Our analysis begins with Synuclein sequences. Fig. 2 shows our alignment
of the three human Synucleins. We divided these sequences into three
domains, and the domains into segments; residues 1-61 comprise the
N-terminus (Nt) domain, residues 62-97 the NAC (Non-Amyloid Component)
domain [1], and residues 98-140 of α-Syn comprise the C-terminus
(Ct) domain. We divided the Nt and NAC domains into eight segments,
Sy1-Sy8. Each segment is 11 residues long except for Sy1 and Sy5, which
are four residues longer. Sy7 is missing in β-Syn, but otherwise no
indels occur in the Nt and NAC domains among the three families of
Synucleins. Our alignment is based primarily on aligning highly
conserved signature sequences (consensus KTKEGV). These are present in
Sy1-5 and Sy7. Sy6 and Sy8 have some characteristics of the Nt segments
but are more hydrophobic and lack the signature sequence. Ct of α-Syn
appears to contain four sequentially similar segments that are nine to
thirteen residues long. We call these segments Ct1, Ct2, Ct3, and Ct4 in
most models. All contain an abundance of negatively charged residues
(residues 104 – 140 have 15 negatively charged carboxyl groups and no
positively charged groups) and a proline at or near their end. Sequences
of Ct1 and Ct2 are poorly conserved between α-Syn and β-Syn with two
short 3-residue insertions in β-Syn; most of the Ct1-Ct2 sequence is
deleted in our alignment of γ-Syn. The Ct3 sequences are similar in
α-Syn and β-Syn, and Ct4 sequences are almost identical. Ct3 and Ct4
sequences are present but not highly conserved in our alignment of
γ-Syn.
Figure
2. Sequence alignments of the three human Synuclein families.
Structures proposed for the segments of Fig.2 differ from one another
among the numerous models presented below. The Nt domain was modeled in
the following ways: (1) in αβ-barrel models Sy1-Sy3 and Sy4-Sy5 each
form an α-helix, (2) in Type A (antiparallel) models Sy1 – Sy5 form a
five stranded antiparallel β-sheet with the signature sequences forming
turn regions, and (3) in Type P (parallel) models Sy2 and Sy4 β-strands
comprise part of a β-barrel and Sy1, Sy3, and Sy5 form a parallel
β-sheet that comprises part of a second surrounding β-barrel. The NAC
domain was usually modeled as two β-strands that form a β-hairpin; for
α-Syn the turn occurs in the polar Q79-K80-T81 region; and for β-Syn,
which has no Sy7 segment, the turn occurs between Sy6 and Sy8 in the
SGAGN segment which has a high propensity for turns or coils [36].
Schematic illustrations of each segment were drawn to scale: the
diameter of the α-helices was 0.12 nm, the rise per residue was 0.15 nm,
and the axial rotation per residue was 100⁰. For β-strands the
translation per residue was 0.35 nm, the perpendicular distance between
adjacent strands was 0.42 nm, and the tilt angle of the β-strands was
predicted by the β-barrel theory of Murzin et al. [37]. Models were
constrained in several ways: (1) All subunits were required to have
identical conformations and interactions with other subunits. The
resulting models have two types of symmerty. All subunits are oriented
in the same direction relative to the radial axis of a core β-barrel in
Type 1 models; these have M-fold radial symmetry where M is the number
of monomers in the assembly. Every other subunit is oriented in the
opposite direction in Type 2 models; these have M/2 radial symmetry and
2-fold perpendicular symmetry. The 2-fold axes of symmetry are located
between each pair of subunits, are perpendicular to the radial axis of
symmetry, and intersect the radial axis at the center of mass of the
assembly; adjacent subunits are related by a 180⁰ rotation about these
axes. (2)The β-barrel theory of Murzin et al. [37] was used to
calculate the number of strands, N, tilt angle, α, diameter, D, and
sheer number, S, of putative β barrels. Values of these parameter are
limited by symmetry constraints and depend upon the number of
strands/monomer in a α barrel. S/N values of the models presented here
are consistent with those that have been commonaly observed; they range
from 0.4 to 1.5 with the most common being 1.0 (see Table S1 of the
appendix and detailed description of theory in Durell et al. [30]).
The structures are even more constrained if the assembly has concentric
β-barrels because all of the barrels must have the same axes of
symmetry. Also, based on known structures with stacked β-sheets or
concentric β-barrels, the distance of ~0.6 nm between
some closely packed β-sheets in some α-Syn fibrils (personal
observation) and in some antifreeze proteins [38], and our
experience in developing atomic scale models, the gap distance between
the walls of the adjacent barrels should be between 0.6 and 1.2 nm.
Residues with small side-chains dominate putative β-strands of Nt and
NAC domins, allowing the gap distances to be less than 1.0 nm in most
models. For well-known physiochemical reasons, we favor models that
maximize hydrogen bonds, salt bridges, interactions among aromatic
side-chains, burial and tight packing of hydrophobic side-chains, and
aqueous solvent exposure of hydrophilic side-chains. Residues with polar
side-chains and/or that have a high propensity for turn or coil
secondary structure [36] were favored for connecting loops. We favor
models in which residues that are identical among distantly related
homologous proteins (e.g. , α-Syn and β-Syn families) cluster in
the interior or at functional sites whereas residues that are
hypervariable and/or where indels occur among families are located
primarily on protein surfaces. We favor models in which interacting
pleats of adjacent β-barrels fit between each other in a manner that
reduces clashes among side-chains and in which all pleats that intersect
the axes of 2-fold symmetry have the same orientation in all concentric
β-barrels. In some cases β-strands were positioned and oriented to
minimize stearic clashes between side-chains in adjacent β-barrels and
to reduce empty cavities between hydrophobic regions of adjacent
β-barrels. The final constraints are experimental: the sizes, shapes,
molecular weights, and secondary structures of assemblies as determined
by EM and other studies.
Atomic scale models were produced with an in-house program consistent
with predicted values of S/N, α, and D and schematic models. Structures
were energy minimized with the CHARMM program [39]. Molecular
graphics images were produced using the UCSF Chimera package from the
Resource for Biocomputing, Visualization, and Informatics at the
University of California, San Francisco (supported by NIH P41
RR-01081)[40]. Image averaging using Adobe Photoshop was used on
some lipoprotein assemblies.
Our β-barrel models are overly structured; i.e. , portions of the
assemblies are likely to be less ordered in vivo than the models
imply. Secondary structures of some regions may be fractionally
occupied; too dynamic for their atomic structures to be resolved, but
sufficiently ordered to contribute to the overall structure and
electrostatic interactions of the assembly. Also, when constraining
models to correspond to sizes and shapes observed in EM studies it seems
prudent to include all of the protein sequence, even if the outer
portions of the models have greater uncertainty, disorder, and only
approximate the general location of the region.