Controlling the morphology of cross β-sheet assemblies by rational design

Article abstract of DOI:10.1021/ja050558c

Low molecular weight peptidomimetics with simple amphiphilic sequences can help to elucidate the structures of cross β-sheet assemblies, such as amyloid fibrils. The peptidomimetics described herein comprise a dibenzofuran template, two peptide strands ma

Full text of DOI:10.1021/ja050558c

Published on Web 05/17/2005
Controlling the Morphology of Cross â-Sheet Assemblies by
Rational Design
Songpon Deechongkit, Evan T. Powers, Shu-Li You, and Jeffery W. Kelly*
Contribution from the Department of Chemistry and The Skaggs Institute for Chemical Biology,
The Scripps Research Institute, La Jolla, California 92037
Received January 27, 2005; E-mail: jkelly@scripps.edu
Abstract: Low molecular weight peptidomimetics with simple amphiphilic sequences can help to elucidate
the structures of cross â-sheet assemblies, such as amyloid fibrils. The peptidomimetics described herein
comprise a dibenzofuran template, two peptide strands made up of alternating hydrophilic and hydrophobic
residues, and carboxyl termini, each of which can be varied to probe the structural requirements for â-sheet
self-assembly processes. The dibenzofuran template positions the strands approximately 10 Å apart, allowing
corresponding hydrophobic side chains in the strands to pack into a collapsed U-shaped structure. This
conformation is stabilized by hydrophobic interactions, not intramolecular hydrogen bonds. Intermolecular
stacking of the collapsed peptidomimetics, enabled by intermolecular hydrogen bonding and hydrophobic
interactions, affords 25-27 Å wide protofilaments having a cross â-sheet structure. Association of
protofilaments, mediated by the dibenzofuran substructures and driven by the hydrophobic effect, affords
50-60 Å wide filaments. These widths can be controlled by changing the length of the peptide strands.
Further assembly of the filaments into fibrils or ribbons can be controlled by modification of the template,
C-terminus, and buffer ion composition.
responsive gels16,19,22,23 and scaffolds for tissue regeneration26,28
to substrates for crystal growth.21 Exploiting the potential of
Introduction
Many peptides and proteins form â-sheet-rich fibrillar ag-
gregates. This phenomenon has attracted the attention of
scientists from many fields because such assemblies both enable
biological function and appear to cause pathology.1 For example,
the formation of fibrillar cross â-sheet assemblies results in
functional materials, such as chorion, which is the main
component of silk moth eggshells,2 or curli fibrils, which are
used by Escherichia coli (E. coli) in the colonization of
surfaces.3 In contrast, the misassembly of peptides or proteins
can lead to cross â-sheet assemblies known as amyloid, the
process of amyloidogenesis being implicated as the cause of
several neurodegenerative diseases, including familial amyloid
polyneuropathy and Alzheimer’s disease.4-7 Many authors have
written about the potential use of ordered peptide assemblies
as materials,8-30 with suggested applications ranging from
â-sheet-based peptide assemblies requires an understanding of
their internal structures that is detailed enough to allow rational
(13) Moses, J. P.; Satheeshkumar, K. S.; Murali, J.; Alli, D.; Jayakumar, R.
Langmuir 2003, 19, 3413-3418.
(14) Sagis, L. M. C.; Veerman, C.; van der Linden, E. Langmuir 2004, 20, 924-
927.
(15) Matsumura, S.; Uemura, S.; Mihara, H. Chem.-Eur. J. 2004, 10, 2789-
2794.
(16) Aggeli, A.; Bell, M.; Boden, N.; Keen, J. N.; Knowles, P. F.; McLeish, T.
C.; Pitkeathly, M.; Radford, S. E. Nature 1997, 386, 259-262.
(17) Aggeli, A.; Nyrkova, I. A.; Bell, M.; Harding, R.; Carrick, L.; McLeish,
T. C.; Semenov, A. N.; Boden, N. Proc. Natl. Acad. Sci. U.S.A. 2001, 98,
11857-11862.
(18) Aggeli, A.; Bell, M.; Carrick, L. M.; Fishwick, C. W.; Harding, R.; Mawer,
P. J.; Radford, S. E.; Strong, A. E.; Boden, N. J. Am. Chem. Soc. 2003,
125, 9619-9628.
(19) Aggeli, A.; Bell, M.; Boden, N.; Carrick, L. M.; Strong, A. E. Angew.
Chem., Int. Ed. 2003, 42, 5603-5606.
(20) Kayser, V.; Turton, D. A.; Aggeli, A.; Beevers, A.; Reid, G. D.; Beddard,
G. S. J. Am. Chem. Soc. 2004, 126, 336-343.
(21) Meegan, J. E.; Aggeli, A.; Boden, N.; Brydson, R.; Brown, A. P.; Carrick,
L.; Brough, A. R.; Hussain, A.; Ansell, R. J. AdV. Funct. Mater. 2004, 14,
31-37.
(1) Huff, M. E.; Balch, W. E.; Kelly, J. W. Curr. Opin. Struct. Biol. 2003, 13,
674-682.
(2) Iconomidou, V. A.; Vriend, G.; Hamodrakas, S. J. FEBS Lett. 2000, 479,
141-145.
(22) Schneider, J. P.; Pochan, D. J.; Ozbas, B.; Rajagopal, K.; Pakstis, L.;
Kretsinger, J. J. Am. Chem. Soc. 2002, 124, 15030-15037.
(23) Pochan, D. J.; Schneider, J. P.; Kretsinger, J.; Ozbas, B.; Rajagopal, K.;
Haines, L. J. Am. Chem. Soc. 2003, 125, 11802-11803.
(24) Rajagopal, K.; Schneider, J. P. Curr. Opin. Struct. Biol. 2004, 14, 480-
486.
(3) Chapman, M. R.; Robinson, L. S.; Pinkner, J. S.; Roth, R.; Heuser, J.;
Hammar, M.; Normark, S.; Hultgren, S. J. Science 2002, 295, 851-855.
(4) Sipe, J. D. Crit. ReV. Clin. Lab. Sci. 1994, 31, 325-354.
(5) Sipe, J. D.; Cohen, A. S. J. Struct. Biol. 2000, 130, 88-98.
(6) Kelly, J. W. Curr. Opin. Struct. Biol. 1996, 6, 11-17.
(7) Selkoe, D. J. Nature 2003, 426, 900-904.
(25) Zhang, S.; Holmes, T.; Lockshin, C.; Rich, A. Proc. Natl. Acad. Sci. U.S.A.
1993, 90, 3334-3338.
(8) Krejchi, M. T.; Atkins, E. D.; Waddon, A. J.; Fournier, M. J.; Mason, T.
L.; Tirrell, D. A. Science 1994, 265, 1427-1432.
(26) Holmes, T. C.; de Lacalle, S.; Su, X.; Liu, G.; Rich, A.; Zhang, S. Proc.
Natl. Acad. Sci. U.S.A. 2000, 97, 6728-6733.
(9) Winningham, M. J.; Sogah, D. Y. Macromolecules 1997, 30, 862-876.
(10) Yamada, N.; Ariga, K.; Naito, M.; Matsubara, K.; Koyama, E. J. Am. Chem.
Soc. 1998, 120, 12192-12199.
(27) Zhang, S.; Marini, D. M.; Hwang, W.; Santoso, S. Curr. Opin. Chem. Biol.
2002, 6, 865-871.
(28) Semino, C. E.; Kasahara, J.; Hayashi, Y.; Zhang, S. G. Tissue Eng. 2004,
10, 643-655.
(11) MacPhee, C. E.; Dobson, C. M. J. Am. Chem. Soc. 2000, 122, 12707-
12713.
(29) Choo, D. W.; Schneider, J. P.; Graciani, N. R.; Kelly, J. W. Macromolecules
1996, 29, 355-366.
(12) Lopez De La Paz, M.; Goldie, K.; Zurdo, J.; Lacroix, E.; Dobson, C. M.;
Hoenger, A.; Serrano, L. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 16052-
16057.
(30) Lashuel, H. A.; Labrenz, S. R.; Woo, L.; Serpell, L. C.; Kelly, J. W. J.
Am. Chem. Soc. 2000, 122, 5262-5277.
9
8562
J. AM. CHEM. SOC. 2005, 127, 8562-8570
10.1021/ja050558c CCC: $30.25 © 2005 American Chemical Society

Morphology of Cross
â
-Sheet Assemblies
A R T I C L E S
Figure 1. (A) Structure of peptidomimetic 1 and the assemblies it forms (protofilaments, filaments, fibrils, and ribbons). The widths (w) and thicknesses
(th) of the assemblies are indicated in the figure. (B) Interdigitated strand model for the assembly of peptidomimetic 1. The dibenzofuran templates are
depicted as gray rectangles. The R-amino acid side chains are depicted as black (Val) or green (Thr) balls, while the C-terminal DMDA groups are depicted
as blue balls. The backbone amides are in the plane of the white rectangles representing the peptide strands; neighboring white rectangles therefore imply
intermolecular hydrogen bonding. (C) Side chain hydrophobic collapse model for the assembly of peptidomimetic 1. The components of the peptidomimetic
are represented as in Figure 1B. Buffer anions are depicted as red balls. See text for detailed descriptions of the models.
control over their supramolecular structures and properties. To
this end, we have previously synthesized peptidomimetic 1
(Figure 1A) and have carried out a preliminary assessment of
the structure of its assemblies.30
Peptidomimetic 1 consists of two peptide strands attached
by their N-termini to propanamide linkers at the 2 and 8
positions of a dibenzofuran substructure. The four residue
peptide strands are composed of alternating valine and threonine
residues (selected because of their high â-sheet propensities).
Peptidomimetic 1 was found to assemble into protofilaments,
filaments, fibrils, and ribbons, with the distribution of species
depending on the aqueous buffer used (Figure 1A).30 The
dimensions of the assemblies were measured using transmission
electron microscopy (TEM) and atomic force microscopy
(AFM). The protofilaments, which were a minor species under
all conditions examined, were 25-27 Å wide. The filaments
were 50-60 Å wide, and the fibrils and ribbons had widths
that were multiples of 50-60 Å. The thicknesses (determined
by AFM height measurements) of the protofilaments, fibrils,
and ribbons were multiples of 10-11 Å. X-ray fibril diffraction,
circular dichroism (CD) spectroscopy, and infrared (IR) spec-
troscopy revealed that assemblies of 1 had a cross â-sheet
structure, in which the peptide strands were oriented perpen-
dicular to the fibril axis. Consistent with this finding, the
assemblies bound dyes known to be selective for cross â-sheet
structures (thioflavin T and Congo red).31,32
9
J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005 8563

A R T I C L E S
Deechongkit et al.
Table 1. Summary of Peptidomimetics Prepared and Their
This information was used to infer a structural model for the
assemblies formed by 1, which will be referred to as the
“interdigitated strand” model (Figure 1B).30 The basic unit of
the assemblies according to this model is a dimer composed of
interdigitated monomers of the peptidomimetic, in which the
dibenzofuran substructure acts as a template for intermolecular
â-sheet formation. These intermolecularly hydrogen bonded
dimers assemble through intermolecular â-sheet formation, in
which the hydrogen bonds are collinear with the fibril axis and
the side chains are oriented perpendicular to the sheets. This
yields a structure that is 25-27 Å wide and 5-6 Å thick, which
will be referred to as a pre-protofilament. The alternating Val
and Thr side chains are on opposite sides of the structure,
creating hydrophobic and hydrophilic faces. Two pre-protofila-
ments can associate by burying their hydrophobic Val faces,
yielding a protofilament that is 25-27 Å wide and 10-11 Å
thick, as observed. The protofilaments can then associate
laterally along their edges, a process that would be governed
by interactions between dibenzofuran substructures (and there-
fore driven by the hydrophobic effect), to produce 50-60
Å-wide filaments. Further edge-to-edge assembly of the fila-
ments would produce wider structures, such as ribbons and
fibrils, with widths that are multiples of 50-60 Å, as observed.
In this model, backbone-backbone hydrogen bonding and
interstrand side-chain-side-chain hydrophobic interactions drive
assembly along the long axis of the assemblies (i.e. the formation
of pre-protofilaments), while hydrophobic interactions drive the
formation of higher order structures (protofilaments, filaments,
fibrils, and ribbons). The interdigitated strand model is consistent
with the existing data on assemblies of 1, but it has a significant
weakness: it cannot explain why the observed filaments, fibrils,
and ribbons had widths that were multiples of 50-60 Å rather
than 25-27 Å. A hypothesis involving “strand swapping”
between protofilaments was proposed to rationalize the 50-60
Å repeat.30
An alternative structural model, which will be referred to as
the “side chain hydrophobic collapse” model (Figure 1C),
recently became apparent with the publication of a structural
model for Aâ amyloid fibrils based on solid-state NMR data
(Aâ amyloidogenesis is thought to cause Alzheimer’s dis-
ease).33,34 In this model, the side chains of Aâ peptides are
turned inward on themselves, so that intramolecular interactions
are mediated by the side chains, not by hydrogen bonding
between peptide amides as would occur if the peptides adopted
a â-hairpin conformation. Assembly then occurs by intermo-
lecular backbone-backbone hydrogen bonding to form cross
â-sheets. The final assembled cross â-sheet structure is also
stabilized by the hydrophobic effect through intermolecular side
chain packing. While the analogy between peptidomimetic 1
and Aâ is imperfect (due to the semirigid spacer that 1 contains),
1 may still assemble like Aâ. The gap between the strands
attached at the 2 and 8 positions of the dibenzofuran template
in 1 is large enough to accommodate a conformation analogous
to that adopted by Aâ subunits in amyloid fibrils, in which the
Val side chains are turned inward, toward each other (Figure
Compositions
peptidominetic
template (R1)
peptide
end group (R2)
1
2
3
4
5
6
7
DBF (H)
DBF (H)
DBF (H)
DBF-EA (CH2CH2NH2)
DBF-PA (CH2CH2CO2H)
DBF (H)
(ValThr)2
(IleThr)2
N(CH3)2
N(CH3)2
N(CH3)2
N(CH3)2
N(CH3)2
N(CH3)2+
N(CH3)3
(ChaThr)2
(ValThr)2
(ValThr)2
(ValThr)4
(ValThr)2
DBF (H)
1C). This conformation would be favored by the hydrophobic
effect. Peptidomimetic 1 could then self-assemble by intermo-
lecular hydrogen bonding in the direction perpendicular to the
strand orientation to yield a protofilament that would be about
25-27 Å wide and 10-11 Å thick. These protofilaments would
have the same dimensions as those from the interdigitated strand
model; however, in contrast, they would have two different
edges: one composed of the hydrophobic dibenzofuran rings
and one composed of the hydrophilic C-termini of the peptide
strands. The protofilaments would be driven to bury their
hydrophobic edges by lateral association in order to minimize
their free energy, forming filaments that were 50-60 Å wide.
These filaments could then further associate through buffer
anion-mediated interactions and/or through hydrogen bonding
between the positively charged C-termini to form fibrils and
ribbons with widths that were multiples of 50-60 Å. In this
model, as in the interdigitated strand model, both backbone-
backbone hydrogen bonding and interstrand side-chain-side-
chain hydrophobic interactions would drive assembly along the
long axis of the cross â-sheet (i.e. the formation of protofila-
ments), and hydrophobic interactions drive the formation of
filaments from protofilaments. In contrast to the interdigitated
strand model, the formation of fibrils and ribbons from filaments
would be driven by buffer anion-mediated electrostatic and/or
hydrogen bonding interactions in the side chain hydrophobic
collapse model. According to the latter model, counterions and
protonation state of the C-termini would play an important role
in higher order assembly. Thus, assembly beyond the filament
stage would be highly buffer-dependent.
(31) Klunk, W. E.; Jacob, R. F.; Mason, R. P. Methods Enzymol. 1999, 309,
285-305.
(32) LeVine, H., 3rd. Methods Enzymol. 1999, 309, 274-284.
(33) Petkova, A. T.; Ishii, Y.; Balbach, J. J.; Antzutkin, O. N.; Leapman, R. D.;
Delaglio, F.; Tycko, R. Proc. Natl. Acad. Sci. U.S.A. 2002, 99, 16742-
16747.
Here, we first describe the design, synthesis, and assembly
characteristics of analogues of 1 that have enabled us to
determine that 1 likely assembles according to the side chain
(34) Tycko, R. Curr. Opin. Struct. Biol. 2004, 14, 96-103.
9
8564 J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005

Morphology of Cross
â
-Sheet Assemblies
A R T I C L E S
Scheme 1. Synthesis of Peptidomimetic 1
hydrophobic collapse model. Then we use this model to design
analogues of 1 (Table 1) that form assemblies with predictable
dimensions and morphologies.
in assemblies of 1 should be antiparallel according to the
interdigitated strand model, but parallel according to the side
chain hydrophobic collapse model. This difference in strand
orientation suggests that it could be possible to distinguish
between the two models using IR spectroscopy. The IR spectra
of assemblies of 1 formed at various pH values have a prominent
peak corresponding to the amide I band at 1629 cm-1 with a
shoulder at higher wavenumbers.30 These data are consistent
with an infinite antiparallel sheet (for which peaks at 1630 and
1690 cm-1 would be expected), but they are also consistent with
an infinite parallel sheet (for which peaks at 1637 and 1651
cm-1 would be expected).41 The far-UV CD spectrum of
assembled peptidomimetic 1 exhibits a broad minimum at 215-
220 nm, consistent with a â-sheet structure.30 However, there
is currently no basis for distinguishing parallel from antiparallel
cross â-sheet assemblies using CD spectroscopy, due to the
paucity of data. Furthermore, CD spectroscopy was unlikely to
enable ordered assemblies to be distinguished from amorphous
aggregates with high â-sheet content. In addition, because both
the hydrogen bonding and the hydrophobic effect drive assembly
in both models, even the temperature dependence of the
assembly stabilities was unlikely to enable the two models to
be distinguished. Because IR and CD spectroscopy and thermal
denaturation were all unlikely to provide the necessary informa-
tion, another means to distinguish between the two models was
sought.
The side chains of the residues in peptidomimetic 1 play
different roles in the interdigitated strand and side chain
hydrophobic collapse models. According to the interdigitated
strand model (Figure 1B), the size of the side chains should
only affect the thickness of the filaments. In contrast, according
to the side chain hydrophobic collapse model (Figure 1C), the
size of the side chains should determine whether assemblies
are observed at all. Peptidomimetics should not be able to
assemble if their side chains are too large to be accommodated
within the gap between the peptide strands attached at the 2
and 8 positions of the dibenzofuran template. Therefore, to
distinguish between these two models, peptidomimetics 2 and
Results and Discussion
Synthesis of Dibenzofuran-Containing Peptidomimetics.
The synthesis of peptidomimetic 1, shown in Scheme 1, begins
with coupling thiopropionic acid to Gly-PAM resin in the
presence of HBTU.35,36 This resin functionalization enables
attachment of Thr(Bzl) to the resin as a thioester, which is then
elaborated into the BocVal-Thr(Bzl)-Val-Thr(Bzl) peptide by
solid-phase peptide synthesis using a Boc/benzyl protection
strategy. The peptide was cleaved from the resin by nucleophilic
attack of N,N-dimethylethylenediamine (DMDA) on the C-
terminal thioester, affording the C-terminal DMDA amide. The
N-terminal Boc protecting group was removed with trifluoro-
acetic acid (TFA), yielding the peptide with a free N-terminal
amine after neutralization. This otherwise protected peptide was
coupled to the dibenzofuran-based template (2,8-dibenzofuran-
bis(3-propionic acid)) using DCC and DMAP, affording two
amide linkages. The protected peptidomimetic was finally
hydrogenated to remove the benzyl groups from the Thr side
chains. This synthesis is analogous to that reported previously,
except that the thiol-derivatized PAM resin was used instead
of Kaiser oxime resin.30,37-40 The disadvantages of using Kaiser
resin were that the coupling of the first residue onto the Kaiser
oxime resin was slow and incomplete and that the peptide-
resin linkage was not completely stable to repetitive TFA
treatments.
Distinguishing between the Interdigitated Strand and Side
Chain Hydrophobic Collapse Assembly Models. The â-sheets
(35) Hackeng, T. M.; Griffin, J. H.; Dawson, P. E. Proc. Natl. Acad. Sci. U.S.A.
1999, 96, 10068-10073.
(36) Hackeng, T. M.; Fernandez, J. A.; Dawson, P. E.; Kent, S. B.; Griffin, J.
H. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 14074-14078.
(37) Degrado, W. F.; Kaiser, E. T. J. Org. Chem. 1980, 45, 1295-1300.
(38) Degrado, W. F.; Kaiser, E. T. J. Org. Chem. 1982, 47, 3258-3261.
(39) Pichette, A.; Voyer, N.; Larouche, R.; Meillon, J. C. Tetrahedron Lett.
1997, 38, 1279-1282.
(40) Mellor, S. L.; Wellings, D. A.; Fehrentz, J.-A.; Paris, M.; Martinez, J.;
Ede, N. J.; Bray, A. M.; Evans, D. J.; Bloomberg, G. B. In Fmoc Solid-
Phase Peptide Synthesis: A Practical Approach; Chan, W. C., White, P.
D., Eds.; Oxford University Press: Oxford, U.K., 2000; pp 137-181.
(41) Barth, A.; Zscherp, C. Q. ReV. Biophys. 2002, 35, 369-430.
9
J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005 8565

A R T I C L E S
Deechongkit et al.
Figure 2. (A-C) Negatively stained (uranyl acetate) EM images of peptidomimetic 1 at pH 5 (acetate buffer), pH 7 (phosphate buffer), and pH 10 (carbonate
buffer). The expansion for A shows a 3-fold magnification of a 100 nm × 100 nm patch of the EM image. This patch contains a 400 Å wide ribbon
composed of eight filaments. (D-F) EM images of peptidomimetic 2 at pH 5 (acetate buffer), pH 7 (phosphate buffer), and pH 10 (carbonate buffer). The
expansion for E shows a 3-fold magnification of a 100 nm × 100 nm patch of the EM image. This patch contains a 55 Å wide filament. (G-I) EM images
of peptidomimetic 3 at pH 5 (acetate buffer), pH 7 (phosphate buffer), and pH 10 (carbonate buffer). The textured appearance of the image shows that the
grid is covered by a negatively stained amorphous film derived from peptidomimetic 3 (the grid has a uniform appearance in the absence of sample). No
ordered structures are apparent.
3 (Table 1) were synthesized, in which the Val residues in the
peptide strands were replaced by Ile and cyclohexylalanine
(Cha), respectively. These peptidomimetics (50 µM) were
incubated without agitation for 16 h at pH 5 (50 mM sodium
acetate buffer, 100 mM NaCl), pH 7 (50 mM sodium phosphate
buffer, 100 mM NaCl), or pH 10 (sodium carbonate buffer, 100
mM NaCl), and the assemblies formed were imaged by
transmission electron microscopy (TEM).
Peptidomimetic 1 formed ribbons composed of laterally
associated filaments at pH 5, consistent with our previous
observations (Figure 2A).30 Some of the ribbons were twisted,
likely because of interactions between the ribbons and the EM
grid. Whatever the interpretation, it is clear that the periodicity
of the twist in the imaged ribbons was highly variable.
Peptidomimetic 1 also formed filaments at pH 7 and 10 (Figure
2B,C), and while these filaments had some tendency to co-align
(especially at pH 7), they were not tightly associated. In contrast
to the variety of morphologies found for 1, peptidomimetic 2
formed predominantly discrete 50-60 Å wide filaments (the
average widths of 12 randomly chosen filaments were 50 ( 6,
60 ( 12, and 62 ( 8 Å at pH 5, 7, and 10, respectively),
regardless of the buffer in which assembly was carried out
(Figure 2D-F). Fewer than 10% of the species observed were
composed of multiple laterally associated filaments. More
strikingly, peptidomimetic 3 did not form well-ordered as-
semblies under any conditions; an amorphous film was observed
instead (Figure 2G-I). These results are not consistent with
the interdigitated strand model, but can be rationalized using
the side chain hydrophobic collapse model as follows. The Ile
side chains in 2, which are only slightly larger than Val side
chains, can still be accommodated in the gap between the two
peptide strands attached to the dibenzofuran template. Thus, 50-
60 Å wide filaments can still form. The space required by the
additional methyl group of the Ile side chain, however, forces
the peptide strands apart to a degree sufficient to prevent further
assembly by the filaments (which requires buffer anion-mediated
electrostatic and/or hydrogen bonding interactions between the
edges of the filaments; see Figure 1C). The Cha side chains in
peptidomimetic 3, in contrast, are too large to occupy the gap
between the peptide strands. Consequently, peptidomimetic 3
9
8566 J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005

Morphology of Cross
â
-Sheet Assemblies
A R T I C L E S
Scheme 2
cannot undergo side chain hydrophobic collapse or self-assembly
(since these processes are likely thermodynamically linked).
These results compel us to prefer the side chain hydrophobic
collapse model to the interdigitated strand model for the structure
of peptidomimetic assemblies, a model further scrutinized below.
We note, however, that a final determination cannot be made
until the structures of the assemblies are studied by a higher
resolution method, such as solid-state NMR. It is conceivable
that 3 could still assemble by the interdigitated strand model.
However, the fact that filamentous assemblies were not observed
suggests that this assembly pathway is not energetically favor-
able in comparison to amorphous aggregate formation, presum-
ably driven by nonspecific hydrophobic interactions.
characteristics of 1. For example, placing hydrophilic groups
at the 4 and 6 positions (on the periphery) of the dibenzofuran
template should reduce the propensity of 25-27 Å wide
protofilaments to assemble into 50-60 Å wide filaments (see
Figure 1C), since this would remove the hydrophobic driving
force for association (i.e. burial of the dibenzofuran-dominated
edge of the protofilament). To test this hypothesis, new
dibenzofuran templates with either ethylamine or propanoic acid
substituents at the 4 and 6 positions were prepared (DBF-EA
and DBF-PA, respectively; see the Supporting Information for
synthetic details) and incorporated into peptidomimetics 4 and
5 (Scheme 2).
The ethylamine and propanoic acid substructures attached to
the dibenzofuran template should make the protofilaments
formed by 4 and 5 about 5 Å wider than those formed by 1.
Thus, protofilaments and filaments of 4 and 5 (if formed) are
expected to be 30-32 and 55-65 Å wide, respectively.
Peptidomimetics 4 and 5 (50 µM) were incubated for 16 h at
Rational Control over Assembly Morphology Using the
Side Chain Hydrophobic Collapse Model as a Guide to
Peptidomimetic Design: Modification of the Dibenzofuran
Template. The dibenzofuran template, the peptide chains, and
the C-termini can all be modified to alter the assembly
Figure 3. (A-C) EM images of peptidomimetics 4 and 5 at pH 7 in phosphate or tris buffer. The expansions are 3-fold magnifications of 100 nm × 100
nm patches of the corresponding EM images. Each expansion contains several typical protofilaments.
9
J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005 8567

A R T I C L E S
Deechongkit et al.
Figure 4. Possible assembly mode for peptidomimetic 5, in which the
positions of charged groups alternate from one layer to the next. Protofila-
ments of peptidomimetic 5 with this structure are expected to be 37-39 Å
wide.
Figure 5. EM image of peptidomimetic 6 at pH 10 (carbonate buffer).
The expansion is a 3-fold magnification of a 100 nm × 100 nm patch of
the EM image, containing several typical filaments.
pH 5 (50 mM sodium acetate buffer, 100 mM NaCl), pH 7 (50
mM sodium phosphate or tris buffer, 100 mM NaCl), or pH 10
(sodium carbonate buffer, 100 mM NaCl), and the assemblies
formed were imaged by TEM.
Figure 3B,C). On the basis of the logic outlined above, these
structures were characterized as protofilaments. In contrast to
those of 4, these protofilaments showed no tendency to assemble
into higher order structures. The absence of both filaments and
looser assemblies of protofilaments demonstrates that assembly
widths can be better controlled using the DBF-PA template.
The lack of lateral assembly by protofilaments of 5 can be
rationalized by noting that, at pH 7, 5 is zwitterionic, whereas
4 bears four positive charges. We speculate that peptidomimetic
5 could assemble as shown in Figure 4, with the template on
alternating edges so that the positively and negatively charged
functional groups could complement each other. Although the
model shown in Figure 4 is speculative, we note that this mode
of assembly should reduce the tendency of the protofilaments
to self-associate further through anion-mediated and/or hydrogen
bonding interactions.
Peptidomimetic 4 did not self-assemble at pH 5 in acetate
buffer, pH 7 in tris buffer, or pH 10 in carbonate buffer, possibly
because the 4,6-ethylamine substituents on the dibenzofuran
template increased the solubility of the monomer. However, it
did assemble at pH 7 in phosphate buffer. The assembly in
phosphate buffer, but not in tris buffer at pH 7, is likely due to
the presence of a divalent anion (HPO42-) in the phosphate
buffer. Divalent ions are known to have stronger effects on
assembly processes than monovalent ions, presumably through
charge neutralization.42 The assemblies formed were 38 ( 5 Å
wide (n ) 12). This width is much closer to that expected for
a protofilament (6 Å too wide) than a filament (17 Å too thin).
We therefore assign the observed structures to be protofilaments.
Roughly 33% of the discrete species formed by 4 consisted of
two or more laterally associated protofilaments. These higher
order structures are not like the filaments formed by 1, because
individual protofilaments can be identified within them (i.e. they
are not condensed; see the inset to Figure 3A), whereas
Control over Assembly Structure by Modification of the
Peptide Strands. According to the side chain hydrophobic
collapse model, the width of the protofilaments and filaments
(and therefore the periodicity of the ribbons composed of these
assemblies) should depend directly on the length of the peptide
strands in a given peptidomimetic. Therefore, it should be
possible to control the assembly widths by altering the length
of the peptide strands. To test this hypothesis, peptidomimetic
6 was prepared (Table 1), which is identical to 1 except that
the peptide strands are twice as long (VTVTVTVT). On the
basis of the side chain hydrophobic collapse assembly model,
protofilaments of 6 should be 13.8 Å wider (39-41 Å wide)
than those of 1 and filaments of 6 should be 27.6 Å wider (78-
88 Å wide) than those of 1 (typical â-strands are about 3.45 Å
long per residue43). Peptidomimetic 6 (50 µM) was incubated
for 16 h at pH 5 (50 mM sodium acetate buffer, 100 mM NaCl),
pH 7 (50 mM sodium phosphate buffer, 100 mM NaCl), or pH
10 (50 mM sodium carbonate buffer, 100 mM NaCl), and its
assemblies were imaged by TEM (Figure 5).
individual protofilaments cannot be resolved in filaments of 1.
2-
The higher order structures formed by 4 likely occur by HPO4
-
mediated interactions between protofilament edges analogous
to those responsible for the formation of ribbons from filaments
of 1, possibly implying that higher order assembly is required
in the case of 4 to achieve stabilization. That condensed
filaments were not observed among the assemblies formed by
4 supports the hypothesis that modification of the dibenzofuran
template can inhibit the formation of filaments from protofila-
ments. The noncondensed lateral association of protofilaments
of 4, however, suggests that the ethylamine substructure is not
ideal for controlling the morphology of the assemblies.
Peptidomimetic 5, like 4, did not self-assemble at pH 5 in
acetate buffer or at pH 10 in carbonate buffer, but, unlike 4, it
self-assembled at pH 7 in either tris or phosphate buffer to form
assemblies that were 34 ( 4 or 38 ( 7 Å wide (n ) 12; see
(42) Israelachvili, J. N. Intermolecular and Surface Forces, 2nd ed.; Academic
(43) Fraser, R. D. B.; MacRae, T. P. Conformation in Fibrous Proteins and
Related Synthetic Polypeptides; Academic Press: London, 1973.
Press: London, 1991.
9
8568 J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005

Morphology of Cross
â
-Sheet Assemblies
A R T I C L E S
Figure 6. (A-C) EM images of peptidomimetic 7 at pH 5 (acetate buffer), pH 7 (phosphate buffer), and pH 10 (carbonate buffer). The expansions are a
3-fold magnification of 100 nm × 100 nm patches of the corresponding EM images. The expansion for A contains several typical filaments, while those for
B and C contain several typical protofilaments.
Scheme 3. Synthesis of Peptidomimetic 7
Peptidomimetic 6 yielded superaggregates of filamentous
assemblies at pH 5 and 7 (data not shown), in which the widths
of individual structures were ill-defined and therefore could not
be measured. However, individual assemblies formed by pep-
tidomimetic 6 could be distinguished at pH 10. These assemblies
were 88 ( 8 Å wide (n ) 12), or roughly 30 Å wider than
filaments of 1, at pH 10. This increase in width is consistent
with the side chain hydrophobic collapse model, but it should
be noted that this peptidomimetic appears to afford a more
heterogeneous distribution of assemblies than the others.
Peptidomimetic 7 (50 µM) was incubated for 16 h at pH 5
(50 mM sodium acetate buffer, 100 mM NaCl), pH 7 (50 mM
sodium phosphate buffer, 100 mM NaCl), or pH 10 (50 mM
sodium carbonate buffer, 100 mM NaCl), and the assemblies
formed were imaged by TEM (Figure 6).
Peptidomimetic 7 self-assembled at pH 5, forming assemblies
with an average width of 61 ( 6 Å (n ) 12), consistent with
the width expected for filaments (Figure 6A). These filaments
had some tendency to co-align to form noncondensed associated
higher order structuressribbons were observed in some samples
(data not shown). In contrast to its behavior at pH 5, peptido-
mimetic 7 formed assemblies with widths of 37 ( 4 and 38 (
4 Å (n ) 12 each) at pH 7 and 10, respectively, consistent with
the width of a protofilament. This unexpected result can be
rationalized by proposing that protofilaments of 7 have an
internal structure similar to that proposed for protofilaments of
5 (see Figure 4), in which the dibenzofuran template is found
on alternating edges of the protofilament. Peptidomimetic 7 may
be driven to adopt this structure to decrease the charge density
at the protofilament edges. We speculate that the high charge
densities in assemblies formed by 1 may perturb the pKa values
Control over Assembly Parameters by Modification of the
C-Termini. According to the side chain hydrophobic collapse
model (Figure 1C), anion-mediated electrostatic and/or hydrogen
bonding interactions between C-terminal DMDA groups on 1
control the lateral assembly of filaments to form structures wider
than filaments (fibrils and ribbons). Altering the properties of
the C-terminus should therefore alter the ability of the filaments
to form fibrils and ribbons. To test this hypothesis, peptidomi-
metic 7, in which the DMDA group was replaced by the
permanently positively charged N,N,N-trimethylethylenediamine
moiety, was prepared (Scheme 3).
9
J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005 8569

A R T I C L E S
Deechongkit et al.
of the DMDA groups, causing them to be partially deprotonated
at pH 7 and completely deprotonated at pH 10. This would
enable peptidomimetic 1 to form the assemblies depicted in
Figure 1C at pH 7 and 10, but, because the charge on
peptidomimetic 7 cannot be removed, peptidomimetic 7 would
have to assemble as shown in Figure 4. It is conceivable that
each of these two modes of assembly could be observed with
a single peptidomimetic under different solution conditions.
assembly process at pH 7 and 10 can be more effectively and
generally halted at the filament stage by exchanging the Val
residues in the strands for Ile residues. Finally, greatly increasing
the size of the R-amino acid side chains can altogether block
ordered hydrophobic collapse and the linked assembly process.
We hope that this work will encourage the development of
analogous, rationally controllable â-sheet assemblies with
desirable and designable properties.
Conclusions
Acknowledgment. We gratefully acknowledge financial
support from the NIH (Grant GM 51105), the Skaggs Institute
for Chemical Biology, the Lita Annenberg Hazen Foundation,
and a Norton B. Gilula Fellowship (S.D.). We thank Dr. Theresa
Fassel for a helpful discussion.
The assembly of peptidomimetic 1 and its analogues is best
described by the side chain hydrophobic collapse model
illustrated in Figure 1C. This structural model provided a rich
variety of approaches by which the assembly morphology of
this class of peptidomimetics could be rationally altered.
Increasing the hydrophilicity of the dibenzofuran template with
propionic acid moieties at the 4 and 6 positions can arrest the
assembly process at the protofilament stage. The assembly
process at pH 5 can be made to favor filaments by changing
the C-terminus from a DMDA to a TMDA group, but this
unexpectedly yields protofilaments at higher pH values. The
Supporting Information Available: All experimental details,
including syntheses of the modified dibenzofuran templates
DBF-EA and DBF-PA (PDF). This material is available free
of charge via the Internet at http://pubs.acs.org.
JA050558C
9
8570 J. AM. CHEM. SOC. VOL. 127, NO. 23, 2005