ARTICLES

Mutations in the channel domain of a neuronal

nicotinicreceptor convert ion selectivity
from cationic to anionic
Jean-Luc Galzi*, Anne Devillers-Thiéry*, Nicolas Hussy', Sonia Bertrand',
Jean-Pierre Changeux* & Daniel Bertrand'
* Neurobiologie
Moléculaire, Unité de Recherche Associée au Centre National de la Recherche Scientifique D1284, lnstitut Pasteur, 25 rue du Dr Roux,
75724 Paris Cedex 15, France
t Département de Physiologie, Centre Médica! Universitaire (Faculté de Médecine), 1211 Geneva 4, Switzerland

lntroduction by site-directed mutagenesis of three amino acids from the Mii segment of glycine or
y-aminobutyric acid (GABAA) receptors into the Mii segment of a7 nicotinic receptor was sufflcient
to convert a cation-selective channel into an anion-selective channel gated by acetylcholine. A critica!
mutation was the insertion of an uncharged residue at the amino-terminal end of Mii, stressing the
importance of protein geometrical constraints on ion selectivity.

NEUROTRANSMITIER-GATED ion channels such as the
nicotinic acetylcholine, glycine, GABAA or 5-hydroxytrypt-
amine (5HT3) receptors are transmembrane allosteric proteins
made up of homologous subunits pseudo-symmetrically
arranged around an axial channel l". The hydropathy profile of
the subunits reveals four potential membrane-spanning hydro-
phobic segments, named MI to MIV, and convergent photo-
affinity labelling"" and mutagenesis experíments'':" with
nicotinic acetylcholine receptors (nAChRs) point to a contribu-
tion of Mii to the walls ofthe ion channel", Chemical kinetics 1º·11
and electrophysíological'<':' experiments show that this receptor
protein may exist in severa! interconvertible states (resting,
conducting and desensitized) which differ by their ligand-
binding and channel-opening properties. In the homo-
oligomeric nAChR composed of the a7 subunit of chick
braín'":", the mutation of a single hydrophobic residue within
Mii (Leu 247 to Thr) causes noticeable modifications of the
response to acetylcholine (ACh), which may represent the con-
version of a high-affinity conformation from a closed desensit-
ized state into a conducting state'?'!". These results suggest that
single amino-acid substitutions may differentially affect the
properties of the ion channel in the various conformations of
the receptor protein.
The anion-selective18 glycine and GABAA receptors'r' display
homologies with the cation-selective nAChRs. In particular, in
Mii (Fig. 1) the following amino acids, which contribute to ion
permeation in the nAChR, are conserved: the cytoplasmic ring
of negatively charged amino acids (position 234 in a7 receptor"),
the two polar rings of' threonines or serines (position 244 in a7,
refs 5, 6, 19, 20, and position 248 in a7, ref. 21) and the
hydrophobic ring ofleucines (position 247 in a7; refs 7, 16, 17,
22). These sequence homologies and the finding that the glycine
receptor (GlyR) Mii peptides form ion channels in mem-
branes/", indicated that Mii participates in anion permeation
in the glycine and GABAA receptors'v', Yet the N-terminal end
of Mii in ali these anionic channels includes an additional
amino acid located in the short segment linking MI to Mii
(between positions 236 and 237 in a7), which could contribute
to the difference of ion selectivity between ACh and
GABAA/ glycine receptors.
Here we show that combined substitutions and/ or addition
of amino acids within (or near) the Mii segment from the
nicotinic a7 receptor with homologous residues from the GlyR
a 1 subunit24 suffice to convert a] ion-channel selectivity from
cationic to anionic. Also, sorne of the arnino-acid permutations
alter monovalent versus divalent cation selectivity, as similarly
500
© 1992 Nature Publishing Group
described for glutamate receptors25-27• Yet in addition to the
alteration in ion selectivity, important changes occur in agon-
ist/ competitive antagonist sensitivity and desensitization
properties of the response to ACh.
Neuronal a.7 channel is permeable to cations
Severa! neuronal nAChRs are permeable to calcium28-30• When
these receptors are expressed in Xenopus oocytes, the calcium
flux through their channel leads to activation of a Ca2+-depen-
dent chloride conductance v"", sensitive to calcium che-
lators ". This chloride current was used here to detect calcium
permeability through nAChRs. Therefore, the chloride contribu-
tion to ACh-evoked ionic responses was determined, at various
potentials, as the difference between currents measured before
and after injection of the calcium chelator BAPTA33 in oocytes
expressing a7 wild-type (WT) or mutant receptors.
When expressed in Xenopus oocytes, a7 WT receptors elicited
large ACh-evoked inward currents reversing at -15.4±4 mV
(Fig. 2a). Substitution of 90% externa! chloride by the large
anion isethionate shifted the reversa! potential close to O mV
(Erev = 1.8 ± 3.7 mV; Table 1), providing evidence for a contribu-
tion of chloride ions to the current. After injection of BAPT A,
ACh-activated currents reversed at a poten ti al (3.8 ± 4.8 m V,
Table 1) similar to that determined on substitution of externa!
chloride, indicating that the current carried by chloride ions,
which represented up to 50% of the ionic response at -40 mV
(Fig. 2b ), had been significantly reduced by chelation of interna!
calcium. Indeed, at -15 m V, the slow onset of chloride in flux
(outward current), which follows an inward cationic current,
was no longer observed after injection of BAPTA (Fig. 2d).
These data thus support the conclusion that a7 WT receptor,
like other neuronal nAChRs28-30, is selective for cations and
permeable to calcium.
Conversion from cationic to anionic selectivity
In a first attempt to convert ion-channel selectivity, we construc-
ted chimaeric complementary DNAs encoding nicotinic a7
receptors in which the segment linking MI to Mii and also half
or ali of the Mii segment (from amino acid a7-234 to 245 or
from 234 to 258) was exchanged with that of the glycine a 1
subunit. No detectable current in response to ACh application
was recorded in oocytes injected with these chimaeric cDNAs.
In a second step, instead of substitution in a block, amino acids
presumed to face the lumen of the ion channel in the GlyR,
assuming that Mii is organized asan a-helix2•6-9•21•22•34•35, were
introduced into a7 nAChR.
NATURE · VOL 359 · 8 OCTOBER 1992
 ARTICLES
TABLE 1 Sequence and properties of the Mii mutants

Selectivlty
Amplitude µA EC50 EC50 Reversa! potenttat mV Cationic/
Type Sequence Mean value A Ch nH DH¡JE nH Control Cont/BAPTA lset lset/BAPTA Anionic
236 237 251 258
a7-WT DSG-EKIS LGITVLLSLT VFM LLVAE 0.63(n~98) 115.00 1.4 C.A -15.4±4.0 3.8±4.8 1.8±3.7 ND e
a7-1 - - - PA- - G - - - - - - - - - - T- - SG - - N 0.21 (n~75) 0.24 1.6 0.30 1.4 -19.0±4.6 -19.0±4.9 27.0±7.6 29.3 ± 3.0 (al A
-23.2±3.4 -18.6±4.8 24.3±8.6 25.2± 6.6 (b)
-22.4±3.1 24.4±9.8 (el
a7-2 - - - PA- - - - - - - - - - - - - T 0.13(n~23I 0.23 1.8 0.94 1.4 -25.8± 5.4 ND 15.3±8.8 12.5±3.0 A
a7-3 - - - PA- - - - - - -- - - - - -
a7-4 - - - - - ---- - T 5.50(n~381 0.63 2.0 9.27 1.4 -17.7 ±6.6 -4.2±5.6 36.3±7.5 2.2±5.7 e
a7-5 - - - •A- - - - - - - - - - - - - - - - - - - - - 0.13(n~231 194.00 1.4 ND ND -14.2± 2.1 -13.5± 1.7 ND ND e
a7-6 - - • A- - - - - - - - - - - - - T 2.90 ln.~101 0.98 2.0 3.69 1.4 -18.5± 1.3 -18.5±3.9 -18.5± 1.3 -19.0±3.4 e
a7-7 - - - p - - - - ----- - - - - - 0.02 n~lOI ND ND ND ND ND ND NO ND ND
a7-8 - - - • A- -G - ---- - -- - - T - -SG- -N 4.27 (n~231 5.45 2.8 ND ND -10.9±2.8 -12.2±2.9 -11.5±3.8 ND e
a7-9 - - - AA- - G - ---- - - - - - T - - SG - - N 0.09(n~09) ND ND ND ND -18.3±3.4 -20.3±3.5 12.6±1.1 29.5±9.5 A

The complete sequence of the Mii segment is glven for a 7 WT. For a 7 -1 to a 7 -9 mutant sequences, only those residues differing from the WT sequence are shown. The gap between positions 236 and
237 of the WT is indicated when present (mutants a7-4, a7-5. a7-6 and a7-8). Mean whole-cell responses were measured at 100 µM ACh. EC50 values for ACh and DHJ3E are gfven in µM. Reversa!
potential of a. 7 -1 mutant was determined after (a) 0.6 s 100 µM ACh application. (b) 0.6 s 0.2 µM ACh application and (e) 3.6 s 100 µM ACh application; Ach pulses lasted 2 s (a, b) or 4 s (c). Similar reversa!
potentials were determined at 1 µM and 100 µM A Ch for mutants a. 7 -2, a 7-4 and a 7 -6, and at 5 µM and 100 µM for mutant a. 7 -8. No ionic response could be recorded in oocytes injected with a. 7-3 cDNA
NO, not determined; CA, competitive antagoníst, IC50 of OHJ3E for WT receptor, 1.6 µM (ref. 36). Similar reconñngs were obtained for every mutant in at least 5 cells from more than one donor. Mutagenesis:
Mutants were prepared as described in ref. 16. Their complete coding sequence were systematically checked. Half- or full-length chimaeras of the Mii segment were prepared by the same method.
Electrophysiology: Oocytes were prepared, injected and recorded as described51 Recordings were typical1y made two or three days after injection. For voltage-clamp measurements, ccus were incubated in
OR2 medium (NaCl, 82.5 mM; KCI. 2.5 mM; CaCl2, 2.5 mM; MgCl2. 1 mM; atroplne. 0.5 µM; HEPES, 5 mM, pH 7.4) and challenged by ACh application. Solution changes were complete within 100 ms. Data from
one oocyte are given in the figures and mean values ±s.e.m. determined from 5-10 oocytes are given in Table 1. Dese-response curves were obtained by plotting the peak of the ACh-evoked currents,
measured for 2-s applications, as a function of the logarithm of the agonist concentration. Cells were held at -100 mV. Mean currents obtained in five cells exoressmg each mutant were normalized to the
vatues determined at saturating egcnlst concentration. The empirical Hill equation, adjusted by welghted least-square algorithrn. yielded the apparent 50% effective concentrations (EC50) and Hill coefficients
(nH) given in Table 1. No residual currents were measured for a 7 WT and a7-1 receptors after 30-min exposure to 100 nM cbungarotoxln. 1-V curves were obtained in the voltage range between -50
and +40 mV and generated by substractlng passive membrane currents from currents evoked in the presence of ACh at each potential. For ion selectivity expenrnents. chloride concentration was oecreased
from 92 mM to 9.5 mM by replacing NaCI by Na-isethionate (experiments denoted isethionate), sodium concentration was decreased from 82.5 mM to 41.25 or 8.25 mM by replacing 50% or 100% of NaCI
by mannitol to preserve osmolarity (experiments denoted 1/2Na-mannitol or mannitol). To minimize changes in junctional potential on ionic substitution. an agar-brtdge was used as a ground erectrode.
Residual potential changes were estimated and corrected for. BAPTA injection: Control responses were first recorded and a third electrode, similar to those used for cONA injection was poked into the
oocyte. This eíectrode was made with a calibrated capillary and fi!led with a solution containing 0.1 M BAPTA in NaCI, 82.5 mM; KC!, 2.5 mM; HEPES, 5 mM, pH 7.4. Small positive pressure pulses, ceuorateo
to produce 10 ni injections were used to inject BAPTA. Four pulses were normally applied for standard injections. The 40 ni BAPTA at 0.1 M must roughly correspond to 7 mM interna! concentration of chelating
agent. The injection was monitored under visual control and effectiveness of BAPTA actíon was checked electrophysiologically. lnjection of larger volumes never showed any significant differences with
standard procedure. Singl«H:hannel recordings: Quality and response of oocytes used for single-channel recordings were first assessed with the dual erectrooc voltage clamp and vitelline membrane was
then removed as described16. Because of apparent clustering of the channels in the oocyte membrane, AChRs were first localized using a puffing electrode and whole-cell recorolng, Pipettes used for
outsjoe-cut peten recordings were sylgarded to reduce capacitance and contained (in mM) KF, 70; EGTA, 20; HEPES, 5; pH 7.4. The bath solution contained OR2 in control experiments and OR2 0.5Na+/0.5
mannitol in ionic substitution experiments. Ground electrode consisted of an agar bridge to minimize changas in junction potential on substitution of CI- in the bath. Data were filtered at 1.5 kHz and sampled
at 5 kHz.

The first mutant investigated (a7-l, Table 1) included one changes of mutant a7-1 was slow compared with WT (Fig. 3b)
additional amino acid (a proline residue) located between a'l- and that dihydro-/3-erythroidine (DH/3E) behaved asan agonist
G236 and a7-E237 ata position now referred to as 236' and six (Table 1 ). This is in contrast with the WT receptor which displays
additional point mutations at positions 237 (Glu ~Ala), 240 a single low-affinity (EC50 = 115 µM) conducting state, referred
(Ser-s Gly), 251 (Val-s Thr), 254 (Leu-v Ser}, 255 (Leu-s Gly) to as the 'active' (A) state, which rapidly desensitizes and is
and 258 (Glu~Asn). competitively inhibited by DHf3E36.
Expression of a7-1 mutant in Xenopus oocytes yielded func- Ionic currents obtained from whole-cell recordings reversed
tional ACh-gated ion channels. The quaternary structure of the at -19.0±4.6 mV. Substitution of90% externa! chloride ions by
WT and mutant receptors was not controlled in this expression isethionate (gluconate or methane sulphonate, not shown),
system, but there were no major differences in Hill coefficient shifted the reversa! potential of ACh-activated currents towards
(Table 1) or sensitivity to a-bungarotoxin (not shown) between positive voltages ( Erev = 27 ± 7 .6 mV) independently of the con-
a'l WT and a7-l, indicating that the overall structure of the centration of ACh or the duration of agonist application (Table
protein is probably not much altered. But severa! other proper- 1). These shifts were well described by the Goldmari-Hodgkin-
ties of this mutant greatly changed, as in the case of the pre- Katz relationship (Fig. 3d) for chloride-specific channels,
viously described L247T mutant=!". The whole-cell current- indicating that the ACh-activated currents were almost entirely
voltage relationship became linear (Fig. 3a ), the <lose-response carried by chloride ions. Furthermore, reversa! potential (as the
curve for ACh was shifted by 500 times towards lower concentra- kinetic properties ofthe response) did not change in the presence
tions (Table 1), and most of the desensitization of the response of interna! BAPTA (Erev=29.3±3 mV, Fig. 3a), excluding a
to ACh was abolished (Fig. 3b, compared with WT responses significant contribution of calcium-activated chloride channels
in Fig. 2c). Thus, in a7-l, as in L247T (ref. 16), a new functional to the ionic response of the oocyte. Replacement of 100% extra-
high-affinity ( effector concentration for half-rnaximal response, cellular sodium chloride by mannitol (Fig. 3c) shifted the
EC50 = 0.24 µM) open state exists, which could correspond to reversa) poten tia! towards positive voltages ( Erev = 43 .6 ±
a desensitized but conducting 'D*' state of the nAChR16• Con- 6.6 mV, n = 8; 8Erev = 62 mV) similar to those determined
sistent with this notion, we noticed that the onset of permeability with substitution of chloride by isethionate, indicating that

FIG. 1 Comparison of Mil sequences from subunits of the Mil

cation-selective nicotinic a 7 receptor with those of anion-
selective glycine a1 (ref. 24) glycine f3 (ref. 49), GABAA a1 234 236 237 240 251 254 255 258
and {31 (ref. 38) receptor subunits. Numbers refer to a 7 AChR a7 D s
G(:®K 1 @L G T V s r®F M(S©V A@
sequence. Amino acids previously mutated in nAChRs (see GlyR a1 D A A P A R V G G T T V T M T T O s S G S R A
text) and conserved in glycine and GABAA receptors are GlyR /3 D A S A A R V p G s V s L A s E e T T A A
indicated by stars. Amino acids mutated in this study are GABAR a1 E s V p A R T V G V T T V T M T T s s A R N
shown in circles (insertion between 236 and 237 and substitu- GABAR /31 D A s A A R V A G T T V T M T T s T H R E
tions at positions 237, 240, 251, 254, 255 and 258). * * * * *
NATURE · VOL 359 · 8 OCTOBER1992 501
© 1992 Nature Publishing Group
ARTICLES

permeability to sodium was not significant (PNa/ Pe1 < 0.05). As Mutant a7-1
a probable consequence of the low amplitude of whole-cell a b Control
-10mV~
currents ( 21 O nA, Table 1), no single-channel recordings could
be obtained (17 oocytes from 7 donors; 122 patches tested) to -40mV~

characterize further a7-l. Yet the available data support the lsethionate
40mV
conclusion that, in the presence of ACh, the current fiows
through a high-affinity D* state of the a 7-1 channel and is mv OmV
-40 o
mostly carried by chloride ions. / _,.. lset-BAPTA
o .. ...-•::::8.-
LI 40mV
To identify further the critica! residues involved in the conver- LI
O Control

sion ofion selectivity from cationic (a7 WT) to anionic (a7-l), -200
!::,. lsethionate
" lset-BAPTA
OmV ~
the number of amino acids exchanged was progressively _J~
2s
reduced. The mutant a7-2 contained a proline at position 236'
and amino acids 237 and 251 were occupied by Ala and Thr, e nA 0 d O Mannitcl
ol
"ít~
-.
respectively. As for a7-l, this mutant formed functional chan- O Control 200
40
6. Mannitol-BAPTA
nels of high apparent affinity for ACh, that can be activated by 6 lsethionate O> lset-BAPTA
o Cont ·BAPTA
/.!,.. Mannitol
DH/3 E (Table 1) and that display weakly desensitizing (Fig. >.§_ 20
3f}, and non-rectifying (Fig. 3e) ionic responses. Also, substitu-
tion of externa! anions in the presence or absence of interna! UJe o
BAPTA yielded results similar to those of a7-1 (Fig. 3e, Table ~~
1); thus a7-2 would, again, conduct ions in the D* state and -20 i§
be selective for anions. 5
C!" Concentration {mM)
100
Further reduction in the number of point mutations (mutants
a7-3 to a7-5) did not invert the sign of ion selectivity of a7
WT. To understand how three amino-acid substitutions (236' ~ Mutant a7-2
Pro, Glu 237 ~Ala and Val 251 ~ Thr) gave the multiple efíects e nA J Control
observed in a7-l and a7-2, we analysed the properties of each 100 -SmV~
__r------._
mutation separately. o

o
/ -45mV~

Desensitization but not selectivityaltered in a7~4 /

mV
As shown in Fig. 4 and Table 1, this single Val 251 ~ Thr point -40 /
mutation (mutant a7-4) accounted for the changes in apparent
----~
o _,.-e
affinity for ACh, sensitivity to DH¡3E (Table 1), response o/ ~ o Control lset-BAPTA
6. lsethionate
desensitization and current rectification observed with the 0 lset-BAPTA
-100
multiple mutants a7-l and a7-2.
Single-channel recordings from outside-out patches (Fig. 4c)
further indicated that mutant a7-4, as the L247T mutant16,

a.7-WT FIG. 3 a, /-V relationship of the a7-1 mutant receptor. The /-V curve was
a nA determined first in control condition (Control) with 2-s ACh applications.
nA b 100
100 Ninety per cent of chloride ions of the perfusion medium were then replaced
by isethionate and the /-V curve determined agaín (lsethionate). lnjection
of BAPTA did not induce significant changes in this /-V relationship (lset-
40 BAPTA). Unes through the data points correspond to the Goldmann-Hodgkin-
Katz (GHK, ref. 50) equation adjusted by weighted least-square algorithm
o Control
assuming a permeability of 5% for isethionate relative to chloride. The
O Control
parameters used correspond, respectively, to: permeability, interna! and
/::, lsethionate + Cont·BAPTA externa! ion concentration (in mM). Control: 0.27, 34, 92; lsethionate: 0.28,
34, 9.5, for CI- and 0.013, O, 82.5 for isethionate; lset-BAPTA: 0.24, 34,
-300 9.5, for c1- and 0.022, O, 82.5 for isethionate. b, Time course of the
ACh-evoked currents recorded in control, isethionate and isethionate-BAPT A
conditions. Traces were recorded at 10 mV increments. Horizontal bars
correspond to ACh applications (100 µM). e, /-V curve determined with 2-s
e d ACh applications in control condition (control), after substitution of 90%
Control chloride by isethionate (lsethionate), and after substitution of 100% sodium
Control
chloride by mannitol (Mannitol). Adjusted parameters for the GHK fit are:
Control: 0.22, 33, 92; lsethionate: 0.24, 33, 9.5 for cr and 0.012, O, 82.5
for isethionate; Mannitol: 0.15, 33, 9.5. d, Reversa! potential values as a
Cont-BAPTA
function of the logarithrn of externa! chloride concentration. /-V curves
<(
were determined in the presence of 92, 50.5, 19.75 and 9.5 mM externa!
_J~
e
o
(\/ chloride, after substitution of NaCI by mannitol (diamonds). Other reversa!
potentials. determined after injection of BAPTA, are: triangles: 9.5, 50.5 and
4s 2s
92 mM externa! chloride (substitution by maonitot); filled circle: 9.5 mM
FIG. 2 a, b, Current-voltage U-V) relationships of «t WT receptor. Peak chloride (substitution by isethionate); open circle: 92 mM chloride. Mean
currents evoked by 1-s exposure to 100 µM ACh were measured at severa! ±s.e.m. reversa! potential values were determined in n = 5 cells. The sol id
holding potentials, first in control conditions (Control) in OR2 medium (see line corresponds to the theoretical Nernst relation. e, /-V relationship of
Table 1 legend), then after substitution of 90% externa! chloride by isethion- the a 7-2 mutant receptor obtained using the protocol in a. Adjusted
ate (lsethionate), or ;;.2 min after injection of 4 nmoles of the Ca2+-chelating parameters for the GHK equation are: Control: 0.16, 32, 92; lsethionate:
agent BAPTA (Cont-BAPTA). e, Time course of the ACh-evoked currents 0.14, 32, 9.5 for chloride and: 0.002, O, 82.5 for isethionate; lset-BAPTA:
recorded under control conditions at tour holding potentials separated by 0.14, 32, 9.5 for chloride and 0.015, O, 82.5 for isethionate. f. Time course
10 mV increments. a Currents recorded at -15 mV, before (Control), or of ACh-evoked currents recorded at severa! holding potentials (10 mV
after BAPTA injection (Cont-BAPTA). Further injection of BAPTA did not increments) in the three conditions of e. Horizontal bars correspond to ACh
induce other detectable changes in the ACh-evoked current. applications (100 µM).

502 NA TURE · VOL 359 · 8 OCTOBER 1992

© 1992 Nature Publishing Group
 ARTICLES

exhibited multiple conducting states, one of S4 ± 3 pS ( n = 6)
comparable to that of the WT A state, and one conductance of
86.3 ± S.4 pS (n = 11), plausibly corresponding to a conducting
D* state. In agreement with this hypothesis, we noticed that
DH/3E behaved as an agonist of a7-4 mutant (Table 1). Also
after injection of BAPT A, ionic responses did not desensitize
and their onset resembled the slow ones observed for a7-l.
Current reversa! potential of mutant a7-4 ionic responses
shifted by S4 mV towards positive voltages on substitution of
90% externa! chloride (Fig. 4a; Table 1), indicating that an
important part of the observed current was carried by chloride
ions. But in contrast to a7-l or a7-2 receptor ionic responses,
after injection of BAPTA, externa! chloride substitution pro-
duced no effect (from E,ev=-4.2±S.6 to E,ev=2.2±S.7mV,
Table 1). Similar results were obtained when Ca2+ was sub-
stituted by Ba2+ (Erev = -3 mV, data not shown). As for the a'l
WT receptor, these data provide evidence for the activation of
calcium-dependent chloride channels. Single-channel current-
voltage relationships (Fig. 4c) indicated that reversa! potentials
of both conducting states were insensitive to changes in externa!
chloride concentrations (not shown) but were shifted towards
negative voltages (<5E,ev = 23 mV) on twofold reduction of exter-
na! sodium (see legend to Fig. 4c ), consistent with a cation-
selective permeability.
Thus two functional features ofthe channel can be dissociated
by mutagenesis: its ability to conduct ions in the D* and/ or A
from -18.S ± 1.3 mV to -27.3 ± l.S mV (n = 4) on substitution
of SO% externa! sodium chloride by mannitol (Fig. Se) as expec-
ted for cation-permeable channels, and thus establishing that
mutant a7-6 is now cation-selective and that conversion of ion
selectivity can be observed for one given conducting state.
Similar results with regard to cation versus anion selectivity
were obtained when deletion of proline 236' was done in the
more complex a7-1 mutant to yield a7-8 (see Table 1). Con-
versely, insertion in the cation-selective a7-8 mutant of either
a proline (a7-l) oran alanine (a7-9) at position 236' reversed
the ion selectivity to anionic, thus suggesting that the length of
the segment spacing MI and Mil is more critica! than the actual
nature of the residue inserted.
No analysable (less than 20 nA) ionic currents could be
recorded from oocytes injected with mutants a7-3 and a7-7
Mutant a7A
b
§------
Control
a nA
2000
o -10mV
·40mV --
lsethionate
o 48mV~
mV
-40 /

;~·
states and its ionic selectivity. o/.
8mV

Divalent/monovalent cation selectivity lset·BAPTA

Glutamate residues at position 237 in a'l WT receptor form the o ti

'interrnediate' ring of negatively charged amino acids previously / O Control 24mV~

/
D. lsethionate
investigated in Torpedo californica receptor9•2º·37. e lset·BAPTA ·34mV"
<l'.
e
In this study with homo-oligomeric a7 neuronal receptor,
mutation of Glu237 to Ala removed the whole set of charges ¡ ·3000
_Jg
2s
from this ring (mutant a7-S). The receptor still responded to
ACh (Fig. Sa) with activation by high agonist concentrations
(EC50 = 194 µM) and rapid desensitízation of ionic responses
e pA
in the presence of physiological concentrations of divalent Control o Control 3
cations (Fig. Sb ). These properties are similar to those of the "' 1 /2Na-mannitol
WT receptor which exhibits a low-affinity A conducting state. o
In contrast to WT, current-voltage relationships were now -80 -60 ' • 20
c.._~c...,.,..,=..-<oo>'---+-~ mV
:-";,,<f'
linear (Fig. Sa) and reversed at -14.2±2.l mV. Also, injection •• ,.. o ..
••• O cPOJO .,..
of BAPT A in the oocytes, did not affect the reversal poten ti al <l'. cP o

(Erev = -13.S ± 1.7 mV) of ACh-evoked currents, suggesting, as !~

1.
SOmsN
proposed for nAChR mutants" and glutamate receptor variants
or mutants25-27, that this channel is permeable to cations, but
has lost its ability to conduct divalent cations (D.B. et al., pA

:~.
o Control 3-
manuscript in preparation). Control
"' 1/2Na-mannitol ¡..\\

Ion selectivityin the same DHPE activated state ,::t.

As seen on the sequence alignments in Fig. 1, one additional
-8~--~- '1t1 ,¡"' _J
20
mV

residue, a Pro or an Ala is present between positions 236 and

237 (position 236') at the N-terminal end of Mil in the anion-
selective Gly and GABAA receptor channels24•38•
Deletion ofthis amino acid (a proline) in the chloride-selective
a7-2 receptor resulted in a functional ACh-gated channel
(mutant a7-6). Mutant a7-6 ionic responses were activated by
ACh and DH/3E with EC50 (0.98 µM and 3.69 µM, respectively)
close to those found for mutant a7-2 (see Table 1), and did not
desensitize either in the presence or absence of BAPTA (Fig.
Sd). Moreover, their onset was slow and no residual current
persisted on removal of the agonist. Altogether, these observa-
tions indicate that a7-6 mutant conducts ions in the same state
as a7-2 mutant, referred to as the D* conducting state.
Current-voltage relationships were linear. The reversa} poten-
tial was not changed on injection of BAPTA (<5E,ev =O mV, Fig.
Se) suggesting that Ca2+ does not permeate, and was no longer
affected by external chloride substitution ( 8E,ev =O m V), thus
indicating that chloride contribution to the current was negli-
gible. Furthermore, whole-cell current reversa} potential shifted
_JJl°
SOmsN -3
FIG. 4 a, 1- V relationship of the a 7-4 mutant receptor obtained using the
protocol in Fig. 3a. b, Time course of ACh-evoked currents were recorded
at severa! holding potentials (10 mV increments) in the three conditions of
a. Horizontal bars correspond to ACh applications (100 µM). e, Single-channel
recordings of a 7-4 mutant receptor. Excised outside-out patches from
oocytes expressing mutant receptors were recorded in OR2 medium. Recep-
tor channels were activated by 100-ms puffs of 20 µM ACh (bars). Mutant
a 7-4 exhibits multiple conductance levels, illustrated here by a 49 pS (top)
channel and an 80 pS (bottom) channel. Recordings are from two different
patches ata membrane potential of -100 mV. Current-voltage relationships
were constructed using a voltage ramp protocol. Leak currents obtained in
the absence of ACh were substracted from traces showing channel activity.
The reversa! potential of both channel types (indicated by arrows) shifted
from -21.9 ± 2.6 mV {n = 5) to -45.5 ± 7 mV {n = 3) for the 49 pS conduct-
ance and from -19.6±3.5mV (n=4) to -42.5 (n=2) for the 80pS on
substitution of 50% of the externa! NaCI with mannitol.

NATURE · VOL 359 · 8 OCTOBER 1992 503

© 1992 Nature Publishing Group
 ARTICLES

(Table 1) containing the additional Pro residue, but not the domain of a chloride-selective GlyRa l (ref. 24) into that of a
Val251 ~ Thr mutation. cation-selective a'l nAChR14'15 allowed the design of an ACh-
gated channel now selective for chloride. Mutations of only
Discussion those residues presumed to face the lumen ofthe channel yielded
The noticeable sequence homologies between nicotinic, functional receptors, whereas exchange of the entire Mii seg-
serotoninergic 5HT3, GABAA and GlyRs (refs 1, 4), suggest that ment of the nicotinic a 7 receptor with that of the glycine a 1
the different properties of these receptors rely on a limited receptor did not give functional channels. Possibly this failure
number of key amino-acid side chains within a common poly- resulted from perturbations of the tertiary or quaternary struc-
peptide backbone. For instance, the fact that invertebrates ture of the receptor associated with amino acids located in the
express chloride channels gated by ACh38-41 and possibly by non-luminal face of Mii segment44. In contrast, the success of
glutarnate'", illustrates the absence of strict relation between the individual amino-acid exchange method indicates that the
neurotransmitter specificity and ion selectivity. Moreover, overall geometry of these oligomers is highly similar.
severa) chloride channels Jet cations permeate together with Ionic selectivity was converted after a small number of amino-
anions (reviewed in ref. 43), indicating that intermediate cation acid additions and permutations within o'l sequence: the addi-
versus anion selectivity may exist and that anion and cation tion of a Pro (or Ala) residue at the N-terminal end of the Mil
channels may share common structural properties. segment (between positions 236 and 237) and two arnino-acid
Consistent with these observations, we show that introducing substitutions at positions 237 (Glu ~Ala) and 251 (Val » Thr).
appropriate amino-acid residues from the putative channel Unexpectedly, mutations of two rings of negatively charged
amino acids (outer (position 258) and intermediate (position
Mutan! a7-5
237) rings") to neutral amino acids led to functional receptors
a nA b
100 (a7-5, a7-6, a7-8) which are still selective for cations. Yet,
Control
unlike the WT receptor, those channels mutated at the intermedi-
a te ring exhibited low calcium permeability. Thus, although
these charged amino acids, in particular from the intermediate
ring, might contribute to the selectivity among cations, they do
not seem to play a critical role in anion versus cation selectivity.
40 The most critica] mutation for inverting the sign of ion selec-
-40mV tivity was the insertion or deletion of a neutral residue modifying
the length of the short segment linking MI to Mil. This effect
D Control is strongly dependent on the nature of the residue at position
+ Cont-BAPTA 251. Indeed, when position 251 is occupied by a valine residue,
-100
-c the receptor displays only a low-affinity conducting state and
_J~ the insertion at position 236' generally does not yield functional
2s receptors (mutants a7-3 and a7-7). When position 251 is
occupied by a Thr residue, the receptor displays a high-affinity
conducting state but is still selective for cations. The insertion
Mutan! a7-6
of a neutral residue then converts the ion selectivity from cationic
e d (mutant a7-6) to anionic (mutant a7-2). Under these conditions,
both anion- and cation-selective channels belong to a high-
Control affinity 'desensitized but conducting' D* state16•17 ofthe receptor.
OmV~~::::=:::===== Propasa] of a detailed structural model for cationic versus
mV ;/ anionic selectivity ofthe ion path is premature at this stage. Yet
-40 r.f 40
-30mV~ it is tempting to speculate that the conversion of ion selectivity
(,l o Control
+ Cont-BAPTA
observed on Pro or Ala insertion in mutant a7-6 and a7-8
¡Jd D. lsethionate -2mV
lsethionate (yielding a7-2, a7-l and a7-9, respectively), is not directly
-1000 o lset-BAPTA related to a defined chemical modification within the channel
-22mV~ lumen but, rather, to slight changes of the geometry of the Mii
e segments which delineate the channel. Such change would then

-~: :@=
lset-BAPTA
40
O Control
A 1/2Na-mannitol
_J§
2s
-800
FIG. 5 a, /-V relationship of a 7-5 mutan! receptor. Same protocol as in
Fig. 2b with 1-s exposure to ACh. Adjusted parameters for the GHK equation
are: Control: 0.07, 1.9, 82.5 for Na+ and 0.11, 89, 2.5 for K+; Cont-BAPTA:
0.04, 1.5, 82.5 for Na+ and 0.075, 89, 2.5 for K+. b, Time course of the
ACh-evoked currents at several holding potentials (10 mV increments). e,
/-V relationship of the a 7 -6 mutant receptor. Same protocol as in Fig. 3a.
Adjusted parameters for the GHK equation are: 1.07, 2.5, 82.5 for Na+ and
1.96, 89, 2.5 for K ". d. Time course of ACh-evoked currents of a 7-6 mutan!
receptor recorded at several holding potentials in control isethionate and
isethionate-BAPT A condition. e, /-V relationships of a 7-6 mutan! were
determined first in control condition (Control) and after substitution of 50%
of NaCI by the mannitol (Na-Mannitol). Adjusted parameters for the GHK fit
are: 0.4, 1.8, 82.5 for Na+ and 0.78, 89, 2.5 for K+ in control conditions and
became 0.55, 1.8, 41.25 for Na+ and 0.78, 89, 2.5 for K+. Horizontal bars
represen! the ACh applications (100 µM).
504
© 1992 Nature Publishing Group
modulate the accessibility of permeant ions to putative selectivity
filters located within or outside Mil.
The slow conformational transitions which take place in the
-ce:
course of desensitization of the Torpedo receptor protein alter
altogether the relative orientation of the subunits45 and the
structure of sites located at their interfaces46'47• Possibly such
dynamic changes in the geometry of the quaternary structure
constrain (or even regulate) the selectivity of the channel, as
long as the ion path is open in the various allosteric states of
the receptor. Consistent with this view is the striking difference
of channel properties between the two conducting states of
L247T (refs 16, 17) or V251T (a7-4) (ref. 52) mutants, as well
as the difference of size or state of hydration between cations
and anions permeating through the various mutant channels.
Also, in the case ofGABAA receptors'", fast and slow transitions
between channel states with different properties have been recor-
ded and display analogies with sorne of those reported here.
Our data further suggest that, as in the case of typical allosteric
proteins, the geometry ofthe quaternary structure ofthe receptor
protein and its ability to undergo conformational transitions
might be important parameters for determining the ion selec-
tivity of the ion channel it contains. O
NATURE · VOL 359 · 8 OCTOBER 1992

Received 13 April: accepted 3 September 1992.
1 Betz, H. Neuron 5, 383-392 (1990).
2. Galzi. J. L., Revah, F., Bessis, A. & Changeux. J. P A. Rev. Pharmac. Tox. 31, 37 - 72 (1991).
3. Seeburg, P. H. et af Cold Spring Harb. Symp. Quant. Biol 55, 29-40 (1991).
4. Maricq, A. V., Peterson, A. S., Brake, A. J., Myers, R.M. & Julius, D. Science 254, 432-437 (1991).
5. Giraudat, J., Dermis. M .. Heidmann. T., Cnang. J. Y. & Changeux, J. P. Proc. natn. Acad. sci. USA
83, 2719-2723 (1986).
6. Hucho, F., Oberthür, W. & Lottspeich, F. FEBS Lett. 205, 137-142 (1986)
7. Revah, F. et al. Proc. natn. Acad. Sci. U.S.A. 87, 4675-4679 (1990)
8. lmoto, K. et al. Nature 324, 670-67 4 (1986).
9. Irnoto. K. et al. Nature 335, 645-648 (1988).
10. Heidmann. T. & Changeux. J. P Eur. J. Biochem. 94, 281-296 (1979).
11. Neubig, R., Boyd, N. & Cohen, J. B. Biochemistry 21, 3460-3467 (1982)
12. Katz. B. & Thesleff. S. J. Physiol., Lond. 138, 63-80 (1957).
13. Sakmann, B., Patlak, J. & Neher. E. Nature 286, 71-73 (1980).
14. Couturier. S. et al. Neuron 5, 84 7-856 (1990).
15. Schoepfer. R .. Conroy, W. G., Whithing, P., Gore. M. & Lindstrom, J. Neuron 5, 35-48 (1990).
16. Revah, F. et al. Nature 353, 846-849 (1991).
17. Bertrand, D. et al. Proc. natn. Acad. SCI. U.S.A. 89, 1261-1265 (1992).
18. Bormann, J., Hammill, O. P & Sakmann, B. J. Physiol, Lond. 385, 243-286 (1987)
19. Charnet, P. et al. Neuron 4, 87-95 (1990).
20. !moto, K. et al. FEBS Lett. 289, 193-200 (1991)
21. Leonard, R. J.. Labarca, C. G., Charnet, P., Davidson, N. & Lester. H. A. Science 242, 1578-1581
(1988).
22. Giraudat, J. et al. Biochemistry 26, 2410-2418 (1987).
23. Langosch, D .. Hartung. K., Grell, E .. Bamberg, E. & Betz. H. Biochim b1ophys. Acta 1063, 36-44
(1991).
24. Grenningloh, G. et al. Nature 328, 215-220 (1987).
25. Hollmann. M., Hartley. M. & Heinemann. S. Science 252, 851-853 (1991)
26. Verdoorn, T. A., Burnashev. N .. Monyer, H . Seeburg, P. H. & Sakmann, 8. Science 252, 1715-1718
(1991).
27. Curutchet, P. et et. Biochem. biophys. Res. Commun. 182, 1089-1093 (1992)
Crystal structure at 1.7 A of the bovine
papillomavirus-1 E2 DNA-binding domain
bound to its DNA target
Rashmi S. Hegde', Steven R. Grossmantt, Laimonis A. Laimlns' & Paul B. Sigler*
* Howard Hughes Medical lnstitute and the Department of Molecular Biophysics and Biochemistry,
t Department of Molecular Genetics and Cell Biology and the Howard Hughes Medical lnstitute, The University of Chicago, Chicago, lllinois 60637, USA
The dominant transcriptional regulator of the papillomaviruses, E2, binds to its specific DNA target
through a previously unobserved dimeric antiparallel p-barrel. The DNA is severely but smoothly
bent over the barrel by the interaction of successive major grooves with a pair of symmetrically
disposed a-helices. The specific interface is an 'interwoven' network of interactions where the
identifyingbase pairs of the targetcontact more than one amino-acidside chain and the discriminating
amino acids interactwith more than one base pair.
THE papillomaviruses are a family of small DNA viruses that
cause hyperplastic epithelial lesions. The most notable
pathogens are the human papillomavirus strains HPV-16 and
18, which have been implicated in cervical carcinomas',
although it is bovine papillomavirus-1 (BPV-1) that has been
the most extensively studied (for reviews, see refs 2 and 3). The
products of the E2 gene, the proteins E2, E2-TR and E8/E2,
are pivota) to the viral life cycle because they regulate viral
transcription from ali viral promoters and are essential for viral
replication'':", The full-length (410 amino acids) E2 gene product
is a trans-acting transcriptional regulator. It is also required, in
conjunction with the El protein, for the initiation ofviral replica-
tion. The two truncated forms of the E2 protein, E2TR and
E8/E2, function as repressors. They both lack an intact amino-
terminal activation domain. Ali forms of the E2 protein are
dimeric and their function requires preferential binding to a
highly conserved palindromic target sequence", termed E2-BS
(Fig. lb), which occurs 17 times in the BPV-1 genome.
:j: Present address: Department of Medicine, Brigham and Women's Hospital,
Boston, Massachusetts 02115, USA.
NATURE · VOL 359 · 8 OCTOBER1992
© 1992 Nature Publishing Group
ARTICLES
28 Mulle, C .. Choquet, D., Korn, H. & cnangeux. J P. Neuron 8, 135-143 (1992).
29. Vernino. S., Amador, M, Luetje. C. W., Patrick. J. & Daru. J. A. Neuron 8, 127-134 (1992)
30. Vernino, S, Amador, M. & Daru, J. A. J. Ce// Biocnem. 16E (suppl.) 217 (1992)
31. Miledi, R. Proc. R. Soc. 8215, 491-497 (1982)
32. Miledi, R. & Parker. l. J. Physiol, Lona. 357, 173-183 (1984)
33 T sien, R. Y. Biocnemistrv 19, 2396-2404 (1980).
34. Pullman, A. crem. Rev 91, 793-812 (1991).
35. Unwin, N. Neuron 3, 665-676 (1989)
36 Bertrand, D., Bertrand. S. & Ba!livet. M. Neurosci. t.ett. {in the pressl
37 Konno, T. et al. Proc. R. Soc. 8244, 69- 79 (1991)
38 Schofield, P R et al. Nature 328, 221-227 (19871.
39. Tauc, L. & Gershenfeld, H. M. J. Neurophysiol 25, 236-262 (1962).
40 Kehoe, J. J. Physiol, Lond. 225, 85-114 (1972)
41. Ascner. P. & Ekulkar, S. in Sicgle-cnennet Recording (eds Sakmann, B. & Neher. E.) 401-406
(Plenum, New York, toncon. 1983).
42 Arena. J. P., Liu, K. K., Pareas, P. S. & Cully, D. F. J. cen. Blocnem. 16E (suppl.) 226 (1992)
43. Franciolini, F. & Petris. A Bioctum. biophys. Acta 1031, 247-259 (1990)
44. Lo, D. C .. Pinkham, J. L. & Stevens C. F. Neuron 6, 31-40 (1991)
45. Unwin, N., Icyosruma. C. & Kubalek, E. J. Gel! Biol. 107, 1123-1138 (1988)
46. GalZt, J. L. et al. Proc. natn. Acad. Sci. U.S.A. 88, 5051-5055 (19911
47. xerun A. tre Harveylectures. Series 85 71-107 (Wiley-Liss, New York, 1991).
48. Newland. C. F.. Colquhoun, D. & Cult-Candy. S. J. Physiol .. Lond. 432, 203-233 (1991)
49 Grenníngtob. G. et al. Neuron 4, 963-970 (1990).
50 Hodgkin, A. L. 8101. Rev. 26, 339-409 (1951).
51 Bertrand, D .. Cooper. E., Velera. S., Rungger. D. & Ballivet. M. Methods in Neuroscience Vol. 4 (ed
Conn. P.M.) 174-193 (Academic. San Diego, 1991)
52. Devillers-Thiéry, A., Ga!zi, J L, Bertrand, S., Changeux, J. P. & Bertrand, D. Neuroreport (in the press).
ACKNOWLEDGEMENTS. We thank R. Miles. C. Mulle and C. Lena ter discussions and reading the
maouscnot. Ttus work was supported by grants from the Association Franéaise centre les Myopathies.
the Collége de France, the Centre National de la Recherche Scientifique. the Ministére de la Recherche,
the Direction des Recherches Etudes et Techniques and the Swiss National Science Foundation (to D.B.).
o
Yate University, New Haven, Connecticut 06510, USA
Specific DNA binding is mediated through a carboxy-terminal
85-amino-acid domain which is also sufficient for dimeriz-
ation 7-9. This domain is highly conserved among the E2 proteins
of the papillomaviruses (30% identity) and does not include a
sequence motif previously associated with DNA binding, The
DNA-binding/ dimerization domain is connected to a 160-
amino-acid, acidic N-terminal activation domain through a pro-
line-rich Iinker of variable length and composition termed the
'hinge' (Fig. la). This arrangement of domains is common to
the E2 protein of al! papillornaviruses.
We describe here the high-resolution crystal structure of the
DNA-binding/dimerization domain, E2(326-410) of BPV-1,
bound to a 16-base-pair (bp) DNA sequence which includes
an idealized E2-binding site (Fig. 2). A dimer of E2(326-410)
forms an eight-stranded antiparallel /3-barrel made up of four
strands from each subunit. A pair of a-helices symmetrically
disposed on the outer circumference of the barrel contain ali
of the amino acids that are directly involved in base-sequence
recognition. The DNA is bent smoothly around the barre) to
enable successive major grooves to engulf these recognition
helices.
505