Calcium binding to gatekeeper residues flanking aggregation-prone segments underlies non-fibrillar amyloid traits in superoxide dismutase 1 (SOD1)

Full text

Calcium binding to gatekeeper residues flanking aggregation-prone
segments underlies non-fibrillar amyloid traits in superoxide dismutase
1(SOD1)
Sílvia G. Estácio
a,b,
⁎, Sónia S. Leal
c
, Joana S. Cristóvão
c
, Patrícia F.N. Faísca
a,b,
⁎, Cláudio M. Gomes
c,
⁎
a
Centro de Física da MatériaCondensada, Universidade de Lisboa, Lisboa, Portugal
b
Departamento de Física, Faculdade deCiências, Universidade de Lisboa, Lisboa, Portugal
c
Instituto de Tecnologia Química e Biológica António Xavier, Universidade Nova de Lisboa, Av. da República, 2780-157 Oeiras, Portugal
abstractarticle info
Article history:
Received 4 September 2014
Received in revised form 14 November 2014
Accepted 18 November 2014
Available online 25 November 2014
Keywords:
Molecular dynamics
Isothermal titration calorimetry
Electrostatic interaction
Protein aggregation
Gatekeeping residue
Protein dynamics
Calcium deregulation isa central feature among neurodegenerative diseases, including amyotrophic lateral scle-
rosis (ALS).Calcium accumulates in the spinal and brain stem motorneurons of ALS patients triggering multiple
pathophysiological processes which have been recently shown to include direct effects on the aggregation cas-
cade of superoxide dismutase 1 (SOD1). SOD1 is a Cu/Zn enzyme whose demetallated form is implicated in
ALS protein deposits, contributing to toxic gain of function phenotypes. Here we undertake a combined experi-
mental and computational study aimed at establishing the molecular details underlying the regulatory effects
of Ca
2+
over SOD1 aggregation potential. Isothermal titration calorimetry indicates entropy driven low affinity
association of Ca
2+
ions to apo SOD1, at pH 7.5 and 37 °C. Molecular dynamics simulations denote a noticeable
loss of native structure upon Ca
2+
association that is especially prominent at the zinc-binding and electrostatic
loops, whose decoupling is known to expose the central SOD1 β-barrel triggering aggregation. Structural map-
ping of the preferential apo SOD1 Ca
2+
binding locations reveals that among the most frequent ligands for
Ca
2+
are negatively-charged gatekeeper residues located in boundary positions with respect to segments highly
prone to edge-to-edge aggregation. Calcium interactions thus diminish gatekeeping roles of these residues, by
shielding repulsive interactions via stacking between aggregating β-sheets, partly blocking fibril formation and
promoting amyloidogenic oligomers such as those found in ALS inclusions. Interestingly, many fALS mutations
occur at these positions, disclosing how Ca
2+
interactions recreate effects similar to those of genetic defects, a
finding with relevance to understand sporadic ALS pathomechanisms.
© 2014 Elsevier B.V. All rights reserved.
1. Introduction
Protein misfolding and aggregation in neurodegenerative diseases
such as Alzheimer's disease, Parkinson's disease or amyotrophic lateral
sclerosis (ALS) result mostly from sporadic factors rather than genetic de-
fects, which actually account for a minority of cases. Therefore, triggers for
protein self-assembly into amyloid oligomer range from variations in the
physical-chemistry of the neuronal environment to modifications in pro-
tein interaction networks [1,2]. Deregulation of brain metal ion homeo-
stasis is emerging as a critical common feature across different
neurodegenerative diseases and cumulating evidence points to
pathological changes in the neuronal balance of metal ions such as zinc,
calcium, iron and copper [3–5] in these diseases. Metal ions are known
modulators of neuronal protein aggregation [6] and the interplay be-
tween metal ions and the amyloid-βpeptide, tau and α-synuclein leading
to toxic by-products, proteotoxic species and metal sequestration has
been substantially investigated in recent years [7,8].
In particular, Ca
2+
deregulation may be central in the selective vul-
nerability of motor neurons in ALS. Indeed, Ca
2+
is found to accumulate
in the spinal and brain stem motor neurons of ALS patients, as well as in
animal and cell models [9–15]. This results from a combination of fac-
tors that include: i) inherently low expression of Ca
2+
buffering pro-
teins; ii) excessive activation of glutamate receptors (excitotoxicity)
leading to an excessive Ca
2+
influx; iii) deregulation of Ca
2+
channels
and iv) disruption of Ca
2+
homeostasis [16]. Calcium dysfunction trig-
gers several pathomechanisms ranging from oxidative stress to mito-
chondrial failure and apoptosis, but the relationship with the protein
aggregation cascade is only now starting to emerge.
Studies on ALS cellular models have shown that Ca
2+
overload pro-
motes and correlates with superoxide dismutase 1 (SOD1) aggregation
Biochimica et Biophysica Acta 1854 (2015) 118–126
⁎Corresponding authors. Tel.: +351 217904862 (S.G. Estácio), +351 217904819
(P.F.N. Faísca), +351 214469332 (C.M. Gomes); fax: +351 217954288 (S.G. Estácio &
P.F.N. Faísca), +351 214411277 (C.M. Gomes). Correspondence to: S.G. Estácio and
P.F.N. Faísca, Departamento de Física, Campo Grande, Ed. C8, 1749-016 Lisboa, Portugal,
or C.M. Gomes, Instituto de Tecnologia Química e Biológica António Xavier, Universidade
Nova de Lisboa ITQB/UNL, Av. da República, 2780-157 Oeiras, Portugal.
E-mail addresses: silvia@cii.fc.ul.pt (S.G. Estácio), patricia.fn.faisca@gmail.com
(P.F.N. Faísca), gomes@itqb.unl.pt (C.M. Gomes).
http://dx.doi.org/10.1016/j.bbapap.2014.11.005
1570-9639/© 2014 Elsevier B.V. All rights reserved.
Contents lists available at ScienceDirect
Biochimica et Biophysica Acta
journal homepage: www.elsevier.com/locate/bbapap

[17,18]. SOD1 is a highly abundant enzyme which is involved in the an-
tioxidant defense. It is a 32 kDa homodimeric β-barrel protein contain-
ing an intramolecular disulfide bond and a binuclear Cu/Zn site that
catalyzes the disproportionation of superoxide to hydrogen peroxide
and dioxygen [19]. Mature SOD1 is a very stable protein (T
m
~ 92 °C)
which is, however, highly destabilized in its metal freeand disulfide re-
duced forms (T
m
~42°C)[20]. A few ALS patients (10–15%) have muta-
tions in the SOD1 gene; however, misfolding of wild type SOD1 is
implicated in the large majority of cases. From these, most SOD1-
linked ALS variants are susceptible to unfolding and loss of post-
translational modifications [21,22]. Indeed, loss of metal ions makes
SOD1 aggregation-prone in vitro and in disease models insoluble
SOD1 has a very low metallation [23,24]. SOD1-linked ALS is not a
loss-of-function disorder since transgenic animal models expressing
ALS-SOD1variants along with endogenous wild type SOD1 still develop
the disease. Likewise, there hasbeen a debate on the roleof the disulfide
crosslinking in SOD1 aggregation: surface-exposed reduced cysteines
could engage in intermolecular disulfide crosslinking reactions leading
to insoluble aggregates but this does not seem to be a critical factor, as
discussed in [21]. However, Cys6 and Cys111 become solvent-exposed
in the metal-free state [25] and modulate SOD1 aggregation [21],so
this question remains open and the involved mechanisms are still un-
clear. Nevertheless, the current scenario is that there may be a common
pathogenic pathwayfor sporadic and familial SOD1-linked ALS and that
this involves the aggregation of recently synthesized immature SOD1
proteins in the metal-free form [21,26]. Following this rationale, apo
SOD1 is, therefore, the most relevant experimental system to investi-
gate SOD1 aggregation mechanisms.
In this context, aberrant protein–metal interactions involving
deregulated metal ions in the nervous system are important modulators
of protein aggregation across most neurodegenerative diseases [6].Our
recent efforts have focused on the analysis of the effect of Ca
2+
over
SOD1 in the context of ALS and we have established that Ca
2+
promotes
SOD1 aggregation into non-fibrillar amyloid [27]. Using a suite of bio-
physical approaches we demonstrated that Ca
2+
induces conforma-
tional changes that increase SOD1 β-sheet content and advance the
onset of aggregation by decreasing the critical concentration and nucle-
ation time. Interestingly, Ca
2+
diverted SOD1 deposition from fibrils to-
wards amorphous aggregates consistingmostly of antiparallel β-sheets,
suggesting that Ca
2+
modulates the aggregation pathway of apo SOD1
[6]. Here, we further pursue the molecular details underlying Ca
2+
ef-
fects over apo SOD1 aggregation combining experimental and compu-
tational methodologies. In this study we ultimately reveal how Ca
2+
interactions recreate effects similar to those of genetic defects, a finding
with relevance to understanding sporadic ALS pathomechanisms.
2. Materials and methods
2.1. SOD1 purification
Human SOD1 was expressed in Escherichia coli BL21(DE3) cells and
purified as in [28], and the plasmid used as a kind gift of M. Oliveberg
(Stockholm University). All SOD1 experiments were performed with
the demetallated SOD1 form (apo SOD1). Preparation of apo SOD1
was made accordingly with previous published protocols [29]. Metal
content (Cu/Zn) of apo SOD1 was confirmed by the colorimetric Zincon
assay [30]. All solutions and buffers were passed through a chelex resin
(Bio-Rad) column to remove contaminant trace metals in order to
maintain apo SOD1 in the demetallated form throughout the experi-
ments. The concentration of SOD1 was determined spectrophotometri-
cally using the extinction coefficient 10,800 cm
−
1·M
−1
at 280 nm.
2.2. Isothermal titration calorimetry
The interaction of Ca
2+
with holo and apo SOD1 was analyzed using
isothermal titration calorimetry (ITC) on a Microcal ITC 200 calorimeter.
Titrations were performed at 37 °C by injecting 2 μl aliquots of a CaCl
2
so-
lution (6 mM) into SOD1 (apo and holo) (300 μM) in the sample cell.
Each injection was made over a period of 5 s with a 120 s spacing interval
between subsequent nineteen injections with a stirring at 1000 rpm.
SOD1 was previously extensively dialyzed against 50 mM HEPES
pH 7.5 buffer. HEPES was chosen because of its negligible interference
with Ca
2+
at the used pH. The SOD1–metal ion titration curves were
corrected using a Ca
2+
to buffer control titration. Data was analyzed
using Origin 7.0 software supplied by Microcal. The best fit parameters
for apo SOD1 were obtained from the sequential binding site model
where dependencies were b0.25 for the first binding site and b0.75 for
the second. In order to drive Ca
2+
titration on apo SOD1 to saturation
and provide reliable fitting parameters, four sequential ITC titrations
were merged with the program ConCat32. Before the beginning of each
sequential ITC titration the excess solution was removed from the over-
flowreservoirandthesameCaCl
2
solution was used in all four titrations.
2.3. Molecular dynamics simulations
The starting structure of the WT apo SOD1 homodimer was extract-
ed from the 2.4 Å crystal structure of the human Cu,Zn superoxide dis-
mutase with PDB accession code 1SPD [31] by removing the
corresponding metal ions, Cu
2+
and Zn
2+
. Apo SOD1 was hydrated
with ~15,000 TIP3P water molecules in a truncated octahedral box to
keep a water layer of at least 12 Å between periodic images of the pro-
tein. The protonationstates of all amino acids (at pH 7) were defined on
the basis of their pK
a
values predicted with the PROPKA 3.1 web server
(http://propka.ki.ku.dk/). The specific protonation states of the histi-
dines' imidazole rings (HIE or HID) were additionally predicted with
the “pdb2gmx”functionality of the GROMACS v4.5.4 suite of programs
[32,33] which performs an optimization of the network of H-bonds.
The negative charge on the protein was neutralized with Na
+
counter
ions and an excess of (i) 2 CaCl
2
and (ii) 4 CaCl
2
was added in order to
obtain two systems with concentrations of Ca
2+
equivalent to
(i) Ca
2+
:apo SOD1 = 2 and (ii) Ca
2+
:apo SOD1 = 4. The positions of
all ions were randomized using the “ptraj”module of the AMBER 11
suite of programs [39], ensuring that the Na
+
/Cl
−
or Ca
2+
ions were
at least 5/8 Å from apo SOD1 and 5 Å from each other. This precludes
the existence of artifacts caused by attraction from the protein closest
negatively/positively charged side-chains. Water molecules were treat-
ed using the TIP3P force field [34] and all protein interactions were de-
scribed by the AMBER ff99SB biomolecular force field [35,36]. Ions were
described using the TIP3P water-model-specific parameters developed
by Joung and Cheatham [37]. For the Ca
2+
ions we used the AMBER-
adapted (to additive combining rules) Åqvist parameters [38] included
in the parm99 parameter set of the AMBER force field. Both systems
containing Ca
2+
,Ca
2+
:apo SOD1 = 2 and Ca
2+
:apo SOD1 = 4, and a
Ca
2+
-free version, Ca
2+
:apo SOD1 = 0, were used as starting points
in MD simulations performed with the AMBER 11 package [39].
SHAKE was applied to constrain the bonds containing hydrogen there-
fore allowing the use of a 2 fstime-step. Simulations were carried out at
310 K and 1 atm using the weak temperature coupling algorithm and
the isotropic pressure scaling method with time constants of 5 ps. A di-
rect space cutoff of 8 Å was used between non-bonded atoms and the
long-range electrostatic interactions were calculated using the
particle-mesh Ewald method in conjunction with periodic boundary
conditions. All simulations started with a standard equilibration proto-
col. Each system was initially energy minimized with the steepest de-
scent method for 5000 steps. Subsequently, each system was
gradually heated to 310 K (for 40 ps) and further heated at 310 K for an-
other 10 ps at constant volume. In both minimization and heating
phases theprotein was keptunder positional restraints with forcecon-
stant of 25 kcal mol
−1
Å
−2
. The system was then equilibrated at constant
pressure for 700 ps using gradually weaker positional restraints (25.0, 5.0,
3.0, 1.0, 0.5, 0.1, and 0.05 kcal mol
−1
Å
−2
). A final equilibration step
consisting of an unconstrained constant-pressure MD run of 50 ns
119S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

followed. Four independent 100 ns long constant-pressure MD simula-
tions of each system were seeded from different snapshots extracted
from this 50 ns trajectory (at 35, 40, 45 and 50 ns) adding up to a total
simulation time of 400 ns per system. For analysis purposes, the last
40 ns of each system's MD trajectories were used.
Root-mean-square deviations (RMSD), root-mean-square fluctua-
tions (RMSF), Ca
2+
–amino acid binding frequencies, Ca
2+
spatial distri-
bution functions (SDFs), and SASA values reported for each system were
computed as averages over an ensemble of ~6400 configurations ex-
tracted from the four independent MD runs of each system with a
time interval of 25 ps.
The reported RMSF values were computed with the g_rmsf tool im-
plemented in GROMACS v4.5.4 [32,33] following the superposition of
the MD snapshots on the starting structure and removal of rotational
and translational degrees of freedom. The mean values of RMSF per
residue-Cαwere computed as averages over the ensemble of ~6400
configurations and the two homodimer subunits.
The spatial distribution functions (SDFs) of the Ca
2+
ions were com-
puted with the g_spatial tool implemented in GROMACS v4.5.4 [32,33].
Successive MD trajectory snapshots were translated and rotated (for an
optimal fitting to the original native structure) so as to minimize the
RMSD between the positions of the apo SOD1 dimer' Cαatoms. The
three-dimensional Ca
2+
density distributions around the apo SOD1
dimer were built by dividing the space in cubic cells of 0.5 Å and by cal-
culating their average occupancy. Simulation snapshots and SDF
isosurfaces were represented graphically using the software VMD.
To evaluate the Ca
2+
-protein binding frequencies, the amino acid
(in one of the homodimer units) closest to each Ca
2+
ion in each ana-
lyzed configuration was identified. The binding frequencies were ob-
tained as counts per amino acid normalized to the total number of
analyzed configurations times Ca
2+
concentration times the number
of units in the homodimer (i.e., 2).
SASA values were calculated in accordance with Connolly's algo-
rithm [40] by using the g_sas routine implemented in GROMACS
v4.5.4 [32,33] and vdW radii from the AMBER ff99SB force field [35,
36]. Reported SASAs represent mean values averaged over the ensemble
of ~6400 configurations extracted from four independent MD runs of
each system and over the two homodimer subunits.
Secondary structure assignments for the individual amino acid resi-
dues focused on beta-sheet content and were performed with the DSSP
algorithm [41]. Reported values depicted represent mean values aver-
aged over the ~6400 configurations extracted from four independent
MD trajectories.
3. Results and discussion
3.1. Ca
2+
interaction with apo SOD1 is entropy driven
We used isothermal titration calorimetry (ITC) to evaluate and char-
acterize the energetics of the SOD1:Ca
2+
interaction. The obtained data
indicates that the process of Ca
2+
binding to apo SOD1 is endothermic,
as shown from the positive peaks observed as apo SOD1 is titrated with
Ca
2+
(Fig. 1a). These peaks are intrinsic to the protein metal–ion inter-
action and do not result from the heat of dilution of CaCl
2
solution on
buffer, which is exothermic (not shown). Since holo SOD1 is not prone
to aggregate, we have used it as a control in a comparative analysis.
The ITC results of Ca
2+
binding to holo SOD1 indicate a low endother-
mic heat effect (Fig. 1b) which precludes fitting to binding models and
is suggestive of residual bindingto the fully metallated protein, via a dis-
tinct binding mechanism from that observed for apo SOD1. Therefore,
the effect of Ca
2+
is mostly significant for the metal-free form, in agree-
ment with the reported effects on its aggregation propensity [27].Anal-
ysis of apo SOD1 ITC data was carried out using a sequential binding
model for two sites and the obtained fitting for both sites indicate that
the favorable positive entropy variation of binding (ΔS1=26±1and
ΔS2 = 39 ± 8 cal/mol/deg) is sufficient to overcome the unfavorable in-
crease in enthalpy (ΔH1 = 1.7 ± 0.6; ΔH2 = 6.5 ± 2.0 kcal), an indica-
tion of hydrophobic and/or conformational changes occurring upon
Ca
2+
binding. Binding proceeds with low affinity (K
D1
=137±
96 μMandK
D2
= 2.4 ± 0.8 mM) but the aberrant interaction of Ca
2+
Fig. 1. ITC analysis of Ca
2+
interaction with apo SOD1 at 37 °C and pH 7.5. (a) The upper panel shows the titration raw data and the lowerpanel the integrated data andbest fit obtained
after subtracting theheat of dilution of the control. (b) Comparison of the integrated ITC thermogram of Ca
2+
bindingto apo SOD1 with theholo SOD1 protein,denoting residual bindingto
the latter. Representative of n = 3 isothermograms.
120 S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

with SOD1 in cells is likely considering the known intracellular Ca
2+
dysregulation in ALS [15,16] and high transient Ca
2+
concentrations
which can reach 200–300 μM during neurotransmitter release [42].
These results suggest that Ca
2+
interaction with apo SOD1 is associated
with a system of entropy-driven “non-specific”interactions, as defined
in [43]. Interestingly, α-synuclein which is another neurodegeneration
related protein whose aggregation is also promoted by Ca
2+
[44],
binds Ca
2+
with a similar endothermic profile as the one here reported
for SOD1 [45].
3.2. Native state dynamics of apo SOD1 is mildly enhanced in the presence
of Ca
2+
In order to gain microscopic insight into theinteraction of Ca
2+
with
apo SOD1 we conducted a series of explicit-solvent molecular dynamics
(MD) simulations starting from the native structure of holo SOD1
(Fig. 2a). To take into account putative effects dependent of Ca
2+
levels,
we considered three Ca
2+
–protein systems equivalent to Ca
2+
:apo
SOD1 = 0, Ca
2+
:apo SOD1 = 2, and Ca
2+
:apo SOD1 = 4. The highest
ratio corresponds to a near-saturation effect over observed secondary
structure changes [27] and is within the physiological ranges of
SOD1 and Ca
2+
[26,42]. We started by investigating protein flexibility
changes induced by Ca
2+
by computing the C
α
atom-positional root-
mean-square fluctuations (RMSF) along the amino acid sequence
(Fig. 2b). Noticeably, the Zn
2+
-binding and electrostatic loops (blue
and orange regions in Fig. 2) display the higher flexibilities across all
tested conditions, in agreement with the acknowledged stabilizing
role of metal ions in both dimeric and monomeric SOD1 [46,47]. The
gain in flexibility and relative disordering upon Ca
2+
addition, especial-
ly of the Zn
2+
-binding loop, naturally leads to an increase in conforma-
tional entropy. This observation partly rationalizes the ITC results,
which are consistent with entropy-driven Ca
2+
–protein interactions.
Next we analyzed Ca
2+
-driven conformational changes of apo SOD1,
by evaluating the root-mean-square deviation (RMSD) with respect to
the original crystallographic structure. We observe an overall increase
of the mean CαRMSD up to 17% (Table 1), which indicates loss of native
structure upon Ca
2+
addition. In particular, the electrostatic loop dis-
plays the highest mean CαRMSD in all studied systems, in line with ex-
perimental data [25]. Unsurprisingly, the increase in RMSD is
particularly pronounced for the flexible Zn
2+
-binding (IV, 18–47%)
and electrostatic (VII, up to 13–18%) loops. Indeed, decoupling of the
Zn
2+
-binding and electrostatic loops is a conformational change exhib-
ited by many pathogenic fALS-associated SOD1 mutants with compro-
mised metal binding affinities [25,47,48],aswellasdemetallatedWT-
SOD1 [46]. Also, it is widely accepted that decoupling of one of those
loops exposes the central SOD1 β-barrel thus triggering aggregation
[46,47]. Interestingly, the presence of Ca
2+
ions increases CαRMSD
values for loops III (up to 10%) and V (up to 13%), denoting a structural
perturbation of the SOD1 β-plug region [49,50].
We then analyzed variations in the distances between some struc-
tural elements in the different SOD1 systems. Analysis of the segments
that define the β-plug region connecting strands β
4
and β
5
reveals
that the mean distance between their centers of mass is ~ 16% higher
than in the original X-ray holo structure. This variation is observed
across the three studied systems. Likewise, the mean distance between
the centers of mass of strands β
5
and β
6
is similar across the three sys-
tems and represents an increase of ~ 11% with regard to the original
holo structure. Interestingly, the de-protection of the edge strands β
5
and β
6
is a known trigger of SOD1 aggregation [50] as this results in a
transient opening of the SOD1 β-barrel andsolvent-exposure of the pro-
tein hydrophobic core. Altogether, this mean-distances analysis pro-
vides evidence for a mild opening of the β-barrel in the three systems.
3.3. Ca
2+
interactions occur at preferential hotspots within apo SOD1 loops
A major advantage of molecular simulations is the possibility to iso-
late every single conformation adopted by the protein with a remark-
ably high temporal resolution. This advantage becomes particularly
handy to address structural challenges such as those resulting from
the distribution of Ca
2+
around apo SOD1. In particular, one can deter-
mine preferential binding spots for the interaction between Ca
2+
ions
and the protein, thus complementing the information provided by ITC
experiments. Three preferential binding regions for Ca
2+
have been
identified in apo SOD1: i) positions within the Zn
2+
binding loop; ii)
on the region encompassing strands β
5
and β
6
plus connecting loop V,
and iii) in the Greek key loop (Fig. 3 and Table 2).
In the first region, within the Zn
2+
binding loop, the association of
Ca
2+
to apo SOD1 is made through residues Glu49, Asp76, Glu77, and
Glu78 in both systems (Fig. 3, orange). In the second region,
encompassing strands β
5
and β
6
plus connecting loop V, Ca
2+
binds to
Asp90 in both systems and differentially to Asp96 and Asp83 (Fig. 3,
pink). Finally, positions in the Greek key loop (VI) comprise amino
acid Asp109, which is a high-frequency binding spot for Ca
2+
in all sim-
ulated conditions, as well as residues Asp101 and Ser102 (Fig. 3, green).
Interestingly, in the crystallographic structure of the pathogenic SOD1
variant H46R/H48Q a Ca
2+
ion, which was presumed to be a crystalliza-
tion artifact, was found to be coordinated to Ser102 and Asn26 [51],two
sites here identified as preferential interaction partners for Ca
2+
(Table 2).
Fig. 2. Assessing Ca
2+
-induced flexibility changes. Cartoon representation of the holo
SOD1 homodimer (PDB ID: 1SPD) (a) and positional root-mean-square fluctuations
(RMSF) of the backbone C
α
atoms around their average positions for the apo SOD1 sys-
tems studied (b). A larger RMSF value indicates higher flexibility. The shaded areas
stand for the standard deviation bars corresponding to differences of RMSF per residue-
C
α
among the four individual trajectories andthe two homodimersubunits. The colored
bars signal the loop regions highlighted in (a). The arrows illustrate the location of β-
strands along the primary sequence. In the RMSF panels of the systems containing Ca
2+
the mean RMSF of the apoSOD1 system is also represented for comparison (black lines).
121S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

3.4. Ca
2+
binding to apo SOD1 triggers hydrophobicity changes favoring
aggregation traits
Next we evaluated variations on solvent-accessible surface areas
(SASA) per residue upon Ca
2+
binding, to infer on the exposure of hy-
drophobic patches, which are known predisposition factors for aggrega-
tion [52].Wefirst compared the behavior of the three systems with
respect to holo SOD1, and then we analyzed effects of Ca
2+
binding
with respect to the apo system with no Ca
2+
.
The first analysis shows that there is an overall in crease of the hydro-
phobicity of the apo and the Ca
2+
loaded SOD1 systems relatively to the
original holo SOD1. In particular, some residues composing the hydro-
phobic core become solvent-exposed upon Ca
2+
binding (Supplemen-
tary Table 1). The Zn
2+
-binding loop, the electrostatic loop, the
Table 1
Mean CαRMSDs of the different systems from the native holo SOD1 structure.
System
(Ca
2+
:SOD1)
Global
(Å)
Loop II
(Å)
Loop III
(Å)
Loop IV
(Å)
Loop V
(Å)
Loop VI
(Å)
Loop VII
(Å)
0 2.80 ± 0.26 –2.38 ± 0.59 –1.65 ± 0.42 –2.80 ± 1.39 –2.42 ± 0.52 –1.65 ± 0.32 4.70 ± 1.95 –
2 3.27 ± 0.54 14% 2.36 ± 0.58 1% 1.82 ± 0.44 10% 3.30± 1.69 18% 2.54 ± 0.60 5% 1.65 ± 0.39 0% 5.56 ± 2.86 18%
4 3.36 ± 0.70 17% 2.54 ± 0.60 7% 1.80 ± 0.43 9% 4.12 ± 2.00 47% 2.73 ± 0.60 13% 1.67 ± 0.36 1% 5.30 ± 2.25 13%
Values obtained for ensembles of ~6400 configurations extracted from four independent MD runs of each system after fitting the protein Cαatoms to the native structure. Relative var-
iations withrespect to the Ca
2+
:apo SOD1 = 0 systemare indicated as percentages. The loopregions have been definedtaking into consideration the DSSP-derived [41] secondary struc-
ture assignment of the first chain(subunit) of the holoSOD1 homodimer asfollows: loop II (residues 22–29), loop III (residues 34–40), loop IV (residues 49–84),loop V (residues 90–96),
loop VI (residues 100–115) and loop VII (residues 121–142).
Fig. 3. Distributionof Ca
2+
ions aroundthe apo SOD1 dimer.Three-dimensional mappingof the spatial distribution function and Ca
2+
bindingfrequencies in the Ca
2+
:apo SOD1= 2 (a,b)
and Ca
2+
:apo SOD1 = 4 (c, d) systems. Inthe three-dimensional mapping, isosurfaces correspond to ion number densities of 0.0012 (blue)/0.0024 (white)/0.0036 (red) ions/Å
3
(Ca
2+
:
apo SOD1 = 4) and 0.0025 (blue)/0.005 (white)/0.0075 (red) ions/Å
3
(Ca
2+
:apo SOD1 = 2). The mean value (and standard deviation) of the distance of each amino acid to Ca
2+
(brCa
2+
N) is also reported in the frequency panels.
122 S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

extended β-plug region (including residues in strands β
4
and β
5
),
strand β
6
, and the adjoining Greek key loop display the higher SASA
values in the three investigated systems (Fig. 4a, shaded areas). The
results of SASA enhancement of residues within the β-plug are in line
with those reported for MD simulations of the apo and Zn
2+
-loaded
SOD1 forms [50] which showed that strands β
4
and β
5
become more ex-
posed in the apo-enzyme and, therefore readily available to establish
aberrant intermolecular interactions. Also, the enhanced SASAs of resi-
dues Asp83,Ala89, Gly93, and Ala95 in the region encompassingstrands
β
5
and β
6
are compatible with the predicted availability of these strands
to establish intermolecular interactions leading to amyloid fibril forma-
tion in fALS-associated mutants [53,54] or apo WT-SOD1 [55].
The second analysis focused on SASA changes upon Ca
2+
binding
with respect to the apo SOD1 system with no calcium. Analysis of the
relative variations shows that Ca
2+
enhances the SASAs of residues in
the Zn
2+
-binding and electrostatic loops and in strand β
6
and in the
Greek key loop. This effect is more pronounced for the system with
the higher ratio Ca
2+
:apo SOD1 (Fig. 4b). Also for this case,the presence
of Ca
2+
mildly enhances the exposure of hydrophobic core residues by
up to 9%, which is in line with the observed enhancement of ANS fluo-
rescence in emission experiments upon incubation of SOD1 with Ca
2+
[27].
Overall, disordering of the electrostatic and Zn
2+
-binding loops in-
duced by Ca
2+
binding to apo SOD1 (Table 1 and Fig. 4b) potentially de-
couples strands β
5
and β
6
which become available to establish
intermolecular interactions with other edge strands or metal-binding
loops [53,55]. It is important to notice that interactions involving resi-
dues on both edge strands and connecting loop play a crucial role in
the proper folding of the enzyme [56]. It is also noteworthy that strand
β
6
and the adjacent Greek key loop, are inserted in one of the three key
regions of SOD1 found in the core of fibrillar aggregates that comprises
residues 90 to 120 [57]. In particular, these structural elements are part
of the Asp101-Ser107 stretch of SOD1 which has been experimentally
identified as a fibril-forming segment [58] as well as of one of the
three SOD1 regions predicted to be highly aggregation-prone according
to the WALTZ algorithm [59], specifically the Ala95-Gly114 segment.
3.5. Ca
2+
increases β-sheet content and counteracts SOD1 aggregation
gatekeepers
We then investigated theeffects of Ca
2+
binding on the β-sheet con-
tent of apo SOD1, a well-established determinant of protein aggregation
[52], as well as impacts on residues located at the boundary of
aggregation-prone regions, the so-called aggregation gatekeepers [60,
61].
The results obtained point to a correspondence between the increase
in β-sheet content and some of the preferential Ca
2+
locations (Fig. 5).
In the systems containing Ca
2+
there is a significant increase of the
mean β-sheet content (ranging from 11 to 14%) in the region
encompassing strands β
5
and β
6
(residues 83–101) (Fig. 5b). Associa-
tion to Ca
2+
has a structuring effect over strands β
5
and β
6
that is espe-
cially pronounced for residues 83–84, 88–89 and 97–99. Although less
significant, there is also an increase of the mean β-sheet content in
strands β
1
and β
4
(Fig. 5b). These results are in linewith the previously
identified experimental correlation between the presence of Ca
2+
ions
and an enhanced content of SOD1 β-sheets [27]. The aforementioned
strands are adjacent to some of the preferential hotspots for Ca
2+
bind-
ing: Asp11 (loop I, near strand β
1
), Glu24 (loop II,near strand β
2
), Glu40
(loop III, near strand β
4
), Asp90 (loop V, near strand β
5
), Asp96 (loop V,
near strand β
6
), Asp101 (loop VI, near strand β
6
), and Asp109 (loop VI,
near strand β
6
)(Fig. 5).
It is important to emphasize the ubiquitous participation of Asp res-
idues in interactions with Ca
2+
, since it has been previously shown that
the neutralization or replacement of Asp (or Asn) residues in a model
peptide leads to a dramatic increase in its β-sheet content. The higher
content of β-sheet structure was furthermore shown to be correlated
with an increased propensity for fibril formation and decreased solubil-
ity at neutral pH [62]. It is known that Asp residues display a consider-
ably low propensity to adopt β-strand conformations, having instead a
Table 2
Most frequent pairings of Ca
2+
-amino acids for the different systems studied.
Ca
2+
:apo SOD1 = 2 Ca
2+
:apo SOD1 = 4
Residue Frequency
(1 × 10
−2
)
brCa
2+
N
(Å)
Frequency
(1 × 10
−2
)
brCa
2+
N
(Å)
Asp11 1.72 6.73 1.48 5.50
Glu24 ––1.60 5.83
Asn26 ––0.90 8.69
Glu40 1.54 7.72 2.46 2.96
Zinc loop (IV)
Glu49 2.03 2.76 7.25 2.61
Asn65 0.59 3.03 ––
Pro66 0.19 5.10 ––
Asp76 1.96 2.63 1.86 2.63
Glu77 3.99 4.93 5.61 2.86
Glu78 2.97 3.32 5.60 2.73
β5/loop V/β6
Asp83 ––0.85 2.60
Asp90 1.68 4.25 1.75 3.85
Asp96 2.21 4.09 ––
Greek key (VI)
Glu100 0.30 9.96 ––
Ser102 0.69 4.74 ––
Asp101 ––2.27 2.58
Asp109 12.05 2.92 6.60 2.91
Glu131 0.24 7.25 ––
Glu132 5.20 4.07 0.83 6.54
Glu133 ––0.59 4.31
Q153 (C-term) ––4.18 3.37
The frequency at which each amino acid–Ca
2+
pairing occurs was calculated for ~6400
configurations extracted from four independent MD runs of each system. For each config-
uration of each system we computed all the distances from eachCa
2+
ion to every amino
acid (irrespective of the homodimer subunit where it is located)and registered the amino
acid for which that distance is a minimum.Frequency values were obtained as counts per
amino acid normalized by the number of configurations times Ca
2+
concentration (2 or
4) times the number of monomers in the homodimer (2). Only the amino acids for
which the mean distance to Ca
2+
is inferior to 10 Å are shown.
Fig. 4. Assessment of hydrophobicitychanges upon Ca
2+
binding.Mean SASAs per residue
in apo SOD1 (a) wherethe thin black line depicts the SASA of the original holo SOD1X-ray
structure (PDB ID: 1SPD)obtained as a meanover the two subunits of the homodimer. The
shadedareas stand for the standarddeviationbars. The differencebetween the mean SASA
per residue in eachof the systems containingCa
2+
and the systemwith no Ca
2+
is repre-
sented in (b). Mean SASA ratios of some hydrophobic core residues to the corresponding
residues in holo SOD1 are indicated in Supplementary Table 1.
123S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

unique ability to form pseudo-turns (the side-chain carboxyl group
forms a H-bond with the main-chain amide proton of the n + 2 residue)
[63]. We observe that Ca
2+
ions located within interaction distance to
residues Asp90, Asp96 and Asp101 cause an extension of strands β
5
and β
6
by inducinga change in the conformationalpreference and back-
bone flexibility of those residues and/or adjacent ones (Fig. 5). These
negatively-charged Asp residues are strategically placed at the extrem-
ities of strands β
5
and β
6
(Fig. 5a), which are particularly prone to edge-
to-edge, β-sheet-to-β-sheet association given their locationat the edges
of the two β-sheets defining the SOD1 β-barrel. Those residues are thus
regarded as aggregation gatekeepers [64] and their action complements
the protecting role played by the Zn
2+
-binding and electrostatic loops
which partially conceal the edges of the β-barrel [64]. Also Asp11,
Glu24 and Glu40 located at the extremities of strands β
1,
β
2
and β
4
are considered aggregation gatekeepers. Indeed, the two former are lo-
cated in the vicinity of one of the WALTZ predicted aggregation-prone
regions of SOD1, specifically the segment Val14-Asn22. These structural
features, which concur to prevent intermolecular interactions between
edge strands, are consistent with the strategies adopted by normal β-
sheet-containing proteins to avoid self-association [65].
The effects here reported have important implications in the molec-
ular understanding of the aggregation process of SOD1 in the presence
of Ca
2+
and of how Ca
2+
interactions recreate effects resembling
those of genetic defects. The binding of Ca
2+
in the cleft created by
strands β
5
and β
6
will shield electrostatic repulsions removing the ag-
gregation protection afforded by the negatively-charged gatekeeper
residues. This will favor aggregation-prone interactions and self-
assembly of amyloids. Indeed, strands β
5
and β
6,
as shown here as
well as by others [53–55], become more likely to participate in intermo-
lecular interactions upon demetallation. In agreement, these strands
and the Zn
2+
-binding and electrostatic loop are involved in non-
native interactions in amyloid-like arrangements of SOD1 dimers in
the X-ray structures of the fALS-associated metal binding-site SOD1 mu-
tants S134N and H46R [53].
In the presence of Ca
2+
the molecular scrambling imposed by metal
ligation to gatekeeper positions is likely to influence the aggregation
pathway resulting in mixed populations of SOD1 amyloidogenic con-
formers, as observed experimentally [27]. This finding is relevant for
ALS pathomechanisms, as an extensive bioinformatic survey of fALS-
associated SOD1 mutations revealed that these pathogenic mutations
occur preferentially on residues with negatively-charged side chains
therefore contributing to weaken electrostatic repulsions between the
negatively-charged individual SOD1 molecules [66,67]. Indeed, the
Ca
2+
binding positions identified in this study have corresponding mu-
tations in familial ALS which also result in the abolishment of charges
(Fig. 5aandTable 3). Such naturally occurring mutations actually back
up our observations: removal of gatekeeper residues does result in an
increased aggregation propensity and probably the reason why we do
not find combinations of disease-related SOD1 mutations in these resi-
dues results from the fact that such multiple changes would have an ex-
tremely deleterious effect on protein structure and on broadening the
Fig. 5. Impact of Ca
2+
binding on the secondary-structure content of apo SOD1 and mapping to gatekeeper residues. Impact of Ca
2+
on the secondary-structure content of apo SOD1
(a) where the black curve represents the β-sheet content of apo SOD1 in the system containing no Ca
2+
.(b)Δβ-Sheet represents the difference between the β-sheet content (in %) of
the systems containing Ca
2+
and the system with no Ca
2+
. Gray bars in (a) and (b) indicate regions with significant Waltz scores (≥80%) [59], whereas dark gray bars in (b) highlight
experimentally validated aggregation prone segments [58]. (c) Cartoon representation of an apo SOD1 monomer extracted from the MD ensemble of Ca
2+
:apo SOD1 = 4 configurations
where gatekeeper residues adjacentto SOD1 β-strands withsignificant frequencies of association to Ca
2+
are highlighted, alongwith fALS mutationsidentified on the same sites(see also
Table 3).
Table 3
Residues with high frequencies of Ca
2+
binding overlap with fALS mutations.
Amino acid Gatekeepers fALS mutation Structural location
I Asp109 Yes Asp109Tyr Greek key (loop VI)
Glu49 No Glu49Lys ZINC LOOP (IV)
Glu77 No n.a. Zinc loop (IV)
Glu78 No n.a. Zinc loop (IV)
Glu132 No n.a. Electrostatic loop (VII)
Q153 (C-term) Yes n.a. C-term/fibril forming
II Glu40 Yes Glu40Gly Loop III/β4 edge
Asp101 Yes Asp101 NGly/Asn/Tyr Greek key VI/β6 edge
Asp96 Yes Asp96 NVal/Asn Loop V/β6 edge
Asp76 No Asp76 NTyr/Val Zinc loop IV
Asp90 Yes Asp90 NVal/Ala Loop V/β5 edge
Asp11 Yes Asp11Tyr NAla Loop I/β1 edge
Glu24 Yes n.a. Loop II/β2 edge
Group I residues have high frequency of Ca
2+
binding (ranging from 0.1205 to 0.0418),
with an average distance of 3.1 Å. Group IIcorresponds to those with medium frequency
of Ca
2+
binding (ranging from 0.0246 to 0.0172) with an average distance of 4.1 Å. Gate-
keepers are defined as charged residues at the edges of aggregation-prone regions. See
Table 2 for further details. fALS mutations were extracted from the ALSoD database
(http://alsod.iop.kcl.ac.uk/). n.a. data not available.
124 S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

conformational landscape of the nascent immature apo-SOD1 proteins
favoring aggregation. Yet, our results indicate that the association of
Ca
2+
to wild typeSOD1 reproduces theeffects of mutations but through
rather subtle molecular effects, thus providing a molecular rationale
that explains the effects of metal ligation of the SOD1 aggregation
pathway.
4. Conclusions
In recent years there has been accumulating evidence that protein
aggregation cascades can be triggered by native-like conformational
states en route to folding [68–73] or even by locally unfolded conforma-
tions resulting from structural fluctuations of the native structure [74].
Many of these conformational changes result from protein–ligand inter-
actions among which metal ions play an important role, especially in
toxic gain-of-function neurodegenerative diseases that involve protein
deposition, considering their significant deregulation across different
pathologies [6]. Such effects have been recently describedin association
with the influence of Ca
2+
on the aggregation of apo SOD1. Specifically,
Ca
2+
was shown to promote the aggregation of apo SOD1 into non-
fibrillar amyloid by inducing conformational changes [27] consistent
with the occurrence of aggregation-prone conformational states.
The investigation of the molecular details underlying how Ca
2+
in-
teracts with apo SOD1 via molecularsimulation in combination with ex-
perimental approaches has elicited the occurrence of conformational
events at the level of the native structure of apo SOD1 that are consis-
tent with an increase of the protein aggregation potential upon Ca
2+
binding. Explicit-solvent molecular dynamics simulations of apo SOD1
systems with Ca
2+
at up to a molar ratio of 4 indicate an enhancement
of native state dynamics upon Ca
2+
binding concomitant with changes
on the β-sheet content. Interestingly, among residues affected by Ca
2+
,
two major groups can be identified which seem to be related to aggre-
gation traits at different molecular levels (Table 3). One group of resi-
dues drives effects which are more prominent within the Zn
2+
binding and electrostatic loops, as in many pathogenic fALS-associated
mutants with compromised metal binding affinities [25,47,48].Indeed
loops are hotspots for Ca
2+
binding induced conformational changes
that also result in increased solvent-accessible surface areas of amino
acids inserted within the protein hydrophobic core andon the boundary
strand β
6
within the SOD1 β-barrel and adjacent Greek key loop. We
used ITC to test if Ca
2+
would prevent Zn
2+
binding to the SOD1 zinc
site and the results obtained suggest that this is not the case: binding
of Zn
2+
still takes place but at a lower affinity (K
D
~70nMversus
0.1 nM as determined in [29]). This agrees with our MD observations
that Ca
2+
binding induces conformational changes within the zinc-
loop, which end up also making SOD1 zinc-metallation less favorable.
The second group comprises residues which are mainly found nearby
SOD segments shown to be fibril-forming both through experimental
[58] and computational [59] studies. These charged residues have
been regarded as aggregation gatekeepers as repulsive interactions
minimize intermolecular contacts between nearby strands that would
otherwise form beta-structure fibrillar aggregates. Interestingly, most
of these residues are found mutated in familial ALS [75], which consti-
tutes a form of validation of our results in respect to the disease
mechanism.
Indeed, our data provides a molecular rationale for the previous
observation that the presence of Ca
2+
promotes the aggregation into
non-fibrillar amyloid. In particular, Ca
2+
-binding to negatively charged
gatekeepers has the potential to shield repulsive interactions facilitating
the self-association of apo SOD1. Nevertheless, the presence of Ca
2+
in
between the stacking β-sheets made up by fibril-prone segments may
block its organized self-assembly into fibrils. This molecular scrambling
imposed by metal ligation diverts the aggregation pathway from fibrils
into amyloid oligomers and agrees with experimental evidence pointing
to the population of aggregation-prone apo SOD1 conformers upon ad-
dition of Ca
2+
under physiological conditions [27]. The relevant
implications for disease is that in general amyloidogenic oligomers are
known to be more cytotoxic than fibrils, and are therefore potentially
more deleterious in the ALS neurodegeneration process. We speculate
that a similar mechanism involving metal–ion interactions with gate-
keeping residues is significant in many amyloid neurodegenerative dis-
eases, which commonly engage the aggregation of proteins with metal
binding properties.
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.bbapap.2014.11.005.
Acknowledgments
This work was supported by the Fundação para a Ciência e
Tecnologia (FCT/MCTES, Portugal) through: research grants PTDC/QUI-
BIQ/117789/2010 (to CMG), PTDC/FIS/113638/2009 (to PFNF), post-
doctoral fellowship SFRH/BPD/46313/2008 to SGE, post-doctoral fel-
lowship SFRH/BPD/47477/2008 to SSL, and by the strategic grants
PEst-OE/EQB/LA0004/2011 (to the ITQB Laboratório Associado) and
Pest-OE/FIS/UI0261/2014 (to CFMC).
References
[1] K. Jellinger, Interaction between pathogenic proteins in neurodegenerative disor-
ders, J. Cell. Mol. Med. 16 (2012) 1166–1183.
[2] M. Jucker, L. Walker, Self-propagation of pathogenic protein aggregates in neurode-
generative diseases, Nature 501 (2013) 45–51.
[3] K. Barnham, A. Bush, Metals in Alzheimer's and Parkinson's diseases, Curr. Opin.
Chem. Biol. 12 (2008) 222–228.
[4] K.A. Jellinger, The relevance of metals in the pathophysiology of neurodegeneration,
pathological considerations, Int. Rev. Neurobiol. 110 (2013) 1–47.
[5] S. Pfaender, A.M. Grabrucker, Characterization of biometal profiles in neurological
disorders, Metallomics 6 (2014) 960–977.
[6] S.S. Leal, H.M. Botelho, C.M. Gomes, Metal ions as modulators of protein conforma-
tion and misfolding in neurodegeneration, Coord. Chem. Rev. 256 (2012)
2253–2270.
[7] M.Savelieff,S.Lee,Y.Liu,M.Lim,Untanglingamyloid-β,tau,andmetalsin
Alzheimer's disease, ACS Chem. Biol. 8 (2013) 856–865.
[8] L. Breydo, V. Uversky, Role of metal ions in aggregation of intrinsically disordered
proteins in neurodegenerative diseases, Metallomics 3 (2011) 1163–1180.
[9] L. Siklos, J.I. Engelhardt, M.E. Alexianu, M.E. Gurney, T. Siddique, S.H. Appel, Intracel-
lular calcium parallels motoneuron degeneration in SOD-1 mutant mice, J.
Neuropathol. Exp. Neurol. 57 (1998) 571–587.
[10] F. von Lewinski, J. Fuchs, B.K. Vanselow, B.U. Keller, Low Ca
2+
buffering in hypoglos-
sal motoneurons of mutant SOD1 (G93A)mice, Neurosci. Lett. 445 (2008) 224–228.
[11] M.K. Jaiswal, B.U. Keller, Cu/Zn superoxide dismutase typical for familial amyotro-
phic lateral sclerosis increases the vulnerability of mitochondria and perturbs
Ca
2+
homeostasis in SOD1G93A mice, Mol. Pharmacol. 75 (2009) 478–489.
[12] L.G. Bilsland, N. Nirmalananthan, J. Yip, L. Greensmith, M.R. Duchen, Expression of
mutant SOD1 in astrocytes induces functional deficits in motoneuron mitochondria,
J. Neurochem. 107 (2008) 1271–1283.
[13] H. Kawamata, G. Manfredi, Mitochondrialdysfunction and intracellularcalcium dys-
regulation in ALS, Mech. Ageing Dev. 131 (2010) 517–526.
[14] L. Siklos, J. Engelhardt, Y. Harati, R.G. Smith, F. Joo, S.H. Appel, Ultrastructural evi-
dence for altered calcium in motor nerve terminals in amyotropic lateral sclerosis,
Ann. Neurol. 39 (1996) 203–216.
[15] J. Grosskreutz, L. Van Den Bosch, B.U. Keller, Calcium dysregulation in amyotrophic
lateral sclerosis, Cell Calcium 47 (2010) 165–174.
[16] P. Marambaud, U. Dreses-Werringloer, V. Vingtdeux, Calcium signaling in neurode-
generation, Mol. Neurodegener. 4 (2009) 20.
[17] M. Tateno, H. Sadakata, M. Tanaka, S. Itohara, R.M. Shin, M. Miura, M. Masuda, T.
Aosaki, M.Urushitani, H. Misawa,R. Takahashi, Calcium-permeable AMPA receptors
promotemisfolding of mutant SOD1protein and developmentof amyotrophic later-
al sclerosis in a transgenic mouse model, Hum. Mol. Genet. 13 (2004) 2183–2196.
[18] M.L. Tradewell, L.A. Cooper, S. Minotti, H.D. Durham, Calcium dysregulation, mito-
chondrial pathology and protein aggregation in a culture model of amyotrophic lat-
eral sclerosis: mechanistic relationship and differential sensitivity to intervention,
Neurobiol. Dis. 42 (2011) 265–275.
[19] J.S. Valentine, P.A. Doucette, S. Zittin Potter, Copper–zinc superoxide dismutase and
amyotrophic lateral sclerosis, Annu. Rev. Biochem. 74 (2005) 563–593.
[20] Y. Furukawa, T.V. O'Halloran, Amyotrophic lateral sclerosis mutations have the
greatest destabilizing effect on the Apo- and reduced form of SOD1, leading to
unfolding and oxidative aggregation, J. Biol. Chem. 280 (2005) 17266–17274.
[21] Y. Sheng, M. Chattopadhyay, J. Whitelegge, J.S. Valentine, SOD1 aggregation and
ALS: role of metallation states and disulfide status, Curr. Top. Med. Chem. 12
(2012) 2560–2572.
[22] Y. Furukawa, K. Kaneko, K. Yamanaka, T.V. O'Halloran, N. Nukina, Complete loss of
post-translationalmodifications triggers fibrillar aggregation of SOD1 in the familial
form of amyotrophic lateral sclerosis, J. Biol. Chem. 283 (2008) 24167–24176.
125S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126

[23] P.A. Jonsson, K.S. Graffmo, P.M. Andersen, T. Brannstrom, M. Lindberg,M. Oliveberg,
S.L. Marklund, Disulphide-reduced superoxide dismutase-1 in CNS of transgenic
amyotrophic lateral sclerosis models, Brain 129 (2006) 451–464.
[24] H.L. Lelie, A. Liba, M.W. Bourassa, M. Chattopadhyay, P.K. Chan, E.B. Gralla, L.M.
Miller, D.R. Borchelt, J.S. Valentine, J.P. Whitelegge, Copper and zinc metallation sta-
tus of copper–zinc superoxide dismutase from amyotrophic lateral sclerosis trans-
genic mice, J. Biol. Chem. 286 (2011) 2795–2806.
[25] L. Banci, I. Bertini, M. Boca, V. Calderone, F. Cantini, S. Girotto, M. Vieru, Structural
and dynamic aspects related to oligomerization of apo SOD1 and its mutants,
Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 6980–6985.
[26] L. Banci, I. Bertini, A. Durazo, S. Girotto, E.B. Gralla, M. Martinelli, J.S. Valentine, M.
Vieru, J.P. Whitelegge, Metal-free superoxide dismutase forms soluble oligomers
under physiological conditions: a possible general mechanism for familial ALS,
Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 11263–11267.
[27] S.S. Leal, I. Cardoso, J.S. Valentine, C.M. Gomes, Calcium ions promote superoxide
dismutase 1 (SOD1) aggregation into non-fibrillar amyloid: a link to toxic effects
of calcium overload in amyotrophic lateral sclerosis (ALS)? J. Biol. Chem. 288
(2013) 25219–25228.
[28] I.M. Ahl, M.J. Lindberg, L.A. Tibell, Coexpression of yeast copper chaperone (yCCS)
and CuZn-superoxide dismutases in Escherichia coli yields protein with high copper
contents, Protein Expr. Purif. 37 (2004) 311–319.
[29] S.Z. Potter, H. Zhu, B.F. Shaw, J.A.Rodriguez, P.A. Doucette, S.H. Sohn, A. Durazo, K.F.
Faull, E.B. Gralla, A.M. Nersissian, J.S. Valentine, Binding of a single zinc ion to one
subunit of copper–zinc superoxide dismutase apoprotein substantially influences
the structure and stability of the entire homodimeric protein, J. Am. Chem. Soc.
129 (2007) 4575–4583.
[30] C.E. Sabel, J.M. Neureuther, S. Siemann, A spectrophotometric method for the deter-
mination of zinc, copper, and cobalt ions in metalloproteins using Zincon, Anal.
Biochem. 397 (2010) 218–226.
[31] H.X.Deng,A.Hentati,J.A.Tainer,Z.Iqbal,A.Cayabyab,W.Y.Hung,E.D.Getzoff,P.Hu,B.
Herzfeldt,R.P.Roos,C.Warner,G.Deng,E.Soriano,C.Smyth,H.E.Parge,A.Ahmed,A.D.
Roses, R.A. Hallewell, M.A. Pericakvance, T. Siddique, Amyotroph ic-lateral-sclerosis and
structural defects in Cu,Zn superoxide-dismutase, Science 261 (1993) 1047–1051.
[32] D. Van Der Spoel, E. Lindahl, B. Hess, G. Groenhof, A.E. Mark, H.J.C. Berendsen,
GROMACS: fast, flexible, and free, J. Comput. Chem. 26 (2005) 1701–1718.
[33] S. Pronk, S. Páll, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, M.R. Shirts, J.C.Smith,
P.M. Kasson, D. van der Spoel, B. Hess, E. Lindahl, GROMACS 4.5: a high-throughput
and highly parallel open source molecular simulation toolkit, Bioinformatics 29
(2013) 845–854.
[34] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L. Klein,Comparison of
simple potential functions for simulating liquid water, J. Chem. Phys. 79 (1983)
926–935.
[35] V. Hornak, R. Abel, A. Okur, B. Strockbine, A. Roitberg, C. Simmerling, Comparison of
multiple Amber force fields anddevelopment of improved proteinbackbone param-
eters, Proteins 65 (2006) 712–725.
[36] L. Wickstrom, A. Okur, C. Simmerling, Evaluating the performance of the ff99SB
force field based on NMR scalar coupling data, Biophys. J. 97 (2009) 853–856.
[37] I.S. Joung,T.E. Cheatham, Molecular dynamics simulations of the dynamic and ener-
getic properties of alkali andhalide ions using water-model-specific ion parameters,
J. Phys. Chem. B 113 (2009) 13279–13290.
[38] J. AQVIST, Ion water interaction potentials derived from free-energy perturbation
simulations, J. Phys. Chem. 94 (1990) 8021–8024.
[39] D.A. Case, T.A. Darden, T.E. Cheatham III, C.L. Simmerling,J. Wang, R.E. Duke, AMBER
11, University of California, San Francisco, 2011.
[40] M. Connolly, Analytical molecular surface calculation, J. Appl. Crystallogr. 16 (1983)
548–558.
[41] W. Kabsch, C. Sander, Dictionary ofprotein secondary structure: pattern recognition
of hydrogen-bonded and geometrical features, Biopolymers 22 (1983) 2577–2637.
[42] H. Tamano,A. Takeda,Dynamic action of neurometals at the synapse,Metallomics 3
(2011) 656–661.
[43] V. Ball, C. Maechling, Isothermal microcalorimetry to investigate non specificinter-
actions in biophysical chemistry, Int. J. Mol. Sci. 10 (2009) 3283–3315.
[44] J. Goodwin, S. Nath, Y. Engelborghs, D.L. Pountney, Raised calcium and oxidative
stress cooperatively promote alpha-synuclein aggregate formation, Neurochem.
Int. 62 (2013) 703–711.
[45] M.K. Jain, R. Bhat, Modulation of human alpha-synuclein aggregation by a combined
effect of calcium and dopamine, Neurobiol. Dis. 63 (2014) 115–128.
[46] F. Ding, N.V. Dokholyan, Dynamical roles of metal ions and the disulfide bond in Cu,
Zn superoxide dismutase folding and aggregation, Proc. Natl. Acad. Sci. U. S. A. 105
(2008) 19696–19701.
[47] K.S. Molnar, N.M. Karabacak, J.L. Johnson, Q. Wang, A. Tiwari, L.J. Hayward, S.J.
Coales, Y. Hamuro, J.N. Agar, A common property of amyotrophic lateral sclerosis-
associated variants: destabilization of the copper/zinc superoxide dismutase elec-
trostatic loop, J. Biol. Chem. 284 (2009) 30965–30973.
[48] L.J. Hayward, J.A. Rodriguez, J.W. Kim, A. Tiwari, J.J. Goto, D.E. Cabelli, J.S. Valentine,
R.H. Brown, Decreased metallation and activity in subsets of mutant superoxide
dismutases associated with familial amyotrophic lateral sclerosis, J. Biol. Chem.
277 (2002) 15923–15931.
[49] P.J. Hart, H. Liu, M. Pellegrini, A.M. Nersissian, E.B. Gralla, J.S. Valentine, D. Eisenberg,
Subunit asymmetry in the three-dimensional structure of a human CuZnSOD mu-
tant found in familial amyotrophic lateral sclerosis, Protein Sci. 7 (1998) 545–555.
[50] R.W. Strange, C.W. Yong, W. Smith, S.S. Hasnain, Molecular dynamics using atomic-
resolution structure reveal structural fluctuations that may lead to polymerization of
human Cu–Zn superoxide dismutase, Proc. Natl. Acad. Sci. U. S. A. 104 (2007)
10040–10044.
[51] D.D. Winkler, J.P. Schuermann, X. Cao, S.P. Holloway, D.R. Borchelt, M.C. Carroll, J.B.
Proescher, V.C. Culotta, P.J. Hart, Structural and biophysical properties of the patho-
genic SOD1 variant H46R/H48Q†, Biochemistry 48 (2009) 3436–3447.
[52] F. Chiti, Relative importance of hydrophobicity, net charge, and secondary structure
propensities in proteinaggregation, Protein Misfolding,Aggregation, and Conforma-
tional Diseases —Protein Reviews, Springer US, Place Published, 2006. 43–59.
[53] J.S. Elam, A.B. Taylor, R. Strange, S. Antonyuk, P.A. Doucette, J.A. Rodriguez, S.S.
Hasnain, L.J. Hayward, J.S. Valentine, T.O. Yeates, P.J. Hart, Amyloid-like filaments
and water-filled nanotubes formed by SOD1 mutant proteins linked to familial
ALS, Nat. Struct. Mol. Biol. 10 (2003) 461–467.
[54] S.D. Khare, N.V. Dokholyan, Common dynamical signatures of familial amyotrophic
lateral sclerosis-associated structurally diverse Cu, Zn superoxide dismutase mu-
tants, Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 3147–3152.
[55] R.W. Strange, S. Antonyuk, M.A. Hough, P.A. Doucette, J.A. Rodriguez, P.J. Hart, L.J.
Hayward, J.S. Valentine, S.S. Hasnain, The structure of holo and metal-deficient
wild-type human cu, zn superoxide dismutase and its relevance to familial amyo-
trophic lateral sclerosis, J. Mol. Biol. 328 (2003) 877–891.
[56] S.D. Khare, F. Ding, N.V. Dokholyan, Folding of Cu, Zn superoxide dismutase and
familial amyotrophic lateral sclerosis, J. Mol. Biol. 334 (2003) 515–525.
[57] Y. Furukawa, K. Kaneko, K. Yamanaka, N. Nukina, Mutation-dependent polymor-
phism of Cu, Zn-Superoxide dismutase aggregates in the familial form of amyotro-
phic lateral sclerosis, J. Biol. Chem. 285 (2010) 22221–22231.
[58] M.I. Ivanova, S.A. Sievers, E.L. Guenther, L.M. Johnson, D.D. Winkler, A. Galaleldeen,
M.R. Sawaya, P.J. Hart, D.S. Eisenberg, Aggregation-triggering segments of SOD1 fi-
bril formation support a common pathway for familial and sporadic ALS, Proc.
Natl. Acad. Sci. U.S. A. 111 (2014) 197–201.
[59] S. Maurer-Stroh, M. Debulpaep, N. Kuemmerer, M. de la Paz, I. Martins, J. Reumers,
K. Morris, A. Copland, L. Serpell, L. Serrano, J. Schymkowitz, F. Rousseau, Exploring
the sequence determinants of amyloid structure using position-specific scoring ma-
trices, Nat. Methods 7 (2010) (237-U109).
[60] J. Beerten, J. Schymkowitz, F. Rousseau, Aggregation prone regions and gatekeeping
residues in protein sequences, Curr. Top. Med. Chem. 12 (2012) 2470–2478.
[61] M. Kurnik, L. Hedberg, J. Danielsson, M. Oliveberg, Folding without charges, Proc.
Natl. Acad. Sci. U. S. A. 109 (2012) 5705–5710.
[62] J. Orpiszewski, M.D. Benson, Induction of β-sheet structure in amyloidogenic pep-
tides by neutralization of aspartate: a model for amyloid nucleation, J. Mol. Biol.
289 (1999) 413–428.
[63] J. Richardson, D. Richardson, Principles and patterns of protein conformation, in: G.
Fasman (Ed.), Prediction of Protein Structure and the Principles of Protein Confor-
mation, Springer US, Place Published, 1989, pp. 1–98.
[64] A. Nordlund, M. Oliveberg, Folding of Cu/Zn superoxide dismutase suggests struc-
tural hotspots for gain of neurotoxic function in ALS: parallels to precursors in am-
yloid disease, Proc. Natl. Acad. Sci. U. S. A. 103 (2006) 10218–10223.
[65] J.S. Richardson, D.C. Richardson, Natural β-sheet proteins use negative design to
avoid edge-to-edge aggregation, Proc. Natl. Acad. Sci. U. S. A. 99 (2002) 2754–2759.
[66] E. Sandelin, A. Nordlund, P. Andersen, S. Marklund, M. Oliveberg, Amyotrophic later-
al sclerosis-associated copper/zinc superoxide dismutase mutations preferentially
reduce the repulsive charge of the proteins, J. Biol. Chem. 282 (2007) 21230–21236.
[67] A. Nordlund, M. Oliveberg, SOD1‐associated ALS: a promising system forelucidating
the origin of protein‐misfolding disease, HFSP J. 2 (2008) 354–364.
[68] T.R. Jahn, M.J. Parker, S.W. Homans, S.E. Radford, Amyloid formation under physio-
logical conditions proceeds via a native-like folding intermediate, Nat. Struct. Mol.
Biol. 13 (2006) 195–201.
[69] P. Neudecker, P. Robustelli, A. Cavalli, P. Walsh, P. Lundström, A. Zarrine-Afsar, S.
Sharpe, M. Vendruscolo, L.E. Kay, Structure of an intermediate state in protein fold-
ing and aggregation, Science 336 (2012) 362–366.
[70] H. Krobath, S.G. Estácio, P.F.N. Faísca, E.I.Shakhnovich, Identification of a conserved
aggregation-prone intermediate state in the folding pathways of Spc-SH3
amyloidogenic variants, J. Mol. Biol. 422 (2012) 705–722.
[71] S.G. Estácio, C.S. Fernandes, H. Krobath, P.F.N. Faísca, E.I. Shakhnovich, Robustness of
atomistic Gōmodels in predicting native-like folding intermediates, J. Chem. Phys.
137 (2012) 085102.
[72] S.G. Estácio, E.I. Shakhnovich, P.F.N. Faísca, Assessing the effect of loop mutations in
the folding space of β2-microglobulin with molecular dynamics simulations, Int. J.
Mol. Sci. 14 (2013) 17256–17278.
[73] S.G. Estácio, H. Krobath, D. Vila-Viçosa, M. Machuqueiro, E.I. Shakhnovich, P.F.N. Faísca,
A simulated intermediate state for folding and aggregation provides insights into ΔN6
β2-microglobulin amyloidogenic behavior, PLoS Comput. Biol. 10 (2014) e1003606.
[74] F. Chiti, C.M. Dobson, Amyloid formation by globular proteins under native condi-
tions, Nat. Chem. Biol. 5 (2009) 15–22.
[75] O. Abel, J.F. Powell, P.M. Andersen, A. Al-Chalabi, ALSoD: a user-friendly online bioinfor-
matics tool for amyotrophic lateral sclerosis genetics, Hum. Mutat. 33 (2012)
1345–1351.
126 S.G. Estácio et al. / Biochimica et Biophysica Acta 1854 (2015) 118–126