Application of this label-ling scheme in combination with cross polarization and magic angle sample spinning CPMAS NMR techniques provides well resolved spectra that exhibit chemical shi
Trang 1after the Schiff base reprotonation of bacteriorhodopsin 15
N CPMAS NMR of D85N/ T170C membranes
A James Mason1, George J Turner2and Clemens Glaubitz1
1 Centre for Biomolecular Magnetic Resonance and Institut fu¨r Biophysikalische Chemie, J.W Goethe Universita¨t, Frankfurt, Germany
2 Department of Chemistry and Biochemistry, Seton Hall University, South Orange, NJ, USA
Bacteriorhodopsin [1] is a 26 kDa seven
transmem-brane helix protein (7TM) found in the extremely
halo-philic archeaon Halobacterium salinarium [2] The
proton pumping ability of this protein is conferred by
the prosthetic retinal attached via a Schiff base to
Lys216 The light-induced isomerization from all-trans
to 13-cis causes the release of a proton from the Schiff
base, which in turn causes a proton to be released at
the extracellular surface The reaction is cyclic and the
photocycle has been characterized spectroscopically
where a series of photointermediates have been
deter-mined:
bR570! K590! L550! M412! N560! O640! bR570
The photocycle can be divided into two phases The
first phase is the K-L-M1-M2-M2¢ sequence, where a
proton is donated from the Schiff base to Asp85 and
another proton is released to the extracellular surface, and the second is the N-N¢-O-bR sequence, where the Schiff base is reprotonated from Asp96 Asp96 is itself reprotonated from the cytoplasmic surface and a pro-ton is transferred from Asp85 to the propro-ton release site Analysis of the photomechanism has been revolu-tionized by the production of a family of high resolu-tion X-ray diffracresolu-tion structures [3,4] The structures and structural changes assigned to the intermediates
of reprotonation reactions remain an area of debate,
as described below
N-state
In the early intermediates of the reprotonation phase, when the protein is in the late M- and N-state, contrasting measurements of the movements in the
Keywords
bacteriorhodopsin; solid-state NMR; N-state;
O-state
Correspondence
C Glaubitz, Institut fu¨r Biophysikalische
Chemie, Centre for Biomolecular Magnetic
Resonance, J.W Goethe Universita¨t,
Marie-Curie Str 9, D-60439 Frankfurt, Germany
Fax: +49 69798 29929
Tel: +49 69798 29927
E-mail: glaubitz@chemie.uni-frankfurt.de
(Received 19 October 2004, revised 10
February 2005, accepted 28 February 2005)
doi:10.1111/j.1742-4658.2005.04633.x
bR, N-like and O-like intermediate states of [15N]methionine-labelled wild type and D85N⁄ T170C bacteriorhodopsin were accumulated in native membranes by controlling the pH of the preparations 15N cross polariza-tion and magic angle sample spinning (CPMAS) NMR spectroscopy allowed resolution of seven out of nine resonances in the bR-state It was possible to assign some of the observed resonances by using13C⁄15N rota-tional echo double resonance (REDOR) NMR and Mn2+ quenching as well as D2O exchange, which helps to identify conformational changes after the bacteriorhodopsin Schiff base reprotonation The significant differences
in chemical shifts and linewidths detected for some of the resonances in N- and O-like samples indicate changes in conformation, structural hetero-geneity or altered molecular dynamics in parts of the protein
Abbreviations
7TM, seven transmembrane helix protein; CPMAS, cross polarization and magic angle sample spinning; DA, dark adapted; LA, light adapted; REDOR, rotational echo double resonance.
Trang 2cytoplasmic half of the protein have been obtained.
Large scale motions have been observed, particularly
in helices E, F and G, and the EF loop, by a variety
of techniques, including electron diffraction [5–7],
X-ray diffraction in projection of purple membranes
[8–10] and Electron Spin Resonance (ESR) spin
label-ling [11,12] An early X-ray diffraction study of F171C
membranes [13], at 7 A˚ resolution, observed fairly
small structural changes with the largest change
invol-ving a movement of helix F and some small
move-ments of helices B and G, whilst two electron
diffraction studies of 2D crystals observed rather large
structural changes in the cytoplasmic region The
structure of a ‘cytoplasmically open’ conformation
found in the D96G⁄ F171C ⁄ F219L triple mutant [6]
revealed displacements of the ends of helices F and G
of 3.5 and 2 A˚, respectively, while the structure of the
N intermediate found in F219L membranes [7] showed
that both helices E and F are displaced by some 3 A˚,
with helix G again moving slightly These results were
in contrast, however, with high resolution structures of
the MN- and N¢-states [14,15], produced from 3D
crys-tals, which do not show the expected tilts or rotations
[16] It has been suggested that, within 3D crystals, the
crystal lattice resists any increase in the unit cell
dimensions preventing such conformational changes
O-state
The bacteriorhodopsin O-state is the least well resolved
conformer of the reprotonation mechanism The most
recent analysis relies on the mutants D85S and
D85S⁄ F219L as O-state models [17] The structures of
D85S and D85S⁄ F219L, at 2.25 and 2.0 A˚ resolution
respectively, reveal important differences between the
bR- and O-like states [17] The most notable
differ-ences are in the extracellular half of the protein and in
the loop regions, particularly the BC, DE and EF
loops A slight repackaging of the transmembrane
ces in the extracellular side results in tilting of the
heli-ces A, B, C and D by approximately 3 and, more
noticeably, helix E by 6.9 relative to the bR-state
The protonation state of Asp85 plays a central role
in the conformational changes and linked proton
movements during the transitions between the M-, N-,
O- and bR-states As illustrated in the discussion of
the O-state models, mutants of Asp85 have been useful
in the study of the reprotonation mechanism [17]
Replacement of Asp85 with asparagine (D85N) allows
study of the intermediate state conformations in which
Asp85 is normally protonated The bR mutant D85N
exists as three spectrally distinct species in a
pH-dependent equilibrium [18] Transitions between these
species regulate the pKa values of Asp96 and the Schiff base in a manner consistent with that observed in the reprotonation phase of the wild-type protein At low-to-neutral pH an O-like species predominates (kmax¼
615 nm), whereas at higher pH values increasing levels
of an N- (kmax¼ 570 nm) and an M-like species (kmax¼ 410 nm) appear [18] D85N, and second site mutants thereof, can be used to isolate the conforma-tional transitions of the reprotonation phase of proton pumping
In this study we exploited the pH-dependent transi-tions of the D85N⁄ T170C double mutant to probe the structures of N- and O-like states Our mutant D85N⁄ T170C behaves similarly to D85N [18] How-ever, the pKaof M accumulation is raised by the addi-tional cysteine mutation and hence, although some M-state remains, it is less populated making this sys-tem more suitable to access N-and O-like states [19]
We applied residue-specific 15N labelling to all methionines in the wild type purple membrane (Fig 1) and D85N⁄ T170C membrane to evaluate the conform-ational flexibility of transmembrane helices in the bR,
163
32
20
54 56
60
68
209
208 118
145 117
Fig 1 The three-dimensional structure of bacteriorhodopsin is shown indicating the positions of the nine15N-labelled methionine residues, present in each sample Three 13 C-labelled residues, [ 13 C1]Ile117, [ 13 C1]Phe54 and [ 13 C1]Phe208 that form spin pairs with labelled methionine residues 118, 56 and 209, present in two separate samples prepared for REDOR experiments are also shown RASMOL was used with coordinates 1c3w [14] from the Pro-tein Data Bank.
Trang 3N- and O-like states Methionines are found at
resi-dues 20, 32, 56, 60, 68, 118, 145, 163 and 209, with
only 68 and 163 located in loops and all others located
in helices A, B, D, E and G Application of this
label-ling scheme in combination with cross polarization
and magic angle sample spinning (CPMAS) NMR
techniques provides well resolved spectra that exhibit
chemical shift differences and resonance line
broaden-ing in N- and O-like states compared with bR An
assignment based on rotational echo double resonance
(REDOR) experiments in conjunction with double
labelling together with Mn2+-induced quenching and
D2O exchange of some of the observed resonances
allows a more detailed view of conformational changes
and motions within these mutants, which serve as
models for the N- and O-like states
Results
The 15N CPMAS spectrum of [15N]Met purple
mem-brane (Fig 2A) allows the resolution of seven
reso-nances out of nine labelled residues Contribution
from the 15N natural abundance (0.37%) from a
26 kDa protein is calculated to be equivalent to 0.9
15N nuclei per nine labelled residues, and is spread
over the full amide spectral region Therefore, it can
be considered to be negligible in contrast to 13C
label-ling Compared to the wild type spectrum and to
each other, both N- and O-like preparations of
D85N⁄ T170C (Fig 2B,C) show marked differences
Both the N-like and O-like state spectra are
character-ized by a number of well resolved resonances In
gen-eral, in the N-like state the resonances appear slightly
broader and there is a greater degree of overlap In the
O-like state, the number of clearly resolved resonances
is reduced with O2 and O4 appearing only as
shoul-ders to the intense O3 resonance A summary of
chem-ical shifts and linewidths resulting from spectral
deconvolution is given in Table 1
Resonances for Met20 (bR7) and Met145 (bR6)
have been assigned previously in bR using the
single-site mutations M20V and M145H [20,21]
(supplement-ary Fig S1), whilst Met32 (bR2) was tentatively
assigned previously as a shoulder resonance using the
exchange [20] Here, we sought to assign the remaining
resolved methionine resonances by making use of
Mn2+-induced line broadening and deuterium
exchange of residues located close to the membrane
surface [22] in addition to REDOR on [15N]Met⁄
[13C1]Ile or [15N]Met⁄ [13C1]Phe membrane
prepara-tions REDOR as an assignment technique has been
used previously as a selective filter [23] and to assign
specific proline residues in bacteriorhodopsin [24] Knowing the primary structure of bR allows the generation of unique 15N-13C1 pairs by colabelling
A
B
C
Fig 2. 15N CPMAS spectra obtained for [15N]methionine-labelled (A) purple membranes (bR) in pH 6 buffered H2O, (B) D85N ⁄ T170C membranes at pH 10 (N-like), and (C) D85N ⁄ T170C at pH 6 (O-like) Resonances are labelled from large to small chemical shifts (Table 1) Resonance assignment is discussed in the text Spectra were acquired at 60.82 MHz 15 N Larmor frequency, 253 K and
8 kHz sample rotation rate Spectra were deconvoluted using PEAK-FIT to obtain the linewidths of overlapping resonances.
Trang 4[15N]Met samples with the upstream residue enriched
with13C1 These 15N-13C1 pairs have a strong dipolar
coupling which can be used to selectively dephase and
therefore assign the related15N-resonances
Applying the REDOR technique to the [15N]Met⁄
[13C1]Ile bR sample using a short dephasing time of
1.4 ms caused a significant reduction in the intensity of
resonance bR1 (Fig 3) Other resonances were
unaffec-ted within the limits of the signal-to-noise ratio of the
spectrum The observed dephasing was mainly due to
the strong dipolar coupling between directly bonded
[13C1]Ile117 and [15N]Met118, which are separated by
only 1.3 A˚ (Fig 1) A small additional contribution
could also arise from [13C1]Ile119 which, in the 3D
crys-tal structure of bR (1c3w [14]), is 4.7 A˚ away
There-fore, bR1 can be assigned to Met118 in bR The only
other spin pair within a 6 A˚ radius of any [13C1]Ile in
this sample would be [15N]Met56 and [13C1]Ile52, which are approximately 4.2 A˚ apart This weak dipolar coup-ling would cause less signal decay at the short dephas-ing time used here We expect this signal reduction to
be below the noise level of this experiment and consider the decay at bR3 as not significant at this stage
Further resonances were assigned in [15N]Met⁄ [13C1]Phe bR, N- and O-like preparations Again using
a short dephasing time of 1.4 ms, a significant signal reduction of bR5 was observed (Fig 4A) in bR mem-branes, while other resonances were unaffected within the noise level of our data In N- and O-like prepara-tions, dephasing was observed for resonances N3 (Fig 4B) and O3 (Fig 4C), respectively The observed dephasing in each case was due to directly bonded [13C1]Phe208-[15N]Met209 Therefore, bR5, N3 and O3 were assigned to Met209 in bR, N-like and O-like states, respectively The N-like preparation was suspen-ded in buffer at pH 10 containing 40 lm Mn2+ to remove the signal from surface-exposed residues (N2, see below) and allow a clearer observation of the dephasing effect on N3 Some additional signal reduc-tion of N4 in the N-like state may be attributed to dephasing of [15N]Met56 by [13C1]Phe54
Extending the REDOR dephasing time from 1.4 to
16 ms causes signal decay for a further resonance, bR3
Table 1 Summary of chemical shift and full width at half height
(FWHH) for all [ 15 N]Met resonances shown in Fig 2 Linewidths
were obtained by deconvolution using PEAKFIT
Resonance
Peak and assignment
d ISO
(p.p.m.)
FWHH (p.p.m.)
bR2⁄ Met32 a
bR4⁄ Met60 b 122.5 0.34
N-like (D85N ⁄ T170C,
pH 10)
N1c (Met118)
O-like (D85N ⁄ T170C,
pH 6)
O1b (Met118)
O6 ⁄ Met20 d 117.5 0.50
a The shoulder down field of bR2 is best approximated by a
Gaus-sian with d ISO ¼ 125.1 p.p.m., FWHH ¼ 0.822 p.p.m b The
shoul-der between bR3 and bR4 is best approximated by a Gaussian with
dISO¼ 123.0 p.p.m., FWHH ¼ 0.67ppm c The best deconvolution
of N1 has been achieved with at least four Lorentzians d The small
peak down field of O6 is best approximated by a Lorentzian with
d ISO ¼ 118.3 p.p.m., FWHH ¼ 0.71 p.p.m.
Fig 3 Dephased (S) and nondephased (S0) 15 N-detected 13 C ⁄ 15 N
CP REDOR spectra of [ 15 N]Met ⁄ [ 13 C1]Ile purple membranes A sig-nificant signal decay of resonance bR1 is observed for a short REDOR dephasing time of 1.4 ms The only directly coupled
15 N– 13 C spin pair is [ 15 N]Met118 ⁄ [ 13 C1]Ile117, which allows the assignment of bR1 to Met118 Intensity variations of the other sig-nals are mainly due to noise and are discussed in the text Spectra were acquired at 40.52 MHz 15 N Larmor frequency, 253 K and
5 kHz sample rotation rate.
Trang 5(Fig 5) In the 1c3w 3D crystal structure of bR, [13C1]Phe54 is located approximately 3.2 A˚ from [15N]Met56 No other15N labels are within 6 A˚ of any [13C1]Phe except [15N]Met209 Therefore, bR3 is assigned to Met56 The sample was suspended in buf-fer containing 40 lm Mn2+to remove the signal from surface-exposed residues (bR4, see below) and allow a clearer observation of the dephasing of bR3 and bR5 For technical reasons, all REDOR experiments pre-sented here were performed at 40.54 MHz 15N Larmor frequency and at a 5 kHz sample rotation rate (Figs 3–6), compared to 60.82 MHz and 8 kHz for the cross polarization spectra presented in Fig 2 and dis-cussed earlier Therefore, a poorer resolution was achieved and bR2 was not clearly resolved under these conditions In addition, the different line shapes and peak intensities obtained by cross polarization and REDOR are caused by different spin relaxation due to the long delays between rf pulses in the REDOR experiment
Mn2+-induced paramagnetic line broadening of NMR signals has been described previously in
Fig 4 Dephased (S) and nondephased (S0) 15 N-detected 13 C ⁄ 15 N
CP REDOR spectra of [ 15 N]Met⁄ [ 13 C 1 ]Phe (A) purple membranes
and D85N ⁄ T170C membranes, at (B) pH 10 (N-like) in presence of
40 l M Mn 2+ , and (C) pH 6 (O-like) Resonances bR5, N3 and O3
show strong decays at 1.4 ms REDOR dephasing time The only
directly coupled 15N-13C spin pair is [15N]Met209 ⁄ [ 13
C 1 ]Phe208 which allows assignment of bR5, N3 and O3 to Met209 Spectra
were acquired at 40.52 MHz 15 N Larmor frequency, 253 K and
5 kHz sample rotation rate.
Fig 5 15 N CPMAS and 15 N-detected 13 C ⁄ 15 N CP REDOR spectra
of [15N]Met ⁄ [ 13
C 1 ]Phe purple membranes suspended in 40 l M
Mn 2+ pH 6 buffer REDOR dephasing was applied for 16 ms which completely dephases the signal from resonance bR5 but also shows signal decay for bR3 In this sample, only [13C 1 ]Phe208 and [ 15 N]Met209 are directly coupled but [ 13 C1]Phe54 is within 3.2 A ˚ of [ 15 N]Met56 causing a slower dephasing due to a weaker dipolar coupling Therefore bR3 is assigned to Met56 Spectra were acquired at 40.52 MHz 15 N Larmor frequency, 253 K and 5 kHz sample rotation rate.
Trang 6bacteriorhodopsin [22], where it was used to assign
resi-dues close to the membrane surface Strong dipole–
dipole interactions induce accelerated spin relaxation
and a concomitant line broadening in excess of 100 Hz,
such that NMR signals from residues close to the
mem-brane surface are suppressed in the CPMAS spectra
Due to their location close to the membrane surface,
signals from Met32, 60, 68 and 163 are expected to
broaden upon the addition of Mn2+ In the bR-state a significant reduction of intensity is observed for bR4 (Fig 6A) However, previous D2O exchange experi-ments [20] which remove signals from exchangeable residues Met32, 68 and 163 [25] did not show an effect
on bR4, which indicates that bR4 is Met60 (Fig 6A) In the N-like state (Fig 6B) a significant reduction of intensity was only seen in the region around N2
15N CPMAS spectra acquired after the incubation
of [15N]Met membranes in D2O reveal solvent-exposed residues due to a reduction in cross polarization by exchanging the amide proton with a deuteron In the N-like state, N2 is effectively removed by deuterium exchange (Fig 7A) as is O2 in the O-like spectrum (Fig 7B) The observed effect on resonance N2 is con-sistent with the detected quenching in the presence of
Mn2+ discussed earlier Therefore this signal must arise from a solvent-accessible residue close to the membrane surface (Met32, 68, 163) Differences between N-like spectra affected by Mn2+ quenching (Fig 6B) and deuterium exchange (Fig 7A) could be caused by Met60, but have not been observed This would suggest that the Met60 resonance is of low intensity and⁄ or largely obscured by the intense reson-ance assigned to Met209 in the N-like state Other resonances are unaffected by deuterium exchange with the exception of O6
In the N-like state spectrum, resonance N7 occurs at the same chemical shift and with similar intensity as bR7 (Met20), but is slightly broader (Table 1) The intensity of N7 is related to that of resonance O6, which is shifted by only )0.5 p.p.m When the mem-branes are suspended in D2O in an O-like state, O6 splits into two resonances O6a and O6b (Fig 7B) The additional resonance O6a occurs at 118 p.p.m as bR7 and N7, while O6b has the same chemical shift as O6 The appearance of resonance O6a appears to cause a signal reduction of O6 An explanation would be a change in equilibrium between the N and O-like states caused by resuspending the samples in D2O with a subsequent change in pH This is supported by the detection of a blue shift of kmax by 7 nm in the absorption spectra of O-like samples in D2O (supple-mentary Fig S2) compared to preparation in H2O N-state samples are shifted by only 4 nm These obser-vations provide further evidence that N7 and O6 cor-respond to the same residue
Discussion
The 15N CPMAS spectrum of [15N]Met purple mem-branes (Fig 2A) allows the resolution of seven reso-nances, which correspond to the seven methionine
Fig 6 Comparison of 15 N CPMAS spectra in the absence or
pres-ence of 40 l M Mn 2+ (A) [ 15 N]Met purple (bR) membranes at pH 6
with (dotted line) and without (solid line) the addition of 40 l M
Mn2+ Resonance bR4 is most affected, which must arise from a
residue close enough to the membrane surface to be broadened in
the presence of Mn 2+ ions Spectra were acquired at 40.52 MHz
15
N Larmor frequency, 253 K and 5 kHz sample rotation rate (B)
[ 15 N]Met D85N ⁄ T170C membranes in water buffered at pH 10
(solid line, top) and 40 l M Mn 2+ solution also buffered at pH 10
(dotted line, top) The reduction in intensity is due to the
broaden-ing of signals resultbroaden-ing from surface-accessible residues Spectra
were acquired at 60.82 MHz 15 N Larmor frequency, 253 K and
8 kHz sample rotation rate.
Trang 7resonances located in transmembrane helices A, B, D,
E and G (bR1–bR7 represent Met118, 32, 56, 60, 209,
145 and 20) The linewidth of the resolved resonances
ranges from 0.37 to 2.1 p.p.m The obtained spectral
resolution was better than in previously published work
[20], which is probably due to the use of a higher
mag-netic field and faster sample spinning Spectra in Fig 2
were deconvoluted with the minimum number of
Gaussian or Lorentzian peaks required to minimize v2
Possible error sources are limited signal-to-noise
(1 : 20–1 : 40 from 3–4 mg sample) as well as a small
amount of potential isotope scrambling which might
account for some background signal Here, no direct
contributions from loop resonances Met68 and Met163
were detected However, deconvolution of the spectrum
in Fig 2A hints at additional signal contributions to
the shoulders seen downfield of bR2 (Met32) and bR4
(Met60) The reduced intensity and line broadening of
these loop residues has been proposed to be due to
fluctuating motions that interfere with the line
narrow-ing processes of MAS or heteronuclear1H decoupling
during acquisition [22] The reduced spectral intensities
of [15N]Met68 and 163 have been also confirmed by
deuterium exchange experiments [20]
Spectra of the N-like (Fig 2B) and O-like (Fig 2C)
states show remarkable differences in line shape and
chemical shift when compared to the ground state
(Fig 2A) Before discussing the potential meaning of
those changes, we need to assess whether they arise
from M-, N- and O-state equilibriums or from clean
intermediates We have chosen the D85N⁄ T170C dou-ble mutant, because N- and O-like states can be popu-lated by controlling the pH while the M-state is much reduced compared to the well characterized D85N bacteriorhodopsin mutant At pH 6, D85N contains
95% O-like state and at pH 10 5% O, 20% N and 75% M [18] By introducing an additional T170C mutation, the pKa of M accumulation is raised as shown in Fig 8B The reason is that the M–N trans-ition is coupled to deprotonation of D96 and protona-tion of the Schiff base For example, in the 3D structure of the N-state [7] T170 faces the cytoplasmic channel at the level of D96 Therefore, a cysteine sub-stitution would alter the hydrophobic pocket and the
pKa of D96 and so shift the M–N transition towards
N By comparing the singular value decomposition (SVD) analysis performed on D85N [18] with our data (Fig 8B), we estimate the contribution of M-state to our sample at pH 10 to be not more than 30% Opposite to M–N, the N–O transition is coupled to the protonation of D96 and to deprotonation of groups at the cytoplasmic surface Therefore, at low
pH, D96 will be protonated and we obtain a sample mainly in O-state as shown for D85N (no M- and very little N-contribution) The O-state shows a characteris-tic absorbance found at 604 nm Raising the pH here also increases contributions from M and N It is known that the N- and O-states have different absorp-tion maxima at kmax 604 nm and 586 nm, respectively (Fig 8A) The N-state extinction coefficient is lower
Fig 7 Comparison of 15 N CPMAS spectra
of [15N]Met D85N ⁄ T170C membranes in water and D2O (A) Buffered at pH 10 (N-like state), (B) buffered at pH 6 (O-like state) The spectral subtractions reveal the intensity of the resonances resulting from exchangeable residues (N2, O2) Spectra were acquired at 60.82 MHz15N Larmor frequency, 253 K and 8 kHz sample rota-tion rate.
Trang 8(70%) than in the O-state Raising the pH from 6 to
10 accumulates a small M-state population but mainly
N-state, which is first seen as a signal reduction of the
O-state resonance The question is now to what extent
both the N- and O-states are mixed at the
experimen-tal conditions we have chosen for our NMR
experi-ments In addition to the fact that the absorption
maxima in Fig 8A are clearly separated for the
N- and O-state, our15N CPMAS spectra in Fig 2 also
show that both states are not significantly mixed The
sharp resonance from Met20 (N7 & O6) is in both
cases very well resolved and appears at 118.1 p.p.m
(N) and 117.5 p.p.m (O) Both lines are only
0.5 p.p.m wide and would be present simultaneously if
samples contained a significant N⁄ O mixture, which is
not the case There is no contribution from O in the
N-like sample (pH 10) The only hint for an N-state
contribution at pH 6 (O-like state) is a small resonance
at 118.3 p.p.m Our deuterium exchange data (Fig 7B)
have shown that a resonance at 118.0 p.p.m occurs
when the N–O equilibrium is shifted towards N We cannot exclude at this point that the resonance at 118.3ppm also arises from N in which case its contri-bution is estimated to 15% based on the fitted peak size Concluding, we can comment that our samples at
pH 6 are mainly found in a clean O-state with only a small potential contribution of 15% N and no M-state
At pH 10 we find an approximately 70 : 30 N-like⁄ M-like mixture but no O-like state Therefore, our spectra are dominated by N- or O-like states, which allows us to discuss the nature of the changes in chem-ical shift and line shape for each individual resonance
in more detail
The resonance bR7 assigned to Met20 in bR [20] is located upfield and well separated from the other peaks (Fig 2A) In the N-like state a resonance N7 appears with identical chemical shift but slightly broadened by 0.2 p.p.m and separated from all other resonances by 2 p.p.m Both bR7 and N7 are unaffec-ted by Mn2+-induced line broadening and D2O exchange Therefore it seems reasonable to assume that resonance N7 is also caused by [15N]Met20 In the O-like state, resonance O6 appears 0.5 p.p.m upfield of bR7 and N7 and is separated by 2.4 p.p.m from other residues Deuterium exchange suggests, as discussed in the results section, that O6 and N7 belong to the same residue, probably Met20 This would mean that Met20 has the same chemical shift in bR- and N-like states and changes only by 0.5 p.p.m in the O-like state Therefore
it is likely that it occupies the same conformation in bR- and N-like states but may experience a subtle change in conformation or an alteration in local hydro-gen bonding on conversion to the O-like state
Further up helix A, Met32 is assigned to bR2 [21] in
bR The fate of this residue on conversion to the N-like state is uncertain as a peak N2 appearing at a similar chemical shift is of a much greater intensity The intensity of resonance N2 is reduced by adding
Mn2+ (Fig 6B) and by deuterium exchange (Fig 7A) which points towards a solvent-accessible residue close
to the membrane surface such as Met32, 68 or 163 Met32 is the only helical resonance that is exchange-able [20,25] and as discussed earlier, Met68 and 163 are difficult to detect in the [15N]Met spectrum of bR
A stronger contribution in the N-like spectrum would only be expected if either loops EF or BC show much reduced molecular motions, which interfere less with the NMR experiments However, this is currently unknown and we cannot safely discriminate between Met32, 68 and 163 Resonance O2 in the O-like state occurs at the same chemical shift as N2 but appears broader with reduced intensity As for N2, deuterium exchange indicates contributions from residues Met32,
Fig 8 UV ⁄ vis spectra obtained for D85N ⁄ T170C membranes
puri-fied by sucrose density gradient centrifugation at 38% (w ⁄ w)
sucrose at different pH values (A) The N-like state contains some
M-state contribution, which is, however, much reduced compared
to D85N (B) and can be estimated to 30% based on the SVD
analysis for D85N D85N analysis results and data were taken from
[18].
Trang 968 or 163 (Fig 7B) Whether the observed line
broad-ening is of homogeneous or nonhomogeneous nature
i.e caused by altered molecular motions on the NMR
time scale or by conformational heterogeneity compared
to the bR state, cannot be concluded from these data
Met56, located on helix B, gives rise to resonance
bR3, as shown by the REDOR experiments discussed
earlier A similar assignment in the N-like and O-like
states was not possible, as experiments with longer
dephasing times were hampered by poor sensitivity
However, our data can be used to limit the number of
possibilities In the N-like state, resonance N4 is not
affected by Mn2+ quenching (Fig 6B) nor deuterium
exchange (Fig 7A) Therefore, Met32, 60, 68 and 163
can be ruled out, assuming that solvent accessibility
and location close to the membrane surface does not
change compared to our wild type samples
Further-more, Met20 and Met209 have been already assigned
to N7 and N3, which leaves us with Met56, 118 and
145 In bR Met118 had been assigned to resonance
bR1 which is 6 p.p.m downfield of N4 It is therefore
considered unlikely that N4 is caused by Met118
However, only a 0.8 p.p.m downfield shift for Met145
or 2.4 p.p.m upfield shift for Met56 would be
required Further down helix B and close to the
extra-cellular surface is Met60 which has been assigned to
bR4 A clear resonance cannot be assigned to Met60
in the N-like state but Mn2+induced line broadening
shows that it is-likely concealed under the intensity N3
assigned to Met209 In the O-like state, the number of
resonances is reduced They appear at different
chem-ical shifts and are generally broadened compared to
bR Especially spectral components O2 and O4
under-lying O3 appear as broad shoulder resonances The
intensities around O2 must belong to deuterium
exchangeable residues (Fig 7B) while O3 has been
assigned to Met209 The small but broad shoulder
resonance O4 (Figs 2C and 7B) could correspond to
residual intensity due to Met56 and⁄ or Met60
The observed broadening of lines could be of
homo-geneous or nonhomohomo-geneous nature Interestingly,
recent research suggests that the cytoplasmic half of
helix B, where both Met56 and Met60 are located,
adopts motional fluctuations after deprotonation of
the Schiff base [26] As discussed previously for loop
resonances Met63 and Met163 these fluctuations might
cause interference with proton decoupling or magic
angle sample spinning [22] Whilst the presence of a
signal (N4, N5 or N6) that can be tentatively assigned
to Met56 in the N-like state suggests that the
fluctu-ating motions proposed in the cytoplasmic half of helix
B above Pro50 do not affect the whole helix in the
N-state, the absence of an intense signal from Met56
or Met60 in the O-like state suggests that such fluctu-ating motions are propagated down helix B as far as Met56 or Met60 in this later stage of the photocycle
In the bR state Met118 is the most downfield and intense of the methionine resonances Met118 is observed as a sharp peak bR1 In the N-like state a broad resonance N1 with the same chemical shift as bR1 occurs Deconvolution of N1 indicates at least four identifiable resonances (Table 1) Because of sim-ilar chemical shifts relative to bR1, their separation by over 3 p.p.m from the other resolved resonances and the fact that bR1, N1 and O1ab are also unaffected by
D2O exchange, we assume that these resonances are also due to Met118 The additional resonances are of similar intensity and are shifted both upfield and down-field in the N-like state This could indicate structural heterogeneity around this residue in the N-like state
in the membrane environment On conversion to the O-like state two resonances O1a and O1b are observed The second of the two methionine residues located close to the retinal binding pocket [27], Met145, gives rise to a comparatively broad resonance bR6 The pro-cess of introducing purple membrane samples into rotors and into the magnet before running experiments for many hours at 253 K will accumulate considerable amounts of dark adapted (DA) bR compared with light adapted (LA) bR in our samples Met145 has already been identified as a key residue in the dark adaptation of bR [28] and the relatively large linewidth
of this resonance is evidence for Met145 being either
in two conformations in DA and LA bR or experien-cing two slightly differing electronic environments At
pH 10, mimicking the N-state, two resolvable reso-nances N5 and N6 are observed which are close to the
bR resonance of Met145 (bR6) The chemical shift of N5 is almost identical to bR6 and N6 is slightly shifted upfield by 0.6 p.p.m In the O-state, resonance O5 appears at the same chemical shift as N6 with a down-field shoulder resonance The small difference in chem-ical shift and the fact that N5, N6 and O5 are also unaffected by Mn2+induced line broadening and D2O exchange suggest that N5, N6 and O5 are due to the same residue, Met145
The final methionine residue is Met209, located on helix G The intense resonances N3 and O3 in the N-like and O-like preparations (Fig 2B,C) were assigned to Met209 using the REDOR technique des-cribed above The resonance asdes-cribed to Met209 shifts downfield by 0.8 p.p.m compared to the ground-state and becomes the most intense resonance The change
in intensity could be due to an increase in cross polar-ization (CP) efficiency which, in combination with the chemical shift change, would indicate a change in
Trang 10conformation and dynamic that is maintained in both
N- and O-like states The exact nature of the alteration
in dynamic and how it is linked to the small observed
changes of orientation of helix G [7,13] is unclear,
however, it is possible that an increase in intensity of
this resonance could be related to a reduced dynamic
in this region as the helix moves from its ground state
orientation
It is interesting to note that in O and also in N, even
when considering small M-state contributions, that
some resonances are broadened while others remain
sharp Resonances from residues Met118 and Met145
are split into a number of differing lines in contrast
with others such as Met20, which remain single peaks
Interestingly, these line-broadend residues are located
around the retinal indicating heterogeneity in this
region Previous solid-state NMR studies of D85N
bacteriorhodopsin [29] and Raman studies of
D85N⁄ F42C [30] showed that in the O-like state at
pH 6, at least two different retinal conformations are
present: 13-cis, 15-syn; and all-trans, 15-anti Despite
earlier reports of a completely all-trans chromophore
at pH 10.8 [31], it was suggested by solid-state NMR
[29] and other studies [32] that a mixture of 13-cis,
15-anti and all-trans chromophore, with a predomination
of the 13-cis, 15-anti form in a bent binding pocket,
exists At pH 10, the residual M-state adds
contribu-tions from deprotonated 13-cis, 15-anti retinal to the
line broadening The retinal structural heterogeneity is
reflected in the chemical shift changes and line
broad-ening that takes place for resonances assigned to
Met118 and Met145, which are located close to the
retinal binding pocket Other resonances such as
Met20 have much smaller linewidths indicating that
the protein structure around those labels is rather
more homogenous It can be seen from the N- and
O-like spectra (Fig 2B,C) that Met118 is strongly
affected by the presence of a structurally
heterogene-ous chromophore Based on the observed linewidths
and line shapes, Met118 seems to be more
heterogene-ous in the N-like spectrum compared with the O- or
bR-state spectra If the assignment of N5, N6 and O5
to Met145 is correct, then our data implies that this
residue adopts two conformations at pH 10, one
sim-ilar to the bR state and one that will persist into the
O-like state At pH 6 the structural heterogeneity
around Met118 would be reduced whilst Met145
would be able to adopt a single conformation
Conclusions
15N CPMAS combined with selective [15N]Met
labelling has provided some observations on the
conformational changes that a number of reporter res-idues in the transmembrane helices undergo The large chemical shift dispersion amongst the 15N-labelled methionine residues allows almost complete resolution
in the bR-state and many interesting spectral features
to be identified in mutant membranes mimicking the
N and O photointermediates The resolution is sufficient to accurately assign some of the residues The observed conformational heterogeneity and the spectral characteristics, as observed previously in
13CO-labelled preparations [22], identifies amino acids
in helix B that undergo fluctuating motions in the last stage of the photocycle before the protein returns to the bR state
While the double mutant used for our study allows
a decent separation of states, the D85N⁄ F42C mutant might be even better suited for future studies A low-ered pKa for Asp96 [30] may well have even less het-erogeneity at pH 10 and could provide an attractive system for further study using this technique, making for an interesting comparison
Recently we have shown that orientational con-straints can be determined, in a site-directed manner,
by specifically labelling bacteriorhodopsin within the purple membrane with 15N-enriched amino acids in combination with magic angle oriented sample spin-ning (MAOSS) [33] solid-state NMR methods [20,34] This will allow the observed chemical shift changes resolved in this study to be correlated with a more extensive study of the helix reorientations during the reprotonation phase of the photocycle within the nat-ural membrane, the results of which will be reported elsewhere
Experimental procedures
Sample preparation
Halobacterium salinarum (S9 or L33 expressing D85N⁄ T170C [19]) were cultured in a synthetic medium (1 L) containing all nutrients requisite for normal growth [35] [15N]l-methionine (0.19 gÆL)1) was added to the medium in place of the usual unlabelled l-methionine After five days incubation in the dark (225 r.p.m., 37C), when D660
measurements peaked, the cells were harvested and the purple membrane purified following published procedures [36] Sucrose density centrifugation was performed using a stepped sucrose gradient of 10 mL of each of 45%, 35% and 25% sucrose (w⁄ w) with centrifugation overnight
(83 000 g, Beckmann SW28 rotor, 15C) Samples contain-ing purified purple membrane were washed and
resuspend-ed in 5 mm Na3citrate, 5 mm KCl buffer (pH 6) Samples containing protein carrying the D85N⁄ T170C mutation