Directed evolution of a bright near-infrared fluorescent
rhodopsin using a synthetic chromophore
Lukas Herwig
a,2
,
Austin J. Rice
a,2
,
Claire N. Bedbrook
a,b
,
Ruijie K. Zhang
a
,
Antti Lignell
a
,
Jackson K. B. Cahn
a
,
Hans Renata
a
,
Sheel C. Dodani
a
,
Inha Cho
a
,
Long Cai
a
,
Viviana
Gradinaru
b
, and
Frances H. Arnold
a,b,1
a
Division of Chemistry and Chemical Engineering, California Institute of Technology, 1200 E.
California Blvd, Pasadena, CA, 91125
b
Division of Biology and Biological Engineering, California Institute of Technology, 1200 E.
California Blvd, Pasadena, CA, 91125
Summary
By engineering a microbial rhodopsin, Archaerhodopsin-3 (Arch), to bind a synthetic
chromophore, merocyanine retinal, in place of the natural chromophore all-
trans
-retinal (ATR), we
generated a protein with exceptionally bright and unprecedentedly red-shifted near-infrared (NIR)
fluorescence. We show that chromophore substitution generates a fluorescent Arch-complex with a
200 nm bathochromic excitation shift relative to ATR-bound wild-type Arch and an emission
maximum at 772 nm. Directed evolution of this complex produced variants with pH-sensitive NIR
fluorescence and molecular brightness 8.5-fold greater than the brightest ATR-bound Arch variant.
The resulting proteins are well suited to bacterial imaging; expression and stability have not been
optimized for mammalian cell imaging. By targeting both the protein and its chromophore we
overcome inherent challenges associated with engineering bright NIR fluorescence into
Archaerhodopsin. This work demonstrates an efficient strategy for engineering non-natural,
tailored properties into microbial opsins, properties relevant for imaging and interrogating
biological systems.
eTOC Blurb
Using a combined approach of chromophore substitution and directed evolution, Herwig
et al.
engineered fluorescent Archaerhodopsin variants with unprecedented near-infrared (NIR)
excitation and emission. Evolved variants display pH sensitivity, enhanced fluorescent molecular
brightness and improved synthetic chromophore affinity.
1
To whom correspondence should be addressed. frances@cheme.caltech.edu, Phone: 626-395-4162.
2
L.H. and A.J.R. contributed equally to this work.
Author Contributions
L.H., A.J.R., and C.N.B. designed and performed research and analyzed data; F.H.A. supervised research; R.K.Z. and H.R.
synthesized and characterized merocyanine retinal; A.L. and I.C. performed research; A.L., L.C., and V.G. provided necessary
analytical tools; J.K.B.C and S.C.D analyzed data; L.H., A.J.R., C.N.B., and F.H.A. wrote the paper. All authors gave final approval
for publication.
Publisher's Disclaimer:
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
HHS Public Access
Author manuscript
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Published in final edited form as:
Cell Chem Biol
. 2017 March 16; 24(3): 415–425. doi:10.1016/j.chembiol.2017.02.008.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
Keywords
Synthetic chromophore substitution; directed evolution; protein engineering; near-infrared
fluorescence; Archaerhodopsin; live-cell imaging
Introduction
Fluorescent proteins have revolutionized our ability to visualize the biological world at the
molecular level (
Davidson and Campbell, 2009
;
Kremers et al., 2011
). The photophysical
properties of a fluorescent protein are dictated by two factors, the light-excitable
chromophore and the interacting protein environment. While nature is very effective at
tuning the protein environment through mutation and selection, the limited library of
reported natural chromophores ultimately constrains engineering possibilities (
Davidson and
Campbell, 2009
;
Shcherbakova and Verkhusha, 2014
). Near-infrared (NIR) fluorescence
(650–900 nm) is desirable for non-invasive deep tissue imaging due to reduced scattering
and low phototoxicity of the longer wavelength light (
Weissleder, 2001
). Engineered
bacterial phytochromes (BphPs) emit NIR fluorescence with their highly conjugated
biliverdin chromophore (
Fischer and Lagarias, 2004
;
Shcherbakova et al., 2015b
;
Shcherbakova and Verkhusha, 2013
;
Yu et al., 2015
), whereas the vitamin A-derived
chromophore all-
trans
-retinal (ATR) enables red- to farred fluorescence in certain microbial
rhodopsins (
Kralj et al., 2011a
;
Kralj et al., 2011b
). Such naturally occurring chromophores
are limited in their ability to produce bright, molecular NIR fluorescence. Even with
substantial protein engineering efforts, peak fluorescence excitation of ATR-bound
rhodopsin variants remain well outside the NIR window (
Engqvist et al., 2014
;
Hochbaum et
al., 2014
). Among BphPs, there seems to be a trade-off between red shift and molecular
brightness, with the brightest variants (brightness equivalent to mCherry (
Kremers et al.,
2011
)) excited outside the NIR (639 nm) and the furthest red-shifted BphPs (peak excitation
at 702 nm) approximately 2.7-fold dimmer than mCherry (
Shcherbakova et al., 2015a
;
Shcherbakova and Verkhusha, 2013
). Thus, the engineering of bright NIR fluorescent
proteins is still an outstanding challenge.
Previous engineering efforts have shown that directed evolution can be used to enhance and
red-shift the fluorescence of ATR-bound microbial rhodopsins, including Archaerhodopsin-3
(Arch) (
Engqvist et al., 2014
;
Hochbaum et al., 2014
;
McIsaac et al., 2014
). Originally from
the halophilic archaea
Halorubrum sodomense
, wild-type (WT) Arch is a yellow light-driven
Herwig et al.
Page 2
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
proton pump that has dim, far-red fluorescence (
Kralj et al., 2011a
;
Mukohata et al., 1999
).
In Arch, as with many rhodopsins, the ATR chromophore is covalently bound to a conserved
lysine residue via a Schiff base (
Ernst et al., 2014
). Protonation of this base modulates the
spectral properties and isomerization of ATR (
Ernst et al., 2014
). As a result, Arch
fluorescence is sensitive to pH and transmembrane voltage gradients (
Maclaurin et al., 2013
)
and has been used to monitor action potentials when expressed in cultured neurons (
Kralj et
al., 2011a
). Directed evolution of the ATR-bound Arch protein generated variants with
brighter and red-shifted fluorescence, enabling fluorescent imaging with lower-power light
(
Flytzanis et al., 2014
;
Hochbaum et al., 2014
;
McIsaac et al., 2014
). However, the brightest
engineered ATR-bound Arch variant was not pH-sensitive and was at least 12-fold dimmer
than mCherry; furthermore its peak excitation (~615 nm) remained outside the NIR window
(
McIsaac et al., 2014
).
In order to access Arch variants with bright NIR emission, we drew inspiration from
previous demonstrations that desirable fluorescent protein properties could be obtained by
expanding the limited repertoire of naturally known chromophores (
Plamont et al., 2016
;
Tamura and Hamachi, 2014
;
Yapici et al., 2015
). Spectral properties of the natural ATR
chromophore (Figure 1
A
, Compound 1) can be modulated by adding electron-withdrawing
groups (
Gaertner et al., 1981
;
Hendrickx et al., 1995
) or changing the length of the
conjugated
π
-bond system (
Albeck et al., 1989
;
Nielsen, 2009
). In particular, retinal analogs
with extended conjugation have been shown to red-shift the absorption maxima of
rhodopsins up to hundreds of nanometers, well into the NIR window (
Asato et al., 1990
;
Sineshchekov et al., 2012
). For example, merocyanine retinal (Figure 1
A
, Compound 2),
with extended conjugation relative to ATR, was shown by Hoischen and coworkers to
bathochromically shift bacteriorhodopsin absorbance by 187 nm (
Hoischen et al., 1997
).
Other chromophore substitutions have modified the spectral and kinetic characteristics of
rhodopsin-based optogenetic tools (
AzimiHashemi et al., 2014
) and modulated the proton
pumping capabilities of proteorhodopsin and
Gloeobacter violaceus
rhodopsin (
Ganapathy
et al., 2015
). Thus chromophore substitution enables rapid introduction of desirable
photophysical properties; however, it is unlikely that a wild-type opsin will bind a non-
natural chromophore preferentially or that its new spectral properties (e.g. NIR fluorescence)
will manifest at optimal levels. We have therefore paired chromophore substitution with
directed evolution in order to build on novel capabilities conferred by the synthetic
chromophore. Here we transcend the natural fluorescent properties of Arch by evolving the
protein around a synthetic chromophore, thereby creating variants with exceptional
molecular brightness and unprecedentedly red-shifted NIR fluorescence.
Results and Discussion
Merocyanine retinal enhances photophysical properties of Arch
Chromophore substitution can dramatically modify the inherent properties of a rhodopsin
and establish a new platform for achieving desirable features by protein engineering. Given
the extensive red-shift conferred on bacteriorhodopsin by merocyanine retinal (
Hoischen et
al., 1997
), we selected this chromophore for substitution in Arch as a first step in developing
a bright NIR fluorescent protein. Merocyanine retinal (aldehyde) was synthesized as
Herwig et al.
Page 3
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
described in supplemental methods (Figure S1). Addition of merocyanine retinal (final
concentration of 1 μM) to
E. coli
cultures expressing the Arch protein produced an opsin-
chromophore complex. A covalent retinal Schiff base (RSB) linkage in the complex was
confirmed by spectral comparison of denatured rhodopsin to free merocyanine retinal Schiff
base (Figure S2). The capacity of merocyanine retinal to form a RSB was anticipated due to
previous observations of a different merocyanine retinal analog (with shorter polyene chain)
binding to the retinoic binding protein CRABPII (
Yapici et al., 2015
). The absorbance and
fluorescence excitation of merocyanine-bound Arch were red-shifted by more than 200 nm
when compared to ATR-bound Arch (ex/em at 556/687 nm) (
McIsaac et al., 2014
) in
purified protein (Figure 1
B
and Table 1) and in whole cells (Figure S3
A
). This represents a
262 nm ‘opsin shift’ for Arch-bound merocyanine, which is the difference between the
absorption maxima of the free aldehyde chromophore (498 nm) and the newly formed Arch
complex (760 nm).
Directed evolution of merocyanine-bound Arch enhances opsin-specific fluorescence
Although capable of doing so, wild-type Arch did not evolve specifically to bind
merocyanine retinal or to allow energy escape in the form of fluorescence. Thus, we sought
to enhance Arch NIR-fluorescence intensity and molecular brightness by directing the
evolution of the Arch protein around this synthetic chromophore. Throughout this work we
distinguish between molecular brightness and fluorescence intensity. Although both
properties quantify the fluorescence light emitted by a fluorescent protein, the former is a
photophysical property of the protein that is determined
in vitro
, while the latter is measured
directly via plate reader or microscopy and is influenced by factors such as expression level
and imaging conditions. Our directed evolution screen selected for greater NIR fluorescence
intensity in
E. coli
.
Random mutagenesis of WT Arch by error-prone PCR and screening 1,700 variants for NIR
fluorescence (ex/em 760/785 nm) identified variant Mero-1 with the P60S mutation (Figure
2). Site-saturation mutagenesis of Mero-1 at position G61 (selected from a marginal hit in
the error-prone PCR library) led to Mero-2, with mutations P60S and G61L (Figure 2).
Mutations that increase fluorescent brightness of microbial rhodopsins (
Engqvist et al.,
2014
;
Hochbaum et al., 2014
;
McIsaac et al., 2014
) and increase occupancy of associated
states in the photocycle (
Maclaurin et al., 2013
;
Wagner et al., 2013
) are known to be located
proximal to ATR or the Schiff base. Thus, to guide further evolution, we generated a
homology model of merocyanine-bound wild-type Arch based on the known structure of
Archaerhodopsin-2 (86% amino acid identity; Figure 3) (
Kouyama et al., 2014
). Sites
located within 5 Å of the indolylidene ring of merocyanine retinal (W148, S151 and P196)
or the Schiff base (M30, V59, A63, T99) in the homology model were selected for site-
saturation mutagenesis in Mero-2. We identified P196G (Mero-3) and S151A. The latter
mutation displayed stronger pigment formation in culture compared to Mero-2, but showed
no significant increase in overall fluorescence intensity. Previous work has shown that
mutation S151A contributes to a hypsochromic absorbance shift in ATR-bound Arch (
Sudo
et al., 2013
). However, combining P196G and S151A in the Mero-2 background yielded
Mero-4 (P60S-G61L-S151A-P196G), which displays greater NIR fluorescence intensity
(Figure 2). Upon screening 1,800 variants of Mero-4 in a final round of error-prone PCR
Herwig et al.
Page 4
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
mutagenesis, we identified Mero-6 that has two additional mutations, T80A and G132S.
Mero-6 exhibited a 10-fold increase in molecular fluorescence intensity over merocyanine-
bound wild-type Arch (Figure 2).
Of the six mutations present in Mero-6 (Table 1), three are predicted to lie in the retinal
binding pocket (G132S, S151A, and P196G), two near the Schiff base (P60S and G61L),
and one on an extra-cellular loop (T80A) (Figure 3). Mutations near the retinal and Schiff
base have been shown to affect the spectral properties of ATR-bound Arch (
Hochbaum et al.,
2014
;
McIsaac et al., 2014
). Our identification of similar sites (non-identical mutations with
the exception of P60S (
Hochbaum et al., 2014
) and S151A (
Sudo et al., 2013
)) suggests that
non-natural retinal analogs can likewise be spectrally modulated by tuning direct protein-
chromophore interactions.
To assess the potential of the merocyanine-bound Arch variants as NIR fluorescent makers
in live cells, we acquired NIR images of bacterial and eukaryotic cells expressing Arch-WT
and Mero-6. For bacterial imaging,
E. coli
expressing either wild-type Arch or Mero-6 in the
presence of merocyanine retinal were readily detected with NIR excitation at 727 nm
(Coherent CUBE laser, 32mW power) and detection within 766–854 nm (Figure 4
A
, bottom
row); moreover, the evolved Mero-6 variant yielded more intense fluorescence visible with
lower contrast (Figure 4
A
, middle row). Opsin expression could be tracked by the fused CFP
tag (Figure 4
A
, top row) and, unlike NIR fluorescence, expression was independent of
merocyanine retinal addition (Figures 4
A
and S3
B
). Fluorescent puncta in cells expressing
wild-type Arch are more pronounced in the absence of merocyanine retinal and may be due
to protein instability when the protein is expressed in the absence of chromophore. In
support of plate-reader measurements (Figure S3
B
), the mean CFP fluorescence intensity
quantified from images reveals that expression was modestly reduced (Figure 4
B
) over the
course of evolution, while the absolute and CFP-normalized NIR fluorescence were
increased significantly (Figure 4
C
and 4
D
). Both NIR results clearly show the advantage of
the evolved Arch variant for bacterial imaging applications. However, comparing the results
of Figures 2 and 4
D
, the evolved fluorescence enhancement appears to be 2-fold greater
when quantified via microscope analysis as opposed to plate reader, which could be due to
the greater light intensity and more sensitive detection involved in microscopy imaging
(
Maclaurin et al., 2013
). Photostability measurements in live
E. coli
indicate that
merocyanine-bound wild-type Arch bleaches 79% faster than CFP, while Mero-6 bleaches
only 49% faster than CFP (ratio of NIR and CFP exponential decay rates; Table 1 and Figure
S4), suggesting that photostability has also increased over the evolutionary trajectory. The
ability to track merocyanine-treated, opsin-expressing bacteria in deep tissue could find
application in a wide range of biological studies. For example, a NIR fluorescent opsin
probe could be useful for tracking bacteria in whole animals or infected tumors (
Berlec et
al., 2015
;
Cronin et al., 2012
).
For NIR fluorescence in eukaryotic cells, GFP-tagged wild type and Mero-6 constructs were
built with optimal codon usage and trafficking signals (
Gradinaru et al., 2010
) (Figure S5
A
).
Representative images and quantitative analysis of transfected human embryonic kidney
(HEK 293T) cells show that Mero-6 expresses in those eukaryotic cells, though at
considerably lower levels than wild-type Arch and with an increased number of fluorescent
Herwig et al.
Page 5
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
puncta and aggregates that may be due to diminished protein stability or trafficking capacity
(Figure S5
B
and
C
). However, normalization of the raw NIR fluorescence (Figure S5
B
, right
column and S5
D
) by total opsin expression level (GFP fluorescence, Figure S5
C
) indicated
that the molecular fluorescence intensity of merocyanine–bound Mero-6 is significantly
greater than that of wild-type protein (Figure S5
E
). These results demonstrate that the
enhancement of Arch molecular NIR fluorescence intensity achieved by evolution in
bacteria transfers to eukaryotic cells. Given the exceptionally red-shifted fluorescence of the
merocyanine-bound Arch variants and their relatively high molecular brightness,
merocyanine-bound Arch variants offer promise for imaging applications in eukaryotic cells.
The absence of ion pumping for Mero-6 in the presence of ATR or merocyanine retinal
(Figure S6) would be beneficial for imaging, where active pumping would perturb the
transmembrane voltage and local pH. For eukaryotic applications, however, merocyanine-
bound Arch variants would need to be further optimized for expression and stability.
In eukaryotic cells, WT Arch and Mero-6 are marginally sensitive to a voltage step with
farred light (Figure S5
F
). In bacteria, the pH sensitivity of merocyanine-bound Arch
fluorescence was retained throughout the course of evolution. When validating each Arch
variant,
E. coli
NIR fluorescence measurements were taken at three pH values (5, 7 and 9).
All selected variants displayed enhanced fluorescence intensity at acidic pHs and dimmer
fluorescence at alkaline pH (Figure 2
A
). This bright, pH-sensitive NIR fluorescence could
be useful in microbiology applications, for instance to monitor bacterial activity and pH
microenvironments in biofilms with dim light (
Guo et al., 2013
;
Hidalgo et al., 2009
;
Schlafer et al., 2015
) or to assess deep-tissue, host-pathogen interactions in disease models
(
Duhring et al., 2015
;
Vande Velde et al., 2014
).
Evolved NIR fluorescent Arch variants bind merocyanine retinal with greater affinity and
lose affinity for ATR
Mutations in the chromophore-binding pocket of Arch modify protein-chromophore
interactions and can enhance the molecular fluorescent brightness (
Hochbaum et al., 2014
;
McIsaac et al., 2014
) and the affinity of Arch for synthetic merocyanine retinal. Improving
the affinity of Arch for merocyanine retinal is important for applications where only lower
concentrations of the synthetic chromophore are available or desirable. For example, limited
availability could be anticipated when the compound is applied indirectly (e.g. in animal
food) or when the natural ATR chromophore is competing for protein binding. Moreover,
due to possible toxicity of merocyanine retinal at high concentrations (Figure S8), the ability
to use lower concentrations is desirable. To determine whether the merocyanine retinal
binding affinity changed during the course of evolution, we first measured how the whole-
cell NIR fluorescence depended on the concentration of chromophore added (Figure 5
A
).
These results allowed us to calculate the concentration of chromophore required for selected
Arch variants to reach half maximal fluorescence after four hours of expression; this
chromophore concentration is referred to as the ‘apparent
K
d
’. In
E. coli
, the apparent
K
d
for
merocyanine retinal decreased 2.5-fold from wild-type Arch to Mero-4 (from 3.25 to 1.33
μM), with the most significant decrease (1.8-fold) observed between WT and Mero-2
(Figure 5
B
). Combined, the two mutations of Mero-2 (P60S and G61L) likely affect the
conformation of Helix 2 (
Vonheijne, 1991
;
Wilman et al., 2014
), possibly modifying the
Herwig et al.
Page 6
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
protein-Schiff base interaction. Of the two additional mutations in Mero-4, P196G may
affect the conformation of Helix 4, and both P196G and S151A could expand the retinal-
binding pocket to better accommodate the indolylidene ring of merocyanine retinal.
Accurate structural information will be required for a more detailed analysis of the
merocyanine conformation and interacting opsin residues.
As a more direct measurement of the affinity of Arch for ATR and merocyanine, we also
monitored the absorbance of purified,
apo
opsin pigmented with increasing concentrations of
each chromophore. Due to differences in binding pocket accessibility of the detergent-
stabilized
apo
protein versus the unpurified membrane-bound opsins (and the likelihood of
co-translational pigmentation in the latter), the absolute binding affinities determined by the
two methods are not expected to be the same. However, the relative affinities between Arch
variants in
E. coli
(Figure 5
B
) and purified protein (Table 1) show a similar trend of
increasing affinity for the new chromophore, particularly in the initial rounds of evolution.
For purified protein, the
K
d
of merocyanine retinal decreased from 99 μM (wild-type Arch)
to 11 μM (Mero-2) and increased slightly in the final rounds of evolution to 17 μM
(Mero-6). Therefore, the
in vitro
results indicate that mutations made in in the initial rounds
of evolution enhanced NIR fluorescence as well as merocyanine affinity. Moreover, the
affinity for ATR decreased during the course of evolution, as the
K
d
increased from 8 μM in
wild-type Arch to 330 μM in Mero-6. Possible differences in the detergent partition
coefficients of merocyanine retinal and ATR prevent direct comparison of calculated binding
affinities of the two lipophilic chromophores; however, the affinities of different Arch
variants for a given chromophore can be compared. The opposite trends in merocyanine
retinal and ATR binding affinity indicate that, over the course of evolution, the binding
pocket was restructured to accommodate the synthetic chromophore at the expense of ATR.
Interestingly, the last round of evolution shows no further improvement in merocyanine
binding affinity despite a 50% increase in fluorescence between Mero-4 and Mero-6 in
E.
coli
(Figure 5
B
). Thus enhanced affinity played an important role in the early stages of Arch
evolution, but the increased
E. coli
fluorescence observed in the later rounds reflects a
change in the photophysical properties of merocyanine-bound opsin.
Spectral properties of evolved merocyanine retinal-bound Arch variants
Merocyanine-bound Arch variants were purified and their spectral properties were
characterized. For all tested variants (Figure S7), the peak excitation and emission values
were well within the NIR window (650–900 nm (
Weissleder, 2001
)). Consistent with other
red-shifted fluorescent proteins (
Shcherbakova et al., 2012
), the merocyanine-bound opsins
have a small Stokes shift (11–17 nm). However, a long emission tail extending past 800 nm
allows for fluorescence detection that is deeper into the NIR region and is well separated
from peak excitation.
The photophysical characteristics of the merocyanine-bound variants were found to have
improved over the course of directed evolution. The quantum yield (QY), extinction
coefficient (
ε
), and molecular brightness ((QY ×
ε
) / 1000) all increased slightly from wild-
type Arch to Mero-2, then decreased slightly between Mero-2 and Mero-4 (Table 1).
Between Mero-4 and Mero-6, the extinction coefficient remained the same, but the QY and
Herwig et al.
Page 7
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
therefore brightness increased significantly (~24% increase over Mero-4). Thus the two new
mutations in Mero-6 (G132S and T80A) do not affect the affinity for merocyanine (Figure
5
B
), but do enhance the QY and brightness of the merocyanine-Arch complex. The
molecular basis of this QY enhancement is difficult to pinpoint without detailed structural
information. However, a T80S mutation was shown to contribute to the fluorescence of
ATR-bound Arch in a previous study (
Hochbaum et al., 2014
), and G132S introduces a polar
group near the indolylidene ring, potentially stabilizing electron delocalization. Mutation
G132S alters a protein residue not previously known to increase Arch fluorescence.
However, a G132V mutation was involved in converting Arch from a light-driven proton
pump into a light-gated proton channel (
Inoue et al., 2015
).
To assess the spectral properties of these Arch variants evolved around a synthetic
chromophore, we drew comparisons to other fluorescent proteins. Compared to the
exceptionally dim fluorescence of wild-type Arch bound to native ATR, merocyanine-bound
Mero-6 is 16- to 200-fold brighter. The range reflects the fact that quantum yield depends on
light intensity, which makes ATR-bound wild-type Arch brighter at high intensity light
(
Maclaurin et al., 2013
) and dimmer at low intensity light (
Kralj et al., 2011a
). An
engineered variant of Arch with seven mutations, termed ‘Arch-7,’ had the highest
molecular brightness of any ATR-bound Arch and was also the furthest red-shifted, with an
excitation peak at 615 nm (
McIsaac et al., 2014
). Compared to Arch-7, merocyanine-bound
Mero-6 is 8.5 times brighter and 143 nm further red-shifted, reaching an unprecedented peak
excitation wavelength of 759 nm. Mero-6 has about 68% the molecular brightness of the
soluble red-fluorescent protein mCherry (
Kremers et al., 2011
), but is excited by light 172
nm further red-shifted. Bacterial phytochrome photoreceptors (BphPs) have been developed
as NIR-fluorescent cellular markers (
Shcherbakova et al., 2015b
). The brightest BphP is
50% brighter than Mero-6; however, Mero-6 is excited by light 120 nm further red
(
Shcherbakova et al., 2015a
). Compared to the most red-shifted BphP, Mero-6 is excited 57
nm further into the NIR and is 84% brighter (
Shcherbakova and Verkhusha, 2013
). These
results demonstrate that chromophore substitution in microbial rhodopsins using a tailored
chromophore and subsequent directed evolution is a powerful and effective way to engineer
desired fluorescent protein properties.
Concluding Remarks
Customized fluorescent proteins can be generated by engineering naturally occurring
scaffolds to bind synthetic chromophores (
Paige et al., 2011
;
Plamont et al., 2016
;
Tamura
and Hamachi, 2014
;
Yapici et al., 2015
). Chromophore-dependent microbial opsins provide
an excellent platform for this approach since modified retinals are well accepted and their
incorporation can dramatically alter and enhance opsin properties (
Albeck et al., 1989
;
Asato et al., 1990
;
AzimiHashemi et al., 2014
;
Gaertner et al., 1981
;
Ganapathy et al., 2015
;
Hoischen et al., 1997
;
Nielsen, 2009
;
Sineshchekov et al., 2012
). We have used a synthetic
chromophore, merocyanine retinal, to generate a near-infrared fluorescent protein that is 200
nm red-shifted and at least 16-fold brighter than WT Arch bound to its natural chromophore,
ATR. Directed evolution of Arch around the highly conjugated merocyanine retinal allowed
us to enhance chromophore affinity, diminish affinity for the native retinal, and further
augment pH-sensitive fluorescent brightness specifically in the NIR window. The Arch
Herwig et al.
Page 8
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
variants were engineered in
E. coli
and are well suited to bacterial imaging applications.
Although we found that merocyanine-dependent NIR fluorescence could be detected in
eukaryotic (HEK) cells, expression and stability of Arch should be optimized for eukaryotic
imaging applications. With appropriate screens, one could use directed evolution to enhance
other properties of interest such as enhanced expression in specific cell types, high voltage
sensitivity, or pumping activity using tailored synthetic retinals that are sensitive to specific
wavelengths. We anticipate that this general approach will continue to provide fine-tuned
properties and functions useful for optogenetic sensors or actuators (
AzimiHashemi et al.,
2014
;
Sineshchekov et al., 2012
), energy harvesting (
Ganapathy et al., 2015
;
Walter et al.,
2010
), or cell labeling and imaging applications (
Albeck et al., 1989
;
Hoischen et al., 1997
).
Significance
At the core of every fluorescent protein, a chromophore absorbs light and emits it as
fluorescence. The molecular brightness and color of this fluorescence are controlled by the
chromophore and surrounding protein environment. Near-infrared (NIR) fluorescent proteins
are desirable for deep tissue imaging yet their development represents a significant
engineering challenge. Here, we engineer both protein and chromophore in
Archaerhodopsin-3 (Arch) to access fluorescent properties in the near-infrared region. By
evolving Arch around a synthetic chromophore, we obtained a near-infrared fluorescent
rhodopsin with exceptional molecular brightness. This engineering strategy provides an
efficient route to develop rhodopsin complexes with properties relevant for optogenetics,
energy harvesting, or
in vivo
imaging applications.
Experimental Procedures
Synthesis and characterization of merocyanine retinal
In brief, merocyanine retinal was synthesized (Figure S1) as described (
Hoischen et al.,
1997
) and characterized by NMR and high-resolution mass spectrometry (Data S1). Full
details of the synthesis are provided in the supplemental information.
Cloning, plasmids and bacterial strains
An
Escherichia coli
codon-optimized version of 6xHis-tagged wild-type Arch (g-block from
Integrated DNA Technologies (IDT)) was used for mutant library construction and
fluorescence screening. Mutations identified in the 6xHis-tagged wild-type Arch construct
were transferred to the previously described Arch construct (pETME14-CFP) (
McIsaac et
al., 2014
) for Arch fluorescence normalization (wild-type sequences are given in the SI). For
cloning and directed evolution, we used electroporation and
E. cloni
®
EXPRESS
BL-21(DE3) cells (Lucigen). Chemically competent NiCo pLEMO cells (NEB) were used
for large-scale expression.
Directed evolution
Library construction—
Random mutagenesis libraries were generated via error-prone
PCR with a range of MnCl
2
concentrations (100 – 600 μM MnCl
2
). For each PCR reaction,
we used a final concentration of [0.08 Units/μL Taq polymerase; 400 μM dNTPs; 1× Taq
Herwig et al.
Page 9
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
standard buffer; 0.2 μM of each primer] in 100 μL total reaction volume. The error-prone
PCR thermocycler conditions were as follows: 2 min at 95 °C (1×); 30 s at 95 °C, 30 s at
55 °C, 1 min at 72 °C (30×); 10 min at 72 °C (1×),10 °C
∞
. This resulted in average library
error rates of 2–3 nucleotides per gene. The mutated Arch PCR products were cloned into
the pET21a expression plasmid (EMD Millipore) by isothermal assembly (
Gibson et al.,
2009
). Ultimately, 1,700 colonies from the 500 μM MnCl
2
library on wild-type Arch and
1,800 colonies from the 600 μM MnCl
2
library on Mero-4 were screened for increased opsin
fluorescence as described below. Site-saturation mutagenesis libraries were generated by
amplifying the parental plasmid DNA (Mero-2) with mutagenic NNK primers as described
previously (
Engqvist et al., 2014
).
Expression and screening of mutant Arch libraries—
Screening of the error-prone
PCR and site-saturation libraries was done using a 6xHis-tagged version of Arch at pH 7.
Putative improved variants were re-screened in sextuples at pH 5, pH 7, and pH 9. Mutations
were identified by Sanger sequencing (Laragen). Single colonies from libraries were
selected with sterile toothpicks and inoculated into 300 μL of Luria broth (LB; 100 μg/mL
carbenicillin) in sterile deep-well 96-well plates. Plates were sealed with EasyApp
microporous films (part no. 2977–6202; USA Scientific, Inc.). Following overnight growth
at 37 °C, 220 rpm and 80% humidity, pre-cultures were diluted 1:20 into 1 mL of fresh LB
(100 μg/mL carbenicillin) and grown for 2 h at 30 °C. Then merocyanine retinal (1 μM final
concentration) and isopropyl
β
-D-1 thiogalactopyranoside (IPTG; 500 μM final
concentration) were added to each well, and proteins were expressed for 4–6 h in the dark.
Next, cells were centrifuged at 4,000 RPM in a swinging bucket rotor (Beckman Coulter,
Allegra TM 25R Centrifuge). The resulting pellets were resuspended in 700 μL of 200 mM
NaCl, and 180 μL of the resuspension was added to 20 μL of 500 mM potassium phosphate
buffer (at pH 7 for the initial screen and pH 5, 7 and 9 in the re-screen) in a measurement
plate (Evergreen Scientific, untreated 96-well microplates, catalog number: 290-8115-01F).
Raw Arch fluorescence was measured using a Tecan Infinite® M200 plate-reader at an
emission wavelength of 785 nm following excitation at 760 nm. The mutations identified
were transferred into the pETME14-CFP construct (
McIsaac et al., 2014
). The C-terminal
cyan fluorescent protein (CFP; ex/em 425/475 nm) tag was used as a proxy for expression
level and to calculate normalized opsin fluorescence. The reported normalized fluorescence
is defined as 1000 × [opsin fluorescence/CFP fluorescence]. Screening of site-saturation
libraries was performed as described above. We analyzed 88 clones for each site, for 94%
coverage of the possible diversity (
Engqvist et al., 2014
). Data from all libraries were
processed and analyzed using Microsoft Excel and GraphPad Prism (version 6.04 for
Windows; GraphPad Software) software. Cells expressing a non-fluorescent protein
ScADH6 (
Saccharomyces cerevisiae
cinnamyl alcohol dehydrogenase) were used as the
negative control in all rounds of directed evolution.
Arch homology model
A homology model, built using the Swiss modeler web server (
Arnold et al., 2006
;
Bordoli
et al., 2008
), was used to identify residues of interest for site-saturation mutagenesis (Figure
3). A three-dimensional model of merocyanine retinal was generated using the NCI
CACTUS server (
http://cactus.nci.nih.gov/translate/
) and manually inserted into the
Herwig et al.
Page 10
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
homology model via Coot (
Emsley and Cowtan, 2004
) and the UCSF Chimera package
(
Pettersen et al., 2004
).
Purification of Arch variants
Arch variants of interest were expressed and purified as fully described in the SI. In brief, 1L
cultures of Arch variants were grown at 37°C, induc ed at mid-log growth by adding
merocyanine retinal and IPTG (final concentrations of 1 μM and 0.5 mM, respectively),
grown post-induction for 3 hours at 30°C in the dark, the n harvested and stored at −80 °C.
Holo
-Arch variants were purified as reported (
McIsaac et al., 2014
) via Ni-affinity
chromatography. Arch-containing elution fractions were pooled and immediately desalted
into ‘DDM desalt buffer’ [20 mM Tris-HCl; pH 6.5; 200 mM NaCl; 0.15% DDM] via
PD-10 desalting columns.
For
in vitro
chromophore binding studies,
apo
-Arch variants were obtained from cultures
induced in the absence of chromophore. All steps were carried out at 4 °C. Thawed cells
were lysed via microfluization in the absence of detergent. A crude membrane fraction was
collected via ultra-centrifugation, stored at −80 °C, and sol ubilized with 1.5% lauryl
maltose neopentyl glycol (LMNG) detergent. Solubilized
apo
-Arch was then purified via Ni-
affinity chromatography. Arch containing fractions were identified via SDS-PAGE and
desalted into [20 mM Tris-HCl; pH 7.5; 200 mM NaCl; 0.015% LMNG] (‘LMNG desalt
buffer’) via PD-10 desalting columns. For the binding assay, purified proteins were
concentrated no more than 2× via spin filtration (Millipore).
K
d
measurements in
E. coli
and of purified proteins
For the
E. coli K
d
measurements, protein expression and fluorescence measurements were
performed as described in the directed evolution methods. For each retinal concentration and
variant, the fluorescence emission was recorded for the opsin (ex/em 760/785 nm) and CFP
(ex/em at 425/475 nm) after 4 hours of post-induction incubation with merocyanine retinal.
Each variant was tested under the following merocyanine retinal concentrations: 32, 16, 4, 1,
0.25, 0.125, 0.063 and 0 μM. The use of an isomeric mixture of merocyanine retinal (3:1
trans
: 16-
cis
) may lead to underestimated chromophore binding affinities; however, trends in
total merocyanine retinal binding affinity can still be assessed. Normalized emission
intensity was calculated in quadruplicate and the average with standard deviation was plotted
versus the retinal concentration. The data were fitted to the following equation F =
(F
max
[retinal]/(
K
d
+ [retinal]), where F is the observed fluorescence and F
max
is the
calculated maximal fluorescence of the retinal-opsin complex.
K
d
values of the selected
Arch variants were determined from the generated plot using GraphPad Prism (version 6.04
for Windows; GraphPad Software) software. Co-translational chromophore binding in this
assay would lead to variable on- and off-rates during the course of protein maturation, which
would complicate the assumption of equilibrium. So the
K
d
values measured in this assay
are referred to as ‘apparent
K
d
.’
In vitro
, the binding of ATR or merocyanine retinal to
apo
-Arch was monitored by the ‘opsin
shift’ in chromophore absorbance (
Baloghnair et al., 1981
;
Booth et al., 1996
). To measure
the binding affinity, purified protein, LMNG desalt buffer, and finally retinal dissolved in
Herwig et al.
Page 11
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
ethanol were added to a 96-well half-area plate for a final volume of 100 μL (final
concentration of ethanol < 1%). With a constant concentration of protein, a range of retinal
concentrations was used (0, 0.1, 0.25, 0.5, 0.75, 1, 2, 5, and 10 molar equivalents of retinal
to protein) so that a binding curve could be determined. The final protein concentration was
set to 0.508 mg/mL for wild-type Arch, Mero-2, and Mero-4 (0.432 mg/mL for Mero-6),
which would allow accurate absorbance measurements of the
holo
protein. At these protein
concentrations, more than a 10:1 molar equivalent of merocyanine retinal to protein resulted
in protein aggregation. For each Arch variant and concentration, the binding affinity of ATR
and merocyanine retinal were measured in triplicate. K
d
values calculated from data
measured after 24 and 36 hours of incubation remain consistent (average percent difference
of 9%) indicating that binding has reached equilibrium by 36 hours; thus the 36 hour data
were used to calculate binding affinities in Table 1. Using a custom fitting function in
Matlab (MATLAB 8.5, The MathWorks Inc., 2015), curves were fitted to the data without
the common assumption that added retinal approximates free retinal (i.e. accounting for
ligand depletion) (
Swillens, 1995
). For this fitting scheme,
holo
protein absorbance was
converted to concentration via Beer’s law; the path-length for this volume of sample was
calculated at 0.6636 cm, the extinction coefficients for merocyanine-bound Arch variants are
determined below, the extinction coefficient for ATR-bound Arch was reported previously
(
McIsaac et al., 2014
), and the extinction coefficient for ATR bound Mero-6 was
approximated by the wild-type value. We found that the
K
d
parameter of the fit was only
marginally influenced by changes in extinction coefficient.
Quantum yield and extinction coefficient determination
Spectral properties of purified, merocyanine-bound variants were characterized as previously
described (
McIsaac et al., 2014
;
Wall et al., 2015
) and detailed in the SI. In brief, emission
spectra (Figure S7A) were collected via plate reader with excitation at 700 nm and emission
detected between 730 and 850 nm; excitation spectra (Figure S7B) were collected with
detection at 810 nm and excitation scanned from 690 to 790 nm. As shown in Figure S9A–
B, the quantum yield for each merocyanine-bound Arch variant was calculated by
comparison to the to the Alexa Fluor® 750 NHS Ester (succinimidyl ester) dye with known
quantum yield of 0.12 (catalog number: A20011, Life Technologies Corporation) (
Wurth et
al., 2012
). The extinction coefficient of free merocyanine retinal (oxime) was determined via
dilution and absorbance measurements. The extinction coefficient of each merocyanine-
bound Arch variant was determined by comparing changes in free and bound chromophore
while chemically bleaching the
in vitro
sample (Figure 9C–F).
Cell maintenance for live cell imaging
Starting from an overnight pre-culture (LB + 100 μg/mL carbenicillin), 5 mL
E. coli
cultures
were grown to early log phase (0.4–0.6 OD
600
) at 37 °C, then induced with 1 μM
merocyanine retinal and 0.5 mM IPTG (final concentrations) at 30 °C for 3 hours. Cells
were harvested via centrifugation and stored on ice. Within three hours, the cells were
resuspended in 1× PBS and aliquoted on freshly prepared agarose pads for imaging.
As fully described in the supplemental methods, HEK 293T cells were grown in D10
medium (Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% (vol/vol)
Herwig et al.
Page 12
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript
FBS, 1% sodium bicarbonate and 1% sodium pyruvate) and transfected with GFP-tagged
Arch variants that had been codon optimized for mammalian expression. Twenty-four hours
after transfection, the D10 medium was replaced with D10 medium supplemented with 1
μM merocyanine retinal. Imaging was done 48 h post transfection. After this prolonged
incubation, HEK cells were adherent with normal morphology and healthy appearance
(Figure S8).
E. coli
and HEK cell imaging and data processing
As fully described in the supplemental methods, cells were imaged with two orthogonal
channels (405 nm and 727 nm for
E. coli
or 473 nm and 727 nm for HEK) with
corresponding filter cube sets (detection at >418 nm, >498 nm, and 766–854 nm for laser
illumination at 405 nm, 473 nm, and 727 nm respectively). Two oil objectives, Olympus NA
1.40 UPlanSApo 100× with additional 1.6× magnification for
E. coli
and Olympus NA 1.35
UPlanSApo 60× for HEK cells were used for imaging. The camera was back-illuminated
CCD Andor iKon-M 934 BEX2-DD, offering high quantum efficiency in the near infrared
region with a pixel size of 13×13 μm. Fluorescence analysis of
E. coli
clusters was done by
masking the background in the CFP image (via an otsu threshold of pixel intensity counts)
and determining the mean pixel intensity within the mask (signal) and outside the mask
(background). The background-corrected mean pixel intensities are reported in Figure 4
B–C
.
The NIR/CFP ratio was determined for each cluster of cells and the ratios were averaged for
the values given in Figure 4
D
. Fluorescence analysis of single HEK cells was done by
manually selecting regions around each cell and separately a background region in open
source ImageJ (version v1.48). Mean fluorescence intensity measurements were recorded for
each region of interest (ROI). Background mean intensity was then used to background
subtract from the cell mean intensity.
Fluorescence photo-bleaching in live
E. coli
As fully described in the SI, fluorescence decay rates were measured for both CFP (111
mW, ex/em at 405/464–500 nm, 500 ms exposure) and merocyanine-bound Arch (32 mW
ex/em at 727/770–840 nm, 250 ms exposure) allowing Arch decay rates to be presented
relative to CFP. The fluorescence decay curves (between 3 and 60 s) were fit with a single
exponential using the scipy.stats module in python. The [NIR fluorescence / CFP
fluorescence] ratio of exponential decay rates was calculated for each spot of clustered
E.
coli
cells. The mean ratio for a given Arch variant (n = 4–5 spots) is given in Table 1.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
Funding
L.H. was supported by a fellowship from the SNSF (Swiss National Science Foundation). The Ruth L. Kirschstein
National Research Service Award supports A.J.R (F32GM116319), C.N.B (F31MH102913), and S.C.D
(5F32GM106618). R.K.Z. was supported by a National Science Foundation Graduate Research Fellowship (NSF
GRFP; DGE-1144469), is a trainee in the Caltech Biotechnology Leadership Program, and has received financial
support from the Donna and Benjamin M. Rosen Bioengineering Center. J.K.B.C. acknowledges the support of the
Herwig et al.
Page 13
Cell Chem Biol
. Author manuscript; available in PMC 2018 March 16.
Author Manuscript
Author Manuscript
Author Manuscript
Author Manuscript