Combination of phage display and molecular grafting generates highly specific tumor-targeting miniproteins

Article abstract of DOI:10.1002/anie.201203857

Peptides are of growing interest as a non-immunoglobulin alternative for the development of new molecular entities. Disulfide-stabilized miniproteins are a class of peptides that represent an ideal template for the development of stable affinity reagents in drug design, molecular diagnostics, and targeted therapy. This is due to their outstanding proteolytic stability, their small size, and especially their tolerance to mutagenesis.

Full text of DOI:10.1002/anie.201203857

.
Angewandte
Communications
DOI: 10.1002/anie.201203857
Peptide Design
Combination of Phage Display and Molecular Grafting Generates
Highly Specific Tumor-Targeting Miniproteins**
Frederic Zoller, Annette Markert, Philippe Barthe, Wenye Zhao, Wilko Weichert,
Vasileios Askoxylakis, Annette Altmann, Walter Mier, and Uwe Haberkorn*
Peptides are of growing interest as a non-immunoglobulin
alternative for the development of new molecular entities.[1]
Disulfide-stabilized miniproteins are a class of peptides that
represent an ideal template for the development of stable
affinity reagents in drug design,[2] molecular diagnostics, and
targeted therapy.[3,4] This is due to their outstanding proteo-
lytic stability, their small size, and especially their tolerance to
mutagenesis.[3,4]
negatively influence the synthesis yield, because autonomous
folding results in regioisomer formation owing to incorrect
disulfide connectivity, and requires the appropriate oxidative
folding methods.[10,12,13] Therefore, we combined the Min-23
based phage-display technology with a different peptide
scaffold—the sunflower trypsin inhibitor (SFTI-I)—to gen-
erate a readily accessible peptide format for hit-to-lead
development in vitro and in vivo.
These structural characteristics also make disulfide-stabi-
lized miniproteins suitable candidates for use in the scaffold
method, that is, the integration of an affinity function into
a well-defined, stably folded structural framework by locally
reshaping its primary structure,[5] which in itself is a promising
approach for the design of new binding motifs and to stabilize
linear peptide binders.[6] Disulfide-stabilized miniproteins
could therefore be used as scaffolds for combinatorial peptide
library construction, such as phage display for the screening of
huge molecular repertoires of potential binders against
virtually any given target.[7,8] Herein, we used the disulfide-
stabilized miniprotein Min-23 as a molecular scaffold in
a phage display approach to generate a combinatorial library
for the identification of new affinity functions against the
angiogenesis marker delta-like ligand 4 (Dll4).
Min-23 is a two disulfide-bridge stabilized scaffold, which
was rationally designed by miniaturization of its parent
knottin Ecballium elaterium trypsin inhibitor II (EETI-II).[9]
Min-23 offers outstanding proteolytic stability and beneficial
pharmacokinetic properties for in vivo applications including
molecular imaging.[10] Furthermore, Min-23 has already been
applied successfully as a structural template for phage display
by insertion of up to ten amino acids into its hypervariable,
surface-exposed GPNG loop.[11] However, random amino
acid substitutions within disulfide-constrained proteins often
SFTI-I is a cyclic 14-residue peptide with a single disulfide
bridge, isolated from sunflower seeds.[14] In various studies,
the suitability of SFTI-I as a scaffold for the development of
peptide-based drug candidates was investigated.[15] It has been
shown, that the KSIPPI-domain, known as the binding loop, is
tolerant to amino acid substitutions[16] and allows for the
integration of binding epitopes.[6] Moreover, backbone modi-
fications and labeling concepts have been developed to
facilitate the in vivo application of SFTI-I derivatives, which
have used the acyclic as well as the head-to-tail cyclic
variants.[17]
The cell membrane protein Dll4 is highly overexpressed
on endothelial and tumor cells and highly involved in
angiogenic balance by binding to the Notch-1 receptor.
Blockage of the Dll4/Notch-1 interaction initiates non-
productive tumor vascularization and causes a delay of
tumor growth.[18,19] Thus, this pathway is a promising target
for cancer therapy and molecular diagnosis of tumor angio-
genesis.[20,21]
Screening of
a combinatorial Min-23 phage-display
library, which incorporates a random peptide sequence of
eight residues between Cys16 and Phe25, a Dll4-specific
binding domain was identified (sequence of the screening
hit:
H2N-LMRCKQDSDCLAGSVCLFHLFIYIFCG-
COOH, binding residues shown in bold, further details are
[*] Dr. F. Zoller,[+] Dr. A. Markert,[+] Dr. V. Askoxylakis, Dr. A. Altmann
DKFZ, Klinische Kooperationseinheit Nuklearmedizin
[+] These authors contributed equally to this work.
[**] This project was supported by the Bundesministerium fꢂr Bildung
und Forschung (Grant-Nos.: 13N10269, 01EZ0807) and the
Deutsche Forschungsgemeinschaft (Grant-No.: HA 2901/6-1). We
are grateful to Thomas Lindner, Annabell Marr, Jennifer Melzer,
Jessica Angel, Ulli Bauder-Wꢂst, Karin Leotta, Uschi Schierbaum,
Sabine Bitter, Gabriela Glensch, Ulrike Hebling, Iris Wolf, Mechthild
Samer, Eileen Gꢁrtner, and the DKFZ Light Microscopy Facility for
technical support. Dr. Matthias Strieker and Dr. Thomas Fleming
are acknowledged for careful revision of the manuscript.
Im Neuenheimer Feld 280, 69120 Heidelberg (Germany)
Dr. P. Barthe
Centre de Biochimie Structurale, Institut National pour la Santꢀ et la
Recherche Mꢀdicale U1054, Centre National pour la Recherche
Scientifique Unitꢀ Mixte de Recherche 5048,
Universitꢀ Montpellier 1 and 2, 34090 Montpellier (France)
Prof. Dr. W. Weichert
Universitꢁtsklinik Heidelberg, Institut fꢂr Pathologie
Im Neuenheimer Feld 220, 69120 Heidelberg (Germany)
Supporting information for this article (experimental details) is
available on the WWW under http://dx.doi.org/10.1002/anie.
201203857. All animal experiments were carried out in accordance
with the national animal guidelines.
W. Zhao, Dr. W. Mier, Prof. Dr. U. Haberkorn
Universitꢁtsklinik Heidelberg
Radiologische Klinik, Abteilung Nuklearmedizin
Im Neuenheimer Feld 400, 69120 Heidelberg (Germany)
E-mail: uwe.haberkorn@med.uni-heidelberg.de
13136
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 13136 –13139

Angewandte
Chemie
given in the Supporting Information). However, the chemical
synthesis of this Min-23 derivative resulted in a poor yield
because of precipitation and formation of regioisomers during
the folding process. To overcome these limitations, the
binding residues of the Min-23 phage-display-evolved peptide
were transferred into the variable loop of the SFTI-I. On the
contrary, phage-display screening using an SFTI-I scaffold
library directly did not yield a Dll4-positive binding species
(Supporting Information).
SFTI-I 10 and the receptor Notch-1 were used as controls.
Specific binding interactions with the recombinant fusion
protein Dll4-Fc were only observed for miniprotein 1, with
a KD value of 22 nm (Figure 1; Fc = a region of human IgG1).
The influence of the individual amino acids on the Dll4-
binding properties was investigated by single-residue substi-
In contrast to making multiple disulfide-constrained
peptides, this approach allowed highly reliable chemical
synthesis of the SFTI-I derivate using automated solid-
phase peptide synthesis (SPPS) followed by iodine oxidation
of the fully deprotected peptide in solution (Scheme 1). The
Figure 1. Surface plasmon resonance (SPR) analysis (RU=response
units). The miniprotein 1 (a,b) and the unmodified scaffold 10 (c,d)
were tested with the recombinant fusion protein Dll4-Fc (a,c) and
Notch-1 (b,d) as control. The proteins were immobilized on a Biacore
CM5 sensor chip and the peptide concentrations were: 0.25, 0.5, 0.75,
1, 1.25 mm (n=2). Kinetic data from the SPR measurements are given
in Table 1.
Scheme 1. Synthesis of SFTI-I miniproteins. The synthesis route as
shown for compound 1 was applied for all SFT-I derivatives in this
study. DIPEA=diisopropylethylamine, Fmoc=fluorenylmethoxy car-
bonyl, HBTU=O-(benzotriazol-1-yl)-N,N,N’,N’-tetramethyluronium
hexafluorophosphate, NMP=N-methylpyrolidone, Pbf=2,2,4,6,7-pen-
tamethyldihydrobenzofuran-5-sulfonyl, tBu=tert-butyl, TFA=trifluoro
acetic acid, TIPS=triisopropyl silane, Trt=trityl.
tution by alanine of the binding domain spanned by the
residues Thr4 to Cys13. SPR analysis was also conducted to
demonstrate the pharmacological activity of peptide 1 as
a potential inhibitor of the Dll4/Notch-1 interaction (Sup-
porting Information).
The resulting eight disulfide-stabilized, alanine substi-
tuted peptides 2–9 were analyzed by SPR and their 125I-
labeled derivatives were used in cell-binding assays with the
Dll4-positive cell line PC-3 (Supporting Information). In both
experiments, substitution with alanine had a significant effect
on the Dll4-binding capability of the miniprotein (Table 1).
Substitution of the lipophilic functions (Phe6,9; peptides 3 and
6, and Ile10,12; peptides 7 and 9) resulted in a loss of binding
potential compared to the unmodified peptide 1. Replace-
ment of Leu5, His7, and Leu8 (peptides 2, 4, and 5) by alanine
resulted in decreased binding affinities (higher KD values) and
moderate cellular binding ratios as compared to 1.
The cellular binding capability of the SFTI-I derivatives
was determined by applying fluorescently labeled (5(6)FAM;
Figure 2) and radiolabeled peptides to different Dll4-express-
ing cell lines (Supporting Information). Fluorescent staining
of HUVECs and PC-3 cells clearly showed the cell-membrane
binding of 5(6)FAM-1b compared to the alanine-mutated
derivative 5(6)FAM-6. Competition for binding with the
unlabeled peptide 1 resulted in a significant decrease in green
fluorescence intensity. These results were confirmed by flow
cytometry (Figure 2). 5(6)FAM-1b was observed to have the
single disulfide bridge was formed within one to two hours at
ambient temperature in dilute acetic acid, resulting in overall
yields up to 39% and high purity (> 95%). The improved
synthesis of SFTI-I peptides also enabled convenient chem-
ical modifications including amino acid substitutions and
fluorescent- or radiolabeling for in vitro and in vivo charac-
terization (Supporting Information).
Hence, the molecular grafting strategy by integration of
an identified binding site into the structurally defined scaffold
SFTI-I represents a promising approach to rationally design
a small, highly stable miniprotein. This concept offers
a reliable hit-to-lead development of new peptide entities
with easy synthesis and high yields as compared to the
complex synthesis of Min-23. Currently, this scaffold strategy
is being pursued for the engineering of SFTI-I binders against
other tumor associated targets, such as B7-H3, that were
identified by in vitro display screening techniques (data not
shown).
Surface plasmon resonance (SPR) spectroscopy was
performed to verify the specificity of 1 for Dll4. The acyclic
Angew. Chem. Int. Ed. 2012, 51, 13136 –13139
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 13137

.
Angewandte
Communications
Table 1: Summary of SPR analysis and cellular binding of the SFTI-I
derivatives 1–11.
Information). These in vitro studies confirm the binding
specificity of 1 for Dll4.
[b]
[b]
[b]
The proteolytic stability of miniprotein 1 was determined
by incubation of the 125I-labeled derivatives in human serum.
Radio-HPLC analysis of the serum aliquots showed high
stability of the full-length peptide 125I-(Tyr11)-1 over a period
of ten days. In contrast, the non-stabilized, linear binding
sequence 125I-(Tyr7)-LFHLFIYI showed a half-life in human
serum of 2.2 hours. Prolonged incubation led to proteolytic
fragmentation, as indicated in the HPLC elution profiles by
the presence of metabolites with shorter retention times
(Supporting Information). The biotinylated peptide 1 showed
a strong membrane-associated immunohistochemical signal
in vascular endothelium and tumor cells of AR42J tumor
tissue. This histological corroboration highlights the potential
of the generated peptide for application in molecular
diagnostics (Figure 3a,b).
The in vivo tumor-targeting potential of the peptides was
determined by biodistribution profiling of the 131I-labeled
SFTI-I derivatives in AR42J tumor-bearing mice (Figure 3c).
The designed miniprotein 1 demonstrated a maximum tumor
accumulation of 0.50 Æ 0.08%ID/g (percent injected dose per
gram of tissue). High concentrations of radioactivity were
found in the liver, as a result of the lipophilic character of 1.
To increase the peptide concentration in the blood pool and to
enable prolonged availability for tumor accumulation, the
lipophilicity of the peptide was reduced by polar isosteric
substitution of the leucine residues Leu5 and Leu8 by
asparagine (11). The Dll4-binding capability of peptide 11
was verified by SPR (Table 1) and cellular binding assays
Peptide Binding
motif[a]
Kon
[mÀ1 sÀ1
Koff
[sÀ1
KD
[nm]
Cellular-binding
ratio[c]
]
]
1
LFHLFIYI
LFHLFIYI
AFHLFIYI
LAHLFIYI
LFALFIYI
LFHAFIYI
LFHLAIYI
LFHLFAYI
LFHLFIAI
LFHLFIYA
KSIPPI
5.56ꢃ105 0.0121 22
1.60ꢃ104 0.0014 88
5.34ꢃ104 0.0273 511
1.00Æ0.02
0.70Æ0.05
0.51Æ0.03
0.38Æ0.02
1.08Æ0.08
0.58Æ0.04
0.27Æ0.01
0.32Æ0.02
–/–
1b[d]
2
3
4
5
6
7
8
9
n.b.
n.b.
n.b.
7.25ꢃ104 0.0121 167
3.22ꢃ104 0.0214 665
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
n.b.
0.42Æ0.01
0.01Æ0.01
0.71Æ0.04
10[e]
11
NFHNFIYI 5.90ꢃ104 0.0012 20
[a] Binding motifs are framed by the residues H2H-Gly-Arg-Cys-Thr- and
-Cys-Phe-Pro-Asp-CONH2 of the disulfide-stabilized SFTI-I scaffold.
[b] Kinetic data from SPR measurments were determined with immobi-
lized fusion protein Dll4-Fc (loading level=8690 RU); n.b.: no binding
detected. Applied peptide concentrations: 0.25, 0.5, 0.75, 1, 1.25 mm
(n=2). [c] Cellular-binding capability to PC-3 cells was determined by
incubation with 125I-labeled peptides. Cellular-binding ratio=(%bind-
ing/dose/106 cells of peptide X)/(%binding/dose/106 cells of 1);
X=peptide 1–11 (meanÆSD; n=3). [d] Peptide 1b was labeled with
5(6)-carboxyfluorescein (5(6)FAM) at the N-terminus. [e] The acyclic
derivative of SFTI-I was used, and the Phe12 residue was substituted by
Tyr12 for radioiodination.
highest cell-bound fluorescence, whereas 5(6)FAM-6 had the
lowest signal at 52 Æ 1%. Competition with unlabeled peptide
1 led to 39 Æ 5% inhibition of cellular binding (Supporting
Figure 3. Immunohistochemical staining and in vivo tumor-targeting
properties of radiolabeled peptides. a,b) Immunostaining with biotiny-
lated peptide 1 on AR42J tumor tissue cryosections taken from mice
(overview, ꢃ20 (a) and high-power view, ꢃ40 (b)). c) Tumor-targeting
properties of the radiotracer 120 min after intravenous administration
were determined by organ biodistribution in AR42J tumor-bearing
mice. Tumor uptake of the miniprotein 131I-(Tyr11)-1, the unmodified
scaffold 131I-(Tyr12)-10, the polar isosteric analogue 131I-(Tyr11)-11, and
its competition with unlabeled 11 are shown (meanÆSD; n=3). Full
biodistribution data are given in the Supporting Information.
Figure 2. Cellular staining of fluorescently labeled miniproteins. Repre-
sentative confocal microscopy images and flow cytometry counts of
5(6)FAM-1b (a–c), 5(6)FAM-6 (d–f), and 5(6)FAM-1b after blocking
with unlabeled 1 (g–i). HUVEC cells (a,d,g) and PC-3 cells
(b,c,e,f,h,i) were used. 5(6)FAM-labeled peptides (green) were co-
stained with DAPI (blue) for visualization of the nuclei. Scale
bar=20 mm. Dll4 expression in the selected cell lines was validated by
RT-PCR and Western blot (Supporting Information).
13138 www.angewandte.org
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Angew. Chem. Int. Ed. 2012, 51, 13136 –13139

Angewandte
Chemie
(Supporting Information). The tumor uptake of 131I-(Tyr11)-11
increased to a maximum of 1.83 Æ 0.34%ID/g, and could be
blocked by co-administration of unlabeled 11 (51% inhibition
of tumor uptake). In contrast, the unmodified scaffold 131I-
(Tyr12)-10 demonstrated a tumor accumulation of only 0.53 Æ
0.10%ID/g at the selected time point (Figure 3c).
In conclusion, this study validates the molecular-grafting
concept for the development of new molecular entities for
therapeutic use. It was shown that the binding specificity
against the selected tumor-associated target was preserved by
transferring the phage-display-identified binding domain
from Min-23 into the SFTI-I peptide. Both the Dll4-binding
specificity and the tumor-targeting capability of the rationally
designed miniprotein were confirmed in vitro and in vivo.
However, further optimization is required to improve the
tumor accumulation of this novel anti-angiogenic drug
candidate. This design approach might facilitate the hit-to-
lead development of other previously identified affinity
functions in drug design.
[6] L. Y. Chan, S. Gunasekera, S. T. Henriques, N. F. Worth, S. J. Le,
R. J. Clark, J. H. Campbell, D. J. Craik, N. L. Daly, Blood 2011,
118, 6709.
[7] G. P. Smith, V. A. Petrenko, Chem. Rev. 1997, 97, 391.
[8] A. Marr, A. Markert, A. Altmann, V. Askoxylakis, U. Haber-
korn, Methods 2011, 55, 215.
[9] A. Heitz, D. Le-Nguyen, L. Chiche, Biochemistry 1999, 38,
10615.
[10] F. Zoller, T. Schwaebel, A. Markert, U. Haberkorn, W. Mier,
ChemMedChem 2012, 7, 237.
[11] C. Souriau, L. Chiche, R. Irving, P. Hudson, Biochemistry 2005,
44, 7143.
[12] N. L. Daly, S. Love, P. F. Alewood, D. J. Craik, Biochemistry
1999, 38, 10606.
[13] O. Avrutina, H. U. Schmoldt, D. Gabrijelcic-Geiger, D. Le
Nguyen, C. P. Sommerhoff, U. Diederichsen, H. Kolmar, Biol.
Chem. 2005, 386, 1301.
[14] S. Luckett, R. S. Garcia, J. J. Barker, A. V. Konarev, P. R.
Shewry, A. R. Clarke, R. L. Brady, J. Mol. Biol. 1999, 290, 525.
[15] M. L. Korsinczky, H. J. Schirra, D. J. Craik, Curr. Protein Pept.
Sci. 2004, 5, 351.
[16] N. L. Daly, Y. K. Chen, F. M. Foley, P. S. Bansal, R. Bharathi,
R. J. Clark, C. P. Sommerhoff, D. J. Craik, J. Biol. Chem. 2006,
281, 23668.
Received: May 18, 2012
Revised: August 22, 2012
Published online: November 13, 2012
[17] R. G. Boy, W. Mier, E. M. Nothelfer, A. Altmann, M. Eisenhut,
H. Kolmar, M. Tomaszowski, S. Kramer, U. Haberkorn, Mol.
Imaging Biol. 2010, 12, 377.
[18] J. Ridgway, G. Zhang, Y. Wu, S. Stawicki, W. C. Liang, Y.
Chanthery, J. Kowalski, R. J. Watts, C. Callahan, I. Kasman, M.
Singh, M. Chien, C. Tan, J. A. Hongo, F. de Sauvage, G.
Plowman, M. Yan, Nature 2006, 444, 1083.
[19] I. Noguera-Troise, C. Daly, N. J. Papadopoulos, S. Coetzee, P.
Boland, N. W. Gale, H. Chieh Lin, G. D. Yancopoulos, G.
Thurston, Nature 2006, 444, 1032.
[20] N. S. Patel, J. L. Li, D. Generali, R. Poulsom, D. W. Cranston,
A. L. Harris, Cancer Res. 2005, 65, 8690.
[21] D. J. Hicklin, Nat. Biotechnol. 2007, 25, 300.
Keywords: diagnostic agents · drug discovery · peptides ·
protein structure · radiochemical analysis
.
[1] H. K. Binz, P. Amstutz, A. Pluckthun, Nat. Biotechnol. 2005, 23,
1257.
[2] D. J. Craik, R. J. Clark, N. L. Daly, Expert Opin. Invest. Drugs
2007, 16, 595.
[3] F. Zoller, U. Haberkorn, W. Mier, Molecules 2011, 16, 2467.
[4] A. Skerra, Curr. Opin. Biotechnol. 2007, 18, 295.
[5] P. A. Nygren, A. Skerra, J. Immunol. Methods 2004, 290, 3.
Angew. Chem. Int. Ed. 2012, 51, 13136 –13139
ꢀ 2012 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.angewandte.org 13139