Anal. Chem.

1994,66, 3723-3726

Protein Epitope Mapping By Mass Spectrometry

Ylngmlng Zhao and Brian T. Chalt’
The Rockefeller University, 1230 York Avenue, New York, New York 10021

A mass spectrometricmethod is described for the rapid mapping antigen bound to an antibody. Alternatively, these workers
of linear epitopes in proteins that are bound by monoclonal subjected the peptide to proteolytic digestion and identified
antibodies. The method consists of three steps. In the first products that bound to the immobilized antibody. In both
step, an antigen protein is digested by a proteolytic enzyme to cases, the peptides of interest were identified by 2 W f plasma
produce an appropriate set of peptide fragments. In the second desorption mass spectrometry.
step, peptide fragments containing the linearepitope are selected Matrix-assisted laser desorption mass spectrometry (MAL-
and separated from the pool of peptide fragments by immu- DI-MS) is a recently developed method for measuring the
noprecipitation with the monoclonal antibody. In the final molecular weights of peptides and proteins.12-14 The technique
step, the immunoprecipitated peptides are identified by matrix- allows the accurate (better than O.l%), rapid (<1 min), and
assisted laser desorption mass spectrometry. The method sensitive (<1 pmol) determination of the molecular weights
allows the rapid determination of antigenicsites without tedious of components of complex mixtures of peptides. MALDI-
peptide synthesis or protein mutagenesis. The approach is MS is finding wide use for the rapid identification of proteins
demonstratedthrough the mapping of epitopes in two peptides and the elucidation of their primary structures (in particular,
(melittin and glucagon-like peptide-1 7-37) against which the definition of posttranslational modifications.)
monoclonal antibodies were raised. In addition to epitope Here, we describe a method for the rapid mapping of linear
mapping, the successful couplingbetween matrix-assistedlaser protein epitopes that are bound by monoclonal antibodies.
desorption mass spectrometry and immunoprecipitationpro- The method, which takes advantage of the purification power
vides a potentially powerful tool for determining binding sites of immunoprecipitation and the high specificity and speed of
between proteins. MALDI-MS, consists of three steps. In the first step, an
A monoclonal antibody (mab) raised against a protein binds antigen protein is digested by a proteolytic enzyme to produce
only to a specific region of the protein, which is called the an appropriate set of peptide fragments. In the second step,
antigenic site or epitope. An antibody can bind either a linear peptide fragments containing the linear epitope are selected
(continuous) epitope or a nonlinear (conformational) epit0pe.l” from the pool of peptide fragments by immunoprecipitation
A linear epitope contains a stretch of contiguous amino acids with the monoclonal antibody. In the final step, the immu-
(typically 5-10 amino acid residues) in the antigen, whereas noprecipitated peptides are identified by MALDI-MS and
a nonlinear epitope is composed of residues that can be distant the antibody-binding peptide region is determined.
in the primary sequence but closein space in the folded protein.
A knowledge of the binding site of an antigen protein to its EXPERIMENTAL SECTION
antibody can improve the utility of the antibody. Materials. Melittin was purchased from the Sigma
Techniques that have been used for antigenic site mapping Chemical Co. (St. Louis, MO) and used without further
of linear epitopes include binding assays of sets of synthetic purification. Sequence-grade chymotrypsin, endoprotease
peptides that span the protein4J and constructs produced by Lys-C, and serine protease inhibitor Pefabloc Sc were
recombinant gene expre~sion.~.~ More recently, epitope purchased from Boehringer Mannheim Biochemical (India-
localizationhas been achieved through the useof bacteriophage napolis, IN), Protein G/protein A agarose was obtained from
peptide librariesa9-’ Although effective, these methods can Oncogene Science (Uniondale, NY). Antimelittin monoclonal
be costly and time-consuming. A different approach to antibody No. 83144 is mouse IgGl subtype antibody gener-
antigenic site mapping has been reported by Przybylski and ously provided by Dr. T. P. King of the Rockefeller University.
co-workers,8who compared the pattern of proteolyticdigestion Glucagon-like peptide-1 7-37 (GLP- 1 7-37) and antiGLP-1
of free peptide antigen with the pattern produced from the 7-37 No. 26.1 mouse monoclonal antibody were kindly
(1) Jin, L.; Fendly, B. M.; Wells, J. A. J . Mol. Biol. 1992, 226, 851-865. provided by Dr. Douglas Buckley of Scios Nova (Mountain
(2) Lane, D. P.; Stephen, C. W. Curr. Opin. Immunol. 1993, 5, 268-271. View, CA).
(3) Harlow, E.; Lane D. Antibodies. u luborutory munuul; Cold Spring Labora-
tory: Cold Spring Harbor, NY, 1988; pp 23-35. Digestion of Peptides. A 1:lOO ratio (w/w) of protease
(4) Wang, Z.; Laursen, R. A. Pep?. Res. 1992, 5, 275-280. and peptide was dissolved in an appropriate buffer (100 mM
( 5 ) Geysen, H. M.; Meloen, R. H.; Barteling, S. J. Proc. Nutl. Acud. Sci. U.S.A.
1984,81, 39984002. Tris-HC1, 10 mM CaC12, pH 8.0 for chymotrypsin; 50 mM
(6) Dias, P.; Parham, D. M.; Shapiro, D. N.; Tapswtt, S. J.; Houghton, P. J. Tris-HC1, 1 mM EDTA, pH 8.0 for endoprotease Lys-C)
Cancer Res. 1992, 52, 6431-6439.
(7) Chen, J.; Marechal, V.; Levine, A. J. Mol. Cell. B i d . 1993,13,4107-4114. with peptide concentration between 10and 20 pM. To obtain
(8) Suckau, D.; Kohl, J.; Karwath, G.; Schneider, K., Casaretto, M.; Bitter- a partial digest of the peptide, the mixture was maintained
Suermann, D.; Przybylski, M . Proc. Nafl.Acud. Sci. U S A . 1990,87,9848-
9852.
(9) Devlin, J. J.; Panganiban, L. C.; Devlin, P. E. Science 1990, 249, 404-406. (12) Hillenkamp, F.; Karas, M.; Beavis, R. C.; Chait, B. T. A n d . Chem. 1991,63,
(10) Scott, J. K.; Smith, G. P. Science 1990, 249, 386-390. 1193-1203.
(11) Cwirl, S. E.; Peters, E. A.; Barrett, R. W.; Dower, W. J. Proc. Nufl. Acud. (13) Beavis, R. C.; Chait, B. T. Proc. Nutl. Acud. Sci. U.S.A. 1990,87,6873-6877.
Sci. U S A . 1990,87, 6378-6382. (14) Chait, B. T.; Kent, S. B. H. Science 1992, 257, 1885-1894.

QQQ3-27QQl94IQ366-3723~Q4.5QlQ AnaEytical Chemistry, Vol. 66, No. 21, November 1, 1994 3723
0 1994 American Chemical Society
at 37 OC for 5-10 min. Alternatively, to obtain a complete
digest of the peptide, the peptide was mixed with the
appropriate protease in a 1:30 ratio (w/w) and digested at 37
OC for 2 h. The digestions were terminated by addition of a
one-tenth volume of 10 mM Pefablock Sc solution (25 "C; 10
min), followed by heating at 90 OC for 15 min.
Immunoprecipitati~n.~~ Monoclonal antibody (2-1 0 pg)
and the mixture of peptides produced by the proteolytic
digestion ( 10-50 pmol) were mixed in 60 pL of 75 mM Tris-
HC1,200 mM NaC1,O. 1-0.5%n-octylglucoside, pH 8.0 (TSO)
solution. After a 2 h incubation at 4 "C with gentle stirring,
2-3 pL of protein G/protein A agarose was added to the
solution and incubated for another 0.5 h at 4 OC. The agarose
A
beads were collected by carefully aspirating the supernatant
after centrifugation of the solution for 1 min at 16000g. The
beads were washed three times with 200 pL of TSO buffer I
MALDI~MSAnatysis stepin
and then three times with 200 pL 10 mM Tris-HCl,200 mM
NaC1,5 mM 8-mercaptoethanol, pH 8.0 (TSM). The laser Flgwe 1. Three-step strategy for linear epitope mapping.
desorption matrix (4 pL of a saturated solution of a-cyano- 1 10 20 26
4-hydroxycinnamic acid in 1% aqueous TFA/ACN (2: 1, (v/ GIGAV LKVLT TGLPA LlSWl KRKRQ Q
v)) and a suitable amount of internal standard peptide Lys-c
8 26
1 26
(dynorphin A 1-13) were mixed with the washed beads. Digestion
Finally, 1 pL of the matrix/agarose bead mixture containing --
I ! . . .-. -. -. -23
the bound peptides was loaded onto the mass spectrometer -
8 .-. -. -. -
23

probe and dried at room temperature with a stream of air for

MALDI-MS analysis. C-trypsin 20 26

Mass Spectrometry. MALDI-MS analysis was carried Digestion 7

-.-.-.- 19
out on a laser desorption time-of-flight instrument constructed 10
-.-.- 19
at the Rockefeller University and described elsewhere.16The Figure 2. Amino acM sequence of mellttin. The peptide fragments
entire triple complexconsisting of proteinG/protein A agarose, produced by endoprotease Lys-Cand chymotrypsindigestionare shown
by lines. The solid lines represent those peptide fragments that were
antibody, and the bound peptide was subjected to mass bound by the antimelittin mab No. 83144 and the dashed lines those
spectrometric analysis. Intense peaks corresponding to the that were not bound by mab No. 83144.
peptides that were bound to the antibodies were observed in
the mass spectra because the MALDI-MS conditions used in Epitope mapping is carried out in three steps (Figure 1).
the present experiments cause the peptide to dissociate from Step I: The antigen peptide is digested by a highly specific
the antibody. This dissociation may occur in the acidified
protease to produce a mixture of component peptide fragments.
matrix and/or during the MALDI process. The mass spectra
The identities of these peptide fragments can be determined
were collected by adding individual spectra obtained from
by accurate measurement of their molecular weights using
200 laser shots to improve the statistics of the measurement. MALDI-MS. Step 2: Peptide fragments that contain the
The spectra were initially calibrated using dynorphin A 1-13
antigenic epitope are purified from the peptide fragment
and oxidized insulin B-chain. Subsequent calibrations were
mixture by immunoprecipitationwith the monoclonalantibody
made using either the undigested parent peptide and dynorphin
against the antigen peptide. In this step, peptides that contain
A 1-13 or an external calibrant. the epitope, form complexes with the antibody-proteinG/
RESULTS AND DISCUSSION protein A agarose. Other peptides fragments in the mixture
Bee venom melittin and GLP- 1 7-37 were selected as model that do not bind to the antibody remain in solution and can
systems for this study because monoclonal antibodies were be removed by washing. Step 3: The molecular weights of
available that bind known linear regions in these peptides. the peptides that are bound by the antibody are determined
Antimelittin monoclonal antibody No. 83 144 was previously by MALDI-MS. The identity of the antibody-bindingpeptides
determined to bind to an epitope located at residues 20-26 of are readily assigned by their molecular weights and the known
melittin,17J8and antiGLP- 1 7-37 monoclonal antibody No. digestion sites of the specific protease.
26.1 was previously determined (by an investigation of the Epitope Mapping of Melittin. Melittin is a 26 amino acid
binding of a series of synthetic peptide analogs) to bind a residue peptide isolated from bee venom (Figure 2). Partial
region that included the first three N-terminal residues of digestion of melittin by endoprotease Lys-C yielded four
GLP-1 7-37 (personal communication from Dr. Douglas peptide fragments that gave intense mass spectral peaks. The
Buckley, Scios Nova, Mountain View, CA). molecular weights of these peptides were found to correspond
(within <1 Da; seeTable 1) to the molecular weights predicted
(1 5) Harlow, E.; Lane D. Antibodies. u luborotory munuul; Cold Spring Labora- for melittin fragments 8-23, 8-26, 1-23, and 1-26 (Figure
tory: Cold Spring Harbor, NY. 1988; pp 421470.
(16) Beavis, R. C.; Chait, B. T. Anal. Chern. 1990,62, 1836-1840. 3A). The other two peaks in the spectrum (designated it and
(1 7) King, T. P.; Kochoumian, L.;Joslyn. A. J . Immunol. 1984,133,2668-2673. i2 in Figure 3A) arise from unidentified impurities that were
(18) Fehlner, P. F.; Berg, R. H.;Tam, J. P.; King, T. P. J. Immunol. 1991, 146,
799-803. present in the melittin sample. After immunoprecipitation

3724 Anatytlcal Chemistry, Vol. 66, No. 21, November 1, 1994

Table 1. Measured and Calculated Values of the Molecular
WeIgMs of Peptide Fragments Produced by PrOtoteolytlc Mg.rtlon i 7-1 9 Melittin
chymotrypsin digestion
of Melittin
peptide fragment calcd MW measd MW A mass (Da) ! i

8-23
Endoprotease Lys-C Digestion
1796.2 1796.1 -0.1
J I 1
8-26 2208.7 2208.0 -0.7
1-23 2435.1 2435.0 -0.1 .-
0
v)
Chymotrypsin Digestion 8 I I I I
I
20-26 956.2 956.0 -0.2 5
.-4

1000 1500 2000 2500 3000

10-19 1058.3 1058.2 -0.1 e,
7-19 1398.7 1398.9 +0.2

- Melittin
Lys-c Digestion 1.23 (A) 1 After immunoprecipitation

I
I I I I '
1000 1500 2000 2500 3000
mh
Flgure 4. Matrixassistedlaser desorptionmass spectra of (A) peptide
- fragments producedby chymotrypslndlpstkm of melittin and (B) peptkie
2 After Immunoprecipitation 1-26 (B) fragments Isolated by immunoprecipitation with antimelittin mab No.
- 83144. The peak labeledwith an asterisk correspondsto the internally
- added caiibrantdynorphin A 1-13, and I refersto an unldentlfted impurity.

Upon immunoprecipitation with mab No. 83 144, fragment

20-26 was selected by the antibody (Figure 4B) while
fragments 7-19 and 10-19 were almost completely washed
I I 1 I I ' away (i.e., the intensity of fragment 7-1 9 relative to fragment
1000 1500 2000 2500 3000
20-26 in Figure 4B was reduced by a factor of 34 compared
mlz to the ratio of peak intensities observed in the original peptide
Flgure 3. Matrix-assisted laser desorptionmass spectra of (A) peptide mixture (Figure 4A)). These findings are summarized in the
fragments produced by endoprotease Lys-C digestion of melittin and
(6)peptide fragments isolated by immunoprecipitationwith mab No. lower portion of Figure 2.
83144. The peak labeledwith an asterisk correspondsto the internally Inspection of the combined data (Figure 2) obtained from
added calibrant dynorphin A 1-13, and 1, and 12 refer to unidentified the experiments shown in Figures 3 and 4 demonstrates that
impurities.
the region encompassing residues 20-26 is sufficient for mab
No. 83144 binding. The present result is in agreement with
with antimelittin mab No. 83144, the two impurities and a previous determination of the binding epitope that used a
peptide fragments 1-23 and 8-23 were washed away, while competitive binding assay of chromatographically separated
fragments 1-26 and 8-26 were identified in the immuno- peptic and tryptic fragments of melittin.17J8 In this earlier
precipitation complex (Figure 3B). As a control, we incubated study, the binding affinity of mab No. 83 144 was determined
the set of melittin peptide fragments with an unrelated anti- to be approximately 10-6 M,17J8which is weaker than that
human basic fibroblast growth factor monoclonal antibody of most commonly used antibodies (typical range 10-7-10-10
and treated the resulting solution in a manner identical to M). Thus the method should be applicable to the mapping
that described above. No significant peptide peaks was found of linear epitopes for most antibodies.
in the mass spectrum of this antibody complex (data not Epitope Mapping of GLP-1 7-37. As a further test of the
shown). The experiment demonstrated that peptide selection present methodology, we mapped the antigenic site of GLP-1
is specified by the antibody used in the immunoprecipitation. 7-37 for antiGLP-17-37 mab No. 26.1. The binding affinity
The results obtained in Figure 3 (summarized diagra- of this antibody is much higher than that of mab No. 83144
matically in the upper portion of Figure 2) indicate that described above. GLP-1 7-37 was subjected to partial
residues 1-7 are not required for antibody binding but that digestion by chymotrypsin. Five major peaks were identified
one or more of residues 24-26 is required. To further define in the mass spectrum, which correspond to GLP-1 7-37
the antigenic site, the melittin sample was digested by fragments 1-13, 1-22, 1-25, 1-31, and 14-22 (Figure 5A).
chymotrypsin. Three peptides were produced, which cor- After immunoprecipitation with antiGLP-1 7-37 mab No.
respond to melittin fragments 20-26,7-19, and 10-1 9 (Table 26.1, fragment 14-22 and the impurity were washed away,
1 and Figure 4A), as well as an unidentified impurity (Figure while fragments 1-13,l-22,l-25, and 1-3 1 remained bound
4A). The peptide fragment 1-7 was not observed in the mass to the antibody (Figure 5B). A summary of the binding
spectrum. In this regard, we note that MALDI-MS sometimes patterns of the chymotryptic peptides (Figure 5C) shows that
does not produce a detectable response for small peptides. the antigenic epitope resides within residues 1-13 of GLP-1

Analytical Chemistry, Vol. 66, No. 21, November 1, 1994 3725

 GLP-I 7-37 (A) 1 applicable to most antibodies that bind linear epitopes. On
Chymotrypsin digestion I
occasion, weak peaks were observed in the mass spectra that
correspond to peptides that bind nonspecifically to protein
Alprotein G agarose or antibody. These nonspecifically
binding peptides could be readily excluded by using harsher
washing conditions for removing them or by comparing the
relative intensities of the peptide peaks in the mass spectra
cc
$ 1 I 1 I I I 1 I
prior to and after immunoprecipitation.
9 1000 1500 2000 2500 3000 3500 The present procedure differs from that of Przybylski and
.-cs! co-workers*in three respects. First, MALDI-MS is consider-
2 After Immunoprecipitation (B) ably more sensitive and less prone to discrimination against
certain peptide components of proteolytic digests than 252Cf
plasma desorption MS. Thus, much less antigen and antibody
is necessary in the present experiment. Second, the detergent
octyl g l u c ~ s i d is
e ~used
~ during immunoprecipitation, which
appears to reduce potentially confusing nonspecific binding
of peptide to antibody. Third, the antibody-antigen complex
(on the agarose beads) is loaded together with the laser
1000 1500 2000 2500 3000 3500 desorption matrix onto the mass spectrometer probe, without
separation of the bound peptides.20 The present procedure
also avoids the addition of high concentrations of salt for eluting
1 10 20 30
the peptides.8
HAEGT FTSDV SSYLE GQAAK EFIAW LVKGR G The present approach to linear epitope mapping is
1 13 considerably more rapid than the widely used conventional
Chymotrypsin 1 72
a p p r o a c h e ~ . ~The
~ ~ Jantigen
~ protein can be digested by a
Digestion 1 25
variety of specific proteases, which can rapidly produce a large
number of peptide fragments. Thus the digestion step is
comparable to fast peptide synthesis or recombinant gene
Flgure5. Matrlx-assisted laser desorption mass spectra of (A) peptide
fragments produced by chymotrypsin digestion of GLP-1 7-37 and (B) expression of deletion mutants. The digestion and immuno-
peptide fragmentsisolated by immunoprecipitationwith antlGLP-17-37 precipitation steps take approximately 6 h and the molecular
mab No. 26.1. The peak labeled with an asterisk corresponds to the weight determination of each sample takes 1 min. Thus, the
internaiiyaddedcalibrantdynorphlnA 1-13,and ireferstoan unidentified
Impurity. The satellite peaks labeled a, and a2arise by adventitious Na epitope for a specific antibody can be located in a short stretch
and Cu adduction to the peptide.22(C). Amino acid sequence of GLP-1 of the protein (typically 6-15 amino acids, depending on the
7-37. The peptide fragments produced by chymotrypsin digestion are available digestion sites in the antigen) in a single day. If
shown by lines. The solid lines represent those peptide fragments that
were bound by the antlGLP-1 7-37 mab No. 26.1 and the dashed lines
I
more detailed localization of the epitope is required, this
those that were not bound by mab No. 26.1. information can be obtained by the synthesis of a small set
of peptides that span the region of interest determined by the
present method.21
7-37. The present result is consistent with the previous finding The present method represents a successful marriage
(personal communication from Douglas Buckley, Scios Nova, between a powerful purification technique (immunoprecipi-
Mountain View, CA) that the epitope includes the first three tation) and an effective characterizationtechnique (MALDI-
N-terminal residues of GLP-1 7-37. MS). We believe that it will be useful for a range of other
biomedical studies such as antigenic site mapping for polyclonal
CONCLUSION antibodies and determination of the sites of protein-protein
The present results demonstrate the feasibility of a method or protein-ligand interactions.
for rapidly mapping linear protein epitopes that combines
immunoprecipitation purification of antibody-binding peptide ACKNOWLEDGMENT
fragments with accurate determination of their molecular We thank Dr. T. P. King of The Rockefeller University
weights by mass spectrometry. Selection and analysis of and Dr. Douglas Buckley of Scios Nova Inc., Mountain View,
antibody-binding peptides provide the basis for localizing the CA, for supplying us with monoclonal antibody and peptide
antigenic sites. The analysis relies on a determination of the used in this study and Drs. T. P. King and Peter Model for
ratioof the relative intensities of thevarious peptide fragments helpful discussions. Y. Z. also thanks Drs. Wenzhu Zhang,
prior to and after immunoprecipitation. The method was Rong Wang, Klaus Schneider, Steven Cohen, James Zhou,
successfully demonstrated for an antibody with a relatively David Fenyo, and Jun Qin for technical help and Zhong Zhong
low binding affinity (loa M) and therefore should be for useful suggestions. Support for this work from theNationa1
Institutes for Health (Grants RR00862 and GM38274) is
(19) Vorm, 0.;Chait, B. T.; Roepstorff, P. In Proceeding41st ASMS Conference
on Mass spectrometry and Allied Topics, 1993; p 621. gratefully acknowledged.
(20) Hutchens, T. W.; Yip, T.-T.Rapid Commun. Mass Spectrom. 1993, 7,576-
580.
(21) Manuscript in preparation. Received for review April 26, 1994. Accepted August 9, 1994."
(22) Beavis, R . C.; Chaudary, T.;Chait, B. T. Org. Mass Spectrom. 1992, 27,
156-158. Abstract published in Aduance ACS Abstracrs, September 15, 1994.

3726 Analytical Chemistty, Vol. 66, No. 21, November 1, 1994