An engineered aryl azide ligase for site-specific mapping of protein-protein interactions through photo-cross-linking

Article abstract of DOI:10.1002/anie.200802088

(Chemical Equation Presented) Labeled and linked: The small-molecule binding site of Escherichia coli lipoic acid ligase was re-engineered to accept a fluorinated aryl azide probe in place of lipoic acid. Labeling with this mutant is highly specific for LAP fusion proteins. In cell lysate, FK506 binding protein was labeled and rapamycin-dependent photocross-linking to its interaction partner was demonstrated.

Full text of DOI:10.1002/anie.200802088

Communications
DOI: 10.1002/anie.200802088
Enzyme Engineering
An Engineered Aryl Azide Ligase for Site-Specific Mapping of
Protein–Protein Interactions through Photo-Cross-Linking**
Hemanta Baruah, Sujiet Puthenveetil, Yoon-Aa Choi, Samit Shah, and Alice Y. Ting*
Protein–protein interactions (PPIs) underlie all signaling
networks in cells, and thus, technologies for identifying and
characterizing them are extremely important. One powerful
approach to PPI detection consists of photo-cross-linking.[1,2]
A photo-cross-linker probe such as an aryl azide (Figure 1a)
or benzophenone is conjugated to a protein of interest, and
irradiation with UV light generates a reactive species that
covalently cross-links to any nearby protein (Figure 1b).
Purification of the cross-linked complexes and analysis by
mass spectrometry, Western blotting, or other techniques
reveal the identities of the interaction partners of the protein
of interest.
PPI detection by photo-cross-linking has considerable
advantages over other more commonly used PPI detection
methods, such as the yeast two-hybrid technique,[3] protein
complementation assays (based on green fluorescent pro-
tein,[4] split b-lactamase,[5] and split luciferase),[6,7] and fluo-
rescence resonance energy transfer (FRET).[8] For example,
photo-cross-linking can detect interactions with endogenous
proteins, whereas other methods detect only interactions with
recombinant reporter-fused proteins. The small size of typical
photo-cross-linking probes minimizes steric interference,
which reduces the frequency of false negatives. Transient
interactions can be captured. Finally, photo-cross-linking is
compatible with all cell types and all subcellular compart-
ments, in contrast to the yeast two-hybrid method, for
example.[9] Yet photo-cross-linking is rarely used in living
cells for PPI detection, principally because chemical photo-
cross-linkers cannot be easily targeted to specific proteins of
interest within the cellular context. The best current method
Figure 1. Photoaffinity probes and photo-cross-linking scheme.
a)Structures of aryl azide probes 1 and 2 and of lipoic acid.
for photo-cross-linker introduction is unnatural amino acid
mutagenesis.[10–17] The non-position-specific form of this
b)Illustration of site-specific aryl azide 1 ligation to a LAP fusion
technique[16,17] introduces high background, reduced signals,
protein, followed by photo-cross-linking to an interacting protein. ATP:
adenosine triphosphate, LplA: lipoic acid ligase, LAP: LplA acceptor
and potential interference with protein function. Position-
peptide.
[*] Dr. H. Baruah, Dr. S. Puthenveetil, Y.-A. Choi, Dr. S. Shah,
specific unnatural amino acid mutagenesis[10–15] is highly
Prof. A. Y. Ting
powerful but limited for mammalian cell applications by the
Department of Chemistry, Room 18-496
Massachusetts Institute of Technology
Cambridge, MA 02139 (USA)
prevalence of natural amber codons,[18] low suppression
efficiencies,[15,19] and the generation of prematurely truncated
Fax: (+1)617-253-7929
E-mail: ating@mit.edu
protein products, which can sometimes produce dominant
negative effects.[20] A simple, robust, high-efficiency, and
minimally invasive method for photo-cross-linker targeting in
cells is required, to bring the advantages of this approach for
PPI detection to cell biology.
As a first step towards this goal, we report herein the
design and characterization of an aryl azide ligase. To produce
this ligase, we re-engineered the active site of the Escherichia
coli enzyme lipoic acid ligase (LplA)[21] to accept a fluorinated
aryl azide photoaffinity probe and catalyze its covalent
[**] This work was supported by the US National Institutes of Health
(grant nos.: R01 GM072670 and P20 GM072029), the Sloan
Foundation, and the Dreyfus Foundation. Samsung Corporation
provided a Lee Kun Hee scholarship to Y.-A.C. We thank Susan M.
Lydon for her assistance with graphics. We thank Kathleen T. Xie
and Jackie Chan for help with site-directed mutagenesis experi-
ments and Marta Fernµndez-Suµrez for assistance and advice.
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/anie.200802088.
7018
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 7018 –7021

Angewandte
Chemie
conjugation to a recognition peptide (Figure 1b). The ligation
reaction is extremely sequence specific; endogenous mam-
malian proteins are not recognized and modified. We
demonstrate the utility of this method by performing site-
specific aryl azide ligation to FK506 binding protein (FKBP)
in mammalian cell lysate and then showing rapamycin-
dependent photo-cross-linking to the proteinꢀs interaction
partner, FKBP–rapamycin binding domain (FRB), also
within lysate.
E. coli LplA is a 38 kDa enzyme that catalyzes the ATP-
dependent ligation of the cofactor lipoic acid (Figure 1a) onto
a specific lysine side chain of one of its three endogenous
protein substrates in E. coli[21,22] or onto a 17-amino acid
unnatural peptide called “LplA acceptor peptide” (LAP).[23]
We previously harnessed LplA for site-specific fluorophore
tagging of LAP fusion proteins on the surface of living
mammalian cells[23] by capitalizing on 1) LplAꢀs extremely
high sequence specificity, which prevents it from recognizing
and modifying any endogenous mammalian proteins and
2) the malleability of LplAꢀs active site, which allows it to
accept and ligate alkyl azide analogues of lipoic acid. Azide-
derivatized cell-surface LAPs were selectively imaged with
cyclooctyne–fluorophore conjugates.[23]
hits, we then produced a second panel of LplA mutants,
shown in Figure 2a. We tested each of these using the same
HPLC assay, with LAP–HP1, and found that W37V LplA
exhibited the highest ligation activity for probe 1, followed by
W37S (Figure 2a). Under identical conditions (saturating
probe 1, ATP, and LAP–HP1, 30 min), W37V gave 20-fold
more product than did W37A, the best enzyme from the
original mutant panel (Figure 2a). It is interesting that W37V
LplA is superior to W37A LplA; we surmise that W37V
better complements the shape and size of probe 1, whereas
W37A may over-enlarge the LplA active site. We selected
W37V as our best mutant and used this clone for all
subsequent experiments.
For site-specific photo-cross-linker introduction, we syn-
thesized two fluorinated aryl azides as candidate substrates
for LplA (Figure 1a; syntheses shown in Figure 1 in the
Supporting Information). These structures represent a com-
promise between high photo-cross-linking efficiency,[1,2]
longer activation wavelength, and small probe size to fit
better within the LplA active site. Using an HPLC assay[23]
with either E2p (a hybrid lipoyl domain derived from one of
LplAꢀs natural protein substrates[22]) or LAP–HP1 fusion[23]
as the protein substrate, we tested the ability of wild-type
LplA to utilize probes 1 and 2. No product conjugate was
detected in the reactions with either probe (data not shown).
Aryl azide probes 1 and 2 are both considerably larger
than lipoic acid or the alkyl azide used in our previous
study,[23] so we attempted to use mutagenesis to increase the
volume of the LplA active site. The cocrystal structure of
E. coli LplA with lipoic acid has been reported,[24] but a
number of unusual features about this structure led us to
examine instead the cocrystal structure of Thermoplasma
acidophilum LplA with lipoyl-adenosine monophosphate[25]
(Figure 2 in the Supporting Information). The structure shows
nine amino acid side chains (L18, D21, Y39, R72, Y73, T74,
H81, L159, and H161) that lie within 7.5 Š of the dithiolane
ring of lipoic acid. We aligned the E. coli and T. acidophilum
LplA protein sequences and individually mutated each of the
corresponding amino acids in the E. coli enzyme (L17, E20,
W37, R70, S71, S72, H79, F147, and H149) into alanine in
order to decrease their size. HPLC testing of this initial panel
of nine LplA mutants showed that W37A, E20A, and S71A
mutants could accept probe 1 to a small degree (Figure 3 in
the Supporting Information). Under identical reaction con-
ditions, W37A gave more product than either E20A or S71A
(Figure 3 in the Supporting Information). No product was
detected for probe 2 with any of the mutants, despite the fact
that it is smaller than probe 1; this indicates that factors other
than size play a role in substrate recognition. Based on these
Figure 2. Engineering of the aryl azide ligase and characterization of
specificity. a)Second-generation LplA mutants screened for aryl azide
1 ligation activity. Wild-type LplA and six mutants were compared
under identical reaction conditions (800 mm aryl azide 1, 400 nm
enzyme, 308C for 30 min), and product (aryl azide–LAP–HP1 conju-
gate)was detected by HPLC. The conversion into product obtained
with W37V LplA was normalized to 100% and other conversion values
are reported relative to this. All measurements were performed in
triplicate. b)HPLC trace showing ligation of aryl azide 1 to LAP–HP1
protein catalyzed by W37V LplA. The results of negative controls with
ATP omitted or W37V replaced by wild-type LplA are also shown. The
eluents of the two starred peaks were purified and analyzed by mass
spectrometry (see Figure 4 in the Supporting Information). c) Aryl
azide ligation performed on mammalian cell lysate to test labeling
specificity. Lysate was generated from HEK cells expressing LAP–CFP
(37 kDa). Lysate was labeled in vitro with LplA and aryl azide 1, then
incubated with phosphine–Flag conjugate to derivatize the aryl azide.
Blotting with anti-Flag antibody is shown to the right of the Coomas-
sie-stained gel. Negative controls are shown with an alanine mutation
in LAP (lane 2), wild-type LplA instead of W37V (lane 4), and untrans-
fected cells (lane 3).
Angew. Chem. Int. Ed. 2008, 47, 7018 –7021
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7019

Communications
We characterized the W37V-catalyzed ligation of aryl
azide 1 in terms of kinetics and specificity. First, we repeated
the HPLC assay, with negative controls in which ATP was
omitted or W37V was replaced by wild-type LplA. As
expected, product was only observed in the presence of
ATP and the W37V mutant (Figure 2b). We collected the
HPLC-purified product and characterized it by mass spec-
trometry. Figure 4 in the Supporting Information shows that
the mass corresponds to the calculated mass for one molecule
of aryl azide 1 plus one LAP–HP1 protein, minus water. Next,
we measured the kinetics of ligation. Using HPLC as readout,
we determined a rate of catalysis kcat = (0.31 Æ 0.04) sÀ1, which
surprisingly is even faster than the natural ligation reaction
(lipoylation of E2p: kcat = 0.25 sÀ1 [23]
Figure 5 in the Support-
;
ing Information). Separate measurements showed that the
Michaelis constant Km was 80 mm or less (data not shown),
which is considerably greater than the Km value of wild-type
lipoic acid for LplA (1.7 mm[21] or 4.5 mm[24]).
We were concerned that remodeling of the LplA active
site might change the peptide specificity of the enzyme.
Therefore, we characterized the specificity of W37V-cata-
lyzed aryl azide ligation in three ways. First, we prepared a
total lysate from HEK cells expressing a LAP fusion with
cyan fluorescent protein (LAP–CFP[23]). We labeled the lysate
under forcing conditions with aryl azide, ATP, and W37V
LplA, and then we derivatized the aryl azide with phosphine–
Flag conjugate by using the Staudinger ligation.[26] Western
blotting with anti-Flag antibody showed that, even though the
LAP–CFP expression level is so low that it cannot be detected
above endogenous proteins in the Coomassie-stained gel,
only LAP–CFP receives the aryl azide, and the endogenous
proteins in the lysate are not modified by W37V LplA
(Figure 2c). Negative controls showed that labeling is specific
for LAP and dependent on the W37V mutation (Figure 2c).
For our second specificity test, we performed ligation with
aryl azide 1 on three non-LAP proteins: FRB-AP, FKBP-
AP,[27] and BCCP-87.[28] We characterized the products by
mass spectrometry (data not shown). None were found to be
modified with aryl azide, despite exposure to reaction
conditions that gave complete conversion of LAP–HP1 into
the aryl azide conjugate.
Third, we tested peptide specificity by using a live cell
imaging assay. A useful feature of the aryl azide probe is that
it not only mediates photo-cross-linking but can also serve as
a functional-group handle for the introduction of fluorescent
probes. Although it is more common to derivatize alkyl azides
with alkynes through [3+2] cycloaddition[23,29] or with phos-
phines through the Staudinger ligation,[30–32] we and other
groups have observed facile reactions of aryl azides with both
cyclooctynes[33] and phosphines.[30–32] To make use of this
reactivity, we expressed a LAP–CFP–TM (transmembrane)
cell-surface construct[23] in HeLa cells and performed extrac-
ellular labeling with W37V LplA, aryl azide, and ATP. After
washing away excess aryl azide, we derivatized the LAP-
ligated azide with cyclooctyne–cyanine 3 (Cy3) conjugate.[23]
Imaging showed that only transfected (CFP-positive) cells
became labeled with Cy3, while neighboring untransfected
cells did not receive Cy3 (Figure 3). Negative controls with
aryl azide omitted from the labeling reaction or with LAP–
Figure 3. Specific fluorophore labeling on living cells through aryl
azide 1 ligation. HeLa cells expressing LAP–CFP–TM were labeled with
W37V LplA and aryl azide 1 for 40 min and, subsequently, with
cyclooctyne–Cy3 conjugate[22] for 20 min, to selectively derivatize the
aryl azide. Imaging shows pink Cy3 staining on the membranes of
transfected CFP-positive cells. Negative controls are shown with
omission of aryl azide (middle row)or with LAP–CFP–TM replaced by
an alanine mutant (bottom row). The CFP and Cy3 images are
merged. DIC: differential interference contrast.
CFP–TM replaced by an alanine point mutant in the LAP
sequence gave no fluorophore labeling (Figure 3). Thus, our
three assays collectively demonstrate that aryl azide ligation
has the same high sequence specificity as lipoylation catalyzed
by wild-type LplA.
To demonstrate the utility of the new aryl azide ligase, we
attempted to detect a known PPI within a complex mixture.
FKBP is known to interact with FRB in the presence but not
absence of the small molecule rapamycin.[34] We prepared an
FKBP–LAP construct and expressed it in HEK cells. In
separate HEK cells, we expressed a CFP–FRB fusion. Total
lysates were prepared from both samples, and they were
mixed. We labeled the combined lysates with aryl azide 1,
ATP, and W37V LplA, before photo-cross-linking with 300–
360 nm light in the presence or absence of rapamycin. We
analyzed the samples by sodium dodecylsulfate polyacryl-
amide gel electrophoresis (SDS PAGE) and detected CFP–
FRB by in-gel CFP fluorescence (Figure 4). Only imaging in
the presence of rapamycin with UV light applied and the LAP
tag intact did we observe a higher-molecular-weight band
corresponding to the covalent FKBP–FRB heterodimer.
Negative controls with either construct omitted gave no
heterodimer band.
In summary, we have engineered an aryl azide ligase that
brings us closer to the dream of routine PPI detection inside
living cells by photo-cross-linking. Our ligase, generated by
active-site engineering of E. coli LplA, catalyzes the
extremely sequence-specific covalent ligation of a fluorinated
aryl azide probe onto LAP fusion proteins, expressed on
7020 www.angewandte.org
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2008, 47, 7018 –7021

Angewandte
Chemie
[2] J. Brunner, Annu. Rev. Biochem. 1993, 62, 483 – 514.
[3] J. R. Parrish, K. D. Gulyas, R. L. Finley, Jr., Curr. Opin. Bio-
technol. 2006, 17, 387 – 393.
[4] C. D. Hu, T. K. Kerppola, Nat. Biotechnol. 2003, 21, 539 – 545.
[5] A. Galarneau, M. Primeau, L. E. Trudeau, S. W. Michnick, Nat.
Biotechnol. 2002, 20, 619 – 622.
[6] K. E. Luker, M. C. Smith, G. D. Luker, S. T. Gammon, H.
Piwnica-Worms, D. Piwnica-Worms, Proc. Natl. Acad. Sci. USA
2004, 101, 12288 – 12293.
[7] E. Stefan, S. Aquin, N. Berger, C. R. Landry, B. Nyfeler, M.
Bouvier, S. W. Michnick, Proc. Natl. Acad. Sci. USA 2007, 104,
16916 – 16921.
[8] K. Truong, M. Ikura, Curr. Opin. Struct. Biol. 2001, 11, 573 – 578.
[9] W. van Criekinge, R. Beyaert, Biol. Proced. Online 1999, 2, 1 –
38.
[10] J. W. Chin, S. W. Santoro, A. B. Martin, D. S. King, L. Wang, P. G.
Schultz, J. Am. Chem. Soc. 2002, 124, 9026 – 9027.
[11] J. W. Chin, A. B. Martin, D. S. King, L. Wang, P. G. Schultz, Proc.
Natl. Acad. Sci. USA 2002, 99, 11020 – 11024.
[12] E. M. Tippmann, W. Liu, D. Summerer, A. V. Mack, P. G.
Schultz, ChemBioChem 2007, 8, 2210 – 2214.
[13] N. Hino, Y. Okazaki, T. Kobayashi, A. Hayashi, K. Sakamoto, S.
Yokoyama, Nat. Methods 2005, 2, 201 – 206.
[14] S. Ye, C. Kohrer, T. Huber, M. Kazmi, P. Sachdev, E. C. Yan, A.
Bhagat, U. L. RajBhandary, T. P. Sakmar, J. Biol. Chem. 2007,
283, 1525 – 1533.
[15] W. Wang, J. K. Takimoto, G. V. Louie, T. J. Baiga, J. P. Noel, K. F.
Lee, P. A. Slesinger, L. Wang, Nat. Neurosci. 2007, 10, 1063 –
1072
[16] M. Suchanek, A. Radzikowska, C. Thiele, Nat. Methods 2005, 2,
261 – 267.
Figure 4. Aryl azide mediated photo-cross-linking of FKBP and FRB in
mammalian cell lysate. Lysate containing FKBP–LAP and CFP–FRB was
labeled with W37V LplA and aryl azide 1. Photo-cross-linking was then
performed with 300–360 nm light for 4 min, in the presence (lane 3)or
absence (lane 4)of rapamycin. The samples were run on SDS PAGE
gels, and cross-linked FKBP-FRB heterodimer was detected by CFP
imaging (right-hand gel). Negative controls are shown with no UV
irradiation (lane 1), an alanine mutation in FKBP–LAP (lane 2), FKBP–
LAP not expressed (lane 5), or CFP–FRB not expressed (lane 6).
Coomassie visualization of the same samples (left-hand gel)shows
equal loading in all lanes. Note that proteins run differently in the CFP
gel because they are not denatured, in order to preserve CFP
fluorescence.
[17] K. Kirshenbaum, I. S. Carrico, D. A. Tirrell, ChemBioChem
2002, 3, 235 – 237.
[18] W. Liu, A. Brock, S. Chen, S. Chen, P. G. Schultz, Nat. Methods
2007, 4, 239 – 244.
[19] L. Wang, J. Xie, P. G. Schultz, Annu. Rev. Biophys. Biomol.
Struct. 2006, 35, 225 – 249.
[20] T. Lu, A. Y. Ting, J. Mainland, L. Y. Jan, P. G. Schultz, J. Yang,
Nat. Neurosci. 2001, 4, 239 – 246.
living cell surfaces or in total lysate. Our immediate goal is to
extend this ligation reaction to the cytosol of living cells.
In addition to photo-cross-linking, the aryl azide ligation
reaction can also be used for fluorophore tagging, as previously
demonstrated with an alkyl azide analogue of lipoic acid.[23]
The aryl azide undergoes facile reaction with cyclooctyne–
fluorophore conjugates on the surface of living cells.
[21] D. E. Green, T. W. Morris, J. Green, J. E. Cronan, Jr., J. R.
Guest, Biochem. J. 1995, 309, 853 – 862.
This study demonstrates the malleability of the LplA
active site. Remarkably, the small-molecule binding site can
be made to accommodate a variety of probe structures,
without sacrificing the extremely high peptide/protein specif-
icity that makes this enzyme such a useful tool for site-specific
labeling applications. We were surprised to find that a single
amino acid mutation (W37V) opened up the LplA active site
to a structure quite dissimilar to lipoic acid. Not all enlarge-
ments of the lipoate binding pocket had equivalent effects,
however; W37A was a much poorer catalyst for aryl azide 1
ligation than W37V. We were also not successful in finding
any catalysts for aryl azide 2, despite its smaller size. These
lessons will guide us in our continued efforts to engineer LplA
for biotechnological applications.
[22] S. T. Ali, J. R. Guest, Biochem. J. 1990, 271, 139 – 145.
[23] M. Fernꢁndez-Suꢁrez, H. Baruah, L. Martꢂnez-Hernꢁndez, K. T.
Xie, J. M. Baskin, C. R. Bertozzi, A. Y. Ting, Nat. Biotechnol.
2007, 25, 1483 – 1487.
[24] K. Fujiwara, S. Toma, K. Okamura-Ikeda, Y. Motokawa, A.
Nakagawa, H. Taniguchi, J. Biol. Chem. 2005, 280, 33645 – 33651.
[25] D. J. Kim, K. H. Kim, H. H. Lee, S. J. Lee, J. Y. Ha, H. J. Yoon,
S. W. Suh, J. Biol. Chem. 2005, 280, 38081 – 38089.
[26] K. L. Klick, E. Saxon, D. A. Tirrell, C. R. Bertozzi, Proc. Natl.
Acad. Sci. USA 2002, 99, 19 – 24.
[27] M. Fernandez-Suarez, T. S. Chen, A. Y. Ting, J. Am. Chem. Soc.
2008, 130, 9251 – 9253.
[28] A. Chapman-Smith, D. L. Turner, J. E. Cronan, Jr., T. W. Morris,
J. C. Wallace, Biochem. J. 1994, 302, 881 – 887.
[29] M. A. Baskin, J. A. Prescher, S. T. Laughlin, N. J. Agard, P. V.
Chang, I. A. Miller, A. Lo, J. A. Codelli, C. R. Bertozzi, Proc.
Natl. Acad. Sci. USA 2007, 104, 16793 – 16797.
[30] F. L. Lin, H. M. Hoyt, H. van Halbeek, R. G. Bergman, C. R.
Bertozzi, J. Am. Chem. Soc. 2005, 127, 2686 – 2695.
[31] S. J. Luchansky, S. Goon, C. R. Bertozzi, ChemBioChem 2004, 5,
371 – 374.
Received: May 3, 2008
Published online: August 4, 2008
Keywords: ligases · photo-cross-linking · protein engineering ·
.
protein interactions · site-specific labeling
[32] J. A. Restituyo, L. R. Comstock, S. G. Petersen, T. Stringfellow,
S. R. Rajski, Org. Lett. 2003, 5, 4357 – 4360.
[33] G. Wittig, A. Krebs, Chem. Ber. 1961, 94, 3260 – 3275.
[34] L. A. Banaszynski, C. W. Liu, T. J. Wandless, J. Am. Chem. Soc.
2005, 127, 4715 – 4721.
[1] Y. Tanaka, M. R. Bond, J. J. Kohler, Mol. Biosyst. 2008, 4, 473 –
480.
Angew. Chem. Int. Ed. 2008, 47, 7018 –7021
ꢀ 2008 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org
7021