(Received for publication, November 1, 1996, and in revised form, February 10, 1997)
From the Laboratory of Biochemistry and Molecular
Pharmacology, Section of Drug Dependence Research, Department of
Clinical Neuroscience, Karolinska Hospital, S-171 76 Stockholm, Sweden,
the § Picower Institute for Medical Research, Manhasset,
New York 11030, and the
Department of Cell and Molecular
Biology, Medical Nobel Institute, Karolinska Institute,
S-171 77 Stockholm, Sweden
We have previously shown that short peptides
incorporating the sequence KLVFF can bind to the ~40amino acid
residue Alzheimer amyloid -peptide (A
) and disrupt amyloid fibril
formation (Tjernberg, L. O., Näslund, J., Lindqvist, F.,
Johansson, J., Karlström, A. R., Thyberg, J., Terenius, L., and
Nordstedt, C. (1996) J. Biol. Chem. 271, 8545-8548).
Here, it is shown that KLVFF binds stereospecifically to the homologous
sequence in A
(i.e. A
16-20). Molecular
modeling suggests that association of the two homologous sequences
leads to the formation of an atypical anti-parallel
-sheet structure
stabilized primarily by interaction between the Lys, Leu, and
COOH-terminal Phe. By screening combinatorial pentapeptide libraries
exclusively composed of D-amino acids, several ligands with
a general motif containing phenylalanine in the second position and
leucine in the third position were identified. Ligands composed of
D-amino acids were not only capable of binding A
but
also prevented formation of amyloid-like fibrils. These ligands are
protease-resistant and may thus be useful as experimental agents
against amyloid fibril formation in vivo.
Accumulating evidence ascribes the deposition of large fibrillar
polymers of the Alzheimer amyloid -peptide
(A
)1 in the brain parenchyma and
vasculature (1-3) as the key step in the pathogenesis of Alzheimer's
disease. In support of this contention, specific mutations in the gene
encoding the
-amyloid precursor protein (4) have been shown to
co-segregate with familial forms of Alzheimer's disease (5, 6). These
mutations have phenotypes with either increased production of the
A
-peptide or formation of structural variants of the A
-peptide
(e.g. terminating at amino acid A
42) that
more easily aggregate and form amyloid fibrils (7-9). Mutations in the
genes encoding the presenilins that also co-segregate with early onset
familial Alzheimer's disease (10-13) lead to an increased production
of A
X-42 (14). Moreover, transgenic mice
expressing mutant human A
precursor protein develop lesions similar
to those found in Alzheimer's disease (15, 16).
A number of pharmacological strategies aimed at decreasing the
production of the A-peptide or precluding its tissue deposition have
been suggested. These include the use of protease inhibitors to prevent
amyloidogenic processing of A
precursor protein (17), altering A
precursor protein metabolism by pharmacological manipulation of signal
transduction pathways (18), and inhibition of amyloid fibril formation
with small molecules (19-23). Inhibition of fibril formation is
essential since fibrils of A
, but not monomeric A
, are toxic to
neurons (21, 24, 25) and protease-resistant (26), and therefore can
accumulate in and probably also damage the tissue.
Previously, it was shown that residues 16-20 in the A-peptide
(KLVFF) are important in the A
-A
interaction (19, 27). The linear
hexapeptide Ac-QKLVFF-NH2 was found to be capable of both
binding to the A
-peptide and preventing its polymerization into
fibrils (19). In the present work we have focused on (i) identifying
the structure or structures in the A
-peptide that bind
KLVFF-containing ligands, (ii) using molecular modeling to study the
interaction between these ligands and model their binding site, (iii)
developing novel peptide ligands based on amino acids in D
configuration, and (iv) investigating the effects of these novel
metabolically stable ligands on amyloid fibril formation in
vitro.
Synthetic A1-40 was obtained from
Dr. David B. Teplow, Biopolymer Laboratory at Harvard University. All
other soluble peptides were obtained from Research Genetics,
Huntsville, AL. Unless otherwise indicated, all reagents were from
Sigma. The peptide KKLVFFA (this peptide will be referred to hereafter
as LBMP1620) and A
1-40 were iodinated using the
Bolton-Hunter technique. Following the reaction, the iodinated peptide
was purified on a Vydac C-4 reverse phase liquid chromatography column
(0.21 × 15 cm) using a solvent system containing 0.1%
trifluoroacetic acid in water (buffer A) and 0.1% trifluoroacetic acid
in acetonitrile (buffer B).
The technique
used is essentially identical to the SPOT technique described by Frank
(28). Cellulose membranes (Whatman 1Chr) were derivatized with
N,N-diisopropylcarbodiimide-activated Fmoc
-alanine. A spacer, consisting of 2 molecules of
-alanine, was
coupled to derivatized filters. The indicated peptides were then
synthesized using Fmoc-protected and pentafluorophenyl- or N,N
-diisopropylcarbodiimide-activated amino
acids dissolved in N-methylpyrrolidone. Coupling efficiency
was monitored by bromphenol blue staining.
Following the blocking of filters with
0.05% Tween 20 in Tris-buffered saline (TBS), they were incubated in
the presence of 2-20 µM 125I-labeled
A1-40 or 125I-labeled LBMP1620 at 20 °C
for 12 h in TBS, pH 7.4, supplemented with 1% bovine serum
albumin. Subsequently, the filters were washed repeatedly in the same
buffer containing 0.5 M NaCl and dried. Radioactivity
bound to the filters was visualized by autoradiography and
quantitated using densitometry.
The docking of a KLVFF (A16-20) pentamer
to an A
13-23 segment was studied using the
Insight/Discover 2.9.7 program suite (Biosym/Molecular Simulations, San
Diego, CA). The two polypeptides were simulated in a periodic box
(47 × 20 × 24 Å) with 607 explicit water molecules and an
8-Å cut-off. Default values were used for all other conventions. The
A
13-23 segment was initialized in a
sheet
conformation.
The ligands were
dissolved in hexafluoroisopropyl alcohol and diluted with TBS to a
final concentration of 100 µM. A was dissolved in
hexafluoroisopropyl alcohol before the addition of the freshly prepared
ligand solution to a final A
concentration of 100 µM.
The final concentration of hexafluoroisopropyl alcohol was 2%. In
another experiment the samples were dissolved directly in TBS. The
samples were thereafter incubated at 37 °C for 48 h.
Subsequently, the peptides were examined by electron microscopy, and
inhibition of fibril formation was estimated semiquantitatively.
The peptides were dissolved in TBS (pH 7.4) and incubated at 37 °C for 48 h. Polymers were sedimented by centrifugation at 20,000 × g for 20 min, the supernatant was aspirated, and the pellet was resuspended in 100 µl of water by a short sonication. Aliquots (8 µl) of the resuspended material were placed on grids covered by a carbon-stabilized Formvar film. After 30 s, excess fluid was withdrawn, and the grids were negatively stained with 3% uranyl acetate in water. Finally, the specimens were examined and photographed in a Jeol electron microscope 100CX at 60 kV.
The previously used peptide sequence Ac-QKLVFF-NH2 has a limited solubility in aqueous buffers. We therefore synthesized a peptide sequence with improved water solubility, LBMP1620.
The 31 possible decamers corresponding to A1-10,
A
2-11, A
3-12, ... , A
31-40 were synthesized on a cellulose membrane with
their COOH terminus covalently bound to the membrane as described (28).
The membrane was incubated overnight at 20 °C in the presence of 2 µM 125I-LBMP1620. Following repeated washing
in TBS, binding of 125I-LBMP1620 to the immobilized
peptides was determined by autoradiography and subsequent densitometry.
The radioligand bound to peptides A
11-20,
A
12-21, A
13-22, A
14-23,
A
15-24, and A
16-25. This binding
pattern is similar to the one found when immobilized peptides were
incubated with 125I-A
1-40 (19). The decamer
A
13-22 (i.e. HHQKLVFFAE) was selected for
further studies. The minimal sequence required for
125I-LBMP1620 binding was determined by systematic
truncation at the NH2- and COOH terminus, respectively
(Fig. 1). The results revealed that the LBMP1620-binding
site had certain distinct features. (i) Only pentapeptides or larger
peptides displayed significant binding, and (ii) binding peptides
contained the sequence KLVFF or LVFFA (i.e.
A
16-20 or A
17-21). It is concluded that
the A
-ligand LBMP1620 binds to the A
-peptide by interacting with
a domain with a homologous primary sequence.
A peptide entirely composed of amino acids in D configuration with the sequence klvff (lowercase marks amino acids in D configuration) was synthesized using the SPOT technique and assayed for 125I-LBMP1620 binding. This peptide failed to bind 125I-LBMP1620 (data not shown) indicating that the KLVFF-KLVFF interaction is stereospecific.
Immobilized AKLVFF was synthesized on cellulose membranes. They were
then incubated overnight at 20 °C with incubation buffer alone, 20 µM A1-40, or LBMP1620. Following washing,
the immobilized peptide was incubated with 20 µM
125I-A
1-40 for 5 h. After the removal
of non-bound radiolabeled peptide by repeated washing, binding of
125I-A
1-40 was determined by
autoradiography and densitometry. Preincubation of KLVFF with
A
1-40 did not affect
125I-A
1-40 binding, whereas preincubation
with LBMP1620 reduced 125I-A
1-40 binding by
49%. This confirms that short KLVFF-containing peptides bind to
A
16-20 and thereby block further A
1-40
binding. Therefore, it is reasonable to assume that a short peptide or
a functionally similar molecule capable of binding to
A
16-20 may inhibit A
polymer growth.
Incubation of
KLVFF in aqueous buffers leads to aggregation and precipitation (19),
which complicates studies of the structural basis for binding using
solution phase techniques (e.g. NMR). The small sizes of the
present peptides made them particularly suitable for computer
simulations using molecular modeling. Repeated runs of molecular
dynamics were followed by conjugate gradient energy minimization.
Peptides with the sequences HHQKLVFFAED (A13-23) and
KLVFF (A
16-20) were docked (Fig. 2). Two
common docking motifs were observed roughly corresponding to parallel
and anti-parallel
-sheet conformations, respectively. The parallel
conformation was considerably higher in energy, and hence the
anti-parallel conformation was chosen for further studies. In dynamic
simulations of up to several picoseconds in duration, a typical docking
pattern involved the approach of the portion of the KLVFF pentamer
comprised of the alkane-like basic side chain of the Lys and the Leu of
the ligand to the Phe (A
20) of A
13-23.
The COOH-terminal Phe of the pentamer typically projected outward. This
motif would suggest a prominent role for Lys, Leu, and the COOH-terminal Phe residues in A
aggregation as observed in earlier experiments (19). Over the course of the run, the A
segment developed an elbow, with the docked pentamer in the interior. Presumably, during longer simulations this conformational shift could
significantly enhance the binding of KLVFF pentamers to A
.
Identification of A
LBMP1620 contains several proteolytic cleavage sites
and is highly sensitive to proteolysis (26). We therefore decided to design a ligand with the same binding properties as LBMP1620 but that
would be resistant to proteolysis. Since proteolytic enzymes are not
capable of hydrolyzing peptide bonds between amino acids in
D configuration, ligands entirely composed of
D-amino acids should be able to withstand enzymatic
proteolysis. Here it was shown that the shortest peptide binding
efficiently to LBMP1620 was a pentamer (see Fig. 1). Therefore,
pentapeptides exclusively composed of D-amino acids were
synthesized. A combinatorial pentapeptide library, theoretically
containing all 2,476,099 permutations of the 19 D-amino
acids (cysteine was excluded to avoid unwanted disulfide bonds), was
synthesized as described (29) and subsequently was assayed for
125I-LBMP1620 binding. By this design, an individual
peptide spot on the membrane contained 130,321 different peptide
sequences. As seen in Fig. 3, all amino acid residues
mediating 125I-LBMP1620 binding had either basic side
chains (h, k, r), aromatic side chains (f, y, w) or non-polar side
chains (i, l, m). No amino acid residues with acidic side chains
promoted binding, strongly suggesting that electrostatic interactions
(i.e. with Lys in LBMP1620) were not critical for binding.
Assaying an identical library for 125I-A1-40 binding yielded similar results
(data not shown).
The three amino acids in each position in the pentapeptide library that most prominently promoted 125I-LBMP1620 binding were selected for additional studies (Table I). The 243 possible pentapeptide combinations of these D-amino acids were synthesized individually and assayed for 125I-LBMP1620 binding. The 39 peptides binding most efficiently to 125I-LBMP1620 are listed in Fig. 4. As seen in the figure, f in the second and l in the third position were critical for binding. The f in position 2 could be replaced with y, and l in position 3 could be replaced with m, but not concurrently. In position 1, l and y showed higher binding than w, whereas l was favored in position 4. The fifth residue could be either one of the three amino acids tested. It is unlikely that binding observed in these experiments was the result of nonspecific protein adhesion since the reaction mixture contained relatively large quantities of bovine serum albumin.
|
Functional Properties of Pentapeptide Ligands Identified in Combinatorial Libraries of All D-Amino Acids
From the
39 125I-LBMP1620 binding peptides showed in Fig. 4, yfllr
and lflrr were selected for further studies. The reason for selecting
these peptides was that they contain one or two D-arginine residues, thus making them more hydrophilic and water-soluble than
other binding peptides. We also selected a non-binding peptide with the
sequence yyrrl as control. The peptides were incubated with synthetic
A1-40 at equimolar concentrations (100 µM) as described (19). When A
1-40 was
incubated alone, a dense network of amyloid fibrils with a diameter of
about 6-8 nm and undefined lengths was formed (Fig. 5A). The addition of the non-binding control
peptide yyrrl had no effect on the polymerization of
A
1-40 (Fig. 5B). On the other hand, lflrr
markedly reduced the amount of polymers produced and instead of
elongated fibrils short rodlike structures were formed (Fig.
5C). The putative toxicity of these structures has not been
studied. However, it has been shown that it is the fibrillar form, not
the amorphous form, of aggregated A
that mediates neurotoxicity
(21). A less distinct effect was obtained with yfllr under these
conditions, and a mixture of elongated fibrils and diffuse aggregates
appeared (Fig. 5D). When the D-pentapeptides
were incubated alone no detectable polymers or aggregates were formed
(not shown), in partial contrast to the previously studied
KLVFF-containing peptide ligand made up of L-amino acids
(19).
In another set of experiments, a lower concentration of
A1-40 (5 µM) was incubated with a 10-fold
excess of the D-pentapeptides (50 µM). In
this case, A
1-40 alone gave rise to fibrils with a
similar width but shorter length than described above. Both lflrr and
yflrr almost completely blocked the formation of fibrils, whereas yyrrl
again had no effect (not shown). No differences between samples
predissolved in hexafluoroisopropyl alcohol and samples dissolved
directly in TBS were detected. The present method is designed to assay
the inhibition of amyloid fibril formation, and hence we cannot exclude
the possibility that small, potentially toxic, aggregates are being
formed in the presence of ligands. However, preliminary results from
fluorescence correlation spectroscopy studies indicate that LBMP1620
dissolves preformed soluble A
aggregates
completely.2
Our previous data indicated that the KLVFF
(A16-20) sequence was critical for A
fibril
formation. However, it was not clear whether fibril formation involved
interaction between homologous sequences or not. Here, we have used
LBMP1620 as the labeled probe for screening defined sequences in
A
1-40, ranging from tripeptides to decapeptides, and
thereby identified the sequence that KLVFF-containing ligands bind to
within the A
1-40 molecule. The identified binding site
docking to KLVFF was then subjected to molecular modeling.
Our next aim was to use the acquired data to search for a
protease-stable inhibitor of A fibril formation. Random
combinatorial pentapeptide libraries made up from D-amino
acids, and consequently protease-stable, were probed with labeled
LBMP1620. The three D-amino acids most efficiently
promoting binding in each position were selected, and out of these all
243 possible pentapeptides were synthesized and assayed for LBMP1620
binding. Two of the binding peptides and one non-binding peptide were
synthesized on a larger scale, and their capacity of inhibiting A
fibril formation in solution was tested.
The first experiments conclusively showed that the KLVFF-containing
ligand binds to the homologous sequence in A (Fig. 1). The data
obtained in binding studies were used in computer simulations of
docking between KLVFF and A
13-23. An anti-parallel
arrangement was shown in the minimum energy conformation. Residues Lys
and Leu in KLVFF interact with the Phe (A
20) in the
homologous sequence in A
13-23, thereby stabilizing the
formation of a
-sheet structure (Fig. 2). This is in agreement with
previous experimental data showing that the substitution of Lys, Leu,
and the COOH-terminal Phe in KLVFF with Ala leads to loss of binding
capacity (19).
Since the first experiment revealed that LBMP1620 binding requires a
sequence of at least five amino acids, we chose to use a combinatorial
library consisting of pentamers of D-amino acids. By using
a short peptide capable of binding to a region critical for
polymerization of A (i.e. LBMP1620), the risk of
identifying D-pentapeptides binding to non-relevant regions
of A
1-40 (i.e. NH2- and
COOH-terminal to A
16-20) was minimized. Of the 243 pentapeptides synthesized from the most efficiently binding amino acids
(Table I), 39 were found to bind 125I-LBMP1620. The
composition of these binding peptides indicates that hydrophobic
interactions are necessary for binding (Fig. 4). Although r can be
found in position 4 or 5, l is more frequent in these positions.
Positions 2 and 3 were most critical for binding, being l and f,
respectively, in most of the binding peptides. In position 1, l and f
were preferred over w. Some of the binding peptides showed similarities
with a reversed KLVFF, e.g. yfllr (Fig. 4), indicating the
formation of a parallel
-sheet (KLVFF forming an anti-parallel
-sheet according to molecular modeling). However, the positions
critical for binding to LBMP1620, 2 (f) and 3 (l), would then
correspond to non-critical residues in KLVFF. Molecular modeling of the
D-amino acid ligands docking to KLVFF might reveal the
nature of this interaction.
An interesting question is how A-ligands, such as those described
here, affect the physiological and pathological properties of
aggregated and non-aggregated A
-peptide. It is well known that
aggregated full-length and truncated variants of the A
-peptide are
toxic to several neuronal and non-neuronal cell types (30). Moreover,
substances capable of inhibiting the polymerization have been shown to
prevent A
-associated toxicity (21-23), which points to the fact
that the higher order structure is important for the toxicity of these
peptides. An A
-ligand has to meet several important criteria to be a
useful inhibitor of fibril formation in vivo. These include
acceptable bioavailability, pharmacokinetics allowing passage of the
blood-brain barrier, access to the desired site of action, and low
toxicity. The present D-ligands are resistant to
proteolysis and should therefore have acceptable bioavailability. Those
that were studied functionally were chosen because of their relatively
high hydrophilicity. Ligands to be used in vivo should probably be more hydrophobic because this increases the probability that they will pass the blood-brain barrier (31). Several of the 39 D-ligands presented in Fig. 4 have a high content of
hydrophobic residues and hence should have properties allowing them to
pass the blood-brain barrier. At present, there is no information
regarding the toxicity of these compounds. On balance, several of the
D-ligands are potentially useful for testing the hypothesis
that A
-amyloidogenesis in vivo may be inhibited by
ligands binding to the KLVFF motif.
A fragment of the A-peptide corresponding to amino acids 25-35
undergoes polymerization and has neurotoxic properties (30, 32, 33). In
these and our previous studies (19), we have not been able to identify
ligands capable of binding efficiently to this region of the peptide.
There are several possibilities for this failure. (i) A ligand
efficiently binding to this region has to be larger than the penta- to
decapeptides tested here and previously (19); (ii) the covalent linkage
of ligands to a matrix via their COOH termini may prevent the adoption
of the conformation necessary for interaction with (i.e.
binding to) amino acids 25-35; and (iii) the polymerization of the
hydrophobic 25-35 segment is not directly related to the
polymerization of full-length A
-peptide.
The overall conclusion from these studies is that not only the KLVFF
peptide (A16-20) but also structurally different
peptides consisting of non-natural amino acids have the capability of
binding A
and preventing its assembly into amyloid fibrils.
Therefore, it is reasonable to assume that pharmacologically useful
organic non-peptide molecules with similar functional properties as the
present ligands can be synthesized.