Anal. Chem.

2003, 75, 1895-1904

Tandem Mass Tags: A Novel Quantification

Strategy for Comparative Analysis of Complex
Protein Mixtures by MS/MS
Andrew Thompson,† Ju1 rgen Scha
1 fer,‡ Karsten Kuhn,‡ Stefan Kienle,‡ Josef Schwarz,‡
1 nter Schmidt, Thomas Neumann,‡ and Christian Hamon*,‡
Gu †

Proteome Sciences§, Coveham House, Downside Bridge Road, Cobham, Surrey, KT11 3EP, U.K., and
Xzillion GmbH, Industriepark Höchst, Building G865a, 65929 Frankfurt am Main, Germany

A novel MS/MS-based analysis strategy using isotopomer gels. Attempts, with the 2D-DIGE (difference gel electrophoresis)
labels, referred to as “tandem mass tags” (TMTs), for the method, for example, to overcome this deficiency by running
accurate quantification of peptides and proteins is de- multiple fluorescently labeled samples on the same gel6 have
scribed. The new tags are designed to ensure that identical improved this, but at the cost of increasing the complexity of the
peptides labeled with different TMTs exactly comigrate in gels; the same protein labeled with different fluorescent dyes does
all separations. The tags require novel methods of quan- not comigrate with itself and multiple labeling products are
tification analysis using tandem mass spectrometry. The produced for each protein. In addition, 2D-PAGE lacks functional
new tags and analysis methods allow peptides from range, because it poorly represents specific classes of proteins,
different samples to be identified by their relative abun- such as very large or very small proteins, extremely acidic or basic
dance with greater ease and accuracy than other methods. proteins, and hydrophobic proteins.7-10 Despite the difficulties
The new TMTs permit simultaneous determination of both inherent in 2D-PAGE, this technique is being used for the analysis
the identity and relative abundances of peptide pairs using of whole proteomes,11 but a method for quantitative global protein
a collision induced dissociation (CID)-based analysis expression analysis that avoids the shortcomings of 2D-PAGE is
method. Relative abundance measurements made in the greatly sought after.
MS/MS mode using the new tags are accurate and Dispensing with 2D-PAGE has, fortunately, become possible
sensitive. Compared to MS-mode measurements, a very with the advent of advanced instrumentation for in-line liquid
high signal-to-noise ratio is achieved with MS/MS based chromatography electrospray ionization mass spectrometry. How-
detection. The new tags should be applicable to a wide ever, to date, MS- and MS/MS-based protein identification has
variety of peptide isolation methods. generally been restricted to smaller peptides, and as a conse-
quence, novel protein analysis techniques compatible with the
The mainstay of protein expression analysis is two-dimensional limitations of currently available instrumentation are required to
polyacrylamide gel electrophoresis (2D-PAGE) for the separation allow proteins in complex mixtures to be identified. One approach
of complex protein mixtures, followed by identification of those is to sample peptides from the proteins in a complex mixtures in
separated components by mass spectrometry using (MS)-based such a way that the peptide sample accurately represents the
peptide mass fingerprinting techniques1-3 or tandem mass spec- constituents of the original mixture. A number of such peptide
trometry (MS/MS)-based peptide sequencing techniques.4 While sampling techniques have been published recently, all based on
it is an effective tool, 2D-PAGE/MS is laborious and difficult to selectively reacting specific functional groups in proteins, such
automate. More significantly, experiments show poor reproduc- as cysteine residues in the isotope coded affinity tags (ICAT)
ibility and limited dynamic range,5 and the proteins, after elec- procedure12 and phosphate groups.13-15 The ICAT procedure, in
trophoretic separation, are not immediately compatible with mass particular, isolates a small number of peptides from each protein,
spectrometry. As a result of these limitations, it is difficult to thus representing almost all proteins with at least one peptide.
perform quantitative comparisons between samples on different
(6) Unlu, M.; Morgan, M. E.; Minden, J. S. Electrophoresis 1997, 18, 2071-
2077.
* Corresponding author. Fax: 0049 69 30544302. E-mail: Christian.Hamon@ (7) Garrels, J. I.; McLaughlin, C. S.; Warner, J. R.; Futcher, B.; Latter, G. I.;
xzillion.com. Kobayashi, R.; Schwender, B.; Volpe, T.; Anderson, D. S.; Mesquita-Fuentes,
†
Proteome Sciences. R.; Payne, W. E. Electrophoresis 1997, 18, 1347-1360.
‡
Xzillion GmbH. (8) Gygi, S. P.; Corthals, G. L.; Zhang, Y.; Rochon, Y.; Aebersold, R. Proc. Natl.
§
Company contact: [email protected]. Acad. Sci. U.S.A. 2000, 97, 9390-9395.
(1) Pappin, D. J. C.; Höjrup, P.; Bleafby, A. J. Curr. Biol. 1993, 3, 372-332. (9) Molloy, M. P. Anal. Biochem. 2000, 280, 1-10.
(2) Mann, M.; Hojrup, P.; Roepstorff, P. Biol. Mass Spectrom. 1993, 22, 338- (10) Santoni, V.; Molloy, M.; Rabilloud, T. Electrophoresis 2000, 21, 1054-1070.
345. (11) Shevchenko, A.; Jensen, O. N.; Podtelejnikov, A. V.; Sagliocco, F.; Wilm,
(3) Yates, J. R., III; Speicher, S.; Griffin, P. R.; Hunkapiller, T. Anal. Biochem. M.; Vorm, O.; Mortensen, P.; Boucherie, H.; Mann, M. Proc. Natl. Acad.
1993, 214, 397-408. Sci. U.S.A. 1996, 93, 14440-14445.
(4) Mann, M.; Wilm, M. Anal. Chem. 1994, 66, 4390-4399. (12) Gygi, S. P.; Rist, B.; Gerber, S. A.; Turecek, F.; Gelb, M. H.; Aebersold, R.
(5) Corthals, G. L.; Wasinger, V. C.; Hochstrasser, D. F.; Sanchez, J. C. Nat. Biotechnol. 1999, 17, 994-999.
Electrophoresis 2000, 21, 1104-1115. (13) Zhou, H.; Watts, J. D.; Aebersold, R. Nat. Biotechnol. 2001, 19, 375-378.

10.1021/ac0262560 CCC: $25.00 © 2003 American Chemical Society Analytical Chemistry, Vol. 75, No. 8, April 15, 2003 1895
Published on Web 03/01/2003
Multidimensional protein identification technology (MudPIT) is isotopically labeled nutrients that will be incorporated into proteins
another related approach to automated and integrated whole into the growth medium of a living organism. Clearly, this
proteome expression analysis exploiting mass spectrometry and approach is limited to the analysis of organisms whose growth
its ability to identify peptides. In this approach, whole proteomes media can be controlled. Moreover, in this approach, the precise
or fractions are digested and separated using a dual-phase liquid amount of incorporation of heavy and light isotopes cannot really
chromatography separation process followed by in-line analysis be predicted. In vivo approaches are typified by the complex
by electrospray ionization mass spectrometry.16,17 This approach spectra of the mass modified peptides. While these patterns can
differs from other approaches in that no sampling of peptides is be useful for the identification of some peptides, the complexities
performed. MudPIT relies on high-resolution multidimensional and limitations of this approach make it laborious and, thus,
chromatography to resolve all the peptide components of a unattractive. In addition, the organism must be grown on minimal
complex mixture. This technique has the advantage of complete media, which has metabolic implications for the experimental
representation of proteins in the original mixture, but at the cost organism; i.e., the protein expression patterns on minimal media
of high redundancy, which results in high sample complexity and will be a specific response to the medium, and these patterns will
limits throughput. bias the response behavior of the model organism. Consequently,
In addition, labeling techniques to allow relative abundances minimal media cannot permit the full range of protein expression
of proteins in different samples to be determined by mass to be explored. In contrast, in vitro labeling allows labeling of
spectrometry are also needed. A number of such labeling virtually any protein sample, it allows control over the degree of
techniques, notably ICAT, have also been published recently.12,18-20 isotope labeling of each peptide, and the conditions under which
These techniques are all based on standard procedures for the sample proteins are produced do not affect the labeling. In
quantification by mass spectrometry in which an analyte is vitro labeling procedures are, thus, more appealing than in vivo
quantified by comparison with an introduced isotopomer of the procedures as long as robust labeling protocols can be developed.
analyte that acts as an internal standard. Typically, the internal Despite some successes with the isotope labeling techniques
standard is differentiated from the analyte by incorporation of discussed above, all of the published approaches that depend
deuterium, but 13C and 15N are also used. It is assumed in these on deuterium labeling suffer from a number of problems. The
techniques that the relative signal intensities of analyte and most significant is that, although the mass modification that results
standard are directly proportional to their relative concentrations, from isotope labeling is small, there is still a detectable shift in
and since the quantity of the standard is known, this means that the mobility of deuterium differentiated peptides in size/mass-
the quantity of the analyte may be determined from the ratios of dependent separation procedures, such as reversed-phase high
their peak intensities. These standard methods of quantification performance liquid chromatography (HPLC).12,22 Typically, the
have been adapted for the purpose of protein expression analysis heavy peptide migrates more rapidly than the light peptide, often
by the introduction of “heavy” and “light” isotope tags that are eluting as a separate fraction.12 This means that in order to
used to label peptides from corresponding proteins in pairs of accurately determine the quantities of each heavy/light peptide
samples under comparison. The isotope tagging procedures pair, it is necessary to allow both peptides to completely elute to
produce pairs of labeled peptide isotopomers that are mass- allow integration of the ion current for each peptide. As a
differentiated and can act as mutual internal standards. The ICAT consequence, the determination of the peptide identity by se-
procedure, in particular, combines a method of sampling a quencing using MS/MS techniques cannot easily be reconciled
complex protein mixture with a method of determining relative with the need for accurate abundance measurements. Moreover,
abundances by using pairs of isotope-differentiated cysteine- since each peptide pair does not coelute, the isotope-tagged
reactive affinity tags and may be regarded as the most advanced peptides do not act as true standards for each other, reducing
alternative to 2D-PAGE for whole proteome analysis to date. The confidence in the accuracy of relative quantification. In particular,
MudPIT procedure is also compatible with the isotope labeling it is possible that one peptide of a pair, but not the other, may
techniques that have been developed recently.21 coelute with another peptide that suppresses its ionization.
The currently published “heavy/light” isotope labeling tech- Another problem for quantitative analysis of peptides labeled
niques fall into two general categories: in vivo19,20 and in vitro12,18 with conventional isotope labels using LC/MS arises from the
labeling. The former approach requires the introduction of different charge states of the labeled peptides that are produced
by electrospray ionization. This means that the mass difference
(14) Ficarro, S. B.; McCleland, M. L.; Stukenberg, P. T.; Burke, D. J.; Ross, M.
between corresponding peptides labeled with conventional isotope
M.; Shabanowitz, J.; Hunt, D. F.; White, F. M. Nat. Biotechnol. 2002, 20,
301-305. tags varies with the charge state of the peptide. Similarly, the
(15) Oda, Y.; Nagasu, T.; Chait, B. T. Nat. Biotechnol. 2001, 19, 379-382. number of tags incorporated into a peptide will alter the mass
(16) Washburn, M. P.; Yates, J. R. Curr. Opin. Microbiol. 2000, 3, 292-297.
difference between each corresponding peptide from a paired
(17) Washburn, M. P.; Wolters, D.; Yates, J. R. Nat. Biotechnol. 2001, 19, 242-
247. sample.
(18) Goodlett, D. R.; Keller, A.; Watts, J. D.; Newitt, R.; Yi, E. C.; Purvine, S.; Improved labels can solve some of the problems with the above
Eng, J. K.; von Haller, P.; Aebersold, R.; Kolker, E. Rapid Commun. Mass
techniques, for example, the novel 13C reagent for ICAT described
Spectrom. 2001, 15, 1214-1221.
(19) Oda, Y.; Huang, K.; Cross, F. R.; Cowburn, D.; Chait, B. T. Proc. Natl. Acad. by ABI avoids the retention time shift.
Sci. U.S.A. 1999, 96, 6591-6596. This paper describes a novel class of reagents termed tandem
(20) Pasa-Tolic, L.; Jensen, P. K.; Anderson, G. A.; Lipton, M. S.; Peden, K. K.;
mass spectrometry tags (TMTs), where the term “tandem” refers
Martinovic, S.; Tolic, N.; Brice, J. E.; Smith, R. D. J. Am. Chem. Soc. 1999,
121, 7949-7950.
(21) Washburn, M. P.; Ulaszek, R.; Deciu, C.; Schieltz, D. M.; Yates, J. R., III. (22) Griffin, T. J.; Han, D. K.; Gygi, S. P.; Rist, B.; Lee, H.; Aebersold, R.; Parker,
Anal. Chem. 2002, 74, 1650-1657. K. C. J. Am. Soc. Mass Spectrom. 2001, 12, 1238-1246.

1896 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003

to the use of MS/MS for the analysis of these tags. The TMTs d3) is available from ISOTEC Inc, Miamisburg, Ohio. An Fmoc-
overcome a number of the above issues for quantitative proteomic Met-d3 reagent for use in a peptide synthesizer must, however,
analyses while offering other advantages, too. be synthesized manually from the unprotected deuterated me-
Tandem Mass Tag Concept and Design. Two pairs of TMT thionine. The guanidino sensitization enhancement group was
reagents are shown in Figure 1. The reagents are peptides synthesized as an N-hydroxysuccinimide ester (NHS-ester) and
comprising one “tag” amino acid linked to a sensitization group,23-25 added to deprotected R-amino groups of synthetic peptides by
which is a guanidino functionality; one “mass normalization” amino conventional methods during automated peptide synthesis.
acid; and in the second pair of tags, a cleavage enhancement After cleavage from the solid-phase synthesis resins, the
group, which is proline in this case.26 These tags are designed so products were purified by HPLC. The identity of each of the
that on analysis by collision-induced dissociation (CID), the TMT peptides was confirmed by mass spectrometry.
fragment is released to give rise to an ion with a specific mass- Synthesis of the Guanidino-NHS Ester. The synthesis of
to-charge ratio.27,28 The N terminal methionine and guanidino the guanidino-active ester linker (6-[bis(tert-butyloxycarbonyl)-
group comprises the TMT fragment and is distinguished from guanidino]hexanoic acid N-hydroxysuccinimide ester) was as
the second methionine, which comprises the mass normalization follows:
group. Each tag can also bear a reactive functionality. In the figure, (1) Synthesis of Aminoiminomethane Sulfonic Acid. A 50-
the reactive functionality is not specified, but could be an mL portion of acetic anhydride and 2 drops of concentrated
N-hydroxysuccinimide ester, for example, which allows for the sulfuric acid were added to 45 g (397 mmol) of 30% aqueous ice-
specific labeling of amino groups. Clearly, this reactive functional- cooled hydrogen peroxide. After 30 min, 100 mL (1157 mmol) of
ity can be easily varied to allow different biological nucleophiles acetic anhydride was added to the solution at 10-12 °C once again.
to be labeled. In addition, the tag design can be readily modified The reaction mixture was stirred overnight and reached room
to accommodate an affinity ligand, such as biotin. Furthermore, temperature (RT) in that time. After adding 150 mL of methanol,
it should be clear that more than two tags can be generated, the solution made from 10 g (131 mmol) thiourea in 500 mL
allowing for comparison of additional samples or for the introduc- methanol was dropped slowly into the reaction at 15-20 °C. The
tion of labeled standards. reaction was stirred at RT for 48 h. After filtration, the solution
The TMT approach is similar in principle to other peptide was condensed to 60 mL. The obtained product was filtered and
isotope labeling techniques and enjoys the same features as these washed with ethanol and purified by crystallization from acetic
other approaches, with some additional advantages. Pairs of TMT- acid (∼1 L). Yield: 37%.
tagged peptides are chemically identical, like the isotope tags used (2) Synthesis of 6-Guanidinohexanoic Acid. A 6.5-g (50
in other methods, but unlike other isotope tags, the TMTs also mmol) portion of 6-aminohexanoic acid and 6.9 g (50 mmol)
have the same overall mass and comigrate in chromatographic
sodium carbonate were dissolved in 50 mL of water. A 6.2-g (50
separations and, thus, will act as more precise reciprocal internal
mmol) portion of aminoiminomethane sulfonic acid was added to
standards, which leads to more accurate quantification. In addition,
the solution with stirring. After 20 h, the product was filtered and
comigration of TMT-labeled species means that the MS signal
washed with acetic acid, methanol, and then ether. Yield: 76%.
for each peptide pair is not split into two peaks, as in conventional
(3) Synthesis of 6-[Bis(tert-butyloxycarbonyl)guanidino]-
isotope labeling, improving sensitivity in the MS mode. As in other
hexanoic Acid N-Hydroxysuccinimide Ester. A 9.5-g (55
approaches, a short sequence of contiguous amino acids from a
mmol) portion of 6-guanidinohexanoic acid and 55 g (270 mmol)
large protein is often sufficient to uniquely identify the protein,4
of N,O-bis-trimethylsilyl acetamide were stirred in 100 mL of
and the TMT reagents are applicable to any peptide isolation
dichloromethane and heated under refluxing until a clear solution
procedure that other in vitro labeling techniques can be used for.
was obtained (the reaction was left for ∼10 h). A 46-g (210 mmol)
The novel TMT strategy relies on a CID-based technique for
portion of di-tert-butyl pyrocarbonate was added to the solution
quantification of tagged analytes, and this feature improves signal-
at RT, and the reaction mixture was heated under refluxing for 3
to-noise ratios by operation of the MS instrumentation entirely in
h after having been stirred at RT for 18 h (overnight). The solution
the MS/MS mode. This allows untagged material to be ignored,
was then cooled to RT and washed with a 10% citric acid solution
greatly improving data quality.
and a sodium chloride solution. After evaporation of the solvent,
EXPERIMENTAL SECTION the pyrocarbonate was distilled at 80-90 °C under vacuum. The
Syntheses of TMT-Labeled Peptides. Peptides were syn- viscous liquid obtained (30 g) was dissolved in 100 mL dichlo-
thesized using conventional automated Fmoc synthesis techniques romethane with 8.6 g (75 mmol) N-hydroxysuccinimide. A 15.5-g
(both starting from commercially available Fmoc-Gly-Trt-PS resin (75 mmol) portion of dicyclohexylcarbodiimide (DCC) was added
from Rapp Polymere, Germany). Deuterated methionine (Met- in portions to the reaction mixture with stirring at RT. After 17 h,
the urea was removed by filtration. The solution was washed with
(23) Brancia, F. L.; Oliver, S. G.; Gaskell, S. J. Rapid Commun. Mass Spectrom.
2000, 14, 2070-2073.
a 10% citric acid solution, and after removing the solvent, the
(24) Roth, K. D.; Huang, Z. H.; Sadagopan, N.; Watson, J. T. Mass Spectrom. product was purified by chromatography (silica gel; solvent,
Rev. 1998, 17, 255-274. dichloromethane/ethyl acetate). The product was then crystallized
(25) Brancia, F. L.; Butt, A.; Beynon, R. J.; Hubbard, S. J.; Gaskell, S. J.; Oliver,
S. G. Electrophoresis 2001, 22, 552-559.
from diisopropyl ether. Yield: 19%. Rf: 0.77 (dichloromethane/
(26) Schwartz, B. L.; Bursey, M. M. Biol. Mass Spectrom. 1992, 21, 92-96. ethyl acetate: 3/1). fp: 108-109 °C
(27) Schlosser, A.; Lehmann, W. D. J. Mass Spectrom. 2000, 35, 1382-1390. MS/MS Analysis of TMT-Labeled Peptides. MS and MS/
(28) Arnott, D.; Kottmeier, D.; Yates, N.; Shabanovitz, J.; Hunt, D. Proceedings of
the 42nd ASMS Conference on Mass Spectrometry and Allied Topics, Chicago MS analyses were performed on a QTOF2 mass spectrometer
IL, 1994, 470. (Micromass, Manchester, U.K.). HPLC analysis was performed
Analytical Chemistry, Vol. 75, No. 8, April 15, 2003 1897
1898
Analytical Chemistry, Vol. 75, No. 8, April 15, 2003
Figure 1. The structures of two versions of the TMT markers are shown in parts 1a and 1b. The tags are modular, comprising different functional components that correspond to individual synthetic
components in the automated synthesis of these reagents. Each tag comprises a sensitization group and a mass differentiated group that together comprise the TMT fragment that is actually
detected. The TMT fragment is linked to a mass-normalization group that ensures that each tag in a pair of tags shares the same overall mass and atomic composition. The first and second generation
tags are distinguished by the presence of an additional fragmentation-enhancing group, proline, in the second generation tag. The tags will additionally comprise a reactive functionality (R) to enable
the tag to be coupled to any peptide, but in the present experiments, R is one of a number of peptide sequences. The proposed TMT fragment that results from the markers, based on current theories
on protonation dependent mechanisms of backbone fragmentation,27,28 is shown in part 1c.
Table 1. Synthetic TMT-Labeled Peptide Pairs Used for MS and MS/MS Analysis

generation 1 generation 2
peptide sequences Mmono ion at m/z (z) Mmono ion at m/z (z)
1A TMTA-GVATVSLPR 1319.7 660.9 (2+) 1415.7 708.9 (2+)
1B TMTB-GVATVSLPR 1319.7 660.9 (2+) 1415.7 708.9 (2+)
2A TMTA-GLGEHNIDVLEGNEQFINAAK 2688.3 897.1 (3+) 2784.3 928.8 (3+)
2B TMTB-GLGEHNIDVLEGNEQFINAAK 2688.3 897.1 (3+) 2784.3 928.8 (3+)
3A TMTA-GNKPGVYTK 1383.7 462.2 (3+) 1479.7 494.3 (3+)
3B TMTB-GNKPGVYTK 1383.7 462.2 (3+) 1479.7 494.3 (3+)
4A TMTA-GDPAALKRARNTEAARRSRARKLQRMKQGGC 3874.6 969.7 (4+) 3970.6 993.7 (4+)
4B TMTB-GDPAALKRARNTEAARRSRARKLQRMKQGGC 3874.6 969.7 (4+) 3970.6 993.7 (4+)

with a CAP-LC HPLC system (Waters Corporation, Milford, MA) fragments have an m/z of 287 or 290, but in the first generation
(column, PepMap C18 HPLC column from Dionex with a 75-µm tags, a second more intense pair of ions with mass-to-charge ratios
i.d. and a length of 150 mm; solvents, 95% water to 95% acetonitrile, of 270 or 273 is observed. These fragments are thought to result
both with 0.2% formic acid) from the loss of ammonia from the expected tag fragments. At
Ion abundance ratios were determined by summation and lower collision energies, the intensities of these two TMT fragment
smoothing of spectra for each peptide, as it was ionized in the ions varied with the sequence of the attached peptide, but at
electrospray source, followed by determination of peak intensities higher CID energies (70 V), the 270/273 fragments are observed
of the TMT fragments. Corrections were applied if TMT fragment almost exclusively. At these higher collision energies, the 270/
peaks overlapped with isotope peaks from other ions. As long as 273 TMT fragments accurately reflect the abundances of the
the parent peak of a non-TMT fragment ion does not directly peptide pairs. This means that to get consistent behavior from
overlap with the TMT fragment peak, it is possible to correct for the first generation tag, MS/MS analysis had to take place at high
isotope peaks that overlap with the TMT fragment by determining collision energies.
the intensities of isotope peaks as a portion of the monoisotopic In the QTOF instrument, however, at higher energies of
peak and subtracting this from the TMT fragment peak. These collision, the series of b or y ion fragments that provide sequence
can be calculated for all of the possible atomic compositions of information are further fragmented to give smaller species so that
the monoisotopic peaks of ions that are 1 or 2 m/z units from the almost no sequence information can be obtained from the peptide
TMT fragment to determine an average composition, assuming (Figure 2c). As a result of the need for high-energy CID to
the overlapping ions are peptides.29 Thus, in our experiments, an guarantee the release of the TMT fragments and to obtain accurate
ion at 289 has a M + 1 isotope peak that overlaps with the 290 quantification, the first generation TMT units cannot be used for
tag fragment m/z peak. The average intensity of this sort of peak identification of the peptide in the same scan; a second lower
is ∼15% of the 289 peak intensity. Thus, 15% of the 289 peak energy scan to determine sequence information would be re-
intensity can be subtracted from the 290 peak to obtain the correct quired. This will also be true of other serial MS/MS instruments.
TMT fragment intensity. Although ion excitation is more selective in ion traps, it is
somewhat limited in its use with TMTs, because it is not possible
RESULTS AND DISCUSSION to detect small CID fragmentation products of larger precursors
In the following examples, peptides have been synthesized as with this type of instrument. In addition, the benefit of consecutive
if they have been completely labeled on the R-amino group with fragmentation to produce the TMT fragment is lost in the ion trap.
the above tags; i.e., the tag was preincorporated during the These results lead to the development of a second generation
synthesis to test the performance of the tags independently of TMT, which has a proline residue in the TMT unit to enhance
the labeling reactions (see Table 1). The tagged peptides were the fragmentation.13 To demonstrate the effect of the proline, a
analyzed by ESI-MS/MS and LC/ESI-MS/MS. 50:50 mixture of a peptide labeled with the first and second
Comparison of First and Second Generation TMT Tags.
generation tags, respectively, was analyzed by MS/MS. The two
To demonstrate the advantages of a tag designed with a fragmen-
resultant peptides (Figure 2) had ions corresponding to the [M
tation-enhancing group, two different TMT designs were explored.
+ 3H]3+ species at mass-to-charge ratios of ∼897 and 929 for the
The tags differ by the inclusion of proline in the second generation
first and second generation tags, respectively. A collision profile
tags (Figure 1). Proline is known to enhance cleavage of the amide
for both peptide species was recorded automatically in one analysis
bond on its N-terminal side.4
in the QTOF, with the quadrupole set to alternatively select ions
Initial experiments on the fragmentation of the first generation
at m/z 897 and 929, with stepwise increases of the collision energy.
TMT in the QTOF instrument showed that the intensity of the
This procedure ensured the same experimental conditions for both
desired TMT fragments was very dependent on the amino acid
precursors.
sequence of the peptide, and at low collision energies (around 30
At collision energies of 20 V or less, no fragmentation was
V), the TMT fragments did not accurately reflect the abundances
observed for either type of TMT. At a collision energy of 35 V-40
of the tagged peptides. As shown in Figure 1c, the expected TMT
V, it is possible to see the expected TMT fragment ions at m/z of
(29) Gay, S.; Binz, P. A.; Hochstrasser, D. F.; Appel, R. D. Electrophoresis 1999, 290 in the CID spectrum for the peptide with the second
20, 3527-3534. generation tag (Figure 2c), but no fragment ions m/z of 273 can
Analytical Chemistry, Vol. 75, No. 8, April 15, 2003 1899
Figure 2. Comparison of the fragmentation behavior of the peptide 2 labeled with TMT1 and TMT2 tag. This figure shows MS and MS/MS
spectra for triply charged ions of the peptide 2 sequence (see Table 1) labeled with the first and second generation TMTs, guanidinocaproyl-
Met(d3)-Met-GLGEHNIDVLEGNEQFINAAK (peptide 2 + TMT1) and guanidinocaproyl-Met(d3)-Pro-Met-GLGEHNIDVLEGNEQFINAAK (peptide
2 + TMT2), respectively. Part a shows the MS-mode TOF spectrum of the 2 labeled peptides. Part b shows the CID spectra of peptide 2 at 40
V and labeled with TMT1 and TMT2. The presence of the expected tag fragment at m/z 273 is not detected for the first generation TMT ( part
b, bottom), but the expected fragment at 290 is clearly observed at 40 V for the second generation TMT (part b, top). Part c shows the CID
spectra of the peptides at 70 V. The 273 TMT fragment ion of the first generation TMT is detected with greatest sensitivity ( part c, bottom) but
sequence information is lost. The 290 ion for the second generation tag is still detected ( part c, top) with a loss of sensitivity. Note that the 273
ion is not seen, because the proline changes the fragmentation behavior of this tag.

be seen in the spectrum for the peptide with the first generation on the energy needed to release the TMT fragment was much
tag at the same energy (Figure 2c). The TMT fragment for the smaller for the second generation TMT.
peptide containing the first generation TMT is not detected with It is worth noting that the charge state of the TMT-labeled
optimal sensitivity until a collision energy of 70 V is used (Figure peptide selected for MS/MS does not affect the appearance of
2c). Comparison of the CID spectra from peptides labeled with the TMT fragments in the CID spectra of the labeled peptides.
TMTs containing proline with peptides labeled with TMTs without The results in Figure 2 were obtained from [M + 3H]3+ ions.
proline shows clearly that the introduction of the proline residue Sequence and tag abundance data have also been obtained from
as a fragmentation enhancer leads to fragmentation in favor of 4+ and 5+ ions (not shown). This is advantageous, because it
the expected TMT tag fragment without resorting to very high means that scanning of the spectrum can take place without
collision energies. In addition, the identification of the peptide via complex adjustments of the scanning software to compensate for
its b- and y-series ions can also be performed at these lower the charge state of each peptide. In other isotope tagging
collision energies. Smaller peptides labeled with the first genera- procedures, such as ICAT, the charge state alters the mass
tion TMT gave rise to the TMT fragment at lower energies, but difference between each tagged ion pair such that for doubly
higher collision energies were required to release the TMT charged ions, the mass difference is halved; for triply charged
fragment from larger peptides. The size dependence of the peptide ions, the mass difference is one-third of that for the singly charged
1900 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003
Figure 3. TMT1 and TMT2 tag fragment ESI-MS/MS data fitted with regression lines for five different expected and observed ratios of the four
peptides listed in Table 1. Peptides with both first and second generation TMTs incorporated into them were analyzed. Abundance ratios were
determined by analyzing the peak maxima at the d3 (A) and d0 (B) of the tag fragment ion peaks after isotope peak correction at m/z of 287
and 290 for TMT2 and m/z of 270 and 273 for TMT1. CID was carried out at ∼70 V for TMT1-labeled peptides and at ∼35 V for TMT2-labeled
peptides.

ions; etc. Software to scan for peptide pairs using conventional Figure 4 shows the elution profiles of two example peptides
isotope labeling techniques, such as ICAT, must therefore monitored at both of the mass-to-charge ratios of the b2 ions from
compensate for these sorts of problems by allowing for the the TMT fragments. It can be clearly seen that the peptide pairs
different possible mass differences or by ignoring certain classes elute as a single fraction. In MS/MS mode, monitoring of the tag
of ions, which either increases the chance of erroneous identifica- fragment ions produces virtually identical results in each case.
tion of peptide pairs or misses out on potential ion pairs that could For each peptide pair, the observed ratios closely matched the
offer useful information. expected ratios.
The improved behavior of the second generation TMT can be Since the tagged peptides exactly coelute, the ratios of the
seen in Figure 2b (see also 6c), which shows a typical CID peptide pairs are conserved throughout the elution profile, which
spectrum of a peptide labeled with these tags. The TMT fragments means that it is not necessary to integrate the total ion current
revealing the abundance ratios are easily seen at the expected for the eluting ions to determine the relative abundance of each
m/z values of 287 and 290. In addition, it is possible to see both peptide pair. In addition, quantification of the peptide pairs could
b-series and y-series ions, allowing the sequence of the peptide be performed with simultaneous sequence determination (data
to be determined. not shown).
ESI-MS/MS analyses of mixtures of the peptides shown in Analysis of the Sensitivity and Robustness of the TMT
Table 1 were carried out to assess the ability of the two types of Technology. To test the dynamic range of the system and to show
tag to determine the relative abundances of the mixtures of peptide that the properties of the TMT labels are consistent over the entire
pairs labeled with the tags. The peptide mixtures with the dynamic range, the conservation of peptide ratios was examined
expected and measured abundance ratios for both the first and at a range of different concentrations of one of the tagged synthetic
second generation tags are shown in Figure 3. It can be seen that peptides (peptides 3A and 3B). As can be seen from Figure 5,
both generations of TMT provide accurate representation of over a serial dilution of peptides 3A and 3B mixed in a ratio of
abundance ratios of the peptides in the mixtures and that the tags 40:60 from 100 pmoles to 100 fmoles, the ratios were reliably
show linear behavior over the entire range of peptide ratios tested. conserved, with a deviation between 5 and 10% in most cases from
In addition, the lower collision energy needed for the second the expected ratio. These and other results (not shown) indicate
generation tag allows simultaneous sequence determination. that the TMT labels do not reduce the intrinsic sensitivity with
Demonstration of Identical Chromatographic Behavior of which a peptide is detected in the MS/MS mode; i.e., the analysis
TMT Tags in LC/MS. A mixture of pairs of synthetic second of TMT-labeled peptides by CID has at least the same sensitivity
generation TMT-labeled peptides was prepared at a ratio of 60: as the MS/MS of untagged peptides. The sensitivity with which
40. The sequences, theoretical monoisotopic masses, and the it is possible to determine the sequence of tagged peptides also
observed ion mass-to-charge ratios are shown in Table 1. The does not seem to be significantly changed in any of the peptides
peptides were loaded independently onto a C-18 reversed-phase tested so far when compared with the unlabeled peptide. Mean-
HPLC column and chromatographed. The purpose of this experi- ingful differences in the ratios of the peptides can be detected
ment was to demonstrate the exact coelution of corresponding over the entire range of concentrations tested.
pairs of peptides with different TMT tags without any other In a further experiment, the ability to detect labeled peptides
complications. The expected ratios of the peptide pairs were in a complex mixture was examined. The peptide pair 3A and 3B
observed by MS/MS and found to be consistent over the entire bearing the second generation TMT was spiked into an excess
elution time for each peptide pair. of a tryptic digest of an unfractionated yeast protein extract. To
Analytical Chemistry, Vol. 75, No. 8, April 15, 2003 1901
Figure 4. The coelution of each peptide pair, peptides A and B for peptides 1 and 3 from Table 1, can clearly be seen in the C18 reversed-
phase HPLC traces shown above. For each example peptide, the ion currents at m/z 290 (top trace) and 287 (middle trace) corresponding to
the TMT fragments from each of the TMTs are recorded. The bottom trace for each peptide is the total ion current. (Example 1 is peptide 1,
example 2 is peptide 3.)

Figure 5. Dynamic range study of TMT peptide pairs 3A/3B, which are present in a ratio of 40:60 and have been analyzed at dilutions in the
range from 100 fmole to 100 pmole.

simulate a global proteomics analysis, the peptide mixture was and the TMTs can help to overcome noise in the sample. In
analyzed by LC/MS/MS, and the five most intense ions from each addition, there do not seem to be any suppression problems; ratios
elution scan were subjected to CID to identify the peptides. The of peptides present in low concentrations can still be determined
TMT-labeled peptides were detected, and the region of spectrum in the presence of other peptides that are in high concentrations.
corresponding to the TMT fragments was analyzed to determine
the abundance ratio of the detected peptides. Analysis by CID CONCLUSIONS AND PERSPECTIVES
(collision energy of 30 V), provides the spectrum shown in Figure The above results clearly demonstrate the advantages of the
6c. The ratio of peptides 3A and 3B was found to be 39.3% to 60.7%. TMT labels for the simultaneous identification and quantification
The expected ratio was 40% 3A to 60% 3B. The quality of the MS/ of the components of complex protein mixtures. The TMT labels
MS spectrum obtained (Figure 6b and 6c) at the low collision overcome a number of problems associated with previously
energy used allows a clear identification of the peptide sequence described isotope labeling techniques. Conventional deuterium
by database searching. This experiment clearly shows that a labeling suffers from the problem that identical peptides labeled
complex mixture of tryptic peptides does not hinder the analysis with different deuterium-isotope tags do not coelute during
of peptide pairs labeled with the second generation TMT labels, chromatographic separations. Although this issue can be solved
1902 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003
Figure 6. This figure shows the results of a spiking experiment in which peptide pairs 3A and 3B (500 fmol in total, in a ratio of 40:60) bearing
a second generation TMT were mixed with a tryptic digest of an unseparated extract of yeast proteins (0.1 µg). Part a shows the base peak
chromatogram from analysis in the MS mode. During the run, the first five most intensive ions analyzed in MS mode were automatically fragmented
in the MS/MS mode. Collision energies are automatically selected by the QTOF instrument on the basis of charge state and peptide mass-to-
charge ratio (typically 20-40 V). The TMT peptide pairs were investigated and located on the base peak chromatogram. Part b shows the
whole MS/MS spectrum of the peptides at mass-to-charge ratio 929.5. Part c presents a zoom of peptide MS/MS data obtained in part b. It
shows the sequence determining y- and b-series fragments of the peptide. The ratio of the TMT2 fragments was then calculated from the
MS/MS spectrum of the [M + 3H]3+ ion by comparing the intensity of the d0 and d3 TMT fragment mass-to-charge ratios (287 and 290) (part
c).

using 13C, as in the new ICAT reagent, there is also a problem analysis of protein samples, because this should allow more
with different charge states of isotope-labeled peptides, because proteins to be identified in a single analysis in the same time as
the mass difference between corresponding peptides with con- other techniques while using conventional instrumentation. Al-
ventional isotope tags varies with the charge state of the peptide. ternatively, an include mass list can be specified for the quantifica-
Both of these problems create difficulties for the accurate tion of peptides with ions of known mass-to-charge ratios.
quantification and identification of peptide pairs, and these In addition, comigration of the TMT-labeled peptide pairs
problems restrict the throughput of tagged peptides that can be avoids the problem of conventional deuterium-based isotope
identified for any given instrument. labeling in which the MS-mode signal is split into two separate
In contrast, the CID analysis of TMT labels gives rise to isotopomer peaks, resulting in loss of sensitivity. Thus, sensitivity
dependable TMT fragment ions that reflect the relative abun- in the MS mode is retained.
dances of the peptides from which they are derived. With MS/ The use of proline containing TMT labels allows low collision
MS-based scanning of the ions eluting from a chromatographic energies to be used to obtain the TMT fragments that represent
separation, TMTs make it possible to easily identify tagged the ratios of the tagged peptides in their source samples. So far,
species, allowing untagged material to be ignored. The tags show the preferential cleavage of the tag fragment has not noticeably
consistent properties over a wide dynamic range, and labeled diminished our ability to sequence the rest of the tagged peptide.
peptides can be detected with a sensitivity that is comparable to It should be noted that two of the peptides tested in the
other labeling procedures. In addition, the charge state of TMT experiments in this paper contain proline, while a third contained
labeled peptides does not affect the ability to detect the abundance an aspartic acid/proline (DP) linkage, and no problems were
ratios of peptide pairs. encountered with these peptides. The DP peptide probably cleaves
Since the tagged peptides exactly coelute, the ratios of the at the DP linkage before the TMT label can fragment, but
peptide pairs are conserved throughout the elution profile, which consecutive fragmentation in the QTOF produces the desired tag
means that it is not necessary to integrate the total ion current fragments. Consequently, we are optimistic that this will be a
for the eluting ions to determine the abundance of peptide pairs, robust tag design for wider use.
as is necessary for ICAT.22 In the QTOF instrument, CID of the Furthermore, the TMT reagents are modular, and variants are
peptide pairs allows differences in ion abundance of the peptide easily produced. Modifications that could be developed are
pairs to be detected while simultaneously producing the CID changes in the fragmentation enhancer, the inclusion of biotin
spectrum necessary to determine the sequence of these peptide into the tag to allow TMTs to be used as affinity ligands, and
pairs. This feature allows TMT-labeled peptides to be quantified modification of the reaction specificity of the tags to allow the
and identified using the automated selection of the most intense tags to be linked to other reactive functionalities in proteins and
ions in the MS-mode spectra. This has advantages for global other biomolecules. The inclusion of a biotin moiety into a peptide
Analytical Chemistry, Vol. 75, No. 8, April 15, 2003 1903
sequence, which can be achieved with standard peptide synthesis in vitro peptide isolation and labeling techniques in that it will be
techniques,30 does not interfere with the quality of the peptide applicable to proteins that are problematic for 2-D gel electro-
identification.8,12 Different tag and mass normalization function- phoresis, such as small proteins, highly basic or acidic proteins,
alities can be used to allow larger sets of tags, with the same mass and hydrophobic proteins.
to be synthesized for multiplexing. This is an advantage of TMTs The use of MS/MS-based tag detection produces high-quality
that is not readily available to conventional isotope labeling data with good sensitivity, excellent signal-to-noise ratios, and a
procedures. In approaches using conventional isotope tags, label- broad dynamic range. These are all features essential to the
ing each sample or set of standards with a different isotope tag development of truly meaningful proteome analysis techniques.
results in an additional peak in the mass spectrum for each peptide In addition, the effective exploitation of the intrinsic properties of
in each sample: if two samples are analyzed together, there will MS instrumentation to maximize throughput and quality of peptide
be twice as many peaks in the spectrum, and similarly, with three identification is crucial to achieving the goal of genuine global
samples, the spectrum will be three times more complex than analysis of cell and tissue samples. Some of the abilities of TMT
for one sample alone. The TMT labels do not increase the labeling to achieve these goals have been demonstrated here, and
complexity in the mass spectrum at all. further advantages will become available with newer reagents.
Because there is no particular restriction on the reactive Thus, TMT labeling provides a new benchmark for automated
functionality that might be used with these tags, it is anticipated quantitative proteome analysis.
that the TMT labeling should be applicable to both global peptide
isolation procedures, such as MudPIT or ICAT, and to specific ACKNOWLEDGMENT
procedures, such as phosphopeptide13-15 and glycopeptide isola- We thank Dr. Linkies and Dr. Reuschling, Aventis R&T,
tion. Therefore, TMT labeling should allow peptide-based protein Germany, for the synthesis of the activated residues (guanidino-
characterization procedures to be fully automated for quantitative, NHS and Fmoc-methionine-d3) and for valuable discussions.
comparative, functional proteome analysis. The TMT strategy,
when applied, will therefore enjoy the same advantages as other Received for review October 23, 2002. Accepted January
30, 2003.
(30) Chersi, A.; Giommi, S.; Rosano, L. Biochem. Biophys. Acta 2000, 1474, 196-
200. AC0262560

1904 Analytical Chemistry, Vol. 75, No. 8, April 15, 2003