Using ion channel-forming peptides to quantify protein-ligand interactions

Article abstract of DOI:10.1021/ja077555f

This paper proposes a method for sensing affinity interactions by triggering disruption of self-assembly of ion channel-forming peptides in planar lipid bilayers. It shows that the binding of a derivative of alamethicin carrying a covalently attached sulfonamide ligand to carbonic anhydrase II (CA II) resulted in the inhibition of ion channel conductance through the bilayer. We propose that the binding of the bulky CA II protein (MW ≈ 30 kD) to the ion channel-forming peptides (MW ≈2.5 kD) either reduced the tendency of these peptides to self-assemble into a pore or extracted them from the bilayer altogether. In both outcomes, the interactions between the protein and the ligand lead to a disruption of self-assembled pores. Addition of a competitive inhibitor, 4-carboxybenzenesulfonamide, to the solution released CA II from the alamethicin-sulfonamide conjugate and restored the current flow across the bilayer by allowing reassembly of the ion channels in the bilayer. Time-averaged recordings of the current over discrete time intervals made it possible to quantify this monovalent ligand binding interaction. This method gave a dissociation constant of ～2 μM for the binding of CA II to alamethicin-sulfonamide in the bilayer recording chamber: this value is consistent with a value obtained independently with CA II and a related sulfonamide derivative by isothermal titration calorimetry.

Full text of DOI:10.1021/ja077555f

Published on Web 01/08/2008
Using Ion Channel-Forming Peptides to Quantify
Protein-Ligand Interactions
Michael Mayer,*,†,‡ Vincent Semetey,‡ Irina Gitlin,‡ Jerry Yang,‡ and
George M. Whitesides*,‡
Department of Chemistry and Chemical Biology, HarVard UniVersity,
Cambridge, Massachusetts 02138, and Department of Biomedical Engineering and
Department of Chemical Engineering, UniVersity of Michigan, Ann Arbor, Michigan 48109
Received October 1, 2007; E-mail: mimayer@umich.edu, gwhitesides@gmwgroup.harvard.edu
Abstract: This paper proposes a method for sensing affinity interactions by triggering disruption of self-
assembly of ion channel-forming peptides in planar lipid bilayers. It shows that the binding of a derivative
of alamethicin carrying a covalently attached sulfonamide ligand to carbonic anhydrase II (CA II) resulted
in the inhibition of ion channel conductance through the bilayer. We propose that the binding of the bulky
CA II protein (MW ≈ 30 kD) to the ion channel-forming peptides (MW ≈ 2.5 kD) either reduced the tendency
of these peptides to self-assemble into a pore or extracted them from the bilayer altogether. In both
outcomes, the interactions between the protein and the ligand lead to a disruption of self-assembled pores.
Addition of a competitive inhibitor, 4-carboxybenzenesulfonamide, to the solution released CA II from the
alamethicin-sulfonamide conjugate and restored the current flow across the bilayer by allowing reassembly
of the ion channels in the bilayer. Time-averaged recordings of the current over discrete time intervals
made it possible to quantify this monovalent ligand binding interaction. This method gave a dissociation
constant of ∼2 µM for the binding of CA II to alamethicin-sulfonamide in the bilayer recording chamber:
this value is consistent with a value obtained independently with CA II and a related sulfonamide derivative
by isothermal titration calorimetry.
Introduction
Ion channels are increasingly being investigated for sensing
applications,4-11 because binding of just one ligand molecule
This paper describes a method to quantify protein-ligand
interactions by measuring the flux of ions through pores of a
synthetically modified, ion channel-forming peptide in planar
lipid bilayers. The method is based on the reduction of ionic
conductivity through self-assembled pores of a derivative of
alamethicin, a naturally occurring ion channel-forming peptide,1-3
which carries a covalently attached benzenesulfonamide group
on its C-terminus (alamethicin sulfonamide, 8, see Scheme 1).
In this assay, reduction of the ion current is triggered by binding
of carbonic anhydrase II (CA II) to the sulfonamide moiety on
the channel-forming peptide. By time-averaging the recorded
current, we were able to compare the amount of transported
charge before and after addition of CA II. Increasing concentra-
tions of CA II reduced ion transport. This reduction made it
possible to estimate binding constants for the interaction between
CA II and the sulfonamide ligand (by plotting transported charge
through the pores as a function of the concentration of CA II).
Addition of a competitive inhibitor, 4-carboxybenzenesulfona-
mide (9), to the solution released CA II from the complex with
8 and restored the current flow through self-assembled ion
channels in the bilayer.
can induce a conformational change of the ion-channel protein
and result in the flux of thousands of ions. In this sense, ion
channels are amplifiers and can amplify signals by factors of
103-105.12 This high-gain amplification makes ion channels well
suited for signal transduction, which is, of course, one of their
critical functions in the membranes of excitable cells.13
On the basis of pioneering work on protein engineering by
the groups of Mutter,14-17 Montal,18,19 DeGrado,20,21 Vogel,4,22-25
(4) Mayer, M.; Terrettaz, S.; Giovangrandi, L.; Vogel, H. Functional Analysis
of Ion Channels: Planar Patch Clamp and Impedance Spectroscopy of
Tethered Lipid Membranes. In Biosensors: A Practical Approach, 2nd
ed.; Cooper, J. M., Cass, A. E. G., Eds.; Oxford University Press: Oxford,
2003; Vol. 1, pp 153-184.
(5) Blake, S.; Mayer, T.; Mayer, M.; Yang, J. ChemBioChem 2006, 7, 433-
435.
(6) Kiwada, T.; Sonomura, K.; Sugiura, Y.; Asami, K.; Futaki, S. J. Am. Chem.
Soc. 2006, 128, 6010-6011.
(7) Bayley, H.; Cremer, P. S. Nature 2001, 413, 226-230.
(8) Das, G.; Talukdar, P.; Matile, S. Science 2002, 298, 1600-1602.
(9) Litvinchuk, S.; Tanaka, H.; Miyatake, T.; Pasini, D.; Tanaka, T.; Bollot,
G.; Mareda, J.; Matile, S. Nat. Mater. 2007, 6, 576-580.
(10) Berkane, E.; Orlik, F.; Charbit, A.; Danelon, C.; Fournier, D.; Benz, R.;
Winterhalter, M. J. Nanobiotechnol. 2005, 3, 3.
(11) Danelon, C.; Lindemann, M.; Borin, C.; Fournier, D.; Winterhalter, M.
IEEE Trans. Nanobiosci. 2004, 3, 46-48.
(12) Aidley, D. J.; Stanfield, P. R. Ion Channels, 1st ed.; Cambridge University
Press: Cambridge, UK, 1996; pp 9-31.
(13) Hille, B. Ion Channels of Excitable Membranes, 3rd ed.; Sinauer Associ-
ates: Sunderland, 2001; pp 25-60.
† University of Michigan.
‡ Harvard University.
(14) Mutter, M.; Altmann, K.-H.; Tuchscherer, G.; Vuilleumier, S. Tetrahedron
Lett. 1988, 44, 771-785.
(1) Fox, R. O., Jr.; Richards, F. M. Nature 1982, 300, 325-330.
(2) Boheim, G.; Benz, R. Biochim. Biophys. Acta 1978, 507, 262-270.
(3) Sansom, M. S. P. Eur. Biophys. J. 1993, 22, 105-124.
(15) Tuchscherer, G.; Steiner, V.; Altmann, K. H.; Mutter, M. Methods Mol.
Biol. 1994, 36, 261-285.
9
10.1021/ja077555f CCC: $40.75 © 2008 American Chemical Society
J. AM. CHEM. SOC. 2008, 130, 1453-1465
1453

A R T I C L E S
Mayer et al.
Scheme 1. Synthesis of an Alamethicin Derivative Carrying a Benzenesulfonamide Groupa
a Reagents and Conditions: (a) (Boc-aminooxy)acetic acid, EDE, DIEA, DMF (b) Pd/C, H2, EtOH (c) 1, EDC, DIEA, DMF (d) TFA (e) Dess-Martin
oxidation, CH2Cl2 (f) 5, CH3OH.
and Bayley,26-30 ion-channel proteins and ion-channel-forming
peptides have received increasing attention for sensing applica-
tions since the 1990s.31 In 1993, Rokitskaya et al. explored
photodynamic inactivation of gramicidin pores in the presence
of phthalocyanines.32 One year later, Bayley’s group demon-
strated the use of R-hemolysin for detecting metal ions26 and
for sensing protease activity,33 whereas Bezrukov et al. used
alamethicin pores to count polymer particles.34 Woolley et al.
used a derivative of alamethicin to form an “ion-activated” ion
channel.35 Cornell et al. employed gramicidin derivatives for
detection of immune reactions;36 and Vogel’s group developed
a synthetic ion channel23 and detected binding of colicin N to
the porin OmpF.24 Since these initial studies, melittin,4,23,25,37,38
R-hemolysin,27,39,40 gramicidin A,5,41-47 and alamethicin6,34,48-56
have become the most extensively studied systems for ion
(16) Altmann, K. H.; Mutter, M. Int. J. Biochem. 1990, 22, 947-956.
(17) Altmann, E.; Altmann, K. H.; Nebel, K.; Mutter, M. Int. J. Pept. Protein
Res. 1988, 32, 344-351.
(18) Oiki, S.; Danho, W.; Montal, M. Proc. Natl. Acad. Sci. U.S.A. 1988, 85,
2393-2397.
(32) Rokitskaya, T. I.; Antonenko, Y. N.; Kotova, E. A. FEBS Lett. 1993, 329,
332-335.
(33) Walker, B.; Bayley, H. Protein Eng. 1994, 7, 91-97.
(34) Bezrukov, S. M.; Vodyanoy, I.; Parsegian, V. A. Nature 1994, 370, 279-
281.
(19) Oiki, S.; Danho, W.; Madison, V.; Montal, M. Proc. Natl. Acad. Sci. U.S.A.
1988, 85, 8703-8707.
(20) DeGrado, W. F.; Wasserman, Z. R.; Lear, J. D. Science 1989, 243, 622-
628.
(21) Akerfeldt, K. S.; Lear, J. D.; Wasserman, Z. R.; Chung, L. A.; DeGrado,
W. F. Acc. Chem. Res. 1993, 26, 191-197.
(35) Woolley, G. A.; Epand, R. M.; Kerr, I. D.; Sansom, M. S.; Wallace, B. A.
Biochemistry 1994, 33, 6850-6858.
(22) Heyse, S.; Stora, T.; Schmid, E.; Lakey, J. H.; Vogel, H. Biochim. Biophys.
Acta 1998, 1376, 319-338.
(36) Cornell, B. A.; Braach-Maksvytis, V. L.; King, L. G.; Osman, P. D.; Raguse,
B.; Wieczorek, L.; Pace, R. J. Nature 1997, 387, 580-583.
(37) Vogel, H.; Jahnig, F. Biophys. J. 1986, 50, 573-582.
(38) Terrettaz, S.; Ulrich, W. P.; Guerrini, R.; Verdini, A.; Vogel, H. Angew.
Chem., Int. Ed. 2001, 40, 1740-1743.
(23) Pawlak, M.; Meseth, U.; Dhanapal, B.; Mutter, M.; Vogel, H. Protein Sci.
1994, 3, 1788-1805.
(24) Stora, T.; Lakey, J. H.; Vogel, H. Angew. Chem., Int. Ed. 1999, 38, 389-
392.
(39) Braha, O.; Walker, B.; Cheley, S.; Kasianowicz, J. J.; Song, L.; Gouaux,
J. E.; Bayley, H. Chem. Biol. 1997, 4, 497-505.
(25) Terrettaz, S.; Mayer, M.; Vogel, H. Langmuir 2003, 19, 5567-5569.
(26) Walker, B.; Kasianowicz, J.; Krishnasastry, M.; Bayley, H. Protein Eng.
1994, 7, 655-662.
(40) Meller, A.; Nivon, L.; Brandin, E.; Golovchenko, J.; Branton, D. Proc.
Natl. Acad. Sci. U.S.A. 2000, 97, 1079-1084.
(27) Gu, L. Q.; Braha, O.; Conlan, S.; Cheley, S.; Bayley, H. Nature 1999,
398, 686-690.
(41) Wallace, B. A. Biophys. J. 1986, 49, 295-306.
(42) Borisenko, V.; Zhang, Z.; Woolley, G. A. Biochim. Biophys. Acta 2002,
1558, 26-33.
(28) Bayley, H.; Jayasinghe, L. Mol. Membr. Biol. 2004, 21, 209-220.
(29) Bayley, H. Curr. Opin. Biotechnol. 1999, 10, 94-103.
(30) Bayley, H. J. Cell. Biochem. 1994, 56, 177-182.
(31) Tien, H. T.; Salamon, Z.; Ottova, A. Crit. ReV. Biomed. Eng. 1991, 18,
323-340.
(43) Woolley, G. A.; Zunic, V.; Karanicolas, J.; Jaikaran, A. S.; Starostin, A.
V. Biophys. J. 1997, 73, 2465-2475.
(44) Futaki, S.; Zhang, Y.; Kiwada, T.; Nakase, I.; Yagami, T.; Oiki, S.; Sugiura,
Y. Bioorg. Med. Chem. 2004, 12, 1343-1350.
9
1454 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
channel-based sensor development. The reasons for the interest
in these four systems are that: (i) they incorporate spontaneously
into bilayers from solution (in contrast to most ion-channel
proteins, which must be reconstituted into the bilayer by
techniques such as proteoliposome fusion);57 and (ii) they are
available commercially.
these authors attached a biotin group to alamethicin and
demonstrated binding of streptavidin and anti-biotin antibodies.51
Here we explored alamethicin as the platform for a sensor
with three goals in mind: (i) to prepare derivatives of alame-
thicin by chemical derivatization of commercially available
alamethicin that would make this type of sensor broadly
available, (ii) to explore if sensors based on ion channel-forming
peptides could be used to quantify protein-ligand interactions
in solution, and (iii) to investigate a “typical”, biologically
relevant, monovalent interaction between a protein and a small
ligand. To accomplish these goals, we chose the well-character-
ized interaction between CA II (E.C. 4.2.1.1)64,65 and a
benzenesulfonamide moiety. This interaction has a typical
dissociation constant for biochemical interactions (∼1 µM).66,67
Alamethicin is an antimicrobial peptide composed of 19
amino acids and one amino alcohol with a molecular weight of
1.96 kD. It is secreted by the fungus Trichoderma Viride.68
Alamethicin adopts an amphipathic, R-helical structure in
biological membranes and forms ion channels by self-assembly
to oligomers. The most accepted model of pore formation by
alamethicin, the so-called barrel-stave model, suggests an
arrangement of transmembrane helices in a circle with a central,
water-filled pore.1,62 In this model, the hydrophilic face on the
R-helix of alamethicin is oriented toward the lumen of the pore,
whereas the hydrophobic face on the helix is in contact with
the surrounding lipid molecules.69
Alamethicin monomers in aqueous solution bind to (or
dissolve in) lipid membranes with partition coefficients of ∼10-3
M. This value results in an equilibrium distribution of alame-
thicin in solution and alamethicin bound to the membrane
(Figure 1).70 It also implies that it is relatively easy energetically
to extract alamethicin from a lipid bilayer in which it is
dissolved. Upon application of a transmembrane voltage,
membrane-associated alamethicin molecules can adopt a trans-
membrane configuration in which the axis of the R-helix is
oriented perpendicular to the plain of the bilayer (alamethicin
has a permanent dipole moment along this axis, which corre-
sponds to a net ∼+0.5 charge at the N-terminus of the helix
and a net ∼-0.5 charge at its C-terminus70). The probability
for alamethicin monomers to adopt this perpendicular orientation
increases strongly (nonlinearly) if the applied transmembrane
voltage exceeds a certain threshold voltage. Once alamethicin
adopts the transmembrane configuration, self-assembly of
alamethicin monomers leads to pores that can comprise up to
11 monomers.3 This number of monomers in a pore fluctuates
dynamically; these fluctuations lead to the characteristic stepwise
changes between discrete conductance levels of single alame-
thicin pores.49,50,71 To perform single-channel recordings of
alamethicin pores, a constant voltage above the threshold voltage
has to be applied (in the work presented here, we used a
In a series of elegant studies, Bayley and co-workers showed
that wild-type or genetically modified R-hemolysin pores can
be used to detect individual molecules, binding interactions,
reversible chemical reactions, or point mutations in DNA
strands.39,58-61 Despite these scientifically compelling results,
the use of R-hemolysin remains limited to specialized research
laboratories. Two factors impede its widespread use: (i)
experiments involving planar lipid bilayers require substantial
technical expertise, and the bilayers are typically only stable
for hours; (ii) rationally designed modification of R-hemolysin,
using methods pioneered by Bayley, requires the tools and
expertise of molecular biology. For a broader application of
these systems, it would be useful if ion channel-forming peptides
could be synthesized chemically at low cost, with high yield,
high purity, and good stability; channel synthesis of (relatively)
low molecular weight species would make it possible to produce
variants at will, and to imagine certain large-volume applications
that would be difficult for proteins.
Because they are relatively small molecules (<2 kD) and
commercially available, melittin, gramicidin A, and alamethi-
cin62 might become interesting building blocks for ion-channel
sensors if they could be modified synthetically. Woolley et al.
demonstrated, for instance, that the ion permeability through
chemically modified gramicidin pores in a lipid bilayer could
be used as a pH sensor.42 In a recent proof of principle study,
the groups of Mayer and Yang showed that gramicidin A can
be engineered to detect specific chemically reactive agents in
solution.5,63 Futaki’s group employed peptide synthesis to
prepare an analogue of alamethicin with a carboxylic acid group
on the C-terminal end that could be modified selectively;6,51-53
(45) Antonenko, Y. N.; Rokitskaya, T. I.; Kotova, E. A.; Agapov, II; Tonevitsky,
A. G. Biochemistry (Mosc) 2004, 69, 220-227.
(46) Antonenko, Y. N.; Rokitskaya, T. I.; Kotova, E. A.; Reznik, G. O.; Sano,
T.; Cantor, C. R. Biochemistry 2004, 43, 4575-4582.
(47) Shapovalov, V. L.; Rokitskaya, T. I.; Kotova, E. A.; Krokhin, O. V.;
Antonenko, Y. N. Photochem. Photobiol. 2001, 74, 1-7.
(48) Sansom, M. S. Prog. Biophys. Mol. Biol. 1991, 55, 139-235.
(49) Mayer, M.; Kriebel, J. K.; Tosteson, M. T.; Whitesides, G. M. Biophys. J.
2003, 85, 2684-2695.
(50) Schmidt, C.; Mayer, M.; Vogel, H. Angew. Chem., Int. Ed. 2000, 39, 3137-
3140.
(51) Zhang, Y.; Futaki, S.; Kiwada, T.; Sugiura, Y. Bioorg. Med. Chem. 2002,
10, 2635-2639.
(52) Futaki, S.; Fukuda, M.; Omote, M.; Yamauchi, K.; Yagami, T.; Niwa, M.;
Sugiura, Y. J. Am. Chem. Soc. 2001, 123, 12127-12134.
(53) Futaki, S.; Zhang, Y. J.; Sugiura, Y. Tetrahedron Lett. 2001, 42, 1563-
1565.
(54) Futaki, S.; Asami, K. Chem. BiodiVers. 2007, 4, 1313-1322.
(55) Yin, P.; Burns, C. J.; Osman, P. D.; Cornell, B. A. Biosens. Bioelectron.
2003, 18, 389-397.
(56) Goldermann, M.; Hanke, W. MicrograVity Sci. Technol. 2001, 13, 35-38.
(57) Miller, C. Ion Channel Reconstitution, 1st ed.; Plenum Press: New York,
1986; pp 131-140.
(64) Krishnamurthy, V. M.; Kaufman, G. K.; Urbach, A. R.; Gitlin, I.; Gudiksen,
K. L.; Weibel, D. B.; Whitesides, G. M. Chem. ReV. 2007, in press.
(65) Gitlin, I.; Mayer, M.; Whitesides, G. M. J. Phys. Chem. B 2003, 107, 1466-
1472.
(58) Shin, S. H.; Luchian, T.; Cheley, S.; Braha, O.; Bayley, H. Angew. Chem.,
Int. Ed. 2002, 41, 3707-3709.
(66) Colton, I. J.; Carbeck, J. D.; Rao, J.; Whitesides, G. M. Electrophoresis
1998, 19, 367-382.
(59) Howorka, S.; Movileanu, L.; Braha, O.; Bayley, H. Proc. Natl. Acad. Sci.
U.S.A. 2001, 98, 12996-13001.
(67) Day, Y. S.; Baird, C. L.; Rich, R. L.; Myszka, D. G. Protein Sci. 2002, 11,
1017-1025.
(60) Movileanu, L.; Howorka, S.; Braha, O.; Bayley, H. Nat. Biotechnol. 2000,
18, 1091-1095.
(68) Woolley, G. A.; Wallace, B. A. J. Membr. Biol. 1992, 129, 109-136.
(69) Salditt, T.; Li, C.; Spaar, A. Biochim. Biophys. Acta 2006, 1758, 1483-
1498.
(61) Gu, L. Q.; Dalla Serra, M.; Vincent, J. B.; Vigh, G.; Cheley, S.; Braha, O.;
Bayley, H. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 3959-3964.
(62) Huang, H. W. Biochim. Biophys. Acta 2006, 1758, 1292-1302.
(63) Capone R.; Blake S.; Rincon Restrepo M.; Yang J.; Mayer, M. J. Am.
Chem. Soc. 2007, 129, 9737-9745.
(70) Bechinger, B. J. Membr. Biol. 1997, 156, 197-211.
(71) Sondermann, M.; George, M.; Fertig, N.; Behrends, J. C. Biochim. Biophys.
Acta 2006, 1758, 545-551.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1455

A R T I C L E S
Mayer et al.
processes. (i) Alamethicin pores are formed by the dynamic self-
assembly of up to eleven monomers (in contrast to gramicidin
pores that form by self-assembly of two peptides or melittin
pores that form by self-assembly of four peptides); the variability
of the composition of the pores results in a cooperative
mechanism of pore formation and hence in a power-law of the
dependence of the current through alamethicin pores on the
concentration of alamethicin (eq 1). This strong (nonlinear)
response of the macroscopic current through alamethicin pores
as a function of the alamethicin concentration is attractive for
sensing, if the reaction that is used for detection can modulate
this concentration and thus yield a strong response from the
sensor system. (ii) The self-assembly of alamethicin can be
monitored with high precision and with a time-resolution below
milliseconds, by using established planar bilayer recording
procedures.49,71 (iii) Alamethicin is commercially available and
can be modified chemically, while preserving its ion channel-
forming properties.51,52,74 The discrete conductance levels
observed from single-channel recordings of alamethicin make
it possible to examine the effect of chemical modification on
the characteristics of the peptide that influence its ability to form
ion channels.75 (iv) Alamethicin is a relatively small molecule,
and thus, binding of a protein with a 10-fold larger molecular
weight may be able to interfere with its self-assembly. (v)
Alamethicin incorporates spontaneously into bilayers; spontane-
ous self-assembly makes its use more practical than possible
uses of most ion-channel proteins. (vi) Alamethicin (or 8 in the
work presented here), once added to the aqueous solution of a
planar bilayer experiment, establishes a chain of equilibria
(Figure 1) between free 8 in solution, 8 bound to the lipid
bilayer, and 8 in an open, conducting pore. Assuming that
molecules of 8 that are bound to CA II cannot participate in
the formation of an open, conductive ion channel (either by
inhibition of self-assembly of the complex of CA II-8 to a
pore or by removal of CA II-8 from the bilayer), we
hypothesized that reducing any one of the concentrations of
Figure 1. Chain of equilibria of alamethicin-sulfonamide (8) in a recording
chamber for planar lipid bilayer recordings. Only the molecules of 8 that
self-assembled to a pore in the bilayer, 8pore, can facilitate current through
the membrane. CA II can bind to molecules of 8 in all “compartments”
(i.e., in solution, in the bilayer, or in the pore) to form the complex CA
II-8. Note, however, that according to the model that we propose in this
work, binding of CA II to molecules of 8 which are in a conducting pore,
effectively removes 8pore from the pore. It is possible that binding of CA II
to 8pore results transiently in the blockage of a pore; we hypothesize,
however, that due to the bulkiness of the CA II-8 complex, it is not likely
that, in steady-state conditions, complexes of CA II-8 remain a part of fully
assembled pores. Instead, we propose that CA II-8 complexes will be present
in one of the other two compartments (i.e., in the solution or in the bilayer).
In addition, binding of CA II to 8 in the solution or in the bilayer, effectively
reduces the concentration of free 8 in all three compartments and hence
shifts the equilibria such that ultimately the concentration of 8pore is reduced
and consequently the transmembrane current is reduced as well.
transmembrane voltage of 0.14 V for all experiments49). To
record ion flux through single alamethicin pores, the concentra-
tion of the peptide, [alamethicin], had to be low (∼5 nM 8 under
the experimental conditions used in this work) to avoid the
occurrence of multiple pores at the same time.49,72 Increasing
the concentration of the peptide, while keeping the transmem-
brane potential constant above threshold, led to an increase of
alamethicin monomers in the transmembrane configuration and
thus to a strong increase in the probability of channel formation
due to the cooperative nature of pore formation.62 The resulting
“macroscopic currents” across the membrane were due to the
opening of multiple (often many) pores at the same time. If we
kept all other experimental parameters constant (applied voltage,
ionic strength, conductance of the electrolyte, lipid composition,
bilayer thickness, surface area of the planar bilayer, etc.), then
the “macroscopic conductance”, G (in Ω-1 or Siemens, S),
through a membrane with many alamethicin pores followed the
relationship:
8solution, 8bilayer, or 8pore by binding to CA II will shift these
equilibria. This shift will ultimately lower the concentration of
alamethicin in open pores, [8pore]. Due to the power law in eq
1, this reduction of [8pore] would lead to a strong reduction in
the recorded transmembrane current.
For sensing applications, the stability of planar lipid bilayers
is often a limiting factor. In order to generate lipid bilayers that
were stable over several hours, we prepared planar membranes
over micropores in Teflon films by the “folding technique” (a
technique that prepares bilayers by apposition of lipid mono-
layers which had previously been spread at an air-water
interface) for all experiments.49,76,77 These folded bilayers have
the advantage that they form readily over small apertures (here,
<50 µm), which increases the mechanical stability and reduces
the electrical noise. Moreover, folded bilayers result in so-called
“solvent-free” membranes, i.e., in lipid bilayers that contain only
a small amount of organic solvent.76 We expected that the
reduced amount of solvent would minimize fluctuations in
bilayer area and thus facilitate reliable quantification of the
current flux across the bilayer.77
G
[alamethicin]n
(1)
where n, the power dependence of G on the peptide concentra-
tion, has previously been used as an estimate of the average
number of alamethicin monomers in the channels. This simple
approach (eq 1) provides, however, at best, only a rough estimate
of the average number of monomers in the channels.73
Experimental Design. We chose alamethicin for this study
because we wanted to explore the possibility of interfering with
a self-assembly process in such a way that it could be monitored
in real time and, hence, exploited for sensing. Six characteristics
of alamethicin make this ion channel-forming peptide an
interesting choice for detecting the modulation of self-assembly
(72) It is also possible to apply a voltage that is below the threshold voltage
while increasing the concentration of alamethicin until single channel events
can be recorded.
(73) Hall, J. E.; Vodyanoy, I.; Balasubramanian, T. M.; Marshall, G. R. Biophys.
J. 1984, 45, 233-247.
(74) Wassner, A. J.; Hurt, J. A.; Lear, J. D.; Akerfeldt, K. S. Org. Lett. 2002,
4, 1647-1649.
(75) Duclohier, H.; Wroblewski, H. J. Membr. Biol. 2001, 184, 1-12.
(76) White, S. H. Biophys. J. 1972, 12, 432-445.
(77) White, S. H.; Thompson, T. E. Biochim. Biophys. Acta 1973, 323, 7-22.
9
1456 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
Figure 2. Comparison of single-channel recordings of 8 with recordings of native alamethicin (6). (A) Single-channel current of native alamethicin (fraction
F30). (B) Single-channel current of 8, recorded under the same conditions as (A) and shown at the same scaling of the axes. Note the well-defined conductance
states, the increased lifetime of individual conductance states (1-4; “C” stands for closed state), and the increased current noise of 8 compared to native
alamethicin. We used a low-pass filter with a cutoff frequency of 10 kHz and a sampling rate of 100 kHz for both recordings.
Results and Discussion
strated that the C-terminal alcohol of alamethicin can be oxidized
under mild conditions to the corresponding aldehyde (7) in 50%
isolated yield (Scheme 1). Reaction of this aldehyde with an
O-alkylated hydroxyl amine carrying a benzenesulfonamide
moiety (5) afforded oxime (alamethicin sulfonamide, 8) in 48%
isolated yield. Although the strategy presented in Scheme 1 is
not ideal due to the requirement for HPLC purification, it is,
we believe, a first step toward a practical synthesis of derivatives
of alamethicin useful in the applications we envision.
Synthesis. To make sensors based on ion channel-forming
peptides practical, a straightforward synthesis of the sensing
element that gives the required molecule (and variants of it) in
good yield and high purity is desirable. Our work constitutes a
step in that direction but is not yet a complete success. Several
groups have reported methods of derivatization of alamethi-
cin74,78-81 through C-terminal modification of native or synthetic
alamethicin.35,82,83 The most common synthetic strategy that
produces alamethicin derivatives with good to moderate ion
channel activity is the formation of an ester or carbamate on
the C-terminal phenylalaninol alcohol moiety of alamethi-
cin.35,82-84 Our own attempts to make derivatives of alamethicin
through esterification of the C-terminal alcohol resulted in
impractically low yields and prompted us to explore new routes
to derivatization of the C-terminus of alamethicin. Using HPLC-
purified, commercially available alamethicin (6), we demon-
Ion-Channel Measurements. To characterize the ion channel
activity of 8, we performed single-channel recordings in a planar
bilayer setup (see Experimental Section for details),49 and
compared the results to native alamethicin (6 in Scheme 1).
Figure 2 shows a representative trace of current versus time for
both molecules. Interestingly, 8 was capable of forming well-
defined conductance states, although its single channel conduc-
tance was slightly reduced compared to native alamethicin
(conductance states or conductance levels refer to assemblies
of alamethicin that formed pores with discrete conductance of
ions through the bilayer). This reduction was most pronounced
at small conductance levels: conductance level 1 of 8 reached
52% of conductance level 1 of native alamethicin, conductance
level 2 reached 61%, conductance level 3 reached 73%, and
conductance level 4 reached 78% of the conductance of the
corresponding level of native alamethicin. In addition, the deriv-
atized molecule resulted in noisier traces than native alamethicin.
(78) Akaji, K.; Tamai, Y.; Kiso, Y. Tetrahedron Lett. 1995, 36, 9341-9344.
(79) Jaikaran, D. C.; Biggin, P. C.; Wenschuh, H.; Sansom, M. S.; Woolley, G.
A. Biochemistry 1997, 36, 13873-13881.
(80) Lougheed, T.; Borisenko, V.; Hennig, T.; Ruck-Braun, K.; Woolley, G.
A. Org. Biomol. Chem. 2004, 2, 2798-2801.
(81) Peggion, C.; Coin, I.; Toniolo, C. Biopolymers 2004, 76, 485-493.
(82) Schmitt, J. D.; Sansom, M. S.; Kerr, I. D.; Lunt, G. G.; Eisenthal, R.
Biochemistry 1997, 36, 1115-1122.
(83) You, S.; Peng, S.; Lien, L.; Breed, J.; Sansom, M. S.; Woolley, G. A.
Biochemistry 1996, 35, 6225-6232.
(84) Woolley, G. A.; Biggin, P. C.; Schultz, A.; Lien, L.; Jaikaran, D. C.; Breed,
J.; Crowhurst, K.; Sansom, M. S. Biophys. J. 1997, 73, 770-778.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1457

A R T I C L E S
Mayer et al.
Figure 3. Simplified model of pores of alamethicin-sulfonamide (8) and
energy minimized representations of 8. (A) Top view of pores of 8
represented as a circular assembly of cylinders that form a central pore in
a lipid bilayer. In this figure, the red disks represent the alamethicin part of
8 and the black triangles represent the sulfonamide moiety with its linker
(5); for clarity, all sulfonamide groups were positioned to face the lumen
of the channel. The figure illustrates that the fraction of the cross-sectional
area of the lumen that is occupied by sulfonamide groups decreases with
increasing number of channel-forming peptides in the assembly. This figure
is not drawn to scale, and the smallest conducting pore of 8 may be
composed of 3-5 molecules of 8.49 (B) Top view representation of an
energy minimized structure of 8 (the carbon atoms of the sulfonamide moiety
are shown in gray). (C) Corresponding side view representation of 8. The
scale bar refers to the structural representations of 8 shown in (B) and (C).
Figure 4. Concept of sensing by binding of analyte to a functionalized
ion channel-forming peptide. (A) Monomers of 8 (red cylinders) with
covalently attached ligands (small black arrows) self-assemble to form pores
in a planar lipid bilayer. The current (flux of ions) through single or multiple
pores is recorded with a patch clamp amplifier. (B) Addition of a
macromolecule (here CA II, represented in blue) that binds to the ligand,
may have two consequences: (1) it may disrupt the pore, either by steric
hindrance, or by removing the peptide from the bilayer (see asterisk), or
(2) it may bind close to the mouth of the pore and block it. In either case,
only a small (or no) current would be recorded as a result of the interaction.
(C) Addition of competitive ligand (here 9, small gray arrows) to the solution
leads to binding of free ligand to the proteins, and releases ion channel-
forming peptides. This action makes it possible for the peptides to self-
assemble again to a conducting pore, or it unblocks an existing pore that
was blocked by the protein.
This additional noise may be caused by structural fluctuations
of the covalently attached sulfonamide group85 (i.e., the attached
group may fold into and out of the pore) or by small-scale
dynamic rearrangements of the self-assembled pore due to the
presence of the sulfonamide moiety. The observation that the
difference in conductance between 8 and native alamethicin was
most pronounced in small pores supports the hypothesis that
the sulfonamide moiety partly occluded the channel by folding
toward or into it; Figure 3 illustrates schematically that such
an occlusion would have a larger effect on small pores than on
large pores. Figure 3B and C shows that the size of the sulfon-
amide group with its linker (5) relative to the size of alamethicin
(6) may affect the ion channel conduction through pores of 8 if
5 were to bend toward the lumen of the pore. The energy-mini-
mized structures in Figure 3B and C represent only one plausible
conformation of 8; the flexible backbone of 5 is expected to
adopt a number of different conformations in solution.
Figure 4 illustrates the concept of an affinity sensor based
on ion channel-forming peptides. The ionic conductance through
a pore is modulated by specific binding of a macromolecule to
the peptides that form the pore. Futaki’s group proposed that
in their experimental system of biotinylated alamethicin, the
reduction in current was due to blocking of alamethicin pores
by streptavidin or anti-biotin antibodies, which bound close to
the mouth of the pore.51 Although we cannot rule out this
mechanism, we believe that blocking of functional (assembled)
channels would rather reduce the conductance instead of
completely abolishing it as Futaki’s group suggested (especially
in the case of large alamethicin pores). We propose another
mechanism as illustrated in Figure 4. In this mechanism, binding
of CA II impedes the self-assembly of alamethicin monomers
due to steric hindrance of the bulky protein that is bound to the
benzenesulfonamide group on this peptide (Futaki’s group had
mentioned this mechanism as a possibility).51 This mechanism
of action, if correct, could lead to complete cessation of ion
channel activity, given that the concentration of CA II is
sufficiently high that the majority of 8 is bound to CA II (and
that no impurities of 6 are present). Another mechanism for
impeding the assembly of the pores might be extraction of
monomers of 8 entirely from the bilayer by formation of a
soluble CA II-8 complex; this complex may be less prone to
membrane interactions than free 8. It is also possible that several
mechanisms, including blockage of pores,51 act in parallel or
sequentially, depending on the concentration of CA II and 8.
Quantitative Analysis of Transported Charge through
Bilayers by Time-Averaging. In the work presented here, we
performed planar lipid bilayer experiments to record macro-
scopic currents through pores from an alamethicin derivative,
8. To determine the dependence of the recorded current, I, on
the concentration of 8, we analyzed the ion channel current over
intervals of time t of 5 min. The area under the resulting
current-time trace corresponded to the transported charge over
the 5 min interval, Q5:
Q ) 5 min I(t)‚dt
(2)
∫
5
t)0
We then used this time-averaging approach to quantify the
binding of CA II to 8 based on the hypothesis that the binding
of the comparably bulky CA II protein (molecular weight ∼30
kD) to the comparably small peptide 8 (molecular weight ∼2.5
kD) would impede the bound molecules of 8 from participating
in functional self-assembly to a conducting pore. By comparing
(85) Woolley, G. A.; Jaikaran, A. S. I.; Zhang, Z. H.; Peng, S. Y. J. Am. Chem.
Soc. 1995, 117, 4448-4454.
9
1458 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
pores of 8, we had to account for variations in the surface area
of the planar lipid bilayers from experiment to experiment (the
solvent torus that surrounds planar bilayers can vary in thickness
and area even if the same aperture is used in repeated
experiments76,86). To account for these variations in bilayer area,
we determined first the value of Q5max by adding 16.1 nM 8 to
both bilayer chambers before starting each binding experiment.
Then, after adding for instance CA II, we normalized the
recorded transported charge Q5 over the particular value of Q5max
that was determined just before addition of CA II and termed
the resulting normalized charge, “relative transported charge,
Q5r”:
Relative transported charge over 5 min:
Q5
Q5r (in %) )
‚100
(3)
Q5max
Figure 6 shows that Q5r depended strongly on the concentration
of 8 in the recording chambers. The increase in Q5r (in %) was
fitted well (R2 ) 0.999, N ) 7) with an equation of the type:
Q5r ) 3.4‚1035‚[8]4.3
(4)
We chose a power law based on eq 1, because current and thus
transported charge are proportional to the conductance G. In
eq 4, the concentration of 8 is expressed in units of M (i.e.,
mol L-1) and the constant 3.4‚1035 is expressed in units of M-4.3
.
Figure 5. Detection of CA II by disruption (or blocking) of self-assembled
ion channels and restoration of ion-channel activity upon addition of
competitive inhibitor 9. A) Original trace of current versus time illustrating
the macroscopic current through many pores of 8 (concentration in the
chamber 40 ng mL-1 or 16.1 nM) before addition of CA II. B) Original
trace of current versus time after addition of ∼1 µM CA II. C) Original
trace of current versus time recorded shortly after addition of 31.2 µM 9.
The applied potential was +140 mV for all recordings. The cartoons next
to the current versus time traces show possible arrangements of 8, CA II,
and 9. Note that the current traces shown here are only 1 min long; to
increase the reproducibility of the quantification of the transported charge,
we increased the interval for time-averaging to 5 min for all quantitative
recordings in this work.
Determination of Equilibrium Binding Constants. Figure
6B shows the effect of the addition of CA II to the compartments
of the recording chamber containing 16.1 nM 8. The channel
activity showed a strong reduction in ionic conductivity and
hence in Q5r upon increasing the concentration of CA II from
0 to 2 µM. We suggest that this reduction in trans-bilayer current
was due either to disruption of the self-assembled pores, or to
blocking of the pores by binding of CA II close to the mouth
of the pore (Figure 4). The line in Figure 6B represents a best
fit to a model that is based on binding of CA II to 8. In this
model, we made the following nine assumptions: (i) Only “free”
molecules of 8 (i.e., 8 that was not bound to CA II) could
contribute to the formation of conductive alamethicin pores. This
assumption does not imply that, at any given time, all free
molecules of 8 actually participated in forming a pore, only the
molecules of 8 that were located in the bilayer and self-
assembled to pores participated in conducting channels, but it
does imply that only free molecules of 8 could participate in
the chain of equilibria that could lead to open pores (illustrated
in Figure 1). Binding of 8 to CA II would thus effectively reduce
the concentration of free 8 and consequently shift the equilibria
in Figure 1 in such a way that less free 8 would be available in
the bilayer to participate in the formation of conducting pores
than before the addition of CA II. (ii) Binding of one molecule
of CA II “removed” only one molecule of 8, it did not remove
aggregated clusters involving multiple molecules of 8. In
contrast to this assumption, if a single molecule of CA II would
block a conducting pore, then binding of each molecule of CA
II would “remove” at least three molecules of 8 since the
smallest conducting alamethicin pores are believed to consist
of at least three (probably four) alamethicin monomers.87 The
model for quantification employed here is hence different from
the value of the transported charge during an interval of 5 min,
Q5, before and after incubating a constant concentration of 8
with increasing concentrations of CA II, we explored the
possibility of determining the binding constant between CA II
and 8.
Figure 5A and B illustrates this time-averaging approach to
the binding of CA II to 8. According to eq 2, we integrated the
area under the curve of current versus time (illustrated by the
yellow area in Figure 5A) to obtain the total charge Q5 (in
Coulomb, C) that was transported through the bilayer over an
interval of 5 min. Figure 5 A shows that a concentration of 40
ng mL-1 (or 16.1 nM) of 8 in the bilayer chamber resulted in
macroscopic currents that could exceed 3 nA. Because the
amplifier with the gain setting that we used cannot measure
currents above 20 nA due to current overload, we chose,
somewhat arbitrarily, the concentration of 16.1 nM of 8 as the
maximum concentration that would ensure, under all conditions,
that even rare current spikes with large amplitude could be
recorded accurately by the amplifier. We termed the charge that
was transported during 5 min through bilayers in the presence
of 16.1 nM 8, the maximum transported charge, Q5max. Because
our goal was to establish a repeatable method for quantifying
the binding of CA II to 8 based on the measured current through
(86) Brullemans, M.; Tancrde, P. Biophys. Chem. 1987, 27, 225-231.
(87) Hanke, W.; Boheim, G. Biochim. Biophys. Acta 1980, 596, 456-462.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1459

A R T I C L E S
Mayer et al.
and no partial binding occurred. (vii) Binding did not alter 8 or
CA II irreversibly. (viii) Binding of 8 to CA II was reversible.
(ix) The binding constant between 8 and CA II was the same
regardless of the location where binding occurred (i.e., in
solution, in the bilayer, or in a pore).
The fit to a power law in Figure 6A yielded an empirical
relationship between Q5r and the total concentration of 8 in the
recording chambers. Because in Figure 6A the total concentra-
tion of 8 was equal to the free concentration of 8 (because no
CA II was present and hence 8 could not be bound to CA II),
we were able to use the measured values of Q5r in Figure 6B to
calculate the concentration of free 8 at each concentration of
CA II in the reaction mixture. With this information, it is
possible to calculate the concentration of 8 that is bound to CA
II and hence to determine a binding constant by fitting the data
in Figure 6B. To derive the function to fit the data in Figure
6B, we used the law of mass action for the binding reaction
between CA II and 8:
CA II + 8 / CA II-8
[CA II - 8]eq
(5)
(6)
Ka8
)
[8]eq‚[CA II]eq
where Ka8 represents the affinity constant of the binding reaction
shown in eq 5, [CA II-8]eq is the molar concentration of the
complex of CA II with 8 at equilibrium, [8]eq is the molar
concentration of free 8 at equilibrium, and [CA II]eq is the molar
concentration of free CA II at equilibrium. Because we
ultimately wanted to fit the data in Figure 6B, we were seeking
a functional relationship between Q5r and the concentration of
CA II that was initially added to the recording chambers, [CA
II]0 (i.e., the concentration of CA II before binding to 8). As
mentioned previously, eq 4 gave a relationship between Q5r and
[8]. Because we assumed that only free 8 (i.e., not 8 bound to
CA II) could participate in pore formation, the concentration
of 8, [8], in eq 4 referred, in this model, to the concentration of
free, unbound 8 at equilibrium, [8]eq. Hence, the relationship
between Q5r and [8]eq can be expressed as:
Figure 6. Ion flux through lipid bilayers after self-assembly of ion channels
and disruption of self-assembly (or blockage of channels) as a function of
increasing concentration of 8 and CA II. (A) Relative transported charge
through the bilayer during a period of 5 min, Q5r ) (Q5/Q5max)‚100%, as a
function of increasing concentrations of 8 in both compartments of the
recording chamber (note, before addition of CA II, [8], [8]0, and [8]eq were
the same). Q5 represents the total charge transported through the bilayer
during 5 min of recording at the concentration of free 8 indicated in the
graph and Q5max represents the charge transported through the bilayer during
5 min of recording when the maximum concentration of free 8 was present
in the chambers (16.1 nM). The line represents a best fit of the data to eq
4. (B) Relative transported charge Q5r as a function of increasing total
concentration of CA II, [CA II]0, in both compartments of the recording
chamber. Both chambers contained 1.61‚10-8 M 8. The solid line represents
a best fit of the data to eq 14 (the point marked with an asterisk was excluded
from this best curve fit). The dashed line represents a best fit of all the data
to eq 14 (including the point marked with an asterisk). (Inset) First five
data in detail. Points represent mean values of Q5r; the number next to each
point indicates how many repetitions were performed to calculate the mean
value of Q5r. Mean values that were calculated from at least seven repetitions
are shown with the standard deviation (error bars). For all other points, the
Q5r ) 3.4‚1035‚[8]eq
(7)
4.3
Rearranging eq 6 yields:
[8]eq
[CA II - 8]eq
Ka8‚[CA II]eq
)
(8)
Conservation of mass requires that eqs 9 and 10 hold ([8]0 )
16.1‚10-9 M):
[CA II-8]eq ) [8]0 - [8]eq
(9)
[CA II]eq ) [CA II]0 - [CA II-8]eq.
(10)
variation in values of Q5r was typically within (20% of Q5max
.
Because the total concentration of 8, [8]0, in the recording
chambers was low (nanomolar range) compared to the concen-
trations of [CA II]0 that we added to the recording chambers
(micromolar range), we assumed that the concentration of free
CA II in equilibrium would be approximately the same as the
initial concentration of CA II:88,89
a model that assumes blockage of alamethicin pores. (iii)
Nonspecific binding of 8 to CA II or to the walls of the recording
chambers or electrodes and stir bars could be neglected. (iv)
All reactions (i.e., partitioning of 8 and binding of CA II to 8)
reached equilibrium. (v) All molecules of 8 were equally
accessible to CA II and vice versa. (vi) Molecules of 8 were
either free or bound to CA II; only one affinity state existed
[CA II]eq ≈ [CA II]0
(11)
9
1460 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
Incorporating eq 9 and 11 into eq 8 yields:
[8]0 - [8]eq
[8]eq
)
(12)
(13)
Ka8‚[CA II]0
and solving eq 12 for [8]eq gives:
[8]0
[8]eq
)
Ka8‚[CA II]0 + 1
Finally, combining eq 13 with eq 7, gives the desired relation-
ship between Q5r and the concentration of CA II that we added
to the recording chambers, [CA II]0:
4.3
[8]0
Q5r ) 3.4‚1035‚
(14)
Figure 7. Relative transported charge Q5r as a function of increasing
concentration of competitive inhibitor 9 in both compartments of the
recording chamber. Both chambers initially contained [8]0 ) 16.1 nM and
[CA II]0 ) 9.3 µM. The solid curve represents a sigmoidal best fit to eq
15. This function approaches Q5r ) 100% for large concentrations of [9]0,
as expected in the presence of an excess of competitor 9, which would
release all 8 from the complex with CA II (Kd9 ) 0.73 ( 0.02 µM67). The
(
)
Ka8‚[CA II]0 + 1
We used eq 14 to fit the data in Figure 6B and obtained Ka8
as the only fitting parameter. We performed this analysis twice;
the first fit included all points in Figure 6B (dotted line). This
curve fit returned a value of Ka8 ) (3.6 ( 1.5)‚105 M-1 (R2 )
0.742, N ) 8) corresponding to a Kd8 ) 2.8 µM. Because the
datum that is marked with an asterisk in Figure 6B appears to
be an outlier (Figure 6B inset), we also performed the best fit
analysis by excluding this point from the analysis (solid line).
This approach returned a best fit with a value of Ka8 ) (6.7 (
1.5)‚105 M-1 (R2 ) 0.963, N ) 7), corresponding to a
data shown as round symbols represent the fraction of 8 bound (Θ8,bound
)
to CA II as a function of [9]0. The values for Θ8,bound were calculated from
Q5r for every value of [9]0 using eqs 7 and 16. The IC50 value for the
displacement of 8 from CA II by adding increasing concentrations of
competetive ligand 9 was determined with a best fit analysis of the round
points with eq 17 (dotted curve). Note the logarithmic scale for [9]0. As
recommended by Motulsky and Christopoulos, we chose, somewhat
arbitrarily, log([9]0) ) -9 for [9]0 ) 0 M, because log(0) is not defined;92
the corresponding data are marked by open symbols. The data in this figure
were recorded only once: hence the absence of error bars.
dissociation constant, Kd8 ) Ka8 ) 1.5 µM.90
-1
To compare these values with a well-established method for
determining binding constants, we transformed the data in Figure
6B to construct a binding isotherm (see Figure S1, Supporting
Information). From the best curve fit analysis of this binding
isotherm, we obtained a value of Kd8 ) 0.9 ( 0.2 µM when we
excluded the apparent outlier (marked with an asterisk in Figure
6B and Figure S1) and a value of Kd8 ) 1.3 ( 0.3 µM when
we included the outlier in the best fit analysis. In summary,
depending on the transformation of the original data and the
corresponding best curve fit analysis, we obtained a mean value
of Kd8 ranging from 0.9 to 2.8 µM for the interaction between
8 and CA II in the recording chambers.
For comparison with an independent method, we measured
the affinity constant of CA II with benzenesulfonamide (5) by
isothermal titration calorimetry (ITC).91 We used precursor 5
for the ITC measurement because it was more soluble in aqueous
solutions than 8; this increased solubility was required to achieve
detectable reagent concentrations for ITC measurements. The
resulting dissociation constant Kd5 by ITC was 2.6 ( 0.6 µM,
which compares well with the range of 0.9-2.8 µM obtained
with the ion channel system presented here.
Competitive Binding Assay: Displacement of 8 from CA
II. To test if it was possible to displace the bound molecules of
8 from CA II by competitive binding, we added increasing
concentrations of a competitive inhibitor, 9 (Scheme 1), to both
compartments of a recording chamber which contained a total
concentration of 16.1 nM of 8 and 9.3 µM of CA II (i.e., [8]0
) 16.1 nM; [CA II]0 ) 9.3 µM). We expected 9 to be able to
displace 8 from CA II because the dissociation constant for
binding of 9 to CA II, Kd9, is 0.73 ( 0.02 µM.67 Figure 7 (data
shown as squares) shows that increasing concentrations of 9
resulted in an increase in the relative transported charge Q5r
through the bilayer. This increase corresponded to an increased
concentration of free 8 in solution as expected if 9 displaced 8
from the CA II-8 complex. The square-shaped points in Figure
7 could be fitted well (R2 ) 0.999, N ) 4) with a dose-response
curve (solid curve) of the following form:93
ymax - ymin
y ) ymin
+
(15)
(88) Motulsky, H.; Christopoulos, A. Fitting Models to Biological Data Using
Linear and Nonlinear Regression: A Practical Guide to CurVe Fitting,
1st ed.; Oxford University Press: New York, 2004; p 193.
(89) Lauffenburger, D. A.; Linderman, J. J. Receptors; Oxford University
Press: New York, 1993; p 23.
1 + 10(log EC
50-x)‚HillSlope
where y corresponds to Q5r, ymin represents the lower asymptotic
value of the sigmoidal function, ymax represents the upper
asymptotic value, log EC50 represents the logarithm of the
concentration that generates 50% response (EC50), x corresponds
to log[9]0, and HillSlope94 determines the steepness of the curve.
Before fitting the data (squares) in Figure 7 to eq 15, we set
ymin to 0.1% because we measured this value of Q5r experi-
(90) We tested the validity of the simplification that we made in eq 11 by
performing the same analysis and fit of the data in Figure 6B without any
simplification. In this case, we used eq 10 to obtain [CA II]eq, which then
lead to a quadratic equation to solve for [8]eq of the form: [8]eq ) (- (1
- Ka8[8]eq + Ka8[CA II]0) + ((1 - Ka8[8]eq + Ka8[CA II]0)2 + 4Ka8[8]eq)0.5)/
(2Ka8). Using this precise analysis, we obtained Ka8 ) (6.71 ( 1.53)‚105
M-1 and, hence, the same mean value for Ka8 as the one we obtained by
using the simplification in eq 11.
(91) Connors, K. A. Binding Constants: The Measurement of Molecular
Complex Stability; John Wiley & Sons: New York, 1987; pp 339-362.
(92) Motulsky, H.; Christopoulos, A. Fitting Models to Biological Data Using
Linear and Nonlinear Regression: A Practical Guide to CurVe Fitting,
1st ed.; Oxford University Press: New York, 2004; pp 210-217.
(93) Motulsky, H.; Christopoulos, A. Fitting Models to Biological Data Using
Linear and Nonlinear Regression: A Practical Guide to CurVe Fitting,
1st ed.; Oxford University Press: New York, 2004; pp 256-260.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1461

A R T I C L E S
Mayer et al.
mentally before adding 9 (i.e., [9]0 ) 0) and we set ymax to
100% because we assumed that at high concentrations of
competitive ligand (here [9]0 > 1 mM) virtually all of 8 would
be released from the CA II-8 complex (i.e., [8]eq ≈ [8]0 ) 16.1
nM) and that consequently the relative transported charge Q5r
would reach 100%.95 The best fit analysis of the square points
returned a value for log EC50 of -4.62 ( 0.01, corresponding
to EC50 ) 24.2 µM and a value for the HillSlope of 4.51 (
0.04.96
To determine the dissociation constant Kd9 for the interaction
between CA II and 9 based on the competitive binding data in
Figure 7, we used an approach developed by Linden in 1982.97
This approach makes it possible to calculate the dissociation
constant from competitive binding curves without the require-
ment that the concentration of any of the involved species has
to remain constant (i.e., it allows for depletion of all species).
We used this approach for accurate determination of Kd9 because
we expected that the addition of micromolar concentrations of
9 to the recording chambers which contained nanomolar
concentrations of 8 and micromolar concentrations of CA II
would lead to significant changes in the equilibrium concentra-
tions of free molecules of 8, 9, and CA II. Simplifications such
as the one made in eq 11 would not be appropriate for
determining Kd9 in this case.
bound to 9 and hence Θ8,bound would be 0. In order to obtain
the value of ymax, we calculated Θ8,bound before addition of 9
(i.e., [9]0 ) 0) in two steps. We first used eq 13 to calculate
[8]eq by employing the previously determined value of Ka8
)
6.7‚105 M-1 and a value of [CA II]0 ) 9.3‚10-9 as well as a
value of [8]0 ) 16.1 nM; with these values we obtained [8]eq
) 2.2‚10-9 M. Second, by using this value for [8]eq and eq 16,
we obtained a value of Θ8,bound ) 0.86. Consequently, we used
this value of 0.86 for ymax to perform the best curve fit of the
blue points in Figure 7 with eq 17. This analysis generated the
following two values: log IC50 ) -4.89 ( 0.04, and HillSlope
) 2.22 ( 0.44 (R2 ) 0.981, N ) 4).96 Hence, the mean value
for the IC50 concentration for displacement of 8 from CA II by
adding 9 was 12.9 µM.
Second, from this value of IC50 we calculated the concentra-
tion of 9 that was free in the equilibrium mixture [9]eq when
the total concentration of 9, [9]0, was equal to the IC50 (i.e.,
[9]0 ) 12.9 µM) with the following equation:97
[9]eq ) IC50 - [CA II]0 +
[CA II]0
2
[8]eq
Kd8
+
(18)
Kd8 + [8]eq
[CA II]0
2
(
)
Kd8 + [8]eq
+
Application of Linden’s approach proceeded in three steps.
First we used the competitive binding data in Figure 7 to
determine the “inhibitory” concentration of 9 that displaced 50%
of 8 from CA II, i.e., the IC50 concentration for competitive
binding of 9 to CA II in the presence of 8 (Figure 7). We
determined the IC50 value by plotting Θ8,bound, which is defined
as
For this calculation, we first computed the value of [8]eq when
[9]0 ) 12.9 µM in two steps: (i) eq 14 yielded a value of Q5r
) 5.7% when [9]0 ) 12.9 µM and (ii) from this value of Q5r,
eq 7 yielded the value of [8]eq ) 8.2 nM. The remaining two
values that we used to calculate [9]eq with eq 18 were Kd8
)
1.5 µM and [CA II]0 ) 9.3 µM. With these parameters, eq 18
returned a value of [9]eq ) 4.8 µM.
And in the third and final step, we used the value of [9]eq
)
[CA II - 8]eq [8]0 - [8]eq
4.8 µM and the following formula derived by Linden to calculate
Θ8,bound
)
)
(16)
the desired value of Kd9:97
[8]0
[8]0
[9]eq
as a function of [9]0 (round symbols in Figure 7) followed by
Kd9
)
(19)
fitting the data with an equation for competitive binding:92
[8]eq
Kd8
+
[8]eq [CA II]0
2
ymax - ymin
1 +
+
(
)
y ) ymin
+
(17)
Kd8
Kd8
Kd8 + [8]eq
1 + 10(x-log IC
50)‚HillSlope
with [8]eq ) 8.2 nM, Kd8 ) 1.5 µM, and [CA II]0 ) 9.3 µM,
we obtained a value for Kd9 ) 0.7 µM. If we repeated this
analysis by employing the range of values for Kd8 that we
determined previously (Kd8 ) 0.9 - 2.8 µM), we obtained a
range from 0.4 to 1.2 µM for the dissociation constant Kd9. The
value of 0.7 µM for Kd9 is in excellent agreement with the
dissociation constant of 0.73 ( 0.02 µM determined for the
same interaction (i.e., for binding of 4-carboxybenzenesulfona-
mide (9) to CA II) by isothermal titration calorimetry.67
Practicability of Quantifying Protein-Ligand Interactions
with Ion Channel-Forming Peptides. In asking if the approach
described here offers a broadly applicable assay platform, we
see three major challenges. First, the preparation of planar
bilayers of high quality requires expertise, experimental skill,
and appropriate equipment, which includes a low noise current
amplifier, a data acquisition board with adequate software, a
Faraday cage, and typically a vibration-damping platform.
Second, planar lipid bilayers are inherently metastable systems;
their lifetime typically ranges from a few minutes to a few hours.
When membranes broke, it was often necessary to wash the
where y corresponds to Θ8,bound, ymin and ymax represent the lower
and the upper asymptotic value of the sigmoidal function,
respectively, x corresponds to log [9]0, log IC50 represents the
logarithm of the IC50 concentration, and HillSlope determines
the steepness of the curve. We set ymin to 0, because we assumed
that at large concentrations of 9, all CA II molecules would be
(94) HillSlope can be interpreted as follows: A HillSlope with a value of 1.0
means that the y-value of a dose-response curve increases from 10 to 90%
of ymax, if the x-value increases by a factor of 81. A dose-response curve
with a HillSlope of 1.0 is called a standard dose-response curve. A HillSlope
greater than 1.0 means that the dose-response curve is steeper than this
standard curve, whereas a HillSlope smaller than 1.0 means that the curve
is shallower than this standard curve.
(95) Motulsky, H.; Christopoulos, A. Fitting Models to Biological Data Using
Linear and Nonlinear Regression: A Practical Guide to CurVe Fitting,
1st ed.; Oxford University Press: New York, 2004; pp 38-39.
(96) Note, a HillSlope with a value significantly different from 1 usually implies
that standard mechanistic mass-action models do not apply and consequently
it is difficult to interpret the IC50 value. Typically this analysis is, however,
performed under the assumption that the concentration of the competitive
inhibitor does not change upon binding (i.e., here this would mean that
[9]eq ) [9]0), whereas in the example we show, a significant fraction of 9
bound to CA II. The fact that Figure 7 was not plotted as a function of
log([9]eq) may contribute to the deviation of the HillSlope from unity.
(97) Linden, J. J. Cyclic Nucleotide Res. 1982, 8, 163-172.
9
1462 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
entire setup and start a new experiment. Because the analysis
presented here was based on equilibrium conditions of binding,
it took several hours to record entire dose-response curves.
Hence, in some cases only a part of the dose-response curves
could be completed with a given bilayer. Third, it can be difficult
to keep the bilayer area of a planar lipid membrane constant
over minutes to hours. Fluctuations in bilayer area (due to
fluctuations of the area and thickness of the torus) could cause
fluctuations in transported charge, and these fluctuations could
occur within one dose-response curve, potentially leading to
inaccuracies of Q5r. For each new bilayer, we measured the
capacitance of the membrane, a quantity that is proportional to
the bilayer area, and waited until the capacitance appeared to
reach a nearly constant value. For some membranes, this waiting
period could take tens of minutes. Then we measured Q5max for
this particular membrane before adding CA II or 9. In some
cases, we measured only a limited number of concentrations of
a dose-response curve and then purposely prepared a new
membrane to minimize possible fluctuations of the membrane
area over long periods of time.
The assays presented here, although promising and remark-
ably accurate with respect to the determined dissociation
constants, are thus, in their current configuration, not practical
for real-world sensor applications; they are best performed under
well-defined conditions in research laboratories relatively
proficient in ion-channel biochemistry. One potential strategy
to reduce problems with changing bilayer areas might be to
perform the assays on apertures with large diameters (e.g., 500
µm) rather than the small diameters that we chose (<50 µm).
If these large apertures would be prepared in very thin polymer
films, then the surface area of the membrane that would be
occupied by the torus would be relatively small compared to
the bilayer area over these large pores. Fluctuations in the area
of the torus may thus have a relatively small effect. Planar lipid
bilayers over large diameter pores are, however, less stable than
bilayers over small pores (which is the reason why we chose
small pores in this work), and the large capacitance of these
membranes generates large current noise in the recordings.49
Recent developments of hydrogel-embedded planar lipid bilayers
may, however, make it possible to extend the lifetime of these
bilayers.98
will have to be simplified, ideally to a degree that it could be
carried out by nonchemists. The work presented here is only a
first step in this direction because it starts from commercially
available building blocks. The overall yield was only moderate,
and the purification required expertise in preparative HPLC
techniques. Further requirements are that the ligands should have
low molecular weight and must be chemically amenable for
covalent coupling to the pore-forming peptide; they also must
be sufficiently stable to withstand the coupling reaction without
changing their physicochemical properties. As with most applied
experiments on planar lipid bilayers, the mechanical49 and
thermodynamic stability99 of the bilayers would have to be
improved before such sensors will yield widely used, robust,
and user-friendly assays. QuantitatiVe analyses of changes in
the transported charge through pores of ion channel-forming
peptides as a function of the addition of an analyte pose an
additional challenge: the area of the lipid bilayer has to be kept
as constant as possible or correction strategies have to be
developed that take into account possible fluctuations in bilayer
area over time. One possible strategy may be frequent measure-
ments of the electrical capacitance of the bilayer and normalizing
the relative transported charge over the membrane capacitance
(or the bilayer area, which is proportional to the membrane
capacitance). Another strategy may be to embed the bilayer into
a hydrogel polymerized in situ. The hydrogel can stabilize the
torus of the membrane and thus minimize fluctuations in area
of the bilayer.98
We believe that the active research efforts in the area of
microfabricated planar lipid bilayer setups,4,25,49,50,71,98,100-110 as
well as in the area of synthesizing functional derivatives of ion
channel-forming peptides,5,7,28,29,35,42,43,63,82,104,111,112 will help to
overcome, at least to some extent, these challenges of affinity
sensors based on ion channels. If so, these sensors might make
it possible to perform quantitative affinity assays in volumes
below 50 µL and at concentrations of ligand and receptor below
10 µM. Their advantage might thus be to require only subna-
nomole amounts of receptor and ligand.
In comparison to existing methods such as isothermal titration
calorimetry, the method presented here has the advantage that
it is a single molecule method in the sense that the flux of ions
through a single assembled pore can be monitored readily and
with a time resolution below milliseconds. Another advantage
Conclusion
Using a system of CA II and alamethicin covalently attached
to a sulfonamide ligand, we demonstrated that binding of a
protein to an ion channel-forming peptide that carries a ligand
can be used to quantify protein-ligand interactions by measur-
ing the inhibition of ion channel activity upon binding. Time-
averaging of the current transported through the bilayer made
it possible to determine the dissociation constant for this
monovalent protein-ligand interaction. The resulting assay
could be carried out in small volumes of solution (here 3 mL
but volumes smaller than 50 µL have been used for planar lipid
bilayer recordings of alamethicin activity50) and required
nanomolar concentrations of the ion channel-forming peptide
(here alamethicin sulfonamide) as well as low micromolar
concentrations of the protein (here CA II).
(99) White, S. H. The physical nature of planar bilayer membranes. In Ion
Channel Reconstitution, 1st ed.; Miller, C., Eds.; Plenum Press: New York,
1986; pp 3-35.
(100) Malmstadt, N.; Nash, M. A.; Purnell, R. F.; Schmidt, J. J. Nano Lett.
2006, 6, 1961-1965.
(101) Fertig, N.; Meyer, C.; Blick, R. H.; Trautmann, C.; Behrends, J. C. Phys.
ReV. E. 2001, 64, 040901-040904.
(102) Fertig, N.; Blick, R. H.; Behrends, J. C. Biophys. J. 2002, 82, 3056-
3062.
(103) Fertig, N.; Klau, M.; George, M.; Blick, R. H.; Behrends, J. C. Appl.
Phys. Lett. 2002, 81, 4865-4867.
(104) Borisenko, V.; Lougheed, T.; Hesse, J.; Fertig, N.; Behrends, J. C.;
Woolley, A.; Schuetz, G. J. Biophys. J. 2003, 84, 612-622.
(105) Wilk, S. J.; Petrossian, L.; Goryll, M.; Thornton, T. J.; Goodnick, S. M.;
Tang, J. M.; Eisenberg, R. S. Biosens. Bioelectron. 2007, 23, 183-190.
(106) Shim, J. W.; Gu, L. Q. Anal. Chem. 2007, 79, 2207-2213.
(107) Quist, A. P.; Chand, A.; Ramachandran, S.; Daraio, C.; Jin, S.; Lal, R.
Langmuir 2007, 23, 1375-1380.
(108) Funakoshi, K.; Suzuki, H.; Takeuchi, S. Anal. Chem. 2006, 78, 8169-
8174.
(109) Suzuki, H.; Tabata, K. V.; Noji, H.; Takeuchi, S. Biosens. Bioelectron.
2007, 22, 1111-1115.
To make this type of ion channel-based affinity sensors
practical, the synthetic route for attaching ligands to alamethicin
(110) Suzuki, H.; Tabata, K.; Kato-Yamada, Y.; Noji, H.; Takeuchi, S. Lab
Chip 2004, 4, 502-505.
(111) Lougheed, T.; Borisenko, V.; Hand, C. E.; Woolley, G. A. Bioconj. Chem.
2001, 12, 594-602.
(98) Jeon, T. J.; Malmstadt, N.; Schmidt, J. J. J. Am. Chem. Soc. 2006, 128,
42-43.
(112) Bayley, H. Nat. Chem. Biol. 2006, 2, 11-13.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1463

A R T I C L E S
Mayer et al.
mL min-1 (UV dectection at 214 and 254 nm). Alamethicin with a
purity >90% (HPLC) was purchased from Sigma (St. Louis, MO). We
started with commercial alamethicin (6) (isolated from Trichoderma
Viride) and separated the two alamethicin variants (F30 and F50), by
HPLC on a 5 µm C18 column (10 × 250 mm) using a linear gradient
of 70% A with 30% B to 100% B over 40 min at a flow rate of 5 mL
min-1. We used the F30 fraction for all experiments; fractions
containing alamethicin F30 were combined and evaporated to dryness
(purity >99%). MS (MALDI-TOF) m/z: 1965.77 [M + H]+, calculated
compared to isothermal titration calorimetry is that sensing with
ion channels requires significantly lower concentrations of ligand
(here 8) and the detection mechanism is, in principle, strikingly
simple (i.e., applying a small constant voltage and measuring a
small current over time). In addition, the footprint of the entire
experimental setup can be small making this method attractive
for potential applications as portable sensors.113-115
Moreover, sensing with ion channels is an amplifying method
and has some analogy to standard affinity assays such as
enzyme-linked immunosorbent assays (ELISAs), which utilize
the amplification provided by enzymatic turnover. This method
takes advantage of the inherent amplification of ion channels
in the sense that one channel can conduct thousands of ions
per millisecond. Another advantage of channel-based sensing
is that it is not sensitive to colored samples or samples that
might contain quenchers of fluorescence or inhibitors of the
enzymes that are used for amplification.63 In addition, ion
channel-based sensing can, presumably, proceed in solution as
long as the affinity interaction reduces the concentration of free
channel-forming peptide and thus shifts the equilibria in Figure
1 in the direction of reduced concentrations of peptide in a
functional pore. ELISA assays, in contrast, are inherently
surface-based assays.
Nonetheless, at this stage, ELISA assays provide superior
robustness and ease of use compared to the channel-based
affinity assay demonstrated in this report. Compared to isother-
mal titration calorimetry, sensing with ion channels has the
disadvantage that it requires derivatization of the ion channel-
forming peptide to attach each ligand of interest. Moreover, as
opposed to isothermal titration calorimetry, sensing based on
ion channels does not provide thermodynamic quantities such
as changes in enthalpy. Compared to both methods, ELISA and
isothermal titration calorimetry, sensing with ion channels, at
this stage of development, is hampered by the stability of the
bilayers, which is limited to periods of minutes to hours.
To carry out the analysis presented here, we proposed that
the disruption and restoration of self-assembly of a derivative
of alamethicin can be used for quantifying receptor ligand
interactions. Although we did not provide definite proof for this
mechanism of detection, we think that the excellent agreement
of the determined dissociation constants with literature values
support this mechanism. On the basis of these results, we suggest
that utilizing the disruption or restoration of self-assembly for
sensing applications is a fundamentally interesting and promising
avenue because self-assembly is often a cooperative process and
may thus respond strongly to specifically engineered distur-
bances.
for C92H151N22O25 [M + H]+ 1965.32.
+
F30 : Ac-Aib-Pro-Aib-Ala-Aib-Ala-Gln-Aib-Val-Aib-Gly-Leu-Aib-
Pro-Val-Aib-Aib-Glu-Gln-Phenylalaninol
F50 : Ac-Aib-Pro-Aib-Ala-Aib-Ala-Gln-Aib-Val-Aib-Gly-Leu-Aib-
Pro-Val-Aib-Aib-Gln-Gln-Phenylalaninol
Compound 1. Acid 1 was prepared as described by Avila et al.116
1
The yield was 89%, and the product was pure. H NMR (400 MHz,
DMSO-d6) ) δ 1.53 (m, 4H), 2.17 (t, 2H), 2.32 (t, 2H), 3.58 (s, 3H),
4.32 (d, 2H), 7.33 (s, 2H), 7.41 (d, 2H), 7.77 (d, 2H), 8.43 (t, 1H).
HRMS (FAB) m/z: 329.1181 [M + H]+, calculated for C14H21N2O5S+
[M + H]+ 329.1171.
Compound 2. Compound 2 was synthesized according to Roy and
Mallik.117 1H NMR (400 MHz, MeOH/CDCl3) ) δ 2.82 (t, 2H), 3.35-
3.39 (m, 2H), 3.53-3.66 (m, 6H), 5.07 (s, 2H), 7.31-7.34 (m, 5H).
Compound 3. To a solution of (Boc-aminooxy)acetic acid (0.437
g, 2.29 mmol), N-ethyl-N′-dimethylaminopropylcarbodiimide.HCl (0.438
g, 2.29 mmol) and N,N-diisopropylethylamine (1.09 mL, 6.24 mmol)
in dimethylformamide (5 mL) was added 2 (0.587 g, 2.08 mmol), and
the solution was stirred at room temperature for 5 h. The solution was
evaporated, and the product partitioned between ethyl acetate and water.
The organic layer was washed with water (2 × 50 mL), dried over
sodium sulfate, and evaporated to dryness. The crude compound was
chromatographed (SiO2, using 100% ethyl acetate as eluent) to yield 3
1
(530 mg, 1.16 mmol, 56% yield) as a clear oil. H NMR (500 MHz,
CD3OD) ) δ 1.46 (s, 9H), 1.57-1.69 (m, 4H), 3.29 (t, 2H), 3.43 (t,
2H), 3.54 (t, 2H), 3.57 (t, 2H), 3.61 (s, 4H), 4.25 (s, 2H), 5.07 (s, 2H),
7.25-7.38 (m, 5H); MS (MALDI-TOF) m/z 478.74 [M + Na]+,
+
calculated for C21H33N3NaO8 [M + Na]+ 478.51.
Coumpound 4. Compound 3 (157 mg, 0.34 mmol) and 10% Pd/C
(20 mg) were combined in ethanol (20 mL). This mixture was
hydrogenated at atmospheric pressure and room temp for 2 h. The
mixture was then filtered through Celite, the filter cake was washed
with 2 × 10 mL of ethanol, and the resulting solution was concentrated
by evaporation. The acid 1 (119 mg, 0.38 mmol), N-ethyl-N′-
dimethylaminopropylcarbodiimide.HCl (73 mg, 0.38 mmol) and N,N-
diisopropylethylamine (180 µL, 1.03 mmol) were combined in 5 mL
of DMF at room temperature with stirring, and the reaction was allowed
to proceed at room temperature overnight. The crude product was
purified by HPLC (linear gradient, 100% A to 80% B, over 40 min)
and lyophilized to afford 4 (111 mg, 0.18 mmol, 52% yield) as a clear
oil. 1H NMR (500 MHz, CD3OD) δ 1.47 (s, 9H), 1.57-1.69 (m, 4H),
2.22 (t, 2H) 2.28 (t, 2H), 3.35 (t, 2H), 3.44 (t, 2H), 3.54 (t, 2H), 3.58
(t, 2H), 3.62 (s, 4H), 4.25 (s, 2H), 4.43 (s, 2H), 7.44 (d, 2H), 7.85 (d,
2H); analytical HPLC tR 12.77 min (linear gradient, 100% A to 100%
B, over 20 min); MS (MALDI-TOF) m/z 618.27 [M + H]+, calculated
for C26H44N5O10S+ [M + H]+ 618.72.
Experimental Section
Materials. All chemicals, including bovine CA II (CA II, pI 5.9,
E.C. 4.2.1.1), were purchased from Sigma-Aldrich (St. Louis, MO)
unless stated otherwise. Mass spectra were performed by matrix-assisted
laser desorption/ionization mass spectrometry (MALDI-TOF) using
R-cyano-4-hydroxycinnamic acid as matrix. All analytical HPLC
separations were run with a Microsorb C18 column 5 µm (4.6 × 250
mm) using a linear gradient from A (100% water containing 0.1% TFA)
to B (100% acetonitrile containing 0.08% TFA), at a flow rate of 1.2
Alamethicin-Sulfonamide, 8. To a solution of alamethicin F30 (8
mg, 4.08 µmol) in dichloromethane, Dess-Martin periodinane (17.3 mg,
40.82 µmol) was added.118 The mixture was filtered and concentrated
to a solid by evaporation. The crude peptide derivative was purified
by preparative HPLC (linear gradient, 70% A with 30% B to 100% B,
over 40 min), and the solution evaporated to afford alamethicin aldehyde
7 (4 mg, 2.04 µmol) as a white solid. No impurities were detected by
(113) Uram, J. D.; Ke, K.; Hunt, A. J.; Mayer, M. Small 2006, 2, 967-972.
(114) Uram, J. D.; Ke, K.; Hunt, A. J.; Mayer, M. Angew. Chem., Int. Ed. 2006,
45, 2281-2285.
(116) Avila, L. Z.; Chu, Y. H.; Blossey, E. C.; Whitesides, G. M. J. Med. Chem.
1993, 36, 126-133.
(117) Roy, B. C.; Mallik, S. J. Org. Chem. 1999, 64, 2969-2974.
(118) Dess, D. B.; Martin, J. C. J. Org. Chem. 1983, 48, 4155-4156.
(115) Uram, J. D.; Mayer, M. Biosens. Bioelectron. 2007, 22, 1556-1560.
9
1464 J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008

Quantifying Protein−Ligand Interactions
A R T I C L E S
analytical HPLC (linear gradient, 100% A to 100% B, over 20 min).The
Boc-protected compound 4 (6.3 mg, 10.21 µmol) was dissolved in 10
mL of trifluoroacetic acid and stirred for 30 min at room temperature.
The solution was evaporated to dryness. The residue was dissolved in
methanol (1 mL) and 7 (4 mg, 2.04 µmol) was added. The reaction
was allowed to proceed for 2 h and then evaporated. The crude peptide
derivative was purified by HPLC (linear gradient, 70% A with 30% B
to 100% B, over 40 min) and evaporated to afford 8 (2.4 mg, 0.97
µmol, 24% yield) as a white solid. HPLC tR ) 14.20 and 14.42 min
(linear gradient, 70% A with 30% B to 100% B, 20 min); MS (MALDI-
TOF) m/z 2484.38 [M + Na]+, calculated for C113H181N27NaO32S+ [M
+ Na]+ 2484.86.
Molecular Modeling. We generated the energy minimized structure
shown in Figure 3B and C using molecular mechanics calculations
employing AMBER force field parameters in water (MacroModel
software, version 7.5, Schroedinger Inc.). We constructed 8 in silico
by modification of the C-terminus of the crystal structure of 6 (1AMT
from the Protein Data Bank) in MacroModel. For this calculation, the
residues of the alamethicin peptide (6) were fixed during the confor-
mational analysis. We allowed the ligand (5) to rotate freely during
these calculations. After performing 5000 iterations of conformational
analysis of 8, we selected the lowest energy conformation.
Planar lipid Bilayer Experiments. We used the same experimental
setup and procedure as described in detail by Mayer et al.49 Briefly,
we prepared Teflon AF films by molding a solution of 6% Teflon AF
in Fluorinert FC-75 solvent (DuPont Fluoroproducts, Wilmington, DE)
around a sharp tip (gold-plated tungsten probe tips with a nominal tip
diameter of 10 µm from Lucas Signatone, Gliroy, CA) that was oriented
perpendicular to, and placed in contact with, a silicon wafer such that
the tip rested on the wafer by gravity. Before casting the Teflon AF
solution, the wafer was treated with an air plasma followed by
silanization with tridecafluoro-1,1,2,2-tetrahydrooctyl-1-trichlorosilane
(United Chemical Technologies, Bristol, PA) under vacuum (60 mmHg)
to facilitate the removal of the film of Teflon AF from the wafer after
evaporation of the solvent).
Before using these Teflon AF sheets for planar lipid bilayer
experiments, we treated119 the area close to the pores with a solution
of 5% squalene (Sigma, St. Louis, MO) in pentane (Sigma-Aldrich,
St. Louis, MO) by dipping the tip of a tissue paper first into the squalene
solution and then moving it over the area close to the pore in the Teflon
film.120 We formed all bilayers by the “folding technique”49,121 in which
two monolayers of lipids at an air-electrolyte interface are raised such
that their acyl chains face each other and span over the aperture in the
Teflon AF support. The result of this procedure is a lipid bilayer that
spans the aperture in the Teflon AF film and separates the two
compartments of the lipid bilayer chamber (each compartment was filled
with 3.5 mL of aqueous electrolyte containing 1.0 M KCl).122 To prepare
the lipid monolayers, we spread a volume of 2-10 µL of a lipid solution
in pentane containing a 1:1 mixture of L-R-phosphatidylserine from
brain and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine, both
from Avanti Polar Lipids, Alabaster, AL with a concentration of ∼12
mg mL-1 of each lipid on the electrolyte solution in each chamber.
We raised and lowered the liquid level of one compartment with a 3
mL syringe until we measured an electrical resistance of g10 GΩ
between both compartments to verify that a bilayer without leak currents
was obtained. We recorded the current (and hence resistance) across
the bilayer membrane from Ag/AgCl electrodes (one in each electrolyte
compartment of the chamber) using an Axopatch 200 B amplifier from
Axon Instruments, Union City, CA connected to an acquisition board
with a sampling rate of 100 kHz (Digidata 1322A, Axon Instruments)
and a computer with Clampex software (Axon Instruments). To reduce
the electric current noise we used either the capacitive feedback
amplification (“patch mode” setting, gain â ) 1) or, typically, the
resistive feedback amplification with the 500 MΩ feedback resistor
(“whole cell mode” setting, gain â ) 1) in combination with the low-
pass filter of the amplifier with a cutoff frequency of 10 kHz. Before
adding alamethicin and its derivative to the bilayer, we confirmed that
the bilayers were stable (i.e., no detectable leak currents or increased
noise levels) for several minutes at transmembrane potentials up to
0.2 V. Both bilayer chambers could be stirred with stir bars (using a
Stir-2 stir plate from Warner Instruments) to ensure rapid mixing after
addition of molecules to the bilayer chambers. To reach equilibrium
conditions of binding, we waited at least 15 min (often more than 30
min) after addition of 8, CA II, or 9 before performing quantitative
recordings of current. We did not wash the recording chambers between
additions of molecule 8, CA II, or 9 in order to keep the total
concentration of all molecules well defined. Throughout the recordings
we monitored the capacitance of the bilayers using the built-in
capacitance compensation of the Axopatch 200 B amplifier.
Acknowledgment. We acknowledge the National Institute
of Health grants GM065364 and GM051559 for support. M.M.
acknowledges the Novartis Foundation for a postdoctoral
fellowship. This material is also based upon work supported
by a National Science Foundation CAREER Award (M.M.,
grant No. 0449088) and by a research grant from IMRA
America and AISIN U.S.A.
Supporting Information Available: Transformation of the
data from Figure 6 to construct a binding isotherm for the
interaction between 8 and CA II. We used the transformed data
(Table S1) and plotted a binding isotherm in Figure S1. A
nonlinear fit to this binding isotherm determined Kd8. This
material is available via the Internet at http://pubs.acs.org.
JA077555F
(119) Schlue, W. R.; Hanke, W. Planar Lipid Bilayers, 1st ed.; Academic
Press: London, 1993; pp 60-78.
(120) Tosteson, M. T.; Chow, M. J. Virol. 1997, 71, 507-511.
(121) Montal, M.; Mueller, P. Proc. Natl. Acad. Sci. U.S.A. 1972, 69, 3561-
3566.
(122) Alvarez, O. How to Set Up a Bilayer System. In Ion Channel Reconstitu-
tion, 1st ed.; Miller, C., Eds.; Plenum Press: New York, 1986; pp 115-
130.
9
J. AM. CHEM. SOC. VOL. 130, NO. 4, 2008 1465