Over the past decade, the field of proteomics has progressed from the identification of proteins in biological samples to protein quantification.1,2 Several strategies have been developed, one of which is stable isotope labeling, where samples are labeled and the relative abundances of proteins are calculated across two or more samples. There are two major types of isotope labeling techniques: 1) metabolic labeling—SILAC (stable isotope labeling with amino acids in cell culture), and 2) chemical labeling—tandem mass tags (TMT®, Thermo Fisher Scientific [Waltham, MA]) and isobaric tags for relative and absolute quantitation (iTRAQ®, AB SCIEX [Framingham, MA]). These two techniques differ in where the labeling occurs. In metabolic labeling, the labeling takes place at the metabolic level, whereas chemical labeling is done after the samples have been enzymatically digested, typically with trypsin. Chemical labeling has become more popular since it is easier to perform and up to eight samples can be multiplexed. Each sample is labeled with a different label and the samples are mixed and subjected to liquid chromatography-mass spectrometry/mass spectrometry (LC-MS/MS) analysis.
As with any quantitative method, high precision and accuracy are very important and can be challenging to achieve. It has been demonstrated that the lack of quantitative precision is common, especially for low-abundance peptide signals.3–5 There are two main reasons for this: 1) the presence of low-molecular-weight/ charge ratio ions that interfere with reporter ion signals, and 2) co-isolation of interferences during the isolation of precursor ions, leading to MS2-generated reporter ions derived from several precursor ions.6
Isotope labeling techniques
Two chemical isobaric labeling tags are commercially available: isobaric tags for relative and absolute quantification (iTRAQ)7 from AB SCIEX and (TMT)8,9 from Thermo Fisher Scientific. Both tagging systems work in a similar fashion. The tags are composed of three parts: 1) a reporter group, 2) a balance group, and 3) a primary amine reactive group. The reactive group reacts with the N-terminus and the lysine ε amino groups of a peptide and thereby tags the peptide.
Upon MS/MS fragmentation, each reporter group produces a unique ion. For the iTRAQ labels, there are two types—a four-plex system and an eight-plex system. The reporter ions for the four-plex are at m/z of 113, 114, 115, and 116. The reporter ions for the eight-plex are at m/z of 113, 114, 115, 116, 117, 118, 119, and 121. The reporter ions for the TMT system are at m/z of 126, 127, 128, 129, 130, and 131. The balance group provides for the necessary molecular weight adjustment so that each tag is nominally the same molecular weight.
Figure 1a shows a flow diagram of how the isobaric labeling experiment works. First, proteins are enzymatically digested to produce peptides. The peptides are then labeled (Figure 1a shows TMT tags as an example). The labeled peptides are then subjected to LC-MS/MS analysis. The MS experiment is generally done as a data-dependent scanning experiment where precursor ions are selected from a full MS scan for further fragmentation. Since all the tags have the same mass, the peptides are indistinguishable in the full mass spectra. The precursor ion is then isolated for fragmentation, which produces fragment ions that can be used for both peptide sequencing and quantification.
Figure 1 – Isobaric labeling, interference, and interference modeling. a) Quantitative mass spectrometry-based proteomics experiments with stable isotope-containing isobaric tags. b) Isolation window showing target peptide ion and contaminating peptide (interference effect). c) Two-proteome model using TMT six-plex labeled human cell line and yeast (
S. cerevisiae) digests. d) An ideal yeast peptide MS2 spectrum without human peptide interference. e) A typical yeast peptide with some interference. (Figures 1, 3, and 4 reprinted by permission from Macmillan Publishers Ltd.:
Nature Methods, reference 10, 2011.)
For quantification to work perfectly, an assumption is made that the precursor ions isolated for fragmentation all belong to the same peptide. In actuality, in complex mixtures, this is not the case. Ions from other peptides can be isolated with the peptide of interest, thereby contaminating the reporter ion signals (Figure 1b). This effect can even be seen in spectra that have been identified as nonlabeled peptides, but contain reporter ion abundances in the spectrum (Figure 2). The effect of adding in signal intensities from interfering peptides is to distort the reporter ion intensities of the peptide of interest. Intensity differences between reporter ions when expressed as ratios tend toward unity owing to interfering peptide ions derived from proteins with unchanging expression.10
Figure 2 – Example of peptide identified as not being labeled. a) Full MS2 spectrum. b) Reporter ion region showing labeled peptide interference.
Studies showing the interferences have been reported by spiking protein standards into background peptide mixtures.3–5 However, one laboratory has reported a unique experiment in which the interference problem was addressed by mixing together two proteome peptide mixtures.10 Six-plex TMT reagents were used to build a model system using a Lys-C protein digest of yeast (Saccharomyces cerevisiae) cells and the human HeLa cell line (Figure 1c). Use of the entire human HeLa cell proteome as the interfering peptide source creates a worst-case scenario for interfering peptides. Not only does this give a clear demonstration of the problem, but also provides a way to evaluate strategies for solving the problem.
Two sets of three different quantities of the yeast digest, in a ratio of 10:4:1, were labeled with the six TMT labels (see Figure 1c). Three samples from the human cell digest were labeled with three of the TMT labels: 126, 127, and 128. When all of the samples were mixed together, the human cell samples interfered with the yeast samples labeled with 126, 127, and 128, but not with samples labeled with the 129, 130, and 131 tags (see Figure 1e). The yeast samples labeled with the 129, 130, and 131 labels become the control set for the experiment. Figure 1d shows the ideal reporter ion relative abundances if there are no interferences.
The sample was worked up using standard procedures for analyzing a complex proteome sample by first fractionating the sample using strong cation exchange (SCX) chromatography11 and analyzing each fraction by LC-MS/MS experiments performed on an LTQ Orbitrap Velos mass spectrometer (Thermo Fisher Scientific), where the MS/MS experiments were performed using higher-energy collision dissociation (HCD)12 to generate high-mass-accuracy MS/MS spectra. This experiment demonstrated that the human proteins mixed into the sample interfered with the quantification accuracy of the yeast protein abundances.
Figure 3a shows an example of a yeast peptide with strong distortion of the reporter ion intensities and an example of a human peptide that has some contamination from yeast peptides. The example of the yeast peptide shows intensity ratios of 1.6:1.2:1 for the 126, 127, and 128 ions. As stated earlier, if no interferences were seen, then the ratios should be 10:4:1, whereas the 131, 130, and 129 ion intensity ratios are 9.6:3.7:1, which is close to the expected ratios. The human peptide displayed a 1:1:1 intensity ratio, as expected; however, there are intensities at the ions at 129, 130, and 131 that must be from yeast peptides since the human sample was not labeled with these tags. In addition, the intensity ratios are very close to what is expected for the yeast samples.
Figure 3 – Evaluation of MS2 method. a) Example of strong distortion of reporter ion intensities for a yeast peptide (left), and human peptide (right). b) Ratio distribution (log2) of yeast peptides in channels with human peptide interference, without human peptide interference and human peptides only. c) Averaged normalized relative intensities for each TMT reporter ion channel for yeast peptides (left) and human peptides (right) from the data set described in (b).
The averaged normalized relative intensities for each TMT reporter ion channel for the yeast peptides and human peptides, shown in Figure 3c, demonstrate results similar to the example peptide. The relative ratios for the yeast peptides with the human peptides interference are 3.6 and 1.8 compared to what they should be—10 and 4, respectively. Figure 3b shows the ratio distributions (log2 scale) of yeast peptides in channels with and without human peptide interferences, as well as human peptides only. In all three of the possible ratios, i.e., 126:127 (2.5:1), 127:128 (4:1), and 126:128 (10:1), the distribution for the yeast peptides with no interference is centered around where it should be, while the distributions for the yeast peptides with interference are shifted toward one. In addition, the standard deviation for the peptide interference distribution increased by 60%10 over the standard distribution for the peptides with no interference. The conclusion that can be drawn from this experiment is that there is a loss of both quantification accuracy as well as a loss in precision due to interfering peptides.10
Two-proteome solution for quantitative proteomics
Using this model system, an MS3 experiment was evaluated for the potential to remove the peptide interference and improve quantitative accuracy and precision. In an MS3 experiment, first the peptide ion is isolated and fragmented, and then a selected fragment ion is again isolated and then fragmented to produce an MS3 spectrum.13 The idea is to select and isolate a fragment ion that is free from any interfering ion and use the reporter ion fragments in the MS3 spectrum to produce the quantitative data.
The instrument used in this experiment was an LTQ Orbitrap Velos mass spectrometer, which is a hybrid mass spectrometer containing a linear ion trap capable of performing collision-induced dissociation (CID)–MS/MS experiments. First, a full mass spectrum Orbitrap scan was performed and used to select a peptide ion for CID in the linear ion trap (Figure 4a), which was used for peptide sequencing and identification. The most intense fragment ion from the MS2 spectrum was then selected for HCD-MS3, yielding the quantitative data. To avoid the selection of neutral loss products and other ions, a selection of fragment ions was restricted to a range of 110–160% relative to the precursor ion m/z.10
Figure 4 – An MS3-based method eliminates the interference effect. a) Precursor ions for MS2 spectra were selected from a high-resolution full mass spectrum acquired on the Orbitrap. Fast LTQ-CID-MS2 experiments were used for peptide sequence assignments, followed by the selection of the most intense fragment ion from MS2 (green) for HCD-MS3, where TMT reporter ion intensities are measured in the Orbitrap. b) Normalized intensities for each TMT reporter ion channel for yeast peptides (left) and human peptides (right). c) Comparison of yeast peptide ratios for average, no interference; MS2, with interference; and MS3, with interference at the indicated ratios. d) Ratio distribution (log2) of yeast peptides in channels with human peptide interference, without human peptide interference, and human peptides only.
This MS3 method was used to analyze the same two-proteome sample previously described. The results showed that the ratios for the yeast peptides with the human peptide interferences were almost the same as the ratios for the yeast peptides with no interference (Figure 4b). Median ratios for yeast peptides with and without human peptide interference, respectively, were 10.5 and 11.7 (expected ratio of 10), and 4.4 and 4.7 (expected ratio of 4)10 (Figure 4c and d). Because each peptide quantified requires two scans to be performed, and the reduced signal intensities of the MS3 spectra, the number of proteins quantified was expected to be reduced. The number of quantified proteins in the MS3 method was 12% lower than the MS2 method.10
In summary, the two-proteome model experiment was able to demonstrate the problem of interfering peptides on quantitative accuracy and precision of isobaric labeling strategies for quantitative proteomics, and also provided a way to evaluate methods to eliminate this type of problem. The MS3 method described here is a very promising technique to eliminate the problem of interference.
- Bantscheff, M.; Schirle, M. et al. Anal. Bioanal. Chem.2007, 389, 1017–31.
- Ong, Se; Mann, M. Nat. Chem. Biol.2005, 1, 252–62.
- Bantscheff, M.; Boesche, M. et al. Mol. Cell Proteomics2008, 7, 1702–13.
- Karp, N.A.; Huber, W. et al. Mol. Cell Proteomics2010, 9, 1885–97.
- Ow, S.Y.; Salim, M. et al. J. Proteome Res.2009, 8, 5347–55.
- Viner, R.; Scigelova, M. et al. Relative quantitation of TMT-labeled proteomics—focus on sensitivity and precision; Application Note #566, Thermo Fisher Scientific.
- Ross, P.; Huang, Y. et al. Mol. Cell Proteomics2004, 3, 1154–69.
- Thompson, A.; Schafer, J., et al. Anal. Chem.2003, 75, 1895–1904.
- Dayon, L.; Hainard, A. et al. Anal. Chem.2008, 80, 2921–31.
- Ting, L.; Rad, R. et al. Nat. Methods20118, 937–40.
- Villén, J.; Gygi, S. Nat. Protoc.2008, 3, 1630–8.
- Olsen, J.; Macek, B. et al. Nat. Methods2007, 4, 709–12.
- Beausoleil, S.; Jedrychowski, M. et al. Proc. Natl. Acad. Sci. USA2004, 101, 12,130–5.
Emanuel M. Schreiber, Ph.D., is Research Specialist, Genomics and Proteomics Core Laboratories, University of Pittsburgh, Biomedical Science Tower 3, 3501 Fifth Ave., Pittsburgh, PA 15260, U.S.A.; tel.: 412-624- 6862; fax: 412-624-1064; e-mail: firstname.lastname@example.org.