Methodology for determining disulfide linkage patterns of closely spaced cysteine residues

Article abstract of DOI:10.1021/ac901161e

We report the development and application of a method for determining bonding patterns in disulfide-linked peptides containing closely spaced cysteine residues. Through the utility of classic N-terminal sequencing chemistry coupled with facile liquid chro

Full text of DOI:10.1021/ac901161e

Anal. Chem. 2009, 81, 7314–7320
Methodology for Determining Disulfide Linkage
Patterns of Closely Spaced Cysteine Residues
Bing Zhang and Steven L. Cockrill*
Analytical Sciences, Amgen, Inc., 4000 Nelson Road, Longmont, Colorado 80503
We report the development and application of a method for
determining bonding patterns in disulfide-linked peptides
containing closely spaced cysteine residues. Through the
utility of classic N-terminal sequencing chemistry coupled
with facile liquid chromatography and mass spectrometric
analysis of the cleavage products, we report the ability to
demonstrate unambiguous assignment of paired cysteine
residues, using human insulin as a model protein. The
conditions of the technique were selected and optimized to
maintain disulfide integrity. In a forthcoming article, we will
present the results of this method as applied to the complete
elucidation of linkages in disulfide variants of a therapeutic
monoclonal antibody of the IgG2 subclass.
Such instances represent significant challenges to elucidation of
the disulfide structure.
The contemporary literature contains several methods that may
provide information for disulfide linkage analysis. Partial reduction
and derivatization of the nascent free thiol group has been applied
for determination of disulfide connectivity in a variety of proteins,
including highly knotted substrates3 and those with closely spaced
cysteines.4 However, for targeted analysis, this approach generally
requires the ability to selectively reduce specific disulfide linkages.
This may involve extensive optimization of the reduction conditions
to afford the selective generation of isomers containing specific
linkage reduction. Such optimization steps are often necessarily time-
intensive, may consume significant quantities of a potentially limited
sample, and are even more challenging when multiple similar
linkages are present in a peptide, each with equivalent susceptibility
to reduction. Additionally, the required denaturation of the protein
prior to the partial reduction step may itself lead to disulfide
scrambling.5 While the data output is readily interpreted, the utility
of multiple steps for each cycle of processing and identification
typically requires several hours of manipulation and analysis.
Automated Edman sequencing has also been utilized for deter-
mining linkage assignments through the detection of di-PTH-cystine.6
Indeed, this approach was used to demonstrate the presence of the
“ladder” structure of cysteine bonding in the hinge region of
therapeutic antibodies of both the IgG46 and IgG27 subclasses.
However, this technique has limitations with respect to the inability
to detect nonparallel linkages, and quantitative distinction between
multiple disulfide variants present in a mixture cannot be made.
Furthermore, issues with cycling efficiency and carryover are of
concern, particularly with closely spaced cysteine residues.
Complete elucidation of the primary structural features of a
protein of interest encompasses a broad array of analytical
methods to assess specific and often separate structural features,
including post-translational modifications such as glycosylation,
processing of terminal variants, and pairing of cysteine residues
to form disulfide bonds. The disulfide linkage pattern is a critical
factor in determining the protein’s conformational properties and
may therefore impart significant impact to the biological function
through the availability of active site(s).1,2
Typically, because of the large size of most proteins, cleavage
of the polypeptide backbone using an appropriate endoprotease
is necessary for characterization of the disulfide linkages. By
comparison of the peptides under nonreducing and reducing
conditions, it is possible to identify the constituent species of a
disulfide-linked peptide. If these species contain only a single
cysteine residue each, then elucidation of the linkage pattern is
simply achieved by identification of those constituent peptides.
For species containing multiple disulfide linkages, additional
cleavage between sequential cysteine residues is required; it is
normally reasonable to affect cleavage between disulfide-linked
cysteine residues due to the broad array of endoproteases and
their associated specificities. However, the amino acid sequence
of the protein of interest may contain closely spaced cysteine
residues with no suitable intermediate cleavage site or even pairs
of cysteine residues that are adjacent in the amino acid sequence.
Recently mass spectrometry-based approaches have been
applied to disulfide characterization.6,8,9 Techniques employing
mass spectrometric detection and analysis are very attractive
options due to the fact that there is typically no need for sample
preparation and manipulation beyond standard digestion and
separation (i.e., peptide mapping). This approach has been
successfully employed in the analysis of various protein substrates,
including therapeutic antibodies.8,9 However, the data presented
in these publications did not afford specific linkage determination
between the four cysteine residues in the dimeric hinge peptide;
it was not possible to differentiate the possible parallel or “crossed”
* Corresponding author. Steven L. Cockrill, e-mail: steven.cockrill@amgen.com.
Fax: (303) 401-4403.
(1) Flynn, J. P.; Pallaghy, P. K.; Lew, M. J.; Murphy, R.; Angus, J. A.; Norton,
(3) Qi, J.; Wu, J.; Somkuti, G. A.; Watson, J. T. Biochemistry 2001, 40, 4531–
4538
.
(4) Martinez, T.; Guo, A.; Allen, M. J.; Han, M.; Pace, D.; Jones, J.; Gillespie,
R. S. Biochim. Biophys. Acta 1999, 1434, 177–190
(2) Dillon, T. M.; Ricci, M. S.; Vezina, C.; Flynn, G. C.; Liu, Y. D.; Rehder,
.
R.; Ketchem, R. R.; Zhang, Y.; Balland, A. Biochemistry 2008, 47, 7496–
7508
(5) Yen, T.-Y.; Joshi, R. K.; Yan, H.; Seto, N. O. L.; Palcic, M. M.; Macher,
B. A. J. Mass Spectrom. 2000, 35, 990–1002
.
D. S.; Plant, M.; Henkle, B.; Li, Y.; Deechongkit, S.; Varnum, B.; Wypych,
J.; Balland, A.; Bondarenko, P. V. J. Biol. Chem. 2008, 283, 16206–16215
7314 Analytical Chemistry, Vol. 81, No. 17, September 1, 2009
.
.
10.1021/ac901161e CCC: $40.75  2009 American Chemical Society
Published on Web 07/23/2009

linkage patterns, i.e., bonds between Cys-229 and Cys-229 and
Cys-232 and Cys-232 (parallel) or Cys-229 and Cys-232 and Cys-
232 and Cys-229 (crossed). No fragmentation was observed
between the closely spaced cysteines of each peptide chain, even
with the adjacent proline residues.9
(i.e., pH 6.5) for digestion conditions has been utilized with specific
intent of assessing disulfide linkages.5,12
Literature reports also show that the expected disulfide structure
of insulin may be perturbed under controlled redox conditions,
resulting in scrambling of the linkages. Fortuitously, these disulfide
variants are readily resolved by chromatographic means.13
In this paper, we describe a novel approach that provides facile
analysis and confirmation of the connectivity of disulfide linkages
using insulin as a model protein. The applications of this
methodology are numerous but perhaps most relevant to the
recently discovered disulfide-mediated structural variants of im-
munoglobulin G2.7
An example application of particular interest is in the hinge
region of therapeutic antibody products, particularly those of the
IgG2 subclass. Determination of the disulfide structure is of
importance due to increasing expectations from regulatory agen-
cies for a more complete understanding of a therapeutic molecule
but also with potential concerns over product safety and efficacy.
For example, recent publications presented the identification of
previously undetermined disulfide variants.2,7 While novel disulfide
mediated structural variants were identified, the specific linkage
patterns between the Fab arm and hinge species were not
determined. The MS-based methodologies described above
face significant challenges in achieving elucidation of the disul-
fide bonding patterns of such complex structures, given the large
number of cysteine residues (up to 16) present in the signature
nonreduced Lys-C peptides characteristic of each structural
variant.7 In a forthcoming paper, we will present complete
elucidation of the disulfide structure through the application of a
novel technique to the analysis of the disulfide-linked peptides
unique to each variant.
MATERIALS AND METHODS
Materials. Recombinant human Insulin was sourced from
Invitrogen (Carlsbad, CA), and insulin oxidized B-chain was from
Sigma-Aldrich (St. Louis, MO). Endoproteinase Glu-C (protease
V 8, Staphylococcus aureus V 8), sequencing grade was purchased
from Roche (Indianapolis, IN). Reagents for manual execution of
Edman sequencing including trifluoroacetic acid (TFA), phenyl-
isothiocyanate (PITC), and N-methylpiperidine/water/methanol
solution were sequencing grade materials obtained from Applied
Biosystems (Foster City, CA). Pyridine was ACS reagent grade
from Fluka (Buchs, Switzerland). Preprepared mobile phases (0.1%
TFA in water and 0.1% TFA in acetonitrile), as well as HPLC-grade
water, were from J.T. Baker, (Phillipsburg, NJ).
In the current manuscript, we describe development of the
novel methodology, with application to insulin as a model protein.
Insulin was selected due to its convenient size, presentation of
three disulfides (two interchain, one intrachain), combination of
adjacent and widely spaced cysteine residues, and a long,
established history of characterization. The interchain disulfide
linkages are between residues Cys-7 and Cys-7 and Cys-20 and
Cys-19 of the A and B chains, respectively. The intrachain disulfide
exists between Cys-6 and Cys-11 of the A chain.10 This structure
of insulin is depicted in the Supporting Information for reference.
Insulin contains both widely spaced, interchain disulfide
linkages with facile sites for proteolytic cleavage by Glu-C on both
chains, as well as adjacent cysteine residues involved in separate
disulfide bonds. As such, insulin is representative of an analytically
challenging substrate that, conveniently, has been extensively
characterized. Treatment of the model protein with endoprotease
Glu-C following treatment with alkylating reagent under denatur-
ing conditions simulates the typical proteolytic processing steps
necessary for generation of peptides appropriate for efficient
sequential release. The utility of an alkylating reagent such as
N-ethylmaleimide (NEM) is employed to “scavenge” free sulfhy-
dryl and thereby inherently prevent disulfide scrambling at the
elevated temperatures and pH ranges typical of most sample
preparation processes.5 Additionally, the use of lower pH buffers
Tris solution (1 M, pH 8.0) was from Calbiochem (La Jolla,
CA), tris(2-carboxyethyl)phosphine (TCEP) was purchased from
Pierce (Rockford, IL), and guanidine hydrochloride solution (8
M), iodoacetic acid (IAA), and N-ethylmaleimide (NEM) were
obtained from Sigma-Aldrich (St. Louis, MO). NAP-5 Sephadex
G-25 columns were sourced from Pharmacia Biotech (Little
Chalfont, U.K.).
Chromatographic Analysis of Commercial Human Insulin.
Commercial human insulin was assessed by reverse-phase HPLC
in order to ascertain the potential presence of multiple disulfide
structures. A 1 µg load of insulin (1 mg/mL in 2 M Gdn HCl, 100
mM Tris, pH 6.5) was applied to a reverse-phase column (Agilent
Zorbax 300SB C8, 2.0 mm × 150 mm, 5 µm particle size) and
separated using mobile phases consisting of (A) 20% acetonitrile
in 200 mM ammonium sulfate and 50 mM sulfuric acid and (B)
40% acetonitrile in 200 mM ammonium sulfate and 50 mM sulfuric
acid. Elution was accomplished using a 30 min gradient from 10%
to 50% mobile phase B at a temperature of 40 °C and a flow rate
of 0.2 mL/min, as reported previously.11
Enzymatic Digestion with Endoprotease Glu-C and Map-
ping of Insulin Peptides. Insulin was reconstituted in 400 µL of
denaturation buffer (6 M guanidine HCl, 200 mM Tris, pH 6.5)
and then treated with 100 µL HPLC grade water and 12 µL of 0.5
M iodoacetic acid for a final protein concentration of 2 mg/mL.
The sample was incubated at room temperature in the dark for
30 min, then buffer exchanged to 50 mM Tris, pH 6.5 using an
equilibrated NAP-5 Sephadex G-25 column, with a resulting protein
concentration of 1 mg/mL.
(6) Wei, Z.; Marzilli, L. A.; Rouse, J. C.; Czupryn, M. J. Anal. Biochem. 2002,
311, 1–9
.
(7) Wypych, J.; Li, M.; Guo, A.; Zhang, Z.; Martinez, T.; Allen, M. J.; Fodor, S.;
Kelner, D. N.; Flynn, G. C.; Liu, Y. D.; Bondarenko, P. V.; Ricci, M. S.;
Dillon, T. M.; Balland, A. J. Biol. Chem. 2008, 283, 16194–16205
(8) Chelius, D.; Huff Wimer, M. E.; Bondarenko, P. V. J. Am. Soc. Mass
Spectrom. 2006, 17, 1590–1598
(9) Wu, S.-L.; Jiang, H.; Lu, Q.; Hancock, W. S.; Karger, B. L. Anal. Chem.
2009, 81, 112–122
(10) Chang, S.-G.; Choi, K.-D.; Jang, S.-H.; Shin, H.-C. Mol. Cells 2003, 16, 323–
330
(11) Hua, Q.-X.; Jia, W.; Frank, B. H.; Phillips, N. F. B.; Weiss, M. A. Biochemistry
2002, 41, 14700–14715
.
.
Digestion using endoprotease Glu-C was accomplished by
incubating 120 µL of 1 mg/mL insulin solution (in 50 mM Tris
.
(12) Foley, S. F.; Sun, Y.; Zheng, T. S.; Wen, D. Anal. Biochem. 2008, 377,
.
95–104
(13) Forsberg, G.; Palm, G.; Ekeracke, A.; Josephson, S.; Hartmanis, M. Biochem.
J. 1990, 271, 357–363
.
.
.
Analytical Chemistry, Vol. 81, No. 17, September 1, 2009 7315

Table 1. Putative Cleavage Products of Insulin Digestion by Endoprotease Glu-C under Nonreducing Conditionsa
peptide
structure/identity
theoretical mass (Da)
observed mass (Da)
mass error (ppm)
A1
GIVE
416.23
416.23
0
3
0
7
7
0
0
0
A2(s-s)B1
QCCTSICSLYQLE(s-s)FVNQHLCGSHLVE
QCCTSICSLYQLE
FVNQHLCGSHLVE
NYCN(s-s)ALYLVCGE
NYCN
2967.33
1489.63
1481.71
1376.59
512.17
2967.32
1489.63
1481.72
1376.58
512.17
A2
B1
A3(s-s)B2
A3
B2
B3
ALYLVCGE
RGFFYTPKT
866.42
1115.58
866.42
1115.58
a Masses are monoisotopic.
buffer, pH 6.5) with Glu-C (substrate to enzyme ratio 3:1)
overnight at room temperature in foil-wrapped sample tubes.
Digestion was quenched by addition of 12 µL of 10% TFA in water.
Separation of the resulting peptides was accomplished using
a 1200 HPLC system (Agilent, Santa Clara, CA) equipped with a
binary pump. A reverse-phase column (Agilent Zorbax 300SB C8,
2.1 mm × 150 mm, 5 µm particle size) was employed with mobile
phases consisting of (A) 0.1% TFA in water and (B) 0.1% TFA in
acetonitrile. A linear gradient between 2 - 50% mobile phase B
was used, following a 5 min system equilibration step at 2% B,
with a flow rate of 0.2 mL/min. Reduced mapping was performed
under identical conditions following treatment of the digest
products with 2 µL of Tris(2-carboxyethyl)phosphine (TCEP) for
10 min at room temperature.
Conversion Reaction. The solution resulting from the cleav-
age step detailed above was cooled and pooled by centrifuging
for 2 min at room temperature. A 15 µL aliquot of HPLC-grade
water was added to reach a final TFA concentration of 25%, after
which the solution was incubated at 64 °C for 10 min. After
cooling, a portion of the samples was loaded directly onto the
LC-MS system for analysis. Note that these conditions were
designed to be identical to those of the automated Procise 494
instrument (Applied Biosystems, Foster City, CA).
Iterative Coupling/Cleavage Cycles. For sequential cycles,
the residual substrate following a given alkylation/coupling/
cleavage/conversion cycle was collected from the chromato-
graphic separation, dried by vacuum centrifugation, and then
subjected to repetition of the alkylation, coupling, cleavage, and
conversion steps followed by LC-MS analysis as described above.
For LC-MS analysis, the outlet of the HPLC separation was
coupled to a MSD-TOF ESI-MS instrument (Agilent, Santa Clara,
CA) running in positive ion mode. Source settings were a scan
range of 200 - 3000 m/z, fragmentor, skimmer, octopole rf, and
capillary voltages of 200, 60, 250, and 4 000 V, respectively. Drying
gas was supplied at 10.0 L/min and 300 °C. Nebulizer pressure
was 40 psig. For fraction collection, the MS instrument was
bypassed, and the HPLC eluent was collected manually and dried
by vacuum centrifugation prior to further processing and analysis.
Coupling Reaction. Derivatization of the primary amines was
accomplished using phenyisothiocyanate (PITC) as traditional
Edman degradation chemistry.14 Briefly, the peptides resulting
from Glu-C digestion of 4 nmol of insulin and isolated by peptide
mapping as described above were reconstituted in 10 µL of 10
mM N-ethylmaleimide NEM in water, then incubated at room
temperature in the dark for 30 min. A 40 µL volume of anhydrous
pyridine, 5 µL of PITC, and a 5 µL volume of N-methylpiperidine/
water/methanol solution were added under dry nitrogen. Coupling
was achieved by incubation at various temperatures (room
temperature, 50 °C) for various times (3-20 min). The solution
was then dried completely by vacuum centrifugation. Complete-
ness of coupling was assessed by resolubilization in 30% pyridine
in water and analysis by LC-MS, using separation and detection
conditions as described for execution of the peptide map.
Cleavage Reaction. Cleavage of the derivatized substrate was
accomplished by acid treatment using anhydrous TFA. Various
modes were assessed, including direct addition of a 10 µL aliquot
of liquid TFA to the dried sample under dry nitrogen, in TFA vapor
(introduced by a filter paper soaked in 20 µL of anhydrous TFA),
and incubating at 50 °C or room temperature for 2-10 min.
Completeness of cleavage was assessed by LC-MS, using the
conditions noted above.
RESULTS AND DISCUSSION
Assessment of Insulin Disulfide Integrity. The disulfide
integrity of recombinant human insulin used as a model protein
was assessed chromatographically. As reported previously,11
scrambled products may be chromatographically resolved from
the native structure using a reverse-phase separation. The
recombinant insulin showed the presence of a single peak (shown
in the Supporting Information). Because of the absence of multiple
peaks in the reverse-phase profile, it is inferred that the model
protein utilized for method development presented the expected
disulfide arrangement.
Isolation of Disulfide-Linked Peptides from Recombinant
Human Insulin. As shown in the Supporting Information, insulin
contains three disulfide linkages between its six cysteine residues.
Digestion with endoprotease Glu-C under nonreducing conditions
putatively generates four proteolysis products, summarized in
Table 1. Comparative peptide mapping of the digest under
nonreducing and reducing conditions (Figure 1) affords identifica-
tion of peaks corresponding to disulfide-containing peptides
through either retention time shift, mass difference, or both. Two
peptides were identified as containing disulfide linkages, peaks
eluting at ∼29 and ∼36 min, as they were absent from the
chromatogram upon treatment with reducing agent. The peak
eluting at ∼29 min was tentatively identified as peptide A3(s-s)B2
by comparison of its observed and theoretical masses (1376.58
Da and 1376.59 Da, respectively, Table 1). Similarly, the peak
eluting at ∼36 min was identified as peptide A2(s-s)B1 (2967.32
Da and 2967.33 Da for observed and theoretical masses, respec-
tively, Table 1).
The disulfide connectivity of nonreduced peptides containing
a single disulfide bond may be determined directly from confirma-
(14) Tarr, G. E. Anal. Biochem. 1975, 63, 361–370
.
7316 Analytical Chemistry, Vol. 81, No. 17, September 1, 2009

oxidized B-chain. Coupling and cleavage steps are summarized
below.
Coupling Reaction (Insulin Oxidized B Chain). Incorporation
of the PITC label on primary amines was assessed by comparing
2% vs 20% of the total reaction volume at 50 °C for 20 min. Insulin
oxidized B chain contains two primary amines: the free N-terminus
and a lysine residue at position 29. Evaluation of extracted ion
chromatograms for the unlabeled, singly-, and doubly labeled
peptide demonstrated that under the coupling conditions, both
2% and 20% proportions generate essentially complete labeling at
primary amines (data not shown). Therefore, in order to accom-
modate more complex disulfide-linked peptides containing more
primary amines, an intermediate value of 10% PITC was selected.
A time-course study to evaluate the labeling requirements
using 10% PITC and 50 °C reaction temperature was performed.
Reaction times of 3, 5, and 20 min were assessed. In each case,
100% label incorporation (i.e., doubly labeled peptide) was
observed; however at 20 min, a side reaction that generated a
+12 Da adduct mass was observed. Therefore 5 min was selected
as the optimal reaction time.
Incubation temperature was also probed. Automated Edman
sequencing programs typically employ 50 °C reaction tempera-
tures; however, in order to mitigate potential thermally induced
disulfide rearrangements, a comparison of incubation at 50 °C and
room temperature was employed for insulin oxidized B chain and
also angiotensin II as a secondary test peptide. The results
indicated that room temperature incubation was sufficient to yield
complete labeling of primary amines (data not shown).
Automated programs for Edman sequencing employ washing
steps with highly hydrophobic solvents (ethyl acetate and heptane)
to remove the small molecule byproduct of the coupling reaction.
However, it has been reported in the literature that exposure to
organic solvents could potentially lead to disulfide scrambling.16
In order to assess the impact on the peptide detectability of the
washing step, the coupling reaction was performed using insulin
oxidized B chain, with and without the washing step. The results
demonstrated that although the reagent peaks were diminished
with washing, recovery of the coupled peptide was reduced by
approximately 50% with the washing step (peak areas of 2 973 728
vs 1 385 883 by the summed extracted ion chromatogram (EIC)
of multiple charge states or 1 368.78 vs 675.20 by UV for exclusion
or inclusion of the wash step, respectively, data not shown). The
issues with compromised recovery as well as the potential for
scrambling of disulfide-linkages was considered to be of sufficient
risk to the value of the method that the washing step was omitted
from the final protocol.
Therefore, the optimized coupling conditions were determined
to be addition of a 40 µL volume of anhydrous pyridine, 5 µL of
PITC (10% of the total reagent volume), and 5 µL of methyl-
piperidine under dry nitrogen to the reconstituted target peptide.
Coupling was achieved by incubation at room temperature for 5
min, after which the solution containing coupled peptide was dried
completely by vacuum centrifugation. Note that coupling was
performed after reconstitution of the target peptide in 10 µL of 10
mM NEM in water and incubation at room temperature in the
dark for 30 min in order to mitigate disulfide scrambling.
Evaluation of the final coupling conditions on insulin Glu-C peptide
Figure 1. Comparative peptide maps of insulin Glu-C digests under
nonreducing and reducing conditions.
tion of the constituent species. Peak A3(s-s)B2 eluting at ∼29 min
yielded two product peaks upon treatment with reducing agent
(peaks A3 and B2 in Figure 1), and these species were identified
by mass spectrometry as NYCN and ALYLVCGE (observed
masses of 512.17 and 866.42 Da, respectively, Table 1). As the
identified peptides contain only a single Cys residue each, the
disulfide linkage of peptide A3(s-s)B2 must be through the two
cysteine residues, i.e., Cys-20 in peptide A3 is bound to Cys-19 in
peptide B2.
Peptide A2(s-s)B1 eluting at ∼36 min also yielded two product
peaks (peptides A2 and B1 in Figure 1). Mass spectrometry
detection of the product peaks confirmed their identities as
QCCTSICSLYQLE and FVNQHLCGSHLVE (observed masses of
1489.63 Da and 1481.72 Da, respectively, Table 1) through
matching of the expected and observed masses. Because of the
fact that one of the peptides contains three cysteine residues,
confirmation of the specific linkage pattern cannot be inferred
from the constituent peptides nor can the presence of coeluting
disulfide variants. In order to ascertain the connectivity of the
disulfide linkages in peptide A2(s-s)B1, a new methodology was
developed, as described in the following sections.
Manual Edman of Glu-C Peptide A2(s-s)B1. Principle of
the Method. The methodology comprises a convergence of
traditional Edman chemistry executed manually with analysis
using LC-MS. Employment of extracted ion chromatograms is a
key aspect, on account of the fact that the UV trace is confounded
by remnant reagent peaks. The mechanisms for Edman chemistry
have been extensively documented.14,15 Under basic conditions,
the phenylisothiocyanate couples with R amino groups (i.e., the
N-terminus) to form a phenylthiocarbamyl (PTC) group. This is
cleaved by treatment with anhydrous acid, releasing the anili-
nothiazolinone (ATZ) derivative of the N-terminal amino acid and
generating a new N-terminus for repetitive cycles. Conversion of
the ATZ derivative to the more stable phenythiohydantoin (PTH)
derivative results in the loss of water and a consequent reduction
in mass.
Optimization Studies of Manual Edman Reaction. Optimization
studies were accomplished using a simpler test molecule, insulin
(15) Creighton, T. E. Proteins: Structure and Molecular Properties, 2nd ed.; W. H.
Freeman and Co.: New York, 1993.
(16) Chang, J. Y.; Li, L. J. Biol. Chem. 2001, 276, 9705–9712.
Analytical Chemistry, Vol. 81, No. 17, September 1, 2009 7317

A2(s-s)B1 showed that the expected incorporation of two PITC
labels was essentially 100% complete.
Cleavage Reaction. Removal of the N-terminal amino acid residue
following the coupling step was accomplished by acid cleavage, based
upon the conditions used with automated instrumentation programs.
Cleavage was found to be complete after incubating with 10 µL of
neat anhydrous TFA for 2 min at 50 °C, i.e., no partial cleavage
products were observed by LC-MS. The utility of room temperature
incubation was found to be not sufficient for complete cleavage (data
not shown). Vapor phase sequencing reactions have been reported
in the literature to afford higher sensitivity;17 however, in our
assessment there was no discernible advantage using vapor phase
TFA over liquid TFA, most likely due to the fact that the remaining
steps of the degradation chemistry are performed in solution phase.
Liquid phase TFA was employed in the final method on account of
the fact that subsequent manipulations were more readily facilitated.
The completeness of the final cleavage reaction conditions was also
demonstrated using insulin peptide A2(s-s)B1 (data not shown).
Conversion. In automated Edman, a conversion step from the
AZT derivative to the more stable PTH derivative is performed.
This was accomplished simply by monitoring the masses of the
resulting products; converted (PTH) and unconverted (AZT) differ
by the loss of water and are therefore readily differentiated by
MS detection. The conversion step is performed under identical
conditions to that established for automated N-terminal sequenc-
ing using a Procise 494 instrument (Applied Biosystems, Foster
City, CA). Conversion was found to be 100% complete after
incubating with 25% TFA at 64 °C for 10 min and therefore was
not optimized further.
Figure 2. EIC traces for expected A2*(s-s)B1* and potential
scrambled products after one cycle.
Confirmation of Disulfide Integrity during Execution of
Optimized Technique. A major concern with any methodology
employed to probe disulfide bonding is the occurrence of linkage
scrambling or rearrangement. Therefore an assessment of scram-
bling potential was made by performing the coupling, cleavage,
and conversion steps in the presence of NEM. By incorporation
of the alkylating reagent during the processing steps, any
rearrangement of the disulfides would be captured through the
labeling of the transitional free sulfhydryl and the resulting
formation of unique products. Five distinct products would indicate
the occurrence of scrambling: peptide B1* with a single NEM
label, peptide A2* with a single NEM label, peptide A2* with three
NEM labels, B1* dimer, or A2* dimer (where the * notation
indicates the loss of the first N-terminal residue from the annotated
peptide). Each of these species has a unique mass and likely a
characteristic retention time.
Insulin peptide A2(s-s)B1 was pretreated with 10 mM NEM
at room temperature for 30 min before execution of the manual
Edman coupling, cleavage, and conversion cycles in order to
identify the propensity for disulfide scrambling. Figure 2 shows
the EIC of possible NEM derivatives of peptides A2* and B1*,
the A2* dimer and B1* dimer species, as well as of the expected
disulfide-linked structure of peptide A2*(s-s)B1*. The fact that
no detectable quantities of the potential scrambled products were
observed indicates that the disulfide structure was not compro-
mised by execution of the coupling, cleavage, and conversion steps
of the Edman reaction. The small signal in the A2* dimer
(theoretical mass 2717.13 Da) EIC trace is related to the sodium
Figure 3. Mass spectrum for doubly charged species eluting at ∼35
min.
adduct of the expected A2*(s-s)B1* structure (theoretical mass
2692.21 Da). The fourth isotope of the [M + H + Na]2+ ion of
the expected A2*(s-s)B1* structure exhibits the same m/z
value as the first isotope of the [M + 2H]2+ ion of the A2*
dimer (Figure 3). The fact that the two signals occur at the same
elution time suggests that they are related to the same structure
(i.e., protonated vs sodiated), which is considerably more likely
than coelution of different structures. The absence of the B1*
dimer also supports this proposal.
Theoretical Distinction of Disulfide Variants for Glu-C
Peptide A2(s-s)B1. Three putative structures (structures 1, 2, and
3) of peptide A2(s-s)B1 resulting from the Glu-C digestion of insulin
are presented in Figure 4, which differ in the connectivity but all
share the same theoretical mass of 2967.33 Da. In structure 1, the
disulfide linkages are Cys-6 to Cys-11 (intrachain on chain A) and
Cys-A7 to Cys-B7 (interchain). This represents the traditionally
accepted linkage pattern.10,11 The other two structures contain
variations of the disulfide linkages; structure 2 contains an interchain
linkage between Cys-6 of the A-chain to Cys-7 of the B-chain and an
intrachain linkage between Cys-7 and Cys-11 of the A-chain. Structure
3 contains an intrachain linkage between Cys-6 and Cys-7 of the
A-chain and an interchain linkage between Cys-11 of the A-chain and
Cys-7 of the B-chain. Note that these represent purely hypothetical
arrangements given the confirmed constituents of peptide A2(s-s)B1,
(17) Brandt, W. F.; Frank, G. Anal. Biochem. 1988, 168, 314–323
.
7318 Analytical Chemistry, Vol. 81, No. 17, September 1, 2009

Figure 4. Differentiation of insulin disulfide variants through characteristic residual and leaving groups.
although both structures 2 and 3 have been generated under
conditions of limited reduction and denaturation.11
The three species are differentiated by their respective residual
and leaving groups following sequential cycles of Edman sequenc-
ing chemistry, as demonstrated in Figure 4. The first cycle does
not permit distinction as each of the species simply loses the
glutamine and valine residues from the A and B chains, respec-
tively. While the products of structures 1 and 3 remain isobaric
after the second cycle, structure 2 is clearly differentiated as two
product species are formed of masses 1256.56 and 1471.59 Da.
Three cycles are necessary to distinguish structures 1 and 3, after
which the A and B chains of structure 1 are cleaved, resulting in
the formation of two product species of mass 1391.57 and 1357.55
Da, respectively. Structure 3 is not cleaved into separate chains
by three cycles of Edman chemistry, and as such the residual
Figure 5. EIC trace for A2*(s-s)B1* peptide (structures 1, 2, and 3)
after one cycle.
mass is 2275.08 Da.
Application of Repetitive Cycles to Insulin Peptide
A2(s-s)B1. As noted in Figure 4, the presence of potential disulfide
variants may be distinguished only after at least two cycles, due to
the location of the first cysteine at the second amino acid position in
peptide A2. Therefore multiple rounds of coupling, cleavage, and
conversion steps were executed in sequential manner.
Cycle 1. The first cycle of N-terminal truncation does not result
in distinguishing species as each of the three structures 1*, 2*,
and 3* possess the same mass. The EIC for peptide A2*(s-s)B1
is shown in Figure 5. These data indicate that the expected
species, A2*(s-s)B1* was detected. As noted above (Figure 2),
potential scrambling products were not detected.
Cycle 2. The second cycle of N-terminal truncation is the first
opportunity at which the discrete disulfide variant structures may
be differentiated. As noted above, structures 1** and 3** (i.e.,
A2**(s-s)B1**), where ** indicates the completion of two Edman
cycles, cannot be resolved by mass at this point, as they still
possess the same constituent amino acid inventory. However,
structures 2a** (peptide A2**) and 2b** (peptide B1**) are readily
differentiated from structures 1** and 3**. EICs for these
structures are presented in Figure 6. The traces demonstrate that
the dominant species identified has a mass that corresponds to
the expected structure (structure 1**), although clearly no
Analytical Chemistry, Vol. 81, No. 17, September 1, 2009 7319

One aspect to note is the reduction in signal-to-noise through
sequential steps (Figures 5-7), which is recognized as a limitation
of the present methodology. This is analogous to the limited
cycling efficiency of automated Edman degradation, which typi-
cally demonstrates a repetitive yield of ∼90%;18 however, it is
apparent that the losses in the manual methodology may be higher
due to the need to recover the residual groups from multiple
chromatographic steps.
CONCLUSIONS
We have presented here a summary of a novel methodology
that affords unambiguous assignment of disulfide linkages in
proteins/peptides containing closely spaced or adjacent cysteine
residues. The method is rapid and sensitive and utilizes estab-
lished methodologies including Edman sequencing chemistry.
Data interpretation is facile, unlike contemporary methods such
as MS/MS fragmentation of disulfide linked peptides. Scrambling
was shown to be well controlled through optimization studies and
judicious selection of reaction parameters consistent with condi-
tions reported in the current literature.
Figure 6. EIC traces for species resulting from structures 1, 2, and
3 after two cycles.
A limitation of the method is the apparent cumulative losses
through repetitive cycling. Therefore, the proximity of the first
cysteine residue to the native N-terminus of the protein or peptide
of interest will dictate the sample requirements for successful
analysis without ambiguity. This highlights the need for judicious
selection of hydrolytic reagent for cleavage as close to the cysteine
residues as possible. Simply pooling multiple rounds of purifica-
tions may afford sufficient material for repetitive analysis. Nev-
ertheless, it is recognized that the cycling efficiency is an area of
potential improvement. The methodology presented in this
manuscript utilizes traditional analytical scale separations. Im-
provements in yield and sensitivity may be afforded through
coupling the separation with a nanospray ionization source capable
of simultaneous fraction collection, such as the Advion NanoMate
Triversa. Preliminary data in our laboratory suggests that this may
afford significant sensitivity improvements due to the properties
of the nanospray interface as well as throughput gains by
eliminating the need for separate chromatographic steps for
analysis and collection. In a forthcoming paper we will present
the application of this methodology in the determination of linkage
connectivity to the recently discovered disulfide variants of
antibodies of the IgG2 subclass.
Figure 7. EIC traces for species resulting from structures 1 and 3
after three cycles.
distinction can be made regarding the presence of structure 3**.
What is evident is the fact that no detectable levels of the species
corresponding to structure 2 are observed; therefore, it is
concluded that this particular unexpected disulfide variant is not
present in the population of insulin molecules.
Cycle 3. The third cycle of N-terminal truncation provides
differentiation of structures 1 and 3. After three cycles, structure
1 dissociates to discrete peptides from each chain (structures
1a*** and 1b*** are peptides A2*** and B1***, respectively).
Completion of three cycles for structure 3 results in the loss of a
di-PTH-cystine group (structure 3b***) and a residual group that
contains peptides A2*** and B1*** bound through a disulfide
linkage. EICs for these structures are presented in Figure 7. Note
that the di-PTH-cystine is not reliably detected due to its small
mass or poor retention on the column.
The traces demonstrate that the dominant species identified
are structures 1a*** and 1b***, which correspond to the expected
structure (structure 1). Note that structure 1b*** and 3a*** were
detected as the pyroglutamylate derivative (with concomitant loss
of 17.03 Da), as the N-terminal residue of each species is a
glutamine. Very minor amounts of species derived from structure
3 were detected, again demonstrating that the native disulfide
structure is essentially preserved throughout multiple rounds of
manual Edman cycling.
SUPPORTING INFORMATION AVAILABLE
Representation of the established insulin amino acid sequence
and disulfide structure (peptide annotations (A1, etc.) refer to
expected Glu-C peptide nomenclature for the corresponding A
or B chain from the N-terminus). In addition a chromatographic
assessment of insulin disulfide integrity by reverse-phase separa-
tion is shown. This material is available free of charge via the
Internet at http://pubs.acs.org.
Received for review May 28, 2009. Accepted July 13,
2009.
AC901161E
(18) Miyashita, M.; Presley, J. M.; Buchholz, B. A.; Lam, K. S.; Lee, Y. M.; Vogel,
J. S.; Hammock, B. D. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 4403–4408
.
7320 Analytical Chemistry, Vol. 81, No. 17, September 1, 2009