Genetically encoded cleavable protein photo-cross-linker

Article abstract of DOI:10.1021/ja504371w

We have developed a genetically encoded, selenium-based cleavable photo-cross-linker that allows for the separation of bait and prey proteins after protein photo-cross-linking. We have further demonstrated the efficient capture of the in situ generated selenenic acid on the cleaved prey proteins. Our strategy involves tagging the selenenic acid with an alkyne-containing dimethoxyaniline molecule and subsequently labeling with an azide-bearing fluorophore or biotin probe. This cleavage-and-capture after protein photo-cross-linking strategy allows for the efficient capture of prey proteins that are readily accessible by two-dimensional gel-based proteomics and mass spectrometry analysis.

Full text of DOI:10.1021/ja504371w

Communication
pubs.acs.org/JACS
Genetically Encoded Cleavable Protein Photo-Cross-Linker
Shixian Lin,† Dan He,† Teng Long,‡ Shuai Zhang,† Rong Meng,† and Peng R. Chen*,†,‡
†Beijing National Laboratory for Molecular Sciences, Synthetic and functional Biomolecules Center, Key Laboratory of Bioorganic
Chemistry and Molecular Engineering of Ministry of Education, College of Chemistry, and Molecular Engineering, Peking
University, Beijing 100871, China
‡Peking-Tsinghua Center for Life Sciences, Beijing, China
S
* Supporting Information
Scheme 1. A General Procedure for Protein Photo-Cross-
Linking Using a Cleavable Photo-Cross-Linker
a
ABSTRACT: We have developed a genetically encoded,
selenium-based cleavable photo-cross-linker that allows for
the separation of bait and prey proteins after protein
photo-cross-linking. We have further demonstrated the
efficient capture of the in situ generated selenenic acid on
the cleaved prey proteins. Our strategy involves tagging
the selenenic acid with an alkyne-containing dimethoxyani-
line molecule and subsequently labeling with an azide-
bearing fluorophore or biotin probe. This cleavage-and-
capture after protein photo-cross-linking strategy allows
for the efficient capture of prey proteins that are readily
accessible by two-dimensional gel-based proteomics and
mass spectrometry analysis.
rotein photo-cross-linking has become a highly valuable
P
technique for the study of protein−protein interactions,
particularly for weak or transient interactions that are difficult to
capture with traditional methodology.1 For example, a panel of
photo affinity moieties (e.g., benzophenones, aryl azide, and
diazirine groups) has been site-specifically incorporated into
“bait” proteins via the genetic code expansion strategy. This
incorporation allows for covalent linkage with its partner proteins
(“prey”) in close proximity under native cellular conditions.1
Such site-specific protein photo-cross-linking methods signifi-
cantly enhanced cross-linking efficiency and fidelity, thus
enabling subsequent proteomics and mass spectrometry (MS)
analysis for target identification. Despite these advantages, these
methods still suffer from high false positive and false negative
rates.2 Nonspecific interactions between the bait and various
cellular proteins have the potential to generate a large amount of
false positive data; what is more, low enrichment efficiency for
prey proteins causes missed hits as well.3 Therefore, a strategy to
separate the bait protein, and thus its nonspecific interacting
partners, from the cross-linked prey proteins is desirable as a
strategy to eliminate false-positive backgrounds. In addition, an
easy capture and label strategy for use on the separated prey
protein pool would further boost proteomic analysis efficiency.
Herein, we have developed a genetically encoded, cleavable
protein photo-cross-linker that is capable of separating bait and
prey proteins after photo-cross-linking. Furthermore, we
demonstrated that the cleavable handle allowed for efficiently
capturing and labeling the prey protein pool that may be later
accessed through proteomics and MS studies (Scheme 1).
Mild cleavage reactions, the reduction of disulfide bonds in
particular, have been successfully applied to small-molecule-
a
Step 1, Photo-cross-linking of bait and prey proteins via a Se-
containing DiZSeK (Nε-3-((3-methyl-3H-diazirine-3-yl))-propamino-
carbonyl-γ-seleno-L-lysine) amino acid before affinity purification of
the protein complex. Step 2, Separation of bait and prey protein
complexes via a H2O2-mediated oxidative cleavage reaction. Step 3,
Capturing the prey proteins harboring an SA moiety generated in situ
by (i) transferring a bioorthogonal handle and (ii) the subsequent
bioorthogonal labeling with a fluorophore or biotin probe that are
suitable for proteomics and MS analysis.
mediated protein photo-cross-linking or chemical cross-linking.4
These methods easily remove the bait protein and facilitate the
subsequent labeling or enrichment of the prey protein pool.
Inspired by these examples, we reasoned that the introduction of
a functional group into our previously developed, alkyl diazirine-
containing photo-cross-linking unnatural amino acid (UAA)-
DiZPK (Figure S1) could produce a genetically encoded
cleavable photo-cross-linker.5 Most current cleavable moieties
are either prohibitively large in size or not chemically inert inside
cells, thus rendering them useless. Indeed, for successful photo-
cross-linking, the protein translational machinery-the amino acid
tRNA-synthetase/tRNA pair must first recognize these UAAs.6
Notably, amino acids that contain a selenium atom at the γ
Received: May 1, 2014
Published: August 1, 2014
© 2014 American Chemical Society
11860
dx.doi.org/10.1021/ja504371w | J. Am. Chem. Soc. 2014, 136, 11860−11863

Journal of the American Chemical Society
Communication
position could undergo the intramolecular β-elimination
reaction in the presence of mild oxidizing agents, generating
dehydroanaline (Dha) and selenenic acid (SA).7 This chemistry
has been successfully employed for the in situ generation of Dha
on a protein of interest (e.g., histone), which can be further
modified to resemble various biologically active lysine or serine
posttranslational modifications.8 In an attempt to create a
cleavable photo-cross-linker, we turned our attention to a Se
atom that could replace the γ position carbon atom of DiZPK.
We showed that our newly designed photoaffinity UAA, termed
DiZSeK (Scheme 1), was successfully encoded by the
pyrrolysine tRNA-synthetase (PylRS)-tRNAPCyUl A pair. Moreover,
we were then able to use this newly designed UAA, in
conjunction with our cleavage-and-capture after protein photo-
cross-linking (CAPP) strategy, to efficiently separate and capture
prey proteins for target identification.
incorporation into proteins produced in E. coli (Figures 1C
and S6). Notably, the GFP-N149DiZSeK can be converted to
GFP-N149Dha upon H2O2-mediated oxidative cleavage; these
results confirm that the DiZSeK probe can be oxidative cleaved
when incorporated into an intact protein (Figure S7). Next, we
examined the photo-cross-linking efficiency of DiZSeK by
capturing the binding proteins of an E. coli acid chaperone
HdeA inside living bacteria. E. coli cells expressing the HdeA
protein carrying DiZSeK at residue S27 (HdeA-S27DiZSeK)
were treated at pH = 2.3 for 30 min to allow HdeA’s binding with
its substrate proteins. Photoirradiation (365 nm) was next
performed for 10 min on live E. coli cells, and the results were
analyzed by SDS-PAGE gel and immunoblotting assay. High
cross-linking efficiency comparable to that of DiZPK-mediated
protein photo-cross-linking was obtained (Figures 1D and S8).
In addition, the ratio of photo-cross-linked HdeA dimer complex
to the monomer at pH 7 was obtained at a similar yield when
DiZSeK or DiZPK was introduced at the protein dimer interface
(residue Phe35). This result further suggests that use of Se to
replace the γ-carbon on DiZPK has no effect on the function of
the diazirine moiety (Figure S9). Taken together, this new
DiZSeK probe showed amber suppression and photo-cross-
linking efficiency similar to that of DiZPK photo-cross-linker.
To determine if the Se-containing handle can be cleaved upon
oxidative treatment at the bait and prey protein interface, we first
produced a covalent-linked HdeA dimer by photo-cross-linking
HdeA at pH 7 with DiZSeK incorporated at residue Phe 35
within its dimer interface. The addition of H2O2 up to 8 mM8b
still did not result in a decrease in the HdeA dimeric band on
SDS-PAGE gel (Figure S10A). We reasoned that since Phe 35 is
located at the center of the highly hydrophobic HdeA dimer, the
H2O2 molecule may not be able to access it (Figure S11). We
then used SDS to denature the cross-linked HdeA-dimer
complex and performed the same oxidative-cleavage reaction
afterward. The HdeA dimer was progressively disappearing on
SDS-PAGE gel upon the treatment with an increasing amount of
H2O2 (Figure S10B). As a control, no change was detected when
DiZPK was used as the photo-cross-linking probe (Figure S10B).
Completion of this cleavage reaction can be achieved in either
Tris buffer (150 mM, pH = 8.0) or PBS buffer (pH 8) in the
presence of 8 mM H2O2. In this composition we observed no
detectable damage to GFP (carrying surface exposed cysteine
and methionine residues) (Figure S7). Therefore, the concen-
tration is compatible with proteomics and MS analysis. We also
found that 0.5% SDS served as a better denaturation reagent than
8 M urea for this purpose (Figure S12). With an optimized
reaction condition in hand, we investigated the separation of
HdeA and its previously identified in vivo binding proteins (e.g.,
DegP and SurA) after in vitro photo-cross-linking under an acidic
condition (pH 2.3).5 The cross-linked low-pH protein
complexes, HdeA/DegP and HdeA/SurA, were clearly detect-
able on the SDS-PAGE gel but disappeared after the addition of 8
mM H2O2 for 1 h (Figure S13). Together, these results confirm
that the H2O2-mediated cleavage reaction may be used on the Se
handle of DiZSeK in order to yield an efficient separation of bait
and prey proteins after photo-cross-linking.
We began by synthesizing DiZSeK through a tristep synthetic
route shown in Figure S1 with an overall yield of 58% (Figures
S2−4). As expected, the Se-containing linker on DiZSeK was
cleaved by H2O2, generating Dha and DiZ-SA (Figure 1A).
Figure 1. Development of a genetically encoded cleavable protein
photo-cross-linker-DiZSeK. (A) The H2O2-mediated cleavage reaction
on DiZSeK to generate DiZ-SA and Dha. The photoaffinity group and
cleavable linker are shown in red and blue, respectively. (B) Western
blot analysis of the amber suppression efficiency of the PylRS mutant-
tRNA pair in recognizing DiZPK or DiZSeK. Full-length GFP carrying
an in-frame amber mutation and a C-terminal Histag was produced only
in the presence but not absence of 1 mM DiZPK or DiZSeK. (C) MS
analysis of GFP-N149DiZSeK (calculated: 27943 Da, found: 27941
Da.) (D) E. coli cells expressing the periplasm-residing HdeA-
V58DiZPK or HdeA-V58DiZSeK protein (carrying a C-terminal
Histag) were incubated at pH 2.3 followed by 365 nm photolysis. Cell
extracts were separated by SDS-PAGE and analyzed by immunobloting.
An anti-His6 antibody was used for both (B) and (D).
DiZPK and DiZSeK are highly similar structurally; we therefore
tested if our previously obtained DiZPK-recognizing PylRS
mutant could incorporate DiZSeK into proteins with its
Pyl
CUA
cognitive tRNA . Indeed, this PylRS mutant showed similar
amber suppression efficiency in the presence of DiZPK or
DiZSeK, which allowed us to site-specifically incorporate
DiZSeK at an in-frame amber mutation site N149 on green
fluorescent protein (Figures 1B and S5). The yield of this
DiZSeK-incorporated GFP, named GFP-N149DiZSeK, pro-
duced by E. coli cells was estimated to be ∼8 mg·l−1 (Supporting
Information). We used ESI-MS analysis on the produced protein
in order to verify the specificity and fidelity of DiZSeK
Next, we developed a CAPP strategy in order to capture the
liberated prey proteins after oxidative-cleavage and separation of
the prey and bait proteins. We designed a covalently linked GFP-
Biotin system as the “prey-bait” model for proof-of-concept
(Figures 2A,B). GFP was chosen as the prey protein because we
have previously gained intensive experience on LC-MS/MS
analysis of this protein containing different unnatural mod-
11861
dx.doi.org/10.1021/ja504371w | J. Am. Chem. Soc. 2014, 136, 11860−11863

Journal of the American Chemical Society
Communication
MS analysis confirmed that ADMA was successfully conjugated
with SA and transferred onto the cleaved GFP template to
generate GFP-Se-ADMA (Figures 2B,C and S16−17). To
further test the specificity of our CAPP strategy, H2O2 and
ADMA were immediately desalted, and the protein samples were
labeled with Fluor 488-azide through CuAAC reaction (Figure
2B). In-gel fluorescence imaging clearly detected the fluores-
cently labeled bands (Figure S21), while no labeled bands were
observed in the lane containing denatured WT-GFP (carrying
two cysteine residues) under the same CAPP protocol (Figure
S22). These observations indicate that proteinogenic amino
acids including cysteine did not react with ADMA under our
experimental conditions. The reaction procedure and the
labeling yield were further optimized by in-gel fluorescence
imaging assay (Figures S21−22). Together, our CAPP strategy
exhibits high efficiency in capturing the in situ generated prey
proteins after cleavage of the cross-linked prey-bait complexes.
We next tested the compatibility of our CAPP strategy for
labeling and profiling the in vivo binding proteins of HdeA after
protein photo-cross-linking. Indeed, we observed a band shift
before and after the addition of H2O2 into the purified HdeA-
substrate protein complex on SDS-PAGE (Figure S23),
indicating the separation of HdeA and its binding proteins
after the cleavage reaction. Furthermore, strong azide-biotin
labeling signal was only detected in the presence of both H2O2
and ADMA (Figure S24), which agreed with our Western blot
analysis after the bait protein was removed by affinity purification
(Figure S25). Noteworthy, the subsequent capturing of the
liberated prey proteins with a biotin tag allowed for the
enrichment of prey proteins with streptavidin beads, which
would significantly improve the detection limit of target proteins.
Finally, we demonstrated direct profiling of in vivo binding
proteins of HdeA under acid stress by using the DiZSeK probe
and CAPP strategy in conjunction with two-dimensional
polyacrylamide gel electrophoresis (2D-PAGE). Although 2D-
PAGE is a method routinely used in proteomics,11 coupling this
method with protein photo-cross-linking remains challenging,
mainly because of low cross-linking efficiency as well as the highly
abundant bait proteins that are difficult to separate even with 2D
gel electrophoresis. In contrast, our newly developed DiZSeK
probe is not only able to completely remove the bait protein but
can also label and/or enrich the binding proteins that are directly
applicable for 2D-PAGE and the subsequent MS analysis.
Following the previously described strategy, E. coli cells
expressing the periplasm-residing HdeA-F35DiZSeK were
treated with acid stress (pH 2.3) and photoirradiated for 10
min. The cross-linked protein complexes were affinity purified
with Ni-NTA column before being applied to CAPP strategy
(Figure S26). The prey protein pool labeled with Cy5-N3
following the CAPP procedure was then analyzed by 2D-
PAGE, which detected ∼80 dots on the fluorescent gel (Figure
S27A). A coomassie gel with a near 4-fold higher loading of the
protein samples without labeling with Cy5-N3 was next used to
afford sufficient samples for LC-MS/MS analysis (Figure 3).
Although, this coomassie gel showed a similar pattern as the
aforementioned fluorescent gel, these two 2D gels were not
completely overlaid (Figure S27). The main reason was likely
due to the different sample loading between the coomassie gel
and the fluorescent gel, which may cause slightly different
mobility change during gel-based analysis. Nevertheless, 2D-
PAGE analysis on this coomassie gel alone is suitable for proof-
of-concept analysis, and overlay with fluorescence gel is not
necessary in this study.
Figure 2. Demonstration of the CAPP strategy on a “prey-bait” model
system. (A) A covalent-linked GFP-Biotin complex containing a Se-
linker was created as a “prey-bait” model system: the Se-linker can be
first oxidative-cleaved to generate SA moiety, which can be subsequently
captured by tagging with an alkynyl-SA scavenger molecule suitable for
CuAAC-labeling with an azide-bearing fluorophore. (B) Detailed
procedure for CAPP strategy is as follows: (1) cleavage of the prey-
bait complex (GFP-Biotin) via a H2O2-mediated oxidative reaction and
(2) capturing the SA that contains prey proteins via (i) tagging the SA
moiety generated in situ with ADMA and (ii) labeling the transferred
ADMA tag with an azide that contains fluorophore. (C) Monitoring the
workflow of the CAPP procedure by LC-MS/MS. The MS data of the
following variants are shown: GFP-Se-Biotin, GFP-Se−OH, and GFP-Se-
ADMA together with the desired peptide sequence, the observed (ob)
and expected (ex) MW. (D) Structures of DMA and its alkynyl analogue
ADMA, with function group in red.
ifications.9 We used a copper-catalyzed azide−alkyne cyclo-
addition (CuAAC) reaction to conjugate a GFP that bears a site-
specifically incorporated alkyne handle with an azido-biotin
molecule carrying a γ-seleno substituted lysine linker at a molar
ratio of 1:4. ESI-MS analysis verified the resulting GFP-Se-Biotin
complex (Figures 2B,C and S14−15). H2O2-mediated cleavage
was then applied to this model system for 1 h before SDS-PAGE
being applied to the cleaved products. In order to successfully
monitor the cleavage process, we used in-gel LC-MS/MS
analysis to detect the desired SA-containing peptide from the
cleaved product GFP-Se-OH (Figures 2B,C and S16−17).
In order to directly capture the cleaved, SA-containing
proteins, we surveyed a series of aniline-based compounds that
have been previously used as SA scavengers during oxidative
elimination of the phenylselenide group (Figure S18).10 To our
delight, we found that 3,5-dimethoxyaniline (DMA) is an
efficient capture reagent that can specifically react with the SA
moiety generated in situ after oxidative cleavage of DiZSeK
(Figures 2D and S19).10 We then designed and synthesized an
alkyne-containing DMA analogue-ADMA (Figures 2D and S20).
To determine the reactivity of ADMA with SA generated in situ
from GFP, we applied the CAPP strategy on denatured GFP-Se-
Biotin by treating the protein sample simultaneously with 8 mM
H2O2 and 500 μM ADMA at 30 °C for 2 h (Figure 2B). LC-MS/
11862
dx.doi.org/10.1021/ja504371w | J. Am. Chem. Soc. 2014, 136, 11860−11863

Journal of the American Chemical Society
Communication
may become a invaluable method in identifying protein−protein
interactions in diverse living species.
ASSOCIATED CONTENT
* Supporting Information
Experimental details and supplemental data. This material is
available free of charge via the Internet at http://pubs.acs.org.
■
S
AUTHOR INFORMATION
Corresponding Author
pengchen@pku.edu.cn
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was supported by the National Key Basic Research
Foundation of China (2010CB912302 and 2012CB917301),
National Natural Science Foundation of China (21225206 and
91313301). We thank supports from Peking University Principal
Foundation. S.F. Reichard edited the manuscript.
Figure 3. 2D-PAGE analysis of the photo-cross-linked HdeA binding
proteins using DiZSeK and CAPP strategy on a coomassie gel. The
HdeA binding proteins after photo-cross-linking with the DiZSeK probe
were H2O2-cleaved, tagged with ADMA for 2D-PAGE analysis. LC-MS/
MS-analyzed spots are indicated in red circles.
REFERENCES
■
(1) (a) Tanaka, Y.; Bond, M. R.; Kohler, J. J. Mol. Biosyst. 2008, 4, 473−
480. (b) Preston, G. W.; Wilson, A. J. Chem. Soc. Rev. 2013, 42, 3289−
3301. (c) Pham, N. D.; Parker, R. B.; Kohler, J. J. Curr. Opin. Chem. Biol.
2013, 17, 90−101.
(2) (a) Wu, H.; Ge, J.; Yang, P.-Y.; Wang, J.; Uttamchandani, M.; Yao,
S. Q. J. Am. Chem. Soc. 2011, 133, 1946−1954. (b) Park, J.; Oh, S.; Park,
S. B. Angew. Chem., Int. Ed. 2012, 51, 5447−5451.
We randomly chose nine spots on the coomassie gel that
overlaid with the spots on fluorescent gel (Figure 3; additional
details regarding these binding proteins from each spot are in
Table S1). LC-MS/MS analysis identified 13 binding proteins
(Table S1, colored in orange) including 10 proteins (Table S2,
colored in blue) that have been previously reported as HdeA
clients, thus verifying the reliability of our method.5 Three new
proteins, PotF (putrescine-binding protein, residing in peri-
plasm), FliY (cystine-binding periplamic protein) and DhsB
(succinate dehydrogenase iron−sulfur subunit, residing on the
inner membrane), were identified for the first time as potential
HdeA binding proteins under acid stress (Table S2, colored in
red), which expanded our knowledge of the in vivo substrate
profiles for this acid chaperone. Taken together, the combination
of DiZSeK cleavable photo-cross-linker, CAPP strategy, and 2D-
PAGE offer a powerful tool for the systematic profiling of the
interaction protein targets of a given protein in living cells.
In summary, we have employed a pyrrolysine-based genetic
code expansion system to encode a Se-containing cleavable
protein photo-cross-linker and developed a cleavage-and-
capturing of interaction CAPP strategy. This cleavable photo-
affinity amino acid not only covalently traps prey proteins under
living conditions but also allows for the subsequent separation of
bait and prey proteins via H2O2-mediated oxidative cleavage. The
released prey proteins carrying the in situ generated SA moiety
could be further captured by (i) tagging with an alkyne-bearing
DMA molecule and (ii) labeling with an azide-containing
fluorophore or biotin probe. This CAPP strategy, in conjunction
with the 2D-PAGE proteomics and MS analysis, enhances the
separation and enrichment efficiency for identifying native
substrates of a given protein after photo-cross-linking. In this
study, we demonstrated this concept by profiling the in vivo
binding proteins of an E. coli acid chaperon HdeA under acid
stress, which revealed potential new substrates that have not been
identified from our previous work using noncleavable photo-
cross-linker. In addition, the strategy we developed here could be
used to improve the identification of the site of cross-linking on
the prey proteins with MS analysis in future. Given the general
applicability of this DiZSeK probe as well as the broad utilization
of the Pyl-based genetic-code expansion system, this strategy
(3) (a) Hino, N.; Okazaki, Y.; Kobayashi, T.; Hayashi, A.; Sakamoto,
K.; Yokoyama, S. Nat. Methods. 2005, 2, 201−206. (b) Hino, N.; Oyama,
M.; Sato, A.; Mukai, T.; Iraha, F.; Hayashi, A.; Kozuka-Hata, H.;
Yamamoto, T.; Yokoyama, S.; Sakamoto, K. J. Mol. Biol. 2011, 406,
343−353.
(4) (a) Cravatt, B. F.; Wright, A. T.; Kozarich, J. W. Annu. Rev. Biochem.
2008, 77, 383−414. (b) Leriche, G.; Chisholm, L.; Wagner, A. Bioorg.
Med. Chem. 2012, 20, 571−582. (c) Tamura, T.; Tsukiji, S.; Hamachi, I.
J. Am. Chem. Soc. 2012, 134, 2216−2226.
(5) (b) Zhang, M.; Lin, S.; Song, X.; Liu, J.; Fu, Y.; Ge, X.; Fu, X.;
Chang, Z.; Chen, P. R. Nat. Chem. Biol. 2011, 7, 671−677. (b) Lin, S.;
Zhang, Z.; Xu, H.; Li, L.; Chen, S.; Li, J.; Hao, Z.; Chen, P. R. J. Am.
Chem. Soc. 2011, 133, 20581−20587.
(6) (a) Xiang, Z.; Ren, H.; Hu, Y. S.; Coin, I.; Wei, J.; Cang, H.; Wang,
L. Nat. Methods 2013, 10, 885−888. (b) Grammel, M.; Hang, H. C. Nat.
Chem. Biol. 2013, 9, 475−484. (c) Chin, J. W. Annu. Rev. Biochem. 2014,
83, 379−408.
(7) Buchardt, O.; Elsner, H. I.; Nielsen, P. E.; Petersen, L. C.; Suenson,
E. Anal. Biochem. 1986, 158, 87−92.
(8) (a) Guo, J.; Wang, J.; Lee, J. S.; Schultz, P. G. Angew. Chem., Int. Ed.
2008, 47, 6399−6401. (b) Wang, Z. U.; Wang, Y.-S.; Pai, P.-J.; Russell,
W. K.; Russell, D. H.; Liu, W. R. Biochemistry 2012, 51, 5232−5234.
(9) Lin, S.; Yan, H.; Li, L.; Yang, M.; Peng, B.; Chen, S.; Li, W.; Chen, P.
R. Angew. Chem., Int. Ed. 2013, 52, 13970−13974.
(10) Tietze, L.-F.; v.Kiedrowski, G.; Berger, B. Tetrahedron. Lett. 1982,
23, 51−54.
(11) Wulfkuhle, J. D.; McLean, K. C.; Paweletz, C. P.; Sgroi, D. C.;
Trock, B. J.; Steeg, P. S.; Petricoin Iii, E. F. Proteomics 2001, 1, 1205−
1215.
11863
dx.doi.org/10.1021/ja504371w | J. Am. Chem. Soc. 2014, 136, 11860−11863