Arginine- and lysine-specific polymers for protein recognition and immobilization

Article abstract of DOI:10.1021/ja0560229

Free radical polymerization of methacrylamide-based bisphosphonates turns weak arginine binders into powerful polymeric protein receptors. Dansyl-labeled homo- and copolymers with excellent water solubility are accessible through a simple copolymerization protocol. Modeling studies point to a striking structural difference between the stiff rodlike densely packed homopolymer 1 and the flexible copolymer 2 with spatially separated bisphosphonate units. Fluorescence titrations in buffered aqueous solution (pH = 7.0) confirm the superior affinity of the homopolymer toward oligoarginine peptides reaching nanomolar K_D values for the Tat peptide. Basic proteins are bound almost equally well by 1 and 2 with micromolar affinities, with the latter producing much more soluble complexes. The Arg selectivity of the monomer is transferred to the polymer, which binds Arg-rich proteins 1 order of magnitude tighter than lysine-rich pendants of comparable pl, size, and (Arg/Lys vs Glu/Asp) ratio. Noncovalent deposition of both polymers on glass substrates via polyethyleneimine layers results in new materials suitable for peptide and protein immobilization. RlfS measurements allow calculation of association constants K_a as well as dissociation kinetics k_D. They generally confirm the trends already found in free solution. Close inspection of electrostatic potential surfaces suggest that basic domains favor protein binding on the flat surface. The high specificity of the bisphosphonate polymers toward basic proteins is demonstrated by comparison with polyvinyl sulfate, which has almost no effect in RlfS experiments. Thus, copolymerization of few different comonomer units without cross-linking enables surface recognition of basic proteins in free solution as well as their effective immobilization on surfaces.

Full text of DOI:10.1021/ja0560229

Published on Web 12/20/2005
Arginine- and Lysine-Specific Polymers for Protein
Recognition and Immobilization
†
‡
,†
Christian Renner, Jacob Piehler, and Thomas Schrader*
Contribution from the Philipps-UniVersit a¨ t Marburg, Fachbereich Chemie,
Hans-Meerwein-Strasse, 35032 Marburg, Germany, and Johann Wolfgang Goethe-UniVersit a¨ t
Frankfurt, Institut f u¨ r Biochemie, Marie Curie Strasse 9, 60439 Frankfurt, Germany
Received September 15, 2005; E-mail: schradet@staff.uni-marburg.de
Abstract: Free radical polymerization of methacrylamide-based bisphosphonates turns weak arginine
binders into powerful polymeric protein receptors. Dansyl-labeled homo- and copolymers with excellent
water solubility are accessible through a simple copolymerization protocol. Modeling studies point to a
striking structural difference between the stiff rodlike densely packed homopolymer 1 and the flexible
copolymer 2 with spatially separated bisphosphonate units. Fluorescence titrations in buffered aqueous
solution (pH ) 7.0) confirm the superior affinity of the homopolymer toward oligoarginine peptides reaching
D
nanomolar K values for the Tat peptide. Basic proteins are bound almost equally well by 1 and 2 with
micromolar affinities, with the latter producing much more soluble complexes. The Arg selectivity of the
monomer is transferred to the polymer, which binds Arg-rich proteins 1 order of magnitude tighter than
lysine-rich pendants of comparable pI, size, and (Arg/Lys vs Glu/Asp) ratio. Noncovalent deposition of
both polymers on glass substrates via polyethyleneimine layers results in new materials suitable for peptide
and protein immobilization. RIfS measurements allow calculation of association constants K
a
as well as
dissociation kinetics k . They generally confirm the trends already found in free solution. Close inspection
D
of electrostatic potential surfaces suggest that basic domains favor protein binding on the flat surface. The
high specificity of the bisphosphonate polymers toward basic proteins is demonstrated by comparison with
polyvinyl sulfate, which has almost no effect in RIfS experiments. Thus, copolymerization of few different
comonomer units without cross-linking enables surface recognition of basic proteins in free solution as
well as their effective immobilization on surfaces.
Introduction
also independently (many enzyme-inhibitor or antibody-
protein complexes), charged residues play a major role, since
Protein-protein interactions play a key role in numerous
they are engaged in dynamic reversible interactions. On the other
hand, permanently homodimeric proteins, which are only
functional in the oligomeric state (homodimer cytochrome c),
recombine via highly hydrophobic interfaces, which are also
biological processes, characterized, inter alia, by hormone-
1
receptor, protease-inhibitor, or antibody-antigen complexes.
Many proteins exist in dimeric or oligomeric states, with
extended surface contact areas. The morphology and composi-
tion of such interfaces have been intensely studied in recent
years, and systematic surveys in structural databases have
5
more closely packed. A database of >2000 Ala mutants has
been compiled; its systematic analysis revealed hot spots on
protein interfaces, enriched in aromatic amino acids and in
arginine. These are often surrounded by energetically less
important residues, that most likely serve to occlude bulk solvent
from the hot spot to lower the local dielectric constant.6
Recently, Sheinerman et al. pointed out that electrostatic aspects
in protein-protein interactions do not only largely determine
the specificity of association but also contribute substantially
to complex stabilization. Structural and mutational analyses
suggest that clusters of charged and polar residues compensate
desolvation penalties by forming networks of ion pairs and
hydrogen bonds. Importantly, these attractive long-range inter-
2
disclosed some general trends and basic rules. Earlier reports
emphasize the factors hydrophobicity and complementarity;
large hydrophobic patches on one protein’s surface, surrounded
3
by charged residues, are matched by its complexation partner.
Argos and many others found that the amino acid composition
in protein interfaces does not resemble the nonpolar protein
interior but is rather intermediate between interior and exterior.
Notably, an unusually high number of arginines has repeatedly
4
been reported in protein-protein complexes. Especially, if these
aggregates are nonobligatory and involve proteins that must exist
7
actions drive protein association kinetically. As a relevant
†
Philipps-Universit a¨ t Marburg.
Johann Wolfgang Goethe-Universit a¨ t Frankfurt.
‡
illustration it was now found that the transcription factor P53,
(
(
(
(
1) Review: Stites, W. E. Chem. ReV. 1997, 97, 1233-1250.
2) Larsen, T. A.; Olsen, A. J.; Goodsell, D. S. Structure 1998, 6, 421-427.
3) Chothia, C.; Janin, J. Nature 1975, 256, 705-708.
(5) Jones, S.; Thornton, J. M. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 13-20.
(6) Bogan, A. A.; Thorn, K. S. J. Mol. Biol. 1998, 280, 1-9.
(7) Sheinerman, F. B.; Norel, R.; Honig, B. Curr. Opin. Struct. Biol. 2000,
10, 153-159.
4) Argos, P. Protein Eng. 1988, 2, 101-113. Janin, J.; Miller, S.; Chothia, C.
J. Mol. Biol. 1988, 204, 155-164.
620
9
J. AM. CHEM. SOC. 2006, 128, 620-628
10.1021/ja0560229 CCC: $33.50 © 2006 American Chemical Society

Arginine- and Lysine-Specific Polymers
A R T I C L E S
an important tumor suppressor, binds cooperatively to DNA only
as a dimer, stabilized by two intermolecular arginine-glutamate
salt bridges.8
High affinity molecular recognition of protein surfaces
combined with high specificity also remains a premier challenge
for artificial receptor systems, especially in light of their solvent-
exposed nature. Pioneering work by Hamilton et al. focused on
the development of aspartate-rich cyclopeptides on calixarene
scaffolds or glutamate-rich peptides on porphyrins for cationic
9
protein surface recognition. Often multiple copies of single
weak binding motifs were used to increase affinities toward
proteins: thus, a tetraguanidinium ligand was indroduced as a
helical protein surface binder for the tetramerization domain of
the above-mentioned protein P53.10 Similarly, linear anionic
1
1
oligomers were reported to adopt heparin-like properties or
efficiently inhibit human leukocyte elastase (Ki values of e 0.2
µM).12 Merritt et al. demonstrated how a high-affinity inhibitor
for cholera toxin evolves, if five copies of an R-D-galactoside
1
3
(
MNPG) are coupled to a pentacyclen core unit (IC50 ∼1 µM).
Specific recognition of phosphorylated peptides and proteins
was achieved by a fluorescent chemosensor carrying two Zn(II)-
dipicolylamine units.14 Strong and selective binding of carbonic
anhydrase was also achieved with multivalent transition metal
1
5
complexes, matching the protein’s histidine surface pattern.
By contrast, anionically functionalized amphiphilic nanoparticles
i.e., monolayer-protected gold clusters - MMPCs) use nonspe-
cific interactions to efficiently inhibit chymotrypsin through
(
1
6
electrostatic binding followed by protein denaturation. Simi-
larly, Kiessling et al. developed postsynthetically modified
(
PSM) polymers in the form of multivalent mannose displays
17
which nonspecifically inhibited hemagglutination. In a mo-
lecular imprinting approach on the protein surface, the shape
of lanthanide ion-carrying liposomes is reconstructed and used
for protein sensing. Finally, protein-protein interactions may
be probed with designed protein binders (generated, e.g., by
combinatorial library screening of “monobodies”)19 or disrupted
with synthetic â-turn mimetics (e.g., the interaction of the nerve
Figure 1. (a) Left: structural key element - arginine residue embraced by
bisphosphonate dianion. Right: schematic depicting the multiplication of
bisphosphonate units for oligoarginine recogntion. (b) Synthesis of monomer
building blocks as well as schematic structure of homo- and copolymer 1a
and 2a.
18
growth factor with its transmembrane tyrosine kinase receptor
TrkA); however, this area is still in its infancy.
(
8) Dehner, A.; Klein, C.; Hansen, S.; M u¨ ller, L.; Buchner, J.; Schwaiger, M.;
Kessler, H. Angew. Chem. 2005, 117, 5381-5386.
20
(
9) (a) Lin, Q.; Park, H. S.; Hamuro, Y.; Lee, C. S.; Hamilton, A. D.
Biopolymers 1998, 47, 285-297. (b) Park, H. S.; Lin, Q.; Hamilton, A. D.
J. Am. Chem. Soc. 1999, 121, 8-13. (c) Jain, R. K.; Hamilton, A. D. Org.
Lett. 2000, 2, 1721-1723. (d) Baldini, L.; Wilson, A. J.; Hong, J.; Hamilton,
A. D. J. Am. Chem. Soc. 2004, 126, 5656-5657.
Concept
Some years ago, we discovered that small bisphosphonate
-1
dianions bind to arginine (Ka ) 86 000 M in d6-DMSO) and
(
10) Salvatella, X.; Martinell, M.; Gair ´ı , M.; Mateu, M. G.; Feliz, M.; Hamilton,
-
1
A. D.; de Medoza, J.; Giralt, E. Angew. Chem., Int. Ed. 2004, 43, 196-
lysine residues (Ka ) 4000 M in d6-DMSO) in a peptidic
environment with remarkable affinity, while almost other amino
acid side chains are completely rejected (Figure 1a).21 However,
1
98.
(
11) Benezra, M.; Vlodavsky, I.; Yayon, A.; Bar-Shavit, R.; Regan, J.; Chang,
M.; Ben-Sasson, S. Cancer Res. 1992, 52, 5656-5662.
12) Regan, J.; McGarry, D.; Bruno, J.; Green, D.; Newman, J.; Hsu, C.-Y.;
Kline, J.; Barton, J.; Travis, J.; Choi, Y. M.; Volz, F.; Pauls, H.; Harrison,
R.; Zilberstein, A.; Ben-Sasson, A. A.; Chang, M. J. Med. Chem. 1997,
(
transition to water causes a drastic drop in free binding energy
-1
(
Ka e 100 M ), because the recognition process relies mainly
4
0, 3408-3422.
on electrostatic attraction, enforced by π-cation attraction. The
original high affinity for basic amino acids could now be
restored and markedly enhanced by applying the concept of
multivalency often found in natural surface recognition processes:
(
13) Merritt, E. A.; Zhang, Z.; Pickens, J. C.; Ahn, M.; Hol, W. G. J.; Fan, E.
J. Am. Chem. Soc. 2002, 124, 8818-8824.
14) Ojida, A.; Mito-oka, Y.; Inoue, M.; Hamachi, I. J. Am. Chem. Soc. 2002,
(
1
24, 6256-6258. Ojida, A.; Inoue, M.; Mito-oka, Y.; Hamachi, I. J. Am.
Chem. Soc. 2003, 125, 10184-10185. Ojida, A.; Kohira, T.; Hamachi, I.
Chem. Lett. 2004, 33, 1024-1025. Review: Ojida, A.; Miyahara, Y.; Kohira,
T.; Hamachi, I. Biopolymers 2004, 76, 177-184.
22
the weak arginine binder was therefore polymerized and thus
(
15) Fazal, Md. A.; Roy, B. C.; Sun, S.; Mallik, S.; Rodgers, K. R. J. Am.
Chem. Soc. 2001, 123, 6283-6290.
16) Fischer, N. O.; McIntosh, C. M.; Simard, J. M.; Rotello, V. M. Proc. Natl.
Acad. Sci. U.S.A. 2002, 99, 5018-5023. Fischer, N. O.; Verma, A.;
Goodman, C. M.; Simard, J. M.; Rotello, V. M. J. Am. Chem. Soc. 2003,
transformed into an efficient receptor site for basic proteins with
KD values in buffered aqueous solution reaching the submicro-
molar regime. To this end, we performed a simple free radical
(
1
25, 13387-13391.
(
(
17) Strong, L. E.; Kiessling, L. L. J. Am. Chem. Soc. 1999, 121, 6193-6196.
18) Santos, M.; Roy, B. C.; Goicoechea, H.; Campiglia, A. D.; Mallik, S. J. J.
Am. Chem. Soc. 2004, 126, 10738-10745.
(20) Burgess, K. Acc. Chem. Res. 2001, 34, 826-835.
(21) Rensing, S.; Springer, A.; Grawe, T.; Schrader, T. J. Org. Chem. 2001,
66, 5814-5821.
(22) Review: Mammen, M.; Choi, S. K.; Whitesides, G. M. Angew. Chem., Int.
Ed. 1998, 37, 2755-2794.
(
19) Koide, A.; Abbateillo, S.; Rothgery, L.; Koide, S. Proc. Natl. Acad. Sci.
U.S.A. 2002, 99, 1253-1258.
J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006 621

A R T I C L E S
Renner et al.
Figure 2. Determination of copolymerization parameters: (a) Decrease of comonomer concentrations; (b) corresponding copolymerization diagram; (c)
Fineman-Ross diagram.
polymerization of the weakly binding monomeric unit, prefer-
ably in its anionic state. We describe in this manuscript its facile
synthesis and the remarkable binding properties of this new class
of bisphosphonate polymers toward basic peptides and proteins,
in free solution and on surfaces. It appears that Nature itself
uses the closely related phosphate-guanidinium interaction
extensively for molecular recognition events between DNA/
RNA and regulatory proteins; the chelate type arrangement
bisphosphonate26 and methacryloyl chloride (Figure 1b). This
could be homopolymerized in DMF or copolymerized in the
same solvent with an amino alcohol-based methacrylamide,27
in order to increase its water solubility. In a final polymer-
analogous reaction, all phosphonic acid methyl esters were
quantitatively cleaved by direct nucleophilic attack in a dipolar
2
8
aprotic solvent. Prior to dealkylation, the polymers were well
soluble in chloroform (>100 mM); afterward, they switched to
perfect water solubility (>100 mM). Alternatively, the bispho-
sphonate unit itself was mildly dealkylated with LiBr and
subsequently copolymerized with a water-soluble initiator in
aqueous solution. The neutral as well as the corresponding
anionic state of the polymers could be well characterized by
-
+
-
Phosphate ...Arginine ...Phosphate has even been coined the
arginine fork” and is largely responsible for the sequence-
selective Tat/TAR recognition at an early critical stage of the
“
2
3
HIV life-cycle. A recent investigation about receptor heter-
omerizations between adenosine A2A/dopamine D2 or glutamate
NMDA/dopamine D2 revealed the paramount importance of
electrostatic interactions between Arg-rich epitopes and phos-
1
31
H and P NMR spectroscopy in CDCl3 and D2O. Experimental
evidence was thus produced for the complete polymerization
of all monomers, and also for the quantitative monodealkylation
of all dimethyl phosphonate ester groups. For an even distribu-
tion of binding motifs along the polymer chain, it was very
important to guarantee a statistical copolymerization in all cases;
to this end, we determined the respective copolymerization
parameters in DMF and water (Figure 2). Monomer concentra-
tions were calculated from their respective NMR integrations,
calibrated with pyridine as an internal standard. Both, the
24
phorylated serines, reaching “amazing covalent-like” stabilities.
With our strategy of developping highly efficient lysine and
2
5
arginine binders we are currently pursuing a 2-fold goal, i.e.,
targeting arginine- and lysine-rich protein domains as well as
immobilization of Arg-tagged proteins. On one hand this would
ultimately allow specific interference with gene regulation
processes such as histone/DNA packing or RNA complexation
by regulatory proteins. Numerous proteins (inter alia, cyto-
chrome c, chymotrypsin, lysozyme) contain active sites sur-
rounded by basic amino acids which might be reversibly blocked
by receptor caps. On the other hand, any Arg-tagged protein
might be immobilized onto surfaces in a functional state by
means of these Arg-tag recognition units.
2
9
integrated form of the copolymerization equation as well the
3
0
Fineman-Ross evaluation method gave comparable results.
With r1 and r2 values between 0.3 and 2.0, an almost ideal
statistical copolymerization occurs in all cases, irrespective of
the solvent. Interestingly, in water, the smaller r2 values
document a certain reluctance of the dianionic bisphosphonate
Synthesis
(
26) Rensing, S.; Schrader, T. Org. Lett. 2002, 4, 2161-2164.
27) Gros, L.; Ringsdorf, H.; Schupp, H. Angew. Chem., Int. Ed. Engl. 1981,
20, 305-325.
The monomeric arginine binding building block was obtained
by amide coupling between a m-amino-substituted xylylene
(
(
28) Karaman, R.; Goldblum, A.; Breuer, E.; Leader, H. J. J. Chem. Soc., Perkin
Trans. 1 1989, 765-774. Krawczyk, H. Synth. Commun. 1997, 27, 3151-
3161.
(
23) Calnan, B. J.; Tidor, B.; Biancalana, S.; Hudson, D.; Frankel, A. D. Science
1
991, 252, 1167-1171.
(
(
24) Woods, A. S.; Ferre, S. J. Proteome Res. 2005, 4, 1397-1402.
25) Arginine hosts: Schrader, T. Chem.sEur. J. 1997, 3, 1537-1541. Lysine
hosts: Fokkens, M.; Schrader, T.; Kl a¨ rner, F.-G. J. Am. Chem. Soc. 2005,
(29) Aguilar, M. R.; Gallardo, A.; del Mar Fern a´ ndez, M.; San Rom a´ n, J.
Macromolecules 2002, 35, 2036-2041. Taden, A.; Tait, A. H.; Kraft, A.
J. Polym. Sci., Part A: Polym. Chem. 2002, 40, 4333-4343.
1
27, 14415-14421.
(30) Fineman, M.; Ross, S. D. J. Polym. Sci. 1950, 5, 259-262.
622 J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006

Arginine- and Lysine-Specific Polymers
A R T I C L E S
Figure 4. Top: side view from molecular mechanics calculations of homo-
and copolymer 1 and 2 in water (3000 steps). Bottom: respective MD
calculations (300 K, water, 10 ps).
approximately 100 bisphosphonate units, whereas the copoly-
mers with twice the molecular weight contain only ∼95
bisphosphonates. The first generation of polymers was used in
UV/vis titrations with proteins carrying a natural chromophore,
such as cytochrome c. To be able to generalize these experi-
ments, several basic proteins were subsequently fluorescence-
labeled selectively at their N-terminus with OregonGreen.31
However, in all cases, the effects remained very small and
unreliable, probably due to the fact that the label was too far
away from the binding region of the polymer. The statistical
distribution of 10% of dansyl monomers within the polymer
chain of the second receptor generation places them in the
immediate vicinity of the complexation region; thus, strong
effects can be expected on binding to a peptide or protein.
Indeed, approach of such a polycationic guest usually leads to
a marked increase in emission intensity of the polymer. As an
exception, the interaction with iron-containing proteins such as
cyt c and ferritin effects a complete fluorescence emission
3
2
quenching, as illustrated in Figure 3b. These binding events
could be quantitatively monitored by fluorescence titrations and
were evaluated by nonlinear regression (Figure 3c).33 All the
experiments described in this paper were conducted in 30 mM
NaH2PO4 buffer (pH 7.0).
Polymer Structures
Figure 3. (a) Molecular weight and polydispersity determination for
homopolymer 1 by GPC; (b) fluorescence spectra from titration of 1a (1
µM) with cytochrome c; (c) corresponding titration curve (inverted intensity
decrease).
Hexadecamer fragments of homo- and copolymer were
subjected to molecular mechanics calculations, followed by MD
simulations (MacroModel7.0, Amber*, water, 3000 steps;
MD: 300 K, 10 ps). The resulting structures and their
conformational flexibility, depicted in Figure 4, are strikingly
different: Homopolymer 1 is characterized by a remarkable
steric congestion with simultaneous electrostatic phosphonate
anion repulsion, resulting in a stiff rodlike overall shape, with
multiple interior amide groups surrounding the backbone and a
high exterior negative charge density. In addition to this
structural reminiscence of DNA, it also reaches a similar
unit to homopolymerize, probably due to electrostatic repulsion.
In DMF, r2 becomes 1 and restores perfect statistical behavior.
For the second generation of polymers a fluorescence label
was incorporated by addition of 10% of a methacrylamide-based
dansyl monomer to the copolymerization mixture to ensure a
general sensitive spectroscopic characterization of the complex-
ation event on protein surfaces (Figure 1b). All polymers were
characterized by GPC, either in DMF (with LiCl additive) or
in water. Molecular weights between 20 000 and 80 000 were
detected, accompanied with the expected monomodal molecular
weight distributions according to Schultz-Flory: MW/Mn
quotients of >2 were observed in all cases, characteristic of
the conventional free radical polymerization mechanism (Figure
(
31) Molecular Probes, http://www.molecularprobes.com/media/pis/mp06153.pdf,
2001.
(
32) Petrat, F.; de Groot, H.; Sustmann, R.; Rauen, U. Biol. Chem. 2002, 383,
4
89-502. Fluorescence quenching by redoxactive iron as a sensitive tool
for the determination of chelatable iron: Petrat, F.; Weisheit, D.; Lensen,
M.; de Groot, H.; Sustmann, R.; Rauen, U. Biochem. J. 2002, 362, 137-
3a). However, the molecular weight of the copolymers was
147.
always a little higher than that of the hompopolymers, a fact
that has to be kept in mind for direct comparisons. Thus, the
homopolymers with a molecular weight of ∼40 000 contain
(33) (a) Schneider, H. J.; Kramer, R.; Simova, S.; Schneider, U. J. Am. Chem.
Soc. 1988, 110, 6442-6448. (b) Wilcox, C. S. In Frontiers in Supramo-
lecular Chemistry; Schneider, H. J., Ed.; Verlag Chemie: Weinheim, 1991;
p 123.
J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006 623

A R T I C L E S
Renner et al.
Table 1. Overview of All Binding Experiments Involving Homo-
and Copolymers with Peptides and Proteins in Free Solution,
a
According to Fluorescence Data
fluorescence titrationb [M-1]
homopolymer
copolymer
MW
c
c
peptide
pItheor
pIexptl
[kDa]
BP:guest
K
a
BP:guest
K
a
Arg
12
12
12
12
12
10
12
12
12
12
12
10
0.2
0.3
0.5
0.6
1.0
1.5
1:1
2:1
4:1
8:1
6e+1
9e + 3
4e + 4
6e + 5
Arg2
Arg3
Arg4
Arg6
2:1
2:1
4:1
4e + 3
1e + 4
1e + 6
5e + 8
d
insoluble >e + 7
e
Tat
∼e + 9
fluorescence titrationb [M-1]
homopolymer copolymer
MW
c
c
protein
pItheor
pIexptl
[kDa]
stoich
K
a
stoich
K
a
Hist H1
Cyt c
lysozyme 9.1 11.0 14.3
ADH
Trypsin
Prot K
BSA
10.4
9.2
22.0
12.3
>e + 6
6e + 5
>e + 5
>e + 5
2e + 5
4e + 6
Figure 5. (a) Schematic depicting the postulated binding mode between
2:1
3:1
4:1
1a and cyt c; (b) Protein topology as determination factor for surface binding
events - BSA’s concave lysine domain is accessible for 1a and 2a only in
8.0 5.4 141
8.3 10.5 22.0
7.7
5.8 5.8 66.0
5.4 6.0 455.3
no shift
no saturation
free solution.36
>e + 5
>e + 4
2e + 5
1e + 6
4:1
4:1
3:2
1:9
4e + 4
2e + 4
2e + 5
1e + 6
28.9
was used as the host, the solubility problem became much more
obvious, because even for the naked eye a considerable number
of protein complexes precipitated from solution (c ≈ µM).
Nevertheless, strong binding could be inferred from the remain-
5:2
1:9
ferritin
a
Dansyl-labeled homo- and copolymers (0.1-1 µM) were titrated with
arg1-arg6 peptides and proteins of varying pI (5-10) and size (8-450 kDa).
5
b
c
ing soluble cases, whose Ka values were in the range of 10
M
Each measurement in 30 mM NaH2PO4. Averaged virtual 1:1 binding
-
1
6
-1
constant, calculated for each complexation step, assuming no cooperativity
between peptide/protein and polymer. In 150 mM NaCl. Alexa 488-
labeled YRRKKQRRRC.
to 10 M , with stoichiometries between 3:2 and 4:1
d
e
indicating the efficient complexation of multiple protein mol-
ecules by one single polymer strand (Table 1, Figure 5a). Two
proteins deserve a special comment: ADH failed to give any
change in fluorescence emission, although its pI of 8.4 is clearly
basic; on the contrary, Ferritin, the least basic protein with a pI
of 5.5, binds very tightly to the polymer with a capacity to hold
diameter of ∼20 Å. By contrast, copolymer 2 with a statistical
number of 2-4 amido alcohol comonomers between each
bisphosphonate pair remains highly flexible and has spatially
well separated receptor units, ideally suited for induced fit
processes on a protein surface.
9
copies at a time. The latter case is easily explained: in order
2
+
to draw hundreds of Fe ions inside the giant molecular
capsule, its interior is covered with numerous aspartate and
glutamate carboxylates, leaving an excess of basic amino acid
Peptide Recognition in Solution. Association constants were
initially calculated for a whole series of oligoarginines, after
establishing their individual stoichiometries on complex forma-
tion with the homo- and copolymer.34 Surprisingly, complex
formation with the homopolymer is highly economic, and on
average each bisphosphonate unit binds exactly one arginine
residue (Job plots, Table 1). A strong synergistic increase in
binding affinity is observed, reaching nanomolar KD values for
the Tat peptide (seven basic residues).35 As a certain drawback
the solubility of these extremely strong complexes drops below
the micromolar regime and, hence, hampers the Ka determina-
tion. In this respect the copolymers proved to be much easier
to handle and produced completely soluble peptide complexes
in all cases. However, the respective binding free energies
remained about one kcal/mol (∼1 order of magnitude in Ka)
behind those of their homopolymer counterparts. Again, the
stoichiometric ratios reflect the statistical distance between two
bisphosphonate receptor units, separated on average by 3
amidopropanol comonomer units.
3
7
residues on the outside. Transition to the copolymer restores
the solubility of most complexes, and this time it is accompanied
only with a marginal decrease in Ka, corresponding to much
less than 1 order of magnitude (Table 1). It should be
emphasized that the molecular weights of most examined
proteins decrease in reciprocal relation to their basicity increase.
Arginine Selectivity. To verify the assumed arginine prefer-
ence of our polymeric hosts, the total number of solvent-exposed
lysine vs arginine residues in each examined protein was drawn
from the crystal structures and visualized in a synoptic view as
3
8
gray and dark blue columns (Figure 6a). Three pairs can be
identified: Cyt c and trypsin contain almost exclusively lysine,
whereas lysozyme and proteinase K carry predominantly
arginine residues. HSA and ADH are mixed in their Arg/Lys
composition. To take into consideration the counterbalancing
acidic residues, another diagram was created which shows the
total number of aspartates and glutamates on each protein
surface, accompanied by a second column with the total number
Protein Recognition in Solution. Encouraged by these
preliminary experiments we then focused our attention on
protein surfaces. Proteins with pI values from slightly acidic
3
8
of basic residues (Figure 6b). Proteins, which differ only in
(
∼5) to strongly basic (∼10) and a broad spectrum of biological
(
36) Instead of BSA’s crystal structure, which has not yet been solved, the known
structure of the related HSA is shown (PDB-code 1BJ5).
37) Reviews: Carrondo, M. A. EMBO J. 2003, 22, 1959-1968. Theil, E. C.;
Ferritin. In Handbook of Metalloproteins; Messerschmidt, A., Poulos, H.
R., Weighardt, K., Eds.; Wiley: Chichester, U.K., 2001; Vol. 2, pp 771-
781.
functions were chosen for comparison. When the homopolymer
(
(
34) (a) Job, P. Compt. Rend. 1925, 180, 928-930. (b) Blanda, M. T.; Horner,
J. H.; Newcomb, M. J. Org. Chem. 1989, 54, 4626-4636.
(35) Cooperation with G o¨ bel, M. Frankfurt University. The K
D
value was
(38) PROVE program (PROtein Volume Evaluation): Pontius, J.; Richelle, J.;
determined by Fluorescence Correlation Spectroscopy (FCS).
Wodak, S. J. J. Mol. Biol. 1996, 264, 121-136.
6
24 J. AM. CHEM. SOC. VOL. 128, NO. 2, 2006
9

Arginine- and Lysine-Specific Polymers
A R T I C L E S
Figure 6. (a) Total number n of Arg and Lys residues displayed on the examined protein surfaces; (b) counterbalance of acidic residues on the same
proteins (total numbers n; net charges ≈ experimental pI values).
one single parameter, can be compared with respect to their
binding strength. Cyt c and lysozyme, e.g., have almost the same
molecular weight and a similar ratio of acidic and basic residues
on their surfaces. However, lysozyme, which is rich in arginines
polymer with (Arg)n and protein solutions produced, in most
cases, a marked increase in optical thickness, proportional to
the offered concentration. For most of the proteins investigated,
the major part of the binding amplitude was caused by reversible
interaction, and only a minor part of the protein was stably
attached to the surface or dissociated only slowly. From the
reversible binding signal titration curves could be derived whose
nonlinear regression analysis furnished association constants for
the surface binding process (Figure 7). These are summarized
in Table 2; in general, they remained ∼0.5 order of magnitude
behind those measured in free solution. This makes sense,
because a considerable number of negative charges have already
been consumed for the electrostatic attraction of the host to the
PEI surface. For Arg-tags, a similar picture evolves as already
found in free solution: a steady increase in affinity was observed
for higher oligomers, with the homopolymer clearly showing a
performance superior to that of the related copolymer.
(Arg: 61% of all basic residues), binds to the polybisphospho-
nate ∼1 order of magnitude more strongly than Cyt c (89%
Lys; Table 1, KCopol). Similarly, proteinase K (60% Arg) and
trypsin (88% Lys) have a comparable size, but trypsin is
definitely more basic than proteinase K. Nevertheless, both bind
to the polymeric tweezer equally well. In both cases, the
arginine-rich protein is superior to the lysine-rich competitor,
establishing a distinct arginine selectivity of the polymeric
receptor. This important feature had hitherto only been proven
for the monomer in organic solution (N/C-protected Arg vs
Lys: Ka ∼14:1).
41
Charge Density. Another important factor governing the
efficiency of protein binding is the sheer size of the target
molecule. This is really an entropy argument: two proteins with
the same absolute number of basic residues and a comparable
pI may differ in their relative sizes. Hence, one protein will
offer a much higher surface charge density which makes it
statistically easier for each bisphosphonate moiety to find an
Arg or Lys counterpart on the protein surface. In our series,
Cyt c and trypsin both carry 14 or 16 lysine residues,
respectively, on their surfaces and have a resembling surface
charge ratio. However, Cyt c is much smaller than trypsin (12
vs 22 kD) and is hence bound much more tightly by the
polybisphosphonate (Table 1, KCopol). This entropy effect is also
reflected in the stoichiometry factors. Transition to larger
proteins such as albumine (66 kD) and ferritin (455 kD)
illustrates this effect even more impressively: although both
proteins are slightly acidic (pI ) 5.5-6.0), they are bound
extremely well in solution, probably due to the high number of
accessible basic residues.
Protein Recognition on Surfaces
The same trend is also true for protein binding to the polymer-
coated surface. In principle, the amount of protein deposition
followed the pI scale, with highly basic histone and lysozyme
taking the lead, followed by moderately basic cyt c and trypsin
(a survey is shown in Figure 8a). Several noteworthy exceptions
remain:
Topology. BSA which was bound tightly in free solution did
not show any affinity toward the immobilized bisphosphonate
polymers, even with the copolymer carrying the hydrophobic
dansyl moieties. This is the first indication of a different binding
4
2
mode sensitive toward the protein topology: while in free
solution, the thin polymer thread can be smoothly wound around
the BSA molecule, even deep inside the cleft, where most of
the basic residues are concentrated, and this place becomes
inaccessible to the flat anionic surface; even worse, several
acidic domains are located on the edges of the heart-shaped
molecule, leading to significant electrostatic repulsion (Figure
Peptide Recognition on Surfaces
5
6
b). Similar reasons may account for the fact, that ADH (Figure
: 52+/44-) and proteinase K (16+/14-) completely refuse
Noncovalent immobilization of Arg-tags and proteins was
accomplished by way of irreversible electrostatic attraction of
a monolayer of the polymeric binders onto a SiO2 surface
densely coated with a polyethyleneimine (PEI) layer. The
consecutive deposition of carrier polymer (PEI) and host
polymer and the formation of stably adsorbed layers were
monitored in a time-resolved manner by RIfS (Reflectometric
Interference Spectroscopy) in a flow-through system (Figure
to interact with the anionic host in its immobilized state.
Crystallization. Ferritin, although in toto acidic, binds most
strongly to the bisphosphonate surface, in an almost entirely
(39) Brecht, A.; Gauglitz, G.; Polster, J. Biosens. Bioelectron. 1993, 8, 387-
392; Gauglitz, G.; Brecht, A.; Kraus, G.; Nahm, W. Sens. Actuators, B
1993, 11, 21-27.
(40) Piehler, J.; Brecht, A.; Gauglitz, G.; Zerlin, M.; Maul, C.; Thiericke, R.;
3
9
7
a). This technique detects interactions at surfaces as an
Grabley, S. Anal. Biochem. 1997, 249, 94-102.
4
0
(41) Connors, K. A. Binding constants; Wiley: New York, 1987.
increase in optical thickness of a thin interference layer.
(42) BSA is the only protein investigated in the above-discussed series which
Subsequent equilibration of the immobilized bisphosphonate
substantially deviates from a globular topology (compare Figure 7).
J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006 625

A R T I C L E S
Renner et al.
Figure 7. (a) Schematic showing the attachment of molecular PEI and PBP layers onto the signal transducer; (b) Corresponding stepswise increase in RIfS
signal. (c) RIfS signal upon injection of trypsin at different concentrations, binding on 1 with reversible and irreversible contributions; (d) corresponding
titration curve furnishing Ka.
Table 2. Overview of All Binding Experiments Involving Homo- and Copolymers with Peptides and Proteins on the PBP-Coated
a
PEI-Surface, According to RIfS Data
RIfS titration (PEI surface)b [M-1]
K
a
reversible
D
k irreversible
MW
peptide
pItheor
pIexptl
[kDa]
homopolymer
copolymer
homopolymer
copolymer
Arg
12
12
12
12
12
10
12
12
12
12
12
10
0.2
0.3
0.5
0.6
1.0
1.5
Arg2
Arg3
Arg4
Arg6
1e + 4
4e + 4
3e + 5
2e + 3
4e + 3
3e + 5
c
Tat
RIfS titration (PEI surface)b [M-1]
K
a
reversible
D
k irreversible
MW
protein
pItheor
pIexptl
[kDa]
homopolymer
copolymer
homopolymer
copolymer
histone H1
Cyt c
lysozyme
ADH
trypsin
Prot K
BSA
10.4
9.2
9.1
8.0
8.3
7.7
5.8
5.4
22.0
12.3
14.3
141
22.0
28.9
66.0
455.3
irreversible
2e + 5
strong
5e + 4
3e + 4
3e - 3
2e - 4
3e - 4
5e - 3
3e - 3
11.0
5.4
10.5
8e + 4
2e - 3
no effect
1e + 5
no effect
no effect
strong
no effect
6e - 3
4e + 4
2e - 1
irreversible
no effect
no effect
5.8
6.0
ferritin
crystallization
a
Unlabeled homo- and copolymers (∼2 µM) were immobilized on the cationic PEI polymer and titrated with arg1-arg6 peptides and proteins of varying
pI (5-10) and size (8-450 kDa). Each measurement in 30 mM NaH2PO4. c Alexa 488-labeled YRRKKQRRRC.
b
irreversible fashion. Due to its self-assembled dodecamer
structure, crystallization is obviously greatly facilitated on the
well-ordered two-dimensional arrangement on the PEI/PBP
layer.
Additional information about the binding mode was drawn
from an analysis of the interaction kinetics. In most cases,
diffusion controlled association kinetics was observed as
expected for electrostatic interaction. An exception was Ferritin,
which associated much slower, in good agreement with the more
complex layer formation on the PBP surface proposed for this
protein. Typically, multiphasic dissociation of the adsorbed
proteins was observed: the major part dissociated very fast,
followed by slower dissociation kinetics, which was character-
istic for the different proteins, and for homo- and copolymer.
To this purpose the dissociation rate constants were determined
by fitting the decay dissociation curve with a mono- or
biexponential decay. Indeed, the homopolymer shows again the
slowest dissociation processes or even binds completely ir-
reversibly (histone H1, Ferritin). To draw as much systematic
information about the immobilization process as possible from
the accumulated data, a graphical survey of all time-dependent
RIfS curves was produced, showing the association and dis-
sociation kinetics of each individual protein immobilization
event (Figure 8a). The same scale was used for the y-axis, indi-
626 J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006

Arginine- and Lysine-Specific Polymers
A R T I C L E S
Figure 8. (a) Protein recognition by homopolymer-coated surface 1 according to RIfS (top series): (histone, ferritin) completely irreversible binding;
cytochrome c, lysozyme, trypsin) partially irreversible binding; (proteinase K, HSA) almost no binding. (b) Bottom series: the Connolly surface of these
(
proteins, depicted in correct relative sizes, is patterned with the electrostatic surface potential (ESP), showing basic (blue) and acidic domains (red) on the
protein surfaces. Histone, cytochrome c, and lysozyme with pI > 9 are uniformly basic, and trypsine has a basic domain at one end and an opposing acidic
domain at the other; proteinase K, human serum albumine (HSA), and ferritin show a statistic distribution of acidic and basic amino acids.
cating absolute optical thicknesses which generally correlate
with immobilized protein material. Histone demonstrates superi-
or affinity (most likely >10 M ) toward the immobilized poly-
with Connolly surfaces,44 colored in blue and red according to
45
their electrostatic surface potential (ESP). Those protein faces
showing the highest density of basic residues are depicted in
Figure 8b. At first glance, it becomes obvious, that especially
lysozyme and, to a lesser extent, also histone H1 and cyt c are
almost evenly covered with basic residues, accompanied with
a high charge density because of the overall small surface area.
Trypsin, however, has a basic and an acidic face, divided by an
essentially neutral gap. In solution, this is a clear disadvantage
6
-1
bisphosphonate by a completely irreversible RIfS profile. Lyso-
5
-1
zyme, trypsin, and cyt c are also strongly bound (∼10 M )
but in an almost fully reversible manner, although with relatively
-
3
slow dissociation kinetics (kD ∼10 ). Ferritin not only binds
irreversibly but also shows again a peculiar stepwise binding
behavior: during the first 5 s, a protein monolayer seems to be
deposited very rapidly on the bisphosphonate tweezer polymer,
(
43) Granier, T.; Langlois d’Estaintot, B.; Gallois, B.; Chevalier, J.-M.;
4
3
followed by slow crystallization of multilayers on top.
Pr e´ cigoux, G.; Santambrogio, P.; Arosio, P. J. Biol. Inorg. Chem. 2003, 8,
1
05-111. Zeth, K.; Offermann, S.; Essen, L.-O.; Oesterhelt, D. Proc. Natl.
Basic Domains. A closer inspection of the arginine and lysine
distribution on the respective protein surfaces gives further
insight into the factors controlling the binding process; to this
end, crystal structures were drawn from the PDB and depicted
Acad. Sci. U.S.A. 2004, 101, 13780-13785.
(44) (a) Connolly, M. L. Science 1983, 221, 709-713. (b) Connolly, M. L. J.
Appl. Crystallogr. 1983, 16, 548-558.
(
45) Sanner, M. F.; Spehner, J. C.; Olson, A. J. Biopolymers 1996, 38, 305-
320.
J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006 627

A R T I C L E S
Renner et al.
5
-1
4
for effective polymer attraction (cyt c, 10 M ; trypsin, 10
the case with histone and lysozyme (Table 1). A copolymer,
however, can minimize repulsive interactions with acidic
residues and by induced fit select just positively charged amino
acids for ion pairing. BSA and Ferritin appear to be good
examples for the latter case. Finally, the combination of a
globular protein surface with the rough but essentially flat
surface of an immobilized polymer on a chip leads to an
enthalpic disadvantage, because the mutual contact areas are
minimized. On the other hand, proteins with large basic domains
find a preorganized even host arrangement, rendering docking
enthalpically and also entropically favorable. Even small
hydrophobic patches on the receiving surface will add to this
attraction (exclusion of solvent molecules, hydrophobic effect)
and most likely lead to irreversible complexation. In our
investigation, trypsin with its distinct basic domain proved to
be even superior to lysozyme with its uniformly high density
of surface arginines.
-1
M ), but on a two-dimensional surface, the one basic face
oriented toward the host suffices and eliminates any discrimina-
tion ability from related basic proteins such as cyt c (both 10
4
-1
M ). Proteinase K’s and ferritin’s EPS look pale, due to
multiple mutual neutralization of neighboring lysine and aspar-
tates; this may account for the extraordinarily low affinity of
the former toward the immobilized bisphosphonate polymer.
In the case of giant ferritin, hydrophobic interactions between
the large contact areas will contribute to strong binding, in
addition to formation of surrounding basic domains of at least
four lysine residues (Figure 8b, blue circle). These domains mark
a path for the anionic polymer chain along the lysines and
arginines across the large ferritin surface, avoiding almost any
contact with acidic residues; thus strong binding can also be
expected in solution.
Specificity
Conclusion and Outlook
An important issue with multiple interactions relying mainly
on Coulomb attraction is the proof of specificity. One could
argue that the monomeric tweezer has already been extensively
examined in this respect and evolved as an arginine- and lysine-
selective receptor. However, this might have been overridden
by the multiplication effect in the polymer. It was therefore an
important additional experimental evidence to find out that those
proteins, which bound strongly to the immobilized bisphospho-
nate polymer, exhibited only very weak effects with a polyvinyl
sulfate layer, which was attached to the surface in the same
manner. This was individually checked with Arg6, histone H1,
and Cyt c.
We conclude that polymeric bisphosphonate hosts are acces-
sible via highly economic synthetic pathways and that they
display high affinities toward Arg oligomers and most basic
proteins in buffered aqueous solution. The unusually effective
combination of homopolymers and Arg-tags suggests a promis-
ing application for the controlled assembly of functional Arg-
tagged proteins on surfaces. The factors governing protein
recognition by the new polybisphosphonate hosts do not follow
the simple pI scale but are much more delicate: among others,
the Arg/Lys ratio, the surface charge density, the occurrence of
basic domains on the outer surface, the overall protein topology,
and the sheer size of the protein guest as well as its related
crystallization tendency play a decisive role in determining its
affinity toward the artificial host. Basic as well as Arg-tagged
proteins can likewise be immobilized on cationic surfaces and
follow similar rules, complemented by retarded dissociation
kinetics for the irreversible contribution to their free binding
energy. In the future, we will investigate whether the tight
complex formation with our polymers even affects the activity
or function of the bound proteins. We also intend to combine
these binding monomers with others tailored for the remaining
classes of amino acid residues, to achieve a highly specific
recognition of protein surfaces, i.e., ultimately create artificial
antibodies. To this end, additional specific monomers with
hydrophilic, hydrophobic, and aromatic binding sites will be
incorporated.
In an exploratory experiment we also checked the intended
selective binding of Arg-tagged proteins to the polybisphos-
phonate surface: A protein of essentially neutral pI was chosen
(maltose binding protein, MBP), equipped with a doubly
orthogonal His4-Arg4-His4-tag at its N-terminus. It stimulated
an intense RIfS signal, with a fully reversible profile. On the
contrary, exchange of the Arg4 for the Ala4 sequence completely
eliminated the protein’s affinity to the anionic surface decorated
with bisphosphonate host polymers, confirming a highly specific
interaction of the protein with PBP through the Arg4 motif.
Thermodynamics
The multivalency of our new hosts leads to stepwise addition
of binding enthalpies and also includes an entropy gain, if the
subsequent binding events take advantage from a favorable
orientation through the previous step.22 More or less one-
dimensional recognition is realized, if (Arg)n oligomers are
bound by homopolymers in solution. Their superiority over
copolymers is easily explained by their perfect complementarity,
enabling both complexation partners to line up along one
another. In the RIfS measurements this effect is greatly
attenuated, because the two-dimensional chip surface per se
produces a substantial degree of preorganization for both
immobilized polymers. However, protein surface recognition
by electrostatic interactions imposes more difficult require-
ments: since arginines and lysines are scattered across the
surface, interspersed with acidic residues, a homopolymer will
win statistically because it may establish attractive contacts to
the maximum number of accessible residues, especially in the
Acknowledgment. We thank the group of Prof. M. G o¨ bel,
Frankfurt University, for the determination of KD values for the
complex between homo- and copolymer and the Tat peptide
by Fluorescence Correlation Spectroscopy (FCS). We are also
indebted to the Deutsche Forschungsgemeinschaft for funding
in the framework of a researcher group entitled: “Synthesis of
functional chemical-biological hybrid compounds”.
Supporting Information Available: Synthetic procedures,
polymerization procedures and characterization, determination
of copolymerization parameters, fluorescence titrations, RIfS
measurements. This material is available free of charge via the
Internet at http://pubs.acs.org.
1
case of small proteins with a high charge density. This is indeed
JA0560229
628 J. AM. CHEM. SOC.
9
VOL. 128, NO. 2, 2006