Click strategies for single-molecule protein fluorescence

Article abstract of DOI:10.1021/ja210587q

Single-molecule methods have matured into central tools for studies in biology. Foerster resonance energy transfer (FRET) techniques, in particular, have been widely applied to study biomolecular structure and dynamics. The major bottleneck for a facile and general application of these studies arises from the need to label biological samples site-specifically with suitable fluorescent dyes. In this work, we present an optimized strategy combining click chemistry and the genetic encoding of unnatural amino acids (UAAs) to overcome this limitation for proteins. We performed a systematic study with a variety of clickable UAAs and explored their potential for high-resolution single-molecule FRET (smFRET). We determined all parameters that are essential for successful single-molecule studies, such as accessibility of the probes, expression yield of proteins, and quantitative labeling. Our multiparameter fluorescence analysis allowed us to gain new insights into the effects and photophysical properties of fluorescent dyes linked to various UAAs for smFRET measurements. This led us to determine that, from the extended tool set that we now present, genetically encoding propargyllysine has major advantages for state-of-the-art measurements compared to other UAAs. Using this optimized system, we present a biocompatible one-step dual-labeling strategy of the regulatory protein RanBP3 with full labeling position freedom. Our technique allowed us then to determine that the region encompassing two FxFG repeat sequences adopts a disordered but collapsed state. RanBP3 serves here as a prototypical protein that, due to its multiple cysteines, size, and partially disordered structure, is not readily accessible to any of the typical structure determination techniques such as smFRET, NMR, and X-ray crystallography.

Full text of DOI:10.1021/ja210587q

Article
pubs.acs.org/JACS
Click Strategies for Single-Molecule Protein Fluorescence
Sigrid Milles, Swati Tyagi, Niccolo Banterle, Christine Koehler, Virginia VanDelinder, Tilman Plass,
̀
Adrian P. Neal, and Edward A. Lemke*
EMBL, Structural and Computational Biology Unit, Meyerhofstrasse 1, 69117 Heidelberg, Germany
S
* Supporting Information
ABSTRACT: Single-molecule methods have matured into
central tools for studies in biology. Foerster resonance energy
transfer (FRET) techniques, in particular, have been widely
applied to study biomolecular structure and dynamics. The
major bottleneck for a facile and general application of these
studies arises from the need to label biological samples site-
specifically with suitable fluorescent dyes. In this work, we
present an optimized strategy combining click chemistry and
the genetic encoding of unnatural amino acids (UAAs) to
overcome this limitation for proteins. We performed a systematic study with a variety of clickable UAAs and explored their
potential for high-resolution single-molecule FRET (smFRET). We determined all parameters that are essential for successful
single-molecule studies, such as accessibility of the probes, expression yield of proteins, and quantitative labeling. Our
multiparameter fluorescence analysis allowed us to gain new insights into the effects and photophysical properties of fluorescent
dyes linked to various UAAs for smFRET measurements. This led us to determine that, from the extended tool set that we now
present, genetically encoding propargyllysine has major advantages for state-of-the-art measurements compared to other UAAs.
Using this optimized system, we present a biocompatible one-step dual-labeling strategy of the regulatory protein RanBP3 with
full labeling position freedom. Our technique allowed us then to determine that the region encompassing two FxFG repeat
sequences adopts a disordered but collapsed state. RanBP3 serves here as a prototypical protein that, due to its multiple cysteines,
size, and partially disordered structure, is not readily accessible to any of the typical structure determination techniques such as
smFRET, NMR, and X-ray crystallography.
strategies exist to economically generate labeled nucleic acid
derivatives, site-specific labeling of proteins represents the
major bottleneck to extend the applications of smFRET
technology in biology and structural biology in particular.5 In
addition, smFRET necessitates exacting labeling methods
compared to conventional ensemble fluorescence approaches,
where low labeling yields can typically be compensated for by
increasing the concentration of the sample. When possible,
labeling of cysteine residues of proteins is a simple and efficient
strategy, but is frequently not viable due to the natural
abundance of this amino acid.
A plethora of chemical biology tools have been developed to
site-specifically label proteins.6−9 However, many of those are
limited to the termini or require introduction of larger
recognition elements, such as small peptide tags, into the
protein. It was recently demonstrated that genetically encoding
the unnatural amino acid (UAA) p-acetylphenylalanine (1) in
Escherichia coli by means of an engineered AMBER suppressor
tRNA/synthetase (tRNA/RS) system from Methanocaldococcus
jannaschii can provide an efficient labeling site for hydroxyl-
amine derivatives of single-molecule dyes.5 However, this
strategy has the major disadvantage that the oxime ligation
optimally proceeds in acidic conditions, rendering it
1. INTRODUCTION
Single-molecule (sm) methods are especially attractive and
powerful for the study of complex biological and chemical
systems and processes.1 In contrast to conventional ensemble
experiments, in single-molecule experiments, molecular proper-
ties are measured one molecule at a time and distributions in
molecular behaviors are measured directly. This is particularly
important when analyzing dynamic systems that cannot be
easily synchronized. Single-molecule observation of such
systems makes it possible to directly determine the connectivity
between different states of complex biomolecular reactions
rather than obtaining a view of the average behavior. Among
single-molecule techniques, the ability to measure distances and
dynamics using Foerster resonance energy transfer (FRET) has
become a powerful tool in structural biology.1−3 This impact is
mainly due to the fact that FRET between a suitable donor dye
and a suitable acceptor dye can be measured efficiently between
30 and 100 Å, and thus falls into size range of most molecular
machines and their conformational changes. The application of
single-molecule FRET (smFRET) technology has been boosted
in recent decades by multiple technological advances that
include optimized hardware, software, and data analysis.1,2,4
However, site-specific labeling with a suitable dye pair, the key
step for all smFRET studies, is still problematic for almost all
proteins. Although typically not a major problem for DNA or
RNA studies, for which many synthetic and semisynthetic
Received: November 10, 2011
Published: February 22, 2012
© 2012 American Chemical Society
5187
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
unattractive for use with many biological systems.10 In contrast,
click chemistry between an azide and a terminal alkyne is a
widely used biocompatible reaction.11 Its biggest advantages are
that the reaction is fast, viable at physiological conditions, and
independent of pH over a wide range. Here we present a
strategy that utilizes genetically encoded clickable UAAs to
make proteins accessible to single-molecule studies, with a
focus on the specific demands necessary for smFRET, as a
precise structural biology tool.
To this end, we performed a quantitative analysis of a range
of clickable UAAs. We first analyzed the UAAs for highly
important practical considerations that are critical to judge the
potential for wide applicability for structural biology studies,
including expression yield in different proteins using various
expression systems, and commercial availability, and/or ease of
synthesis. We systematically labeled a set of disordered and
structured proteins with different incorporated UAAs and then
performed smFRET multiparameter analysis of the samples,
which differ from each other by only the choice of UAA.
Intrinsically disordered proteins, which account for a large
fraction of the eukaryotic proteome, were specifically included
in this study, as they are typically more challenging proteins to
express due to their lack of stable structure and high sensitivity
toward proteases.12 Furthermore, smFRET is one of the few
techniques that can actually provide insights into IDP
function,3 which allows us to present a technology that goes
clearly beyond demonstrating idealized model systems.
2. MATERIALS AND METHODS
2.1. Protein Expression and Purification. For quantitative
expression tests, a GFP with an AMBER mutation (i.e., mutation to a
TAG stop codon) at position 39 and an Maltose Binding Protein
(MBP) with an AMBER mutation at position 11, both under the
control of a pBAD promoter, were used following the procedure
described by Plass et al.16 For labeling studies, a single-cysteine-
containing T4Lysozyme (T4L) mutant gene with an AMBER
mutation at position 38 was used.5 A single cysteine mutant of the
hNup153 domain (residues 875−1475) with an AMBER mutation at
position 938 was used analogously to the procedure described in
Milles et al.17 The intrinsically disordered protein (IDP) domain of the
Saccharomyces cerevisiae Nup49,18,19 residues 1−260, was expressed
and purified as an intein-6His fusion construct. Similarly to T4L and
hNup153, the yNup49FG construct contained a unique cysteine at
position 250 and an AMBER TAG mutation at site 191 (Figure 1a).
Highly precise measurements of FRET at the single-molecule
level require that the dyes are attached via flexible, long linkers
such that the dipole interaction, from which the FRET arises,
averages isotropically. At the same time, long linkers result in a
relatively large distance between the dye and the protein
backbone. Correlating the measured smFRET distance with
actual protein backbone distance is thus still a major
problem,3,13−15 and has been studied mainly on model systems
(synthetic small DNA, RNA, or polyproline) that are also
accessible to complementary computational methods. There-
fore, even precise smFRET measurements are currently mainly
applicable for relative comparison. Genetically encoding
different UAAs and linking dyes to them also now generates
an unexplored plethora of chemically and structurally diverse
linkers. Using multiparameter smFRET technology, that is,
recording fluorescence lifetimes, intensities, and anisotropies at
the same time, we were able to access the individual
consequences of the various UAA-dye linkages. Our results
show that structure and size of the linker can affect the
smFRET measurement in an unexpected way. Our data allow
the individual effects to be distinguished and highlight the
potential pitfalls when applying state-of-the-art protein
engineering methods to smFRET experiments.
Overall, this work provides a resource on the advantages of
different labeling systems for smFRET measurments, with a
focus on biochemical considerations and effects on high-
precision measurements in structural biology using smFRET. In
fact, we were able to build upon the new strategy to show how
a key regulatory protein of the nuclear transport pathway can
now be studied with smFRET technology. Because of its amino
acid composition, size, and complex structure, RanBP3 is
typically not amenable to structural biology tools such as
smFRET, NMR, or X-ray determination.
Figure 1. Proteins (a) and UAAs (b) used in this study for smFRET
measurements.
T4L and yNup49 were cloned into pBad vectors and hNup153 into
a pTXB3 vector. All vectors contained an ampicillin resistance gene
and were cotransformed with plasmids harboring the tRNA/RS system
for the respective UAA into E. coli Bl21(AI) (Invitrogen, Carlsbad,
CA). For selective expression with 1, 2 (p-azidophenylalanine), and 3
(p-(propargyloxy)-phenylanlanine) (Figure 1b), the recently reported
optimized mutant variants of the pEvol system from the tyrosine
synthetase of M. jannaschii were used, and termed pEvol-pAcF -pAzF
and -PrF, for encoding compounds 1, 2, and 3, respectively.20
For expression with 4 and 5,21,22 the recently introduced analogous
pEvol system of the pyrrolysine tRNA/RS system from Methanosarci-
na mazei was used, termed pEvol PylRS.16 Expression and purification
of all proteins was performed as previously described for Nup153,17
GFP,16 and T4L, but more details and composition of standard
purification buffers are given also in the Supporting Information
Figures S1 and S2.23,24 In short, expression was done in terrific broth
(TB) medium at 37 °C. At OD600 = 0.2−0.5, 1 mM of the UAA was
added. The cells were induced at OD600 = 0.4−1 with 0.02% arabinose
and additional 1 mM IPTG in case of hNup153 and incubated until
harvesting. Depending on the constructs, Ni and/or chitin beads were
used for the purification in phosphate buffered saline (PBS, pH 7.4),
and for the IDPs with 2 M urea following the manufacturer’s protocols
for Ni (Qiagen, Dusseldorf, Germany) and chitin (NEB, Ipswich, MA)
beads, respectively. All purification buffers contained 0.2 mM tris(2-
carboxyethyl)phosphine (TCEP) and 1 mM phenylmethanesulfonyl-
fluoride (PMSF).
The gene encoding for RanBP3 was cloned into a pTXB1 vector as
an intein-6His fusion construct and expressed in Bl21(AI) (Invitrogen,
Carlsbad, CA) cotransformed with the pEvol PylRS to genetically
encode 5 twice at the AMBER sites 181 and 294. Cultures were grown
in TB medium at 37 °C and induced at OD600 = 0.4−0.6 with
5188
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
arabinose and IPTG. Bacteria were harvested and purified under native
conditions according to standard purification protocols for His-tagged
proteins.
in 2D τ versus EFRET and r versus τ histograms. The populations in
these histograms of several hundred events were fit with 2D Gaussian
functions, which determine EFRET, r and τ for the respective
population. The Perrin curve, described by the equation
Expression efficiency using Amber suppression is not only UAA
dependent, but also depends on the incorporation site as well as on the
protein itself (see Figure 2 for comparison between relative GFP and
MBP expressions). Typical expression yields for 1 L of expression
cultures were >20 mg for 1, ∼6 mg for 2, ∼10 mg for 3, ∼13 mg for 4,
and ∼12 mg 5 incorporated into GFP. For hNup153, which is very
difficult to expresses due to its large size and intrinsically disordered
structure, we typically obtained ∼1 mg (1), >0.5 mg (3), and >0.5 mg
(5); for yNup49, 8 mg (1), >1 mg (3), and >2 mg (5); and for T4L,
>20 mg (1), >2 mg (3), and >10 mg (5), respectively.
⎛
⎞
⎟
τ
r = r / 1 +
⎜
0
⎝
ρ ⎠
with the fundamental anisotropy r0 = 0.4 and the rotational correlation
coefficient ρ was plotted onto the r versus τ histogram to visualize the
behavior of these two parameters in an ideal FRET system. 2D
stoichiometry histograms were plotted from EFRET and the
stoichiometry
2.2. Site-Specific Dual Labeling. For quantitative comparison of
UAA and dye linker effects on the quality of single-molecule
measurements, all proteins (incorporating 1, 3, and 5) were prepared
in the same manner with Alexa 594 attached to equivalent cysteines
using standard maleimide chemistry.24 The ketone groups of proteins
containing 1 were labeled under denaturing conditions (4 M
guanidinium chloride) at pH 4 in sodiumacetate buffer with a
hydroxylamine derivative of Alexa 488 as previously reported.5,23,24
Labeling of the alkyne functionality (proteins containing 3 and 5,
respectively) with commercially available azide derivatives of the Alexa
488 fluorophore was performed using copper-catalyzed alkyne−azide
cycloaddition reaction (CuAAC). The exact labeling conditions, as
well as kinetics and yields are given in detail in Supporting Information
Figures S1 and S2.
RanBP3 was labeled using a 5 molar excess of Alexa 488 azide and 6
molar excess of Alexa 594 azide dyes in a single reaction pot. The
reaction was quenched with 5 mM EDTA. All proteins were further
purified on a Superdex-200. Concentrations were determined by UV−
vis spectroscopy and/or BCA assay (Fisher Scientific, Rockford, IL)
and labeling efficiencies were above 70% per labeling site (Supporting
Information Figure S2).
2.3. Multiparameter Single-Molecule Fluorescence Spec-
troscopy. For smFRET experiments, the doubly labeled proteins
were diluted to a final concentration of 50−200 pM in PBS, 2 mM
dithiothreitol (DTT), pH 7.4, and measured on a custom built
multiparameter (MP) spectrometer centered around an Olympus
IX81 microscope (Hamburg, Germany) equipped with a high
numerical water objective (60×, 1.2 NA) as previously described.17
A laser diode (LDH 485, Picoquant, Berlin, Germany) and a white
light laser (SuperK Extreme, NKT Photonics, Denmark) filtered
through a z578/10 excitation filter (Chroma, Olching, Germany),
which were pulsed alternately at 80 MHz total, were used to excite
freely diffusing labeled proteins with linearly polarized light.
Fluorescence originating from single molecules was first spatially
filtered by a 100 μm pinhole and then split into parallel and
perpendicular polarization directions before separation into green
donor (D) and red acceptor (A) fluorescence light. D and A
fluorescence was detected with micro photon counting devices (MPDs
for the green (Picoquant, Berlin, Germany) and avalanche photo
diodes for the red (Perkin-Elmer, Waltham, MA). Acquired data were
subjected to multiparameter fluorescence analysis.25,26 After burst
recognition, a threshold of 100 photons was applied over ID and IA.
Only bursts below a maximum brightness of 600 photons and a
maximum length of 8 ms were considered for data evaluation. Foerster
resonance energy transfer,
I
+ I
D
A
S =
dir
A
I
+ I + I
D
A
with IAdir being the signal from the directly excited acceptor molecules
using orange excitation light.27,28
All data sets were measured in triplicates and corrected for leakage
of D fluorescence into the A channel. Single-molecule data were
processed and analyzed with a custom written code in IgorPro
(Wavemetrics, Lake Oswego, OR). Error bars represent standard
deviations unless noted otherwise.
2.4. Distance Determination from Multiparameter Single-
Molecule Fluorescence Spectroscopy. Supporting Information
Table S2 lists the detailed protocol of how the distances were
calculated from MP smFRET data. The FRET distances were
calculated from the fluorescence lifetimes as lifetime-based FRET is
less affected by measurement artifacts than intensity-based FRET,4
with the equation
τ
τ
DA
E
= 1 −
FRET
τ
D(0)
where τD(0) and τDA are the fluorescence lifetimes of the donor only
population and of the FRET-species. Since yNup49 and hNup153 are
intrinsically disordered proteins and thus highly flexible, all their FRET
efficiencies underlie fast fluctuations and can therefore be described
according to the equation
∞
⟨E⟩ =
E(r)P(r) dr
∫
0
where
τ
⟨E⟩ = E
FRET
and the radial probability distribution of a Gaussian chain (see
Supporting Information).17,29−33
3/2
2
⎛
⎝
⎞
⎛
⎞
⎠
3
3r
2
⎜
⎜
⎟
⎜
exp −
⎜
⎟
⎟
P(r) = 4πr
⎟
2
2
2π⟨r ⟩
2⟨r ⟩
⎠
⎝
End-to-end distance values RE = (⟨r2⟩)1/2 satisfying this equation
were determined numerically using IgorPro (Wavemetrics, Lake
Oswego, OR).
3. RESULTS
3.1. Choice of System and Unnatural Amino Acids. To
date, more than 80 UAA have been genetically encoded into E.
coli.34 A few of those offer unique chemical handles that can be
reacted with commercially available derivatives of single-
molecule-suitable dyes. Commercial availability or easy
accessibility are essential criteria, since only very few
fluorophores actually fulfill the high demands of single-
molecule studies, specifically high photostability, low blinking,
and ideally single exponential lifetime decay kinetics. If not
commercially available, they are typically only accessible to
particularly skilled chemistry laboratories. For copper-catalyzed
alkyne−azide click (CuAAC) chemistries, genetically encoded
I
A
E
=
FRET
I
+ I
A
D
and fluorescence anisotropies of D,
I − GI
I + 2GI
⊥
r =
⊥
were determined for every data set as described in Milles et al.17 IA and
ID describe the A and D intensities, respectively. I∥ and I⊥ are parallel
and perpendicular polarized light intensities and G is the G-factor of
the instrument. τ, r, and EFRET were summed up burst-wise and plotted
5189
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
alkyne and azide functionalities are most suitable. The M.
jannaschii system that was used for incorporation of 1 has
additionally been used for the genetic encoding of compounds
2 and 3.35,36 The structure of all those amino acids contains an
aromatic, rather stiff phenylalanine derivative side chain.
Aliphatic alkyne residues have been added to the repertoire
of clickable UAAs by exploring the promiscuity of the natural
AMBER suppressor of the pyrrolysine tRNAPyl/PylRS pair
from M. mazei or Methanosarcina bakeri.21 This allows the site-
specific encoding of 4 and 5.22 The four UAAs, 2−5, can be
reacted in CuAAC reactions with the respective alkyne and
azide single-molecule dyes. Furthermore, all four UAAs are
readily available on a large scale, either commercially (2) or by
simple, few step, efficient synthesis (see also Discussion below
and Supporting Information).
3.2. Protein Engineering and Yield. Arguably, the most
important practical discriminator for the suitability of
genetically encoding UAAs for single-molecule research is the
yield of protein expression. Although sample consumption in
smFRET experiments is small, proteins first need to be
prepared to high purity, which is complicated with low-
expressing proteins. Furthermore, for most proteins other than
classical model systems, expression is the limiting factor, and a
drop in expression due to inefficient AMBER suppression often
jeopardizes the experiment. Thus, knowing the best expressing
system is of utmost importance for the success of the
technology in biology studies. Thus, we first aimed to
quantitatively compare the AMBER suppression efficiency of
the various systems.
Genetically encoding UAAs requires cotransformation of the
expression host E. coli with a plasmid harboring an orthogonal
tRNA/RS pair along with a plasmid carrying the protein of
interest, which typically contains an AMBER stop codon at the
site of UAA incorporation. A particular concern is if the
available systems to encode clickable UAAs maintain high
incorporation fidelity compared to the established systems
using 1 and if labeling ultimately yields a protein suitable for
single-molecule observation in a straightforward way. We thus
included 1 as the current “gold standard” in all our
analysis.5,17,23,24
For determination of the expression yield, we used two
reporter systems in order to ensure a representative readout. A
GFP containing a unique AMBER TAG site at amino acid
position 39 is a convenient check of protein expression by
means of the fluorescence from either a cell culture suspension
or purified GFP via a C-terminal 6His handle using a Ni-affinity
resin.16 As AMBER suppression heterogeneity for different
proteins is possible, we also expressed maltose binding protein
(MBP) with a TAG site in an artificial N-terminal linker. Figure
2 shows that for the M. jannaschii system, 3 expresses
GFPTAG→UAA better than 2 and MBP only marginally worse.
While 2 is commercially available and has also been used before
for protein engineering,37,38 3 can be easily synthesized (within
about a day) in gram quantities using basic laboratory
procedures convenient also for labs not specialized in chemistry
(see Supporting Information). Most importantly, azides in
general, and in particular aromatic azides,39 are often reduction-
sensitive, and are less stable in culture and during biochemical
purification than their alkyne counterparts, which potentially
impacts labeling yields. To experimentally verify the potential
advantages or disadvantages of both chemistries, we expressed
T4L encoding 2−5 side-by-side and also purified the proteins
under standard biochemical conditions. Labeling was then
Figure 2. AMBER suppression expression test. (a) Fluorescence image
of E. coli suspension expressing GFPTAG→UAA; (b) corresponding
SDS−PAGE gel of purified GFPTAG→UAA; (c) SDS−PAGE gel of
purified MBPTAG→UAA
.
carried out at exactly the same protein concentrations using 5-
fold excess of the complementary clickable dye derivative.
Supporting Information Figure S1 confirms the major intuitive
disadvantage of encoding azides compared to alkynes. Although
all proteins behave well during labeling (no degradation), only
3 and 5 can be efficiently labeled under the tested standard
conditions. It is therefore advisable to genetically encode the
alkyne and perform the comparatively quick CuAAC labeling
(see Supporting Information Figure S2 for labeling rates) in
controlled clean buffer conditions with azido dyes rather than
the other way around. Consequently, we discarded 2 and 4
from our further analysis and focused our comparative study on
1, 3, and 5. Compound 5 is equally well synthetically accessible
as compound 3 and can be synthesized in gram quantities
within a day also by laboratories not specialized in chemistry
(see Supporting Information).
3.3. Single-Molecule Multiparameter Spectroscopy.
Single-molecule FRET is used in a variety of fields to study
structure and dynamics of not only folded, but also disordered
proteins.1,3 In particular, the study of IDPs had recently gained
much attention, since only very few techniques exist that can
quantitatively provide insights into IDP plasticity. We selected
three different proteins in order to extract general properties of
site-specific labeling of various UAAs and to compare the
different labeling strategies based on protein conformational
studies. Therefore, in all experiments, the acceptor dye Alexa
594 (A) was attached via a single cysteine using standard
maleimide chemistry serving as a reference in all measurements
and the donor dye Alexa 488 (D) was reacted with the
respective UAA. Since lifetime-based smFRET experiments
provide considerably more information about the photophysical
properties of D than A, this labeling strategy affords maximum
insight into distinct UAA properties attached as a donor.
Moreover, Alexa 488 is available in all the reactive forms
required for this study and Alexa 488−Alexa 594 is a well-
established (and probably most frequently used) dye pair for
studying protein structure and dynamics.17,32,40−43 Prior to
labeling, the three proteins were reacted with maleimide A on
their unique cysteines, and D was introduced under optimized
experimental labeling conditions (see Materials and Methods).
1 was always reacted under denaturing conditions at pH 4 as in
previously published procedures,5,23,24 while the CuAAC for 3
and 5 was performed under physiological conditions (Figure
3a,b). See Supporting Information Figure S2 for more details
on labeling conditions, kinetics, and yields.
5190
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
used to record fluorescence intensities IA and ID, fluorescence
lifetimes τ, and anisotropies r of D for all three proteins.4,17
Figure 4 shows exemplary experimental data for hNup153, and
Supporting Information Figure S3 for yNup49 and Supporting
Information Figure S4 for T4L.
An important feature when analyzing protein conformation is
the brightness of the labeling dye. Alexa fluorophores, in
particular Alexa 488, have high quantum yields and relatively
long fluorescence lifetimes. These two connected properties
might, however, change when the fluorophore is conjugated to
a protein and are likely to also depend on the linker properties.
In addition, fluorescence lifetimes and EFRET are also directly
related to each other. The longer the initial fluorescence
lifetime of the donor (τ0), the more sensitive the lifetime-based
FRET efficiency is because small distance changes result in
larger lifetime changes. The correlation between these two
parameters is visible in the 2D τ versus EFRET histograms,
shown in Figure 4 for hNup153 (Supporting Information
Figure S3 for yNup49 and Supporting Information Figure S4
for T4L). The two populations discernible in the 2D τ vs EFRET
histograms in Figure 4 correspond to the 0-peak population
(inactive or absent acceptor, high lifetime) and to the FRET
population (lower lifetime). The 0-peak and FRET populations
are fitted with Gaussian functions from which EFRET and τ can
be obtained. Each protein displays a consistent trend with
respect to UAA in both EFRET (based on fluorescence intensity
recordings) and τ (i.e., τ0, τDA), as can be seen in the summary
Figure 5. Alexa 488 has a shorter τ and higher EFRET when
attached to 3 than when attached to 1 or 5, which behave rather
similarly.
While both 1 and 3 are based on a rather stiff phenylalanine
backbone, 5 is derived from the much more flexible lysine. On
the other hand, the triazole ring originating from the labeling of
3 and 5 might render additional stiffness to the donor molecule
conjugated via this reaction compared to the oxime linker
between 1 and the dye. Fluorescence anisotropies report on
differences in rotational freedom of dyes and might therefore
reflect such variability of linker properties. Figure 4 shows the r
versus τ plots (for the FRET population) of the 1, 3, and 5
conjugated D fluorescence of hNup153 (Supporting Informa-
tion Figure S3 for yNup49 and Supporting Information Figure
S4 for T4L). Since r is an observable that depends on the
observation time of a single molecule after a laser excitation
pulse, it is directly related to the fluorescence lifetime through
the Perrin equation. ρ, on the contrary, is an intrinsic property
of the fluorophore.4 To draw reliable conclusions about ρ, both
the 0-peak population and the FRET population have to lie on
the same Perrin curve. This is true for hNup153 (Figure 4) and
for yNup49 (Supporting Information Figure S3). T4L
(Supporting Information Figure S4) shows slight deviations
from the Perrin curve under all labeling conditions (with all
UAAs). This observation indicates possible interference of the
folded structure with dye flexibility, and not necessarily UAA
properties. In general, care should be taken when evaluating
FRET data from proteins that do not follow the prediction of
the Perrin curve: elevated rotational correlation times can also
affect the assumption that the rotational averaging factor κ2 =
2/3, which can affect the accuracy of distance estimates.14 To
avoid any bias in this study, Table 1 and Supporting
Information Table S2 list the actual distances that were
extracted only for the IDPs. Figure 5d summarizes the
rotational correlation coefficients of D labeled to proteins.
IDPs labeled with UAA 3 show highest ρ (i.e., lowest rotational
Figure 3. Scheme for labeling proteins via oxime ligation (a) and via
CuAAC (b). (c) Energy minimized representation (using Chem-
Bio3D, Cambridgesoft, MA) of chemical structures of UAA linked to
dye derivative (truncated after linker C6 of Alexa 488). Distance drawn
from Cα to linker-C of dye (see Supporting Information Figure S3 for
more information).
Since commonly proteins are double-labeled using maleimide
chemistry at two cysteines, little is known on how different
dye−linker structure and compositions (Figure 3c, and
Supporting Information Figure S5) can effect smFRET
measurements. However, this is an important topic since
FRET distance measurements can be affected by a perturbed
local environment due to quenching or constrained mobility of
the dyes.2,4 Using MP smFRET, those effects can be measured.
To assess these, the three proteins were subjected to
multiparameter (MP) spectroscopy under identical conditions,
which can detect lifetime anisotropy decays with high
resolution.17 The two-domain T4L serves as a model system
for a folded protein.5,23 However, amino acids in structured
proteins are known to be exposed to a less homogeneous
environment than those in rapidly fluctuating, polymer-like
intrinsically disordered proteins (IDPs).44 In IDPs, the dye is
exposed to a more isotropic environment, so potential
differences in a multiparameter single-molecule measurement
due to changes in the UAA−dye architecture are more likely to
be detected and interpreted correctly compared to a folded
protein. We selected the yeast yNup49 intrinsically disordered
protein19 and the large human IDP hNup153.45 These IDPs are
well suited for our high-resolution study, as they lack tyrosine
and tryptophan, which can dramatically alter dye photophysics
and bias results.46 For the IDPs, D and A were placed 40−70
amino acids away, so that FRET efficiencies were recorded
under the most sensitive regime of the technique. Multi-
parameter spectroscopy of freely diffusing single molecules was
5191
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
Figure 4. MP smFRET data for hNup153 with D labeled via 1, 3 and 5, respectively. In each column, burst integrated fluorescence lifetime (BIFL)
analysis is in the upper panel and corresponding anisotropy r vs τ in the lower panel. The 2D plots are gray scale coded for frequency of occurrence,
while the top and right histograms are maximum projections of the data along the vertical τ axis and along the horizontal EFRET axis. In the lower row,
the black line shows the expected trend according to the Perrin equation. Dotted lines outline fit results.
a
Table 1. Linker Distances and Measured FRET Distances
RE (Å)
ΔRE (Å)
linker (Å)
Δlinker (Å)
yNup49TAG→1
yNup49TAG→3
yNup49TAG→5
hNup153TAG→1
hNup153TAG→3
hNup153TAG→5
63.8
57.8
62.4
57.6
53.2
55.6
0.0
−6.0
−1.4
0.0
11.5
12.8
14.8
11.5
12.8
14.8
0.0
1.3
3.3
0.0
1.3
3.3
−4.4
−2.0
a
ΔRE and Δlinker correspond to measured value of UAA 1.
labeled under physiological conditions and is therefore the
choice for general single-molecule applications.
3.3. FRET-Labeling of the Large and Cysteine-Rich
RanBP3. Having now an optimized system in hand, we aimed
to apply this technology to a common problem for smFRET in
biology: the site-specific dual labeling with full choice of
labeling sites (i.e., not just the termini) of large and cysteine-
rich proteins that require mild, physiological labeling
conditions, ideally in one step. RanBP3, a key player in
regulation of nucleocytoplasmic transport, is such a protein;
since partially disordered, no structural information can be
obtained from X-ray crystallography, and due to its size (>50
kDa), it is essentially inaccessible by NMR.47,48 While these
limitations are not a concern for smFRET measurements in
general, the multiple cysteines (>5), that could be of functional
relevance, render RanBP3 also inaccessible to standard
smFRET approaches. As a consequence, only a small region
of RanBP3 has been amenable to crystallographic analysis but
the overall architecture of RanBP3 is still elusive.48 RanBP3,
however, can undergo a variety of tasks, and regulate nuclear
Figure 5. Summary of data measured for all UAA containing proteins.
(a) EFRET (from intensity measurement); (b) τDA; (c) τ0; and (d) ρ.
freedom), while dyes attached to 1 and 5 indicate faster
rotation (low ρ), following the same pattern as for the other
two parameters (EFRET and τDA) reported in Figure 5, with 3
behaving different than the other two.
Since both, high fluorescence lifetimes of D and rotational
freedom of the D and A are important for high-resolution
single-molecule FRET, UAAs 1 and 5 seem the ideal handles
on which to attach fluorophores for multiparameter fluo-
rescence studies. In addition, compared to 1, 5 can also be
5192
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
export of various cargos, including mRNA, in the form of a
tetrameric complex with RanGTP, the cargo and major export
factor Crm1.47−49 The part of RanBP3 that is not accessible to
X-ray crystallography is heterogeneous in sequence and
contains an unusual nuclear localization sequence (NLS)
(Figure 6). Particularly interesting is also the existence of two
evidence of the existence of a native collapsed state, which is
distinct from a fully unfolded or folded state.30,32 In such a
state, intramolecular interactions are favored versus a fully
unfolded population and could function in presenting the FxFG
motives for complex formation as part of the still unknown
molecular transport mechanism. This result demonstrates that
click labeling of two site-specifically encoded 5 is a reliable and
efficient tool to label even complex proteins in order to study
protein folding and conformational changes. Further mutagenic
screening following the same procedure will now allow us to
fully map the disordered domains and the conformational
motions of RanBP3 in its various active states and during active
transport. This will make the disorder-mediated regulatory
mechanism of this proteins and how it can facilitate nuclear
transport by binding to various proteins such as Importin-α,
Importin-β, Crm1, and RanGTP finally amenable to structural
biology.
4. DISCUSSION
Encoding UAAs for site-specific protein labeling is one of the
rare tools for accessing structurally complex and demanding
proteins in single-molecule FRET studies. However, not all
potentially available UAAs fulfill the requirements for future
high-resolution conformational studies. As also summarized in
Supporting Information Table S3, in this study we started out
from five known UAAs that can in principle be used for
biochemical labeling. Because of the reduction sensitivity of
azides during standard protein expression and purification of
proteins, we quickly discarded the further use of azides 2 and 4
(Supporting Information Figure S1). This is a particularly
noteworthy result, since 2, likely owing to its commercial
availability, has been used frequently to engineer proteins in
applications where, for example, inefficient labeling was not
detrimental.38
Apart from a good expression yield and the possibility of
labeling under physiological conditions, the way dyes are
attached to proteins can have a major impact on the
interpretation of sm data. A possible explanation for the
different ρ measured with the various UAAs is that 1 and 5
contain only one rigid aromatic ring (benzene ring from the
phenylalanine and triazole from the click reaction, respectively),
while 3 contains two (both, benzene and triazole rings).
smFRET requires the isotropic averaging of dye orientation
during the lifetime, so that the orientational averaging factor κ2
can be assumed to be 2/3. Any deviation from this would
jeopardize accurate distance information.50 This is particularly
important, since no easy way exists to directly measure κ2 and
thus the assumption needs to be valid in order not to
compromise the experiment. In principle, all the tested UAAs
fulfill these requirements. Since the κ2 criterion requires high
flexibility and thus long linkers, a direct correlation of measured
FRET distance with actual residue distance within a protein
labeled via very short (no) linkers is not possible. A conversion
between these two distances is thus necessary. For simple
model systems that are also accessible to computational
complementation using accessible volume or molecular
dynamics calculations, such as labeled polyproline, DNA, and
RNA, it has been shown that very high, sometimes even
effectively atomic resolution can be achieved.13−15,29 As these
procedures cannot be easily extended to large proteins with
unknown structure, an approximation of dye−linker contribu-
tions in protein structure determination by smFRET is
nontrivial. We thus calculated for the highly flexible intrinsically
Figure 6. (a) GdmCl titration of RanBP3 clearly shows a gradual
decrease in EFRET when transferred to denaturing condition. (b)
Cartoon of overall RanBP3 architecture showing disordered N- and C-
terminal domains, as well as the structured Ran binding domain
(RBD) and labeling sites. The position of the nuclear localization
sequence (NLS) is depicted in gray and FxFG-repeats in orange. (c)
EFRET obtained for RanBP3 under different GdmCl concentrations can
be described by an exponential fit (black line) and decreases from ∼0.4
in PBS to ∼0.15 in 4 M GdmCl.
FxFG repeats within this region. These FG repeats are usually
highly enriched in nucleoporins that line the inner channel of
the nuclear pore complex and constitute the permeability
barrier. The function of these FG repeats in a protein that itself
can shuttle the pore is yet completely unknown, as well as the
structure of this region. To demonstrate the utility of our
technology, we introduced TAG mutations at AA site 181 and
294 and expressed RanBP3 in the appropriate E. coli strain to
twice genetically encode 5. A single reaction with a mixture of
Alexa 488 and Alexa 594 azide derivative dyes yielded efficient
and site-specific labeling under physiological conditions. The
doubly labeled RanBP3 gives a FRET around EFRET = 0.4 under
physiological conditions (Figure 6a). The FRET peak moves to
EFRET = 0.15 in the presence of increasing concentrations of
guanidinium chloride (GdmCl). Thus, despite predicted to be
an intrinsically disordered region, the region can be further
“unfolded” through denaturation. Notably, the corresponding
EFRET and GdmCl-concentrations can be fit with an exponential
decay, which indicates the absence of stable secondary
structure, but points toward the existence of transient
secondary structure.30 Taken together, the occurrence of a
single EFRET population that can be gradually denatured is
5193
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
disordered proteins yNup49 and hNup153 the dye-to-dye
distance, also termed the end-to-distance (RE). To accurately
take the flexibility of the molecule into account, the basic
Foerster equation cannot be used but needs to be expanded to
take the fast fluctuating character of the protein into account
(see Materials and Methods and Supporting Information Table
S2 for more information).17,29−33 In Table 1, we summarize the
obtained RE distances and also plot the relative difference to the
measurement with 1, which serves as a reference in this study.
Our data (Figure 5a, Table 1, and Supporting Information
Table S2) already show that, while the energy minimized
structure of for example 5 is >3 Å longer than for 1, proteins
that were labeled via 1 show consistently longer distances
compared to when they were labeled via 5. As another example,
a measurement of yNup49TAG→1 yields a distance of 63.8 Å. It
would be common to assume that due to the approximately 1.3
Å-longer linker of 3 compared to 1, a measurment of
yNup49TAG→3 should yield a distance of 64.1 Å. In fact,
however, a distance of 57.8 Å is measured and thus the total
error could be more than 7 Å. Although we can measure very
small but consistent differences between the proteins labeled
on the different UAAs (Figure 5), accurate absolute protein
backbone distance measurements still need to be developed in
general. However, our data permit relative, but precise,
comparison between different experiments and different
UAAs, and show that 1, 3, and 5 are suitable for FRET
distance measurements. Furthermore, our data show how the
different linkers that “counterintuitively” affect the FRET
distance measurement can be used as a reference to relate
measurements with different UAAs to another. Compared to 3,
1 and 5 obey higher lifetimes and lower anisotropy and thus
clearly depict advantages with respect to higher dynamic range
and higher measurement precision for smFRET measurements
(Figure 5, Supporting Information Table S1).
Furthermore, the availability to use chemically distinct UAAs
with phenylalanine- (3, aromatic, hydrophobic) and lysine-like
(5, aliphatic, hydrophilic) backbones can be useful when the
protein backbone structure requires minimal perturbance, such
as folding sensitive proteins, where a canonical Phe could be
replaced with either 1 or 3, or a canonical Lys with 5. 3 might
also have advantages when large changes in anisotropy need to
be recorded with high precision, such as in drug screening.51
Whenever the protein is not denaturation-sensitive and can be
labeled under acidic conditions, 1 seems to be the ideal choice,
as it offers similar photophysical properties as 5 but a higher
expression yield. We note that sometimes proteins cannot be
labeled using copper-based click chemistry typically due to
intolerance against copper. In those cases, our recently
demonstrated use of genetically encoding strained alkynes
that allow for catalyst-free click chemistry can in principle
present a solution.16 However, strained alkynes are difficult to
synthesize and it would not be easy to express large cultures as
necessary for the preparation of high purity samples of very
challenging to express IDPs such as hNup153.
Finally, our optimized system allowed us to make the
challenging protein RanBP3 accessible to smFRET studies by
encoding two UAAs at any position we wanted, followed by a
simple single-step reaction. We showed a particular region in
the IDP-domain adopts collapsed conformation. This now
affords the possibility to study also the regulatory mechanism of
this multifunctional, partially unfolded protein at single-
molecule resolution and to map conformational changes and
changes of the protein dynamics induced through protein−
protein interactions that are relevant in nuclear transport.
Combining our approach in the future with recent develop-
ments that facilitate multifunctional and multisite incorporation
of UAAs will likely influence the power of smFRET in biology
studies dramatically.52−55 In more general applications, our
strategy will also facilitate the labeling with other reporters such
as spin markers for NMR, EPR, and MRI as well as post-
translational modifications such as glycosylation and ubiquiti-
nylation.
5. CONCLUSION
We have shown in this work that genetically encoded clickable
UAAs can be used for an efficient and general strategy to finally
make proteins accessible to fluorescence and, in particular, to
high-resolution single-molecule FRET measurements, inde-
pendent of size or if the protein is folded or disordered. Taking
the accumulated information together from this systematic
study, for high-resolution smFRET measurements the tRNAPyl/
PylRS system from M. mazei genetically encoding 5 is the best
system for several reasons: (i) most importantly, expression
using our optimized pEvol system guarantees efficient and
selective AMBER suppression and high-yielding protein
production even in ultrarich media such as terrific broth.
Good expression thus also allowed double AMBER suppres-
sion, which can then, at the cost of random labeling, be used to
label a protein with full labeling freedom at any site in only one
step. (ii) 5 can easily be synthesized even with basic chemical
tools in gram quantities and reacted with commercially available
compounds; (iii) the alkyne moiety is less prone to reduction
compared to the azide (particularly important during long-term
expression in living cultures); (iv) the D dye maintains a long
fluorescence lifetime, supporting a high dynamic range for
FRET measurements; (v) CuAAC proceeds readily under
physiological conditions; (vi) the initial anisotropy of labeled
proteins is very low, encouraging the application of the key
assumption for all high-resolution smFRET measurements of κ2
being 2/3.
ASSOCIATED CONTENT
* Supporting Information
Supporting Information including additional methods and
results. This material is available free of charge via the Internet
at http://pubs.acs.org.
■
S
AUTHOR INFORMATION
Corresponding Author
lemke@embl.de
■
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was funded by the Emmy Noether program of the
DFG. S.M. acknowledges a fellowship from the BIF; S.T. from
the Darwin Trust foundation; and T.P. from the VCI. We thank
P.G. Schultz (Scripps) for the pEvol system to incorporate 1, 2,
and 3.
REFERENCES
■
(1) Deniz, A. A.; Mukhopadhyay, S.; Lemke, E. A. J. R. Soc. Interface
2008, 5, 15.
(2) Roy, R.; Hohng, S.; Ha, T. Nat. Methods 2008, 5, 507.
5194
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195

Journal of the American Chemical Society
Article
(3) Schuler, B.; Eaton, W. A. Curr. Opin. Struct. Biol. 2008, 18, 16.
(4) Sisamakis, E.; Valeri, A.; Kalinin, S.; Rothwell, P. J.; Seidel, C. A.
Methods Enzymol. 2010, 475, 455.
(38) Sommer, S.; Weikart, N. D.; Brockmeyer, A.; Janning, P.;
Mootz, H. D. Angew. Chem., Int. Ed. 2011, 50, 9888.
(39) Staros, J. V.; Bayley, H.; Standring, D. N.; Knowles, J. R.
Biochem. Biophys. Res. Commun. 1978, 80, 568.
(5) Brustad, E. M.; Lemke, E. A.; Schultz, P. G.; Deniz, A. A. J. Am.
(40) Gambin, Y.; VanDelinder, V.; Ferreon, A. C.; Lemke, E. A.;
Groisman, A.; Deniz, A. A. Nat. Methods 2011, 8, 239.
(41) Ferreon, A. C.; Gambin, Y.; Lemke, E. A.; Deniz, A. A. Proc.
Natl. Acad. Sci. U.S.A. 2009, 106, 5645.
Chem. Soc. 2008, 130, 17664.
(6) Blaschke, U. K.; Silberstein, J.; Muir, T. W. Methods Enzymol.
2000, 328, 478.
(7) Scheck, R. A.; Dedeo, M. T.; Iavarone, A. T.; Francis, M. B. J. Am.
(42) Muller-Spath, S.; Soranno, A.; Hirschfeld, V.; Hofmann, H.;
Ruegger, S.; Reymond, L.; Nettels, D.; Schuler, B. Proc. Natl. Acad. Sci.
U.S.A. 2010, 107, 14609.
(43) Nettels, D.; Muller-Spath, S.; Kuster, F.; Hofmann, H.; Haenni,
D.; Ruegger, S.; Reymond, L.; Hoffmann, A.; Kubelka, J.; Heinz, B.;
Gast, K.; Best, R. B.; Schuler, B. Proc. Natl. Acad. Sci. U.S.A. 2009, 106,
20740.
(44) Wright, P. E.; Dyson, H. J. J. Mol. Biol. 1999, 293, 321.
(45) Lim, R. Y.; Huang, N. P.; Koser, J.; Deng, J.; Lau, K. H.;
Schwarz-Herion, K.; Fahrenkrog, B.; Aebi, U. Proc. Natl. Acad. Sci.
U.S.A. 2006, 103, 9512.
(46) Chen, H.; Ahsan, S. S.; Santiago-Berrios, M. B.; Abruna, H. D.;
Webb, W. W. J. Am. Chem. Soc. 2010, 132, 7244.
(47) Lindsay, M. E.; Holaska, J. M.; Welch, K.; Paschal, B. M.;
Macara, I. G. J. Cell Biol. 2001, 153, 1391.
(48) Langer, K.; Dian, C.; Rybin, V.; Muller, C. W.; Petosa, C. PLoS
One 2011, 6, e17011.
(49) Mueller, L.; Cordes, V. C.; Bischoff, F. R.; Ponstingl, H. FEBS
Lett. 1998, 427, 330.
Chem. Soc. 2008, 130, 11762.
(8) Sun, X.; Zhang, A.; Baker, B.; Sun, L.; Howard, A.; Buswell, J.;
Maurel, D.; Masharina, A.; Johnsson, K.; Noren, C. J.; Xu, M. Q.;
Correa, I. R. Jr. ChemBioChem 2011, 12, 2217.
(9) Yi, L.; Sun, H.; Itzen, A.; Triola, G.; Waldmann, H.; Goody, R. S.;
Wu, Y. W. Angew. Chem., Int. Ed. 2011, 50, 8287.
(10) Dirksen, A.; Hackeng, T. M.; Dawson, P. E. Angew. Chem., Int.
Ed. 2006, 45, 7581.
(11) Hackenberger, C. P.; Schwarzer, D. Angew. Chem., Int. Ed. 2008,
47, 10030.
(12) Tompa, P. Structure and Function of Intrinsiclaly Disordered
Proteins; CRC PRESS: Boca Raton, FL, 2010.
(13) Hoefling, M.; Lima, N.; Haenni, D.; Seidel, C. A.; Schuler, B.;
Grubmuller, H. PLoS One 2011, 6, e19791.
(14) Sindbert, S.; Kalinin, S.; Nguyen, H.; Kienzler, A.; Clima, L.;
Bannwarth, W.; Appel, B.; Muller, S.; Seidel, C. A. J. Am. Chem. Soc.
2011, 133, 2463.
(15) Muschielok, A.; Andrecka, J.; Jawhari, A.; Bruckner, F.; Cramer,
P.; Michaelis, J. Nat. Methods 2008, 5, 965.
(16) Plass, T.; Milles, S.; Koehler, C.; Schultz, C.; Lemke, E. A.
Angew. Chem., Int. Ed. 2011, 50, 3878.
(17) Milles, S.; Lemke, E. A. Biophys. J. 2011, 101, 1710.
(18) Denning, D. P.; Patel, S. S.; Uversky, V.; Fink, A. L.; Rexach, M.
Proc. Natl. Acad. Sci. U.S.A. 2003, 100, 2450.
(50) Iqbal, A.; Arslan, S.; Okumus, B.; Wilson, T. J.; Giraud, G.;
Norman, D. G.; Ha, T.; Lilley, D. M. Proc. Natl. Acad. Sci. U.S.A. 2008,
105, 11176.
(51) Owicki, J. C. J. Biomol. Screening 2000, 5, 297.
(52) Johnson, D. B.; Xu, J.; Shen, Z.; Takimoto, J. K.; Schultz, M. D.;
Schmitz, R. J.; Xiang, Z.; Ecker, J. R.; Briggs, S. P.; Wang, L. Nat. Chem.
Biol. 2011, 7, 779.
(19) Strawn, L. A.; Shen, T.; Shulga, N.; Goldfarb, D. S.; Wente, S. R.
Nat. Cell Biol. 2004, 6, 197.
(53) Mukai, T.; Hayashi, A.; Iraha, F.; Sato, A.; Ohtake, K.;
Yokoyama, S.; Sakamoto, K. Nucleic Acids Res. 2010, 38, 8188.
(54) Wan, W.; Huang, Y.; Wang, Z. Y.; Russell, W. K.; Pai, P. J.;
Russell, D. H.; Liu, W. R. Angew. Chem., Int. Ed. 2010, 49, 3211.
(55) Neumann, H.; Wang, K. H.; Davis, L.; Garcia-Alai, M.; Chin, J.
W. Nature 2010, 464, 441.
(20) Young, T. S.; Ahmad, I.; Yin, J. A.; Schultz, P. G. J. Mol. Biol.
2010, 395, 361.
(21) Neumann, H.; Peak-Chew, S. Y.; Chin, J. W. Nat. Chem. Biol.
2008, 4, 232.
(22) Nguyen, D. P.; Lusic, H.; Neumann, H.; Kapadnis, P. B.;
Deiters, A.; Chin, J. W. J. Am. Chem. Soc. 2009, 131, 8720.
(23) Lemke, E. A.; Gambin, Y.; Vandelinder, V.; Brustad, E. M.; Liu,
H. W.; Schultz, P. G.; Groisman, A.; Deniz, A. A. J. Am. Chem. Soc.
2009, 131, 13610.
(24) Lemke, E. A. Methods Mol. Biol. 2011, 751, 3.
(25) Koshioka, M.; Sasaki, K.; Masuhara, H. Appl. Spectrosc. 1995, 49,
224.
(26) Schaffer, J.; Volkmer, A.; Eggeling, C.; Subramaniam, V.; Striker,
G.; Seidel, C. A. M. J. Phys. Chem. A 1999, 103, 331.
(27) Kapanidis, A. N.; Lee, N. K.; Laurence, T. A.; Doose, S.;
Margeat, E.; Weiss, S. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 8936.
(28) Muller, B. K.; Zaychikov, E.; Brauchle, C.; Lamb, D. C. Biophys.
J. 2005, 89, 3508.
(29) Best, R. B.; Merchant, K. A.; Gopich, I. V.; Schuler, B.; Bax, A.;
Eaton, W. A. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 18964.
(30) Hoffmann, A.; Kane, A.; Nettels, D.; Hertzog, D. E.; Baumgartel,
P.; Lengefeld, J.; Reichardt, G.; Horsley, D. A.; Seckler, R.; Bakajin, O.;
Schuler, B. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 105.
(31) Laurence, T. A.; Kong, X.; Jager, M.; Weiss, S. Proc. Natl. Acad.
Sci. U.S.A. 2005, 102, 17348.
(32) Mukhopadhyay, S.; Krishnan, R.; Lemke, E. A.; Lindquist, S.;
Deniz, A. A. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 2649.
(33) Gopich, I. V.; Szabo, A. J. Phys. Chem. B 2007, 111, 12925.
(34) Liu, C. C.; Schultz, P. G. Annu. Rev. Biochem. 2010, 79, 413.
(35) Chin, J. W.; Santoro, S. W.; Martin, A. B.; King, D. S.; Wang, L.;
Schultz, P. G. J. Am. Chem. Soc. 2002, 124, 9026.
(36) Deiters, A.; Schultz, P. G. Bioorg. Med. Chem. Lett. 2005, 15,
1521.
(37) Seo, M. H.; Lee, T. S.; Kim, E.; Cho, Y. L.; Park, H. S.; Yoon, T.
Y.; Kim, H. S. Anal. Chem. 2011, 83, 8849.
5195
dx.doi.org/10.1021/ja210587q | J. Am. Chem. Soc. 2012, 134, 5187−5195