High-affinity peptide-based collagen targeting using synthetic phage mimics: From phage display to dendrimer display

Article abstract of DOI:10.1021/ja902285m

(Figure Presented) Peptides derived from phage display typically show significantly weaker binding than their respective high affinity phage, which can bind to protein surfaces in a multivalent fashion. Here we show that mimicking key aspects of the multi

Full text of DOI:10.1021/ja902285m

Published on Web 07/31/2009
High-Affinity Peptide-Based Collagen Targeting Using Synthetic Phage
Mimics: From Phage Display to Dendrimer Display
Brett A. Helms, Sanne W. A. Reulen, Sebastiaan Nijhuis, Peggy T. H. M. de Graaf-Heuvelmans,
Maarten Merkx, and E. W. Meijer*
Laboratory of Chemical Biology and Institute for Complex Molecular Systems, EindhoVen UniVersity of Technology,
P.O. Box 513, 5600 MB EindhoVen, The Netherlands
Received March 23, 2009; E-mail: E.W.Meijer@tue.nl
Collagens are extracellular matrix (ECM) proteins that provide
mechanical strength to tissues and are important in cell attachment,
differentiation, and migration.1,2 Visualization of collagen networks
in tissues is of great interest, particularly in light of the many human
ailments that are characterized by ECM degradation or turnover
such as angiogenesis, myocardial infarction, and atherosclerosis.3,4
Naturally occurring collagen binding proteins such as the von
Willebrand Factor (vWF) or CNA355-7 labeled with fluorescent
dyes or collagen-specific antibodies8 are frequently employed for
collagen imaging. Protein-based targeting strategies have a few key
limitations, however: tissue penetration depth decreases with
increasing size, and naturally occurring collagen binding proteins
Figure 1. From phage display to dendrimer display: phage display to
collagen reveals a consensus binding sequence that is translated into a high
are often promiscuous in binding to collagen’s many subtypes,
effectively reducing the structural information gleaned with these
probes.5,6 The use of short peptides as targeting ligands presents
an attractive alternative: their activity often does not require folding,
and their chemical synthesis is inexpensive in comparison to, e.g.,
antibodies.
affinity, versatile synthetic collagen-specific probe by mimicking the original
pentavalent phage architecture on a dendritic wedge.
demonstrated using a collagen-specific 7-mer peptide sequence
against rat tail collagen type I but is likely to be broadly applicable
to other multivalent peptide or protein architectures.
Collagen binding phages were identified from commercially
available M13 phage libraries containing linear 7-mer, linear 12-
mer, or cyclic 7-mer peptides. Initially, strong background binding
was observed between the phage and rat tail collagen type I coated
on 96-well plates, probably due to Coulombic interactions between
the polyanionic phage and the positively charged collagen. Screen-
ing of several buffer conditions revealed that this background
binding was efficiently suppressed at 0.5 M NaCl. ELISA experi-
ments showed clear enrichment with collagen binding phage for
the linear 7-mer library after two rounds of panning, whereas little
enrichment was observed for either of the other two libraries (Figure
S1). Subsequent panning experiments were done with the linear
7-mer library using either low pH elution or competitive elution
with the collagen binding protein CNA35. Highly homologous
peptide sequences were obtained for all panning experiments with
nearly all clones containing the consensus sequence H-V-F/W-Q/
M-Q-P/A-P/K (Figure S1). This newly discovered consensus
sequence is markedly different from any of the previously reported
collagen binding peptides.11,13,16 To mimic the multivalent peptide
presentation on the M13 phage head, a pentavalent dendritic
platform was constructed, which was designed to incorporate
various imaging modalities at the focal point in addition to five
copies of the collagen binding phage peptide (Scheme 1).
Phage display has proven to be a particularly efficient method
to rapidly screen large peptide libraries for targeting ligands to a
broad variety of substrates.9 Peptides derived from phage display,
however, suffer from a serious drawback in that they typically show
significantly weaker binding than their respective high affinity
phage. This applies in particular to peptides targeting protein
surfaces and results from the intrinsic flexible nature and limited
binding interface of peptides.10 For example, the WREPSFCALS
peptide derived from vWF has been reported to bind bovine
collagen type I with an apparent Kd of only 100 µM.11 In related
work, Caravan et al. reported the development of a disulfide bond
constrained collagen binding peptide, but its low µM affinity was
obtained only after extensive affinity maturation via the incorpora-
tion of non-natural amino acids.12,13
As collagens and other ECM proteins have repetitive primary
sequences and elaborate structural hierarchies permitting multivalent
interactions with cells, a targeting strategy based on multivalent
peptide binding is highly attractive. In fact, each phage in the widely
used M13 libraries presents five peptide ligands fused to the
N-terminus of the phage pIII coat protein at one end of the phage
particle. Affinity selection of peptide ligands using these libraries
might thus rely on not only the specific amino acid sequence
displayed at the phage head but also their presentation in multiple
copies capable of interacting with collagens in a multivalent fashion.
While a large number of multivalent approaches have been disclosed
in recent years, most of them are designed for structurally well-
characterized targets such as bacterial toxins and lectins.14,15 Here
we report a multivalent approach to enhance the affinity of phage-
display derived peptides through reconstruction of the phage’s
multivalent architecture using carefully designed dendritic wedges
as synthetic multivalent scaffolds (Figure 1). The approach is
The branched core of the pentameric structure was built up
via a double Michael addition of the bivalent acrylamide 10 to
the linear nitroalkane 6 (Scheme 1 and Supporting Information).
This AB5 synthon is versatile, incorporating oligoethyleneglycol-
based flexible linkers to peptide ligands at the periphery and
nitro, biotin, or fluorescein moieties at the focal point of the
wedge. Final construction of the collagen binding phage mimics
was accomplished by native chemical ligation of H2N-H(Dnp)-
9
10.1021/ja902285m CCC: $40.75  2009 American Chemical Society
J. AM. CHEM. SOC. 2009, 131, 11683–11685 11683

C O M M U N I C A T I O N S
Scheme 1. Synthesis of Pentavalent Dendritic Wedges as Scaffolds for Phage Peptide Presentationa
a For reaction conditions, see Figure S2.
VWMQAPGGG-MPAL thioesters to wedges bearing five N-
terminal cysteines (19, 24, and 27). Complete conversion to the
pentameric collagen binding constructs (20, 25, and 28) was
observed after 16 h at 37 °C in the presence of benzyl mercaptan
with thiophenol as the catalyst.17
The effect of multivalent display on collagen affinity was studied
by ELISA assays using biotinylated monovalent and pentavalent
constructs and substrates coated with rat tail collagen type I (Figure
2A). Monovalent H2N-HVWMQAPGGGK(Biotin)-NH2 yielded a
Kd ) 61 ( 5 µM, while a scrambled or reversed sequence showed
no significant binding above background. Titration experiments for
the pentameric phage analogue 25 revealed a 100-fold improvement
in collagen affinity, showing the same affinity (Kd ) 550 ( 100
nM) as that of the natural collagen protein CNA35. Moreover, the
plot for 25 seems to show a more cooperative affinitysa more steep
increase at the low concentration part of the titrationsthan a simple
1:1 Langmuir binding model (see Supporting Information). To
understand how the increase in affinity is reflected in the association
and dissociation kinetics, surface plasmon resonance (SPR) was
employed. Not surprisingly, the monovalent peptide showed fast
association and dissociation kinetics to rat tail collagen type I
immobilized on a dextran-coated biosensor surface (Figure 2B).
Association of the pentamer 20, in contrast, was slower, possibly
due to its larger size. The most dramatic effect of the pentamer
manifests in its dissociation kinetics, where persistent adhesion of
20 to collagen was observed (Figure 2C). Low pH elution, similar
to that used for the initial phage panning, was required to release
the pentamer from the collagen-coated SPR surface. The enhance-
ment in binding over 2 orders of magnitude between the free peptide
and the peptide pentamer suggests that dendrimer display is capable
of harnessing at least part of the favorable binding characteristics
of the peptide-displaying phage.
Figure 2. Binding properties of pentavalent and monovalent collagen
binding peptides. (a) Binding of biotinylated monovalent and pentavalent
peptides to immobilized rat tail collagen type I was detected using HRP-
conjugated antibiotin antibody. Solid lines represent fits to a 1:1 Langmuir
binding model. (b/c) SPR analyses of the association (i) and dissociation
(ii) kinetics of monovalent (b) and pentavalent (c) peptides binding to rat
tail collagen type 1. Plots show reference subtracted responses obtained by
flowing 1 mM H2N-HVWMQAPGGGC-NH2 (b) or 25 µM 20 (c) over
CM5 chips immobilized with 5000 RU (b) and 1500 RU (c) of collagen
type I. Conditions: HBS-EP buffer pH 7.4 with 1 mM DTT at 25 °C.
Regeneration and re-equilibration (iii): 10 mM glycine pH 1.5 for 2 × 30 s
then running buffer for 5 min.
together provide an average distance that is similar to the 2-3 nm
spacing between two adjacent pIII domains on the phage. Unlike
the peptides displayed on the phage, which are restricted by the
pIII proteins, the peptide ligands on the dendritic wedge most likely
assume a random coil conformation, acting independently in their
interaction with collagen. For example, tryptophan fluorescence and
circular dichroism experiments (Figures S17 and S18) did not show
any evidence for the formation of secondary or tertiary structure
upon multimerization of the collagen binding peptide onto the
dendritic wedge. To provide a starting point for understanding the
effect of multivalency in these systems, we analyzed the binding
using the minimal model of a bivalent interaction. The enhanced
affinity can be understood by assuming that binding of one peptide
The AB5 dendritic wedge mimics several key features of the
phage, including its overall valency and its conjugation of peptide
ligands via their C-terminus. In this case, the collagen binding
peptides are linked using a short GGG spacer and PEG chains that
9
11684 J. AM. CHEM. SOC. VOL. 131, NO. 33, 2009

C O M M U N I C A T I O N S
are thus consistent with previous findings that CNA35 actually binds
to various types of collagen with similar affinity6 and suggest that
phage display is a suitable approach to identify more specific
targeting ligands.23
In summary, the AB5 dendritic wedge represents a well-defined,
highly versatile platform for the affinity enhancement of phage-
display derived peptides by mimicking key aspects of the multi-
valent architecture of the phage head. The collagen targeting ligands
presented here provide attractive alternatives for antibody- and
protein-mediated targeting of collagen remodeling in a number of
disease processes. These results thus emphasize the advantage of
combining the strength of biological display methods for the affinity
selection of peptide ligands with the ability of synthetic chemistry
to provide a wide variety of functional groups and 3D topologies.
Acknowledgment. We thank Katy Nash Krahn, Ingrid van Baal,
Tilman Hackeng (Maastricht), and Jos Raats (Nijmegen) for
experimental support. This work was supported by NWO VIDI
Grant 700.56.428 to M.M., a SPINOZA grant to E.W.M., and the
BSIK Program BSIK03033.
Figure 3. Laser scanning confocal microscopy (LSCM) images of 1 × 1
cm porcine pericardium sections showing the performance of dendrimer-
displayed collagen binding peptides in the ex vivo imaging of collagen in
native tissue. Tissues were incubated for 3 h in HBS pH 7.4 with 0.6 µM
28 (a) or H2N-HVWMQAPGGGK(fluorescein)-NH2 at concentrations of
0.6 µM (b) and 60 µM (c). (d-f) Costaining of a tissue section with 6 µM
AlexaFluor568-labeled CNA35 in HBS pH 7.4 for 3 h, followed by 0.6
µM 28 for an additional 2 h (d: green channel showing 28: e: red channel
showing AlexaFluor568-labeled CNA35; f: overlay of green and red
channels). The scale bar represents 50 µm.
Supporting Information Available: Experimental details regarding
phage display, dendrimer and peptide synthesis, in Vitro binding
experiments, and collagen imaging. This material is available free of
charge via the Internet at http://pubs.acs.org.
ligand increases the effective molarity (EM) to 0.34 ( 0.07 mM
for binding to a second receptor site.18 This number is consistent
with the length of the flexible linker used here and an average
distance of ∼30 Å between two peptide binding sites.19 This
analysis also indicates that multivalent binding is only effective
when EM > Kinter and that statistical factors contribute to enhance
the effect of multivalent interactions. A more detailed characteriza-
tion of the binding specificity of phage peptide dendrimers and its
dependency on the molecular architecture, valency of the wedge,
and density of binding sites is ongoing.
References
(1) Even-Ram, S.; Yamada, K. M. Curr. Opin. Cell Biol. 2005, 17, 524–532.
(2) Gelse, K.; Poschl, E.; Aigner, T. AdV. Drug DeliVery ReV. 2003, 55, 1531–
1546.
(3) Adiguzel, E.; Ahmad, P. J.; Franco, C.; Bendeck, M. P. Vasc. Med. 2009,
14, 73–89.
(4) Myllyharju, J.; Kivirikko, K. I. Ann. Med. 2001, 33, 7–21.
(5) Boerboom, R. A.; Krahn, K. N.; Megens, R. T.; van Zandvoort, M. A.;
Merkx, M.; Bouten, C. V. J. Struct. Biol. 2007, 159, 392–399.
(6) Krahn, K. N.; Bouten, C. V.; van Tuijl, S.; van Zandvoort, M. A.; Merkx,
M. Anal. Biochem. 2006, 350, 177–185.
(7) Megens, R. T.; Oude Egbrink, M. G.; Cleutjens, J. P.; Kuijpers, M. J.;
Schiffers, P. H.; Merkx, M.; Slaaf, D. W.; van Zandvoort, M. A. Mol.
Imaging 2007, 6, 247–260.
Having established the favorable binding characteristics of the
phage mimics on collagen-coated substrates in vitro, we next
assessed their efficacy for selective collagen staining in the more
heterogeneous environment of native collagen-containing tissues.
We tested the performance of fluorescein-labeled monomers and
pentamers for collagen imaging in the parietal pericardium, a fibrous
sac around the heart that consists mainly of collagen type I and
some type III.20,21 Incubation of pig parietal pericardium with 0.6
µM fluorescein-labeled peptide pentamer 28 showed a distinct
fibrous network with fibers of an ∼2 µm thickness (Figure 3).
Consistent with the in Vitro binding studies, no collagen was
detected using the same concentration of the monovalent collagen
binding peptide. Similar collagen architectures as observed for the
pentamer at 0.6 µM were observed when using 60 µM of
the monovalent peptide, but the amount of background fluorescence
was significantly increased. To compare the properties of the peptide
pentamer to those of CNA35, the same tissue was costained with
0.6 µM 28 (green) and 6 µM AlexaFluor568-labeled CNA35 (red).
Both probes visualize collagen fibers oriented in the same direction,
but a more specific stain was observed for the peptide pentamer
compared to CNA35 (Figure 3D-F). The wave-like collagen
structures observed with the CNA35 agree with other imaging
studies of total collagen content in pericardial tissues.22 These results
(8) Hayes, A. J.; Hughes, C. E.; Caterson, B. Methods 2008, 45, 10–21.
(9) Smith, G. P.; petrenko, V. A. Chem. ReV. 1997, 97, 391–410.
(10) Sergeeva, A.; Kolonin, M. G.; Molldrem, J. J.; Pasqualini, R.; Arap, W.
AdV. Drug DeliVery ReV. 2006, 58, 1622–1654.
(11) Takagi, J.; Asai, H.; Saito, Y. Biochemistry 1992, 31, 8530–8534.
(12) Caravan, P.; Das, B.; Deng, Q.; Dumas, S.; Jacques, V.; Koerner, S.;
Kolodziej, A.; Looby, R. J.; Sun, W. C.; Zhang, Z. Chem. Commun. 2009,
430–432.
(13) Caravan, P.; Das, B.; Dumas, S.; Epstein, F. H.; Helm, P. A.; Jacques, V.;
Koerner, S.; Kolodziej, A.; Shen, L.; Sun, W. C.; Zhang, Z. Angew. Chem.,
Int. Ed. 2007, 46, 8171–8173.
(14) Choi, S.-K. Synthetic multiValent molecules. Concept and biomedical
applications; Wiley: Hoboken, NJ, 2004.
(15) Kitov, P. L.; Sadowska, J. M.; Mulvey, G.; Amstrong, G. D.; Ling, H.;
Pannu, N. S.; Read, R. J.; Bundle, D. R. Nature 2000, 403, 669–672.
(16) Depraetere, H.; Viaene, A.; Deroo, S.; Vauterin, S.; Deckmyn, H. Blood
1998, 92, 4207–4211.
(17) van Baal, I.; Malda, H.; Van Dongen, J. L. J.; Merkx, M.; Meijer, E. W.;
Synowsky, S. A.; Hackeng, T. M. Angew. Chem., Int. Ed. 2005, 44, 5052–
5057.
(18) Krishnamurthy, V. M.; Semetey, V.; Bracher, P. J.; Shen, N.; Whitesides,
G. M. J. Am. Chem. Soc. 2007, 129, 1312–1320.
(19) van Dongen, E. M. W. M.; Evers, T. H.; Dekkers, L. M.; Meijer, E. W.;
Klomp, L. W. J.; Merkx, M. J. Am. Chem. Soc. 2007, 129, 3494–3495.
(20) Keene, D. R.; Sakai, L. Y.; Bachinger, H. P.; Burgeson, R. E. J. Cell Biol.
1987, 105, 2393–2402.
(21) Simionescu, D.; Iozzo, R. V.; Kefalides, N. A. Matrix 1898, 9, 301–310.
(22) Rodriguez-Feo, J. A.; Sluijter, J. P. G.; de Kleijn, D. P. V.; Pasterkamp,
G. Curr. Pharm. Des. 2005, 11, 2501–2514.
(23) Lusvarghi, S.; Kim, J. M.; Creeger, Y.; Armitage, B. A. Org. Biomol. Chem.
2009, 7, 1815–1820.
JA902285M
9
J. AM. CHEM. SOC. VOL. 131, NO. 33, 2009 11685