Mass defect labeling of cysteine for improving peptide assignment in shotgun proteomic analyses

Article abstract of DOI:10.1021/ac0600407

A method for improving the identification of peptides in a shotgun proteome analysis using accurate mass measurement has been developed. The improvement is based upon the derivatization of cysteine residues with a novel reagent, 2,4-dibromo-(2′-iodo)acetanilide. The derivitization changes the mass defect of cysteine-containing proteolytic peptides in a manner that increases their identification specificity. Peptide masses were measured using matrix-assisted laser desorption/ionization Fourier transform ion cyclotron mass spectrometry. Reactions with protein standards show that the derivatization of cysteine is rapid and quantitative, and the data suggest that the derivatized peptides are more easily ionized or detected than unlabeled cysteine-containing peptides. The reagent was tested on a ¹⁵N-metabolically labeled proteome from M. maripaludis. Proteins were identified by their accurate mass values and from their nitrogen stoichiometry. A total of 47% of the labeled peptides are identified versus 27% for the unlabeled peptides. This procedure permits the identification of proteins from the M. maripaludis proteome that are not usually observed by the standard protocol and shows that better protein coverage is obtained with this methodology.

Full text of DOI:10.1021/ac0600407

Anal. Chem. 2006, 78, 3417-3423
Mass Defect Labeling of Cysteine for Improving
Peptide Assignment in Shotgun Proteomic
Analyses
Hilda Hernandez, Sarah Niehauser, Stacey A. Boltz, Vijay Gawandi, Robert S. Phillips, and
I. Jonathan Amster*
Department of Chemistry, University of Georgia, Athens, Georgia 30602
ment of shotgun proteomic methods.5-9 These methods identify
and quantify proteins that have not been separated prior to
digestion. The basis of this approach is to perform a batch
digestion of an unseparated protein mixture, to separate the
resulting peptides by one or more dimensions of liquid chroma-
tography, and to identify the proteins from which the peptides
derive by mass spectrometry analysis.8
A method for improving the identification of peptides in
a shotgun proteome analysis using accurate mass mea-
surement has been developed. The improvement is based
upon the derivatization of cysteine residues with a novel
reagent, 2,4-dibromo-(2′-iodo)acetanilide. The derivitiza-
tion changes the mass defect of cysteine-containing pro-
teolytic peptides in a manner that increases their identi-
fication specificity. Peptide masses were measured using
matrix-assisted laser desorption/ionization Fourier trans-
form ion cyclotron mass spectrometry. Reactions with
protein standards show that the derivatization of cysteine
is rapid and quantitative, and the data suggest that the
derivatized peptides are more easily ionized or detected
than unlabeled cysteine-containing peptides. The reagent
was tested on a 15N-metabolically labeled proteome from
M. maripaludis. Proteins were identified by their ac-
curate mass values and from their nitrogen stoichiometry.
A total of 47% of the labeled peptides are identified versus
27% for the unlabeled peptides. This procedure permits
the identification of proteins from the M. maripaludis
proteome that are not usually observed by the standard
protocol and shows that better protein coverage is ob-
tained with this methodology.
Two mass spectrometry approaches for shotgun proteomic
analysis have been reported. First is the use of tandem mass
spectrometry to generate fragmentation data, which can be used
by search engines to identify the protein origin of the pep-
tides.2,6,8,10,11 These methods are able to detect and identify a wide
variety of protein classes including those with extremes in
isoelectric point, molecular weight, abundance, and hydrophobic-
ity. However, these methods are time-consuming and produce very
large data sets, as they require the generation of a fragmentation
spectrum for each peptide in a mixture that contains thousands
of components. A second approach is the use of accurate mass
measurement to identify proteins. If the molecular masses of the
peptides from a batch digest are measured with high enough mass
measurement accuracy (MMA), a reasonable fraction of their
masses can uniquely identify them by comparison to a list of
masses for all of the possible proteolytic peptides predicted from
an in silico digest of the genome. Other experimental information
can be used to increase the fraction of identified peptides, for
example, HPLC retention time.11 Methods that combine MMA
with the MS/MS capabilities have also been reported.11,12
In this paper, we describe a new method for improving the
specificity of protein identification by accurate mass measurement
of peptides. The improvement is based upon the derivatization of
The primary goal of a proteomic analysis is to be able to
systematically identify and quantify the majority of proteins
expressed in a cell or tissue.1,2 The conventional approach for
conducting proteome-wide studies is two-dimensional polyacryl-
amide gel electrophoresis (2D-PAGE),3 where a large number of
proteins can be separated on the basis of their isoelectric point
and molecular weight. Although 2D-PAGE technology has been
the chief technology for proteomic analysis to date, it has
recognized limitations, such as a bias toward the most abundant
proteins and dynamic range and protein solubility issues that
complicate the detection and separation of low-abundance and
hydrophobic proteins.4 In recent years, a number of researchers
have focused on improving proteomic analyses via the develop-
(5) Conrads, T. P.; Anderson, G. A.; Veenstra, T. D.; Pasa-Tolic, L.; Smith, R.
D. Anal. Chem. 2000, 72, 3349-3354.
(6) Link, A. J.; Eng, J.; Schieltz, D. M.; Carmack, E.; Mize, G. J.; Morris, D. R.;
Garvik, B. M.; Yates, J. R., III. Nat. Am. 1999, 17, 676-682.
(7) Nepomuceno, A. I.; Muddiman, D. C.; Bergen, R.; Craighead, J. R.; Burke,
M. J.; Caskey, P. E.; Allan, J. A. Anal. Chem. 2003, 75, 3411-3418.
(8) Wolters, D. A.; Washburn, M. P.; Yates, J. R., III. Anal. Chem. 2001, 73,
5683-5690.
(9) McDonald, H.; Yates, J. R., III. Disease Markers 2002, 18, 99-105.
(10) Eng, J. K.; McCormack, A. L.; Yates, J. R., III. J. Am. Soc. Mass Spectrom.
1994, 5, 976-989.
* To whom correspondence should be addressed. Phone: (706) 542-2001.
E-mail: jamster@uga.edu.
(1) Gygi, S. P.; Rist, B.; Aebersold, R. Curr. Opin. Biotechnol. 2000, 11, 396-
401.
(2) Aebersold, R.; Goodlett, D. R. Chem. Rev. 2001, 101, 269-295.
(3) Peng, J.; Gygi, S. P. J. Mass Spectrom. 2001, 36, 1083-1091.
(4) Beranova-Giorgianni, S. Trends Anal. Chem. 2003, 22, 273-281.
(11) Smith, R. D.; Anderson, G. A.; Lipton, M. S.; Pasa-Tolic, L.; Shen, Y.; Conrads,
T. P.; Veenstra, T. D.; Udseth, H. R. Proteomics 2002, 2, 513-523.
(12) Lipton, M. S.; Pasa-Tolic, L.; Anderson, G. A.; Anderson, D. J.; Auberry, D.
L.; Battista, J. R.; Daly, M. J.; Smith, R. D. Proc. Natl. Acad. Sci. U.S.A. 2002,
99, 11049-11054.
10.1021/ac0600407 CCC: $33.50 © 2006 American Chemical Society
Published on Web 04/19/2006
Analytical Chemistry, Vol. 78, No. 10, May 15, 2006 3417

Figure 1. Histogram of the molecular mass distribution of the predicted tryptic peptides of M. maripaludis over the range 1500-1503 Da,
Illustrating the distribution of mass defects of peptides. The bin size is 0.01 amu. Peptide masses are observed to cluster in approximately
one-third of the available mass space.
a specific amino acid with a reagent that changes the mass defect
of the peptide. For the purpose of discussion, we refer to the mass
defect as the difference between the exact monoisotopic mass of
a compound and its nominal molecular weight, that is, the weight
based on the nucleon values of the most abundant isotope of each
element, e.g., 12 amu for C, 16 amu for O, etc. Peptides are
composed principally of elements from the first two rows of the
periodic table. These elements have mass defects that lie in the
range of ( 0.008 amu. The mass defect of peptide molecules is
∼+0.05 amu/100 amu of molecular weight, i.e., a 1-kDa peptide
has a mass defect of ∼+0.5 amu, and a 2-kDa peptide has a mass
defect of ∼1 amu. The positive mass defect is a result of the high
stoichiometric proportion of hydrogen atoms in a peptide molec-
ular formula (the hydrogen mass defect is +0.0078 amu).
Although peptide molecules have significant mass defects because
of the large number of atoms from which they are assembled,
the distribution of mass defects is generally narrow, causing
peptide molecular weights at any given nominal mass to occupy
only a small portion of a unit mass. This is illustrated in Figure 1,
which shows a histogram of monoisotopic masses for the 125
possible tryptic peptides (up to 1 missed cleavage) with molecular
weights between 1500 and 1503 that one predicts for all proteins
in the sequence database for the organism Methanococcus mari-
paludis. This organism has 1722 open reading frames, which is
about average for a single-cell organism, and has ∼95 700
predicted tryptic peptides with molecular weights above 700 amu
(allowing up to 1 missed cleavage), and the predicted peptide mass
distribution is similar to that of any organism. Because of the
narrow distribution of mass defects for peptides, their molecular
weights cluster into one-third of the total mass space causing
masses to overlap and reducing the specificity of a peptide mass
for identifying the protein origin. Greater specificity would be
possible if the peptide masses were distributed more evenly across
the mass scale.
classes.14-19 Labeling the N-terminus of a protein with a compound
that alters the mass defect is used to distinguish the N-terminal
peptide fragments from C-terminal and internal fragments pro-
duced by nozzle-skimmer dissociation of intact proteins and is
the basis of a commercial reagent (IDBEST) and process.20,21 We
report here a method for altering the mass defects of a selected
fraction of the peptides in a batch digest of a proteome so that
the resulting peptides can be more readily identified by accurate
mass measurement.
EXPERIMENTAL SECTION
Reagent Synthesis. The cysteine-alkylating reagent, 2,4-
dibromo-(2′-iodo)acetanilide, was prepared by addition of 5.4 mmol
(0.46 mL) of oxalyl chloride (Acros Organics, Morris Plains, NJ)
in 2.7 mL of dry dichloromethane to 1 equiv (1 g) of 2-iodoacetic
acid (Acros Organics) in 4 mL of dry dichloromethane. This
mixture was stirred for 3 h at 0 °C under nitrogen to yield a pink
solution (2-iodoacetyl chloride.) This solution was added dropwise
with stirring to 1 equiv (1.3 g) of 2,4-dibromoaniline (Acros
Organics) in 10 mL of dry dichloromethane. A white crude solid
appeared as a precipitate and was collected by filtration and
purified by recrystallization from hot water to give the final product
in 70% yield. The structure of the purified 2,4-dibromo-(2′-
1
iodo)acetanilide was confirmed by H NMR and mass spectrom-
etry (NMR and MS spectra are included as Supporting Informa-
tion.). All reagents and solvents were used as purchased without
further purification.
Protein Labeling. The labeling of cysteine before versus after
tryptic digestion was compared for a number of proteins. We
consistently find that the best results are obtained by labeling
before digestion, as it is easier to remove the excess labeling
(13) Beynon, J. H. Advances in Mass Spectrometry; Pergamnon Press: New York,
1963.
(14) Kendrick, E. Anal. Chem. 1963, 35, 2146-2154.
(15) Wu, Z. G.; Rodgers, R. P.; Marshall, A. G. Anal. Chem. 2004, 76, 2511-
2516.
(16) Marshall, A. G.; Rodgers, R. P. Acc. Chem. Res. 2004, 37, 53-59.
(17) Stenson, A. C.; Marshall, A. G.; Cooper, W. T. Anal. Chem. 2003, 75, 1275-
1284.
(18) Hughey, C. A.; Hendrickson, C. L.; Rodgers, R. P.; Marshall, A. G.; Qian,
K. N. Anal. Chem. 2001, 73, 4676-4681.
The narrow distribution of mass defects for a compound class
has been noted previously by other researchers in mass spec-
trometry. Perfluoroalkanes have distinctly different mass defects
that do not overlap those of most other organic compounds and
have long been employed as internal calibrants for exact mass
measurements.13 The components of complex mixtures of small
molecules can be assigned a Kendrick mass defect value, which
allows homologous series to be assigned to various compound
(19) Jones, J. J.; Stump, M. J.; Fleming, R. C.; Jackson O. Lay, J.; Wilkins, C. L.
J. Am. Soc. Mass Spectrom. 2004, 15, 1665-1674.
(20) Hall, M. P.; Schneider, L. V. Expert Rev. Proteomics 2004, 1, 421-431.
(21) Schneider, L. V.; Hall, M. P.; Petesch, R. Target Discovery: U. S.; 2002.
3418 Analytical Chemistry, Vol. 78, No. 10, May 15, 2006

reagent from a protein solution than from a solution of lower
molecular weight peptides. Each protein standard was dissolved
in alkaline solution (10 mM ammonium bicarbonate) to make a 1
mg/mL solution and denatured by heating at 95 °C. Disulfide
bonds were reduced by addition of tris(2-carboxyethyl)phosphine
(Pierce Biotechnology, Rockford, IL). The protein then underwent
reaction with a 100-fold molar excess of 2,4-dibromo-(2′-iodo)
acetanilide at pH 8 for 90 min in the dark at room temperature.
Prior to trypsin digestion, the derivatized protein was subjected
to centrifugal size exclusion chromatography using a 3-mL spin
column packed with Sephadex G-25 (Aldrich, St. Louis, MO) to
remove excess 2,4-dibromo-(2′-iodo) acetanilide. Trypsin digestion
was performed under standard conditions (Promega, Madison
WI), i.e., at 37 °C, pH 7, for 18 h.
Proteome Labeling. Whole cell lysates were extracted from
M. maripaludis that was grown on minimal media with ammonium
sulfate as the sole source of nitrogen. Cells were grown using
ammonium sulfate both with the naturally occurring isotopic
composition (99.6% 14N, 0.4% 15N) and with 98% 15N-enrichment.
The cells were concentrated by centrifugation at 10 000 rpm for
30 min; lysis of the cells was performed with a French press. DNA
was digested and removed from the extract by adding DNAase
to the sample followed by centrifugation. Equal amounts of protein
extracts were mixed together before batch trypsinolysis. Prior to
denaturing and labeling of the proteome, small molecules were
removed by centrifugal size exclusion spin columns packed with
Sephadex G-25. Subsequent treatment of the sample followed the
procedure described above for labeling of the protein standards.
Total protein concentrations were determined spectrophotometri-
cally measuring at 562 nm using a bicinchoninic acid protein assay
kit (Pierce).
Table 1. Mass Difference from Nucleon Value of the
Most Abundant Isotope of the Elements Found in
Proteins
element
mass defect (amu)
12C
1H
0
0.0078
-0.0051
0.0031
-0.0279
16O
15N
32S
Protein Identification. The molecular weight of the peptides
and their nitrogen stoichiometry were determined from the
MALDI-FTICR mass spectrum. The number of nitrogen atoms
in each peptide was determined from the mass separation between
the monoisotopic peak of the peptide containing the natural
distribution of 14N/15N and the monoisotopic peak of the 15N-
enriched counterpart. The data were analyzed using software that
was developed in-house to identify the proteins from which the
peptides were derived. The software compares the experimentally
determined molecular weight and nitrogen stoichiometry with
values in a look-up table that is populated with the predicted tryptic
fragments (up to 1 missed cleavage) for all protein sequences for
the organism in question. A peptide is considered to be identified
when there is only one predicted peptide that meets the following
match criteria: the predicted peptide has a mass that lies within
a specified mass tolerance of the measured molecular weight, and
it has the same nitrogen stoichiometry as the measured value.
Peptide identifications were made using a mass tolerance of 10
ppm.
Mass Spectrometry. Samples were analyzed by matrix-
assisted laser desorption/ionization (MALDI) Fourier transform
ion cyclotron resonance (FTICR) mass spectrometry using a 7-T
magnet (Bruker Daltonics Inc, Billerica, MA). This instrument is
equipped with a SCOUT 100 MALDI source, which desorbs ions
at elevated pressure (∼1 mTorr) to suppress metastable decom-
position. Conditions for operation of the FTICR MS were similar
to those reported previously,22 and external mass calibration was
established using a peptide mixture generated by tryptic digestion
of chicken egg albumin (Sigma, St. Louis, MO). The MALDI
matrix was 2,5-dihydroxybenzoic acid (DHB) (Lancaster, Pelham,
NH).
High-Performance Liquid Chromatography. Separations of
peptide mixtures were performed on an UltiMate Plus, FAMOS
by Dionex (Sunnyvale, CA). Reversed-phase columns used were
as follows: (1) 75 µm i.d. × 15 cm, C18 PepMap100, 3 µm, 100
Å; (2) 75 µm i.d. × 15 cm, C8 PepMap100, 3 µm, 100 Å (LC
Packings-Dionex). Mobile phase A was water/acetonitrile/tri-
fluoroacetic acid (98:2:0.1 by volume), and mobile phase B was
acetonitrile. A gradient from 0 to 100% B over 90 min was used at
an approximate column flow of 300 nL/min; the total run time
was 120 min. The eluate was collected onto a stainless steel
MALDI target at 60-s intervals using a Probot Micro Fraction
Collector (LC Packings-Dionex). The MALDI matrix was added
after the fraction collection was completed, requiring resuspension
of the dried, fractionated peptides in 0.5 µL of the matrix solution
(1 M DHB in 50:50:0.1 water/acetonitrile/trifluoroacetic acid.)
RESULTS AND DISCUSSION
Mass Defect Labels. The narrow distribution of mass defects
that is characteristic of peptides arises in part from the small mass
defect of their component elements and from the uniform
stoichiometry of peptides. Table 1 shows the mass defect of the
elements that comprise proteins. As can be seen, their mass
defects are small (less than 10 mmu for H, C, N, and O, and ∼28
mmu for S). The average elemental ratio for an amino acid residue
is C4.9384H7.7583N1.3577O1.4773S0.0417.23 Given that nitrogen (mass defect,
+3.1 mmu) and oxygen (mass defect, -5.1 mmu) have compa-
rable stoichiometric values, their mass defects tend to cancel in
a peptide. One can see that the mass defect of a peptide is
principally due to hydrogen and that the distribution of mass
defects comes from the narrow distribution of elemental stoichi-
ometries. Figure 1 suggests that the distribution of mass defects
at any nominal mass is roughly one-third of an amu. One can
calculate the distribution of mass defects at each nominal mass
for the tryptic peptides of all proteins in a database, and we have
done this for peptides with masses from 700 to 3000 that derive
from the proteins in the M. maripaludis database. The composite
distribution of mass defects around the average value at each
nominal mass is shown in Figure 2A. As can be seen, peptides
masses occupy only one-third of the available mass scale, which
causes some of the predicted masses to overlap, even at a mass
(22) Wong, R. L.; Amster, I. J. J. Am. Soc. Mass Spectrom. 2006, 17, 205-212.
(23) Senko, M. W.; Beu, S. C.; McLafferty, F. W. J. Am. Soc. Mass Spectrom.
1995, 6, 52-56.
Analytical Chemistry, Vol. 78, No. 10, May 15, 2006 3419

by underivatized peptides. To achieve this end, we have synthe-
sized a reagent that we refer to as a mass defect label (MDL),
which derivatizes a specific type of amino acid and which changes
the mass of the resulting product in a manner that makes it easy
to distinguish derivatized peptides from other peptides of the same
nominal mass. The ideal tagging reagent will (1) have high
reaction specificity for a low-abundance amino acid such as
cysteine or tryptophan, (2) introduce a mass defect shift of 0.3-
0.6 amu, (3) be stable to the chemical and physical conditions
necessary for derivatization and mass spectral characterization,
and (4) have no deleterious effects on peptide solubility or
ionization efficiency. The MDL reported here is a derivative of
iodoacetamide and reacts specifically with cysteine, as shown in
Figure 3. Figure 2B illustrates the change in the mass defect
distribution for the tryptic peptides that is expected from cysteine
derivatization of all the proteins in the M. maripaludis sequence
database. The derivatized peptide masses occupy a region in which
no unlabeled peptides are found, ∼0.3 amu below the unlabeled
peptides. Because only 15-20% of tryptic peptides contain cysteine,
fewer peptides will occupy the new region of mass, and therefore,
there is a lower probability of mass overlap for predicted peptides.
This suggests that a higher proportion of derivatized peptides can
be identified by their mass compared to underivatized peptides.
Labeling of Protein Standards. Several protein standards
were tested, including bovine serum albumin (BSA), â-lactoglo-
bulin, ovalbumin, and carbonic anhydrase. These proteins under-
went derivatization of their cysteine residues with the MDL,
digestion by trypsin, and analysis by MALDI-FTMS. Mass defect
labeled peptides could be identified both by their mass defect
values and by the isotope pattern that is characteristic of the
presence of two bromine atoms. Figure 4 shows the calculated
isotopic distribution of a peptide (BSA 445-458) that is labeled
by the mass defect reagent and compares the distribution to that
of the corresponding unlabeled peptide. The use of chlorine
isotope patterns to identify derivatized cysteine-containing peptides
in a proteomics assay has been reported previously.24 Here, we
do not use the isotopic pattern to establish that derivatization has
occurred. The mass defect of the resulting peptide provides this
information. However, it is important that the unusual isotopic
pattern is taken into consideration when assigning the monoiso-
topic peak. Figure 5 shows a mass spectrum of the tryptic peptides
of bovine serum albumin that has been derivatized with the MDL;
peaks corresponding to labeled peptides are identified with a
square. As can be seen in the mass spectrum, many of the
abundant peaks in the mass spectrum are from derivatized
Figure 2. (A) Composite distribution of mass defects for all tryptic
peptides of M. maripaludis with molecular weights between 700 and
3500 amu. The horizontal axis is the mass difference (amu) between
a peptide’s mass defect and the average mass defect for all peptides
of the same nominal mass. (B) The composite distribution when all
the cysteine-containing peptides have been labeled. The central
distribution corresponds to all peptides that do not contain cysteine.
All singly labeled cysteine-containing peptides appear in the smaller
distribution centered at -0.30 amu. Doubly labeled cysteine-contain-
ing peptides appear at +0.40 amu.
tolerance of 10 ppm. To shift some of the peptide masses to the
region of the mass scale that is unpopulated, we alter the mass
defects of a portion of the peptides by derivatizing a less frequently
occurring amino acid, cysteine, with a reagent that introduces a
large mass defect. This is accomplished by introducing a heavy
element with a large mass defect into the elemental composition,
in this case, bromine.
Derivatization of a specific amino acid with a compound that
affects the mass defect will yield two sets of peptides: unlabeled
peptides with typical mass defects, and labeled peptides with
masses that lie in a region of the mass scale that is unoccupied
Figure 3. Mechanism of an alkylation reaction of a cysteine-containing peptide with 2,4-dibromo-(2′-iodo)acetanilide.
3420 Analytical Chemistry, Vol. 78, No. 10, May 15, 2006

digest. Overall, these data suggest that the reaction of the MDL
reagent is specific for cysteine residues, is quantitative in reactivity
(no underivatized cysteines were observed), and has no adverse
effect on the detectability of the derivatized peptides by MALDI
mass spectrometry.
Protein Identification. To test the effectiveness of this
method for improving protein identification, we derivatized a
proteome sample from the organism M. maripaludis. For this
experiment, we also use endogenous 15N labeling of protein
mixtures to improve the specificity of the protein identification.
All proteins from two whole-cell lysates are isolated from two
identical cultures, one grown using a nitrogen source (ammonium
sulfate) with the natural abundance of 15N and the other with 98%
15N. Equal amounts of protein are then collected from each culture
and combined.25 This method is a useful tool to assist with protein
identification. Previously, we have found a significant improvement
in the ability to identify peptides by accurate mass measurement
when nitrogen stoichiometry is used as a search constraint.
(Parks, B. A.; Amster, I. J., manuscript in preparation).
M. maripaludis contains 1722 open reading frames (ORFs),26
and 18% of the 95 719 predicted tryptic peptides with up to 1
missed cleavage contain cysteine. The utility of this approach (15N
and MDL labeling) to protein identification by accurate mass
measurement has been estimated for this organism at a mass
search tolerance of 10 ppm; the fraction of unique peptides
increases from 8% (unlabeled peptides) to 43% (labeled peptides)
when all the possible peptides up to m/z 3500 are taken into
account. If only the mass defect labeled cysteine-containing
peptides are considered, 75% of the masses are unique (database
searching with 10 ppm mass tolerance and using the nitrogen
stoichiometry as a search constraint).
Figure 4. Calculated isotopic pattern for the peptide MPCTEDYLS-
LILNR from BSA (residues 445-458) (A) without and (B) with the
dibromoacetanilide mass defect label.
Increasing the percentage of identified peptides should in-
crease the number of identified proteins. This was examined for
the M. maripaludis proteome. Whole-cell lysates from M. mari-
paludis were derivatized and digested by trypsin and subsequently
fractionated by nano-LC using a C18 column. The fractions were
analyzed by MALDI-FTMS. Analysis of the spectra resulted in
the assignment of 1449 nonredundant peptides masses. Out of
these, 156 (11%) were found to be mass defect labeled peptides.
Using these data, a search was made against a list of predicted
M. maripaludis tryptic peptide masses. Using a mass tolerance
of 10 ppm, this resulted in the identification of 304 proteins using
both nitrogen stoichiometry and mass defect labeling, which is
an improvement of 14% over the 268 proteins identified when the
search is made against a list that does not include the MDL
peptides. We have previously analyzed the same proteome (but
without mass defect labeling or cysteine alkylation) several times
under similar conditions, and we typically identify 250 ( 25
proteins. We attribute the improvement in proteome coverage to
the fact that mass defect labeling increases both the detectability
and the identification specificity of cysteine-containing peptides.
To check the effect of the MDL on the detectability of cysteine-
Figure 5. MALDI-FTICR mass spectrum obtained of a BSA digest.
Mass defect labeled-peptides are denoted with a box. Inset shows a
mass scale expansion of the peaks near m/z 1957, identified as the
peptide MPCTEDYLSLILNR, whose predicted isotope pattern is
shown in Figure 4B.
peptides, demonstrating that the MDL does not adversely affect
the detectability of the peptides. No nonderivatized cysteine-
containing peptides were found in the mass spectra for any of
the protein tryptic digests that were tested, suggesting that the
derivatization reaction was complete. Bovine serum albumin
contains 35 cysteines, and 32 labeled cysteine residues were
observed in the mass spectra of the tryptic peptides. Interestingly,
in the underivatized control spectrum, only five cysteine-containing
peptides were observed, suggesting that this derivatization in-
creases the detectability of the cysteine-containing peptides. For
â-lactoglobulin, five out of seven possible cysteines were observed
in their labeled state, and for ovalbumin, three out of six labeled
cysteines were observed. Bovine carbonic anhydrase II, which
does not have a cysteine residue, served as a negative control.
No labeled peptides were found in the mass spectrum of its tryptic
(25) Oda, Y.; Cross, F. R.; Cowbum, D.; Chait, B. T. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 6591-6596.
(26) Hendrickson, E. L.; Kaul, R.; Zhou, Y.; Bovee, D.; Chapman, P.; Chung, J.;
Conway de Macario, E.; Dodsworth, J. A.; Gillett, W.; Graham, D. E.; Hackett,
M.; Haydock, A. K.; Kang, A.; Land, M. L.; Levy, R.; Lie, T. J.; Major, T. A.;
Moore, B. C.; Porat, I.; Palmeiri, A.; Rouse, G.; Saenphimmachak, C.; Soll,
D.; Van Dien, S.; Wang, T.; Whitman, W. B.; Xia, Q.; Zhang, Y.; Larimer, F.
W.; Olson, M. V.; Leigh, J. A. J. Bacteriol. 2004, 186, 6956-6969.
(24) Goodlett, D. R.; Bruce, J. E.; Anderson, G. A.; Rist, B.; Pasa-Tolic, L.; Fiehn,
O.; Smith, R. D.; Aebersold, R. Anal. Chem. 2000, 72, 1112-1118.
Analytical Chemistry, Vol. 78, No. 10, May 15, 2006 3421

Figure 7. Histogram for all possible tryptic peptides from M.
maripaludis within 700 and 3500 amu. Gray bars represent the
number of MDL peptides and black bars the total number of peptides
for each 100 amu mass bin.
Figure 6. Chromatogram and plot of percentage of labeled peptides
versus elution time for (A) C18 column proteome separation and (B)
C8 column proteome separation. Percent of labeled peptides was
calculated using the total number of peptides observed and the
number of MDL peptides found for each fraction collected and then
analyzed by MALDI-FTICR mass spectrometry.
Combining both sets of data, the total number of observed
peptide pairs (14N/15N) is 6146; 475 of these were found to be
labeled with the cysteine-specific reagent. It is useful to use this
large data set to examine the improvement in database searching
that results from mass defect labeling and metabolic 15N labeling.
For peptides without a mass defect label, the fraction of unique
peptides goes from 7% when using only the molecular weight to
search the database (i.e., no nitrogen stoichiometry data used in
search) to 27% for the non-MDL peptides when the nitrogen
stoichiometry constraint is used. For the mass defect labeled
proteome, the number of unique peptides increases to 2108, which
represents 34% of the total number of peptides. If one considers
only the peptides labeled by 2,4-dibromoacetanilide, 47% of the
peptides are identified. Having a higher percentage of unique
peptides increases the number of identified proteins. Indeed,
identification of proteins shows that if only the nonlabeled peptides
are used, 377 proteins are identified compared to 425 proteins
identified when all the found peptides masses are used. These
“extra” 48 proteins are not usually identified from the complex
mixture of proteins from M. maripaludis by the standard protocol
(no cysteine alkylation), demonstrating that better protein cover-
age is obtained by using the accurately measured masses of mass
defect labeled cysteine-containing peptides to identify proteins.
We anticipate significant improvement in this method by
refinement of this technique. For example, we note that the
percentage of identified peptides obtained in these experiments
is lower than one would predict from a statistical analysis of the
proteome. The expected identification specificity mentioned above
(43% identification for non-MDL peptides, searching at 10 ppm
mass tolerance and using the nitrogen stoichiometry as a
constraint; 75% identification for MDL-peptides) was calculated
using all the possible tryptic peptides in the mass range of 700-
3500 amu. Figure 7 shows a plot of the number of peptides
observed versus their mass-to-charge ratio for the experiment
using a C18 analytical column. Most peptides are found in the
range between 700 and 2500 amu. The calculated fraction of
unique peptides for the tryptic peptides within this mass range is
36%, which corresponds well with the observed experimental result
of 34%. Detection of higher mass peptides can be achieved by
containing peptides, we have made MALDI-FTMS measurements
of the tryptic digest products of BSA prepared using three different
methods: (1) with no alkylation of cysteine; (2) with alkylation
by iodoacetamide (carbamidomethylation); (3) with alkylation by
the mass defect label. Each of the three digests were analyzed
four times. Of the 35 cysteine residues in BSA, we observe 4-6
(average equals 5) when the cysteines are not alkylated, 8-15
(average 11.3) when cysteines are alkylated by iodoacetamide,
and 12-20 (average 15.5) when the mass defect label is used.
These data show that the MDL procedure provides 50% better
detectability for cysteine-containing peptides compared to carba-
midomethylation and 300% improvement compared to peptides
with unalkylated cysteines.
Detailed analysis of the data gave some insight into the
hydrophobicity of the labeled peptides. Figure 6A shows a graph
of the percentage of labeled peptides found per fraction versus
the retention time. Most of the labeled peptides eluted from the
column after the gradient reaches 50% organic composition. These
data suggest that the labeled cysteine-containing peptides are
more hydrophobic, consistent with the structure of the mass defect
label. Earlier elution and better separation of this sample can be
achieved by using a column with a less hydrophobic stationary
phase. We have examined the same labeled proteome using a C8
column. Analysis of the data resulted in assignment of 1195 pairs
of nonredundant peptides masses, of which 126 (11%) were mass
defect labeled peptides. Figure 6B shows the percentage of labeled
peptides per fraction versus retention time with the C8 column.
The labeled peptides are found to be distributed more evenly
throughout the LC separation with the C8 column. Nevertheless,
the total number of identified proteins shows a slight decrease
when compared with the data obtained using a C18 column (279
versus 304 identified proteins). Based on the results obtained, it
appears that earlier elution of mass defect labeled peptides does
not seem to positively affect the total number of those peptides
observed by MALDI-FTMS.
3422 Analytical Chemistry, Vol. 78, No. 10, May 15, 2006

optimizing the operational conditions of the instrument MALDI-
FTMS. For the instrument used in these studies, by optimizing
the higher mass region of the mass range, the sensitivity of the
lower mass region is reduced. Recently, it has been demonstrated
in our laboratory that by combining data collected using two
different sets of tuning conditions the dynamic range for the
analysis of a proteome can be improved.22 Another approach to
increasing the number of mass defect labeled peptides observed
is analyzing them by ESI-MS. It has been found in previous studies
that ESI is more favorable for the ionization and detection of
hydrophobic peptides than is MALDI.27,28 Therefore, more mass
defect labeled cysteine-containing peptides are expected to be
observed by using ESI compared to MALDI. This will be the
subject of future studies in our laboratory. Combining both MALDI
and ESI results could lead to gaining the most information possible
out of a particular sample due to their complementary nature.29
are analyzed simultaneously, which eliminates the need for
separation prior to analysis and allows the detection of proteins
that do not contain cysteine. Second, improvement in specificity
arises from the decongestion of the mass spectrum, meaning that
regions of the mass space that were previously unoccupied will
be populated by the labeled cysteine-containing peptides. Another
important advantage of using this approach constitutes the
identification of proteins usually missed by other methods; in this
case, 48 extra proteins were identified by adding a mass defect
tag to the cysteine-containing peptides, as this is found to improve
both their detectability and their identification specificity. In
addition, the analysis of these samples was performed by MALDI-
FTMS without requiring the use of tandem MS, which demands
the acquisition of much larger data sets and requires significantly
more computational analysis of the data. This approach can be
extended to the labeling of other amino acids that occur with lower
than average frequency, such as tryptophan or histidine, by using
labeling reactions that are specific for these amino acids. Such
work is currently under investigation in our laboratory.31
CONCLUSIONS
The method presented here provides a way to improve the
specificity of peptide identification based on accurate mass
measurement, which leads to an increase in the number of
proteins that can be identified in an organism with a small genome
(<5000 ORFs). This approach has several significant differences
from methods that use derivatives with affinity tags, such as ICAT
reagents.30 First, both unlabeled and mass defect labeled peptides
ACKNOWLEDGMENT
The authors thank Dr. Iris Porat and Prof. William B. Whitman
for the metabolically labeled M. maripaludis proteome. We
gratefully acknowledge generous financial support from the
National Institutes of Health (R01 RR019767-03) and the National
Science Foundation (CHE-0316002).
(27) Medzihradszky, K. F.; Leffler, H.; Baldwin, M. A.; Burlingame, A. L. J. Am.
Soc. Mass Spectrom. 2001, 12, 215-221.
(28) Bodnar, W. M.; Blackburn, R. K.; Krise, J. M.; Moseley, M. A. J. Am. Soc.
SUPPORTING INFORMATION AVAILABLE
Additional information as noted in text. This material is
available free of charge via the Internet at http://pubs.acs.org.
Mass Spectrom. 2003, 14, 971-979.
(29) Stapels, M. D.; Barofsky, D. F. Anal. Chem. 2004, 76, 5423-5430.
(30) Gygi, S. P.; Rist, B.; Griffin, T. J.; Eng, J.; Aebersold, R. J. Proteome Res.
2002, 1, 47-54.
Received for review January 6, 2006. Accepted March 23,
2006.
(31) Hernandez, H.; Li, C.; Niehauser, S.; Gawandi, V.; Phillips, R. S.; Amster, I.
J. San Antonio, TX, June 5-9 2005.
AC0600407
Analytical Chemistry, Vol. 78, No. 10, May 15, 2006 3423