1
Introduction
Organisms of all domains
of life use photoreceptor proteins to
sense and respond to light. The light-sensitivity of photoreceptor
proteins arises from bound chromophores such as retinal in retinylidene
proteins, bilin in biliproteins, and flavin in flavoproteins. Rhodopsins
found in Eukaryotes, Bacteria, and Archaea consist of opsin apoproteins and a covalently
linked
retinal which is employed to absorb photons for energy conversion
or the initiation of intra- or intercellular signaling.
1
Both functions are important for organisms to
survive and to adapt to the environment. While lower organisms utilize
the family of microbial rhodopsins for both purposes, animals solely
use a different family of rhodopsins, a specialized subset of G-protein-coupled
receptors (GPCRs).
1,2
Animal rhodopsins, for example,
are employed in visual and nonvisual phototransduction, in the maintenance
of the circadian clock and as photoisomerases.
3,4
While
sharing practically no sequence similarity, microbial and animal rhodopsins,
also termed type-I and type-II rhodopsins, respectively, share a common
architecture of seven transmembrane α-helices (TM) with the
N- and C-terminus facing out- and inside of the cell, respectively
(Figure 1).
1,5
Retinal is
attached by a Schiff base linkage to the ε-amino group of a
lysine side chain in the middle of TM7 (Figures 1 and 2). The retinal Schiff base
(RSB) is
protonated (RSBH+) in most cases, and changes in protonation
state are integral to the signaling or transport activity of rhodopsins.
Figure 1
Topology
of the retinal proteins. (A) These membrane proteins contain
seven α-helices (typically denoted helix A to G in microbial
opsins and TM1 to 7 in the animal opsins) spanning the lipid bilayer.
The N-terminus faces the outside of the cell and the C-terminus the
inside. Retinal is covalently attached to a lysine side chain on helix
G or TM7, respectively. (B) Cartoon representation of the helical
arrangement of a microbial rhodopsin with attached all-trans-retinal (bacteriorhodopsin,
PDB ID: 1C3W).
Figure 2
Genesis of the chromophore of microbial and animal rhodopsins.
Cleavage of β-carotene is the source of the chromophore. The
ground state of microbial and animal rhodopsins possesses all-trans- and 11-cis-retinal
as its chromophore,
respectively, bound to a Lys residue via a Schiff base, which is normally
protonated and exists in the 15-anti configuration.
It should be noted that microbial rhodopsins depend exclusively on
all-trans-retinal, while some animal rhodopsins possess
vitamin A2 (C3=C4 double bond for fish visual pigments) and
hydroxyl (C3—OH for insect visual pigments) forms of 11-cis-retinal. Usually, photoactivation
isomerizes microbial
rhodopsin selectively at the C13=C14 double bond and animal
rhodopsin at the C11=C12 double bond.
Retinal, the aldehyde of vitamin A, is derived from β-carotene
and is utilized in the all-trans/13-cis configurations in microbial rhodopsins and
the 11-cis/all-trans configurations in animal rhodopsins (Figure 2).
1,6
For optimal light to energy or
light to signal conversion, defined chromophore–protein interactions
in rhodopsins direct the unique photophysical and photochemical processes,
which start with specific retinal isomerization and culminate with
distinct protein conformational changes. The protein environment is
typically optimized for light-induced retinal isomerization from all-trans → 13-cis
in microbial rhodopsins
and for 11-cis → all-trans in animal rhodopsins. Variations in this isomerization
pattern are
discussed in sections 3 and 4.
The 7TM protein scaffold of microbial rhodopsins
is designed for
light-driven ion pumps, light-gated ion channels, and light sensors
which couple to transducer proteins (Figure 3).
7
Microbial rhodopsins were first found
in the Archaea (Halobacterium salinarum, historically referred to as Halobacterium
halobium)
8
and were therefore initially termed
archaeal rhodopsins. H. salinarum contains bacteriorhodopsin
(BR)
8
and halorhodopsin (HR)
9
that act as a light-driven outward proton pump
and inward chloride pump, respectively. As ion pumps, they contribute
to the formation of a membrane potential and thus have their function
in light–energy conversion. The other two H. salinarum rhodopsins are sensory rhodopsin
I and II (SRI and SRII),
10
which act as positive and negative phototaxis
sensors, respectively. Since the original discovery of BR in H. salinarum, similar
rhodopsins have been found in Eubacteria and lower Eukaryota, leading
to the name microbial rhodopsins. For example, Anabaena sensory rhodopsin (ASR), the
first sensory rhodopsin observed in
the Eubacteria,
11
is a
sensor that activates a soluble transducer (Figure 3).
Figure 3
Microbial rhodopsins can function as pumps, channels, and light-sensors.
Arrows indicate the direction of transport or flow of signal: (A)
light-driven inward chloride pump (halorhodopsin (HR), PDB ID: 1E12), (B) light-driven
outward proton pump (bacteriorhodopsin (BR), PDB ID: 1C3W), (C) light-gated
cation channel (channelrhodopsin (ChR), PDB ID: 3UG9), (D) light-sensor
activating transmembrane transducer protein (sensory rhodopsin II
(SRII), PDB ID: 1JGJ), (E) light-sensor activating soluble transducer protein (Anabaena
sensory rhodopsin (ASR), PDB ID: 1XIO).
Channelrhodopsins (ChRs), another group of microbial
rhodopsins,
were discovered in green algae where they function as light-gated
cation channels within the algal eye to depolarize the plasma membrane
upon light absorption.
12,13
The primary depolarization of
the eyespot membrane is transferred to the flagellar membrane and
results in a reorientation of the alga toward a light source (photophobic
responses and phototaxis). Thus, ChRs naturally function as signaling
photoreceptors as well. Discovery of ChR led to an emergence of a
new field, optogenetics,
14
in which light-gated
ion channels and light-driven ion pumps are used to depolarize and
hyperpolarize selected cells of neuronal networks, e.g., for therapeutic
reasons
15
or in order to understand the
circuitry of the brain.
16,17
Thus, studies on microbial
rhodopsins are beneficial not only for our basic understanding of
retinal proteins, but also for providing a toolset to study neuronal
signaling through optogenetics.
Animal rhodopsins belong to
the superfamily of GPCRs which detect
extracellular signals, typically by binding small molecule ligands
like hormones and neurotransmitters.
18,19
By a ligand-induced
conformational change, GPCRs become activated and capable of transducing
the activation signal by catalyzing GDP/GTP exchange on membrane-bound
heterotrimeric G proteins within the cell, thus initiating G protein-mediated
signaling cascades (Figure 4).
19
After activation-dependent phosphorylation by a G-protein-coupled
receptor kinase, active GPCRs can also interact with arrestin to effect
G protein-independent signaling and attenuation of the ligand-mediated
activation signal.
20
Animal rhodopsins
are typically specialized GPCRs, capable of detecting single photons
as a physical stimulus.
21
Because the 11-cis-retinal ligand is covalently bound within the retinal-binding
pocket of the receptor, photon absorption and the ensuing retinal cis → trans isomerization
convert
an inactivating ligand (the inverse agonist 11-cis-retinal) into an activating ligand
(the agonist all-trans-retinal) in situ. Vertebrate rhodopsin was discovered
more than 130 years ago and has long been used as a prototypical GPCR.
22
Due to the relative ease of purification from
native material, it has been studied extensively.
2
Figure 4
Animal rhodopsins are specialized G-protein-coupled receptors (GPCRs).
(A) Binding of extracellular ligands stabilizes certain GPCR conformations
which enable the GPCR to catalyze GDP/GTP exchange in heterotrimeric
G proteins (Gαβγ) and/or to induce G-protein-independent,
arrestin-mediated signaling. (B) Typical GPCR fold shown in cartoon
representation for bovine rhodopsin (PDB ID: 1U19). Structures of
animal and microbial rhodopsins differ largely (cf. Figure 1B) and are drawn in opposite
orientations with respect
to the membrane. As model for the GPCR family, animal rhodopsin is
shown in the orientation commonly used for GPCRs. In a large number
of publications, animal rhodopsins are shown for historical reasons
in the orientation of microbial rhodopsins (C-terminus up).
The goal of this review is to
provide mechanistic insights from
biophysical and structural studies into the function of microbial
and animal rhodopsins, with the latter as representatives of GPCRs.
After a general description of retinal photoisomerization in section 2, the functional
mechanisms of various
basic types of microbial rhodopsins will be discussed in section 3. In section 4,
animal rhodopsins will be reviewed with a focus on bovine visual rhodopsin
(abbreviated as Rho in some cases), the photoreceptor in retinal rod
cells, followed by some extension to color visual rhodopsins and invertebrate
rhodopsins.
2
Light Absorption and Photoisomerization
2.1
Color Tuning
Light absorption initiates
functions of both microbial and animal rhodopsins,
23,24
and the wavelength dependence of the absorption efficiency determines
the colors of the proteins (Figure 5). The
length of the π-conjugated polyene chain in the retinal chromophore
as well as the protonation of the RSB linkage determine the energy
gap of the π–π* transition,
25
so that the absorption of most rhodopsins is within the
visible region (400–700 nm). Humans have a single photoreceptor
for dim light vision (rhodopsin, λmax ∼500
nm) and three receptors for color vision (blue, λmax ∼425 nm; green, λmax ∼530 nm;
red,
λmax ∼560 nm),
26,27
whereas some
shrimp species contain up to 16 rhodopsins covering the spectral range
from 300 to 700 nm.
28
While the chromophore
molecule is usually the same in all pigments (retinal bound via a
(protonated) Schiff base), the absorption maxima differ significantly,
implying an active protein control of the energy gap between the ground
and excited states of the retinal chromophore. The mechanism of color
tuning has fascinated researchers for a long time, and several factors
have been determined to be responsible for it.
Figure 5
Microbial rhodopsins
exhibit a wide range of absorption maxima.
Colors of microbial rhodopsins (A) and their absorption spectra (B).
The following rhodopsins are shown: (1) a blue-proteorhodopsin (LC1-200,
pH 7), (2) Q105L mutant of LC1-200 (pH 7), (3) a green-proteorhodopsin
(EBAC31A08, pH 7), (4) A178R mutant of green-proteorhodopsin (pH 7),
(5) bacteriorhodopsin (pH 7), (6) H. salinarum sensory
rhodopsin I (pH 4).
The protonation state
of the chromophore plays a crucial role in
color tuning; the unprotonated RSB absorbs in the UV region (λmax ∼360–380 nm), and
this absorption is quite
insensitive to the environment in contrast to the RSBH+, which exhibits a large variation
in absorption covering the entire
visible light spectrum. Other factors defining the spectral tuning
of individual rhodopsins are given by chromophore–protein interactions
such as electrostatic interactions with charged and polar amino acids,
termed electrostatic tuning and extensively studied, first using retinal
analogues,
29−32
and, later, site-directed mutagenesis.
33−35
Electrostatic
tuning was elegantly demonstrated in a model system based on cellular
retinol-binding protein II. This system was engineered to covalently
bind all-trans-retinal and its absorption maximum
was changed from 425 to 644 nm via mutations that changed the electrostatic
potential within the retinal-binding pocket.
36
Interactions of retinal with charged, polar, and aromatic amino
acids play a role in changing the electronic energy levels, as do
hydrogen-bonding interactions and steric contact effects. Strong hydrogen
bonds can lead to charge transfer, and steric contacts can lead to
a twist of retinal. All these tuning processes in concert shape the
absorbance maxima of retinal in microbial and animal rhodopsins.
One of the most prominent factors in color tuning is the interaction
of retinal with the counterion(s) (Figure 6). In the ground state, the retinal chromophore
is positively charged
due to RSBH+ (C=NH+). The excited state
has strong charge transfer character where the positive charge is
displaced toward the β-ionone ring, leading to a neutralization
of the RSBH+ (Figure 6B).
37,38
Interaction of the RSBH+ with the negatively charged
counterion(s) in microbial and animal rhodopsins leads to an electrostatic
stabilization in the electronic ground state of retinal accompanied
by an increase of the RSB pKa (Figure 6A). The resulting larger energy gap between
ground and excited
states causes a blue-shift of the absorption (Figure 6D, compare cases [A] and [B]).
39,40
If a negative
charge is located near the β-ionone ring, the excited state
is energetically stabilized compared to the ground state (Figure 6C,D), which leads
to a smaller energy gap and therefore
to a red-shift of the wavelength of electronic excitation. As the
absorption maximum of isolated all-trans RSBH+ in gas phase is 610 nm (Figure 6B,D,
case [B]),
40
in principle, absorption in
the deep red range (λmax > 600 nm) should be possible
for case [C] in Figure 6, while λmax < 600 nm is expected for case [A]. However, microbial
and animal rhodopsins only exhibit λmax < 600
nm, with some exceptions in Crustaceae that absorb
beyond 600 nm, indicating that case [C] is less favorable, and indeed,
a negative charge near the RSBH+ (case [A]) is found in
the crystal structures of microbial and animal rhodopsins, where water-containing
hydrogen-bonding networks constitute complex counterions.
Figure 6
Color tuning
exemplified by visual rhodopsins (containing 11-cis-retinal and RSBH+). Photoexcitation
causes
bond alteration, which leads in the electronically excited state to
the movement of the positive charge from the RSBH+ to the
β-ionone ring. (A) Excitation when a negatively charged counterion
is close to the positively charged RSBH+. (B) Excitation
in the absence of a negatively charged counterion. (C) Excitation
when the chromophore counterion is located near the β-ionone
ring. (D) Electrostatic interactions with the counterion lower the
energy level of the ground state (case [A]) or excited state (case
[C]), yielding a spectral blue- or red-shift, respectively.
It should be noted that, historically,
the protein-induced spectral
shift of retinal absorption (so-called opsin shift) has been considered
bathochromic (i.e., red), because the absorption maximum of RSBH+ in organic solvents
(∼440 nm),
41
rather than the more recently measured absorption in the
gas phase (∼610 nm),
40
served as
a reference point. The interaction with individual polar and polarizable
residues exerts a much smaller influence, on the order of several
nanometers. The overall effect of the opsin is, in most cases, to
counteract the blue-shift induced by the counterion. The delicate
interplay of these electrostatic effects can now be understood in
detail.
42−45
Further spectral tuning can be achieved by other factors.
Figure 2 shows that the retinal chromophore
in microbial
and animal rhodopsins differs not only in configuration of the C11=C12
double bond, but also in the conformation of the C6–C7 single
bond. For bovine rhodopsin it is known that the C6–C7 bond
is in a 6-s-cis conformation,
46,47
making the polyene chain and β-ionone ring not planar due
to steric hindrance between the C8-hydrogen and the C5-methyl group.
As a result, the conjugation of π-electrons does not extend
to the β-ionone ring efficiently. For microbial rhodopsins,
however, the C6–C7 bond is 6-s-trans,
1
although the C6–C7 6-s-cis conformer is more stable in solution.
48
As a consequence, an extended conjugation of π-electrons becomes
possible from the polyene chain to the β-ionone ring, which
presumably contributes to the considerable spectral red-shift observed
in microbial rhodopsins. In fact, while absorbance spectra of protonated
Schiff bases of all-trans- and 11-cis-retinal in solution are similar (λmax ∼450
nm),
49
most microbial and animal rhodopsins
typically possess λmax in 520–580 nm and 480–525
nm ranges, respectively,
50
which can in
part be explained by the differences in the C6–C7 bond conformation.
In the 480–525 nm range, a C6–C7 6-s-cis-retinal conformer is found predominantly.
However, it should be
noted that SRII and ChR absorb maximally at 470–500 nm, even
though these microbial rhodopsins possess the C6–C7 6-s-trans conformation,
51−53
while some visual pigments are
substantially red-shifted.
45
The presence
of a different chromophore (11-cis-3,4-didehydroretinal,
e.g., in some fish species) or a different counterion (e.g., where
Glu181 (in bovine Rho) is replaced by a His residue to form a chloride
binding site in green and red color pigments) can contribute to this.
These facts reveal the complexity of color tuning mechanism in microbial
and animal rhodopsins, and the importance of structural information
in understanding the mechanistic basis of color tuning in rhodopsins.
Pioneering calculations by Birge, Schulten, Tavan, and Warshel
gave interesting qualitative insights into the electronic structure
of retinal and basic mechanisms of color tuning.
54−56
However, only
the combination of high-resolution structural data for many rhodopsins
with more accurate computational methods has allowed estimating the
contributions of the different mechanisms of color tuning. Retinal
proteins are highly challenging for study by quantum chemical methods,
57
necessitating QM/MM approaches where the retinal
and a few surrounding amino acids are included in the quantum mechanical
(QM) description and the remainder of the system is treated by molecular
mechanics (MM). This coupled analysis leads to a description of steric
effects, but most importantly, the electrostatic environment of retinal
is provided by fixed point charges on the atoms treated by MM. For
very accurate predictions, the fixed point charge representation of
MM may even become too simplistic, and a more complicated model is
required.
58−60
QM/MM calculations have been applied to many
microbial rhodopsins
such as BR,
61−64
HR,
58,65
SRII,
66,67
and ASR.
68
The color tuning mechanism has been examined
for both bovine Rho
44,64,69−73
and squid rhodopsin.
74
In the case of
color visual pigments, computational studies based on the homology
models with bovine Rho have been reported,
75−77
and a similar
approach was applied to green- and blue-absorbing proteorhodopsins.
43,78
These QM/MM studies gave detailed insight into the mechanisms of
color tuning, since the calculations allow the different mechanistic
factors to be separated. The environment tunes the absorption wavelength
of retinal, but there are limits to color tuning, as other properties
of retinal are affected by the environment, as well. The pK
a of the RSBH+ is affected by steric
and electrostatic factors,
57
as is the
isomerization efficiency of retinal, which are both crucial for retinal
protein activation.
73
Therefore, extensive
color tuning can be in conflict with other functions of rhodopsins,
limiting the possible protein modifications for efficient color tuning.
2.2
Primary Photochemical Reactions
The
initial photochemical reaction in microbial and animal rhodopsins
is known to be one of the fastest and most efficient chemical events
in biology. The fate of the photoexcited molecules is determined within
several hundreds of femtoseconds, and to avoid nonproductive de-excitation
back to the ground state, microbial and animal rhodopsins have optimized
the primary photochemical reaction in a specific manner.
79−82
The quantum yield of the photoreaction of bovine Rho is very high
and is essentially independent of temperature and excitation wavelength,
83−85
while the fluorescence quantum yield is quite low (Φ ∼
10–5).
86
Since the radiative
lifetime of nearly all rhodopsins is expected to be within the range
1–10 ns (from the Strickler–Berg equation),
87
these extremely low fluorescence quantum efficiencies
led to the conclusion that the ultrafast reaction occurs through a
barrierless excited-state potential surface.
88
Because the formation of the primary photoproduct takes place for
rhodopsins even at liquid nitrogen (77 K) and liquid helium (4 K)
temperatures, where molecular motions are essentially arrested,
89,90
the cis→trans isomerization
as the primary event in vision was initially questioned.
91,92
Figure 7 shows the pathways of photochemical
reactions in microbial and animal rhodopsins, represented as a combination
of transitions between the ground and excited electronic states and
evolution along the isomerization reaction coordinate (dihedral angle
of the respective bond). As the energy barrier for the rotation around
the double C=C bond is prohibitively high, isomerization becomes
possible only in the excited state, where the chromophore becomes
twisted before relaxing into the ground state through the so-called
conical intersection (CI), the point of the closest approach of the
energy surfaces. In microbial rhodopsins, all-trans-retinal is typically isomerized
into the 13-cis form, even though the reverse reaction is possible for 13-cis-forms
of several rhodopsins. The selectivity of isomerization
is 100%, and the quantum yield was found to be 0.64 for BR.
93,94
In animal rhodopsins, such as bovine rhodopsin, the isomerization
of 11-cis-retinal to all-trans-retinal
occurs with 100% selectivity and a quantum yield of 0.67.
95
It should be noted that some photoisomerases,
such as retinochromes, and circadian and visual bistable pigments,
such as melanopsin and squid rhodopsin, can perform photoisomerization
in the opposite direction.
96−98
Photoisomerization in both microbial
and animal rhodopsins has much higher efficiency (4–5 times)
and selectivity than in solution, suggesting that rhodopsins developed
highly efficient isomerization pathways which favor their respective
photoproducts.
Figure 7
Suggested mechanism of retinal photoisomerization in rhodopsins.
Potential energy profiles along the reaction coordinate (the dihedral
angle of C11=C12 and C13=C14 bonds of animal (left)
and microbial (right) rhodopsins): S0, ground state; S1, the first electronic excited
state. Colored arrows represent
excitation by visible light. C.I. represents conical intersection,
the point of the closest approach of the energy surfaces of the ground
and the excited states, through which transitions are the most probable.
In the case of animal rhodopsins,
low-temperature spectroscopy
of bovine Rho at 87 K detected a trapped, red-shifted photointermediate,
now called bathorhodopsin.
89
Bathorhodopsin
converts to lumirhodopsin upon warming and ultimately decomposes through
several intermediates to all-trans-retinal and opsin.
It was therefore proposed that bathorhodopsin has a higher potential
energy than dark state bovine Rho and subsequent intermediates and
contains a “highly constrained and distorted” all-trans form of the chromophore.
90
According to this assumption, the process of converting dark state
Rho to bathorhodopsin would be a cis→trans isomerization of retinal (Figure 7). When
the first experiment with picosecond resolution was
performed on bovine Rho in 1972,
91
it revealed
that bathorhodopsin forms within 6 ps after excitation at room temperature,
and the interpretation was that its formation was too fast to arise
solely from a conformational change as large as the cis→trans isomerization of the
retinal chromophore,
92
but this perspective has been entirely revised by more recent experiments.
Time-resolved studies on bovine Rho containing 11-cis-ring-locked retinal analogues
provided the experimental evidence
for 11-cis to all-trans isomerization
as the primary reaction.
99
The cis→trans isomerization process
has also been observed in real time by femtosecond transient absorption
spectroscopy of bovine Rho. These experiments revealed that within
200 fs the all-trans photoproduct formation is complete,
suggesting that isomerization is a femtosecond time scale event.
100
Further, oscillatory features with a period
of 550 fs (60 cm–1) were observed in the kinetics
of the primary photoproduct of bovine Rho, whose phase and amplitude
demonstrated that they are the result of nonstationary vibrational
motions in the ground state of photorhodopsin, the precursor of bathorhodopsin.
101
Coherent vibrational motions in the isomerized
product support the idea that the primary step in vision is a vibrationally
coherent process and that the high reaction quantum yield in bovine
Rho is a result of the extremely fast excited-state torsional motion.
101
Ultrafast photoisomerization in bovine Rho
was also confirmed by other experimental techniques, such as femtosecond
fluorescence spectroscopy
102
and femtosecond
stimulated Raman spectroscopy.
103
In the case of microbial rhodopsins, light absorption by BR at
77 K causes formation of a red-shifted primary intermediate, the K
intermediate (λmax ∼590 nm). The presence
of a red-shifted K precursor termed the J intermediate (λmax ∼625 nm) was revealed
by time-resolved visible spectroscopy.
104
The excited state of BR possesses a blue-shifted
absorption, which decays in two components of about 200 and 500 fs.
105
As expected from the formation of the J intermediate
within 500 fs, photoisomerization must take place on the femtosecond
time scale.
105,106
Although the results obtained
from BR containing all-trans-ring-locked 5-membered
retinal were not as easy to interpret as those for bovine Rho,
107,108
the common principle of retinal isomerization for animal and microbial
rhodopsins has been established, namely, that photoisomerization takes
place in the excited state (Figure 7). Using
femtosecond visible-pump and infrared-probe spectroscopy enabling
higher temporal resolution, the 13-cis characteristic
vibrational band at 1190 cm–1 was observed with
a time constant of 500 fs.
109
Thus, isomerization
of all-trans to 13-cis occurs on
the femtosecond time scale, which is coincident with the formation
of the J intermediate. The appearance of the 13-cis form in less than 1 ps was also
shown through the Fourier-transform
of the transient absorption data with less than 5-fs resolution, supporting
that the all-trans → 13-cis isomerization takes place within femtoseconds.
110,111
Additionally, previous anti-Stokes resonance Raman spectroscopy
data suggested that the J intermediate is a vibrationally hot state
of the K intermediate, showing that its chromophore is highly twisted
and thermally excited.
112
Thus, isomerization
in the excited state is supported by experimental data for microbial
rhodopsins, too (Figure 7). More recently,
the role of vibrational coherence in the efficient excited-state photoisomerization
suggested for bovine Rho (see above) has been demonstrated for BR
via coherent control experiments.
113
While Rho and BR share a light-driven isomerization event in the
excited state, they may utilize different relaxation pathways from
the Franck–Condon state (Figure 7).
In Rho, the excited wavepacket slides down a barrierless potential
surface. For BR, however, several experiments favor a 3-state model
that postulates a small potential barrier along the isomerization
coordinate.
110,114−116
Isomerization does not occur instantly as shown by real-time spectroscopy
of BR with sub-5-fs pulse, but involves the transient formation of
a tumbling state,
110
consistent with this
3-state model. Interestingly, recent spectroscopic measurements performed
on the 13-cis forms of microbial rhodopsins BR and
ASR show an ultrafast ballistic barrierless wavepacket movement on
the excited state surface, similar to that observed in bovine Rho,
suggesting that the isomeric state of retinal may define the dynamics
of the excited state.
117
By ultrafast
spectroscopy, reduced rates and efficiency of photoisomerization
were measured for various BR mutants where charged counterion residues
had been replaced with neutral residues.
37
Thus, the efficient primary photoisomerization mechanism involves
a charged counterion complex. In addition to electrostatic interactions,
steric interactions with the methyl groups at the C9 and C13 positions
seem to be of importance. For example, upon photoisomerization in
bovine Rho, the C13-methyl group moves whereas the C9-methyl group
does not. Furthermore, to avoid collision of retinal’s C13-methyl
group and the C10-hydrogen, the C11=C12 and C12—C13
bonds are pretwisted even in the dark state. Experimental and theoretical
evidence has been provided that the C13-methyl is important for efficient
isomerization.
118,119
The C9-methyl group appears
to function as a scaffold for the primary isomerization event,
120
while its motion is important for G protein
activation in the late photointermediates.
121,122
The isomerization reaction mechanism has been further detailed
with the help of quantum chemical simulations, first of retinal in
gas phase, then with QM/MM methods that include the surrounding protein.
In both animal and microbial rhodopsins, after the photoexcitation,
retinal leaves the Franck–Condon region by relaxing the stretching
and torsional vibrations and moving into the S1 potential
minimum (Figure 7). The evolution follows two
reaction coordinates: first, the bond length alternation along the
retinal chain changes, and then, the motion follows the torsional
coordinate at C11=C12 in animal rhodopsins and C13=C14
in microbial rhodopsins leading to a conical intersection funnel (C.I.
in Figure 7), where the chromophore displays
an approximately 90° twisted double bond, and very efficient
decay to the ground state of the primary photointermediate or to the
parent state occurs.
56,123−125
Most of the earlier studies used energy minimization techniques,
i.e., basically following the molecular forces acting on the chromophore
structure after excitation, moving down the excited states surface
toward the conical intersection as shown in Figure 7. Using Molecular Dynamics (MD)
techniques in combination
with QM/MM methods, the fast femtosecond dynamics of the excited chromophore
can also be directly studied.
126−136
With the help of these computational models an atomistic understanding
of the primary events in different rhodopsins can be achieved.
137
In particular, the role of the hydrogen-out-of-plane
(HOOP) vibrations of retinal, the isomerization pathway, the nature
of the fast photointermediate, and the isomerization efficiency can
be understood in greater detail for different rhodopsins.
2.3
Primary Energy Storage and Relaxation
Functions of
microbial and animal rhodopsins are initiated by light
absorption, followed by the efficient isomerization of retinal (see
above). Since the photoisomerization occurs on a femtosecond time
scale (Figure 8), the protein environment does
not have time to show a large structural response, which means that
the chromophore pocket does not change on this time scale, even though
the neighboring amino acids were shown to respond to the charge motions
inside the retinal.
138,139
Restricting the degrees of freedom
of the retinal promotes the fast isomerization reaction and assures
that part of the energy of light is stored through structural changes
of the chromophore and its hydrogen bonding interactions.
Figure 8
Time scale
related to activation of microbial and animal rhodopsins.
Light absorption, retinal isomerization, proton transfer, and local
and global protein structural changes take place hierarchically, leading
to functional activity.
Most (rhodopsin-like) GPCRs exist in equilibrium between
inactive
and active states, where thermal fluctuations are sufficient for transitions
and ligand binding shifts the equilibrium. Unlike other GPCRs (with
the exception of a handful of protease-activated receptors), animal
rhodopsins are covalently bound to an inverse agonist, 11-cis-retinal, such that the
equilibrium is essentially irreversibly
shifted to the inactive state until the photoisomerization of the
chromophore occurs. Contrary to what could be expected from a simple
rotation around the individual C=C bond, the RSB does not actually
rotate upon photoisomerization. Experimentally this has been shown
by resonance Raman spectroscopy of bovine Rho,
140
which revealed no change of the RSB hydrogen bond upon
retinal isomerization. This view is also consistent with the X-ray
structure of bathorhodopsin at 105 K,
120
where the displacement of the retinal chromophore is very small
at both ends, (i.e., at the RSB and the β-ionone ring). In contrast,
isomerization is clearly indicated by the change of the dihedral angle
around the C11=C12 bond from −40° to −155°.
Thus, a large local distortion of the polyene chain must be produced
to keep the minimally changed overall chromophore geometry, an idea
which is supported by the results of femtosecond stimulated Raman
spectroscopy
103
and quantum chemical calculations.
128,130,133
In bathorhodopsin, about
60% of the photon energy (∼150
kJ/mol) is stored as revealed by low-temperature photocalorimetry,
141
and confirmed by a room-temperature transient
grating method applied to octopus rhodopsin.
142
Energy is stored in the highly twisted retinal chromophore and the
enhanced HOOP vibrational modes at 1000–800 cm–1 allow monitoring the structural deformation
of the polyene chain.
140,143
It should be noted that, in the Rho → bathorhodopsin conversion,
changes in hydrogen-bonding interactions of the retinal are small,
as retinal isomerization does not affect the hydrogen-bonding strength
of the RSB.
140
This trend seems to be general
for animal rhodopsins, since similarly enhanced HOOP modes were observed
for squid rhodopsin
144
and color visual
pigments.
145
Relaxation of the distorted
polyene chain leads to the formation of lumirhodopsin, signifying
energy transfer to the opsin moiety, as was clearly observed by X-ray
crystallography.
146
The bathointermediate
is ∼150 kJ/mol higher in energy than the dark state of rhodopsin,
and activation energy for the isomerization (Figure 7) is estimated to be larger than
180 kJ/mol. Such a high energy
barrier is consistent with the rigid structure of the retinal binding
site, leading to a low noise that is a prerequisite for highly sensitive
dim light vision (see section 4).
The
amount of photon energy stored in early photointermediates
of microbial rhodopsins is smaller than in animal rhodopsins, as low-temperature
photocalorimetric studies reported the energy stored to be ∼67
kJ/mol for the K state of BR,
147
which
is only about 30% of the absorbed photon energy (about half of that
stored in bovine Rho).
141
Despite a loss
of 70% of photon energy upon formation of the K intermediate, the
remaining energy is still sufficient for the pumping activity of BR
(one proton per photon), assuming that pumping a proton normally requires
a free energy of ∼25 kJ/mol.
148
Since
the photoisomerization quantum yields are very similar for bovine
Rho (0.67) and BR (0.64), the differences in energy storage efficiency
must correlate with the structures of the primary intermediates, Batho
and K, respectively. Interestingly, the trend of lower energy storage
in the ion pumps compared to photosensors is supported by the finding
that the K intermediate of sensory rhodopsin II in its proton-pumping
transducer-free form stores less energy than in its transducer-bound
signaling form (88 vs 134 kJ/mol).
149
As expected, X-ray crystallographic studies of the K intermediate
of BR show that the protein structure is not changed significantly
upon isomerization.
150−152
According to vibrational spectroscopy, the
enhanced HOOP vibration bands in the K intermediate mainly come from
the H–D exchangeable groups, implicating that the chromophore
is distorted at the RSB region.
153
The
stretching vibration of the RSBH+ is largely up-shifted
upon photoisomerization (350 cm–1 and ∼500
cm–1 for the N–D and N–H stretches,
respectively), indicating that the hydrogen bond has been weakened.
154
Thus, retinal isomerization induces rotational
motion of the RSB dipole in the case of BR, so that its hydrogen bond
is remarkably weakened.
The hydrogen-bonding acceptor of the
RSBH+ in BR is
a water molecule, and this water bridges the RSBH+ with
the primary proton acceptor Asp85. Rotation of the NH+ group
of the RSBH+ causes rearrangement of the water-containing
hydrogen-bonding network in the RSB active site and weakens hydrogen
bonds of water molecules upon K formation,
155,156
contributing to the high-energy state. The contribution of the weakened
hydrogen bonds to the energy storage was estimated to be 46 kJ/mol,
157,158
which is more than half of the total energy storage (67 kJ/mol).
While the idea of energy storage through the distortion of retinal
has been well-accepted, energy storage through hydrogen-bonding destabilization
was novel. Another computational study has analyzed this aspect in
more detail, reporting that about half of the energy is stored in
the twist of retinal while the other half results from changed interactions
of retinal with the protein environment.
159
The importance of hydrogen-bonding alteration in the energy storage
leads to an unexpected finding for microbial rhodopsins, i.e., positive
correlation between strong hydrogen bond of water near the RSB and
proton pumping activity (examined in depth in section
3).
153,160
Figure 8 summarizes the events occurring
in microbial and animal rhodopsins by placing them on a single time
scale. Light absorption and photoisomerization take place in 10–15 and 10–14–10–12
s, respectively, and converted light energy in each protein, as
steric constraints in animal rhodopsins, or as hydrogen-bonding alterations
and steric constraints in microbial rhodopsins, drive protein structural
changes in later time domains. The primary stable photoproducts are
red-shifted from the original absorption, being called “K intermediate”
and “Batho intermediate” for microbial and animal rhodopsins,
respectively. Relaxation of these high energy states of the chromophore
and its immediate vicinity trigger global protein structural changes
necessary for function. For many microbial and animal rhodopsins,
such changes accompany deprotonation of the RSBH+, forming
the “M intermediate” and “Meta-II intermediate”,
respectively. These intermediates are the key states for function,
which are described in detail in sections 3 and 4. Figure 8 also
contains a time domain of evolution, the time of natural design of
protein architecture, by which both microbial and animal rhodopsins
have been functionally optimized.
Microbial and animal rhodopsins
have further striking differences
in the fate of their photoproducts (see the relevant sections below).
In microbial rhodopsin, all-trans → 13-cis photoisomerization usually triggers a cyclic
sequence
of reactions linking a series of intermediates and including thermal
reisomerization of retinal, which is not released from opsin during
this photocycle. This is important for organisms in which the rhodopsins
are not embedded within a specialized organelle and enzymatic reisomerization
is not possible. In animal rhodopsins, retinal photoisomerization
from the 11-cis to the all-trans configuration triggers a reaction pathway that also
comprises a
series of intermediates, whereas the reaction is not necessarily cyclic
and often requires the retinal molecule to be released from the opsin
apoprotein and reisomerized either enzymatically
161,162
or photochemically.
163
It appears that,
depending on the biological function of a rhodopsin, the specific
design of a protein/retinal combination determines whether the light-induced
conformational changes in the RSB region are large enough for RSB
hydrolysis to occur.
3
Microbial Rhodopsins
3.1
General Features of Microbial Rhodopsins
For about
25 years since the early 1970s, microbial rhodopsins
were epitomized by haloarchaeal proteins, the first-discovered and
best-studied light-driven proton pump bacteriorhodopsin (BR) and its
close relatives halorhodopsin and sensory rhodopsin I and II (HR,
SRI and SRII). But in the last 15 years, thousands of related photoactive
proteins with similar or different functions have been identified
in Archaea, in marine, freshwater, and terrestrial Eubacteria and Eukaryota (Figure
9).
1,7,164−166
Thus, a small group of proteins, believed
to be present only in halobacteria and having a limited number of
functions (proton and chloride pumps and phototactic/photophobic receptors),
grew into a confounding variety of species with versatile sequences,
ecology, taxonomy, and functions. Among the new functions are light-gated
cation channels (see the section on channelrhodopsins below), light-switchable enzymes,
light-driven sodium pumps, and
novel photosensors distinct from the haloarchaeal SRs (Figures 3 and 9).
167−170
Figure 9
Phylogenetic
tree of selected microbial rhodopsins. Four main functions
of microbial rhodopsins are shown in different colors: blue, proton
pumps; green, chloride ion pumps; red, light-gated cation channels;
yellow, photosensors. Proton pumps are widely distributed among Archaea, Eubacteria,
and Eukaryota. Two additional poorly studied functional groups (sodium ion pumps
and enzymerhodopsins) are not included. See Supporting
Information Figure 1 for additional information on genus and
species of the microbial rhodopsins.
Despite the variety of sequences and functions, the common
structural
and mechanistic principles of microbial rhodopsin architecture can
be deduced on the basis of about a dozen of structures of unique proteins.
X-ray and NMR structures are available for various proton and chloride
pumps, as well as for photosensors and a light-gated channel. While
all of the known structures of microbial rhodopsins show a common
tight α-helical bundle of 7TM helices surrounding the retinal
chromophore, there is substantial variation in the arrangement of
side chains, the structure of the interfacial regions and loops, as
well as in the positions of internal water molecules.
7,165,171,172
The retinal binding pocket is the most conserved common element
of the structure. Figure 10 illustrates the
overall structure of BR with the conserved aromatic amino acids important
to function shown. Strongly conserved Trp86 and Trp182 constitute
an important part of the chromophore binding site by sandwiching all-trans-retinal
(Figure 10). The presence
of these bulky groups possibly determines the isomerization pathway
from the all-trans to 13-cis form,
and the interaction of photoisomerized retinal with Trp182 may serve
as the mechanical transducer for passing the energy stored in retinal
deformation into the functionally important changes of the helical
tilts necessary for function.
173
Another
important position occupied by aromatic amino acids in the retinal
binding pocket is that of Tyr185 in BR (Figure 10), which participates in hydrogen-bonding
stabilization of the RSBH+ counterion for many rhodopsins, but is replaced by a Phe
in channelrhodopsins (Supporting Information Figure 1), suggesting that the lack of
hydrogen-bonding interaction
at this position is important for channel function. Interestingly,
an aromatic pair similar to Trp182/Tyr185 is found in retinal binding
pockets of many animal visual rhodopsins (Trp265/Tyr268 in bovine
Rho), suggesting a conceptually similar mechanism of energy transfer
between retinal and opsin.
174
Figure 10
(A) Structure
of bacteriorhodopsin (BR), with conserved aromatic
residues highlighted (PDB ID: 1QM8). Tyr83, Trp86, and Trp182 are strongly
conserved among microbial rhodopsins (orange). Aromatic amino acids
are strongly conserved at the position of Tyr185, Trp189, and Phe219
(yellow). In BR, Trp86, Trp182, Tyr185, and Trp189 constitute the
chromophore binding pocket for all-trans-retinal
(gray). (B) Crystallographically observed internal water molecules
of BR (shown as green spheres). Note much higher hydration of the
extracellular half compared to the cytoplasmic one.
In addition to the retinal polyene chain–aromatic
side chain
rings interactions described above, electrostatic and hydrogen-bonding
interactions in the proximal part of retinal are crucial in defining
the functionality of microbial rhodopsins. The side chain of Lys216
of BR (or its equivalent in other microbial rhodopsins) forms a covalent
bond with the retinal molecule through the Schiff base (Figures 2 and 11). Since the
RSB is
usually protonated, Lys216 and superconserved Arg of helix C (Arg82
in BR) provide two positive charges within the protein, which require
two negative charges for electrostatic stabilization. This dictates
the most common configuration of the RSBH+ counterion,
that with two carboxylic acids (Asp85 and Asp212 in BR), which are
perfectly conserved for proton-pumping microbial rhodopsins (Figure 11). Any deviation
from this arrangement has strong
functional consequences (Supporting Information Figure 1). For example, in the chloride-pumping
HR, the negatively
charged Asp at position 85 (of BR) is replaced with Thr, and in the
recently discovered sodium-pumping rhodopsins, it is replaced with
an Asn.
168,175
Similarly, in Anabaena sensory
rhodopsin (ASR), the negatively charged Asp at position 212 of BR
is replaced with Pro, which prevents light-induced transfer of the
RSBH+ proton to the homologue of Asp85 of BR.
176,177
Importantly, all microbial rhodopsins contain protein-bound water
molecules near the RSB, presumably contributing to the stabilization
of the RSBH+ in the hydrophobic protein interior (Figure 11). These water molecules
play crucial roles in
protein function, and have been extensively studied by X-ray crystallography
of photointermediates, Fourier transform infrared spectroscopy (FTIR),
and computational methods (see below).
Figure 11
X-ray crystallographic
structures of the RSB region for eight representative
microbial rhodopsins with various transport/signaling functions. BR,
AR2 (archaerhodopsin 2), and XR function as proton pumps. NpSRII is a phototaxis sensor,
but pumps protons in the
absence of its transmembrane transducer. HsHR and NpHR are chloride pumps, ASR is
a photochromic sensor, and
ChR is a light-gated cation channel. Membrane normal is approximately
in the vertical direction of the figure. Upper and lower regions correspond
to the cytoplasmic and extracellular sides, respectively. Green spheres
denote ordered water molecules observed crystallographically. For
abbreviations of microbial rhodopsins, see Supporting
Information Figure 1.
As described in section 2, photoisomerization
(which usually proceeds from the all-trans- to 13-cis-retinal in microbial rhodopsins)
disrupts the finely
tuned retinal–opsin interactions, inducing functionally important
changes in the protein moiety. Most importantly, this results in changes
in ion affinities and/or conformational changes of the backbone (e.g.,
helical tilts), that are required for ion transport and photosensory
transduction (see the respective subsections below for a detailed
description of these events). It should be noted that, in some microbial
rhodopsins, most notably in BR, all-trans-retinal
is not stable in the dark (Figure 2), and thermal
compact isomerization from the all-trans- to 13-cis,15-syn-form takes place (so-called
dark-adaptation).
178−181
The thermally stable 13-cis,15-syn-form is different from the metastable 13-cis,15-anti-form
(C=N trans) that appears
during the photocycle of microbial rhodopsins. The converse light-induced
conversion from the 13-cis,15-syn- to all-trans-form is called light-adaptation and
restores the functionally active state for most microbial rhodopsins.
182
There are some notable exceptions among sensory
rhodopsins, as the direction of light- and dark-adaptation are reversed
in ASR, which also presents an exception to the normally cyclic character
of photoreactions of microbial rhodopsins.
183,184
Most often, photoisomerization from the all-trans- to 13-cis-form of microbial rhodopsins
triggers
a cyclic series of reactions (the photocycle) comprising a series
of photointermediates, which have been studied by various methods
in a time-resolved manner. Figure 12 shows
the typical photocycle for a microbial rhodopsin, exemplified by BR.
185−189
Spectroscopic properties of photointermediates reflect isomeric
configuration, planarity, and protonation state of the retinal, as
well as the position of surrounding protein and ion charges and water
molecules. The primary red-shifted K intermediate with twisted retinal,
introduced in detail in section 2, is usually
followed by the blue-shifted L intermediate.
190
For proton-pumping (and some of the photosensory) rhodopsins, the
L intermediate serves as the precursor of the proton transfer reaction
from the RSBH+ to its primary carboxylic proton acceptor,
by which the M intermediate is formed, the key step in proton transport.
Since the M intermediate has a deprotonated 13-cis chromophore,
191
it exhibits a characteristic
strongly blue-shifted absorption (λmax at 360–410
nm), well-isolated from that of other intermediates. In the case of
chloride pumps, the RSBH+ does not deprotonate during the
photocycle and the L intermediate converts directly to the N intermediate,
192
which in proton-pumping rhodopsins arises as
a result of reprotonation of the RSB from the cytoplasmic side (detected
as the M intermediate decay). The N intermediate is often characterized
by the largest changes in the protein backbone conformation, most
notably, outward tilts of the cytoplasmic end of helix F, which are
functionally significant both for ion transport and interactions with
transducers of sensory rhodopsins.
193
The
photocycle usually ends with another red-shifted intermediate, known
as O, serving as a last step in resetting the original unphotolyzed
conformation. A more detailed description of the individual photocycles
for different classes of microbial rhodopsins can be found in the
subsequent sections.
Figure 12
Typical photocycle of microbial rhodopsins showing isomeric
and
protonation state of the retinal. Names of the photocycle intermediates
and their characteristics were originally established for BR. In the
case of BR, K and O are the red-shifted intermediates, while L, M,
and N are all blue-shifted intermediates. The primary photoreaction
is the retinal isomerization from the all-trans,15-anti- to 13-cis,15-anti-isomer.
The RSBH+ deprotonates upon M formation and is
reprotonated upon M decay. Thermal reisomerization occurs upon O formation
from the 13-cis,15-anti- to reform
the all-trans,15-anti-state.
3.2
Bacteriorhodopsin,
the Prototypical Ion Pump
Bacteriorhodopsin from Halobacterium salinarum, the first discovered microbial
rhodopsin,
194
was the first membrane protein
whose structure was found to be composed
of seven helices by electron microscopy,
195
and was also the first membrane protein to have its amino acid sequence
determined.
196
As the best studied microbial
rhodopsin, it serves as a paradigm of a light-driven retinal-binding
ion pump and aids in studies of novel rhodopsins. The photocycle of
BR, during which a proton is vectorially translocated across the membrane
from the cytoplasmic to the extracellular side, is shown in Figure 12, and the main
proton transfer steps with the respective
transitions between photocycle intermediates are shown in Figure 13. These intermediate
states were first identified
by visible absorption spectroscopy, while the structure of the retinal
chromophore in each photointermediate was revealed by resonance Raman
and FTIR spectroscopy.
189,197−199
In the early days of BR research, it became clear that all-trans → 13-cis isomerization
occurs
upon photoexcitation of BR, and that the RSB is deprotonated in the
M intermediate and reprotonates upon formation of the N intermediate.
191,200−202
In combination with site-directed mutagenesis
and photoelectric measurements, this indicated that the RSBH+ proton is transferred
to the extracellular side upon M formation,
and is taken up from the cytoplasmic side by the RSB upon M decay
(Figure 13).
203−205
Since then, the intricate
details of the proton transport mechanism and structure of proton-conducting
pathways of BR have been elucidated by the concerted efforts of many
research groups, with seminal contributions made by site-directed
mutagenesis, vibrational spectroscopy, X-ray crystallography, solid-state
NMR, and molecular dynamics simulations.
Figure 13
Main proton transfers
in the bacteriorhodopsin photocycle. Protonatable
groups and bound water molecules important for transport activity
are shown as stick representation and blue spheres, respectively (PDB
ID: 1C3W). Numbers
with arrows represent the sequence of proton transfer reactions, the
corresponding transitions between the photointermediates are indicated
in the inset. The TM helices are shown in the following colors: A,
blue; B, teal; C, green; D, lime green; E, yellow; F, orange; G, red;
and the chromophore is depicted as black sticks. ① Proton transfer
from the RSBH+ to the primary proton acceptor Asp85; ②
proton release to the extracellular medium from the proton-releasing
complex; ③ reprotonation of the RSB from the primary proton
donor Asp96; ④ reprotonation of Asp96 from the cytoplasmic
medium; ⑤ proton transfer from Asp85 to the proton-releasing
complex.
The molecular mechanism of vectorial
transport by pumps has been
a long-standing challenge in bioenergetics, and proton transport by
BR is not an exception to this rule. An ion pump must be fundamentally
different from a channel, where the ion pathway is continuous between
the cytoplasmic and extracellular aqueous phases, because a
pump should be able to prevent ion backflow for transport against
an electrochemical gradient. Therefore, the alternating access model
has been developed for BR, wherein the RSB accessibility is switched
between the cytoplasmic and extracellular sides (Figures 13 and 14). The nature of
this so-called reprotonation switch has been disputed for many years,
with models ranging from the pure structural accessibility (conformational
changes of the retinal and/or opsin) to pure affinity (changes in
the pK
a’s of proton donors and
acceptors) mechanisms, but is most likely a combination of both.
187,189,206−210
As can be seen in Figure 12, the isomeric
states of the retinal chromophore are identical between the L and
N intermediates, whereas the RSB accessibility must be switched between
them, from the extracellular to the cytoplasmic side, to ensure vectorial
transport. Thus, in addition to the two unidirectional steps in the
photocycle (the primary isomerization reaction and the last step,
the O-to-BR reaction), a third unidirectional step is often considered
between the two M substates, M1 and M2, to account
for the vectorial transport (Figure 14). The
presence of multiple M states has been confirmed by several techniques
and is consistent with this idea.
211−217
While M1 exists in equilibrium with the L intermediate whose accessibility
is to the extracellular side, M2 is in equilibrium with the N intermediate
whose accessibility is to the cytoplasmic side. Thus, the switch should
occur between the M1 and M2 states, preceded by the proton release
to the extracellular surface (Figure 13). It
appears that the conformational changes associated with the switch
must be very subtle, as there is no significant difference in the
vibrational signals of retinal and protein between M1 and M2, with
the only detectable difference being in the water FTIR signals (so-called
IR continuum).
212,218
Figure 14
Sequence of the main
molecular events in the bacteriorhodopsin
photocycle and accessibility of RSB: ① absorption of a photon
by all-trans-retinal, photoisomerization to twisted
13-cis form; ② relaxation of retinal twist,
strengthening of water-mediated hydrogen-bonding between RSBH+ and Asp85; ③ proton
transfer from RSBH+ to the primary proton acceptor Asp85; ④ proton release to
the extracellular medium from the proton-releasing complex and switch
of the accessibility of RSB to the cytoplasmic side; ⑤ conformational
change of the backbone in the cytoplasmic half and reprotonation of
RSB from the primary proton donor Asp96; ⑥ reprotonation of
Asp96 from the cytoplasmic medium and thermal reisomerization of retinal
to all-trans; ⑦ proton transfer from Asp85
to the proton-releasing complex and restoration of the initial conformation.
Extensive studies by FTIR spectroscopy
in combination with site-directed
mutagenesis and other techniques revealed the capture and donation
of protons by specific amino acids during the pumping process. For
the analysis of proton pathways, the C=O stretching vibrations
of protonated carboxylic acids in the 1800–1700 cm–1 spectral region provided much
needed information on protonation
and hydrogen-bonding changes of proton donors and acceptors during
the photocycle.
198,199,219−222
Transient deprotonation of Asp96 in the N intermediate and transient
protonation of Asp85 in the M, N, and O states have been observed,
indicating that a proton is translocated via Asp96 (the donor), the
RSB, and Asp85 (the acceptor) (Figure 13).
The temporal coincidence between the deprotonation of the RSB and
protonation of Asp85 directly demonstrated that the primary proton
transfer occurs from the RSBH+ to Asp85 (process 1 in Figure 13).
200,223
Similarly, the kinetic coincidence
between the reprotonation of the RSB and deprotonation of Asp96 signifies
proton transfer from Asp96 to the RSB (process 3 in Figure 13), and reprotonation
of Asp96 upon decay of the
N state coincides with proton uptake from the cytoplasmic bulk phase
(process 4 in Figure 13).
205,220,224
The persistence of Asp85 in
the protonated state even after proton release upon M formation suggests
that there is another proton release group (or complex) located close
to the extracellular surface of BR. While site-directed mutagenesis
studies pointed to the involvement of several protein side chains
in the proton-releasing mechanism (most notably, Glu204, Glu194, Arg82,
and Tyr57), the exact identity of the proton-releasing group remained
controversial for a long time, due to the lack of a clear deprotonation
signal.
225−229
The observation of the infrared continuum band (presumably of a
protonated water cluster) at 2000–1800 cm–1 by time-resolved FTIR provided the data
needed to confirm the presence
of a highly delocalized proton,
230
and
the involvement of water vibrations in this continuum band was proven
directly using
18
O water.
231
As the negative continuum band disappears in mutants deficient
in proton release, it can serve as an experimental signature for this
proton transfer.
232
A theoretical study
proposed that a low-barrier hydrogen bond between Glu194 and Glu204
explains the presence of this continuum band, reflecting the existence
of a delocalized excess proton in the water cluster stabilized by
these residues, which may serve as the proton-releasing group.
233
It should be noted that the essential
character of the early proton
release has not been proven, as BR retains its proton-pumping activity
even when proton release occurs after uptake, as happens in a number
of BR mutants, at low pH and in some close homologues of BR.
164,234
Nevertheless, it has been postulated that early release is important
for the functionality of the reprotonation switch, as it ensures irreversibility
of the M1 → M2 reaction by changing the proton affinity of
Asp85.
235,236
FTIR spectroscopy in combination with mutagenesis
showed that strongly hydrogen-bonded (as judged from the low O–D
stretching frequency) water molecule in the RSB vicinity is a prerequisite
for efficient proton pumping.
153,155
In the “hydration
switch model”, it was proposed that the light-induced shift
of this water molecule from Asp85 to Asp212 ensures efficient deprotonation
of the RSBH+ (Figure 15), but a
further increase in the proton affinity of Asp85 is required to make
it unidirectional, and this is exactly where the early proton release
may become important.
237
This might be
achieved through the electrostatic coupling of Asp85 with the proton
releasing complex, demonstrated by both mutagenesis and crystallography.
When Asp85 becomes protonated, the positive charge of Arg82 is reoriented
toward the proton-releasing water cluster and the destabilized proton
is released to the extracellular bulk solvent.
238,239
Conversely, this stabilizes the proton on Asp85 and prevents the
backflow of protons to the RSB, ensuring its eventual reprotonation
from the cytoplasmic side.
Figure 15
(A) Photocycle of Natronomonas
pharaonis halorhodopsin (NpHR). (B)
Light-induced hydrogen-bonding
alteration in the RSB region of NpHR (left) and bacteriorhodopsin
(BR, right), suggesting the mechanism of proton and chloride ion translocation
(refs (306, 566), respectively).
The BR crystal structure exhibits
an asymmetric pattern of hydration,
where the extracellular half contains seven internal water molecules,
but the cytoplasmic half contains only two (Figures 10 and 13).
165,240
Such asymmetry makes sense in view of BR function, as the water
molecules build a hydrogen-bonding network on the extracellular side
for fast proton release, while the cytoplasmic side should be inaccessible
in the dark and allow proton uptake only after the light-induced accessibility
switch. To make proton conduction in the cytoplasmic region possible,
an additional conformational change allowing the entrance of water
into the vicinity of Asp96 should take place. Such a conformational
alteration is realized mainly by changes in helical tilts (especially
that of the cytoplasmic half of helix F) in the late M and early N
intermediates (Figure 14), as was observed
by neutron diffraction, X-ray diffraction, electron diffraction, and
EPR spectroscopy, and most recently by high-speed atomic force microscopy.
216,241−246
Interestingly, the cytoplasmically open conformation of BR can be
obtained via mutagenesis,
247,248
for example, by neutralizing
the charge of Asp85 and/or by promoting deprotonation of Asp96, emulating
the N intermediate.
249
It should be noted
that a similar outward opening motion of the TM6 helix takes place
for bovine Rho (see section 4 below), as well
as for some of the sensory rhodopsins,
193
once again suggesting a common mechanism of energy transduction.
X-ray crystallography has been very successful in revealing the
sequence of structural alterations in each of the photocycle intermediates
through finding an appropriate illumination regime of BR crystals
under suitable temperature and pH conditions or by using mutants which
stabilize or emulate particular intermediate states.
185,250,251
The reader is referred to excellent
and extensive crystallographic reviews, which give an atomistic level
of detail of the conformational changes and mechanistic insights (many
of which have been mentioned above).
165,188,252−254
In conclusion, it should
be mentioned that an alternative model
of BR function has been proposed, postulating that BR works as an
inward-directed hydroxide (OH–) pump rather than
the commonly accepted outward-directed proton (H+) pump.
The model was based on the presence of the strongly polarized water
hydrogen-bonded to the RSB in the dark state of BR and its absence
in M intermediate structures, and supported by the fact that in homologous
halorhodopsins and some BR mutants other anions can also be transported.
187,255−257
While this is an interesting model, the
Grotthus mechanism, a concerted proton transfer through hydrogen bonds,
is sufficient to explain the proton transfer reactions in BR, but
it should be however noted that there has been no strong experimental
evidence to rule out this alternative hydroxide pump model either.
3.3
Other Light-Driven Proton Pumps: Variations
on the Common Mechanism
During the past decade, it became
clear that BR-like light-driven proton pumps exist not only in Archaea, but also are
abundant among Eubacteria and lower Eukaryota (Figure 9).
7,164,166
Even though
the molecular mechanism of proton pumping has been thoroughly investigated
for BR, these additional rhodopsins present many interesting variations
assisting in our understanding of the common principles of proton
transport. While it would be impossible to survey all the details
of the numerous (the count goes by thousands) species of proton-pumping
rhodopsins in the framework of this review, we would like to outline
the major classes of such rhodopsins and stress the interesting mechanistic
and structural differences and commonalities with BR. Even though
the number of taxonomic species bearing genes encoding proton-pumping
rhodopsins is large, these rhodopsins can be conveniently classified
into a small number of subtypes with substantial sequence similarity
within each class. The main subtypes are bacteriorhodopsins (mostly
haloarchaeal), proteorhodopsins (mainly eubacterial but also found
in some Archaea), xanthorhodopsins (and related actinorhodopsins),
and fungal/algal proton pumps.
Green-absorbing proteorhodopsin
(PR) was the first eubacterial proton pump identified by metagenomic
analysis of marine proteobacteria.
258
Since
this groundbreaking discovery in 2000, it became clear that proteorhodopsins
are likely to be the most abundant and widely distributed of the microbial
rhodopsins, now found in many different bacterial taxa and not only
in marine organisms.
259−263
Spectrally, PRs can be classified into green-absorbing PRs (GPR,
λmax ∼ 525 nm) and blue-absorbing PRs (BPR,
λmax ∼ 490 nm), thought to be characteristic
for marine bacteria living near the surface and in the deep sea, respectively.
264,265
It was found that a single amino acid at position 105 (equivalent
to Leu93 in BR) is the determinant of the specific color (Leu in GPR
and Gln in BPR). Although the proton-pumping function was clearly
demonstrated for GPR heterologously expressed in E. coli,
258,266,267
its native
physiological significance was not clear until recently. Initially,
it was reported that several PR-harboring species exhibited no differences
between the growth rates of cultures grown in the dark or under illumination,
but it eventually became clear that PR expression can give significant
advantages to some bacteria under the conditions of carbon starvation.
268−271
Additionally, many PR variants (especially BPRs) possess very slow
photocycles and weak photoelectric signals, making it unlikely that
they function as proton pumps. Alternative physiological functions
(e.g. photosensory) have been suggested.
266,272−274
Another group of related (but distinct from
PR) eubacterial pumps is represented by xanthorhodopsin (XR) from Salinibacter ruber
and its homologues, including the cyanobacterial
rhodopsin from Gloeobacter (GR) and numerous actinorhodopsins.
172,275,276
Actinorhodopsins were originally
identified as numerous metagenomic PR-like sequences in predominantly
freshwater habitats, but further analysis revealed that they represent
a separate subgroup along with XR.
261,276,277
The most novel and distinctive feature of XR is its
second chromophore, a carotenoid salinixanthin, which serves as a
light-harvesting antenna that transfers the absorbed light energy
to retinal with a quantum efficiency of ∼40%.
278,279
GR can bind carotenoids as well, and it has been suggested that
the structural prerequisite of such binding is the replacement of
Trp near the β-ionone ring of retinal (Trp138 of BR) with Gly.
280
The crystal structure of XR provides
many structural insights into
the mechanism as well as unique features of eubacterial proton pumps.
172,279
On the extracellular side of the RSB, one observes at least four
important differences from BR. First, there is only a single water
molecule observed in the RSB region (Figure 11), considerably different from the pentagonal
water cluster (comprising
three water molecules and two aspartates) and adjacent extended hydrogen-bonding
network (seven water molecules in total) of the archaeal rhodopsins
(Figures 10 and 11).
155
Second, the carboxylic counterion exists as
a His-Asp complex in XR and most other eubacterial pumps. As a result,
the carboxylic counterion is no longer electrostatically coupled to
the conserved equivalent of Arg82, in contrast to BR.
279,281−284
The higher pK
a of the RSB counterion
is characteristic for GPR and other eubacterial rhodopsins, and interactions
with a conserved His may modulate this.
281,285−287
Third, the proton-releasing complex of BR
(the glutamate dyad with intercalating water) is not present in eubacterial
proton pumps, and early proton release is not observed, at least under
physiological conditions.
275,285,288
This may imply that early proton release is not essential for a
proton-pumping mechanism in general.
164
Finally, the β-stranded BC loop is displaced, opening a hydrophilic
cavity on the extracellular side, which may account for the alternative
proton release mechanism.
279
On the cytoplasmic
side, the proton donating Asp96 of BR, which is usually paired with
a conserved Thr residue (Thr46 of BR), is replaced by a Glu paired
with Ser (Supporting Information Figure
1). It appears that this Glu residue is more tightly coupled with
the RSBH+ in the dark state of XR than in BR and that the
proton transfer pathway between them is partially prebuilt.
275,279,289
One important exception to this
rule is a unique rhodopsin from Exiguobacterium (ESR),
in which the carboxylic proton donor is replaced by a Lys residue,
suggesting a completely different mechanism of RSB reprotonation.
290
While a few proton-pumping microbial
rhodopsins in lower eukaryotes
(mainly dinoflagellates) are homologous to XR or PR,
166,291
many rhodopsin-bearing eukaryotes, such as fungi and green algae,
have BR-like pumps. Fungi demonstrate many different forms of rhodopsins,
some of which have been shown to be proton pumps, starting with rhodopsin
from Leptosphaeria (LR).
292−294
Fungal proton-pumping rhodopsins are similar to BR in many respects,
but their exact physiological role is unclear.
295
The marine green algae Acetabularia acetabulum contains a proton-pumping microbial
rhodopsin (Ace2), and recently
its crystal structure was determined to 3.2 Å resolution, providing
the first glimpse of a eukaryotic retinal-binding proton pump.
296,297
Similar to fungal proton pumps, proton uptake in Ace2 occurs earlier
than release, which was expected as it lacks the proton-releasing
complex of BR (Supporting Information Figure
1). While its overall structure is similar to that of BR, the most
distinct feature of this rhodopsin is the unique interaction of the
carboxylic cytoplasmic proton donor with a Cys residue on helix G,
replacing typical pairing with Thr or Ser on helix B found in other
microbial rhodopsins.
Despite the unique features and differences
described above, all
of the eubacterial and eukaryotic proton pumps conserve the carboxylic
primary proton acceptor (homologous to Asp85 of BR). Importantly,
they also seem to preserve the strongly hydrogen-bonded water molecule
between the RSBH+ and this counterion, further confirming
the importance of hydrogen-bonding strength of protein-bound waters
for proton pumps.
153
In fact, strongly
hydrogen-bonded water molecules were found in proton-pumping fungal
rhodopsins from Leptosphaeria and Phaeosphaeria, but not in their nonpumping homologues
from Neurospora (NR),
298
and also found for GPR at high
pH (proton-pumping active), but not at low pH.
299
3.4
Changing the Ion Specificity:
Light-Driven
Chloride and Sodium Pumps
It appears that, by changing the
finely tuned structural template of BR-like proton pumps a little
further, one can produce ion pumps with different ion specificities,
namely, chloride and sodium pumps.
168,300
Sometimes,
these function-altering changes of the template can be achieved not
only by evolution but simply by applying a few mutations to key residues in vitro.
301,302
A good example of such functional
conversion is HR, initially identified in Halobacterium salinarum and shown to transport
chloride ions in the cytoplasmic direction.
While the overall architecture of HRs is BR-like, the crystal structures
of two chloride pumps (HsHR and NpHR) clearly show the presence of a chloride ion
in the RSB region
(Figure 11), where it occupies the position
of Asp85 carboxylate in BR.
303,304
In HR, this unprotonated
carboxylate is replaced by Thr (Supporting Information Figure 1), suggesting that
the electric quadrupole of the RSBH+ with its counterion complex (Asp85, Asp212, and
Arg82 in
BR) lacks a negative charge and the charge balance is compensated
for by the binding of the negatively charged chloride ion. Hydrogen
bonds of the RSB and of protein-bound water molecules in HR are weak,
suggesting that the transported chloride ion is not involved in strong
hydrogen-bonding (Figure 15), in stark contrast
to the case for proton transport in BR.
305,306
The photocycle of HR has no M intermediate reflecting the
absence of RSBH+ deprotonation (Figure 15), which might, however, occur as a side
reaction.
307,308
Instead, the existence of two L intermediates (L1 and
L2, sometimes called L and N) was observed by various methods,
suggesting an extracellular to intracellular accessibility change
during their interconversion, analogous to the M1 and M2 intermediates of BR.
192,309−311
The photoisomerization induces changes in the electric and hydrogen-bonding
environment of the Cl– ion, driving its movement
to the cytoplasmic side of the RSBH+. The hydrogen bond
of the RSB is strengthened in the L intermediate, but the hydrogen-bonding
acceptor is not the chloride ion, but most likely a water molecule
(Figure 15).
306,312
Rearrangement
of the water-containing hydrogen-bonding network most likely opens
the valve to the cytoplasmic region, and the chloride ion is released
to the cytoplasmic side during the transition to the O intermediate.
313,314
Surprisingly, BR can be converted into an HR-like chloride
pump
(albeit, a very inefficient one) in vitro by replacing
Asp85 with Thr, suggesting that the amino acid at position 85 is a
determinant for ion specificity.
301,302
In contrast,
the reverse Thr-to-Asp mutations of HR, such as Thr108Asp of HsHR and Thr126Asp of
NpHR, fail to convert
HR into a BR-like outward proton pump.
315,316
This may
imply that the molecular determinants of a proton pump are more demanding
than those of a chloride pump, and indeed, an NpHR
mutated to contain 10 key BR-like amino acids still fails to pump
protons (Supporting Information Figure
3), consistent with its lack of a strongly hydrogen-bonded water.
160
Nevertheless, the conversion of HR into a proton
pump can be achieved by the simple addition of sodium azide, which
most likely serves as an artificial proton shuttle, suggesting common
elements in the transport mechanism of proton and chloride pumps.
316
The restoration of a strongly hydrogen-bonded
water for the azide-bound HR is completely consistent with these results.
317
The recent discovery of a new group of
rhodopsins homologous to
XR, but with different ion specificities, gives very strong confirmation
for the ideas gleaned from the comparisons of HR to BR.
168,318
In this new group of eubacterial rhodopsins the main carboxylic
proton acceptor and donor (Asp and Glu homologous to Asp 85 and Asp
96 of BR, respectively) are replaced by Asn and Gln (Supporting Information Figure
1), giving them a name of NQ
rhodopsins. A subgroup of these rhodopsins introduces a new proton
acceptor for the RSBH+, at the position homologous to Thr89
of BR (Figure 11), creating the so-called NDQ
motif. It appears that the NDQ rhodopsins can selectively transport
sodium in the extracellular direction and that RSBH+ deprotonation
is a required step for such transport. This once again demonstrates
that new functions can be generated on the BR template through a limited
number of amino acid substitutions, and that the energy stored during
photoisomerization can be channeled into various processes.
3.5
Light-Signal Transduction by Microbial Rhodopsins
Originally,
it was believed that the photosensory functional class
of microbial rhodopsins was typified by SRI and SRII, signaling via
the classical two-component transduction pathway.
1,319
In the past decade, it became clear that there are several more
types of sensory rhodopsins present in Eubacteria and lower Eukaryota, which employ
dramatically
different signaling mechanisms.
11,167,171,274
While in halobacterial SRs,
primary signal transduction occurs via interaction with a membrane-embedded
transducer, other novel types of signaling employ soluble transducers
or fused signaling domains, or are performed via changes in membrane
potential by light-gated channels (see section
3.6 on channelrhodopsins).
SRI and SRII were the third
and fourth species of microbial rhodopsins found in Halobacterium
salinarum, and their homologues have been discovered in Eubacteria, such as Salinibacter.
320−322
They function as phototactic (and photophobic) receptors controlling
the cell’s swimming behavior in response to changes in light
intensity and color. SRI and SRII form 2:2 complexes with their cognate
transducers, HtrI and HtrII (Figure 16).
274,319,323
These transducers share homology
with bacterial chemotaxis receptors, and comprise two transmembrane
helices and a large cytoplasmic domain that binds to the histidine
kinase CheA. CheA kinase activity is modulated by these SR–Htr
complexes, culminating in control of the flagellar-motor switch. SRI
and SRII exhibit no proton pumping activity in the SR–Htr complex,
whereas they pump protons in the absence of transducer, suggesting
their close structural relationship to BR.
324−326
Indeed, both transducer-free SRI and SRII show photocycles similar
to that of BR, but with long-lived M intermediates. The conserved
cytoplasmic carboxylic proton donor of proton pumps (Asp96 in BR, Supporting Information
Figure 1) is replaced
by aromatic amino acids in SRI and SRII. Consequently, the photocycles
of SRs are long-lived, as would be expected for a sensory photoreceptor
to allow time for the signal to move from receptor to transducer.
Figure 16
X-ray
crystallographic structure of the transmembrane part of the NpSRII–NpHtrII complex
(PDB ID: 1H2S). NpSRII helices are shown in purple, and NpHtrII helices
are shown in green. (A) Side view of the complex. (B) Complex viewed
from the cytoplasmic side. Inset: Illustration of the light-induced
conformational changes of helix F of SRII and TM2 of HtrII. Because
of the tight interaction between helix F and TM2, the outward movement
of helix F in the receptor (arrow) causes a clockwise rotary motion
of TM2 in the transducer.
Dependent on the wavelength of stimulating light, SRI is
a dual
function attractant–repellent phototaxis receptor.
10,327
SRI uses an intrinsic mechanism to discriminate between colors,
thus allowing the cells to be attracted by orange light or be repelled
by UV light. SRI absorbs maximally at 587 nm, and absorbance of a
photon at 587 nm results in the formation of the M intermediate, triggering
the attractant signal. However, if the M intermediate itself is photoexcited
(by near UV light), a repellent-signaling intermediate is generated.
Therefore, excitation with orange light attracts cells, whereas subsequent
excitation with a second photon of near-UV light repels the cells.
The molecular mechanism of this dual attractant–repellent function
is intriguing, and it was suggested that the direction of the behavioral
response directly correlates with the state of the RSB accessibility
switch (see above).
328
SRII (also called
phoborhodopsin) absorbs in the midvisible range (λmax = 490–500 nm) and only performs
repellent signaling.
Structural studies are much more advanced for SRII than for SRI
owing to the determination of crystal structures of SRII from Natronomonas pharaonis
(NpSRII) for both
the receptor alone and the receptor–transducer complex (Figure 16).
52,53,329
The structure of NpSRII is largely the same regardless
of the presence of its transducer, NpHtrII, consistent
with the lack of significant changes in color, the RSBH+ pK
a, or photocycle upon transducer binding.
326,330
The M intermediate structure of the receptor–transducer complex
revealed several light-induced structural alterations of both proteins,
most notably, in TM2 of the transducer, which interacts with the receptor
closely.
331
It should be noted that EPR
studies suggest much larger conformational changes in the activated
receptor than those observed in the crystal structure.
193,332,333
In particular, significant outward
tilting of the cytoplasmic half of helix F was detected (Figure 16), similar to other
microbial and animal rhodopsins.
The mechanism of signal transduction was also extensively studied
by FTIR spectroscopy, which revealed that steric interaction between
the retinal chromophore and conserved Thr204 (Ala215 in BR; Supporting Information
Figure 1) is a prerequisite
for light-signal transduction in SRII.
334,335
The common
pattern of structural changes upon the light activation of both haloarchaeal
photosensors and proton pumps was elegantly demonstrated by introducing
three strategically placed hydrogen-bonding residues of NpSRII into BR, inducing a
functional interaction with NpHtrII.
336
The first identified eubacterial
sensory rhodopsin (ASR) was from
the freshwater cyanobacterium, Anabaena.
11
In contrast to SRI and SRII, ASR activates a
soluble transducer protein (ASRT), possibly leading to transcriptional
control of several genes.
169,337
Even though ASRT was
shown to interact with DNA in vitro, the exact mechanism
of photosensory transduction is not clear, especially in view of the
existence of many close homologues of ASR which do not coexist with
ASRT-like transducers.
338
In addition,
a direct interaction of the C-terminus of ASR with DNA has been reported,
suggesting that ASR itself may function as a transcription factor
as well.
339
The X-ray structure of ASR
has been determined, and it reveals an unusually polar water-filled
cytoplasmic half, distinct from that of BR and other proton pumps.
184
Unique to ASR, the highly conserved Asp of
the RSBH+ counterion (Asp212 position in BR; Supporting Information Figure 1) is replaced
by Pro, dramatically modifying the hydrogen-bonding pattern in the
SB region.
340
One of the most unique aspects
of ASR is its photochromism. Its chromophore binding site accommodates
both all-trans- and 13-cis,15-syn-retinal in the dark, which can be interconverted
by
choosing an appropriate wavelength of illumination due to the difference
in the spectral maxima of these two isomeric forms.
183,184,341
The working (and very preliminary)
model of ASRT activation includes release of the transducer from the
cytoplasmic side of the receptor during the formation of the M intermediate
arising from the excitation of all-trans-ASR.
177,342
Green algae, such as Chlamydomonas, give
us another
example of the versatility of sensory mechanisms in microbial rhodopsins.
167,343
In addition to channelrhodopsins (see below), the unusual histidine
kinase rhodopsins (HKR) have been recently identified. HKRs are modular
proteins consisting of a rhodopsin domain, a histidine kinase domain,
a response regulator domain, and in some cases an effector domain
such as an adenylyl or guanylyl cyclase, reminiscent of signaling
by LOV (light–oxygen–voltage) domain-bearing proteins.
Surprisingly, the heterologously expressed rhodopsin fragment acts
as a bistable UVA receptor which can be switched by UV and blue light
between 380 and 490 nm absorbing forms.
344
The chromophore of the former was found to be in the 13-cis,15-anti configuration,
which is normally observed
only during the photocycle of other rhodopsins. It was suggested that
the photochromic HKR1 plays a role in the adaptation of behavioral
responses in the presence of UVA light.
3.6
Channelrhodopsin,
a Light-Gated Cation Channel
3.6.1
Basic Discoveries
The work of many
scientists studying the light responses and swimming behavior of the
motile microalgae Euglena gracilis and Chlamydomonas
reinhardtii provided the basis for the discovery of channelrhodopsin.
345
The unicellular Chlamydomonas has an equatorial diameter of about 8 μm, two flagella,
and
a 1 μm orange eye. About 100 years ago, it was discovered that
the eye senses light to modulate the flagellar beating for helical
swimming of the alga.
346
In the 1950s it
was found that behavioral responses depend on Mg2+ and
Ca2+.
347
Next, a correlation
between Ca2+ influx and changes of the flagellar beat frequency
was observed.
348
The function of rhodopsin
as the sensory photoreceptor in microalgae was interpreted from published
action spectra for phototactic movement,
349
and was further supported by the capability of retinal and retinal
analogues to restore behavioral light responses in blind Chlamydomonas mutants.
350
Detailed retinal complementation
studies revealed an all-trans configuration as characteristic
for microbial rhodopsins, rather than the 11-cis configuration
characteristic of the animal (visual) rhodopsins.
351,352
Several years later, electric responses were measured from
single Chlamydomonas cells and action spectra were
recorded which led to the proposal that the photocurrents were mediated
by the photoreceptor rhodopsin to control phototaxis and phobic responses.
353
On the basis of the appearance of the photoreceptor
current in Chlamydomonas reinhardtii and Volvox carteri within 50 μs after a brief
light flash, it was postulated
that their photoreceptors and ion channel(s) are closely linked, forming
a tight complex.
353,354
In 2001, three groups identified
novel DNA sequences that encode microbial type rhodopsins in Chlamydomonas.
12,355−357
By in vivo analysis of photoreceptor electrical
current and RNA interference technology in Chlamydomonas, two rhodopsin genes were
found to mediate phototaxis.
356
The concept of a light-gated ion channel was
finally demonstrated through the expression of ChR in Xenopus oocytes and in HEK cells
by voltage clamp measurements.
12
Further experiments showed that both rhodopsin
DNAs from Chlamydomonas directly encode light-gated
ion channels, henceforth termed channel-rhodopsin-1 (ChR1) and channelrhodopsin-2
(ChR2).
12,13
In the alga, ChR1 is more dominant than
ChR2 but the amount of ChR1 strongly varies with environmental conditions,
358
whereas ChR2 is more or less constitutively
expressed, albeit at a lower level. ChR2 absorbs maximally at 460
nm whereas for ChR1 the absorption shifts from 465 nm at high pH to
505 nm at low pH with a pK
a of 7.5.
359
ChR2 inactivates (reduction of the conductance)
more strongly in continuous light compared to ChR1, and its selectivity
for sodium is 2-fold higher than that of ChR1.
3.6.2
Transfer to Neuroscience, Birth of Optogenetics
On
the basis of the ability to express light-gated ChR2 in HEK293
and BHK cells, it was also proposed for neuronal cells that expression
of ChR2 might allow depolarization of the cells by illumination.
13
Several groups began to apply ChRs in neuroscience,
mainly using a truncated version of ChR2, due to an expression level
twice as high as the full-length protein and more than 10 times better
than ChR1. With a set of seminal publications, five groups almost
in parallel demonstrated the applicability of ChR in hippocampal neurons,
PC12 cells, the spine of living chicken embryos, mouse brain slices,
the retina of blind mice and transgenic C. elegans.
360−364
These publications marked the genesis of what we term today, optogenetics.
In optogenetics, researchers target well-defined neuronal cell subpopulations
by using cell-specific promoters to express light-activatable proteins
and thus are able to selectively activate or silence (depolarize or
hyperpolarize) cells through the application of short light pulses.
Most surprisingly, the mammalian brain contains sufficient retinoid
levels to allow wild-type ChR2 to function without addition of exogenous
retinal. The affinity of the apoprotein for retinal is in the nanomolar
range but varies widely in ChR isotypes and mutants, partially explaining
why some variants work better than others in neurons, even though
expression and membrane targeting is equivalent.
365
More recently, channelrhodopsins from algal species other
than Chlamydomonas, most notably, from Volvox, have been similarly adopted for use
in optogenetics.
170,366,367
3.6.3
Channelrhodopsin
Architecture
ChRs
are microbial rhodopsins with long C-terminal extensions. These extensions,
unusual for a rhodopsin, appear to be important for targeting the
channel to the algal eyespot overlaying part of the plasma membrane
but are not required for ion channel function. Despite remarkable
achievements in protein engineering, a mechanistic understanding of
activation and ion transport in channelrhodopsins has been hampered
due to a lack of accurate structural data. Although ChR1 is the dominant
photoreceptor in the alga, our knowledge about ChR1 and its relatives
in other algae is scarce since the heterologous expression in host
cells is difficult.
356,359
ChR2 expresses reasonably well
in mammalian COS-1 and HEK293 cells as well as in the yeast Pichia pastoris, allowing
purification and spectroscopic
characterization.
368−370
Recently, two new structural models
have been proposed independently, in addition to the earlier model
built using low-resolution electron crystallography data.
371
The first one was based on a 2.3 Å X-ray
crystal structure (PDB ID: 3UG9), which revealed the dark state of C1C2, a ChR1/ChR2
chimera variant mainly consisting of ChR1 with the last two helices
derived from ChR2.
51
The C1C2 hybrid
372
(also termed ChrGR
373
and ChEF
374
in the literature) has not
been well-characterized spectroscopically, but as judged from electrical
studies its properties are close to those of ChR1. The other model
was built for ChR2 using homology modeling on the ASR template, global
sampling with classical force-field MD simulations, and structural
refinement by combined QM/MM methods.
375
These models agree in many aspects such as orientations of common
charged residues, hydrogen-bonding patterns, and the active site.
From the seven TM helices, helices C–G of ChR1 and ChR2 have
sufficient homology to other microbial rhodopsins (Supporting Information Figure 1)
so that a homology model
is readily built; however, ambiguities in the model arise from helices
A and B which both contain a large number of unique charged residues,
unlike BR or ASR.
Figure 17 shows the
structure of ChR2 as
derived from homology modeling based on the C1C2 crystal structure.
376
The model was embedded into the membrane and
aqueous solution including sodium and chloride ions. The computational
model displays a larger water density in the protein core compared
to the crystal structure, probably because many of these waters are
highly mobile and may not contribute to a defined electron density
in the X-ray structure. The water is mainly distributed along helix
B, where a large number of charged amino acids are lined up, and a
continuous water distribution can also be observed at the cytoplasmic
side, stabilized by a cluster of hydrophilic residues consisting of
Glu82, Glu83, His134, His265, and Arg268. The two continuous water
densities are separated by the residues Ser63 and Glu90, which are
highly conserved among the ChRs, and may define constriction sites
as discussed previously.
51
Glu90 deprotonates
during the photocycle; however, the functional importance and exact
timing of this event in the photocycle is controversial.
377,378
Figure 17
Structural model of ChR2 based on the C1C2 chimera crystal structure
as derived from extensive MD simulations.
376
Shown are the relevant amino acids as discussed in the text and
the calculated water (blue) and sodium ion (yellow) distributions
averaged over the course of MD simulations. Insets: cation binding
site (top) and cation uptake pathway observed in MD simulations with
sodium trajectory (bottom). Cytoplasmic side of ChR2 is facing up.
The active site in ChR2 is defined
by Glu123 (Asp85 in BR) and
Asp253 (Asp212 in BR). The hydrogen-bonded network as observed in
the C1C2 crystal structure differs from that of BR, with fewer water
molecules and a direct salt-bridge between retinal and the counterion
(Figure 17). Classical MD simulations predict
water molecules entering the active site. In the case of the ChR2
dark-adapted state, this could lead to the formation of a hydrogen-bonding
network similar to the pentagonal cluster found in BR, since the ultrafast
reaction dynamics after photoexcitation is similar for these proteins,
379
but with faster energy transfer from retinal
to protein and water molecules in ChR2.
139
However, due to imperfections in present classical force field methods,
375,376,380
the precise active site water
arrangement could not be determined until now. In the C1C2 X-ray structure,
Asp253 (ChR2 numbering) is in close proximity with the RSBH+ compared to Glu123, and
the electrical studies on various ChR2 mutants
revealed that Glu123 (but not Asp253) is dispensable for function.
51,381
This suggests that either Asp253 is the preferred primary proton
acceptor or it takes over this function in the absence of Glu123.
Interestingly, Glu123 serves as the voltage sensor that regulates
the photocycle speed at different membrane voltages,
382
as replacement of Glu123 with Gln, Thr, or Ala completely
eliminates the voltage sensitivity of the reaction cycle thus allowing
ultrafast action potential firing when expressed in neurons.
382
3.6.4
Channel Function, Gating,
and Selectivity
Electrical studies in combination with structural
data identified
two gates as key elements of the channel, the central gate defined
by Ser63, Glu90, Asn258, and, second, the inner gate built by Glu82,
Glu83, His134, and His265. Density attributed to a sodium ion has
been found by computer modeling in a cavity defined by these residues
(Figure 17, inset). Conformational changes
open both gates and mediate pore formation and ion conductance, but
our knowledge of the conducting state is still vague since the structural
information about the conducting state(s) is lacking. The key residue
of the central gate, Glu90, is in close proximity to the RSBH+, but the mechanistic
rationale for this connection is unclear.
Both the central gate and inner gate serve as selectivity filters,
and mutagenesis of the participating residues changes the selectivity
quite substantially. For example, both H134R (inner gate) and E90Q
mutants (central gate) conduct more Na+ than the wild-type
and give larger currents, especially under alkaline conditions.
374,383
The two gates are the objectives of intensive research in expectation
that modification of the gates will reveal novel and useful ChR properties.
Despite the large volume of extant electrophysiological data, in the
future, X-ray crystallography, NMR, theoretical modeling, and time-resolved
vibrational spectroscopy must be employed to fully understand the
gating process.
The ultrafast retinal isomerization, similar
to other rhodopsins, activates the protein and leads, ∼1 ms
later, to an opening of the ion pore. The structural rearrangement
of the chromophore is expected to be very minor, but the NH+ dipole of the RSB is
switched upon isomerization, leading to a rearrangement
of the H-bond network surrounding the RSB. The induced rearrangement
of the protein structure is a multistep process, and at present, only
a few intermediate states can be identified as defined by their optical
spectra. The on-kinetics of the photocurrents (on-gating) are defined
by this sequence of reactions. So far, two early reaction intermediates,
P500 and P380 (corresponding to the L and M states of bacteriorhodopsin),
have been assigned, preceding the conducting state P520 which relates
to the N state of BR (Figure 18).
368,369,379
The structural changes involved
in the opening are reversed for a closure of the conducting pore,
completing the photocycle by leading back to the dark state. Opening
and closure of the ion pore follow different reaction pathways because
the kinetics of channel opening and dark state recovery differ by
many orders of magnitudes. By electron paramagnetic resonance spectroscopy,
a light-induced movement of helix B was identified as an important
step in this process.
384,385
Strictly speaking, the ion selectivity
of the channel immediately after the onset of illumination and under
steady-state conditions differs significantly suggesting the existence
of two conducting states, O1 and O2. These states are populated differently
depending on the light intensity, wavelength, and duration. At the
moment there are no spectroscopic criteria available to discriminate
O1 from O2, and they are both subsumed under the photocyle intermediate
P520 in Figure 18.
Figure 18
Simplified scheme of
the ChR2 photocycle with D470 as dark-adapted
state and P520 as conducting state based on data from refs (369, 370, 379, 386). It
is worth noting that during
bright continuous illumination or repetitive flashing the photocycle
is fed by photoconversion of the late photocycle intermediate Des480,
the so-called desensitized or light-adapted state (blue dotted arrow).
Green and UV light photoconverts the conducting state P520 and the
early state P390, respectively, back to the light-adapted state (green
and purple dotted arrows).
The residues Cys128 and Asp156, which are in contact with
each
other via a water molecule (Figure 17),
51,376
play a fundamental role in on- and off-gating as evidenced by the
fact that mutation of either Cys128 or Asp156 causes a dramatic slowdown
of reaction kinetics and results in a great increase of the lifetime
of the open state(s).
386,387
Illumination of the open state
(P520) with longer wavelength light results in photochemical conversion
back to the dark state and closes the channel. Since these ChR variants
are bimodal and switchable with light they were named “Step
Function Rhodopsins” or simply SFRs.
386−388
Asp156 deprotonates during the photocycle with kinetics that closely
correlate with the reprotonation of the RSB nitrogen suggesting that
this Asp156 is the proton donor.
378
Cys128
is hydrogen-bonded via Thr127 to the side chain of the active site
residue Glu123, a feature that has been considered to be crucial for
the structural changes occurring during the photocycle. Simulations
suggest that this hydrogen bond in the dark state switches to a backbone
oxygen of Glu123 (intrahelical hydrogen bond) after its protonation.
376
This finding suggests a mechanism where the
protonation change of Glu123 is propagated via Thr127 to Cys128 and
Asp156, which change conformations accordingly. This event is believed
to cause breakage of the water-mediated hydrogen bond between them,
the central event for channel opening according to spectroscopic measurements.
389−391
However, this model is challenged by the fact that Glu123 is dispensable
for ChR2 activation,
382,392
and in E123T or E123A mutants
Asp253 functions as the negative RSBH+ counterion. Certainly,
additional residues are involved in the activation process.
375,393
During the lifetime of the light-induced conducting state,
P520,
up to 100 ions are conducted in wild-type ChR2 with a unitary conductance
of 40–100 fs.
374,394,395
Ions approach the inner gate via the access channel that is mainly
defined by the charged residues of helix B (Figure 17). The individual mutations of
up to three glutamates on helix
B (Glu101, Glu97, and Glu90) reduced the cation conductance gradually,
but did not abolish transport completely.
383,396−398
MD simulations predict the access channel
entrance to be defined by Asn53, Gln56, Glu97, and Glu101, due to
the high ion density found in this region.
376
3.6.5
Perspectives for Optogenetics Development
The hopes for future applications of optogenetics are high, but
ChR is not the optogenetic prodigy as some advertise it, as it shows
clear limitations for optogenetic use: (i) Nature employs ChR for
gradual membrane depolarization, resulting in small conductance and
not in the all or none responses required for efficient optogenetic
use. (ii) Molecular engineering may lead to wider pores (and thus
greater depolarization), but only at the risk of destabilization and
thermal activation in the dark. (iii) Although the ion selectivity
can be changed toward higher or even exclusive H+ conductance
(as, e.g., found naturally for ChR of the halotolerant alga Dunaliella salina)
170
or tuned
toward higher selectivity for monovalent or divalent cations, it will
be an enormous challenge to reach the high selectivity ratio for K+ over Na+ necessary
for light-controlled hyperpolarization
of host cells. (iv) The upper limit for possible shift of the absorption
maximum will be reached at approximately 630 nm due to an increase
of thermal activation with longer wavelengths.
399
Nevertheless, engineering of ChR and other microbial
rhodopsins will continue and the ongoing sequencing of hundreds of
algal genomes will provide countless new ChR variants with more advantageous
characteristics.
400
Furthermore, development
of improved techniques to target ChRs to organelles and specific membrane
subareas, to make ChR bimodally switchable, to control expression
more accurately, and to guarantee a better turnover of the protein
when used as retinal prostheses in bright light vision will continue.
Several ChRs with blue- and red-shifted absorption were described
recently and await further characterization.
366,401,402
A set of ChRs that covers a
spectral range similar to that of rhodopsins of animal eyes is within
reach. Also, ChRs will be further optimized for two-photon microscopy.
403
Nevertheless, ChRs are basically analytical
tools and will only be used for treatment of diseases in patients
in a few very special cases, such as deep brain stimulation or retinal
prosthesis.
404−406
From our perspective, the greatest
potential for optogenetic application
lies unexploited in a smart combination of light-gated channels or
light-driven pumps with endogenous channels of host cells in the sense
that both primary outward or inward directed transport of Ca2+ or H+ can be used to
activate Ca2+- or H+-sensitive endogenous channels of high selectivity and high
conductance (two-component optogenetics). The question will be how
to connect these two units to ensure that the on-gating of the endogenous
channel is fast enough for the specific needs. Such constructs will
move the field a great step forward. Similarly, gene-fusion strategies
for combining excitatory ChR2 and various inhibitory ion pumps or
bovine Rho and a G protein-activated potassium channel have been explored.
407,408
4
Animal Rhodopsins
4.1
Animal Rhodopsins are Prototypical GPCRs
In the previous
sections we have seen how microbial (or type I)
rhodopsins are employed in microorganisms in light-dependent functions
as ion pumps, channels, and sensors. Animals employ in their eyes
and other organs a different type of rhodopsin, which was optimized
during evolution to perform a variety of roles pertaining to vision,
sensation of light for nonvisual reasons (e.g., circadian rhythms,
sensing dawn/dusk, determining the horizon, pupillary constriction,
body color change, and seasonal reproduction), and direct utilization
for the isomerization of retinal.
3,4,409−412
These animal (or type II) rhodopsins are
GPCRs which are activated by light to catalyze GDP → GTP nucleotide
exchange in heterotrimeric G proteins. Opsins belong to the largest
GPCR family (the family of rhodopsin-like GPCRs with ∼700 members
in humans)
413
and can be roughly subdivided
into ciliary and rhabdomeric opsins, which are further diversified
by their G protein subtypes, and photoisomerases.
3,410,414
Typically, expression of the ciliary rhodopsins
occurs in vertebrate ciliary photoreceptor cells with cyclic nucleotide
signaling cascades or in invertebrate rhabdomeric photoreceptor cells
with phosphoinositol signaling cascade. However, this is a simplified
view, as ciliary and rhabdomeric opsins are present in both vertebrates
and invertebrates, and expression is also known for other cell types,
such as intrinsically photosensitive retinal ganglion cells and human
epidermal melanocytes.
3,4,409,414−416
Although the
7TM architecture appears at first glance to be similar for microbial
and animal rhodopsins (Figures 1 and 4), the primary and tertiary structures of microbial
and animal rhodopsins differ largely.
252,413,417−419
As GPCRs, animal rhodopsins
feature kinked TM helices, a strongly tilted TM3, longer cytoplasmic
loops connecting TMs, a cytoplasmic helix 8 (H8) with palmitoylated
Cys residue at its end, a disulfide bridge linking TM3 with the extracellular
loop 2 connecting TM4 and TM5, as revealed by the crystal structures
of bovine Rho,
46,420,421
and squid rhodopsin (Figures 19 and 20 and Supporting Information Figure 4).
422,423
Unique to rhodopsin is a compact
extracellular domain that is formed by the N-terminus and extracellular
loops connecting TMs, with a β-sheet in extracellular loop 2
forming a plug to cover the hydrophobic retinal and the RSB. This
“retinal plug” occludes the chromophore-binding site,
424,425
whereas retinal binds in microbial rhodopsins closer to the center
of the protein such that a plug is not needed (cf. Figures 1 and 4). Due to limited
data,
it is not clear whether microbial rhodopsins and animal rhodopsins
(as GPCRs) have a common ancestry or acquired their photosensitivity
and topology independently. Although the sequence similarity between
microbial and animal rhodopsins is low, they share in TM7 a specific
Lys residue for retinal attachment, and in TM6 they share homology
for two aromatic residues that serve to interact with the retinal
(Trp265 and Tyr268 in bovine Rho). It has therefore been proposed
that animal rhodopsins may have developed via exon shuffling caused
by recombination of microbial rhodopsins and GPCRs. Exons for TM6
and 7 would have been provided by microbial rhodopsins and for TM1–5
by GPCRs.
415
Figure 19
Global changes in the
structure of rhodopsin and other GPCRs upon
attainment of the active state. (A) Structural superposition of inactive
and active G-protein-interacting state of bovine Rho reveals structural
rearrangement of TM5 and TM6 to accommodate binding of the C-terminus
of the α-subunit of transducin (Gtα peptide,
shown in gray). Inactive rhodopsin (dark state; PDB ID: 1U19) is denoted in red,
and active rhodopsin (Meta II; PDB ID: 3PQR) is denoted in yellow. (B) Model of the
complex between a rhodopsin dimer and transducin built into electron
microscopy map derived from native source purified bovine rhodopsin/transducin
complex. GDP/GTP binding site is denoted by a yellow hexagon in parts
B and D although both structures are solved in the nucleotide-free
state. (C) Structural superposition of antagonist bound (inactive
state, denoted in green) and agonist bound (active state, denoted
in orange) β2-adrenergic receptors (β2AR). Structural displacement of TM5 and TM6
is similar to that seen
in the comparison shown in part A. A nanobody (shown in gray) was
utilized to stabilize the agonist bound state, and a nanobody loop
protrudes into a similar position as seen for the Gtα peptide which stabilizes the
Meta II state in part A. A T4 lysozyme
(T4L) domain used to facilitate crystallization is not shown for clarity.
The antagonist bound structure is the carazolol bound β2AR-T4L fusion (PDB ID: 2RH1),
and the agonist bound structure is
the nanobody stabilized, BI-167107 high affinity agonist bound structure
(PDB ID: 3POG). (D) Crystal structure of an agonist bound β2AR-T4L
fusion (T4L not shown) in complex with its cognate heterotrimeric
G protein, Gsαβγ, and a stabilizing nanobody
(not shown) reveals the mode of Gs binding to monomeric β2AR (PDB ID: 3SN6). In this
complex, the Gsα C-terminus also binds
in the cleft formed by the outward movement of TM5 and TM6. All representations
are in approximately the same orientations, and all superpositions
were performed with TM1 to TM4, TM7, and cytoplasmic helix H8 to accurately
portray the differences in the positions of TM5 and TM6.
Figure 20
Sequence and motif conservation in GPCRs extends to ordered
bound
water molecules. Sequence conservation among rhodopsin-like (class
A) GPCR sequences was mapped onto the backbone of rhodopsin as reported
in ref (463); greater
“tube” thickness and ramping from blue to red indicate
greater residue conservation at that position. Because of considerably
lower sequence conservation outside of the TM region, these regions
were not included in the analysis and are denoted in white. Structural
superposition of all antagonist bound structures of GPCRs with a resolution
2.7 Å or higher reveals a subset of ordered water molecules that
are found within the transmembrane bundle (shown in light blue). As
indicated by the color and thickness of the cartoon representation,
these waters are found in close proximity to positions within the
TM region that have high homology throughout all class A GPCRs. Water
molecules shared among four or more different receptor structures
are shown here. In addition, density best represented by a bound octahedrally
coordinated sodium ion has been found within the TM bundle of A2A-adenosine and PAR1
receptor structures (PDB ID: 4EIY, 3VW7; shown as a black
sphere) in a similar position to a water observed in the bovine Rho
structure (PDB ID: 1U19). Three motifs important for GPCR activation
are denoted by shaded ovals.
The GPCR 7TM scaffold allows animal rhodopsins to undergo
conformational
changes upon activation by light which are larger than the light-induced
changes in microbial rhodopsins. Key for G protein coupling to active
rhodopsin is a 2–3 Å motion of TM5 and a 6–7 Å
motion of TM6 which open up the cytoplasmic surface for G protein
binding and catalysis of GDP/GTP exchange in the G protein.
426
A summary of the current knowledge of the structural
basis of GPCR/G protein coupling is shown in Figure 19. Movement of TM5 and TM6 was
shown first in the crystal structure
of bovine opsin, the apoprotein of Rho.
427,428
This crystal structure is consistent with TM movements upon transition
from the resting state to the active receptor state as concluded from
electron paramagnetic resonance (EPR) spectroscopy
429,430
and biochemical cross-linking experiments.
430,431
On the basis of these results, it was proposed that the opsin conformation
(Ops*) observed in these crystals and subsequent crystallographic
studies of opsin bound to G-protein-derived peptides was analogous
to that of the active state.
428,432
Indeed, the protein
backbone conformation found in the opsin crystal is almost perfectly
superposable with the structure of the active Meta II photoproduct
(Figure 19A),
426
as well as with the structures of constitutively active rhodopsin
mutants.
433−435
The presence of a G-protein-interacting
conformation in Ops* or
Meta II structures was established by cocrystallization experiments
with synthetic peptides representing one of the key interaction sites
on rhodopsin’s cognate heterotrimeric G protein, transducin
(Gt).
426,428,433
Similar to
these peptides derived from the distal C-terminus of the Gtα-subunit, suitable nanobodies
can stabilize the active conformation
of GPCRs as was demonstrated for the human β2-adrenergic
receptor (β2AR) in an active agonist-bound form,
crystallized in complex with a nanobody binding in the cytoplasmic
cleft opened for the G protein (Figure 19C).
436,437
In the subsequently solved crystal structure of the nucleotide-free
complex between the β2AR and the G protein Gs, the
binding mode for the Gsα C-terminus was revealed
to be similar to that of the Gtα-derived peptide
bound to Meta II.
438
The β2AR–Gs complex structure provided important insight into the
interface and the relative orientation of the receptor and G protein
and confirmed the role of the Gα C-terminal helix
in coupling the Gα nucleotide binding site to the
GPCR. GPCR-catalyzed nucleotide exchange in heterotrimeric G proteins
is a complex multistep process which is not fully understood yet;
the reader is referred to the literature for greater detail.
438−441
For the Rho/Gt pair, a high-resolution structure of the nucleotide-free
complex remains elusive, but a model similar to the β2AR-Gs complex was built into
the 3-D envelope calculated from single
particle analysis of negative-stained electron microscopy images of
bovine Rho–Gt complexes purified directly from native source
(Figure 19B).
442
The proposed Rho–Gt complex model, however, differs from
the β2AR–Gs complex crystal structure by comprising
one G protein and two Rho molecules, with one protomer interacting
with the active Rho* conformation, although binding to Gt occurs in
a similar orientation.
442−444
Dimerization of rhodopsin and
opsin as well as of other GPCRs (for review see ref (445)) can occur and affect
receptor function and signaling allosterically.
446−448
Animal rhodopsins thus can be seen as both prototypic and
highly
specialized GPCRs which differ from diffusible ligand-activated GPCRs
by their unique photoactivation mechanism. On one hand, the GPCR 7TM
scaffold of animal rhodopsins has been optimized by evolution, for
fast and selective retinal isomerization with the high quantum yield
(section 2) necessary for fast millisecond
time scale receptor activation. Also, on the other hand, visual rhodopsins
couple to specific G proteins, transducins, specific to rod or cone
cells, which were co-optimized for high fidelity and fast nucleotide
exchange. For bovine Rho in native membranes, a nucleotide exchange
rate of several hundred Gt turnovers per single light-activated rhodopsin
molecule has been determined.
449
This value
is much higher than that computed for other G protein signaling systems.
450
It is likely related to the fact that vertebrate
photoreceptor cells are responsive to a large range of photons, spanning
8 orders of magnitude,
451
starting from
single photons, where fast nucleotide exchange is mandatory for sufficient
signal amplification. Due to the common architecture found for all
GPCR structures determined to date and common gross conformational
changes observed in bovine Rho and other GPCR structures upon activation,
452
knowledge of the activation steps in Rho’s
photochemical core and signal propagation from the retinal binding
site to the G protein coupling cytoplasmic surface is likely to be
directly applicable to understanding of the activation of other GPCRs.
4.2
Bovine Visual Rhodopsin as a Model System
A wealth of information on the structure and function of rhodopsin
has been gained from the bovine visual signal transduction system.
2
Rhodopsin is the most abundant membrane protein
in the outer segment of retinal rod photoreceptor cells. Because of
the high sensititvity of these photoreceptors, single photons are
sufficient to couple retinal cis → trans isomerization via rhodopsin conformational
changes
and GDP–GTP exchange catalysis to Gt activation. Activated
(GTP-bound) Gt further amplifies the signal via activation of a cGMP-specific
phosphodiesterase and the resulting closure of cGMP-gated cation channels
in the plasma membrane to generate an electrical signal (see, e.g.,
ref (453)). Visual
signal transduction in rod cells has been studied extensively by biochemical
and biophysical means, because of the relative ease of obtaining pure
preparations of protein from bovine eye tissue. This enabled bovine
rhodopsin to be the first GPCR to be conventionally sequenced
454,455
and cloned.
456
Its sequence homology
to the β2AR was observed in 1986,
457
making rhodopsin the eponym of the largest class of GPCRs
which comprises much of the ∼800 GPCRs in the human genome.
413
Rhodopsin was also chosen for the first chemical
synthesis of a eukaryotic GPCR gene for expression in mammalian cells
allowing extensive structure/function studies.
458−461
A milestone in both rhodopsin and GPCR research was the elucidation
of the crystal structure of bovine rhodopsin by Palczewski and colleagues
in 2000.
421
This first GPCR structure served
and continues to serve as a framework to interpret much of the biochemical
and biophysical studies on GPCRs. The structure also revealed constraints
which stabilize the inactive dark state of rhodopsin. Among these
constraints are hydrogen-bonding networks involving the RSB region
and conserved residues in the cytoplasmic part of TMs (Figures 20 and 25, and Supporting
Information Figure 4). Comparison
of inactive and active Rho and GPCR structures indicate general functional
roles for microdomains comprising residues of motifs conserved among
rhodopsin-like GPCRs (Figure 25). The TM3/6
network comprises four residues with a central salt bridge between
Arg135 of the (D/E)RY motif in TM3 and Glu247 on TM6 known as the
“ionic lock”, and was initially proposed for the β2AR as a feature that stabilized
the inactive receptor conformation.
462
The NPxxY(x)5,6F motif constrains
the end of TM7 with H8 by electrostatic interaction between aromatic
rings of Tyr306 on TM7 and Phe313 on H8. Asn302 of the NPxxY(x)5,6F motif is linked
to the most conserved residues on TM1
(Asn55) and TM2 (Asp83) forming the TM1/2/7 network. The model of
rhodopsin in Figure 20 also reflects results
from GPCR sequence analyses,
463
with the
ribbon thickness coding for amino acid conservation, and results from
GPCR structure analyses, where structurally bound waters are observed
in “homologous” positions. This homology of water positions
couples to the high degree of conservation seen in residues interacting
with these waters, suggesting a functional role for water in the activation
process.
464,465
Figure 21
Activation states of
rhodopsin and GPCRs. (A) In the photochemical
core process, photon energy is used to convert the inverse agonist
11-cis-retinal into the full agonist all-trans-retinal. Energy stored in the initially
twisted all-trans-retinylidene-Lys296 chromophore is gradually released
via local protein (side chain) conformational changes in the Batho
and Lumi photoproducts. Conformational changes in more distant parts
of the protein begin within microseconds of when Meta I forms, the
first intermediate of several Meta states in equilibrium, which as
GPCR funtional states interact with G protein, GRK1 (rhodopsin kinase),
and arrestin. Deprotonation of the RSBH+ and protonation
of its counterion Glu113 lead to the formation of Meta II substates
which develop sequentially. The largest conformational changes (inset
and Figure 19A) are observed in the transition
from Meta IIa to Meta IIb; the latter intermediate is further stabilized
by proton uptake to Glu134 of the (D/E)RY motif at the cytoplasmic
end of TM3. The retinal-free apoprotein opsin exists also in an equilibrium
between inactive (rhodopsin-like) and active (Meta-II-like) conformations,
426,427,530
termed Ops and Ops*. (B) Diffusible
ligand activated GPCRs similarly exist in equilibrium between inactive
and active conformations in which similar activating conformational
changes as in the Meta states are thought to occur.
503
Ligand binding shifts the equilibrium toward ligand type
specific energetic states.
567,568
Figure 22
Spectroscopically detected intermediates of photoactivated bovine
rhodopsin. Photoisomerization of the retinal 11-cis double bond leads within femtoseconds
to photorhodopsin with a highly
distorted 11-trans bond. Via thermal relaxation,
several intermediates form with distinct λmax values,
distinguishable by low-temperature or time-resolved spectroscopy.
476,569
Gradual release of the strain in the chromophore leads through Batho
and Lumi to Meta I, as seen by the different absorption maxima that
arise from changes in chromophore/protein interaction. A transient
blue-shifted intermediate (BSI) cannot be trapped at low temperatures.
Time-resolved UV–vis measurements revealed the existence of
additional transient forms of Lumi (Lumi II),
570,571
and Meta I (Meta I380;
572,573
Meta Ib).
574
The RSBH+ remains protonated up
through Meta I, probably due to the low pK
a of the stabilizing counterion Glu113. Larger protein conformational
changes lead to Meta II (comprising substates Meta IIa and Meta IIb)
which is in equilibrium with its predecessor Meta I. Meta II is the
agonist-bound active receptor state capable of catalyzing GDP/GTP
nucleotide exchange in the G protein transducin. Meta II is characterized
by a deprotonated RSB resulting in a large blue-shifted value for
λmax (380 nm). As a result of RSB hydrolysis Meta
II decays to the apoprotein opsin and all-trans-retinal.
Meta I can also form Meta III, involving thermal isomerization of
the RSBH+ (λmax = 465 nm) from all-trans,15-anti to all-trans,15-syn. Meta III decays
to opsin and all-trans-retinal, but can also be photoconverted to Meta I
and Meta II.
575
Unlike in invertebrates,
bovine rhodopsin cannot be regenerated in situ by
reisomerization of retinal with a second photon. All-trans-retinal is reduced to retinol
by retinol dehydrogenase and transported
out of the photoreceptor cell to adjacent retinal pigment epithelial
cells, where 11-cis-retinal is regenerated (for details,
see ref (551)). Adapted
with permission from ref (576). Copyright 2002 John Wiley & Sons, Inc.
Figure 23
Isomerization, elongation, and rotation of retinal upon
light activation
of rhodopsin. (A) Superposition of retinal and the Lys296 and Trp265
side chains in bovine Rho (11-cis-retinal, PDB ID: 1U19, red sticks), Batho
(twisted all-trans-retinal, PDB ID: 2G87, gray sticks), Lumi
(partially relaxed all-trans-retinal, PDB ID: 2HPY, white sticks),
and Meta II (relaxed all-trans-retinal, PDB ID: 3PXO, yellow sticks)
structures. Note that from Lumi to Meta II retinal undergoes a large
rotation along its long axis. (B) Overlay of rhodopsin and Meta II
structures showing differences in the positions of the TM helices.
Note that retinal movement induces TM5 motion and rotational tilt
of TM6.
426,435
View from cytoplasmic side.
Figure 24
Conformational changes upon rhodopsin activation leading
to the
Meta II activated state. ① Photon absorption causes retinal cis → trans isomerization
and small
scale changes in structure in the immediate vicinity of the retinal,
driving all subsequent activation steps. ② Deprotonation of
the RSBH+ along with further small-scale changes within
the TM region. ③ Signal propagation to two regions almost universally
conserved in class A GPCRs, the (D/E)RY and NPxxY(x)5,6F motifs. Changes in the (D/E)RY
motif (in TM3; Glu134, Arg135, Tyr136
of bovine Rho), resulting in disruption of the “ionic lock”
between Arg135 and Glu247 (on TM6), and changes in the NPxxY(x)5,6F region (TM7/H8)
which rearranges. ④ Proton uptake
from the cytoplasm onto Glu134. The TM helices are depicted in the
following colors: TM1, blue; TM2, teal; TM3, green; TM4, lime green;
TM5, yellow; TM6, orange; TM7, red; and H8, purple.
Figure 25
Structural changes in the chromophore binding site and
conserved
motifs that accompany bovine Rho activation. Upon light-induced activation
of Rho a series of small scale structural changes result in the release
of restraints enabling attainment of the fully active Meta II state.
(A, B) Glu181 is found hydrogen bonded to a water molecule in both
the dark state and Meta II structures, and it appears that this water
functions as a noncovalently bound cofactor which moves along with
Glu181 to stabilize the deprotonated RSB in Meta II. (C, D) TM3/TM6
restraints in the dark state due to the “ionic lock”
formed by Arg135, Glu134, Glu247, and Thr251 are released upon Rho
activation. In Meta II new interactions are formed between TM3–TM5
(Arg135–Tyr223) and TM6–TM5 (Glu247–Lys231).
(E, F) Structural changes within the NPxxY(x)5,6F motif
upon activation of Rho entail a remodeling of solvent-mediated hydrogen
bonding of two water molecules as well as a 180° change of rotamer
for Tyr306 and a concomitant shift of the conserved Phe313 residue.
For ease of interpretation, helices are depicted in the following
colors: TM1, blue; TM2, teal; TM3, green; TM4, lime green; TM5, yellow;
TM6, orange; TM7, red; and H8, purple. For all comparisons, PDB ID: 1U19 was used
for dark
state and PDB ID: 3PQR was used for Meta II state.
4.2.1
Photochemical Core and GPCR Conformations
in Equilibrium
Rhodopsin differs from diffusible ligand-activated
GPCRs by its covalently linked retinal and photochemical core enabling
photosensory function. 11-cis-retinal serves together
with the constraints described above in highly efficient stabilization
of the inactive dark state of rhodopsin. As a consequence, thermal
activation of a single bovine rhodopsin molecule occurs only after
420 years on average,
466
which is a prerequisite
for rhodopsin’s high photosensitivity and the ability of rod
photoreceptor cells to function as a single photon detector. Thermal
rhodopsin activation causes discrete electrical signals in the dark
(dark events) and is thought to occur from retinal isomerization.
399,467,468
The measured thermal activation
barrier of 80–110 kJ/mol is lower than the barrier for light-dependent
activation of bovine Rho (≥180 kJ/mol, Figure 7 and section 2).
469−471
For a value of >180 kJ/mol, thermal isomerization of a rhodopsin
molecule once every 1010 years would be expected, implying
that thermal and light-dependent processes follow different pathways.
A more recent theoretical study proposed a pathway of thermal isomerization
in rhodopsin with a transition state displaying the same charge-transfer
character as the electronically excited state of Rho.
472
From a quantitative relation between rhodopsin’s
photoactivation energy and its peak absorption, λmax, it was proposed that dark noise
arises from thermal retinal isomerization
which needs to overcome the same energy barrier as in the photoisomerization
process.
399
On the basis of the slow hydrogen/deuterium
exchange of Thr118 in bovine Rho, it was suggested that local protein
structural fluctuations transiently widen the retinal binding pocket
for thermal retinal isomerization.
473
It
is interesting to note that Drosophila rhodopsin
also has a light-independent role in temperature discrimination in
larvae which may be related to thermal retinal isomerization.
474
To effectively release the structural
constraints that stabilize the inactive rhodopsin state, photon energy
is absorbed and used for retinal isomerization (cf. section 2), driving subsequent
protein conformational changes.
About 150 kJ/mol of the initially absorbed photon energy are stored
in the “distorted” all-trans-retinal
of the Batho photointermediate and gradually dissipated via a transient
blue-shifted intermediate (BSI)
475
and
the Lumi intermediate, concomitantly with a release of strain in retinal
(Figures 21–23), eventually yielding Meta I after a few microseconds. The “early”
photointermediates Batho and Lumi can be trapped by low temperature
and have been studied structurally,
120,146
as well as
spectroscopically.
476−478
The distinct absorption maxima of each intermediate
reflect the gradual changes in chromophore–protein interaction.
It is not until the formation of Meta I that significant backbone
structural changes occur.
479,480
In Batho and to a
lesser extent in Lumi intermediates, protein conformational adjustments
to retinal relaxation are limited to a few amino acid side chains
within the retinal binding pocket.
120,146
When
the Meta I state is attained, an equilibrium between Meta
I and Meta II states develops (Figure 21),
which is dependent upon pH and temperature, with lower pH and higher
temperature favoring Meta II.
481,482
Deprotonation of the
RSBH+ upon Meta II formation results in the characteristic
100 nm blue-shift of the absorption maximum to the near UV region
(λmax = 380 nm).
481,483
Meta II,
as defined by its 380 nm absorption, comprises both the isochromic
Meta IIa and Meta IIb substates which develop sequentially from Meta
I.
484
In Meta IIb, proton uptake occurs
to Glu134 of the (D/E)RY motif in TM3,
485,486
explaining
why low pH favors Meta II despite the loss of a proton from the RSBH+.
487
Time-resolved EPR studies
with bovine Rho in dodecylmaltoside detergent revealed that the large
TM6 movement occurs during transition from Meta IIa to Meta IIb and
led to the reaction scheme for the Meta states shown in Figure 21.
488
FTIR studies confirmed
this reaction scheme for bovine Rho in its native membrane environment.
489,490
Electron crystallography on bovine Rho 2-D crystals that were illuminated
and trapped in the Meta I photointermediate by the crystal lattice
demonstrated that up until Meta I the protein backbone remains in
a conformation similar to that of the rhodopsin dark state.
491
Illumination of 3-D bovine Rho crystals yielded
the spectral shift characteristic to Meta II, but only revealed a
small TM6 movement and some rearrangement of the cytoplasmic surface.
492,493
On the basis of absorption maximum and the extent of TM6 movement,
the structure of this photoactivated Rho most likely represents the
Meta IIa state. The Meta IIb state with fully opened cytoplasmic domain
is represented in the bovine Meta II structures obtained by reconstitution
of Ops* crystals with all-trans-retinal or by illumination
of the constitutively active M257Y rhodopsin mutant before crystallization.
426,433
Examination of agonist-bound GPCR structures reveals that
GPCRs
exist in conformations with differing extents of TM6 movement, with
no or little TM6 movement for inverse agonists and larger TM6 movement
for partial agonists and full agonists.
418
However, stabilizing mutations, truncations of loops, and in many
cases insertion of T4 lysozyme or apocytochrome b562 fusion
partners into cytoplasmic loop 3 (connecting TMs 5 and 6) or at the
N-terminus have been necessary for the crystallization of all nonrhodopsin
GPCRs to have their structures determined to date.
419,494,495
This affects affinity for agonist
or antagonist binding and likely exerts some influence upon the degree
of movement of TM6.
496−498
NMR and hydrogen/deuterium exchange studies
on β2AR lacking fusion partners provide evidence
for substantial conformational heterogeneity of agonist- and inverse
agonist-bound β2-AR preparations.
499−502
The heterogeneity supports the view of conformational equilibria
of GPCRs that can be shifted to either side depending on the type
of ligand and are comparable to the Meta I/Meta II equilibrium of
rhodopsin (Figures 21 and 22).
503,504
A difference between diffusible
ligand-activated GPCRs and the activation of rhodopsin by light is
reflected in the mode by which the ligand acts in the activation process.
It was proposed that diffusible ligands might select the suitable
conformation from the equilibrium of inactive and active GPCR conformations,
whereas in an induced-fit scenario the ligand would bind an inactive
conformation and induce a conformational change toward the active
conformation.
503
Rhodopsin with its activation
by retinal photoisomerization is likely to correspond to an induced-fit
scenario for initial events, whereas both scenarios are conceivable
for later activation phases as well as for GPCRs activated by diffusible
ligands.
4.2.2
Rhodopsin Activation
The steps
involved in rhodopsin activation are illustrated in Figures 24–26. In the rhodopsin
dark state, 11-cis-retinal is tightly bound as an
inverse agonist in its binding pocket and covalently fixed by the
RSBH+ to Lys296 on TM7. The positively charged RSBH+ is stabilized by a complex counterion
comprising negatively
charged residues Glu113 on TM3 and Glu181 on extracellular loop 2,
505,506
with the former functioning as the primary counterion (Figure 25A). The β-ionone ring
at the other end of
the retinal is ensconced in a hydrophobic pocket formed by aromatic
side chains, and the conjugated double-bond system linking the two
is tightly engaged by a constriction in the retinal binding site which
forces a negative 6-s-cis twist of the β-ionone
ring about the C6—C7 single bond and a twist about the C11=C12
double bond.
46,420,421
A twist about the C12—C13 single bond results from a steric
interaction between a proton at C10 and the methyl group at C20. In
addition to the pretwist of the C11=C12 double bond, the proximity
of the negatively charged Glu181, which was predicted from modeling,
507
reduces the bond order further to enable selective
isomerization around the C11=C12 double bond in the direction
shown in Figure 23. Following the gradual release
of the potential energy stored in the distorted retinal–protein
complex via Batho and the transient BSI, it is not until Lumi that
a displacement of the β-ionone ring is observed as a result
of the elongation of the retinal. The local perturbations of amino
acid side chains in Lumi increase slightly when compared with Batho,
but the protein structural changes are still limited to the residues
making up the retinal binding site and do not propagate to the cytoplasmic
surface.
146
FTIR studies using site-directed
infrared labels suggest that the first global movement of the protein
backbone is a small rotation of TM5 and TM6 which occurs upon Meta
I formation.
479
Solid-state NMR experiments
on Meta I and changes in electron density in TM6 on the side facing
retinal in the Meta I electron crystallography structure are consistent
with a slight motion of TM6
491
which can
be described as rotation, but not outward movement.
480
Also consistent with this, another rhodopsin-specific constraint,
the TM3/TM5 hydrogen bonding network including Glu122 and Trp126 on
TM3 and His211 on TM5, changes upon formation of Meta I as concluded
from FTIR data.
508
Straightening of the
retinal due to isomerization is thought to move the β-ionone
ring toward the region of Met207 to Phe212 on TM5 and thus driving
TM rearrangement.
480,509,510
Figure 26
Structural and functional changes in the activation pathway of
bovine Rho based on structural and complementary biophysical data
discussed in the text and ref (490) and described in Figures 23 and 24.
Until formation of Meta I, the RSBH+ remains protonated,
but structural changes in the RSB region occur. In the rhodopsin dark
state, RSBH+ interacts with the negatively charged Glu113
counterion (Figure 25A)
511−513
from which a hydrogen bonding network extends to Glu181 on extracellular
loop 2.
46
On the basis of both UV–vis
and Raman spectroscopy it was hypothesized that in Meta I Glu181 transfers
a proton to Glu113 and the RSBH+ switches counterions from
Glu113 to Glu181.
514
This view was later
modified on the basis of FTIR,
505,515
NMR,
516
and molecular dynamics simulations
517
whose data argue for a complex-counterion made up by Glu113
and Glu181, with both residues being deprotonated and giving the retinal
binding pocket a net negative charge.
505,516
In the complex-counterion
switch model, Glu113 functions as the primary counterion in the photointermediates
up to Lumi; in Meta I, the conformational changes in retinal lead
to a shift of the counterion from Glu113 to Glu181.
505,515,517
With the deprotonation
of the RSB, an equilibrium of Meta II states
is reached, in which the RSB nitrogen is deprotonated prior to the
occurrence of TM movements.
488,490
As a result of retinal
isomerization, the RSBH+ reorients (Figure 23), and its high pK
a (determined
experimentally to be above 16 in the dark state
518
and suggested to have contributions from the negatively
charged Glu181
505,506
) drops so that the proton dissociates
for the formation of the first Meta II state, Meta IIa. Concomitantly,
Glu113 becomes protonated, and a direct internal proton transfer from
the RSBH+ is likely to be the source of this proton.
519
According to FTIR studies,
490
RSBH+ deprotonation upon transition to Meta
IIa leads to a rearrangement of the TM1/2/7 network as seen by changes
of infrared bands assigned to Asp83 on TM2. In this water-mediated
hydrogen-bonding network, Asp83 (on TM2) links Asn55 (on TM1) with
the NPxxY(x)5,6F motif (TM7), and extends to Trp265 (on
TM6) adjacent to retinal’s β-ionone ring. Activating
conformational changes in the TM3/TM5 network occur in the Meta IIa
→ Meta IIb transition, as seen by changes of infrared bands
assigned to Glu122.
490
This hydrogen bonding
network links Glu122 and Trp126 (both on TM3) with His211 (on TM5)
which is in contact with the β-ionone ring of retinal. Retinal
movement toward TM5 exerts its full effect upon transition to Meta
IIb, reflected in a weakening of the hydrogen bond to Glu122.
508,520
In the Meta IIa → Meta IIb transition, changes in infrared
amide I marker bands indicate structural changes in the protein backbone.
490
According to Meta II structures, these structural
alterations include an elongation of TM5 by 1.5 to 2.5 helix turns
depending on the dark state reference structure (PDB ID: 1GZM/3C9L or 1U19, respectively)
and
a rotational tilt of the kinked TM6 where the TM rotation results
in the 6–7 Å outward movement of the cytoplasmic end of
TM6 (Figure 23).
426,429,433,435
Time-resolved EPR studies with a spin label sensor at position 227
on TM5, designed to detect TM6 movement, probed the Meta IIa →
Meta IIb transition for this event, although the sensor might have
detected movements of both TM5 and TM6.
488
TM6 rotation is facilitated by the breakage of the (D/E)RY
ionic
lock between TM3 and TM6 consisting of Glu134 and Arg135 (on TM3)
and Glu247 and Thr251 (on TM6). The cytoplasmic end of TM5 contains
two residues, Tyr223 and Lys231, which function as microswitches.
521
These residues face the lipid environment in
the dark state, but stabilize the Meta IIb state when the Arg135-Glu247
ionic lock has been disrupted and TM5 and TM6 rearrangements have
occurred (Figure 25C,D). By swiveling inward,
Tyr223 can hydrogen bond to Arg135 of the ionic lock, thereby linking
TM3 and TM5. Glu247 (on TM6) is freed in Meta IIb from Arg135 and
can form a salt bridge to Lys231 on TM5, thus stabilizing TM6 in the
outward position. Also in the transition to Meta II when TM6 moves,
the electrostatic interaction between Tyr306 and Phe313 becomes disrupted
allowing Tyr306 to swivel into the vacated space below Arg135 where
it becomes part of an extended water-mediated hydrogen bonding network
reaching from the retinal binding site via Ser298, Asp83, Asn302,
Met257, Tyr306, and Tyr223 to Arg135.
435
Studies with mutants of the NPxxY(x)5,6F motif suggested
dual roles for the NP and the Y(x)5,6F submotifs.
522
Whereas the first has a structural role related
to the hydrogen bonding network, the latter is involved in the interaction
with G protein. The crystal structure revealed how Tyr306 stabilizes
Arg135 and thus the cytoplasmic cleft for binding of the Gtα C-terminus.
In the structure of the Meta II·GαCT2
peptide complex
the backbone carbonyls of Cys347 and Lys345 (on GαCT2) form
hydrogen bonds to Arg135 and Gln312, respectively. The presence of
the GαCT2 peptide in the cytoplasmic G protein binding cleft
moves the cytoplasmic end of TM7 slightly toward the center, leading
to a somewhat narrower binding cleft. The proposed water-mediated
hydrogen bonding network provides an answer regarding how the retinal
binding site and the G protein binding site 30 Å apart are connected.
426,435,464,465
4.2.3
Retinal Channeling in Rhodopsin
The structure of bovine Ops* revealed that the retinal binding pocket
has two openings toward the lipid bilayer, one opening between TM1
and 7, the other between TM5 and 6, which form the two halves of the
channel leading to the retinal binding pocket (Figure 27). Using molecular docking
of retinal, a 3–12 Å
wide continuous channel through opsin with the retinal binding pocket
as the central part was found.
523,524
The openings to these
channels are lined with aromatic residues while the central part is
more polar. A major constriction of the channel is around Lys296 which
enforces a 90° elbow-like kink in the channel. Passage of that
restriction would be easier for the kinked 11-cis-retinal, whereas the more elongated
and rigid all-trans-retinal would require global conformational changes. A study with
rhodopsins containing mutations throughout the channel failed to correlate
specific opening(s) with retinal entry and exit.
525
However, the study suggested that the ease of retinal passage
through constrictions in the channel is not rate-limiting for rhodopsin
reconstitution or Meta II decay, but for RSB formation and hydrolysis,
respectively. Upon rhodopsin activation, when the helix bundle opens
up, bulk solvent molecules obtain access to the RSB,
526−528
most likely from the cytoplasmic side of rhodopsin for RSB hydrolysis
and retinal release.
526
This hypothetical
solvent channel appears to be quite narrow because only small nucleophiles
such as water and hydroxylamine have access to the RSB for retinal
hydrolysis.
526,529
Figure 27
Retinal channel in the
Ops*/Meta II conformation. (A) Meta II structure
(PDB ID: 3PXO) with the putative retinal channel indicated, rotated to face opening
one (red arrow and half channel indicated in red mesh) which is located
between TM1 and TM7, and (B) rotated to face opening two (green arrow
and half channel indicated in green mesh) which is located between
TM5 and TM6. Channels were determined using MOLE
577
on the Ops* structure (PDB ID: 3CAP). The Meta II structure was used for
the figure so that the retinal could be shown as it is absent in the
Ops* structure.
The size of the opening
of the retinal channel varies in different
Ops*, Meta II, and Meta II-like structures due to variations in side
chain rotamers. In the rhodopsin dark state with its compact helix
bundle, and presumably in the conformation of inactive opsin, which
is similar to dark state rhodopsin,
530
no
opening is observed and the 11-cis-retinal appears
to reside in a hermetically sealed retinal binding pocket (Figure 28A).
46,420,421
Opsin crystals in the Ops* conformation with an open channel therefore
allowed reconstitution of Meta II in soaking experiments with all-trans-retinal.
426
All-trans-retinal can bind tightly in its binding pocket in
Meta II with only slight adjustment of the surrounding amino acid
side chains. The binding pocket, however, is flexible enough to allow
retinal rotation along its long axis upon rhodopsin activation as
concluded from the Meta II crystal structures.
426,433
A limited data set of intramolecular distances obtained by solid-state
NMR experiments on Meta II is mostly consistent with the crystal structure.
A different degree of retinal rotation in the NMR experiment, however,
cannot be ruled out, but could be explained by the equilibrium of
Meta II substates.
174,426,509,510,531
Figure 28
Retinal binding site of bovine rhodopsin. (A) Crystal structure
of inactive dark state (PDB ID: 1U19) where 11-cis-retinal
is tightly bound deep in the protein with no openings of the retinal
binding site toward the lipidic environment. (B) In the active Ops*
(or Meta II) conformation the retinal channel allows all-trans-retinal access and
egress, and some detergents like β-d-octylglucoside, mimicking the all-trans-retinal
chromophore, to enter the retinal binding site. Shown is an overlay
of the two ligands in the retinal binding pocket as observed in the
crystal structures of Meta II (all-trans-retinal
depicted in yellow; PDB ID: 3PXO) and Ops* in complex with β-d-octylglucoside
(depicted in green/red; two rotamers of Lys296 are shown in red; PDB
ID: 4J4Q). Note
that the ring moieties of chromophore and detergent are oriented in
opposite directions. Whereas all-trans-retinal is
covalently linked by the RSB to Lys296, β-d-octylglucoside
is fixed in the ligand binding site by hydrogen bonding of its hydroxyl
groups to the opsin environment.
After uptake of 11-cis- or 9-cis-retinal by opsin (likely by the Ops* conformation
when openings
to the retinal channel are present) and RSB formation, a conformational
change yields inactive rhodopsin or isorhodopsin, respectively, both
of which generate the same photoproducts after photon absorption.
46,420,421,532
For the formation of artifical rhodopsin pigments, the retinal channel
in opsin is flexible enough to take up a wide variety of bulkier retinal
analogues, e.g., featuring larger or additional alkyl groups or alkyl
rings which prevent isomerization around the C11=C12 double
bond.
533
Recently, crystallographic evidence
was provided that the detergent, β-d-octylglucoside,
can bind within rhodopsin’s retinal binding pocket thus stabilizing
the Ops* conformation (Figure 28B).
432
A study on rhodopsin reconstitution from bovine
opsin and 11-cis-retinal in the presence of various
glucose- or maltose-containing detergents suggested that even some
maltoside detergents can enter the retinal channel, while the affinity
of various detergents for opsin was dependent on the alkyl chain length.
432
The varied hydrogen-bonding possibilities of
the glucose hydroxyl groups to opsin within the retinal binding pocket
is reminiscent of the proposed dynamic binding of odorants within
olfactory receptors.
534
With its lateral
ligand entry and its active GPCR conformation, Ops* was suggested
to be ideal for homology modeling of olfactory receptors binding olfactant
agonists.
432
4.3
Mechanistic
Variations in Other Rhodopsins
4.3.1
Color
Pigments
Rod and cone cells
are responsible for scotopic and photopic vision, i.e., vision under
low light and daylight conditions, respectively. Rods are characterized
by high sensitivity, slow response, slow dark adaptation, and a single
type of rhodopsin, whereas cones have low sensitivity, fast response,
fast dark adaptation, and several types of cone rhodopsins.
3,535
Some of the features of cone cells can be attributed to the cone
visual pigments, which absorb at different wavelengths [red (λmax = 560 nm), green
(λmax = 530 nm), and
blue (λmax = 420 nm) rhodopsins in humans] necessary
to discriminate colors for color vision. The vertebrate cone opsins
and rod rhodopsins (λmax = 500 nm) form a single
family of homologous proteins,
26
where
rod rhodopsins have evolved from cone visual pigments, being closer
to the short-wavelength cone opsins.
536,537
The extinction
coefficient and quantum yield and thus photosensitivity of bovine
Rho and chicken green pigment are comparable.
538
Cone rhodopsins are also thought to undergo a photoactivation
process similar to rod rhodopsin, with some variation for the long
wavelength pigments.
539,540
A difference, however, is a
shorter lifetime of the Meta II state, with parallel formation of
Meta III (a late nonproductive photointermediate involved in the decay
of the photoactivated state to the apoprotein opsin and retinal) from
Meta I (which is in equilibrium with Meta II, Figure 22).
538,541,542
Additionally, regeneration of cone opsins with 11-cis-retinal is faster than that
of rod rhodopsins.
538,543
The faster kinetics of cone opsins (by 1–2 orders of magnitude)
of the active state decay and rhodopsin regeneration are optimal and
necessary for the high light levels present during the daytime.
Site-directed mutagenesis studies identified amino acids at position
122 and 189 (Glu122 and Ile189 in bovine Rho, replaced by Gln/Ile
and Pro in cone opsins) to be responsible for this functional difference.
544,545
In bovine Rho, Glu122 is a part of a hydrogen bonding network with
His211 (TM3/5 network).
546
Glu122 and Ile189,
present in rhodopsin but not in cone pigments, also potentiate a more
efficient G protein activation compared with cone pigments.
542
The lower Gt activation capacity of cone pigments
and faster Meta II decay contribute to the lower photosensitivity
of cones compared with rods. Formation and decay of Meta II are related
to TM6 movements as outlined above, and EPR experiments could potentially
give insights into structural dynamics and conformational changes
of color pigments. Unfortunately, because of the difficulties in the
sample isolation and preparation of cone pigments, structural studies
on these pigments have lagged significantly behind those of bovine
Rho. From resonance Raman spectroscopy it is known that the chromophore
structure is similar between human green and red sensitive pigments,
with both being similar to that of rhodopsin.
547
From FTIR spectroscopy it was concluded that the protein
structure of monkey green and red pigments is clearly different from
that of rhodopsin, and that hydrogen-bonding networks differ between
green and red pigments.
145
It should be
noted that in human red and green color vision pigments Glu181 (in
bovine Rho) is replaced by a His residue that functions as a chloride
binding site,
548,549
suggesting that participation
of an anion is a prerequisite for seeing red light. Future structural
and modeling studies will be needed to further elucidate the molecular
mechanism(s) of color tuning and structural dynamics in cone pigments.
550
4.3.2
Bistable Rhodopsins and
Photoisomerases
The rhodopsins we have discussed so far are
stable in the dark,
but once retinal isomerization has occurred and metarhodopsin states
are reached, the eventual decay into opsin and all-trans-retinal takes place. These
rhodopsins are also called monostable
rhodopsins, and their opsins must be regenerated with new 11-cis-retinal produced
in the adjacent retinal pigment epithelium
cells.
551
Vertebrates and invertebrates
possess in addition to these monostable rhodopsins, a second type
of rhodopsins, the bistable rhodopsins,
552
which mostly couple to the G protein, Gq, to initiate phosphoinositol
signaling cascades. While this Gq signaling is typical for the majority
of bistable rhodopsins including those of squid and octopus, other
bistable rhodopsins have been shown to couple to other G proteins
such as Go.
3,552
A characteristic of bistable
rhodopsins is that the dark state and the active states are both thermally
stable; i.e., RSB hydrolysis does not occur.
553
Furthermore, a secondary absorption of a photon is utilized to photoregenerate
the dark state.
554
These opsins lack the
conserved RSBH+ stabilizing counterion found in monostable
rhodopsins (Glu113 on TM3 in bovine Rho). Instead, in squid rhodopsin
the RSBH+ forms a hydrogen bond to Asn87 or Tyr111 side
chains (the equivalent positions in bovine Rho are Gly89 on TM2 and
Glu113 on TM3).
422
In other bistable opsins,
position 113 is occupied by neutral amino acid residues such as Tyr,
Phe, or Met, which can explain why in contrast to bovine Rho, RSBH+ deprotonation
of the active state is not required.
555
The conserved Glu181 (bovine Rho numbering)
may serve as the sole negatively charged residue near the RSBH+, which is, however,
not close enough for direct interaction
with the RSBH+.
422
Molecular
evolutionary analysis implies that the counterion has been switched
from Glu181 to Glu113 during the evolution of vertebrate opsins and
that Glu181 serves as a counterion in bistable pigments.
3,552,556,557
Another distinctive feature of bistable pigments is that their opsins
can bind all-trans-retinal to form the pigment.
3,552
Rhodopsin from the Japanese flying squid, T. pacificus is the only invertebrate opsin
(and incidentally the only other
GPCR purified from native source apart from bovine rhodopsin) to have
its structure determined. This structural information has been instrumental
in understanding its chromophore–protein interactions in the
dark and bathorhodopsin states as well as in the artificial isorhodopsin
pigment (containing 9-cis-retinal).
422,423,558
The crystal structures of dark
state bovine Rho and squid rhodopsin exhibit similar features including
the presence of the disulfide bridge at the extracellular side of
the retinal pocket, a hydrophobic aromatic cage surrounding the retinal
and the presence of the (D/E)RY (ionic lock) and NPxxY(x)5,6F motifs.
422,423
As in bovine Rho, the retinal
is attached through a RSBH+ linkage to the conserved Lys
residue on TM7, but retinal itself takes a more relaxed configuration
in squid rhodopsin as opposed to the distortions observed in bovine
Rho.
422
The structure for squid bathorhodopsin
indicates that just as in the bovine Rho early intermediate states,
there is little structural change except for a change in the twist
of the retinal.
558
It has been postulated
from protein structure and sequence analysis that the extended TM5
and TM6 observed in the squid rhodopsin structure may explain the
selectivity of coupling to Gq proteins.
422,559
Squid rhodopsin contains an additional C-terminal domain that might
be involved in G protein binding, but it was not structurally characterized
as it was necessary to proteolytically remove this domain comprising
the last 90 amino acids for crystallization.
422,423
Another reason for the functional difference may be the extent of
TM6 movement upon photoactivation of bistable rhodopsins as proposed
from site-directed fluorescence labeling measurements. A comparison
of bovine Rho and parapinopsin, a bistable nonvisual Gt-coupled rhodopsin,
revealed much smaller light-induced TM5 and TM6 movement for parapinopsin
which correlates with its reduced capability to activate G protein.
560
Of further interest among Gq-coupled bistable
rhodopsins is melanopsin, which has been identified in various vertebrates.
552,561
In mammals, melanopsin is localized in intrinsically photosensitive
retinal ganglion cells and is involved in nonvisual functions, including
photoentrainment of the circadian clock, pupillary light reflex, and
sleep.
4
Another, more divergent grouping
of opsins are the photoisomerases,
which function to bind all-trans retinoids and photoisomerize
them to their 11-cis-forms.
410
Structural information remains elusive for these opsins. The best
characterized of these photoisomerases are retinochrome from molluscan
species and the mammalian retinal G-protein-coupled receptor (RGR).
97,562,563
In mollusks, retinochrome functions
to provide the 11-cis-retinal to newly synthesized
apoprotein opsin, whereas in mammals, RGR appears to play more of
a regulatory role in the mobilization of all-trans-retinyl esters into the retinoid
cycle.
562,564
These photoisomerases lack the NPxxY(x)5,6Y motif and
thus may be deficient in G protein coupling.
3
A close relative is peropsin from mammalian retinal pigment epithelium
(RPE) cells. Peropsin shows photoisomerase activity, but contains
the (D/E)RY and NPxxY(x)5,6Y motifs and therefore may couple
to G protein and be involved in activation of signaling cascades.
565
5
Conclusions
and Perspective
In the past, the impact of research on both
microbial and animal
rhodopsins propagated far beyond the boundaries of the retinal-binding
protein field. It will suffice to give just a few of the most striking
examples. Structural work on BR and Rho has greatly contributed to
our understanding of the structural principles of membrane proteins
in general. The detailed understanding of the proton transport mechanism
by BR has been extremely useful to the bioenergetics community, who
extended these principles to important systems such as cytochrome
oxidases and ATPases. The great structural and mechanistic advances
in the understanding of visual signal transduction significantly enriched
the GPCR field, and for many years bovine Rho served (and continues
to serve) as a model GPCR. More recently, the discovery of proteorhodopsins
gave a strong push to the field of metagenomics, while the discovery
of channelrhodopsins gave birth to optogenetics. Finally, rhodopsins
have been serving as testing grounds for many cutting-edge biophysical
techniques, aiding, for example, in the development of time-resolved
and low-temperature FTIR spectroscopy, ultrafast spectroscopy, advanced
Raman techniques, new methods for 2-D and 3-D crystallization of membrane
proteins, protein solid-state NMR, and high-field EPR, including site-directed
spin-labeling techniques.
But what are the next exciting steps
in rhodopsin research? Expanding
on these recent trends, we can predict many more interesting developments
without being too speculative. Judging from the large number of new
rhodopsin variants unearthed by genomic and metagenomic sequencing,
new interesting functions of retinal proteins will continue to be
discovered. This applies to both microbial and animal rhodopsins,
and will lead to new breakthroughs in understanding microbial, invertebrate,
and vertebrate physiology and evolution. These new functional variants
of rhodopsins may also find use in optogenetics, enriching its arsenal
of tools. New advanced techniques of structural biology and biophysics
will be applied to these new rhodopsins leading to new insights, capitalizing
on such emerging methods as, for example, high-speed AFM, ultrafast
time-resolved crystallography, structural mass spectroscopy, and dynamic
nuclear polarization NMR.
From the point of view of structural
biology, the next challenging
frontier in the field will be to understand the mechanisms of rhodopsin–protein
interactions. While structural methods have enjoyed great success
in determining structures of isolated rhodopsins and their binding
partners, structures of inactive and activated protein complexes,
especially those of membrane and soluble proteins, remain elusive.
This is especially important for visual rhodopsins with their multiple
interacting partners, and bears on the entirety of the GPCR field,
contributing to a better understanding of GPCR signaling cascade,
activation, and regulation.