In contrast, calcium sen-sors have a lower affinity Kd¼ 105 to 107m Keywords calcium-binding proteins; molecular dynamics; NMR; nuclear relaxation; solution structure Correspondence C.. T
Trang 1a compact calcium-binding protein
Ghada Rabah1, Razvan Popescu1, Jos A Cox2, Yves Engelborghs3and Constantin T Craescu1
1 INSERM & Institut Curie, Centre Universitaire, Orsay, France
2 De´partement de Biochimie, Universite´ de Gene`ve, Suisse
3 Katholieke Universiteit, Leuven, Belgium
Calcium is a universal cellular secondary messenger
involved in the regulation of a large variety of vital
processes from the fertilization and development of a
cell to its death [1] Its great versatility is mainly due
to multiple specific interactions within various
mole-cular networks, the precise response being finally
deter-mined by the local ion concentration as well as its time
and spatial evolution Some proteins (such as enzymes
and chaperones) bind Ca2+and change their own
bio-logical activity, whereas others convey the Ca2+signal
through a functional change in another (target)
mole-cule Calmodulin and troponin C, the best studied
members of the latter category, belong to the EF-hand
superfamily, which is characterized by a well-conserved
Ca2+-binding motif (helix–loop–helix) In addition to the mediator, or sensor-type activity, proteins of this family may also interact with the metal ion for buffer-ing, uptake, or transport purposes [2] The ion-binding parameters (affinity, binding kinetics, selectivity, co-operativity) and the structural response on ion binding are strongly related to the biological role played by the protein Thus, Ca2+ buffers (e.g parvalbumin, calbin-din D9k) generally have a high affinity (Kd< 10)7m), may bind Mg2+equally well, and are conformationally less sensitive to ion binding In contrast, calcium sen-sors have a lower affinity (Kd¼ 10)5 to 10)7m)
Keywords
calcium-binding proteins; molecular
dynamics; NMR; nuclear relaxation; solution
structure
Correspondence
C T Craescu, INSERM & Institut
Curie-Recherche, Centre Universitaire, Baˆtiments
110–112, 91405 Orsay, France
Fax: +33 1 69 07 53 27
Tel: +33 1 69 86 31 63
E-mail: Gil.Craescu@curie.u-psud.fr
(Received 15 December 2004, revised 14
February 2005, accepted 24 February 2005)
doi:10.1111/j.1742-4658.2005.04629.x
The solution structure of Nereis diversicolor sarcoplasmic calcium-binding protein (NSCP) in the calcium-bound form was determined by NMR spec-troscopy, distance geometry and simulated annealing Based on 1859 NOE restraints and 262 angular restraints, 17 structures were generated with a rmsd of 0.87 A˚ from the mean structure The solution structure, which is highly similar to the structure obtained by X-ray crystallography, includes two open EF-hand domains, which are in close contact through their hydrophobic surfaces The internal dynamics of the protein backbone were determined by studying amide hydrogen⁄ deuterium exchange rates and 15N nuclear relaxation The two methods revealed a highly compact and rigid structure, with greatly restricted mobility at the two termini For most of the amide protons, the free energy of exchange-compatible structural open-ing is similar to the free energy of structural stability, suggestopen-ing that iso-tope exchange of these protons takes place through global unfolding of the protein Enhanced conformational flexibility was noted in the unoccupied
Ca2+-binding site II, as well as the neighbouring helices Analysis of the experimental nuclear relaxation and the molecular dynamics simulations give very similar profiles for the backbone generalized order parameter (S2), a parameter related to the amplitude of fast (picosecond to nanosec-ond) movements of NH-H vectors We also noted a significant correlation between this parameter, the exchange rate, and the crystallographic B fac-tor along the sequence
Abbreviations
MD, molecular dynamics; NSCP, Nereis diversicolor sarcoplasmic calcium-binding protein; SCP, sarcoplasmic calcium-binding protein.
Trang 2parative analysis of a large diversity of sequences and
biological functions within the EF-hand family is
therefore required to better understand the structural
basis of the various biological functions
Nereis diversicolorsarcoplasmic calcium-binding
pro-tein (NSCP) is an acidic calcium buffer propro-tein, which
is very abundant in the sarcoplasmic reticulum from
the annelid N diversicolor It belongs to the
sarcoplas-mic calcium-binding protein (SCP) subfamily, also
including the invertebrate functional analogs of the
vertebrate parvalbumin [4] Similar to other SCPs,
NSCP has four potential ion-binding sites, but only
three of them (sites I, III and IV) have a high affinity
for Ca2+or Mg2+[5]
The 3D crystal structure, solved at 2 A˚ resolution
[6], revealed a globular structure in which the two
EF-hand pairs, constituting an EF-hand domain, are
close to each other, in contrast with the bi-lobal,
extended conformation of calmodulin or troponin C
Spatial proximity of the binding sites makes functional
communication between them possible, measured in
terms of a strong positive co-operativity in Ca2+
bind-ing [7] In addition, physicochemical experiments
per-formed in our laboratory revealed that metal binding
induces transition from a molten globule state into a
well-defined, and highly stable conformation [8,9] A
rational understanding of these specific properties
requires structure and dynamic characterization in
solution of the various functionally relevant states
To achieve this aim, we initiated a structural and
dynamics analysis of Ca2+-saturated NSCP in
solu-tion, based on NMR spectroscopy, nuclear relaxation
measurements, and molecular dynamics (MD)
simula-tions Using purified wild-type samples, as well as
15N-labeled samples overexpressed in Escherichia coli,
we have assigned the 1H and 15N resonances of the
protein (BioMagResBank, accession number 4129)
[10], and collected distance and angle restraints for
structural determination The solution structure, based
on 1859 NOE distance restraints and additional
experi-mental and chemical information, is very similar to the
previously reported crystallographic structure [6] In
order to better understand the great structural stability
eter profiles that were in close agreement with the experimental parameters
Results and Discussion
Solution structure
We have shown previously that apo-NSCP in solution
is highly disordered and gives poor NMR spectra, characterized by a low chemical shift dispersion and absence of NOE interactions [8] In the presence of saturating Ca2+ions, the spectra extend over 12 p.p.m
in the proton dimension and exhibit many interproton dipolar interactions, and the sample is stable enough for long lasting 3D experiments Spin systems for 171
of 174 residues ( 98%) were partially or completely assigned [10] Unassigned resonances correspond to a peptide fragment situated at the C-terminus of the pro-tein (D163–T165) Combined analysis of an ensemble
of NMR parameters in the Ca2+-bound state, inclu-ding short-range and medium-range NOE interactions,
Ha secondary chemical shifts, 3JHNHa coupling con-stants, and amide proton exchange rates, enabled us to delineate eight a-helices consisting of fragments 3–15, 25–38, 45–59, 72–82, 90–103, 113–122, 130–137 and 147–159, representing 56% of the residues The posi-tion of the helices corresponds closely to those observed in the crystal structure [6], but the helices are often shorter by 1–3 residues Four short b-strands (22–24, 69–71, 110–112, 144–146) were also identified
by the low-field-shifted Ha protons, strong daN(i,i+1) sequential connectivities and large 3JHNHa couplings The strands could be grouped into two antiparallel b-sheets based on the strong dipolar interactions observed between Ha and HN protons from opposite chains
A total of 1859 interproton distance restraints and
262 dihedral angle restraints were used to fold these secondary-structure elements into a 3D conformation, using distance geometry and simulated annealing com-putations Figure 1A shows the backbone superimposition
of the final 17 structures, and Fig 1B the ribbon repre-sentation of the best representative of the ensemble for
Trang 3the Ca2+-bound NSCP form In contrast with the
bilobal aspect of calmodulin, the prototype of the
EF-hand superfamily, NSCP exhibits a globular shape
characterized by a large contact area and multiple
side-chain contacts between the two EF-hand domains
Consequently, residues from the two terminal
frag-ments are very close to each other The structural
cohesion of the two EF-hand domains is mainly deter-mined by a highly hydrophobic core including 15 Phe and three Trp side chains contributed roughly equally
by the two molecular halves (Fig 2) The structure of individual EF-hand domains is close to the canonical geometry [11], observed for the family of proteins ana-lyzed so far
The aqua and procheck-nmr programs [12] were used to assess the quality of the restraints and to deter-mine the geometry regularity of the final structures (Table 1) More than 87% of (F, Y) angle pairs of the
17 final structures lie in the most-favored region of the Ramachandran plot, and about 99% of them lie in the allowed regions The segment 162–165 is mostly found in the disallowed regions of the Ramachandran plot These loop residues are not engaged in a detect-able hydrogen bond and presumably undergo some geometry constraint from the neighboring amino acids The global solution structure of NSCP is very close
to the previously determined crystal structure [6] The rmsd calculated for the heavy atoms in regular secon-dary-structure elements is 1.14 A˚, with a large asym-metry between the two halves: 1.05 A˚ and 0.77 A˚ for the N-terminal and C-terminal halves, respectively This may partially reflect the lack of metal binding
in the second loop and the longer linker region between EF-hand I and II As in aequorin [13], these features may result in increased flexibility of the first half
It was generally observed that the relative position
of the two helices in EF-hand motifs changes signifi-cantly upon Ca2+ binding [14], from an almost anti-parallel configuration to a perpendicular arrangement
In a domain containing a pair of EF-hand motifs, this movement creates a large exposed hydrophobic surface which, in the case of regulatory proteins, constitutes the target binding site The values of the interhelix angles in NSCP, measured over the NMR ensemble (Table 2), are similar to those observed in regulatory EF-hand proteins, in the Ca2+-bound state, and are centered around 90 Compared with the calmodulin, where the two lobes are spatially separated, the com-pact SCPs exhibit a larger variability among the inter-helix motifs, with significantly lower values for the first two motifs Indeed, these two EF-hand domains in SCPs are more open, than in other Ca2+ buffer pro-teins, such as calbindin D9k [15] or parvalbumin [16] These differences may be due to the compact structure and the tight interactions between the two EF-hand domains, inducing supplementary constraints on the interhelical angles
As can be seen in Table 2, some binding loops lost the high affinity for the metal ion, but this still
main-A
B
C
Fig 1 Global representation of the solution structure of
Ca 2+ -NSCP (A) Backbone stereo view of the 17 final structures
superimposed based on the main-chain heavy atoms in regular
secondary-structure elements The N-terminal EF-hand domain is
shown in red, the C-terminal EF-hand domain is shown in blue,
the linker is green, and the C-terminal fragment is in magenta.
(B) Ribbon representation of the best structure selected as the
clo-sest to the ensemble average The color code is the same as in
(A) The binding loops are noted from I to IV, and the a-helices and
some residue positions are labeled to facilitate the chain-folding
pathway The figure was prepared with MOLSCRIPT [55] and RASTER 3 D
[56] (C) Electrostatic potential calculated at the molecular surface
of the best structure of Ca 2+ -NSCP using the GRASP program [57].
On the left side the molecule is oriented as in (B), whereas on the
right side it is rotated through 180 around the vertical axis The
positive and negative potential are conventionally coded in blue and
red, respectively.
Trang 4tains the corresponding motif in an open
confor-mation The presence of Ca2+ in the three active
EF-hand motifs is nonambiguously confirmed by the
chemical-shift signature of the Hb protons in the Asp
residue occupying the first position in the binding loop
D1 [17] Indeed, the close proximity to the conserved
Phe residue in position)4 (relative to D1) accounts for
the large upfield shift of one of the b protons in D1: 1.58, 1.73, and 1.97 p.p.m in EF-hand motif I, III, and IV, respectively
The chemical shift of the amide nitrogen in the resi-due occupying position 8 of the Ca2+loop was shown
to be larger in the Ca2+-bound form (124–127 p.p.m.), relative to the apo form (111–120 p.p.m.) of EF-hand proteins [18], because of a decreased electronic density around the amide nitrogen nucleus This parameter was therefore proposed as a sensitive probe for the metal occupancy of a given motif In NSCP(Ca2+)3 the amide 15N chemical shift at the corresponding positions (I23, I70, I111, L145) are 124.7, 122.0, 123.1 and 120.8 p.p.m., respectively, with the metal-bound motif IV showing a smaller value than the empty site
II According to the above classification, only the first site shows a chemical shift compatible with a bound loop These observations suggest that the nitrogen
elec-Fig 2 Stereoview of the aromatic residue cluster in Ca 2+ -NSCP Phe, Trp and Tyr side chains are shown in red, blue and green, respect-ively.
Table 1 Experimental restraints and structural statistics for the 17
simulated annealing structures of (Ca 2+ )3-NSCP.
Restraint statistics
Medium range (2 £ |i-j| < 5) 342 18.4%
Hydrogen bond restraints 188
Dihedral angle restraints (F,Y) 262
Average no of NOE restraint violations
Average of NOE upper restraint violations 0.0045 A ˚
Averahe of NOE lower restraint violations 0.0009 A˚
Average rmsd from the average structure (A ˚ )
Residues 3–15, 22–37, 46–57, 69–82,
89–102, 110–122, 129–137, 144–159 a
0.87 ± 0.13
Ensemble Ramachandran plot
Residues in the most-favored region 87.1%
Residues in additional allowed regions 10.8%
Residues in generously allowed regions 1.4%
Residues in disallowed regions 0.8%
a Backbone atoms (N, C¢, Ca).
Table 2 Statistics of the interhelix angles within EF-hand motifs in NSCP, related EF-hand proteins and calmodulin An asterisk marks EF-hand motif that had lost Ca 2+ -binding capacity.
Helix pair
Angle () NSCPa (NMR)
NSCPb (X-ray)
BlSCPc (X-ray)
SeCaBPd (NMR)
Calmoduline (X-ray)
a
This work, mean ± SD over the 17 molecule ensemble. b[6], 2SCP.pdb c B lanceolatum SCP [53], 2SAS.pdb d Bacterial EF-hand protein, calerythrin [54], 1NYA.pdb e [20], 1CLL.pdb.
Trang 5tron density is not the unique determining factor for
its chemical shift, and question the utilization of this
NMR parameter as a probe for the metal binding to a
given site
The electrostatic potential at the protein surface is
dominantly negative (Fig 1C), in agreement with the
acidic character of the EF-hand proteins It may be
noted that the surface area encompassing the Ca2+
-binding sites III and IV exhibits a more homogeneous
and intense negative potential, as compared with the
corresponding area of the first two binding loops This
may contribute to the low cation-binding affinity of
site II
The question may be raised as to how the sequence
and structure of EF-hand proteins, classified as
cal-cium buffers or transporters, account for the lack of
regulatory capacity, essentially expressed by a Ca2+
-modulated interaction with cellular targets Structural
analysis of calmodulin complexes revealed that the
interaction interface is constituted by a large
hydro-phobic surface created by the opening of the two
EF-hand domains on Ca2+ binding In the case of
SCPs, the highly apolar surface of the two EF-hand
domains exhibit a greater affinity for each other [19],
yielding a compact globular fold, which precludes the
recognition and binding to the hydrophobic surface of
the target proteins This is made possible by extensive
bending of the interdomain linker, which is usually
found to be almost linear in the crystal structure of
calmodulin [20] However, a recent crystallographic
study showed that calmodulin can equally form a
com-pact structure [21], as suggested by previous biophysical
and biochemical studies in solution [22,23] In fact, the
D⁄ E interhelix angle, calculated as in [14], is very close
in the compact calmodulin (1PRW.pdb) and NSCP
(solution structure):)111 and )116, respectively
Sequence and structural comparison between
cal-modulin and three different compact EF-hand proteins
(NSCP, Botrychium lanceolatum SCP and
Saccharo-polyspora erythraea calcium-binding protein) may
reveal some factors contributing to the preference for
the globular shape A Pro residue between the D and
E helices, which exists only in NSCP, could induce a
tight turn in this region and render the compact
struc-ture more stable More significantly, the number of
long-chain hydrophobic residues is distinctly higher in
the domains that associate to form compact structures
Thus, the total number of Phe, Trp, Leu and Ile
resi-dues is 25 in calmodulin and 38, 37, and 35 in NSCP,
B lanceolatum SCP and S erythraea calcium-binding
protein, respectively Therefore, the collapse of the two
EF-hand halves, with the formation of a more stable
apolar core (Fig 2), is preferred over the extended,
highly solvent-exposed structure In addition, a 9–10 residue insertion in the C-helix of NSCP, B lanceola-tum SCP and S erythraea calcium-binding protein, including five hydrophobic side chains, ensures a larger and tighter contact surface between the N-terminal and C-terminal domains (Fig 1B) This tendency to form a very stable hydrophobic core is very well illus-trated by the fact that the isolated N-terminal and C-terminal halves of NSCP form homodimers, but when mixed, they form a complex with a conformation
as of intact NSCP [19] A more detailed structural and thermodynamic investigation of the interdomain inter-face in the compact EF-hand proteins should be very useful for a quantitative explanation of the conforma-tional preference
Conformational flexibility studied by amide exchange kinetics
Analysis of the amide exchange kinetics provides a site-specific description of global or local conforma-tional dynamics of a protein in solution Description
of the exchange process in terms of protection factors [24] enables us to eliminate the influence of the solu-tion properties (pH, ionic strength, temperature, etc.), and of the chemical environment of amide groups (the sequence context) Therefore, the protection factors may be directly related to the relative attenuation of the hydrogen exchange rate in given main-chain posi-tions of the native structure, relative to the random-coil state We were able to quantitate this parameter for 72 out of the total of 169 amide protons and for three indole amino groups from Trp side chains Five missing values corresponding to the amide protons with high exchange rates (kex> 10)2min)1), K19, F35, L49, M122, and V168, are indicated by the down-ward arrows in Fig 3 The intensity of the remaining peaks could not be accurately measured due to overlap
in the HSQC spectra
Most of the measured protection factors have relat-ively high values (mean 106.5), and are associated with the a-helices and b-strands, except for the empty binding motif II (Fig 3A) The strong protection of the amide protons in Ca2+-saturated NSCP is in good agreement with its high structural stability [9] Thus, the free energy of conformational opening estimated here from hydrogen isotopic exchange measurements are centered around 37.7 kJÆmol)1 (9 kcalÆmol)1) for the three bound EF-hands, which is close to the free energy of the structural stability calculated from dena-turation experiments [9] This strongly suggests that the conformational fluctuations, enabling the measured proton exchange, correspond to global dynamics, and
Trang 6are highly similar to those accompanying the
co-opera-tive unfolding of the whole structure
Clearly, the number of measurable amide protection
factors and their magnitude are smaller for the second,
unbound EF-hand motif, but also for the neighboring
helices (B and E) The unbound loop induces more
rapid conformational fluctuations that extend outside
the motif, in both directions of the main chain, as also
reflected in the crystallographic B factors of the N
atoms (Fig 3)
The side-chain Trp protons exhibit remarkably high
protection factors, comparable to the backbone amide
protons in the more flexible helices and in the
C-terminal fragment (Fig 3) Among the three Trp
resi-dues, the Ne1 proton of Trp4 belongs to a hydrogen
bond with the carbonyl oxygen of Phe158, observed
both in the NMR and the previous X-ray structure [6]
Owing to the deshielding effect of this interaction, the
proton chemical shift is significantly low-field-shifted
(10.52 p.p.m.) relative to the random coil value (10.22
p.p.m.) The corresponding proton in Trp170 may
form an aromatic hydrogen bond with the side chain
of Phe157, as suggested by the structure, and
suppor-ted by the large high-field shift of its proton resonance (7.28 p.p.m.) induced by the phenyl ring current The low exchange rate of the indole protons in these two residues may be explained by the protection provided
by the hydrogen-bond formation In contrast, no explanation is actually available for the exchange pro-tection in Trp57, which is largely exposed to the sol-vent, and shows no detectable intramolecular hydrogen bond
Internal MD studied by relaxation measurements Computation of the principal components of the iner-tia momentum, based on the NMR-derived solution structure, gives (1.00 : 0.83 : 0.66) According to these values, only a modest degree of anisotropy is expected, which may not influence significantly the microdynamic parameters (at least the order parameters) [25] Owing
to cross-peak overlap, reliable analysis of peak inten-sities and relaxation parameters was limited to 121
HN-N vectors, including 118 amide groups (out of the total of 169 observable amide protons) and three indole amino groups from the Trp side chains
Fig 3 Dynamics analysis by amide proton
exchange kinetics of Ca 2+ -NSCP (Top)
1 H ⁄ 2 H exchange kinetics expressed as the
logarithm of the protection factor [log (P)]
and the free energy of isotope exchange
(DG ex ) The mean log (P) is indicated by the
horizontal line Downward arrows designate
the amino protons with exchange kinetics
faster than 10)2min)1 The exchange
parameter of the Trp indole protons (W4,
W57 and W170) are shown at the end of
the sequence, by the grey bars The
secon-dary-structure elements and occupation of
the binding loops are represented
sche-matically at the top of the figure (Bottom)
crystallographic B factors of N H atoms
determined by the X-ray approach [6].
Trang 7Figure 4 shows the relaxation parameters (R1, R2,
NOE) and their uncertainty plotted as a function of
residue number The dynamic analysis started with
the determination of the rotational correlation time
(sc) of the whole protein using a procedure designed
to minimize the effects of heterogeneous local
move-ments [26] For a given arbitrary value of sc, the
dynamic parameters (S2 and se) are computed using
R1 and heteronuclear NOE for each amide vector
within the most rigid segments (84 sites) in the frame
of the Lipari-Szabo approach [27] Then, R2 values
may be reconstructed (from the spectral density
func-tions) and compared with the experimental
counter-parts for the selected sites The final value of the
correlation time, corresponding to the minimum of
the v2(R2) function, was 6.93 nsÆrad)1 Using a simp-ler method based on the independence of the R2⁄ R1 ratio from S2 and se [28], we obtained a very close value for sr (6.86 ±0.34 ns) The value of the rota-tional correlation time is indicative of a mainly mono-meric form of NSCP
With the value for the global correlation time obtained by the first method, we analyzed the relaxa-tion parameters in terms of the simple or extended Lipari-Szabo model-free methods [27,29] A Monte-Carlo simulation with 500 steps was used to estimate the standard error of the microdynamic parameters The data for all the studied vectors (except amides in A32 and S112) could be fitted to the simple Lipari-Szabo procedure giving the generalized order
param-Fig 4 Relaxation parameters (R1, R2, g) measured in Ca 2+ -saturated NSCP at 308 K The elements of secondary structure and the occu-pancy of Ca 2+ -binding sites are shown at the top of the first panel The last three values correspond to the N e1 -H vector in Trp side chains (W4, W57, W170).
Trang 8eter (S2), the internal correlation time (se) and the
exchange contribution to the transversal relaxation
(R2ex) The extended procedure of relaxation analysis
did not improve the fitting quality of any amide or
amine system
The generalized order parameter S2 at a given
back-bone site is a measure of the amplitude of the fast
(picosecond to nanosecond) movement of the HN-N
vector Figure 5 shows the order parameters and their
standard deviations, estimated by Monte-Carlo
simula-tions, as a function of the residue number The
param-eter varies between 0.66 and 0.93, with a mean value
of 0.83 ± 0.04, which is close to the average value
usually observed for native globular proteins [30]
Overall, the estimated order parameters indicate that
the backbone amide vectors undergo
picosecond-to-nanosecond movements of low amplitude, reflecting a
compact and rigid fold, with well-structured end
frag-ments Larger S2 values (from 0.84 to 0.93) are
grouped in the loop I, helix F and helix H, while the
intermotif linkers and the unoccupied Ca2+-binding
loop exhibit lower S2 values (down to 0.66), attesting
to larger amplitude fast movements The end
frag-ments are characterized by high S2 values, meaning
that their movement is significantly restricted Of the
four EF-hand motifs, the first and the fourth exhibit
the most restricted picosecond-to-nanosecond mobility
of the backbone vectors The absence of Ca2+binding
to the second EF-hand induces an irregular pattern of
S2 values in the loop residues that extends over the
neighboring helices It is worth noting that the helices
characterized by a larger fast movement amplitude and
a high amide proton exchange (B, C, E) belong to the interface between the two EF-hand domains
The relatively high values (0.75–0.78) of S2observed for the amino group of the Trp side chains indicate that the fast movements of these indole moieties are restricted to a similar extent to the backbone amide vectors (Table 3) Inspection of the calculated structure (Fig 2) shows that the three indole groups do not have comparable environments in the 3D structure: whereas W4 and W57 are highly exposed to the solvent at the protein surface, W170 is deeply embedded in the hydrophobic core created by helices E, F and H The order parameters of these side chains appear to be independent of this environmental context
The large majority of the residues display fast rate librational motions characterized by an internal corre-lation time se< 50 ps, with a dozen (mainly localized
in linker fragments) having a correlation time in the range 50–100 ps (not shown) Five residues (V5, E40, G67, S90, and D156) exhibit R2ex values between 1 and 2 s)1, reflecting microsecond–millisecond internal motions in their environment Again, they are associ-ated with end fragments (V5, D156), linkers (E40, S90)
or the empty calcium-binding loop (G67)
Fig 5 Generalized order parameters along
the sequence of Ca 2+ -saturated NSCP The
experimental values obtained from the NMR
relaxation experiments, together with the
standard deviations, are shown in red The
order parameters estimated from the MD
simulations are in black The last three
values, at the end of the sequence,
corres-pond to the Trp4, Trp57, and Trp170 indole
N-H vectors (boxed).
Table 3 Experimental (S 2
NMR ) and calculated (S 2
MD ) order parame-ters for the Trp side chains.
MD
Trang 9Comparison with the crystallographic B factor
The crystallographic B factor is considered to reflect
mainly the molecular flexibility at the atomic level, but
other factors related to static and dynamic disorders in
the crystal may give important contributions as well
[31] Intuitively, it should be inversely correlated with
the order parameters estimated from the relaxation
measurements, but in practice the relationship is more
complex [32] Overall, in NSCP the regions of
high-order parameters exhibit lower B factors for amide
nitrogens (Fig 3) As the range of values for the order
parameter is about three times lower than that for the
B factor, the quantitative correlation between the two
parameters along the sequence is only moderate (the
correlation coefficient is )0.32) Discrepancies also
arise from the fact that S2 reflects only fast motions,
whereas the B factor is sensitive to both fast and slow
movements [33] In a similar approach for
ribonuc-lease, Mandel et al [34] found similar low values
between )0.36 and )0.64, depending on the X-ray
structure considered This variability illustrates the
dependence of the thermal factors on the
crystalliza-tion state and the intermolecular contacts within the
crystal
MD simulation
Simulation of the protein internal dynamics under an
appropriate physical force field provide a detailed
atomic picture of the movements underlying the
nuc-lear relaxation parameters and the corresponding order
parameters Only 1 ns (the second half) from the 2-ns
very stable trajectory of the Ca2+-saturated form of
NSCP was used in the theoretical analysis The
mean ± SD temperature was 299.99 ± 4.47 K, and
the total energy was 1722.5 ± 21 kJÆmol)1 (411.4 ±
5 kcalÆmol)1) Correlation functions for backbone
NH-H vectors were computed from the selected
traject-ory using an interval of 400 ps, which provides reliable
sampling of the fast ( 100 ps) motions [35]
Correla-tion funcCorrela-tions were computed from the trajectory for
168 residues from the total of 174 (the N-terminus and
the six prolines do not have an sp2NH-H bond)
The correlation functions of the internal motions,
CI(t) display different patterns along the sequence, and
only 132 could be considered completely convergent
For 25 other residues, the correlation function exhibits
a slow decay towards a lower plateau and oscillatory
behavior, suggesting the presence of slower motions,
which may not be accounted for correctly by the
length of the present trajectory [35] Finally, for 11
resi-dues no plateau value could be reached within 400 ps
The plateau values of the correlation functions (inclu-ding both rapidly and slowly convergent functions), which represent the theoretical order parameters S2
MD, were computed for 157 amino-acid residues (Fig 5)
A total of 151 of the 157 simulated S2
MD values are larger than 0.7, confirming the high rigidity of the NSCP backbone structure As shown in Fig 5, there
is a good correlation between the order parameter profiles established by theoretical and experimental approaches
The average value of S2 obtained by MD (S2ave¼ 0.80, average over 157 points) is slightly lower than that calculated from nuclear relaxation data (S2ave¼ 0.83, average over 116 points), as also noted for apo-neocarzinostatin [36] It may be noted that there is a remarkable parallelism between the two S2 profiles at the limits between secondary-structure elements and linkers, where this parameter exhibits larger variations Amide vectors in the unoccupied binding loop II, in the linker between helices F and G, as well as in the last loop show markedly decreased order parameters both in the relaxation experiments and the simulation, permitting cross validation of the underlying MD (Fig 5) The MD simulations predict slightly larger amplitude motions in the more flexible protein seg-ments than estimated by relaxation measureseg-ments It must be stressed that the overall shift between the NMR and MD values for the order parameter could
be decreased by choosing a slightly smaller value for the initially estimated global correlation time [37] Previous comparative studies on the simulation and experimental dynamic parameters of globular proteins [35–37] have always noted some differences in the order parameters determined by the two methods One
of the reasons is that both MD and NMR analysis involve several approximations In the case of the MD simulations, these include the parameters of the empir-ical force field, insufficient sampling because of the finite length of the trajectories, and the simplified treat-ment of the solvent The main simplification in the NMR approach concerns the relaxation data analysis, usually performed under the assumption of a unique overall correlation time and the independence of move-ments on different time scales
In addition to the amide vector dynamics, we also analyzed the fast movements of the Ne1-H vectors in the three Trp side chains The order parameters of the picosecond motions of these bonds (Table 3) are only slightly lower than the backbone values (as in the relaxation results), indicating that the side chains also move in a highly restricted regime [37] Trp57 senses the molecular environment of the empty binding site and shows significantly larger amplitude than the other
Trang 10and co-ordinate rmsd profiles, the apo and holo
dynamics are of comparable quality Figure 6A
com-pares the two final simulated structures and the crystal
structure, using the pseudo-dihedral angles defined by
four successive Ca atoms along the sequence As may
be noted from the upper panel, in presence of the
bound Ca2+ions the final protein secondary structure
(blue symbols) shows no significant change relative to
the crystal conformation (red bars) In contrast, the
final structure of the apo simulation (Fig 6A, lower
panel) shows significant secondary-structure differences
relative to the starting (holo-type) conformation,
mainly localized to the first binding loop, the N-side of
the C and D helices, and the C-end of the F helix
Comparison of the tertiary structures including these
areas (Fig 6B) shows that removing the metal ion
from the first site determines an opening of the binding
loop, and induces important structural changes in the
first half of the protein The rearrangement of this
domain has remote consequences on the F helix, in the
second protein half, which may be accounted for by
the close contacts existing between the B and F helices
(Fig 6B) Indeed, this interhelix space belongs to the
large interdomain surface which plays an important
role in the structural and functional coupling between
the two halves The perturbation propagation
path-way, observed in the present simulation, includes the
higher flexibility secondary-structure elements,
high-lighted by the nuclear relaxation and proton exchange
experiments
A longer, and more complete, MD simulation of
apo-NSCP, in an explicit water environment, should
provide valuable insight into the metal-induced
con-formational changes and the characteristics of the apo
molten globule state This approach is currently being
taken in our laboratory
Experimental procedures
Protein preparation and labeling
Wild-type NSCP was purified from the Nereis muscle by
the method of Cox & Stein [5], modified as described by
NMR samples (1.0–1.2 mm) were prepared in deuterated
20 mm Tris⁄ HCl buffer ⁄ 5 mm CaCl2, pH 6.5, in 95%1H2O⁄ 5%2H2O or in 100%2H2O
NMR spectroscopy All NMR spectra were acquired at 308 K on a Varian Unity-500 spectrometer, equipped with a triple-resonance probe and a Z-field gradient Standard methods were used
to obtain 2D NOESY and 3D15N-NOESY-HSQC spectra [40,41], with mixing times of 100 ms The 3D spectra were acquired as 128 (t1)· 32 (t2)· 512 (t3) complex points with
a spectral width of 1500 Hz in the nitrogen dimension,
3200 Hz in the amide proton dimension, and 7000 Hz in the all-proton dimension Dihedral angle restraints were obtained by analyzing the HMQC-J spectrum by the method proposed by Wishart & Wang [42] Data processing and restraint collection were performed using felix97 soft-ware (Accelrys, San Diego, CA, USA), running on a Silicon Graphics Octane workstation
15N nuclear relaxation The heteronuclear relaxation experiments were performed
at 308 K and 11.74 T (500 MHz proton resonance fre-quency) The R1 relaxation rate was measured using the inversion recovery method, modified to obtain decreasing signal intensities as a function of the relaxation delay Measurement of the transverse relaxation rate (R2) was based on the Carr–Purcell–Meiboom–Gill pulse sequence with a delay between15N 180 pulses during the relaxation period of 0.9 ms Recycle delays of 2.5 s were used at the beginning of R1 and R2 pulse sequences Spectra for R1 measurements were acquired using 11.04 (· 2), 55.22, 165.66, 220.88 (· 2), 386.54, 552.2, 662.64, 828.3 and 1104.4 ms as relaxation delays R2 data were recorded with delays of 31.4, 47.1, 62.8, 78.5 (· 2), 125.6, 157, 188.4 and 219.8 ms Steady-state 1H-15N NOE was determined from spectra pairs with and without proton saturation The two experiments start with a 5 s recycle delay, but during the last 3 s of the saturation experiment the protons are irradi-ated by 120 pulses every 5 ms The pulse sequences used in the present experiments were adapted from those kindly