Experimental evidence for the functional relevance of anion-π interactions

Article abstract of DOI:10.1038/nchem.657

Attractive in theory and confirmed to exist, anion-π interactions have never really been seen at work. To catch them in action, we prepared a collection of monomeric, cyclic and rod-shaped naphthalenediimide transporters. Their ability to exert anion-π interactions was demonstrated by electrospray tandem mass spectrometry in combination with theoretical calculations. To relate this structural evidence to transport activity in bilayer membranes, affinity and selectivity sequences were recorded. π-acidification and active-site decrowding increased binding, transport and chloride > bromide > iodide selectivity, and supramolecular organization inverted acetate nitrate to nitrate acetate selectivity. We conclude that anion-π interactions on monomeric surfaces are ideal for chloride recognition, whereas their supramolecular enhancement by π,π-interactions appears perfect to target nitrate. Chloride transporters are relevant to treat channelopathies, and nitrate sensors to monitor cellular signaling and cardiovascular diseases. A big impact on organocatalysis can be expected from the stabilization of anionic transition states on chiral π-acidic surfaces.

Full text of DOI:10.1038/nchem.657

ARTICLES
PUBLISHED ONLINE: 16 MAY 2010 | DOI: 10.1038/NCHEM.657
Experimental evidence for the functional relevance
of anion–p interactions
Ryan E. Dawson1†, Andreas Hennig1†, Dominik P. Weimann2, Daniel Emery1, Velayutham Ravikumar1,
Javier Montenegro1, Toshihide Takeuchi1, Sandro Gabutti3, Marcel Mayor3,4, Jiri Mareda1,
Christoph A. Schalley2 and Stefan Matile1
*
Attractive in theory and confirmed to exist, anion–p interactions have never really been seen at work. To catch them in
action, we prepared a collection of monomeric, cyclic and rod-shaped naphthalenediimide transporters. Their ability to
exert anion–p interactions was demonstrated by electrospray tandem mass spectrometry in combination with theoretical
calculations. To relate this structural evidence to transport activity in bilayer membranes, affinity and selectivity sequences
were recorded. p-acidification and active-site decrowding increased binding, transport and chloride > bromide > iodide
selectivity, and supramolecular organization inverted acetate > nitrate to nitrate > acetate selectivity. We conclude that
anion–p interactions on monomeric surfaces are ideal for chloride recognition, whereas their supramolecular enhancement
by p,p-interactions appears perfect to target nitrate. Chloride transporters are relevant to treat channelopathies, and
nitrate sensors to monitor cellular signaling and cardiovascular diseases. A big impact on organocatalysis can be expected
from the stabilization of anionic transition states on chiral p-acidic surfaces.
nion–p interactions take place between anions and p-acidic
aromatic surfaces with positive quadrupole moments1–11
They complement cation–p interactions between cations
Results
.
Design. To study anion–p interactions at work, we designed and
synthesized, as far as possible, monomeric, macrocyclic and rod-
shaped O-NDIs 2–11, where anion–p interactions were
systematically modulated and interferences from unrelated effects
vigorously excluded (Fig. 1a). The original amphiphile 1 was
composed of a transmembrane O-NDI tail for transport by anion–p
interactions and a trihistidine head for delivery to the membrane
(Fig. 1b)13. In amphiphile 2, the cationic trihistidine was replaced by
an anionic triglutamate to determine the contributions from the
peptide head to anion selectivity.
A
and p-basic aromatic surfaces with inverted, negative quadrupole
moments12. The experimentally observed contacts between
p-acidic aromatic rings and anions are the non-covalent anion–p
interaction, the strongly covalent or Meisenheimer complex, the
2
. . .
weakly covalent or charge-transfer interaction, and the C–H
X
hydrogen bond10. Contrary to the ubiquitous cation–p interactions,
p-acidic aromatic surfaces and their interaction with anions are
relatively rare and hard to access11. It is in part for this reason
that the functional relevance of anion-p interactions remains to
be shown.
To deconstruct rods 1 and 2 to extract anion–p interactions in
the purest possible form, the minimalist NDI 3 was defined as the
ideal starting point. The ortho-methyl groups of the peripheral
mesitylenes point into the anion–p binding site on the pyridine-
dione surface and thus contribute to active-site confinement and
Recently, the functional significance of anion–p interactions in
the context of anion transport across lipid bilayer membranes has
been reported13,14. Important with regard to basic science as well
as applications in organic, biological, medicinal and materials
chemistry, anion recognition and transport is otherwise achieved
today in elegant compositions of ion pairing, hydrogen bonding
and, very recently, also anion–macrodipole interactions15–20. To
use anion–p interactions for transport, three naphthalenediimides
(NDIs)21–25 were lined up in rigid-rod oligomer (O-NDI) 1
(Fig. 1). The obtained p-slide was thought to be long enough to
span a lipid bilayer membrane and allow anions to move along
the p-acidic surfaces (Fig. 1). However, although excellent activity
and anion selectivity were found, it was not possible to prove exper-
imentally that anion–p interactions that account for functional be-
haviour really exist; or, at best, the acquired evidence is indirect. For
example, contributions from amides or potentially charged peptide
chains at both termini of the slides to anion binding and transport
could never be fully ruled out13,14. Herein, unequivocal, clear-cut
experimental evidence for the functional relevance of anion–p
interactions is presented.
. . .
C–H anion bonding. Moreover, they should obstruct intermole-
cular p,p-interactions and thus interfere with self-organization into
supramolecular architectures. From there, active-site decrowding
and p-acidity are systematically maximized by o-methyl removal
in 4 and 6 and addition of cyano acceptors in the core of 5 and 6.
In dicyano NDIs, the lowest unoccupied molecular orbital drops
from –4.0 to –4.2 eV (ref. 22) and the quadrupole moment increases
from Qzz ≈ þ20 B to Qzz ≈ þ39 B (ref. 13); four cyano groups give
Qzz ≈ þ56 B (ref. 13). For comparison, hexafluorobenzene has
Qzz ¼ þ9.5 B and trinitrotoluene has Qzz ≈ þ20 B (ref. 1).
Active suprastructures in lipid bilayer membranes are intrin-
sically complex and dynamic, usually thermodynamically
unstable20,26, and include not only NDI–anion but also NDI–lipid
and NDI–NDI interactions (Fig. 1). As a simplified working
hypothesis, NDI monomers were assumed to align along the lipid
tails in one leaflet of the lipid bilayer. Preferred NDI–NDI over
NDI–lipid interactions gives single-leaflet NDI bundles, which
1Department of Organic Chemistry, University of Geneva, Geneva, Switzerland, 2Institut fu¨r Chemie und Biochemie der Freien Universita¨t, Berlin, Germany,
3Department of Chemistry, University of Basel, Basel, Switzerland, 4Institute of Nanotechnology, Karlsruhe Institute of Technology, Germany; †These authors
contributed equally to this work. e-mail: stefan.matile@unige.ch
*
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
533
© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.657
ARTICLES
nucleophilic substitution with CuCN gave 5. This approach to
dicyano NDIs was not applicable to NDI 6. The problem was that
with the more reactive aniline, core substitution of NDA 12
coincided with diimide formation, and all attempts to establish
chemoselectivity failed. However, results from anion transport
with 3–5 suggested that NDI 6 would have perfect properties.
These expectations justified an in depth search of new approaches
to 6. Best results were obtained when NDA 12 was reacted first
with ammonium acetate under acidic conditions to prevent core
substitution. The primary diimide 15 was then reacted with
CuCN to afford 16, where the peripheral phenyls could be added
a
R
O
O
O
N
N
3
O
O
O
O
N
N
4
O
O
O
O
N
N
5
O
O
O
O
N
N
6
O
O
O
O
N
N
7
O
O
O
NH
CN NC
NC
CN
CN
CN
NC
O
NC
HN
O
N
O
O
O
N
N
O
O
(Hypothetical molecule)
by
a transition-metal-catalysed C–N coupling with phenyl
+
–
O3S
H3N
boronate 17, and traces of the desired NDI 6 could be extracted
and purified from nearly intractable product mixtures.
O
NH
O
NH
N
Both the classical approach to 5 and the alternative approach to 6
were not applicable to the synthesis of tetracyano NDI 7. This per-
sistent failure in synthesis despite massive industrial interest25
suggested that tetracyano NDIs may be either too electron deficient
to exist or too intractable to be observed.
N
O
O
N
N
O
O
O
N
N
O
O
O
N O
O
O
N
N
O
O
O
N
N
O
O
O
O
N
N
O
O
O
O
N
O
O
ESI-FTICR-MS-MS experiments. The use of nuclear magnetic
resonance (NMR) spectroscopy and isothermal titration
calorimetry (ITC) to detect anion–p interactions was not
meaningful with NDI monomers because binding was too weak
and NMR shifts too small. In this demanding situation,
electrospray ionization Fourier-transform ion cyclotron resonance
tandem mass spectrometry (ESI-FTICR-MS-MS) experiments27
appeared ideal to deliver the desired direct experimental evidence
for anion–p interactions. Mass spectrometry offers the additional
advantage that non-solvated complexes are monitored, which can
be more easily compared to quantum chemical calculations than
the corresponding complexes in solution. Furthermore, the gas
phase likely resembles more closely the situation in the unpolar
core of the lipid bilayers used in the following than the situation
in competing polar solvents.
O
N
O
O
8
9
HN
1: R = -His-His-His-NH2
10
11
NHFmoc 2: R = -Glu-Glu-Glu-H
c
e
d
b
1, 2
3–6, 9
8
10, 11
Equimolar solutions of NDIs 3–6 and 8 and salts of different
anions were electrosprayed from acetonitrile under as mild as poss-
ible ionization conditions. Quite fragile NDI–anion complexes of
Figure 1 | Structures and hypothesized active suprastructures of anion–p
transporters. a, NDIs 1–11. b–e, These are thought to self-assemble into
transmembrane dimers of single-leaflet bundles (c) that are stabilized by
hydrophilic heads (b,e) and vertical (b) or horizontal (d) crosslinking.
2
3–6 were observed for Cl2, Br2 and NO3 (Supplementary
Fig. S4). With dimer 8, an additional series of anions was observed
2
to bind. Adduct detectability decreased with NO3 ≈ Cl2 ≈ Br2
could further dimerize into transmembrane dimers (Fig. 1c). From
there, we explore the impact of vertical crosslinking (Fig. 1b), hori-
zontal crosslinking24 (Fig. 1d), hydrophilic anchoring13 (Fig. 1b,e),
p-acidification and peripheral decrowding on the self-organization
of the active transmembrane dimers.
(1:1 and 2:1 adducts) . I2 ≈ H2PO4 ≈ OTf2 ≈ ClO4
(1:1
2
2
2
2
2
adducts) . OAc2 ≈ BF4 (weak 1:1 adducts) . PF6 ≈ BPh4
≈
22
2
SO4 ≈ MnO4
(not detected, Supplementary Table S1).
Although qualitative as long as no ESI response factors are
known27,28, the preference for chloride among halides and nitrate
Synthesis. The synthesis of most target molecules was accomplished among oxyanions was interesting (see below).
on the basis of or in analogy to reported procedures (Supplementary
The formation of dimers was used for competition experiments
Figs S1–S3). The synthesis of NDIs with cyano groups in the core with NDI monomers 3–6 and chloride as the preferred anion. An
required special attention (Fig. 2). Routine bromination of equimolar mixture of NDIs 3 and 4 was electrosprayed together
naphthalene dianhydride (NDA) with dibromoisocyanuric acid with one equivalent of NEt4Cl, and the corresponding heterodimer
followed by reaction of crude 12 with 13 gave NDI 14, and 3 þ 4 þ Cl2 was isolated (Fig. 3a). Fragmentation of heterodimer
H
N
H
N
O
O
N
N
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
(v)
Br
(iv)
(iii)
Br
(i)
(ii)
CN
Br
5
6
Br
NC
Br
Br
NH2
HO
OH
B
N
H
N
H
O
16
15
12
14
13
17
Figure 2 | Synthesis of dicyano NDIs. (i) Acetic acid (AcOH), 80 8C, 12 h, .28% (two steps); (ii) CuCN, N-methyl-2-pyrrolidinone (NMP), 100 8C, 5 h,
63%; (iii) NH4OAc, AcOH, reflux, 1 h, 77%; (iv) CuCN, NMP, 100 8C; (v) Cu(OAc)2, triethylamine, dimethylacetamide, 4.2% (two steps).
534
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.657
ARTICLES
3+4+Cl–
Charge-transfer complexes. Addition of tetrabutylammonium
iodide (TBAI) to NDI 3 in a 9:1 mixture of acetonitrile and
chloroform produced an absorption band at 400–650 nm. This
transition was assigned to a charge transfer between the p-acidic
NDI and the iodide anion (Fig. 3d). This assignment was
supported by a red shift from 470 to 510 nm with decreasing
solvent polarity (Supplementary Fig. S9) and a maximum that
was blueshifted compared with charge transfer with the more
easily reducible tetracyanopyrazine6. Rose–Drago analysis6 of the
dose response of NDI 3 to TBAI gave a dissociation constant
KD ≈ 5.9 M (Fig. 3e, open circles, assuming 1:1 stoichiometry6; see
Supplementary Information). Similar results were found for NDI
4 (Fig. 3e, filled circles). The redshifted charge transfer band at
lmax ≈ 520 nm observed with TBABr was in agreement with the
high reducibility of dicyano NDI 5 (Fig. 3e, squares).
a
b
c
4+Cl–
Computational chemistry. Electrostatic potential surfaces were
computed at the MP2/6-311 þþG**//PBE1PBE/6-311G** level
as described previously14. For NDI 6, they confirmed that the
pyridinedione heterocycles are the regions of highest electron
deficiency where anion binding is expected to occur (Fig. 4).
Depending on interference from peripheral o-methyls or
hydrogens, the anion was displaced from the pyridinedione
binding site. The weakening of anion–p interactions was reflected
in an increasing distance between the NDI plane and anion (Re
3+Cl–
200
400
600
m/z
800
1,000
d
e
0.04
0.02
0.0
distance, Table 1) and only partially compensated by stabilizing
0.1
0.05
0.0
2
X
. . .
C–H
interactions14 (see Supplementary Information).
The N-aryl planes were not perpendicular to the NDI planes;
dihedral angles found were between 868 to 628, a twist that can
0.0
0.5
c (M)
1.0
cause anions to move away from the arene centroid to maximize
2
X
. . .
C–H
interactions. Their main ipso–para axis was
occasionally bent out of the NDI plane for the same reason.
The anion binding energies computed at PBE1PBE/6-311++G**//
PBE1PBE/6-311G** level with BSSE (basis set superposition error)
correction14,29–31 increased with decreasing steric hindrances and
increasing p-acidity of the pyridinedione surface (Table 1). For
monomeric NDIs, binding energies increased with decreasing dis-
tance from the NDI surface and displacement from the pyridine-
dione binding site. The computed chloride . bromide . nitrate
affinity sequences in 1:1 complexes corresponded with results
from tandem mass spectrometry and transmembrane transport.
Anion binding by cyclophane 8 was stronger than with monomers
3 and 4 but weaker than with p-acidified monomers 5 and 6.
According to computational results, the properties of tetracyano
NDI 7 should further improve on those of dicyano NDI 6 with
regard to chloride recognition (Table 1, entry 5). However, these
predictions could not be confirmed because the synthesis of tetra-
400
500
600
700
λ (nm)
Figure 3 | Laser-induced ESI-MS-MS fragmentation of heterodimer
complexes and charge-transfer absorption. a–c, The spectra show the
isolated heterodimer 3 þ 4 with one chloride before (a) and after
fragmentation induced by a 100 ms (b) and a 200 ms laser pulse (c).
d, Changes in UV–vis absorbance A of a solution of 300 mM NDI 3 in
acetonitrile/chloroform 9:1 in the presence of 0.0, 0.25, 0.5, 0.75 and 1.0 M
TBAI (see open circles in e). e, Absorbance A at the maximum of the
charge transfer band as a function of TBAX concentration for 300
(open circles, X ¼ I), 300 M 4 (filled circles, X ¼ I) and 150 M 5
(squares, 150
M, X ¼ Br).
mM 3
m
m
m
3 þ 4 þ Cl2 was induced by irradiation with a 25 W infrared laser. cyano NDI 7 was unsuccessful.
After 100 ms irradiation, Cl2 started to appear together with mono-
meric NDI 4 alone (Fig. 3b). After 200 ms irradiation, the peak for Anion transport. The transport activity of NDIs 2–11 was
3 þ 4 þ Cl2 nearly disappeared and a new peak for 3 þ Cl2 determined in large unilamellar vesicles (LUVs) composed of egg
appeared beside the dominant peak for 4 þ Cl2 (Fig. 3c). These yolk phosphatidylcholine (EYPC) using different fluorescent
observations demonstrated that NDI 4 binds chloride better than probes26. Assays of the fluorophores 8-hydroxy-1,3,6-
NDI 3 (Table 1, entry 1 versus 2). The same tandem ESI experiment pyrenetrisulfonate (HPTS), 5(6)-carboxyfluorescein (CF) and
revealed that chloride prefers NDI 5 over NDI 4 and NDI 6 over lucigenin (LG) were prepared. Vesicles loaded with the pH-
NDI 5 (Supplementary Figs S6, S7, Table 1, entries 1–4). The sensitive fluorophore HPTS were exposed to a transmembrane pH
found selectivity sequence 6 . 5 . 4 . 3 demonstrated increasing gradient13,14,20,26. The dissipation of this pH gradient reported by
anion affinity with increasing p-acidity and decrowding of the EYPC LUVs . HPTS can be attributed to Hþ/Mþ antiport,
anion–p binding site (see below). In contrast to the more qualitative OH2/A2 antiport, Hþ/A2 symport, OH2/Mþ symport, HPTS
results from routine ESI-MS measurements, intensities found in efflux or vesicle destruction. In the CF assay, the vesicles were
ESI-FTICR-MS-MS relate to true gas phase experiments and can loaded with CF at concentrations high enough to assure self-
be used to determine anion affinity sequences quantitatively28. quenching26. EYPC LUVs . CF report CF efflux or vesicle
These results from ESI-FTICR-MS-MS were thus of the highest destruction as fluorescence recovery. Vesicles loaded with the
importance because they provide direct experimental evidence for chloride-sensitive fluorophore LG in nitrate buffer were exposed
anion binding to minimalist NDI models that contain not much to external chloride32. EYPC LUVs . LG selectively report
else to bind to than a p-acidic surface.
chloride influx or vesicle destruction as fluorescence quenching.
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
535
© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.657
Table 1 | Summary of anion binding data for NDIs 2–11 from MS-MSa, molecular modellingb,c and transmembrane transportd–i
ARTICLES
.
CI2
int
Br2
int
NO3
int
2
Entry NDIj CI2
(MS)a
E
E
E
Re (Å)c
CI2/I2d CI2/Br2e NO32/AcO2f Na1/ EC50
EC50
(CF)
(m
b
b
b
(kJ mol21
)
(kJ mol21
)
(kJ mol21
)
K1
g
(HPTS)
M)h
(
m
M)i
1
3
4
5
6
7
2
8
9
10
11
þ
283.3
292.1
2136.4
272.0
279.9
2123.1
2129.8
2173.7
269.9
269.5
2110.9
2110.1
2143.6
2.86 (2.97)
2.85 (2.87)
2.70 (2.86)
2.63 (2.79)
2.50 (2.74)
2
1.0
1.1
1.1
2
1.0
1.0
1.0
1.0
150+20
27+1
.100
17+2
10+1
2
3
4
5
6
7
8
9
10
þþ
2.2
2.4
2.1
0.7
0.8
6.0
þþþ
37+2
þþþþ 2142.7
2188.8
1.5
0.33+0.03k .100
1.8
2
1.6
9.0
1.1
1.4
1.2
1.0
1.3
1.1
2.8
0.7
0.8
0.4
1.0
1.0
1.0
1.0
1.3
1.0
22+2
110+20
32+1
7.8+0.7
8.7+1.4
95+8
2
8.0+0.7
4.5+0.1
8.2+0.2
þþþ
2121.0
104.6
295.4
3.02 (2.93)l
aRelative selectivity sequence for chloride binding from tandem MS experiments with mass-selected heterodimer/Cl2 complexes, see Fig. 3 for original data; bInteraction energy in monomeric NDI anion
complexes computed with the PBE1PBE/6-311 þþG**//PBE1PBE/6-311G** method and BSSE correction (see Fig. 4 for structures of 6); cEquilibrium distance between chloride (nitrate) and NDI plane in
monomeric complexes; dRelative NDI activity in EYPC LUVs obtained in the HPTS assay with different extravesicular ions, for example: dExternal chloride (NaCl) compared with external iodide (NaI);
f
eExternal chloride (NaCl) compared with external bromide (NaBr); External nitrate (NaNO3) compared with external acetate (NaOAc); gExternal sodium (NaCl) compared with external potassium (KCl,
100 mM MX, all at constant 100 mM internal NaCl), see Fig. 5 and Supplementary Figs S11, S14–S24; hEffective NDI concentration needed to reach 50% activity in the HPTS assay, data from Hill
i
analysis of dose response curves (Supplementary Fig. S11); Effective NDI concentration needed to reach 50% activity in the CF assay, data from Hill analysis of dose response curves (Supplementary
Fig. S12); jSee Fig. 1; kLucigenin assay: EC50 ¼ 0.75+0.01
m
M (Supplementary Fig. S13); lAverage distance from the two NDI planes of cyclophane 8.
Results from HPTS, CF and LG assays were quantified with EC50 binding to the transporter to overcompensate the cost of at least
and Hill (n) coefficients, which were both obtained from dose partial dehydration13,26, although other explanations are possible.
response curves13,20,26. EC50 is the ‘effective’ monomer concentration These results are compatible with transmembrane dimers of
needed to reach 50% activity. These values are always overestimates single-leaflet NDI bundles as plausible active suprastructures
because delivery, partitioning, self-organization and self-assembly (Fig. 1c), ordered enough to generate significant anion selectivity
are never quantitative. EC50 (HPTS) , EC50 (CF) demonstrates and dynamic enough to let CF pass as well.
that ion transport is more efficient than dye export or vesicle
Improving activity with decrowding in 4 and p-acidification in 5
destruction; EC50 (HPTS) , EC50 (LG, CF) can support cation suggested that the decrowded and p-acidic NDI 6 should be perfect.
selectivity; and EC50 (HPTS, LG) , EC50 (CF) demonstrates chlor- Because of this perspective, difficulties to synthesize 6 were tackled
ide selectivity, and so on.
and overcome (Fig. 2). This effort was worthwhile as NDI 6 exhib-
NDI 3 was essentially inactive in the HPTS and CF assays; ited nanomolar activity in both HPTS and lucigenin assays
decrowded and p-acidified NDIs 4 and 5 were clearly more active (Supplementary Figs S13, S18), whereas non-specific activity in
(Table 1, entries 1–3, Supplementary Figs S11, S12). Sensitivity the CF assay was not detectable (Table 1, entry 4). Cations remained
towards external anion but not cation exchange implied anion- uninvolved in the transport, Cl2/Br2 discrimination improved to
selective transport (Fig. 5a,b). The anion selectivity sequence of 1.5 without significant losses in Cl2/I2 selectivity, and a spectacular
NDIs 4 and 5 showed increasing activity with increasing hydration NO32/AcO2 selectivity of 6.0 appeared (Table 1, entry 4, Fig. 5c).
energy (Fig. 5b, Supplementary Figs S16, S17). This so-called anti-
The sequence NO32 . ClO42 . SO422 . AcO2 (obtained from
Hofmeister behaviour can be interpreted as support for strong the difference to a nearly linear anti-Hofmeister Cl2 . Br2 . I2
correlation, Fig. 5b) was compatible with operational p,p-inter-
actions in p–anion–p complexes, selecting for planar oxyanions
with many p-bonds and rejecting tetrahedral a-carbons (Fig. 5c).
The emergence of nitrate selectivity together with CF exclusion
suggested that high self-organization of the active suprastructure
Cl–
NO3
–
is essential for p,p-enhanced anion–p interactions to occur. This
interpretation was supported by the poor sensitivity of Cl2
.
Br2 . I2 sequences of NDIs 4–6 to supramolecular organization
(fluoride exclusion with NDI 6 was attributed to Meisenheimer
complexes; see Supplementary Information). Moreover, stabiliz-
ation of the active dimeric bundles by vertical crosslinking in
O-NDI rod 2 resulted in an intermediate NO32/AcO2 selectivity
of 2.8 at intermediate EC50 (HPTS) , EC50 (CF) (Table 1, entry 6;
Figs 1a and 5d). Formal horizontal crosslinking with O-NDI cyclo-
phane 8 caused the corresponding loss in activity (Table 1, entry 7;
Supplementary Fig. S19; Fig. 1d). This result further demonstrated
that NDIs do not act as anion carriers, a conclusion that was fully
supported by inactivity in U-tube experiments.
6
1.93
1.36
3.01
2.84
2.79
2.63
The addition of negative charges in NDI 10 gave outstanding
halide selectivity without oxyanion recognition or CF exclusion
(Table 1, entry 9; Figs 5e and 1d). This suggested that charge repul-
sion at the termini would loosen the active suprastructure, minimize
the p,p-enhanced anion–p interactions accounting for nitrate
selectivity but provide free access to anion–p binding sites on
monomer surfaces, accounting for chloride selectivity. The comp-
lementary addition of positive charges in NDI 11 deleted all mean-
ingful selectivity (Table 1, entry 10; Fig. 5f). Selectivity inversion to a
–142.7 kJ mol–1
–110.1 kJ mol–1
Figure 4 | Molecular modelling of anion–p interactions. Electrostatic
potential surface (blue positive, red negative, +150 kJ mol21, MP2/
6-311 þþG**//PBE1PBE/6-311G**) and density functional theory optimized
structures of 1:1 chloride and nitrate complexes of NDI 6 (C, cyan; H, white;
O, red; N, blue; Cl2, green) with indication of anion location (Å) and
interaction energies.
536
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.657
ARTICLES
Comparison of binding with transport data reveals identical
a
2– F–
b
Cl–
0.5
0.4
0.3
0.2
0.1
0
SO4
OAc–
trends. p-acidity and active-site decrowding are found to govern
K+
Rb+
Na+
0.5
0.4
0.3
0.2
0.1
0
halide selectivity (Cl2 . Br2 . I2), whereas supramolecular organ-
Br–
NO3
Li+
2
ization accounts for oxyanion selectivity (NO32 . ClO4
.
Cs+
–
SO422 . AcO2 obtained from the difference to a nearly linear
anti-Hofmeister Cl2 . Br2 . I2 background). This change
occurs because the p,p-enhanced anion–p interactions thought to
account for nitrate recognition are amplified in sandwich com-
plexes, whereas normal anion–p interactions accounting for chlor-
ide recognition are not. Decreasing functional relevance with more
easily oxidizable anions suggested that charge-transfer contributions
to anion–p interactions are less relevant for function.
I–
–
ClO4
–500 –400
–300
–500 –400
ΔGhyd (kJ mol–1
–300
ΔGhyd (kJ mol–1
)
)
The perfect correspondence of transport data with affinity and
selectivity sequences determined for anion–p interactions by
tandem mass spectrometry and computational simulation demon-
strates that anion–p interactions account for function. Particularly
promising is the possibility to gain control of the nitrate recognition
by p,p-enhanced anion–p interactions on the supramolecular level
and chloride recognition by active-site engineering on the molecular
level. Modular access to tunable chloride, carbonate and nitrate
sensors and transporters is of interest in biology and medicine.
Channel replacement therapy has been proposed to rectify chloride
c
0.8
0.7
0.6
0.5
0.4
0.3
0.2
d
0.7
0.6
0.5
0.4
0.3
0.2
0.1
–
2–
2–
NO3
SO4
SO4
Cl–
Cl–
–
–
ClO4
NO3
Br–
I–
Br–
–
l–
–300
ClO4
OAc–
F–
OAc–
F–
channel malfunction in channelopathies such as cystic fibrosis15–18
,
and carbonate transport and sensing is important in cellular respir-
ation, cellular signalling, homeostasis and channelopathies19. The
biological and therapeutic importance of nitrate and nitrite
include cardiovascular diseases, global NO signalling, regulation
of mitochondrial function, vasodilator action and tissue protec-
tion33. Nitrate, nitrite and NO research suffers from the unavailabil-
ity of fluorescent nitrate/nitrite sensors that work in vivo33. p-p-
enhanced anion–p interactions combined with core-substituted
NDI fluorophores of different colours34 offers an approach to the
rational design of nitrate sensors. With regard to broader perspec-
tives, our results on molecular recognition and translocation
invite a shift of attention towards molecular transformation.
Complementary to the central importance of cation–p interactions
in carbocation chemistry12, anion–p interactions have the potential
to catalyse enolate chemistry in a similarly fundamental manner.
–500
–400
–500
–400
–300
ΔGhyd (kJ mol–1
)
ΔGhyd (kJ mol–1
)
e
0.6
f
0.6
0.5
0.4
0.3
0.2
0.1
0
F–
OAc–
Cl–
2–
Cl–
SO4
0.5
0.4
0.3
0.2
I–
Br–
OAc–
–
Br–
–
ClO4
–
F–
NO3
2–
NO3
SO4
0.1
0
I–
–
ClO4
–300
ΔGhyd (kJ mol–1
–500 –400
–500 –400
–300
)
ΔGhyd (kJ mol–1
)
Methods
Figure 5 | Transport selectivity. a–f, Dependence of the fractional transport
activity Y of NDIs 4 (a,b), 6 (c), 2 (d), 10 (e) and 11 (f) in the HPTS assay
on the nature of extravesicular cations (a; 100 mM MCl, inside 100 mM
NaCl) and anions (b–f; 100 mM NaX, inside 100 mM NaCl), see
Supplementary Information for details.
The (tandem) mass spectrometric experiments described herein were conducted
with an Ionspec QFT-7 FTICR mass spectrometer (Varian), equipped with a 7 T
superconducting magnet and a Micromass Z-Spray electrospray ionization source
(Waters). The samples were introduced into the source as 50-mM solutions of NDI
and the tetrabutyl ammonium salt of the corresponding anion in acetonitrile at flow
rates of 1–2 ml min21. A constant spray and highest intensities were achieved with a
capillary voltage of 3,000–3,800 V (depending on the used NDI and anion) and at a
very weak Hofmeister series testified for destructive interference source temperature of 40 8C. The parameters for the sample cone (20–45 V) and
extractor cone (8–10 V) voltages were optimized for maximum intensities of the
desired complexes. Multiple scans (20–50) were recorded and averaged for each
spectrum to improve the signal-to-noise ratio. After accumulation and transfer into
from non-specific anion binding to the ammonium cations.
Discussion
the instrument’s FTICR analyser cell, the ions were detected by a standard excitation
The objective of this study was to catch anion–p interactions at
work. This evidence is secured from a collection of corroborative
data on anion recognition and translocation from infrared multi-
photon dissociation (IRMPD) MS-MS experiments, charge-transfer
absorption, molecular modelling and transport in bilayer mem-
branes. Ultimate evidence, as pure as possible, for the existence of
anion–p interactions as such is obtained by (tandem) mass spectro-
metric experiments with NDI models where only the p-acidic
surface is left for anions to interact with (Fig. 3). To connect data
from different methods, series of binding and selectivity sequences
are recorded and compared. MS sequences demonstrate preferences
for (1) chloride among halides, (2) nitrate among oxyanions, and (3)
p-acidic, (4) decrowded and (5) dimeric binding sites. Molecular
modelling confirms trends concerning p-acidity and active-site
decrowding; models of 1:1 complexes confirm the expected
chloride . bromide . nitrate selectivity.
and detection sequence. For the fragmentation experiments, the ions of interest were
mass selected in the ICR cell and irradiated with a 25 W CO2 laser in the infrared
region (infrared multiphoton dissociation, 10.6 mm wavelength) to
induce fragmentation.
Selectivity sequences for anion transport were determined with the HPTS
assay26. In brief, 25 ml EYPC LUVs . HPTS were added to 1,950 ml of gently stirred
thermostated buffer (10 mM Hepes, 100 mM MCl (Mþ ¼ Liþ, Naþ, Kþ, Rbþ, Csþ)
or 100 mM NaX (X2 ¼ F2, SO422 (66 mM), OAc2, Cl2, NO32, Br2, I2, SCN2,
ClO42), pH 7.0) in a disposable plastic cuvette. The time-dependent change in
fluorescence intensity (lem ¼ 510 nm) was monitored ratiometrically (lex
¼
450 nm, lex ¼ 405 nm) during the addition of base (20 ml 0.5 M NaOH), NDI
(various concentrations in various solvents; see Supplementary Information), and
gramicidin A (100 mM, 20 ml DMSO) for calibration. Fluorescence kinetics were
normalized to fractional activities Y and plotted against hydration energies or as
dose response curves for Hill analysis to extract EC50 and n (Table 1 and
Supplementary Table S3).
Full experimental details on synthesis, MS-MS anion binding studies, charge-
transfer absorption, molecular modelling and HPTS, CF and LG transport assays
can be found in the Supplementary Information.
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
537
© 2010 Macmillan Publishers Limited. All rights reserved.

NATURE CHEMISTRY DOI: 10.1038/NCHEM.657
ARTICLES
21. Bhosale, S. V., Jani, C. H. & Langford, S. J. Chemistry of naphthalene diimides.
Chem. Soc. Rev. 37, 331–342 (2008).
Received 4 December 2009; accepted 30 March 2010;
published online 16 May 2010
22. Jones, R. A., Facchetti, A., Wasielewski, M. R. & Marks, T. J. Tuning orbital
energetics in arylene diimide semiconductors. Materials design for ambient
stability of n-type charge transport. J. Am. Chem. Soc. 129, 15259–15278 (2007).
23. Bhosale, S. et al. Photoproduction of proton gradients with p-stacked
fluorophore scaffolds in lipid bilayers. Science 313, 84–86 (2006).
24. Gabutti, S. et al. A rigid sublimable naphthalenediimide cyclophane as
model compound for UHV STM experiments. Chem. Commun.
2370–2372 (2008).
25. Ko¨nemann, M. Naphthalenetetracarboxylic acid derivatives and their use as
semiconductors. WO/2007/074137.
26. Matile, S. & Sakai, N. in Analytical Methods in Supramolecular Chemistry (ed.
Schalley, C. A.) 391–418 (Wiley, 2007).
27. Kogej, M. & Schalley, C. A. in Analytical Methods in Supramolecular Chemistry
(ed. Schalley, C. A.) 104–162 (Wiley, 2007).
28. Cooks, R. G., Patrick, J. S., Kotiaho, T. & McLuckey, S. A. Thermochemical
determinations by the kinetic method. Mass Spectrom. Rev. 13, 287–339 (1994).
29. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation
made simple. Phys. Rev. Lett. 77, 3865–3868 (1996).
30. Ernzerhof, M. & Scuseria, G. E. Assessment of the Perdew-Burke-Ernzenhof
exchange-correlation functional. J. Chem. Phys. 110, 5029–5036 (1999).
31. Frisch, M. J. et al. Gaussian 03, Revision C.02 (Gaussian, 2004).
32. McNally, B. A., Koulov, A. V., Smith, B. D., Joos, J-B. & Davis, A. P. A fluorescent
assay for chloride transport; identification of a synthetic anionophore with
improved activity. Chem. Commun. 1087–1089 (2005).
33. Butler, A. R. & Feelisch, M. Therapeutic uses of inorganic nitrite and nitrate -
from the past to the future. Circulation 117, 2151–2159 (2008).
34. Kishore, R. S. K. et al. Ordered and oriented supramolecular n/p-heterojunction
surface architectures: Completion of the primary color collection.
J. Am. Chem. Soc. 131, 11106–11116 (2009).
References
1. Schottel, B. L., Chifotides, H. T. & Dunbar, K. R. Anion-p interactions.
Chem. Soc. Rev. 37, 68–83 (2008).
2. Gamez, P., Mooibroek, T. J., Teat, J. S. & Reedijk, J. Anion binding involving
p-acidic heteroaromatic rings. Acc. Chem. Res. 40, 435–444 (2007).
3. Quinonero, D. et al. Anion-p interactions: Do they exist? Angew. Chem. Int. Ed.
41, 3389–3392 (2002).
4. Mascal, M., Armstrong, A. & Bartberger, M. D. Anion-aromatic bonding:
A case for anion recognition by p-acidic rings. J. Am. Chem. Soc. 124,
6274–6276 (2002).
5. Alkorta, I., Rozas, I. & Elguero, J. Interaction of anions with perfluoro aromatic
compounds. J. Am. Chem. Soc. 124, 8593–8598 (2002).
6. Rosokha, Y. S., Lindeman, S. V., Rosokha, S. V. & Kochi, J. K. Halide recognition
through diagnostic anion-p interactions: Molecular complexes of Cl2, Br2,
and I2 with olefinic and aromatic receptors. Angew. Chem. Int. Ed. 43,
4650–4652 (2004).
´
´
7. Gil-Ramırez, G., Escudero-Adan, E. C., Benet-Buchholz, J. & Ballester, P.
Quantitative evaluation of anion-p interactions in solution.
Angew. Chem. Int. Ed. 47, 4114–4118 (2008).
8. Zaccheddu, M., Filippi, C. & Buda, F. Anion-p and p-p cooperative
interactions regulating the self-assembly of nitrate-triazine-triazine complexes.
J. Phys. Chem. A 112, 1627–1632 (2008).
9. Berryman, O. B., Bryantsev, V. S., Stay, D. P., Johnson, D. W. & Hay, B. P.
Structural criteria for the design of anion receptors: The interaction of halides
with electron-deficient arenes. J. Am. Chem. Soc. 129, 48–58 (2007).
10. Hay, B. P. & Bryantsev, V. S. Anion-arene adducts: C-H hydrogen bonding,
anion-p interaction, and carbon bonding motifs. Chem. Commun.
2417–2428 (2008).
11. Hay, B. P. & Custelcean, R. Anion-p interactions in crystal structures:
Commonplace or extraordinary? Cryst. Growth Design 9, 2539–2545 (2009).
12. Ma, J. C. & Dougherty, D. A. The cation-p interaction. Chem. Rev. 97,
1303–1324 (1997).
13. Mareda, J. & Matile, S. Anion-p slides for transmembrane transport.
Chem. Eur. J. 15, 28–37 (2009).
14. Gorteau, V., Bollot, G., Mareda, J. & Matile, S. Rigid-rod anion-p slides for
multiion hopping across lipid bilayers. Org. Biomol. Chem. 5, 3000–3012 (2007).
15. Davis, A. P., Sheppard, D. N. & Smith, B. D. Development of synthetic
membrane transporters for anions. Chem. Soc. Rev. 36, 348–357 (2007).
16. Gokel, G. W. & Barkey, N. Transport of chloride ion through phospholipid
bilayers mediated by synthetic ionophores. New J. Chem. 33, 947–963 (2009).
17. Li, X., Shen, B., Yao, X. Q. & Yang, D. Synthetic chloride channel regulates cell
membrane potentials and voltage-gated calcium channels. J. Am. Chem. Soc.
131, 13676–13680 (2009).
Acknowledgements
We thank D.-H. Tran, J. Praz and L. Maffiolo for contributions to synthesis, D. Jeannerat,
A. Pinto and S. Grass for NMR measurements, P. Perrottet, N. Oudry and G. Hopfgartner
for mass spectrometry, the Swiss National Supercomputing Center in Manno for CPU time,
and the University of Geneva (S.M., J.M.), the University of Basel (M.M.), the Deutsche
Forschungsgemeinschaft (C.A.S.), the Fonds der Chemischen Industrie (C.A.S.), the
NCCR Nanoscale Sciences of the SNF (S.G., M.M.) and the Swiss NSF (S.M., M.M.) for
financial support.
Author contributions
R.E.D., V.R., J.M., T.T., S.G. and A.H. synthesized compounds, A.H. and R.E.D. transported
ions across membranes, D.P.W. performed the (tandem) mass spectrometric experiments
and evaluated the mass spectrometry data, D.E. prepared computational models, M.M.,
J.M., C.A.S. and S.M. directed the study, contributing to design, execution and
interpretation of experiments, computer modelling and manuscript writing.
18. Deng, G., Dewa, T. & Regen, S. L. A synthetic ionophore that recognizes
negatively charged phospholipid membranes. J. Am. Chem. Soc. 118,
8975–8976 (1996).
19. Davis, J. T. et al. Using “small” molecules to facilitate exchange of bicarbonate
and chloride anions across liposomal membranes. Nature Chem. 1, 138–144 (2009).
20. Hennig, A., Fischer, L., Guichard, G. & Matile, S. Anion-macrodipole
interactions: self-assembling, macrocyclic oligourea/amide anion transporters
that respond to membrane polarization. J. Am. Chem. Soc. 131,
16889–16895 (2009).
Additional information
The authors declare no competing financial interests. Supplementary information and
chemical compound information accompany this paper at www.nature.com/
naturechemistry. Reprints and permission information is available online at http://npg.nature.
com/reprintsandpermissions/. Correspondence and requests for materials should be
addressed to S.M.
538
NATURE CHEMISTRY | VOL 2 | JULY 2010 | www.nature.com/naturechemistry
© 2010 Macmillan Publishers Limited. All rights reserved.