteomics studies demanding large sample size, such as those in systems biology and clinical biomarker discovery (Fig.3C). As elaborated above, synchronous precursor selection (SPS), mass defect isobaric labeling and other recently developed techniques have solved many issues in traditional isobaric labeling methodology, and proven to be powerful tools in relative quantitative proteomics.
3
Progress in target protein quantitation: parallel reaction monitoring (PRM) and multiplexing acquisition (MSX)
SRM filters and monitors the precursor and product ions using low-resolution triple quadrupoles. For small molecular compounds whose structures often differ significantly from each other, SRM is extraordinarily successful in differentiating target compound and matrix background[35]. However, owing to the close similarity of peptides, SRM method fails to completely eliminate matrix interferences in proteomics, undermining the accuracy and specificity of quantitation[17,18], not to mention that selection and optimization of SRM transitions can be very timeconsuming[36]. SRM method also benefits from the technical advances of mass spectrometry. For example, high resolution SRM (H-SRM) shrinks the precursor isolation window from 0.7 m/z down to 0.2 m/z and thereby reduces the matrix interferences[37,38]. However, selection efficiency also decreases with isolation window, rendering H-SRM more suitable for small molecule analysis. Another option for increasing selectivity is to incorporate an additional stage of mass filtering to form MRM3 scanning[39]. But unfortunately, MRM3 faces the same problems-decreased sensitivity and increased cycle time. Development of hybrid quadrupole-high resolution mass spectrometers (e.g., Q-Orbitrap) and enhancement of scan rate provides brand new avenues for quantitative proteomics. Parallel reaction monitoring (PRM) method running on
Fig.3 Increasing labeling capacity by mass defect isotopes[33] (A) Principle of mass defect labeling (TMT-10); (B) Minimum resolution for discrimination of 6 mDa mass difference. (C) Mass spectrum of mass defect reporter ions (TMT-10)
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
platforms such as Q-Orbitrap is just the high resolution ion monitoring technique in contrast to the classic SRM[40,41]. Different from SRM, PRM scans all product ions of the precursor, i.e. it monitors every transition of the precursor ion in parallel. PRM first selects precursor ion in the quadrupole (Q1) with a window ≤ 2 m/z, then fragments the precursor in collision cell (Q2) to generate product ions, and finally uses Orbitrap in place of Q3 to conduct high resolution-accurate mass (HR/AM) full scan for all product ions and complete the PRM acquisition (Fig.4)[42]. PRM displays several advantages over classic SRM: (1) High resolution product ion monitoring. With a ppm level mass accuracy (usually < 3 ppm when using external standards, and < 1 ppm using internal standards), the interferences can be effectively eliminated without losing sensitivity; (2) Full scan at MS2. The optimization of transition and collision energy is avoided, and the chromatogram peak of target compound can be extracted simply by selecting the best product ion (s) during data processing stage. Moreover, linear quantitative dynamic range of PRM spans 5‒6 orders of magnitude; (3) Simultaneous identification and quantitation. The MS2 full scan spectrum supports peptide identification, while product ion (s) with the best signal can be used as transition (s) for quantitation. Peterson et al[40] compared the selectivity and sensitivity of SRM and PRM using 14 peptide standards and found that both methods could reach a limit of quantitation (LOQ) around amol level in neat sample. However, in yeast matrix, the LOQ of SRM method raised to fmol level, one order of magnitude higher than the amol LOQ of PRM method (Fig.5A), demonstrating the vulnerability of SRM method to matrix interferences (Fig.5B). In addition, Gallien et al[42] examined the LOQ of 175 transitions from 35 heavy isotope-labeled peptides in urine matrix. Their results showed that SRM achieved better LOQ for only 19% of the transitions. Owing to its advantages, PRM has found its way into life sciences research in recent years. Tsuchiya et al[43] employed PRM monitoring and quantitation to compare the ubiquitylation
level of ubiquitin-proline-β-galactosidase (Ub-P-bgal) in wildtype and mutant type with UFD4 (ubiquitin fusion degradation pathway gene) knockout. By investigating the change of linkage sites between ubiquitin chain and protein in these two types of cell, the authors demonstrated that UFD4 is responsible for the K29-linked ubiquitin chains. In another study, Tang et al[44] applied PRM to verify 55 sites for methylation, acetylation, propionylation and other modifications of histone H3 and H4. They also quantitated the modification levels, and identified several new modification sites, such as H3 K18 methylation sites, and H4 K20 acetylation sites which had not been discovered before. However, PRM falls short on analytical throughput compared to SRM. Since the typical effective scan rate of high-resolution mass spectrometer is only 10–15 Hz, far below that of SRM which can easily handle hundreds of transitions, while PRM can simultaneously monitor 10‒15 precursor ions at best. Fortunately, the solution to this problem is emerging. Multiplexing (MSX) technique is one exemplary way to improve PRM throughput. MSX takes advantage of the C-trap in front of Orbitrap, and different precursor ions can be sequentially selected by quadrupole and stored in C-trap, ready for Orbitrap to finish last scan. The mixture of precursor ions in C-trap is then injected into Orbitrap for analysis.
Fig.4
Principle of selected reaction monitoring (SRM) (A) and parallel reaction monitoring (PRM) (B)[40]
Fig.5 Comparison of sensitivity and selectivity of SRM/PRM[40] (A) Comparison of LOQs of peptides in neat sample and yeast matrix. (B) Comparison of extracted ion chromatograms of peptide GVSAFSTWEk in yeast matrix
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
In this fashion, simultaneous monitoring of as many as 10 precursor ions can be achieved, with a 10-times increase in method throughput[45]. Gallien et al[46] combined MSX and PRM to establish the MSX-PRM method. By quadruplexing acquisition and retention time segmenting (into 1.5–2.5 min windows), they quantitatively monitored 770 target peptides (corresponding to 436 proteins) in yeast digest under a 60-min LC gradient elution. In their study, as many as 60 co-monitoring peptides with the same retention time window were observed. The authors managed to keep a cycle time of less than 2 s, and the throughput matched that of SRM method. Throughput of target peptide can be further boosted by optimizing the parameters such as resolution and maximum ion injection time. For example, simultaneous monitoring of 128 peptides is possible when MSX is set to 8[42,45]. In summary, the advent of high-resolution PRM and MSX-PRM technology has solved many issues in classic SRM-based quantitative proteomics, and provides powerful tools for target protein verification and absolute quantitation.
4
Progress in data-independent acquisition: a variety of Orbitrap-based DIA techniques
Shotgun proteomics, the classic strategy and primary relatively quantitative approach in proteomics, is based on data-dependent acquisition (DDA), in which MS2 acquisition depends on the MS1 result, leading to loss of low abundance peptides, stochasticity in analysis, and un-even distribution of scan points. Targeted approaches such as SRM, on the other hand, fail to collect the information outside of the target list, let alone the low throughput limit. To solve these problems, data-independent acquisition (DIA), a data acquisition strategy combining the features of DDA and SRM, is emerging in recent years. DIA divides the entire scan range into several isolation windows, then sequentially isolates and fragments in each window, acquiring data on the product ions of all precursor ions in the window. DIA offers an array of advantages over DDA and SRM[47]. DIA does not require target peptide, has no throughput limit, and generates scan points that are evenly distributed. Sequence verification and selection of product ions for quantitation can be achieved with the aid of spectrum library, and the data can be reviewed post-acquisition retrospectively. Venable and colleagues[48] first demonstrated the use of DIA strategy in relative quantitation of 15N-labeled yeast proteome by sequentially isolating and fragmenting precursor windows of 10 m/z in a LTQ ion trap mass spectrometer. Their results showed that chromatogram extracted from mass spectra obtained through the DIA strategy offered significantly better S/N ratio, reproducibility, and quantitative accuracy than those through DDA strategy, and the number of quantitated proteins increased by 87%. Gillet et al[49]
developed a Q-TOF (TripleTOF)-based SWATH technique which utilized 32 of consecutive 25 m/z windows, and showed that the selectivity of SWATH is comparable to that of SRM. Development of SWATH expanded the use of DIA in quantitative proteomics[50–52]. Meanwhile, Q-Orbitrap (Q Exactive)-based DIA method, using the same 25 m/z step size, improved selectivity in quantitating complex samples owing to its higher resolution. Un-segmented DIA methods, such as MSE and AIF, also found their way in quantitative proteomics. However, due to fragmenting and monitoring all ions at the same time without dividing them into different m/z windows, these methods suffer from matrix interferences and are un-suitable for quantitative analysis of complex samples[49,53]. Limited scan rate is the Achilles heel of high-resolutionMS-based DIA methods. These methods usually require a wide isolation window of 25 m/z to ensure sufficient scan points for quantitation, which on the other hand sacrificed quantitation results by allowing for more interferences (Fig.6A)[54]. Recent development in DIA methods helps narrow DIA isolation window, further improving DIA selectivity, sensitivity, and reproducibility. The MSX technique described earlier has also been applied in DIA methods. Egertson et al[54] developed a MSX-DIA method using the MSX function of Q-Orbitrap, wherein five separate, random 4-m/z isolation windows were sequentially selected, accumulated, and injected into Orbitrap for analysis. In this fashion, the total step size of MSX-DIA was 20-m/z-wide, leaving scan rate unaffected, but the actual isolation window was only 4-m/z-wide (Fig.6A). The resulting data were deconvoluted using the software Skyline to organize each fragment into specific windows, realizing a selectivity matching that of 4-m/z isolation window (Fig.6B). MSX-DIA can offer selectivity comparable to that of DDA, minimizing interferences from co-eluting peptides and contaminants. The emergence of Q-OT-qIT tribrid mass spectrometer made it possible for DIA advancement[55]. Since the LTQ and Orbitrap of Q-OT-qIT MS work independently in parallel, Orbitrap can be dedicated to MS1 scan, while the faster LTQ with a scan rate of 20 Hz is devoted to MS2 DIA scan. The high resolution MS1 spectra are then used for chromatogram peak extraction and quantitation, whereas the MS2 spectra only support peptide identification and confirmation. This arrangement significantly improves scan rate, and allows for narrower isolation windows by relieving the scan point concern from MS2 DIA scans. WiSIM-DIA and Full MS-DIA are precisely the DIA methods adopting such workflow (Fig.7)[56,57]. WiSIM-DIA uses precursor ions from 200 m/z wide isolation window-SIM (WiSIM) scans for quantification and ion trap MS2 DIA scan spectra collected at 12 m/z step size in parallel for confirmation. The ultra-high resolution (240000) WiSIM scans significantly improves precursor
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
Fig.6 Schematic elucidation of DIA and MSX-DIA[54] (A) Comparison of DIA and MSX-DIA; (B) Deconvolution of MSX-DIA spectra for 4 m/z selectivity
Fig.7 Schematic elucidation of WiSIM-DIA and Full MS-DIA[57] (A) Workflow of WiSIM-DIA; (B) Workflow of Full MS-DIA
selectivity, while the accurate precursor mass based chromatogram extraction and quantitation offers better sensitivity than classic DIA product ion-based quantitation (Fig.7A). As for peptide confirmation, despite that WiSIM-DIA MS2 spectrum are acquired at lower resolution compared to classic DIA, the decrease of isolation window size from 25 m/z to 12 m/z significantly eliminated matrix interferences. Full MS-DIA pushes this strategy further, its MS2 acquisition is conducted at 3 m/z step size, with a few ultra-high resolution
(240k) Orbitrap MS full scans inserted to guarantee sufficient data points for precursor scans. Similar to WiSIM-DIA, Full MS-DIA also uses precursor chromatogram for quantitation and MS2 DIA scans for confirmation (Fig.7B). It is worth pointing out that the selectivity of Full MS-DIA conducted with 3 m/z isolation window is comparable to that of DDA, allowing Full MS-DIA results to be used as low resolution data (precursor ±1.5 Da) in database searching, waiving the need for a spectrum library, and realizing the integration of
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
DDA and DIA[57]. In support of DIA development, algorithms using DIA data directly for database searching as well as methods for result validation are also being developed, allowing for direct DIA-data-based database searching and peptide identification. It has been shown that better performance can be obtained using DIA methods than DDA in protein identification. Last but not least, the recently developed pSMART technique achieves WiSIM-DIA-type of acquisition on a Q-Orbitrap platform by inserting several high-resolution Orbitrap MS scans into MS2 scans collected with 5 m/z isolation windows[58], yielding better sensitivity, selectivity, and reproducibility than classic DIA in both qualitative and quantitative studies. DIA surpasses DDA shotgun method and SRM-based targeted monitoring in both relative and absolute protein quantitation, looking into a bright future in quantitative proteomics. However, DIA suffers from long cycle time currently. It is only compatible with nano-LC and requires slow gradient to obtain sufficient LC peak width. This drawback limits DIA application and puts forward a challenge for its advancement.
5
Conclusions
Isobaric labeling, target ion monitoring, and data-independent acquisition have dominated quantitative proteomics. Table 1 summarizes and compares the principle of these three approaches. MS technology is constantly facing challenges presented by proteomics as the field advances. Currently, stable isotope labeling-based relative quantitation and SRM-based absolute quantitation methods both suffer from the limitations such as matrix interferences and insufficient throughput (Table 1). To solve these problems, a variety of high-resolution mass spectrometer-based methods are developed: synchronous precursor selection (SPS) and mass defect isobaric labeling may solve the interference and throughput problems for relative quantitation; parallel reaction monitoring (PRM) and multiplexing acquisition (MSX) can improve SRM selectivity and have become the new pathway to absolute quantitation; while data-independent acquisition (DIA) strategies combine the advantages of DDA and SRM; multiplexing and tribrid mass spectrometers allow for smaller step size in DIA scans, empowering DIA to expand its application in high-throughput quantitative proteomics. It is not hard to foresee that these new technologies will be used more and more to solve the problems in quantitative proteomics in place of conventional MS techniques, such as elucidating protein interaction and discovering clinical biomarkers.
Table 1 Principle and progress of quantitative proteomic methods Reporter ion quantification Major techniques
iTRAQ, TMT
Method principle
Data-dependent acquisition of isobaric labeled samples, and comparison of reporter ion intensities in MS2
Application fields Technical limitations Technique progress
Large scale relative quantification and comparison Inaccurate ratios for complex samples, limited labeling capacity Synchronous precursor selection (SPS), Mass defect isotopic labeling
Target ion monitoring SRM/MRM Selection and acquisition of specific fragment ions isolated from target precursor ion, and extraction of transitions for quantification Target proteins validation and absolute quantification Severe interference from complicated matrix Parallel reaction monitoring (PRM), Multiplexed PRM (MSX-PRM)
References
376–386 [8]
[1]
Ong S E, Mann M. Nat. Chem. Biol., 2005, 1(5): 252–262
[2]
Veenstra T D. J. Chromatogr. B, 2007, 847(1): 3–11
[3]
Zhou Y, Shan Y C, Zhang L H, Zhang Y K. Chinese Journal of Chromatography, 2013, 31(6): 496–502
[4]
Bantscheff M, Schirle M, Sweetman G, Rick J, Kuster B. Anal. Bioanal. Chem., 2007, 389(4): 1017–1031
[5]
Zhu J L, Zhang K, He X W, Zhang Y K. Chinese J. Anal. Chem., 2010, 38(3): 434–441
[6] [7]
Lange V, Picotti P, Domon B, Aebersold R. Mol. Syst. Biol.,
Data-independent acquisition SWATH, DIA Untargeted acquisition of all fragment ions from all precursor ions within scan range, and extraction of transitions for quantification Large scale, high-throughput relative and absolute quantification Severe interference due to wide isolation window (> 20 m/z) Multiplexed DIA (MSX-DIA), WiSIM-DIA, Full MS-DIA, pSMART
Capelo J L, Carreira R J, Fernandes L, Lodeiro C, Santos H M, Simal-Gandara J. Talanta, 2010, 80(4): 1476–1486
[9]
Boersema P J, Raijmakers R, Lemeer S, Mohammed S, Heck AJ. Nat. Protoc., 2009, 4(4): 484–494
[10] Koehler C J, Strozynski M, Kozielski F, Treumann A, Thiede B. J. Proteome Res., 2009, 8(9): 4333–4341 [11] Thompson A, Schafer J, Kuhn K, Kienle S, Schwarz J, Schmidt G, Neumann T, Johnstone R, Mohammed A K, Hamon C. Anal. Chem., 2003, 75(8): 1895–1904 [12] Ross P L, Huang Y N, Marchese J N, Williamson B, Parker K,
2008, 4: 222
Hattan S, Khainovski N, Pillai S, Dey S, Daniels S,
Ong S E, Blagoev B, Kratchmarova I, Kristensen D B, Steen H,
Purkayastha S, Juhasz P, Martin S, Bartlet-Jones M, He F,
Pandey A, Mann M. Mol. Cell. Proteomics, 2002, 1(5):
Jacobson A, Pappin D J. Mol. Cell. Proteomics, 2004, 3(12):
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
1154–1169 [13] Karp N A, Huber W, Sadowski P G, Charles P D, Hester S V, Lilley K S. Mol. Cell. Proteomics, 2010, 9(9): 1885–1897 [14] Ow S Y, Salim M, Noirel J, Evans C, Rehman I, Wright P C. J. Proteome Res., 2009, 8(11): 5347–5355 [15] Ting L, Rad R, Gygi S P, Haas W. Nat. Methods, 2011, 8(11): 937–940 [16] Zhao Y, Ying W T, Qian X H. Chem. Life, 2008, 28(2): 210–213 [17] Sherman J, McKay M J, Ashman K, Molloy M P. Proteomics, 2009, 9(5): 1120–1123 [18] Abbatiello S E, Mani D R, Keshishian H, Carr S A. Clin. Chem., 2010, 56(2): 291–305
[35] Gallien S, Duriez E, Demeure K, Domon B. J. Proteomics, 2013, 9(81): 148–158 [36] Karlsson C, Malmstrom L, Aebersold R, Malmstrom J. Nat. Commun., 2012, 3: 1301 [37] Gallart-Ayala H, Moyano E, Galceran M T. J. Chromatogr A., 2008, 1208(1-2): 182–188 [38] Martínez-Villalba A, Moyano E, Martins C P, Galceran M T. Anal. Bioanal. Chem., 2010, 397(7): 2893–2901 [39] Fortin T, Salvador A, Charrier J P, Lenz C, Bettsworth F, Lacoux X, Choquet-Kastylevsky G, Lemoine J. Anal. Chem., 2009, 81(22): 9343–9352 [40] Peterson A C, Russell J D, Bailey D J, Westphall M S, Coon J J. Mol. Cell. Proteomics, 2012, 11(11): 1475–1488
[19] Kiyonami R, Schoen A, Prakash A, Peterman S, Zabrouskov V,
[41] Schiffmann C, Hansen R, Baumann S, Kublik A, Nielsen P H,
Picotti P, Aebersold R, Huhmer A, Domon B. Mol. Cell.
Adrian L, von Bergen M, Jehmlich N, Seifert J. Anal. Bioanal.
Proteomics, 2011, 10(2): M110.002931 [20] Cima I, Schiess R, Wild P, Kaelin M, Schüffler P, Lange V, Picotti P, Ossola R, Templeton A, Schubert O, Fuchs T, Leippold T, Wyler S, Zehetner J, Jochum W, Buhmann J, Cerny T, Moch H, Gillessen S, Aebersold R, Krek W. Proc. Natl. Acad. Sci. USA, 2011, 108(8): 3342–3347 [21] Picotti P, Bodenmiller B, Mueller LN, Domon B, Aebersold R. Cell, 2009, 138(4): 795–806 [22] Pichler P, Kocher T, Holzmann J, Mazanek M, Taus T, Ammerer G, Mechtler K. Anal. Chem., 2010, 82(15): 6549–6558 [23] Thingholm T E, Palmisano G, Kjeldsen F, Larsen M R. J. Proteome Res., 2010, 9(8): 4045–4052 [24] McAlister G C, Nusinow D P, Jedrychowski M P, Wühr M, Huttlin E L, Erickson B K, Rad R, Haas W, Gygi S P. Anal. Chem., 2014, 86(14): 7150–7158 [25] Wuhr M, Haas W, McAlister G C, Peshkin L, Rad R, Kirschner M W, Gygi S P. Anal. Chem., 2012, 84(21): 9214–9221
Chem., 2014, 406(1): 283–291 [42] Gallien S, Duriez E, Demeure K, Domon B. J. Proteomics, 2013, 81: 148–158 [43] Tsuchiya H, Tanaka K, Saeki Y. Biochem. Biophys. Res. Commun., 2013, 436(2): 223–229 [44] Tang H, Fang H, Yin E, Brasier A R, Sowers L C, Zhang K. Anal. Chem., 2014, 86(11): 5526–5534 [45] Gallien S, Bourmaud A, Kim S Y, Domon B. J. Proteomics, 2014, 100: 147–159 [46] Gallien S, Duriez E, Crone C, Kellmann M, Moehring T, Domon B. Mol. Cell. Proteomics, 2012, 11(12): 1709–1723 [47] Law K P, Lim Y P. Expert Rev. Proteomics, 2013, 10(6): 551–566 [48] Venable J D, Dong M Q, Wohlschlegel J, Dillin A, Yates J R. Nat. Methods, 2004, 1(1): 39–45 [49] Gillet L C, Navarro P, Tate S, Rost H, Selevsek N, Reiter L, Bonner R, Aebersold R. Mol. Cell. Proteomics, 2012, 11(6): O111.016717
[26] Wenger C D, Lee M V, Hebert A S, McAlister G C, Phanstiel D
[50] Liu Y, Huttenhain R, Surinova S, Gillet L C, Mouritsen J,
H, Westphall M S, Coon J J. Nat. Methods, 2011, 8(11):
Brunner R, Navarro P, Aebersold R. Proteomics, 2013, 13(8):
933–935 [27] Goeringer D E, Asano K G, McLuckey S A. Anal. Chem., 1994, 66(3): 313–318 [28] Viner R, Bomgarden R, Blank M, Rogers J. 61st ASMS, 2013, Poster W617 [29] Blank M, Bomgarden R, Rogers J, Jacobs R, Fong J, Puri N, Zabrouskov V, Viner R. 61st ASMS, 2013, Poster Th449 [30] Weekes M P, Tomasec P, Huttlin EL, Fielding C A, Nusinow D, Stanton R J, Wang E C, Aicheler R, Murrell I, Wilkinson G W, Lehner P J, Gygi S P. Cell, 2014, 157(6): 1460–1472 [31] Dephoure N, Gygi S P. Sci. Signal, 2012, 5(217): rs2 [32] Werner T, Becher I, Sweetman G, Doce C, Savitski M M, Bantscheff M. Anal. Chem., 2012, 84(16): 7188–7194 [33] McAlister G C, Huttlin E L, Haas W, Ting L, Jedrychowski M P, Rogers J C, Kuhn K, Pike I, Grothe R A, Blethrow J D, Gygi SP. Anal. Chem., 2012, 84(17): 7469–7478 [34] Werner T, Sweetman G, Savitski M F, Mathieson T, Bantscheff M, Savitski M M. Anal. Chem., 2014, 86(7): 3594–3601
1247–1256 [51] Collins B C, Gillet L C, Rosenberger G, Rost H L, Vichalkovski A, Gstaiger M, Aebersold R. Nat. Methods, 2013, 10(12): 1246–1253 [52] Lambert J P, Ivosev G, Couzens A L, Larsen B, Taipale M, Lin Z Y, Zhong Q, Lindquist S, Vidal M, Aebersold R, Pawson T, Bonner R, Tate S, Gingras A C. Nat. Methods, 2013, 10(12): 1239–1245 [53] Chapman J D, Goodlett D R, Masselon C D. Mass Spectrom. Rev., 2013: 10.1002/mas.21400 [54] Egertson J D, Kuehn A, Merrihew G E, Bateman N W, MacLean B X, Ting Y S, Canterbury J D, Marsh D M, Kellmann M, Zabrouskov V, Wu C C, MacCoss M J. Nat. Methods, 2013, 10(8): 744–746 [55] Senko M W, Remes P M, Canterbury J D, Mathur R, Song Q, Eliuk S M, Mullen C, Earley L, Hardman M, Blethrow J D, Bui H, Specht A, Lange O, Denisov E, Makarov A, Horning S, Zabrouskov V. Anal. Chem., 2013, 85(24): 11710–11714
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
[56] Kiyonami R, Patel B, Senko M, Zabrouskov V, Egertson J, Ting S, MacCoss M, Rogers J, Huhmer A. Large Scale Targeted
[57] Zhang W, Reiko K, Jiang Z, Chen W. Chinese J. Anal. Chem., 2014, 42(12): 1750–1758
Protein Quantification Using WiSIM-DIA Workflow on a
[58] Prakash A, Peterman S, Ahmad S, Sarracino D, Frewen B,
Orbitrap Fusion Tribrid Mass Spectrometer. ASMS, 2014,
Vogelsang M, Byram G, Krastins B, Vadali G, Lopez M. J.
W737
Proteome Res., 2014, DOI:10.1021/PR5003017
Cite this article as: Chin J Anal Chem, 2014, 42(12), 1859–1868.
REVIEW
Progress in Mass Spectrometry Acquisition Approach for Quantitative Proteomics ZHANG Wei* Thermo Fisher Scientific, Shanghai 201206, China
Abstract: As the major approach for quantitative proteomics, classic quantitative mass spectrometry (MS) faces new challenges, such as interferences from complex matrices and limits on analytical throughput. Recent progresses in MS technologies, including development of synchronous precursor selection (SPS), mass defect isobaric labeling, parallel reaction monitoring (PRM), multiplexing acquisition (MSX), and various novel data-independent acquisition (DIA) strategies, provide viable solutions for problems in relative and absolute quantification in proteomics. This review analyzed the current bottlenecks in the field of quantitative proteomics, summarized the most recent advances in quantitative MS acquisition, and highlighted the characteristics and the advantages of these new techniques in quantitative proteomics. Key Words: Quantitative proteomics; Synchronous precursor selection; Parallel reaction monitoring; Data-independent acquisition; Review
1
Introduction
Currently, the focus of proteomics has shifted from qualitative to quantitative studies. Quantitative proteomics requires quantitation of the proteome of a cell, tissue, or even entire organism, and plays a significant role in investigation of mechanism of basic biological processes as well as biomarker identification and validation[1,2]. Proteome quantitation can be divided into relative and absolute quantification[3]. Relative quantitation is often applied in comparative studies, which involves large-scale, high-throughput quantitative MS analyses of samples collected under normal or pathological conditions aiming to identify the precise proteome differences. Two major techniques, stable isotope labeling or label-free strategies, can be employed in such studies[4,5]. Absolute quantitation seeks to determine absolute quantity of specific protein, which is typically achieved by monitoring unique peptides of target protein, and calculating concentrations based on peak area ratios to known amount of peptide standards (external method) or isotope-labeled “heavy” peptide standards (internal method). Selected/multiple
reaction monitoring (SRM/MRM) of unique peptides is the most frequently employed MS approach[6]. Stable isotope labeling is the classic approach for relative quantitative proteomics. The labeled samples are mixed prior to MS analysis, so that relative quantitation can be achieved in one run, avoiding the instability in LC separation and minimizing error. Stable isotopes can be introduced metabolically (SILAC)[7], enzymatically (18O labeling)[8], and chemically (dimethyl labeling)[9], whereas relative quantitation is usually based on the chromatogram peak area ratio of precursor ions. However, MS1 based quantitation suffers from low labeling efficiency, narrow dynamic range, and poor sensitivity, and thus isobaric labeling based MS2 quantitation is gaining more and more attention[10]. Using isobaric labeling, the same peptide originated from different samples displays identical m/z at MS1 level, and thereby improves method sensitivity significantly. Meanwhile the reporter ions with different m/z are generated at MS2 level, providing quantitation information as well as b/y ion identification. In addition, a wider dynamic range can be attained with this approach. iTRAQ[11] and TMT[12] are two major isobaric
Received 10 September 2014; accepted 18 October 2014 * Corresponding author. Email: [email protected] Copyright © 2014, Changchun Institute of Applied Chemistry, Chinese Academy of Sciences. Published by Elsevier Limited. All rights reserved. DOI: 10.1016/S1872-2040(14)60788-X
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
labeling reagents, with labeling capacity of 8-plex and 6-plex, respectively. Despite these improvements, isobaric labeling methods face interference issues caused by co-eluting peptides. Specifically, on the LC side, the matrices of proteomic samples are usually very complex, with a significant number of peptides co-eluted from the column; whereas on the MS side, an isolation window of 2 m/z is typically used when filtering precursor ions for MS2 analysis. In combination, these technical pitfalls result in the lack of discrimination between target peptide and co-eluting near-m/z peptides, leading to erroneous incorporation of the reporter ions of co-eluting peptides, compromising quantitative accuracy[13,14]. Ting et al[15] has shown that the co-eluting peptides in complex samples can dramatically affect reporter ion intensity, distort peptide and protein ratios, leading to underestimation of proteome changes. Such problems have become the bottleneck in isobaric labeling method development. Triple quadrupole-based SRM (or MRM), the gold standard for MS quantitation, is widely used for absolute quantitation of proteins[6]. SRM monitors the precursor and product ion mass of specific peptides. The first quadrupole (Q1) filters for the precursor ions, and Q3 selects for product ions after precursor fragmentation in Q2. By monitoring the signal of unique precursor-product transition, interfering ions can be mostly excluded. Peptide confirmation and quantitation can be achieved either with external standard using calibration curves, or by spiking with known amount of stable isotope-labeled peptide internal standards[6,16]. There is no doubt that SRM method offers high sensitivity and wide dynamic range, and is indispensible for target protein verification and quantitation. However, accompanying the progress of quantitative proteomics, SRM is challenged with more complex sample matrices, lower abundance of target proteins and ease of suppression by high abundance components. In addition, the low resolution SRM fails to effectively exclude interferences from complex matrices, and is prone to false positive results[17,18]. On the other hand, the demand on analysis throughput also increases rapidly, with the need to monitor tens of thousands of transitions in one analytical run, while SRM method can only manage a limited number of transitions simultaneously due to its speed and sensitivity limit[19]. Moreover, optimization of SRM transition and collision energy can be overwhelming for proteomic research, especially for those biomarker and systems biology studies requiring large sample sizes[20,21]. Therefore, absolute proteome quantitation also faces intimidating technical challenges. Recently, technical improvement of high-resolution mass spectrometers (such as Orbitrap), and innovations in data acquisition strategies, including isobaric labeling-based synchronous precursor selection (SPS) and mass defect isobaric labeling, as well as high-resolution parallel reaction monitoring (PRM) and multiplexing acquisition (MSX) for
SRM type analysis, and various novel data-independent acquisition (DIA) strategies, have brought new opportunities to solve these challenges in quantitative proteomics. 2
Progress in isobaric labeling technique: synchronous precursor selection (SPS) and mass defect isobaric labeling
As mentioned earlier, isobaric labeling techniques represented by iTRAQ and TMT face the challenges on two fronts: (1) massive co-eluting near-m/z peptides co-fragment with target peptide, which affect the measurement of reporter ion intensity, and distort quantitative results[13‒15]; (2) isobaric tags have limited labeling capacity. Taking iTRAQ as an example, its reporter group (N-methylpiperazine) consists 6 C and 2 N, with limiting labeling capacity to 8-plex, thus unsuitable for higher sample throughput[22]. Although larger reporter group supports higher labeling capacity, it hurts sensitivity[23]. To solve the problem caused by co-eluting peptides, Ting et al[15] analyzed TMT-labeled samples using MS3 scan in LTQ-Orbitrap (Fig.1). They first employed relatively low collision energy for MS2 CID (35%) fragmentation to generate b/y ions for sequence identification without over-fragmenting the TMT tags; then within the 110%–160% m/z range of parent ion, selected the product ion with highest intensity to undergo thorough reporter group fragmentation by high energy HCD (60%) and to determine reporter ion ratios with MS3 detection. They demonstrated that, through selection on both precursor ion and product ion level, elimination of co-eluting interference can be achieved, yielding a quantitative ratio matching the theoretical one. For example, in a model two-proteome peptide mixture sample from Lys-C protein digests of yeast (mimicking sample) and the human HeLa cell line (functioned as interference), when the theoretical ratio was 10:1, classic MS2 analysis only got about 3, while MS3 reported a ratio of 11.7, which was very close to the theoretical ratio. However, signal intensity drops significantly in MS3 results compared to MS2, leading to a much lower sensitivity of MS3 in quantitation despite its high quantitation accuracy, thus limiting its application in real sample analysis[24]. Although the recently developed TMTc[25] and QuantMode [26] methods are effective in dealing with co-eluting peptides, these methods involve tedious procedures without offering a quantitative accuracy comparable to that of MS3. Synchronous precursor selection (SPS) technique completely solved the signal intensity drawback of MS3. SPS uses MultiNotch technique which allows for synchronous isolation of multiple ions (≤ 15, Fig.2A) in a linear trap[27]. With this technique, SPS method can select multiple b/y fragments from the target peptide in MS2 for MS3 experiment, resulting in accumulation of reporter ions and a substantial improvement
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
Fig.1 Schematic elucidation of MS3 method[15]. (A) Acquisition workflow of MS3, (B) Comparison of MS3 and classic MS2 results
Fig.2 Schematic elucidation of SPS MS3 method[24,28] 3
(A) Acquisition workflow of SPS MS ; (B) Influence of classic MS2, MS3 and SPS MS3 for reporter ion intensity; (C) Comparison of sensitivity of MS3 and SPS MS3; (D) Comparison of accuracy of classic MS2 and SPS MS3
of MS3 signal (Fig.2B)[24,28]. Experimental results have demonstrated that the spectra generated using SPS method for MS3 quantitation (CID fragmentation in MS2, then select multiple fragment ions for MS3 HCD fragmentation and Orbitrap detection) are comparable to MS2 spectra, and the success rate of quantitation is over 70% which is similar to that of classic MS2 (Fig.2C)[24,28]. Moreover, compared to the reporter ion from the single product ion, the accumulation of reporter ion from multiple product ions in MS3 gives rise to more stable ratios, and consequently better reproducibility and accuracy (Fig.2D). Besides, SPS MS3 quantitation is compatible with trypsin digested samples which are more common than Lys-C digestion required in traditional MS3, empowering SPS MS3 with wider applicability in routine
quantitative proteomics[29]. Weekes et al[30] investigated the process of CMV infection of fibroblasts and underlying mechanism using SPS MS3 method. They obtained precise quantitative data on over 8000 proteins (including 1184 cell surface proteins) at 8 time points during infection, conducted real-time dynamic quantitative analysis, and established the field of quantitative temporal viromics. As an attempt to improve the throughput limit of isobaric labeling tags, Dephoure et al[31] combined the triplex metabolic labeling SILAC and 6-plex isobaric tag TMT to quantitate 18 samples in parallel and obtained satisfactory results. But the expansion of labeling capacity was not achieved fundamentally until the introduction of mass defect isobaric labeling in recent years. More specifically, subtle
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
mass defect differences exist between 13C/12C and 15N/14N due to nuclear binding energy, and mass defect isobaric labeling utilizes the 6.32 mDa difference between 13C14N and 12C15N to realize higher labeling capacity by replacing 13C with 15N (Fig.3A). For example, by replacing one 13C atom with 15N in TMT127 and TMT129 reporter group of TMT 6-plex, one will get TMT127L (12C8H1615N1+, 127.1247608 Da)/TMT127H (12C713C1H1614N1+, 127.1310808 Da) and TMT129L (12C613C2H1615N1+, 129.1314705 Da)/TMT129H (12C513C3H1614N1+, 129.1377904 Da), raising labeling capacity to 8-plex[32]. Similarly, TMT 10-plex and 18-plex can also be accomplished without altering reporter group structure, thus providing an effective solution to the multiplexing capacity issue of parallel samples in quantitative proteomics[33,34]. On the other hand, in the range of m/z 100–200, as to the 6.32 mDa difference, a resolution of over 50000 is required to achieve baseline separation (Fig.3B), which is beyond the capacity of common high resolution mass spectrometers[33]. But for the fourier-transformation-based ultra-high resolution Orbitrap and FT-ICR whose resolution is inversely proportional to m/z, discrimination of mass defect labels can be easily realized[34]. Currently, mass defect labeling based-TMT 10-plex has been commercialized, and widely applied in quantitative proteomics studies demanding large sample size, such as those in systems biology and clinical biomarker discovery (Fig.3C). As elaborated above, synchronous precursor selection (SPS), mass defect isobaric labeling and other recently developed techniques have solved many issues in traditional isobaric labeling methodology, and proven to be powerful tools in relative quantitative proteomics.
3
Progress in target protein quantitation: parallel reaction monitoring (PRM) and multiplexing acquisition (MSX)
SRM filters and monitors the precursor and product ions using low-resolution triple quadrupoles. For small molecular compounds whose structures often differ significantly from each other, SRM is extraordinarily successful in differentiating target compound and matrix background[35]. However, owing to the close similarity of peptides, SRM method fails to completely eliminate matrix interferences in proteomics, undermining the accuracy and specificity of quantitation[17,18], not to mention that selection and optimization of SRM transitions can be very timeconsuming[36]. SRM method also benefits from the technical advances of mass spectrometry. For example, high resolution SRM (H-SRM) shrinks the precursor isolation window from 0.7 m/z down to 0.2 m/z and thereby reduces the matrix interferences[37,38]. However, selection efficiency also decreases with isolation window, rendering H-SRM more suitable for small molecule analysis. Another option for increasing selectivity is to incorporate an additional stage of mass filtering to form MRM3 scanning[39]. But unfortunately, MRM3 faces the same problems-decreased sensitivity and increased cycle time. Development of hybrid quadrupole-high resolution mass spectrometers (e.g., Q-Orbitrap) and enhancement of scan rate provides brand new avenues for quantitative proteomics. Parallel reaction monitoring (PRM) method running on
Fig.3 Increasing labeling capacity by mass defect isotopes[33] (A) Principle of mass defect labeling (TMT-10); (B) Minimum resolution for discrimination of 6 mDa mass difference. (C) Mass spectrum of mass defect reporter ions (TMT-10)
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
platforms such as Q-Orbitrap is just the high resolution ion monitoring technique in contrast to the classic SRM[40,41]. Different from SRM, PRM scans all product ions of the precursor, i.e. it monitors every transition of the precursor ion in parallel. PRM first selects precursor ion in the quadrupole (Q1) with a window ≤ 2 m/z, then fragments the precursor in collision cell (Q2) to generate product ions, and finally uses Orbitrap in place of Q3 to conduct high resolution-accurate mass (HR/AM) full scan for all product ions and complete the PRM acquisition (Fig.4)[42]. PRM displays several advantages over classic SRM: (1) High resolution product ion monitoring. With a ppm level mass accuracy (usually < 3 ppm when using external standards, and < 1 ppm using internal standards), the interferences can be effectively eliminated without losing sensitivity; (2) Full scan at MS2. The optimization of transition and collision energy is avoided, and the chromatogram peak of target compound can be extracted simply by selecting the best product ion (s) during data processing stage. Moreover, linear quantitative dynamic range of PRM spans 5‒6 orders of magnitude; (3) Simultaneous identification and quantitation. The MS2 full scan spectrum supports peptide identification, while product ion (s) with the best signal can be used as transition (s) for quantitation. Peterson et al[40] compared the selectivity and sensitivity of SRM and PRM using 14 peptide standards and found that both methods could reach a limit of quantitation (LOQ) around amol level in neat sample. However, in yeast matrix, the LOQ of SRM method raised to fmol level, one order of magnitude higher than the amol LOQ of PRM method (Fig.5A), demonstrating the vulnerability of SRM method to matrix interferences (Fig.5B). In addition, Gallien et al[42] examined the LOQ of 175 transitions from 35 heavy isotope-labeled peptides in urine matrix. Their results showed that SRM achieved better LOQ for only 19% of the transitions. Owing to its advantages, PRM has found its way into life sciences research in recent years. Tsuchiya et al[43] employed PRM monitoring and quantitation to compare the ubiquitylation
level of ubiquitin-proline-β-galactosidase (Ub-P-bgal) in wildtype and mutant type with UFD4 (ubiquitin fusion degradation pathway gene) knockout. By investigating the change of linkage sites between ubiquitin chain and protein in these two types of cell, the authors demonstrated that UFD4 is responsible for the K29-linked ubiquitin chains. In another study, Tang et al[44] applied PRM to verify 55 sites for methylation, acetylation, propionylation and other modifications of histone H3 and H4. They also quantitated the modification levels, and identified several new modification sites, such as H3 K18 methylation sites, and H4 K20 acetylation sites which had not been discovered before. However, PRM falls short on analytical throughput compared to SRM. Since the typical effective scan rate of high-resolution mass spectrometer is only 10–15 Hz, far below that of SRM which can easily handle hundreds of transitions, while PRM can simultaneously monitor 10‒15 precursor ions at best. Fortunately, the solution to this problem is emerging. Multiplexing (MSX) technique is one exemplary way to improve PRM throughput. MSX takes advantage of the C-trap in front of Orbitrap, and different precursor ions can be sequentially selected by quadrupole and stored in C-trap, ready for Orbitrap to finish last scan. The mixture of precursor ions in C-trap is then injected into Orbitrap for analysis.
Fig.4
Principle of selected reaction monitoring (SRM) (A) and parallel reaction monitoring (PRM) (B)[40]
Fig.5 Comparison of sensitivity and selectivity of SRM/PRM[40] (A) Comparison of LOQs of peptides in neat sample and yeast matrix. (B) Comparison of extracted ion chromatograms of peptide GVSAFSTWEk in yeast matrix
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
In this fashion, simultaneous monitoring of as many as 10 precursor ions can be achieved, with a 10-times increase in method throughput[45]. Gallien et al[46] combined MSX and PRM to establish the MSX-PRM method. By quadruplexing acquisition and retention time segmenting (into 1.5–2.5 min windows), they quantitatively monitored 770 target peptides (corresponding to 436 proteins) in yeast digest under a 60-min LC gradient elution. In their study, as many as 60 co-monitoring peptides with the same retention time window were observed. The authors managed to keep a cycle time of less than 2 s, and the throughput matched that of SRM method. Throughput of target peptide can be further boosted by optimizing the parameters such as resolution and maximum ion injection time. For example, simultaneous monitoring of 128 peptides is possible when MSX is set to 8[42,45]. In summary, the advent of high-resolution PRM and MSX-PRM technology has solved many issues in classic SRM-based quantitative proteomics, and provides powerful tools for target protein verification and absolute quantitation.
4
Progress in data-independent acquisition: a variety of Orbitrap-based DIA techniques
Shotgun proteomics, the classic strategy and primary relatively quantitative approach in proteomics, is based on data-dependent acquisition (DDA), in which MS2 acquisition depends on the MS1 result, leading to loss of low abundance peptides, stochasticity in analysis, and un-even distribution of scan points. Targeted approaches such as SRM, on the other hand, fail to collect the information outside of the target list, let alone the low throughput limit. To solve these problems, data-independent acquisition (DIA), a data acquisition strategy combining the features of DDA and SRM, is emerging in recent years. DIA divides the entire scan range into several isolation windows, then sequentially isolates and fragments in each window, acquiring data on the product ions of all precursor ions in the window. DIA offers an array of advantages over DDA and SRM[47]. DIA does not require target peptide, has no throughput limit, and generates scan points that are evenly distributed. Sequence verification and selection of product ions for quantitation can be achieved with the aid of spectrum library, and the data can be reviewed post-acquisition retrospectively. Venable and colleagues[48] first demonstrated the use of DIA strategy in relative quantitation of 15N-labeled yeast proteome by sequentially isolating and fragmenting precursor windows of 10 m/z in a LTQ ion trap mass spectrometer. Their results showed that chromatogram extracted from mass spectra obtained through the DIA strategy offered significantly better S/N ratio, reproducibility, and quantitative accuracy than those through DDA strategy, and the number of quantitated proteins increased by 87%. Gillet et al[49]
developed a Q-TOF (TripleTOF)-based SWATH technique which utilized 32 of consecutive 25 m/z windows, and showed that the selectivity of SWATH is comparable to that of SRM. Development of SWATH expanded the use of DIA in quantitative proteomics[50–52]. Meanwhile, Q-Orbitrap (Q Exactive)-based DIA method, using the same 25 m/z step size, improved selectivity in quantitating complex samples owing to its higher resolution. Un-segmented DIA methods, such as MSE and AIF, also found their way in quantitative proteomics. However, due to fragmenting and monitoring all ions at the same time without dividing them into different m/z windows, these methods suffer from matrix interferences and are un-suitable for quantitative analysis of complex samples[49,53]. Limited scan rate is the Achilles heel of high-resolutionMS-based DIA methods. These methods usually require a wide isolation window of 25 m/z to ensure sufficient scan points for quantitation, which on the other hand sacrificed quantitation results by allowing for more interferences (Fig.6A)[54]. Recent development in DIA methods helps narrow DIA isolation window, further improving DIA selectivity, sensitivity, and reproducibility. The MSX technique described earlier has also been applied in DIA methods. Egertson et al[54] developed a MSX-DIA method using the MSX function of Q-Orbitrap, wherein five separate, random 4-m/z isolation windows were sequentially selected, accumulated, and injected into Orbitrap for analysis. In this fashion, the total step size of MSX-DIA was 20-m/z-wide, leaving scan rate unaffected, but the actual isolation window was only 4-m/z-wide (Fig.6A). The resulting data were deconvoluted using the software Skyline to organize each fragment into specific windows, realizing a selectivity matching that of 4-m/z isolation window (Fig.6B). MSX-DIA can offer selectivity comparable to that of DDA, minimizing interferences from co-eluting peptides and contaminants. The emergence of Q-OT-qIT tribrid mass spectrometer made it possible for DIA advancement[55]. Since the LTQ and Orbitrap of Q-OT-qIT MS work independently in parallel, Orbitrap can be dedicated to MS1 scan, while the faster LTQ with a scan rate of 20 Hz is devoted to MS2 DIA scan. The high resolution MS1 spectra are then used for chromatogram peak extraction and quantitation, whereas the MS2 spectra only support peptide identification and confirmation. This arrangement significantly improves scan rate, and allows for narrower isolation windows by relieving the scan point concern from MS2 DIA scans. WiSIM-DIA and Full MS-DIA are precisely the DIA methods adopting such workflow (Fig.7)[56,57]. WiSIM-DIA uses precursor ions from 200 m/z wide isolation window-SIM (WiSIM) scans for quantification and ion trap MS2 DIA scan spectra collected at 12 m/z step size in parallel for confirmation. The ultra-high resolution (240000) WiSIM scans significantly improves precursor
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
Fig.6 Schematic elucidation of DIA and MSX-DIA[54] (A) Comparison of DIA and MSX-DIA; (B) Deconvolution of MSX-DIA spectra for 4 m/z selectivity
Fig.7 Schematic elucidation of WiSIM-DIA and Full MS-DIA[57] (A) Workflow of WiSIM-DIA; (B) Workflow of Full MS-DIA
selectivity, while the accurate precursor mass based chromatogram extraction and quantitation offers better sensitivity than classic DIA product ion-based quantitation (Fig.7A). As for peptide confirmation, despite that WiSIM-DIA MS2 spectrum are acquired at lower resolution compared to classic DIA, the decrease of isolation window size from 25 m/z to 12 m/z significantly eliminated matrix interferences. Full MS-DIA pushes this strategy further, its MS2 acquisition is conducted at 3 m/z step size, with a few ultra-high resolution
(240k) Orbitrap MS full scans inserted to guarantee sufficient data points for precursor scans. Similar to WiSIM-DIA, Full MS-DIA also uses precursor chromatogram for quantitation and MS2 DIA scans for confirmation (Fig.7B). It is worth pointing out that the selectivity of Full MS-DIA conducted with 3 m/z isolation window is comparable to that of DDA, allowing Full MS-DIA results to be used as low resolution data (precursor ±1.5 Da) in database searching, waiving the need for a spectrum library, and realizing the integration of
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
DDA and DIA[57]. In support of DIA development, algorithms using DIA data directly for database searching as well as methods for result validation are also being developed, allowing for direct DIA-data-based database searching and peptide identification. It has been shown that better performance can be obtained using DIA methods than DDA in protein identification. Last but not least, the recently developed pSMART technique achieves WiSIM-DIA-type of acquisition on a Q-Orbitrap platform by inserting several high-resolution Orbitrap MS scans into MS2 scans collected with 5 m/z isolation windows[58], yielding better sensitivity, selectivity, and reproducibility than classic DIA in both qualitative and quantitative studies. DIA surpasses DDA shotgun method and SRM-based targeted monitoring in both relative and absolute protein quantitation, looking into a bright future in quantitative proteomics. However, DIA suffers from long cycle time currently. It is only compatible with nano-LC and requires slow gradient to obtain sufficient LC peak width. This drawback limits DIA application and puts forward a challenge for its advancement.
5
Conclusions
Isobaric labeling, target ion monitoring, and data-independent acquisition have dominated quantitative proteomics. Table 1 summarizes and compares the principle of these three approaches. MS technology is constantly facing challenges presented by proteomics as the field advances. Currently, stable isotope labeling-based relative quantitation and SRM-based absolute quantitation methods both suffer from the limitations such as matrix interferences and insufficient throughput (Table 1). To solve these problems, a variety of high-resolution mass spectrometer-based methods are developed: synchronous precursor selection (SPS) and mass defect isobaric labeling may solve the interference and throughput problems for relative quantitation; parallel reaction monitoring (PRM) and multiplexing acquisition (MSX) can improve SRM selectivity and have become the new pathway to absolute quantitation; while data-independent acquisition (DIA) strategies combine the advantages of DDA and SRM; multiplexing and tribrid mass spectrometers allow for smaller step size in DIA scans, empowering DIA to expand its application in high-throughput quantitative proteomics. It is not hard to foresee that these new technologies will be used more and more to solve the problems in quantitative proteomics in place of conventional MS techniques, such as elucidating protein interaction and discovering clinical biomarkers.
Table 1 Principle and progress of quantitative proteomic methods Reporter ion quantification Major techniques
iTRAQ, TMT
Method principle
Data-dependent acquisition of isobaric labeled samples, and comparison of reporter ion intensities in MS2
Application fields Technical limitations Technique progress
Large scale relative quantification and comparison Inaccurate ratios for complex samples, limited labeling capacity Synchronous precursor selection (SPS), Mass defect isotopic labeling
Target ion monitoring SRM/MRM Selection and acquisition of specific fragment ions isolated from target precursor ion, and extraction of transitions for quantification Target proteins validation and absolute quantification Severe interference from complicated matrix Parallel reaction monitoring (PRM), Multiplexed PRM (MSX-PRM)
References
376–386 [8]
[1]
Ong S E, Mann M. Nat. Chem. Biol., 2005, 1(5): 252–262
[2]
Veenstra T D. J. Chromatogr. B, 2007, 847(1): 3–11
[3]
Zhou Y, Shan Y C, Zhang L H, Zhang Y K. Chinese Journal of Chromatography, 2013, 31(6): 496–502
[4]
Bantscheff M, Schirle M, Sweetman G, Rick J, Kuster B. Anal. Bioanal. Chem., 2007, 389(4): 1017–1031
[5]
Zhu J L, Zhang K, He X W, Zhang Y K. Chinese J. Anal. Chem., 2010, 38(3): 434–441
[6] [7]
Lange V, Picotti P, Domon B, Aebersold R. Mol. Syst. Biol.,
Data-independent acquisition SWATH, DIA Untargeted acquisition of all fragment ions from all precursor ions within scan range, and extraction of transitions for quantification Large scale, high-throughput relative and absolute quantification Severe interference due to wide isolation window (> 20 m/z) Multiplexed DIA (MSX-DIA), WiSIM-DIA, Full MS-DIA, pSMART
Capelo J L, Carreira R J, Fernandes L, Lodeiro C, Santos H M, Simal-Gandara J. Talanta, 2010, 80(4): 1476–1486
[9]
Boersema P J, Raijmakers R, Lemeer S, Mohammed S, Heck AJ. Nat. Protoc., 2009, 4(4): 484–494
[10] Koehler C J, Strozynski M, Kozielski F, Treumann A, Thiede B. J. Proteome Res., 2009, 8(9): 4333–4341 [11] Thompson A, Schafer J, Kuhn K, Kienle S, Schwarz J, Schmidt G, Neumann T, Johnstone R, Mohammed A K, Hamon C. Anal. Chem., 2003, 75(8): 1895–1904 [12] Ross P L, Huang Y N, Marchese J N, Williamson B, Parker K,
2008, 4: 222
Hattan S, Khainovski N, Pillai S, Dey S, Daniels S,
Ong S E, Blagoev B, Kratchmarova I, Kristensen D B, Steen H,
Purkayastha S, Juhasz P, Martin S, Bartlet-Jones M, He F,
Pandey A, Mann M. Mol. Cell. Proteomics, 2002, 1(5):
Jacobson A, Pappin D J. Mol. Cell. Proteomics, 2004, 3(12):
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
1154–1169 [13] Karp N A, Huber W, Sadowski P G, Charles P D, Hester S V, Lilley K S. Mol. Cell. Proteomics, 2010, 9(9): 1885–1897 [14] Ow S Y, Salim M, Noirel J, Evans C, Rehman I, Wright P C. J. Proteome Res., 2009, 8(11): 5347–5355 [15] Ting L, Rad R, Gygi S P, Haas W. Nat. Methods, 2011, 8(11): 937–940 [16] Zhao Y, Ying W T, Qian X H. Chem. Life, 2008, 28(2): 210–213 [17] Sherman J, McKay M J, Ashman K, Molloy M P. Proteomics, 2009, 9(5): 1120–1123 [18] Abbatiello S E, Mani D R, Keshishian H, Carr S A. Clin. Chem., 2010, 56(2): 291–305
[35] Gallien S, Duriez E, Demeure K, Domon B. J. Proteomics, 2013, 9(81): 148–158 [36] Karlsson C, Malmstrom L, Aebersold R, Malmstrom J. Nat. Commun., 2012, 3: 1301 [37] Gallart-Ayala H, Moyano E, Galceran M T. J. Chromatogr A., 2008, 1208(1-2): 182–188 [38] Martínez-Villalba A, Moyano E, Martins C P, Galceran M T. Anal. Bioanal. Chem., 2010, 397(7): 2893–2901 [39] Fortin T, Salvador A, Charrier J P, Lenz C, Bettsworth F, Lacoux X, Choquet-Kastylevsky G, Lemoine J. Anal. Chem., 2009, 81(22): 9343–9352 [40] Peterson A C, Russell J D, Bailey D J, Westphall M S, Coon J J. Mol. Cell. Proteomics, 2012, 11(11): 1475–1488
[19] Kiyonami R, Schoen A, Prakash A, Peterman S, Zabrouskov V,
[41] Schiffmann C, Hansen R, Baumann S, Kublik A, Nielsen P H,
Picotti P, Aebersold R, Huhmer A, Domon B. Mol. Cell.
Adrian L, von Bergen M, Jehmlich N, Seifert J. Anal. Bioanal.
Proteomics, 2011, 10(2): M110.002931 [20] Cima I, Schiess R, Wild P, Kaelin M, Schüffler P, Lange V, Picotti P, Ossola R, Templeton A, Schubert O, Fuchs T, Leippold T, Wyler S, Zehetner J, Jochum W, Buhmann J, Cerny T, Moch H, Gillessen S, Aebersold R, Krek W. Proc. Natl. Acad. Sci. USA, 2011, 108(8): 3342–3347 [21] Picotti P, Bodenmiller B, Mueller LN, Domon B, Aebersold R. Cell, 2009, 138(4): 795–806 [22] Pichler P, Kocher T, Holzmann J, Mazanek M, Taus T, Ammerer G, Mechtler K. Anal. Chem., 2010, 82(15): 6549–6558 [23] Thingholm T E, Palmisano G, Kjeldsen F, Larsen M R. J. Proteome Res., 2010, 9(8): 4045–4052 [24] McAlister G C, Nusinow D P, Jedrychowski M P, Wühr M, Huttlin E L, Erickson B K, Rad R, Haas W, Gygi S P. Anal. Chem., 2014, 86(14): 7150–7158 [25] Wuhr M, Haas W, McAlister G C, Peshkin L, Rad R, Kirschner M W, Gygi S P. Anal. Chem., 2012, 84(21): 9214–9221
Chem., 2014, 406(1): 283–291 [42] Gallien S, Duriez E, Demeure K, Domon B. J. Proteomics, 2013, 81: 148–158 [43] Tsuchiya H, Tanaka K, Saeki Y. Biochem. Biophys. Res. Commun., 2013, 436(2): 223–229 [44] Tang H, Fang H, Yin E, Brasier A R, Sowers L C, Zhang K. Anal. Chem., 2014, 86(11): 5526–5534 [45] Gallien S, Bourmaud A, Kim S Y, Domon B. J. Proteomics, 2014, 100: 147–159 [46] Gallien S, Duriez E, Crone C, Kellmann M, Moehring T, Domon B. Mol. Cell. Proteomics, 2012, 11(12): 1709–1723 [47] Law K P, Lim Y P. Expert Rev. Proteomics, 2013, 10(6): 551–566 [48] Venable J D, Dong M Q, Wohlschlegel J, Dillin A, Yates J R. Nat. Methods, 2004, 1(1): 39–45 [49] Gillet L C, Navarro P, Tate S, Rost H, Selevsek N, Reiter L, Bonner R, Aebersold R. Mol. Cell. Proteomics, 2012, 11(6): O111.016717
[26] Wenger C D, Lee M V, Hebert A S, McAlister G C, Phanstiel D
[50] Liu Y, Huttenhain R, Surinova S, Gillet L C, Mouritsen J,
H, Westphall M S, Coon J J. Nat. Methods, 2011, 8(11):
Brunner R, Navarro P, Aebersold R. Proteomics, 2013, 13(8):
933–935 [27] Goeringer D E, Asano K G, McLuckey S A. Anal. Chem., 1994, 66(3): 313–318 [28] Viner R, Bomgarden R, Blank M, Rogers J. 61st ASMS, 2013, Poster W617 [29] Blank M, Bomgarden R, Rogers J, Jacobs R, Fong J, Puri N, Zabrouskov V, Viner R. 61st ASMS, 2013, Poster Th449 [30] Weekes M P, Tomasec P, Huttlin EL, Fielding C A, Nusinow D, Stanton R J, Wang E C, Aicheler R, Murrell I, Wilkinson G W, Lehner P J, Gygi S P. Cell, 2014, 157(6): 1460–1472 [31] Dephoure N, Gygi S P. Sci. Signal, 2012, 5(217): rs2 [32] Werner T, Becher I, Sweetman G, Doce C, Savitski M M, Bantscheff M. Anal. Chem., 2012, 84(16): 7188–7194 [33] McAlister G C, Huttlin E L, Haas W, Ting L, Jedrychowski M P, Rogers J C, Kuhn K, Pike I, Grothe R A, Blethrow J D, Gygi SP. Anal. Chem., 2012, 84(17): 7469–7478 [34] Werner T, Sweetman G, Savitski M F, Mathieson T, Bantscheff M, Savitski M M. Anal. Chem., 2014, 86(7): 3594–3601
1247–1256 [51] Collins B C, Gillet L C, Rosenberger G, Rost H L, Vichalkovski A, Gstaiger M, Aebersold R. Nat. Methods, 2013, 10(12): 1246–1253 [52] Lambert J P, Ivosev G, Couzens A L, Larsen B, Taipale M, Lin Z Y, Zhong Q, Lindquist S, Vidal M, Aebersold R, Pawson T, Bonner R, Tate S, Gingras A C. Nat. Methods, 2013, 10(12): 1239–1245 [53] Chapman J D, Goodlett D R, Masselon C D. Mass Spectrom. Rev., 2013: 10.1002/mas.21400 [54] Egertson J D, Kuehn A, Merrihew G E, Bateman N W, MacLean B X, Ting Y S, Canterbury J D, Marsh D M, Kellmann M, Zabrouskov V, Wu C C, MacCoss M J. Nat. Methods, 2013, 10(8): 744–746 [55] Senko M W, Remes P M, Canterbury J D, Mathur R, Song Q, Eliuk S M, Mullen C, Earley L, Hardman M, Blethrow J D, Bui H, Specht A, Lange O, Denisov E, Makarov A, Horning S, Zabrouskov V. Anal. Chem., 2013, 85(24): 11710–11714
ZHANG Wei et al. / Chinese Journal of Analytical Chemistry, 2014, 42(12): 1859–1868
[56] Kiyonami R, Patel B, Senko M, Zabrouskov V, Egertson J, Ting S, MacCoss M, Rogers J, Huhmer A. Large Scale Targeted
[57] Zhang W, Reiko K, Jiang Z, Chen W. Chinese J. Anal. Chem., 2014, 42(12): 1750–1758
Protein Quantification Using WiSIM-DIA Workflow on a
[58] Prakash A, Peterman S, Ahmad S, Sarracino D, Frewen B,
Orbitrap Fusion Tribrid Mass Spectrometer. ASMS, 2014,
Vogelsang M, Byram G, Krastins B, Vadali G, Lopez M. J.
W737
Proteome Res., 2014, DOI:10.1021/PR5003017