Characterization of a three-component coupling reaction on proteins by isotopic labeling and nuclear magnetic resonance spectroscopy

Article abstract of DOI:10.1021/ja710927q

A three-component Mannich-type electrophilic aromatic substitution reaction was previously developed to target the phenolic side chain of tyrosine residues on proteins. This reaction proceeds under mild conditions and provides a convenient alternative to lysine-targeting strategies. However, the use of reactive aldehydes, such as formaldehyde, warrants careful inspection of the reaction products to ensure that other modifications have not occurred. Through the use of isotopically enriched reagents, nuclear magnetic resonance (NMR)-based studies were used to obtain structural confirmation of the tyrosine-modification products. These experiments also revealed the formation of a reaction byproduct arising from the indole ring of tryptophan residues. Cysteine residues were shown to not participate in the reaction, except in the case of a reduced disulfide, which formed a dithioacetal. We anticipate that this analysis method will prove useful for the detailed study of a number of bioconjugation reactions.

Full text of DOI:10.1021/ja710927q

Published on Web 05/23/2008
Characterization of a Three-Component Coupling Reaction on
Proteins by Isotopic Labeling and Nuclear Magnetic
Resonance Spectroscopy
Jesse M. McFarland, Neel S. Joshi, and Matthew B. Francis*
Department of Chemistry, UniVersity of California, Berkeley, California 94720-1460, and
Materials Sciences DiVision, Lawrence Berkeley National Labs, Berkeley, California 94720-1460
Received December 7, 2007; E-mail: francis@cchem.berkeley.edu
Abstract: A three-component Mannich-type electrophilic aromatic substitution reaction was previously
developed to target the phenolic side chain of tyrosine residues on proteins. This reaction proceeds under
mild conditions and provides a convenient alternative to lysine-targeting strategies. However, the use of
reactive aldehydes, such as formaldehyde, warrants careful inspection of the reaction products to ensure
that other modifications have not occurred. Through the use of isotopically enriched reagents, nuclear
magnetic resonance (NMR)-based studies were used to obtain structural confirmation of the tyrosine-
modification products. These experiments also revealed the formation of a reaction byproduct arising from
the indole ring of tryptophan residues. Cysteine residues were shown to not participate in the reaction,
except in the case of a reduced disulfide, which formed a dithioacetal. We anticipate that this analysis
method will prove useful for the detailed study of a number of bioconjugation reactions.
Introduction
NMR could be a valuable tool for the development of protein
bioconjugation reactions; however, the complexity of protein
Due to the importance of protein bioconjugates to the fields
of chemical biology, medicine and materials, there have been
significant efforts to develop new reactions that can modify
proteins in aqueous media under mild reaction conditions.1 In
most cases, the reactions are designed to exhibit chemoselec-
tivity for a particular amino acid side chain. However, with these
techniques comes the significant challenge of characterizing the
products that are formed. While many methods can be used to
detect protein modification in the overall sense, including gel
electrophoresis, affinity-based assays, mass spectrometry (MS)
and combinations of those techniques, the full understanding
of bioconjugation reactions requires more exact chemical
information than these methods generally provide. In particular,
it is difficult to identify regio-isomers and unanticipated reaction
products. Although the MS/MS analysis of proteolytic digests
can confirm site-selectivity to an extent, the incomplete sequence
coverage typically observed makes it challenging to be certain
that undesired species are, in fact, not present.
spectra, coupled with the limited access of most researchers to
high-field instruments, significantly limits the throughput that
can be realized. To simplify this, we hypothesized that by using
isotopically enriched reagents we could enhance only the signals
associated with the new products that are formed. Thus, we
would be able to use NMR for the efficient and accurate
characterization of protein bioconjugation reactions.
Previously, we have developed a three-component Mannich-
type electrophilic aromatic substitution (EAS) reaction that
targets tyrosine residues on proteins (Figure 1a). The low
numbers of solvent-accessible tyrosine residues on most proteins
presents an attractive alternative to cysteine- or lysine-targeting
chemistries for generating well-defined bioconjugates from
native proteins.4,5 Initial studies were performed primarily on
R-chymotrypsinogen A, and the reaction products were detected
by both mass spectrometry and gel electrophoresis. These
techniques were used to identify formaldehyde and electron-
rich anilines as the optimal reagents (Figure 1). Given the
availability of 13C-formaldehyde and the chemical nature of the
products that are formed on proteins, we chose this reaction as
a platform to explore the viability of NMR-based characteriza-
tion of isotopically enriched bioconjugates, eq 1. In this report
we confirmed the previously reported tyrosine modification
product, as well as identified two side products formed in the
Mannich reaction on certain proteins. The first is an aminal with
Nuclear magnetic resonance spectroscopy (NMR) has long
been applied to protein structure determination using experi-
ments that probe through-bond and through-space couplings.2
In addition, previous reports have demonstrated the selective
incorporation of NMR active nuclei as environmental probes
to monitor changes in proteins.3 These successes suggest that
(1) Hermanson, G. T. Bioconjugate Techniques, 1st ed.; Academic Press:
San Diego, 1996. Francis, M. B. New Chemical Tools for Protein
Modification. In Chemical Biology, Schreiber, S. L., Kapoor, T., Wess,
G., Eds.; Wiley-VCH: Weinheim, 2007; pp593-635. Antos, J. M.;
Francis, M. B. Curr. Opin. Chem. Biol. 2006, 10, 253–262. Cowburn,
D.; Shekhtman, A.; Xu, R.; Ottesen, J. J.; Muir, T. W. Methods Mol.
Biol. 2004, 278, 47–56.
(3) Jackson, J. C.; Hammill, J. T.; Mehl, R. A. J. Am. Chem. Soc. 2007,
127, 1160–1166. Vaughan, M. D.; Cleve, P.; Robinson, V.; Duewel,
H. S.; Honek, J. F. J. Am. Chem. Soc. 1999, 121, 8475–8478.
(4) Joshi, N. S.; Whitaker, L. R.; Francis, M. B. J. Am. Chem. Soc. 2004,
126, 15942–15943.
(5) Tilley, S. D.; Francis, M. B. J. Am. Chem. Soc. 2006, 128, 1080–
1081. Hooker, J. M.; Kovacs, E. W.; Francis, M. B. J. Am. Chem.
Soc. 2004, 126, 3718–3719.
(2) Clore, G. M.; Gronenborn, A. M. Annu. ReV. Biophys. Biophys. Chem.
1991, 20, 29–63.
9
10.1021/ja710927q CCC: $40.75
2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 7639–7644 7639

A R T I C L E S
McFarland et al.
Scheme 1a
Figure 1. Three-component Mannich-type coupling on tyrosine residues.
(a) Typical reaction conditions employed for protein modification. (b) ESI-
MS analysis of chymotrypsinogen A (expected m/z ) 25666), after exposure
to reaction conditions with fluorescent aniline 1 (product expected m/z )
26411). (c) SDS-PAGE analysis of sample shown in (b). The modification
was not sufficient in mass to allow for resolution of modified and unmodified
protein.
a Conditions: (a) 100 mM phosphate, pH 6.5, 40 °C, 16 h. (b) 1:1 MeCN:
H2O with 50 mM phosphate, pH 6.5, RT, 3 h.
conditions for the small-molecule reactions were similar to the
protein reaction conditions, except that the mixtures were
warmed to 40 °C to accelerate the reaction rates for convenience
(Scheme 1). While the tyrosine and tryptophan analogues were
formed in moderate yields, the low yields of the other products
indicated that those adducts were unlikely to be important on
protein substrates. It was also noted that cysteine analogue 7
was sensitive to the residual acid present in chloroform.
Nonetheless, these efforts yielded the most likely set of reaction
products that could be formed. While indole would be expected
to react at the 2- and 3-positions of the ring, reports documenting
the nucleophilicity of the nitrogen have also appeared, both in
the context of small-molecule6 and protein substrates.7,8
Gratifyingly, heteronuclear correlation experiments revealed
that the chemical shifts of the methylene resulting from
formaldehyde were different in both the carbon (F1) and proton
(F2) dimensions for each molecule (Figure 2 and Table 1).
1H-13C HMBC experiments on tyrosine analogue (3) to detect
long-range coupling showed 3- and 4-bond correlations from
the benzylic methylene to the protons on the phenolic ring
(Figure 2b, green arrows). In addition, a previously reported
species (6) corresponding to the addition of a second formal-
dehyde molecule to the Mannich product was characterized by
NMR.4
the nitrogen of the indole ring of tryptophan residues, and the
second byproduct is a dithioacetal with adjacent sulfhydryl
groups. No other byproducts were identified, despite the
potential reactivity of formaldehyde toward many nucleophilic
side-chain groups.
Results and Discussion
The three-component coupling reaction developed previously
in our laboratory involves the in situ formation of imines that
react with the phenolic side chain of tyrosine residues in aqueous
buffer at neutral pH (Figure 1). Initial efforts demonstrated that
a variety of proteins at concentrations ranging from 10 to 200
µM could be modified with good conversion using a variety of
anilines and formaldehyde. While concentrated formaldehyde
solutions are typically used for the chemical cross-linking of
proteins, the low concentration used in the labeling reactions
does not lead to protein oligimerization. Furthermore, anilines
were found to be competent in the reaction, while aliphatic
amines did not participate. The selectivity of the reaction for
tyrosine residues was demonstrated in the case of R-chymot-
rypsinogen A by proteolytic digest of a modified sample,
followed by MS/MS analysis. However, as noted above, this
technique yields limited information about the nature of the
reaction products formed on proteins. Because this reaction
could provide additional modification species that could have
been missed using MS analysis alone, we chose to explore an
NMR-based analysis of the products of this reaction.
A molecule that was also isolated from the reaction mixtures
was assigned by NMR to be an adduct of two aniline molecules
and one formaldehyde molecule. The aminal is stable to column
chromatography but is hydrolyzed upon exposure to protic
solvents, such as methanol. The aminal can also be identified
by thin-layer chromatography of the reaction mixtures, which
indicates that the equilibrium state likely contains a significant
(6) Swaminathan, S.; Ranganathan, S. J. Org. Chem. 1957, 22, 70–72.
(7) Foettinger, A.; Melmer, M.; Leitner, A.; Lindner, W. Bioconjugate
Chem. 2007, 18, 1678–1683.
Small-Molecule Reaction Characterization. Before analyzing
protein substrates containing 13C-labels, small-molecule models
of expected tyrosine, tryptophan and cysteine residue conjugates
were synthesized and fully characterized for reference. The
(8) Antos, J. M.; Francis, M. B. J. Am. Chem. Soc. 2004, 126, 10256–
10257.
9
7640 J. AM. CHEM. SOC. VOL. 130, NO. 24, 2008

Characterizing Products from Protein Bioconjugation
A R T I C L E S
Table 1. Chemical shifts for Methylene Protons of Coupling
Products, Corresponding to the Spectra in Figure 2.
compound
1H δ (ppm)
13C δ (ppm)
3 (CD3CN)
4.26
4.36
4.30
5.40
5.57
5.27
4.39
3.82
4.65
45.9
49.0
43.8
56.0
69.6
80.1
46.5
35.5
82.1
3 (CDCl3)a
3 (buffer)a,b
4 (CDCl3)
5 (CDCl3)
6 (DMSO-d6)
7 (DMSO-d6)
8 (buffer)a,b
13C-formaldehydea,b
a Not shown in Figure 2. b Spectrum recorded in D2O with 10 mM
phosphate, pH 6.5.
Table 2. Percent Modification of Proteins for NMR Analysis As
Quantified by ESI-LCMS.
protein
% +0
% +1
% +2
% +3
chymotrypsinogen A
lysozyme
7
25
51
40
42
27
trace
8
in the sample, which was likely due to the ongoing hydrolysis
of semistable imine and (hemi)aminal species. Treatment of the
reaction mixtures with hydroxylamine, a commonly used reagent
for the cleavage of transient species after acylation of lysine
residues with NHS-esters, was identified as an important step
to aid in the removal of these adducts on the protein. After
reaction and treatment with hydroxylamine, protein samples
were purified by spin concentration. The resulting solutions were
lyophilized to remove water, and the powder was reconstituted
in deuterium oxide for NMR analysis. Analysis of the modified
protein sample by ESI-MS revealed high levels of conversion
to products with up to three modifications per protein (Table
2).
After optimization of sample purification, an HSQC experi-
ment on modified chymotrypsinogen was acquired to probe for
hydrogen atoms directly bonded to 13C atoms. Using this
technique, a single major peak was observed in the region
corresponding to the benzylic methylene of the tyrosine residue
modification (Figure 3b). The chemical shift of the peak
correlated well with the chemical shift of the indicated meth-
ylene in tyrosine analogue 3, supporting the observation of
tyrosine residue modification. Previous experiments have shown
that more than one tyrosine residue in the protein will participate
in the reaction, so the single peak is likely to be composed of
at least two signals from different positions in the primary
sequence, which are not resolved under the experimental
conditions. Another peak observed at 70 ppm in the 13C
dimension appeared related to the native protein, as an HSQC
spectrum of unmodified protein contained a peak with nearly
identical shifts.9 There were no peaks in the spectrum that could
be assigned to tryptophan residue modification, which was
consistent with previous proteolytic digest data. An HMBC
experiment gave one large and one small peak in the spectrum,
corresponding to the two protons on the phenolic ring that would
be expected to couple to the introduced 13C-methylene (Figure
3c). As the presence of both of these peaks is characteristic of
tyrosine modification, the data provide an unambiguous con-
firmation of the Mannich modification product.
Figure 2. 1H-13C correlation spectra of model compounds. Methylenes
resulting from the formaldehyde carbon are indicated with blue arrows and
asterisks; chemical shifts for these signals are tabulated in Table 1. (a)
Tyrosine analogue product 3 HSQC spectrum. (b) Tyrosine analogue 3
HMBC spectrum. Green arrows and asterisks indicate 3- and 4-bond
heteronuclear coupling with the formaldehyde carbon. (c) Tryptophan
analogue product 4 HSQC spectrum. (d) Tyrosine analogue aminal 6 HSQC
spectrum. (e) Cysteine product 7 HSQC spectrum. (f) 3-Methyl indole
hemiaminal 5 HSQC spectrum.
amount of this species. However, this compound does not
correspond to adducts formed on proteins.
The aniline should, in principle, be an excellent partner in
an EAS reaction, thus suggesting that the Mannich product on
proteins could itself serve as a site for additional modifications.
However, all attempts to identify or isolate an EAS product on
the anilines themselves were unsuccessful. Therefore, it is
unlikely that similar reaction pathways occur on proteins.
NMR Analysis of 13C-Labeled Chymotrypsinogen and
Lysozyme. We reasoned that using 13C-enriched formaldehyde
would generate unique, isotopically labeled methylenes on the
protein that could subsequently be detected using 13C-filtered
1
heteronuclear NMR experiments, such as H-13C HSQC and
1H-13C HMBC. To compare protein reaction products to the
model compounds, we undertook the modification of chymot-
rypsinogen with 13C-formaldehyde and anline 2. Previous mass
spectrometry and gel-based analyses have shown that proteins
do not react with anilines without the formaldehyde present.
Therefore, all protein conjugates formed in the reaction should
be observable using 13C-filtered experiments.
(9) Two additional peaks appear at 32 ppm in F1 at 2.6 and 1.9 ppm,
respectively. These peaks are much lower in intensity and do not
correlate well to any model structures. Thus, they could not be assigned
in this study (see Figure S4).
Several 1H 1D and 1H-13C 2D spectra of modified proteins
revealed significant amounts of residual formaldehyde hydrate
9
J. AM. CHEM. SOC. VOL. 130, NO. 24, 2008 7641

A R T I C L E S
McFarland et al.
Figure 4. NMR analysis of modification products for lysozyme. (a) 1H-13C
HSQC of unmodified lyszyme. (b) H-13C HSQC of modified lysozyme.
1
Figure 3. NMR analysis of modification products for R-chymotrypsinogen
A. (a) HSQC of unmodified chymotrypsinogen A. (b) 1H-13C HSQC of
modified chymotrypsinogen A. The peak appearing at 3.64 ppm is in all
samples and thus is independent of this reaction. (c) 1H-13C HMBC of
modified chymotrypsinogen A. (d) Observed couplings (i) by HSQC and
(ii) by HMBC.
The peak appearing at 3.15 ppm is in all samples and thus is independent
of this reaction. (c) 1H-13C HMBC of modified lysozyme. (d) Observed
coupling for peaks 1 and 2. Peaks 3-4 are assigned to tyrosine modification
as in Figure 3.
be discerned. Comparison of the modified and unmodified
spectra easily identify most peaks that result from the native
side chains, although residues that are close in space to the
modification are likely to be shifted in the modified spectra,
complicating their assignment. While the sensitivity will be
affected by signal diffusion resulting from the specific dynamics
of the label location, further improvements in the signal-to-noise
ratio can be achieved by increasing the sample concentration,
the inherent sensitivity of the instrument being used, and the
number of scans that are acquired.
Modification and NMR Analysis of Thioredoxin. Thioredoxin
is a redox-active protein that has recently attracted interest due
to its multiple biological functions. It is emerging as a key player
in several human disease states, including cancer, AIDS, and
Alzheimer’s disease.11,12 Thioredoxin’s 12 lysines and 5 cys-
teines, some of which are necessary for catalytic activity, make
it difficult to produce well-defined bioconjugates of this protein.
In contrast, the protein contains only one naturally occurring
tyrosine that is positioned remotely from the active site, making
it an attractive substrate for the three-component coupling
reaction.
The protein was overexpressed in Escherichia coli, purified
by affinity chromatography, and characterized by SDS-PAGE
(Figure S8). To test the reactivity of the tyrosine residue, the
protein was treated with 13C-formaldehyde and aniline 2.
Unfortunately, we were unable to obtain mass spectral data of
the modified protein to assess the overall conversion and the
number of times that the modification occurred. We therefore
turned to NMR to identify the products that were formed.
HSQC analysis of the sample revealed four peaks in the
spectrum (Figure 5a). We were able to assign the peaks labeled
1 and 2 as resulting from modification of the tryptophan and
tyrosine residues, respectively. Integration of those peaks
The analysis of lysozyme, which was modified under identical
conditions (Table 2), gave three major peaks by HSQC. By
chemical shift, it appeared that two peaks corresponded to
tyrosine residue modification (peaks at 46.9 and 49.5 ppm in
F1) and one to tryptophan residue modification (Figure 4b).
However, examination of the HMBC spectrum indicated that
only one of the 13C labels coupled to aryl protons (at 46.9 ppm
in F1), which was characteristic of tyrosine modification. Thus,
the use of both correlation experiments provides an excellent
way to confirm peak identity. There was no secondary peak
observed for the coupling of the 13C label to the indole ring of
the tryptophan residue (Figure 4c). Integration of the peak
volumes in the HSQC spectrum indicated that approximately
40% of the modifications were associated with tyrosine residues,
while the remaining 60% were split between two tryptophan
modifications (Table 3). This result is significant, as it is the
first confirmation of tryptophan residues participating in this
coupling reaction.
Examination of the spectra provides some insight into the
sensitivity of the technique. Signal to noise (S/N) calculations
for the assigned peaks range from 90 to 300.10 In the case of
lysozyme, the peak assigned to tyrosine modification has S/N
≈ 200. Peaks with S/N ≈ 10 are present and are readily
detectable, although the assignment of the peaks is difficult.
Thus, modification levels as low as 5% of major species can
Table 3. Relative Percent Modification of Protein Residues As
Quantified by NMR.a
protein
peak #
residue
% vol.
lysozyme
1
2
3
1
2
W
W
Y
W31
Y49
20
40
40
50
50
thioredoxin
(10) The S/N is approximated as 2.5 × (peak intensity/peak-to-peak noise).
(11) For a review on the role of thioredoxin in several disease states, see:
Powis, G.; Montfort, W. R Annu. ReV. Pharmacol. Toxicol. 2001, 41,
261–295.
a Peak numbers refer to labels on the appropriate HSQC spectra.
Residue assignment was made on the basis of chemical shift compared
to that of a reference compound. Percent volumes were approximated by
volume integration of HSQC spectra.
(12) Mitchell, D. A.; Marletta, M. A. Nat. Chem. Biol. 2005, 1, 154–158.
9
7642 J. AM. CHEM. SOC. VOL. 130, NO. 24, 2008

Characterizing Products from Protein Bioconjugation
A R T I C L E S
question in our laboratory. In a final experiment, papain was
first treated with N-ethyl maleimide to cap reactive cysteines
and then modified as above.13 Comparison of the resulting
HSQC spectra showed no significant differences. To further
probe the compatibility of this reaction with cysteine residues,
a similar experiment was performed on the S123C mutant of
recombinant tobacco mosiac virus coat protein (rTMV).14 The
protein was modified to ∼25% conversion with the three-
component coupling reaction, as determined by ESI-MS.
Subsequent treatment with N-ethyl maleimide gave high levels
of conversion to the expected products (Figure S9). While not
predictive in all cases, these results confirm that free thiols are
generally compatible with the reaction conditions.
Figure 5. NMR analysis of modification products of thioredoxin. (a)
1H-13C HSQC of modified thioredoxin. (b) Crystal structure of thioredoxin
(PDB ID: 1ERT) - W31 in green, Y49 in orange. (c) Observed coupling
for peak 3.
Conclusions
The unique reactivity of each protein is emphasized by this
analytical method. Chymotrypsinogen, which has four tyrosines
and eight tryptophans in the primary sequence, shows differences
in reactivity when compared to lysozyme, which has three
tyrosines and six tryptophans. Examination of the crystal
structures does not clearly indicate that one or more tryptophan
residues on lysozyme is more solvent accessible than any of
the tryptophans on chymotrypsinogen. The crystal structure of
thioredoxin is more predictive, showing that the indole ring is
positioned adjacent to the active site and appears to be highly
accessible to the solvent. Thus, in the three-component coupling
reaction, the participation of tryptophan residues should be
considered, but is not necessarily prohibitive in the selective
modification of tyrosine residues. As there are very few methods
that can be used to introduce new functional groups on
tryptophans, the development of aniline substrates that can target
this residue selectively with respect to tyrosines is underway.
We are also attempting to correlate indole proton-exchange rates
and fluorescence emission spectra to the observed levels of
reactivity. However, we lack sufficient data to develop a
predictive model at this time.
These studies also show that the use of formaldehyde at
moderate concentrations as a reagent for bioconjugation reac-
tions does not appear to give rise to significant amounts of
byproducts, except in the unique case of the dithioacetal formed
on thioredoxin. Thus, the development of this methodology to
observe the products formed on proteins by NMR through the
use of 13C-enriched reagents and 13C-filtered two-dimensional
experiments has been important in the identification of new
bioconjugates, as well as the confirmation of existing data. We
are currently synthesizing isotopically labeled probes for ad-
ditional bioconjugation reactions for similar characterization
purposes.
Figure 6. NMR analysis of modification products for papain. (a) 1H-13C
HSQC spectrum of unmodified papain. (b) 1H-13C HSQC spectrum of
modified papain. The peak appearing at 2.93 ppm is in all samples and
thus is independent of this reaction.
showed similar levels of modification for each residue (Table
3). As one of the most hydrophobic residues, tryptophan is
typically buried in the interior of the protein, making it difficult
to predict whether the residue will participate in bioconjugation
reactions. In the crystal structure of thioredoxin, the single
tryptophan residue is located adjacent to the active site and is
highly solvent exposed, perhaps allowing this result to be
anticipated (Figure 5b).
Finally, by comparison of the shifts in the carbon dimension,
peak 3 can be assigned as dithioacetal formation with the
reduced active-site cysteines. Djenkolic acid, 8, showed a shift
of 35.5 ppm for the methylene carbon, compared to 36.6 ppm
for the protein sample, although there is a large difference in
the proton dimension. Dithioacetals are know to be exceptionally
stable to hydrolysis, and thus the persistence of this species
despite treatment with hydroxylamine is understandable. The
final peak resulted from a residual amount of 13C-formaldehyde
that remained after purification. Because of the relatively small
amount of protein in the sample, an HMBC experiment was
not attempted due to instrument constraints.
Materials and Methods
General Procedures and Instrumentation. Unless otherwise
noted, all chemicals and solvents were of analytical grade and used
as received from commercial sources. Water (dd-H2O) used in
biological procedures or as reaction solvent was deionized using a
NANOpure purification system (Barnstead, U.S.A.). Small mol-
ecules were synthesized and fully characterized as described in the
Supporting Information.
NMR Analysis of 13C-Labeled Papain. To test the inherent
potential of thiol groups to participate in this reaction, we chose
to modifiy a protein containing an active-site cysteine. Papain,
a cysteine protease, was chosen for this study and was modified
using the conditions described above. Analysis of the HSQC
spectrum showed a single large peak, indicating tyrosine
modification (Figure 6). In addition, a minor peak at 59.1 ppm
in the F1 dimension was observed, showing a minimal amount
of participation by tryptophan residues. The peak appearing at
2.93 ppm in F2 and 38.9 ppm in F1 was also present in the
HSQC spectrum of unmodified protein samples, and thus was
not a result of this reaction. Importantly, no peak corresponding
to cysteine modification was observed, clarifying a long-standing
Electrospray LC/MS analysis was performed using an API
150EX system (Applied Biosystems, U.S.A.) equipped with a
(13) Papain was inactivated with N-ethyl maleimide in a manner similar
to that described in Anderson, B. M.; Vasini, E. C. Biochemistry 1970,
17, 3348–3352.
(14) Miller, R. A.; Presley, A. D.; Francis, M. B. J. Am. Chem. Soc. 2007,
129, 3104–3109.
9
J. AM. CHEM. SOC. VOL. 130, NO. 24, 2008 7643

A R T I C L E S
McFarland et al.
Turbospray source and an Agilent 1100 series LC pump. Centrifu-
gation was conducted using a Legend Mach 1.6R Tabletop
Centrifuge (Sorvall, U.S.A.) or a RC 5C Plus Superspeed Centrifuge
(Sorvall, U.S.A.).
sufficiently large quantities for assignment. When possible, product
formation was quantified by ESI-LCMS after purification (Figure
S2, S3). The purification of the protein samples is described below.
Thioredoxin Modification Procedure for NMR Analysis. In
a 15 mL Falcon tube were combined 0.34 µmol of thioredoxin
(dissolved in 0.9 mL of 50 mM phosphate buffer, pH 6.5), 0.9 mL
of 100 mM phosphate buffer, pH 6.5, 6 mg of 4-aminophenethyl
alcohol dissolved in 18 µL of DMF, 5.4 µL of 13C-formaldehyde
(20% w/w in H2O), and 1 mg of TCEP (3 µmol). The solution was
mixed gently for 40 h at room temperature. Product formation could
not be observed by ESI-LCMS as the protein did not ionize well.
Expression and Purification of Thioredoxin. A wild-type
human thioredoxin clone was received as a gift from the group of
Professor Michael Marletta at the University of California, Berkeley.
Thioredoxin was expressed according to a modified literature
procedure.12 Tuner DE3pLysS competent cells (Novagen, U.S.A.)
were transformed with pET20b plasmid vector containing the
thioredoxin gene and an ampicillin resistance gene. All steps were
carried out under ampicillin (50 µg/mL) selection. Colonies were
selected for inoculation of Luria Broth cultures. When cultures
reached midlog phase as determined by OD600, expression was
induced by addition of 100 µM IPTG (Invitrogen, U.S.A.). Cultures
were grown for 4 h at 37 °C and then harvested by centrifugation.
Cell pellets were resuspended in lysis buffer (50 mM sodium
phosphate, 300 mM sodium chloride, 10 mM imidazole, 5%
glycerol, pH 8.0) and stored at -80 °C until further purification.
The cells were lysed by ultrasonification, and the resulting
cellular debris was pelleted by centrifugation. The supernatant was
applied to a nickel-nitrolotriacetic acid (Ni-NTA) (Qiagen, USA)
column. After loading, the column was washed with 20 volumes
of wash buffer (50 mM sodium phosphate, 500 mM sodium
chloride, 20 mM imidazole, 5% glycerol, pH 8.0), followed by 10
volumes of elution buffer (50 mM sodium phosphate, 500 mM
sodium chloride, 500 mM imidazole, 5% glycerol, pH 8.0).
Fractions containing the desired thioredoxin protein (as determined
by SDS-PAGE) were pooled and concentrated using spin concen-
trators (Amicon Ultra 15, MWCO 10k) (Millipore, USA). The
protein was further purified using a Sephadex S200 16/60 gel
filtration column equilibrated with 50 mM sodium phosphate, 300
mM sodium chloride, 100 µM tris-(2-carboxyethyl)phosphine
(TCEP), 5% glycerol, pH 8.0. Fractions containing thioredoxin were
pooled, concentrated, and stored at -80 °C.
Procedure for Protein Reaction Purification for NMR
Analysis. After the Mannich reaction, H2NOH·HCl was added to
a final concentration of 100 mM. The solution was mixed for a
minimum of 15 min, although longer mixing times were not
detrimental to the samples. The small molecules were removed from
the protein by three rounds of spin concentration (15 mL, MWCO
10k, prewashed with 0.1 M sodium hydroxide and water to remove
glycerol) into 10 mM phosphate buffer, pH 6.5, to a final volume
of approximately 0.6 mL. The resulting solution was lyophilized
to yield a fluffy white powder. The powder was reconstituted with
deuterium oxide and analyzed by NMR.
NMR Analysis of Protein Conjugates. Single bond couplings
were determined by gradient 1H-13C HSQC performed on a Bruker
AV-500 fitted with an inverse broadband probe in the cases of
chymotrypsinogen A, lysozyme, and papain or a Bruker DRX-500
with inverse triple band probe in the case of thioredoxin.15 The
lysozyme and chymotrypsinogen spectra were signal averaged over
16 h. The papain spectra were signal averaged over 12 h. The
thioredoxin spectrum was signal averaged over 5 days. Multiple
bond couplings on chymotrypsinogen and lysozyme were detected
by 1H-13C CIGAR-HMBC16 experiments, which were signal
averaged over 16 h (lysozyme) or 48 h (chymotrypsinogen). All
protein spectra were acquired at 300 K and calibrated to acetone
1
in 10 mM phosphate, pH 6.5, in D2O: H - δ 2.22, 13C - δ 30.89.
NMR Sample Preparation and Analysis. NMR characterization
of small molecules for comparison to protein conjugates was
performed on a Bruker AV-500 instrument fitted with an inverse
broadband probe. Due to poor solubility in deuterium oxide, small-
molecule models were characterized in organic solvent as noted.
In the case of the tyrosine analogue, 3, and djenkolic acid,
characterization in deuterium oxide with 10 mM phosphate, pH
6.5 was also performed.
Acknowledgment. We thank Prof. Dave Wemmer and Dr. Jeff
Pelton for helpful discussions and donation of instrument time, Prof.
Michael Marletta and Dr. Doug Mitchell for help with thioredoxin,
and Andrew Presley for a generous gift of rTMV coat protein. N.S.J.
gratefully acknowledges the Saegebarth family for generous fel-
lowship support. This work was supported by the NIH (GM072700).
Procedure for Protein Modification for NMR Analysis.
Protein samples for NMR analysis were prepared by the method
first described by Joshi, et al.4 In a 15 mL Falcon tube were
combined 2.0 µmol of chymotrypsinogen, lysozyme, or papain, in
10 mL of 100 mM phosphate buffer, pH 6.5, 34 mg of 4-ami-
nophenethyl alcohol (2) dissolved in 100 µL of DMF, and 30 µL
of 13C-formaldehyde (20% w/w) in water (Cambridge Isotopes).
The solution was mixed gently at room temperature until high levels
of labeling were obtained as estimated by ESI-MS, usually 1-7
days. Although such long reaction times are not typically necessary
for simple protein labeling using this reaction, they were used to
be sure that even minor modification products would be present in
Supporting Information Available: Full experimental proce-
dures and characterization data are available for all compounds.
This material is available free of charge via the Internet at http://
pubs.acs.org.
JA710927Q
(15) Palmer, A. G.; Cavanagh, J.; Wright, P. E.; Rance, M. J. Magn. Reson.
1991, 93, 151–170.
(16) Hadden, C. E.; Martin, G. E.; Krishnamurthy, V. V. Magn. Reson.
Chem. 2000, 38, 143–147.
9
7644 J. AM. CHEM. SOC. VOL. 130, NO. 24, 2008