letters to nature
(Fig. 1c). Therefore, Ni3Al remained nanocrystalline throughout the
deformation. This observation has signi®cance regarding the defor-
mation mechanism. The common explanation of strain hardening
during superplasticity at constant strain rate and temperature is that
grain growth increases the ¯ow stress through grain-size sensitivity,
where stress is proportional to grain size. But at 650 8C, the ¯ow
stress for nanocrystalline Ni3Al increased six-fold between the yield
stress and the peak stress, while the grain size increased only two-
fold. This suggests that mechanistic details of superplasticity in
nanocrystalline materials are fundamentally different from those in
microcrystalline materials. A detailed microstructural investigation
is required to explain the high strain hardening observed during
superplasticity.
From the stress±strain curves, it is clear that the maximum
elongation for each material was found at higher temperatures
than those discussed above. The reason for focusing on the lower
temperatures is to investigate the effect of nanostructure, and to
probe the low-temperature limits of superplasticity. At higher
temperatures, grain growth shifts the microstructure into the sub-
microcrystalline range. Furthermore, the practical advantages of
superplastic deformation are not necessarily lost by operating
away from the ductility maximum, as deformation in the range of
a few hundred per cent is all that is required for many commercial
forming operations.
Our experimental results allow us to make some observations
about superplasticity in nanocrystalline metals and alloys. The
single most signi®cant observation is the large reduction in super-
plastic temperatures. In the case of nickel, an extremely low
normalized temperature was attained. But even at very low tem-
peratures, grain growth during deformation of nanostructured
nickel resulted in a grain size larger than the commonly used
de®nition for nanostructured materials of 100 nm or less. The
primary reason for such growth is a large driving force resulting
from grain-boundary energy. These results appear to eliminate the
hope of obtaining superplasticity in pure metals having a grain size
of 100 nm or less, because the reduction of superplastic temperature
is offset by a reduction in the grain-growth temperature. However,
secondary factors such as internal strain energy may in¯uence the
grain-growth temperature through contributions to the driving
force, and these effects have not yet been evaluated.
Even if secondary contributions to grain growth could be elimi-
nated, grain-growth and superplastic temperatures cannot be
treated independently because both are thermally activated pro-
cesses linked closely through solid-state diffusion. Therefore, in the
limiting case of nanostructured pure metals, grain growth will occur
if suf®cient thermal activation is present for superplastic deforma-
tion. This competition between grain growth and superplasticity
was established early in the development of the ®eld, but has not
hitherto been investigated at such small starting grain size.
Grain growth is commonly controlled with multiple-phase alloys,
where second phases are present as particles or individual grains.
Particles inhibit grain growth by pinning the boundaries, while two-
phase systems require diffusion through neighbouring grains of
different composition to accommodate growth. These effects can be
seen by comparing the grain growth during deformation between
nanocrystalline nickel and 1420-Al (data not shown), where the
increase in grain size of nickel was more than that of 1420-Al even
though nickel was at a lower normalized temperature. In the case of
ordered intermetallics, such as Ni3Al, grain growth is inhibited by the
kinetic barrier of preferred atomic pairing between the species present,
and the grain-size stability of the Ni3Al is attributed to this effect.
Investigation of superplasticity in nanostructured materials is still
in its infancy, and many systems exist that hold promise for super-
plastic behaviour while retaining a grain size of 100 nm or less. The
small number of experimental results on superplasticity obtained
from nanostructured materials have upheld predictions of low-
temperature deformation, while at the same time the dif®culty
686
© 1999 Macmillan Magazines Ltd
of maintaining a nanocrystalline grain size in pure nickel has
emphasized the interaction of thermally activated processes in
these materials. However, the high ¯ow stresses and signi®cant
work hardening observed in Ni3Al and 1420-Al have suggested that
details of superplasticity are fundamentally different in nanostruc-
tured materials. M
Received 3 November 1998; accepted 19 February 1999.
1. Mukherjee, A. K. in Materials Science and Technology Vol. 6, Plastic Deformation and Fracture of
Materials (ed. Mughrabi, H.) Ch. 9 (VCH, New York, 1993).
2. Floreen, S. Superplasticity in pure nickel. Scripta Metall. 1, 19±23 (1967).
3. Mukhopadhyay, J., Kashner, G. & Mukherjee, A. K. Superplasticity in boron doped Ni3Al alloy. Scripta
Metall. Mater. 24, 857±862 (1990).
4. Gleiter, H. Nanocrystalline materials. Prog. Mater. Sci. 33, 223±315 (1989).
5. Suryanarayana, C. Nanocrystalline materials. Int. Mater. Rev. 40, 41±64 (1995).
6. Valiev, R. Z. Ultra®ne-grained materials produced by severe plastic deformation. Ann. Chim. 21, 6±7,
369±378 (1996).
7. Birringer, R. & Gleiter, H. in Encyclopedia of Material Science and Engineering: Supplement 1 (eds
Cahn, R. W. & Beaver, M. B.) 339±349 (Pergamon, Oxford, 1988).
8. El-Sherik, A. M. & Erb, U. Synthesis of bulk nanocrystalline nickel by pulsed electrodeposition.
J. Mater. Sci. 30, 5743±5749 (1995).
9. Wang, N., Wang, Z., Aust, K. T. & Erb, U. Isokinetic analysis of nanocrystalline nickel electrodeposits
upon annealing. Acta Mater. 45, 1655±1669 (1997).
10. Natter, H., Schmeltzer, M. & Hempelmann, R. Nanocrystalline nickel and nickel copper alloys:
Synthesis, characterisation, and thermal stability. J. Mater. Res. 13, 1186±1197 (1998).
11. Everhart, J. L. Engineering Properties of Nickel and Nickel Alloys 20 (Plenum, New York, 1971).
12. Liu, V. T. & Sikka, V. Nickel aluminides for structural use. J. Metals 38, 19±21 (1987).
13. Berbon, P. B. et al. Fabrication of bulk ultra®ne-grained materials through intense plastic straining.
Metall. Mater. Trans. A 29, 2237±2243 (1998).
14. Kaybyshev, O. A. & Mareklov, A. A. Structural changes during superplastic deformation of nickel and
chromium. Fiz. Metal. Metalloved. 41, 190±196 (1976).
Acknowledgements. This work was supported by the US National Science Foundation and the Civil
Research and Development Foundation.
Correspondence and requests for materials should be addressed to A.K.M. (e-mail: akmukherjee@
ucdavis.edu).
Stochastic sensing of organic
analytes by a pore-forming
protein containing
a molecular adapter
Li-Qun Gu*, Orit Braha*, Sean Conlan*, Stephen Cheley*
& Hagan Bayley*²
* Department of Medical Biochemistry & Genetics, Texas A&M University Health
Science Center, 440 Reynolds Medical Building, College Station, Texas 77843-
1114, USA
² Department of Chemistry, Texas A&M University, College Station,
Texas 77843-3255, USA
.........................................................................................................................
The detection of organic molecules is important in many areas,
including medicine, environmental monitoring and defence1±5.
Stochastic sensing is an approach that relies on the observation of
individual binding events between analyte molecules and a single
receptor6. Engineered transmembrane protein pores are promis-
ing sensor elements for stochastic detection6, and in their simplest
manifestation they produce a ¯uctuating binary (`on/off ') res-
ponse in the transmembrane electrical current. The frequency of
occurrence of the ¯uctuations reveals the concentration of the
analyte, and its identity can be deduced from the characteristic
magnitude and/or duration of the ¯uctuations. Genetically
engineered versions of the bacterial pore-forming protein a-
haemolysin have been used to identify and quantify divalent
metal ions in solution6. But it is not immediately obvious how
versatile binding sites for organic ligands might be obtained by
engineering of the pore structure. Here we show that stochastic
sensing of organic molecules can be procured from a-haemolysin
by equipping the channel with an internal, non-covalently bound
molecular `adapter' which mediates channel blocking by the
analyte. We use cyclodextrins as the adapters because these ®t
NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com

comfortably inside the pore and present a hydrophobic cavity
suitable for binding a variety of organic analytes. Moreover, a
single sensing element of this sort can be used to analyse a mixture
of organic molecules with different binding characteristics. We
envisage the use of other adapters, so that the pore could be
`programmed' for a range of sensing functions.
a-Haemolysin is an exotoxin secreted by the bacterium
Staphylococcus aureus7. The monomeric 293-amino-acid polypep-
tide can self-assemble on lipid bilayers to form a heptameric pore.
Transported molecules move through a 100 AÊ-long channel centred
on the molecular 7-fold axis8. The opening of the channel on the cis
side of the bilayer (the side of assembly) is ,70 AÊ above the
membrane surface and 29 AÊ in diameter. The channel widens into
a roughly spherical vestibule of ,42 AÊ in diameter. About 20 AÊ
above the membrane plane, the vestibule narrows and becomes a
14-stranded b-barrel, 52 AÊ in length, which continues through
the membrane with an average internal diameter of about 20 AÊ.
The trans entrance to the channel lies close to the bilayer surface.
Roughly globular molecules of relative molecular mass up to ,2,000
(2K; refs 9, 10), or larger, elongated polymers such as single-
stranded nucleic acids11, can pass through the a-haemolysin pore.
We found that a-, b- and g-cyclodextrins at micromolar con-
centrations enter the a-haemolysin (aHL) channel and produce
letters to nature
reversible partial blocks of the ionic current. b-Cyclodextrin (bCD;
M r ˆ 1:135K) was investigated further. The block was established
when b-cyclodextrin was added from the trans side of a planar
bilayer (Fig. 1a, b), but not from the cis side. The kinetics of blocking
were consistent with a simple scheme in which there is a single
binding site for b-cyclodextrin inside the lumen of the channel, for
which kCon ˆ 5:46 3 104 M 2 1 s 2 1 , kCoff ˆ 1:15 3 102 s 2 1 , and the
formation constant K Cf ˆ 4:75 3 102 M 2 1 in 1 M NaCl, 10 mM
sodium phosphate and 5 mM EDTA at pH 3.0 (trans). The con-
ductance of a-haemolysin was 750 pS (s:d: ˆ 21, n ˆ 35) in the
absence of b-cyclodextrin, and 272 pS (s:d: ˆ 9, n ˆ 35) in its
presence. Molecular modelling revealed that b-cyclodextrin could
®t into the lumen of the a-haemolysin channel, with its molecular
7-fold axis coincident with the 7-fold axis of the heptameric pore.
On entering from the trans side, b-cyclodextrin might be retained
by the ring of seven leucine residues (Leu 135) that are found about
half-way through the transmembrane domain or, after squeezing
past this restriction, it might be stopped by a second restriction
near the cis end of the b-barrel which comprises (from the trans
side) the amino-acid residues Met 113, Lys 147 and Glu 111. We
distinguished between these two possibilities by site-directed muta-
genesis. The kinetics of block by b-cyclodextrin in the mutant
L135N (carrying an asparagine instead of a leucine residue at

a
0 cis
I( pA )

-15 α-HL

-30 β-CD
Level 1
trans
b 2- adamantanamine
0 cis hydrochloride (A1 )
I( pA )

-15 Level 2
1- adamantane
carboxylic acid (A 2 )
-30 Level 1
trans
c
0 cis
Level 3
I( pA )

-15 Level 2

Level 1 αHL
-30 Level 1
trans
d Level 2 αHL•βCD
0 Level 4 cis
Level 3 Level 3 αHL•βCD•A1
I( pA )

-15 Level 2

Level 4 αHL•βCD•A2
-30 Level 1
Time 20ms trans

Figure 1 Bilayer recordings showing the interaction of a single a-haemolysin
(aHL) pore with b-cyclodextrin (bCD) and the model analytes 2-adamantanamine
(A1) and 1-adamantanecarboxylic acid (A2). All traces were recorded at -40 mV
(cis at ground). aHL was added to the cis chamber and bCD and the adamantane
derivatives to the trans chamber. a, Single a-HL pore continuously open, -31.5 pA
(level 1); b, bCD (20 mM, trans) produces transient partial blockades of the
channel, -11.5 pA (level 2); c, 2-adamantanamine (80 mM, trans) does not affect the
fully open channel (level 1), but produces an additional block of aHL × bCD,
-5.7 pA (level 3); d, 1-adamantanecarboxylic acid (20 mM, trans) produces
additional blockades, -4.7 pA (level 4), of longer duration than those produced
by 2-adamantanamine (level 3).

a b c

Figure 2 Molecular graphics representation of the interaction between aHL and in the lumen of the transmembrane channel. On the basis of mutagenesis data,
bCD. a, View into the channel from the trans side, bCD removed. b, View into the bCD is shown placed against a constriction in the transmembrane b-barrel
channel from the trans side with bCD (gold) bound. c, Sagittal section through the formed by the ring of seven Met 113 side chains.
heptameric a-haemolysin pore showing b-cyclodextrin (yellow and gold) lodged

NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com

© 1999 Macmillan Magazines Ltd 687
letters to nature
position 135) were similar to those of wild-type (WT) a-haemo- Because the block of wild-type a-haemolysin by b-cyclodextrin
lysin (pH 7.5 (trans): WT, kCoff ˆ 1:08 3 103 s 2 1 ; L135N, kCoff ˆ was partial (64%) and because cyclodextrins act as hosts for a variety
4:7 3 102 s 2 1 ). By contrast, the blocking events were greatly of guest molecules, we investigated whether guests could further
prolonged for both M113N and the double mutant M113N/ reduce single-channel currents. If they could, then the system would
L135N (pH 7.5 (trans): M113N, kCoff . 3:7 3 10 2 2 s 2 1 ; M113N/ constitute a stochastic sensor for organic molecules. A large number
L135N, kCoff . 0:13 s 2 1 ). These results suggest that b-cyclodextrin of readily available adamantane derivatives bind to cyclodextrins12
binds near Met 113 in wild-type a-haemolysin (Fig. 2) and that and were therefore chosen as model analytes. Strikingly, 2-adaman-
when this residue is mutated the binding af®nity is increased at the tanamine (A1, 80 mM trans) reduced the conductance of the
second restriction, which is now formed by Lys 147 and Glu 111. partially blocked channel to 126.5 pS (s:d: ˆ 0:5; n ˆ 7) with a
residence time (tA1) of 2.54 ms (s:d: ˆ 0:21), but had no effect
on the completely open channel (Fig. 1c). A second guest, 1-
P adamantanecarboxylic acid (A2, 20 mM trans), also reduced
the conductance of the partially blocked channel, this time to
C
koff C• Ai
kon Ai '
112.2 pS (s:d: ˆ 3:2, n ˆ 7) with a residence time (tA2) of 14.0 ms
kon (s:d: ˆ 0:8). During each b-cyclodextrin blocking event, guest
C
kon koC•ff Ai C + Ai C Ai
Ai '
koff residence times were independent of guest concentration and the
Ai
kon rates of guest association were linearly dependent on concentration,
PC P C Ai suggestive of a bimolecular interaction. Events due to each guest in a
koAffi
mixture could be distinguished, indicating that they compete for a
Figure 3 A kinetic scheme for the interactions between a-haemolysin (aHL), b- single binding site in the a-haemolysin × b-cyclodextrin complex
cyclodextrin (bCD) and guests. P, aHL pore; C, the adapter bCD; Ai, guests (Fig. 1d). A simple kinetic scheme13 is suf®cient to explain the
(analytes); C×Ai , P×C, P×C×Ai, the various non-covalent complexes of P, C and Ai. observed interactions of analytes in simple solutions or as mixtures

a
-3
-3
Level 3 0 µM
-8
-8

-13
Level 2
-13
-3
-3
Level 3 40 µM
-8
-8

Level 2
I (pA )

-13
-13
-3
-3
Level 3 160 µM
-8
-8

-13
Level 2
-13
-3
-3
Level 3 240 µM
-8
-8

Level 2
-13
-13
Time 20ms
concentrations of A 1 and A 2 (µM)

b c d
A1 A1 A1
40
Experimentally-derived

12 A2 300 A2
A2
PP • C •A(P P • C) -1

[C • A]( µM )

30
8 200
20

4 100
10

0 0 0
0 100 200 300 400 0 1 2 3 4 5 0 100 200 300
[ A] (µM ) [ C ] •[ A] ( x1000 µM 2 ) Actual concentration of A1( µM)
(A 2 concentration fixed)
Figure 4 Response of aHL × bCD at different analyte concentrations. a, aHL × bCD unoccupied aHL pore), [A]0 and [C]0 (the total concentrations of analyte and bCD)
(level 2) to aHL × bCD×A1 (level 3) transitions at various 2-adamantanamine (A1) and KCf (the equilibrium formation constant for the aHL × bCD complex), which can
concentrations. Representative segments of traces are shown depicting entire be measured separately (see Supplementary Information). c, Plots of [C×A] versus
bCD occupancy events at 0, 40, 160 and 240 mM 2-adamantanamine. The [C][A] for 2-adamantanamine (A1) and 1-adamantanecarboxylic acid (A2) using
conditions were as for Fig.1, but with 40 mM bCD. b, Plots of PP×C×A/PP×C versus [A] data from a. The slope yields the formation constant KA9
f for bCD×A from analyte

for 2-adamantanamine (A1) and 1-adamantanecarboxylic acid (A2) using data and bCD in solution. [C×A] ˆ KA9
f [C][A], where values for [C×A], [C] and [A], the

from a. The slope yields the formation constant KAf for aHL × bCD×A from analyte A
and aHL × bCD. For a single analyte, PP×C×A =PP×C ˆ KAf ‰AŠ, where PP×C×A and PP×C are
the experimentally determined probabilities of occurrence of the states of the aHL
pore with both bCD and analyte bound or with only bCD bound, KAf is the
equilibrium formation constant for aHL × bCD×A from analyte A and aHL × bCD,
and [A] is the concentration of free analyte. [A] can be determined from the
experimental values of PP×C and PP (PP is the probability of occurrence of the
concentrations of the bCD×A complex, free bCD and free analyte, respectively,
can be obtained from experimental or known values of PP×C, PA, [A]0, [C]0 and KCf
(see Supplementary Information). d, Analysis of currents from binary solutions of
analytes. The experimentally derived concentrations of 2-adamantanamine (A1)
and 1-adamantanecarboxylic acid (A2), determined by applying equation (1) to
current amplitude histograms, are plotted against the actual concentration of A1.
Conditions are as for Fig. 1, but with 40 mM bCD.

688
© 1999 Macmillan Magazines Ltd NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com

(Fig. 3; details in Supplementary Information). In the case of b-
cyclodextrin and the adamantane derivatives described here, the
residence time of the cyclodextrin in the lumen of the pore is long
compared with the residence time of the analyte in the cyclodextrin.
Therefore, b-cyclodextrin acts as a non-covalent molecular adapter.
When the residence time of the analyte within the cyclodextrin host
is relatively long, the host acts instead as a carrier.
The signals can be used not only to identify analytes but also to
quantify them. As expected, the frequency of occurrence of
aHL×bCD occupancy events by an analyte increases with analyte
concentration (Fig. 4a). Interestingly, the residence time of bCD×2-
adamantanamine on a-haemolysin is greater than b-cyclodextrin
itself (Fig. 4a), which is true for all bCD×A observed in this study.
From the slope of a linear plot (Fig. 4b), KAf values for 1-
adamantanecarboxylic acid and 2-adamantanamine (at pH 3.0)
were found to be, respectively, 1:356 0:19 3 105 M 2 1 and
1:03 6 0:15 3 104 M 2 1 (mean 6 s:d:, n ˆ 7), corresponding to
DG values of -7.0 kcal mol-1 and -5.5 kcal mol-1. Values of KA9 f ,
the equilibrium formation constant for the analyte (A) with
b-cyclodextrin in solution, can be obtained from a second plot
(Fig. 4c). KA9
f values at pH 3.0 for 1-adamantanecarboxylic acid and
2-adamantanamine were, respectively, 5:76 6 1:50 3 104 M 2 1 and
9:84 6 2:19 3 103 M 2 1 (mean 6 s:d:, n ˆ 7), corresponding to DG
values of -6.5 kcal mol-1 and -5.5 kcal mol-1, which are close to the
values found when b-cyclodextrin is bound to the a-haemolysin pore.
Literature values for DG in solution are in the same range as those
we have determined: 1-adamantanecarboxylic acid, -7.5 kcal mol-1
(pH 4.05)14; 2-adamantanamine, -5.3 kcal mol 2 1 (pH 2.5)12. In a
working sensor, the total concentration [Ai]0 of analyte Ai of known
KAi Ai9
f and Kf would be determined from:
‰Ai Š0 ˆ …1=K Ai Ai9 C
f †…1=PP×C ‡ Kf =…P P K f ††P P×C×Ai …
where PP×C×A and PP×C are the experimentally determined probabil-
ities of occurrence of the states of the a-haemolysin pore with both
b-cyclodextrin and analyte bound, or with only b-cyclodextrin
letters to nature
bound, and PP is the probability of occurrence of the unoccupied
a-haemolysin pore (see Supplementary Information for details).
Another important attribute of stochastic sensing is that two or
more analytes, only one of which can occupy the receptor at a given
moment, can be identi®ed and quanti®ed `simultaneously' by a
single sensor element. In a mixture, signals from different analytes
are recognized by their characteristic extents of channel block
and residence times15 (for example, see Table 1 of Supplementary
Information for values for seven adamantanes). To illustrate this
with a-haemolysin and the b-cyclodextrin adapter, we did an
experiment in which 1-adamantanecarboxylic acid (A2) was kept
constant at 20 mM, while 2-adamantanamine (A1) was varied.
Current amplitude histograms, which revealed PP, PP×C and PP×C×Ai,
and equation (1) were used to generate the total concentrations of
the two analytes, [A1]0 and [A2]0. The values obtained were in close
agreement with the actual concentrations in the mixtures (Fig. 4d).
The principle of stochastic sensing with adapter molecules can be
expanded to molecules of biological interest. For example, many
pharmaceuticals interact with cyclodextrins12 and many of those
we have tested give useful signals in our system. For instance, the
tricyclic antidepressant imipramine can be distinguished by using
residence-time histograms from the antihistaminic promethazine,
which has a similar structure (Fig. 5a±c). The use of alternative
cyclodextrins, as illustrated here by the interaction of imipramine
with g-cyclodextrin (Fig. 5d), demonstrates that the system is
programmable. Stochastic sensing with adapter molecules is there-
fore a versatile approach to analyte identi®cation and quanti®ca-
tion. As in olfaction, in which carrier molecules (olfactory binding
proteins) deliver odorants to membrane-bound receptors16, we
could vary both the adapter/carrier and the receptor: for example,
the range of detectable analytes will be extended by using naturally
1† occurring and chemically modi®ed cyclodextrins12 as adapters;
synthetic adapter/carriers could also be considered17±19. Genetically
engineered a-haemolysin pores20, or pores other than a-
haemolysin21,22 could be used to accommodate the alternative

a
100
Number of events

β−CD / 100 µM Promethazine

N 0 pA t =0.3ms

N 10

Promethazine 1
0 1 2 3 4 5 6
b
100
β-CD / 100 µM Imipramine
Number of events

0 pA τ=1.4ms
N
10
N

Imipramine
1
0 1 2 3 4 5 6

c β-CD / Mixture of 100µM Promethaz ine and 100µM Imipramine 100

Number of events

0 pA τ1=0.3ms
τ2=1.4ms
10

1
0 1 2 3 4 5 6
d
γ-CD / 100 µM Imipramine
Number of events

1000
0 pA τ2=0.2ms
100

10
15pA

1
0 1 2 3 4 5 6
5ms Residence time (ms)

Figure 5 Analysis of drug molecules by stochastic sensing with aHL and a bCD or to that of promethazine, but a characteristic signature is provided by the
gCD adapter. a, Signal from 100 mM promethazine in the presence of 40 mM bCD. longer residence time of 1.4 ms. c, Mixture of 100 mM promethazine and 100 mM
Many events with a mean duration of 0.3 ms are observed. b, Signal from 100 mM imipramine in the presence of 40 mM bCD. d, Signal from 100 mM imipramine in the
imipramine in the presence of 40 mM bCD. The extent of channel block is similar presence of 20 mM gCD.

NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com

© 1999 Macmillan Magazines Ltd 689
letters to nature
adapters, which might also be covalently attached to the pores.
Single-molecule ¯uorescence23 or force-detection methods24 might
be feasible for read-out. For applications in the ®eld, rugged
stochastic sensors must be developed, as was recently achieved for
a sensor based on macroscopic channel currents25. M 26. Bhakdi, S., FuÈssle, R. & Tranum-Jensen, J. Staphylococcal a-toxin: oligomerization of hydrophilic
.........................................................................................................................
Methods
Pore formation. Heptameric wild-type a-haemolysin formed by treating the
monomer, puri®ed from S. aureus, with deoxycholate26,27 was isolated from
SDS-polyacrylamide gels as described6.
Planar bilayer recordings. A bilayer of 1,2-diphytanoylphosphatidylcholine
(Avanti Polar Lipids) was formed on a 100- to 150-mm ori®ce in a 25-mm-thick
Te¯on ®lm (Goodfellow) separating two compartments (2 ml each) of a planar
bilayer apparatus28. Solutions in the compartments contained 1 M NaCl, 5 mM
EDTA, 10 mM sodium phosphate, at pH 7.5 in the cis chamber and pH 3.0 in
the trans chamber. Neutral-pH solutions in the cis chamber prevent closure of
a-haemolysin channels; low-pH solutions in the trans chamber increase the
residence time of cyclodextrins. Heptameric a-haemolysin (0.5 to 1 ml at 50 to
200 ng ml-1) was added to the cis compartment, which was stirred until a single
channel inserted into the bilayer. Currents were recorded at a holding potential
of -40 mV (cis at ground) by using a patch-clamp ampli®er (Axopatch 200B,
Axon Instruments). Currents were low-pass ®ltered with a built-in 4-pole
Bessel ®lter at 5 kHz and sampled at 20 kHz by computer with a Digidata 1200
A/D converter (Axon Instruments). Cyclodextrins (Aldrich) and analytes
(Aldrich) were added to the trans chamber.
Data analysis. Current amplitude and residence (dwell) time histograms were
constructed using pClamp 6.0 software (Axon Instruments). Probabilities, PP,
PP×C and PP×C×A, were obtained from the amplitude histograms after ®tting the
peaks to gaussian functions. Mean residence times (t values) were obtained
from residence time histograms by ®tting the distributions to exponential
functions. Measurements are given as the mean 6 s.d.
Molecular graphics. The structures of a-haemolysin were generated from the
coordinates8 by using SPOCK 6.3 software29. The b-cyclodextrin structure was
generated with Insight II 97.0. The two structures were scaled and super-
imposed in Adobe Photoshop 4.0.
Received 6 January; accepted 3 March 1999.
1. Dickinson, T. A., White, J., Kauer, J. S. & White, D. R. A chemical-detecting system based on a cross-
reactive optical sensor array. Nature 382, 687±700 (1996).
2. Lonergan, M. C. et al. Array-based vapor sensing using chemically sensitive, carbon black-polymer
resistors. Chem. Mat. 8, 2298±2312 (1996).
3. Hellinga, H. W. & Marvin, J. S. Protein engineering and the development of generic biosensors. Trends
Biotechnol. 16, 183±189 (1998).
4. Czarnik, A. W. A sense for landmines. Nature 394, 417±418 (1998).
5. Crooks, R. M. & Ricco, A. J. Special issue on chemical sensors. Acc. Chem. Res. 31, 199±324 (1998).
6. Braha, O. et al. Designed protein pores as components for biosensors. Chem. Biol. 4, 497±505 (1997).
7. Gouaux, E. a-Hemolysin from Staphylococcus aureus: an archetype of b-barrel, channel-forming
toxins. J. Struct. Biol. 121, 110±122 (1998).
8. Song, L. et al. Structure of staphylococcal a-hemolysin, a heptameric transmembrane pore. Science
274, 1859±1865 (1996).
9. Krasilnikov, O. V., Sabirov, R. Z., Ternovsky, V. I., Merzliak, P. G. & Muratkhodjaev, J. N. A simple
method for the determination of the pore radius of ions channels in planar lipid bilayer membranes.
FEMS Microbiol. Immunol. 105, 93±100 (1992).
10. Bezrukov, S. M., Vodyanoy, I., Brutyan, R. A. & Kasianowicz, J. J. Dynamics and free energy of polymer
partitioning into a nanoscale pore. Macromolecules 29, 8517±8522 (1996).
11. Kasianowicz, J. J., Brandin, E., Branton, D. & Deamer, D. W. Characterization of individual
polynucleotide molecules using a membrane channel. Proc. Natl Acad. Sci. USA 93, 13770±13773
(1996).
12. Rekharsky, M. V. & Inoue, Y. Complexation thermodynamics of cyclodextrins. Chem. Rev. 98, 1875±
1917 (1998).
13. Colquhoun, D. & Hawkes, A. G. On the stochastic properties of bursts of single ion channel openings
and of clusters of bursts. Phil. Trans. R. Soc. Lond. B 300, 1±59 (1982).
14. Inoue, Y. et al. Thermodynamics of molecular recognition by cyclodextrins. 1. Calorimetric titration
of inclusion complexation of naphthalenesulfonates with a-, b-, and g-cyclodextrins: enthalphy±
entropy compensation. J. Am. Chem. Soc. 115, 475±481 (1993).
15. Lucchesi, K., Ravindran, A., Young, H. & Moczydlowski, E. Analysis of the blocking activity of
charybdotoxin homologs and iodinated derivatives against Ca2+-activated K+ channels. J. Membr.
Biol. 109, 269±281 (1989).
16. Bianchet, M. A. et al. The three-dimensional structure of bovine odorant binding protein and its
mechanism of odor recognition. Nature Struct. Biol. 3, 934±939 (1996).
17. Chen, H., Weiner, W. S. & Hamilton, A. D. Recognition of neutral species with synthetic receptors.
Curr. Opin. Chem. Biol. 1, 458±466 (1997).
18. Arduini, A., Casnati, A., Pochini, A. & Ungaro, R. Recognition of cationic species with synthetic
receptors. Curr. Opin. Chem. Biol. 1, 467±474 (1997).
19. Beer, P. D. & Schmitt, P. Molecular recognition of anions by synthetic receptors. Curr. Opin. Chem.
Biol. 1, 475±482 (1997).
20. Bayley, H. Building doors into cells. Sci. Am. 277, (Sept.) 62±67 (1997).
21. Hartgerink, J. D., Clark, T. D. & Ghadiri, M. R. Peptide nanotubes and beyond. Chem. Eur. J. 4, 1367±
1372 (1998).
690
© 1999 Macmillan Magazines Ltd
22. Schmid, B., Maveyraud, L., Kromer, M. & Schulz, G. E. Porin mutants with new channel properties.
Protein Sci. 7, 1603±1611 (1998).
23. Lu, H. P., Xun, L. & Xie, X. S. Single-molecule enzymatic dynamics. Science 282, 1877±1882 (1998).
24. Oberhauser, A. F., Marszalek, P. E., Erickson, H. P. & Fernandez, J. M. The molecular elasticity of the
extracellular matrix protein tenascin. Nature 393, 181±185 (1998).
25. Cornell, B. A. et al. A biosensor that uses ion-channel switches. Nature 387, 580±583 (1997).
monomers to form amphiphilic hexamers induced through contact with deoxycholate micelles. Proc.
Natl Acad. Sci. USA 78, 5475±5479 (1981).
27. Walker, B. J., Krishnasastry, M., Zorn, L., Kasianowicz, J. J. & Bayley, H. Functional expression of the
a-hemolysin of Staphylococcus aureus in intact Escherichia coli and in cell lysates. J. Biol. Chem. 267,
10902±10909 (1992).
28. Montal, M. & Mueller, P. Formation of bimolecular membranes from lipid monolayers and study of
their electrical properties. Proc. Natl Acad. Sci. USA 69, 3561±3566 (1972).
29. Christopher, J. A. SPOCK: The Structural Properties Observation and Calculation Kit (Program
Manual) (Center for Macromolecular Design, Texas A&M Univ., College Station, 1998).
Supplementary information is available on Nature's World-Wide Web site (http://www.nature.com) or
as paper copy from the London editorial of®ce of Nature.
Acknowledgements. This work was supported by DARPA, DOE and ONR. We thank E. Gouaux for
comments on the manuscript.
Correspondence and requests for materials should be addressed to O.B. (e-mail: obraha@medicine.
tamu.edu) or H.B. (e-mail: bayley@tamu.edu).
Present and future trends in
the atmospheric burden of
ozone-depleting halogens
S. A. Montzka*, J. H. Butler*, J .W. Elkins*, T. M. Thompson*,
A. D. Clarke²³ & L. T. Lock²
* National Oceanic and Atmospheric Administration, Climate Monitoring and
Diagnostics Laboratory, 325 Broadway, Boulder, Colorado 80303, USA
² Cooperative Institute for Research in Environmental Science, University of
Colorado, Boulder, Colorado 80309, USA
.........................................................................................................................
The burden of ozone-depleting chemicals in the lower atmosphere
has been decreasing since 1994 as a result of the Montreal
Protocol1±3. Here we show how individual chemicals have in¯u-
enced this decline, in order to estimate how the burden could
change in the near future. Our measurements of atmospheric
concentrations of the persistent, anthropogenic chemicals that
account for most ozone-depleting halogens in today's strato-
sphere show that the decline stems predominantly from the
decrease in the atmospheric load of trichloroethane (CH3CCl3),
a previously common cleaning solvent. The in¯uence of this
chemical on the decline has now peaked, however, and will
become much smaller over the next ®ve to ten years. As this
in¯uence lessens, a decrease in the burden of ozone-depleting
halogen will be sustained only if emissions of other halocarbons
fall. Although emissions of most gases regulated by the Montreal
Protocol have decreased substantially over the past ten years (refs
4±11), emissions of the potent ozone-depleting gas CBrClF2
(halon-1211) have remained fairly constant during this
period12,29, despite stringent limits on production in developed
countries since 1994. The consequent atmospheric accumulation
of this halon is retarding the decline of ozone-depleting halogens
in the atmosphere more than any other persistent gas.
Measurements of chlorinated and brominated trace gases at
remote locations on the Earth's surface provide constraints on the
burden of ozone-depleting chemicals in the future strato-
sphere1,4,13,14. Here we relate tropospheric halocarbon burdens to
the future stratospheric abundance of ozone-depleting halogens by
considering halocarbon destruction rates in the stratosphere and by
recognizing that bromine is 50 times more effective than chlorine
in ozone-depleting reactions4 (Table 1). The aggregate halocarbon
burden weighted in this way, which is called ``effective equivalent
³ Present address: National Oceanic and Atmospheric Administration, Climate Monitoring and Diag-
nostics Laboratory, South Pole, Antarctica.
NATURE | VOL 398 | 22 APRIL 1999 | www.nature.com