Full text of DOI:10.1038/sj.bjp.0704660

British Journal of Pharmacology (2002) 135, 1943 ± 1950
ã
2002 Nature Publishing Group All rights reserved 0007 ± 1188/02 $25.00
www.nature.com/bjp
Direct visualization of ligand-protein interactions using atomic
force microscopy
2
1
1Calum S. Neish, Ian L. Martin, Robert M. Henderson & *,1J. Michael Edwardson
2
1Department of Pharmacology, University of Cambridge, Cambridge CB2 1PD and Pharmaceutical Sciences Research
Institute, School of Life and Health Sciences, Aston University, Aston Triangle, Birmingham B4 7ET
1
Streptavidin is a 60-kDa tetramer which binds four molecules of biotin with extremely high
anity (KA*1014 71). We have used atomic force microscopy (AFM) to visualize this ligand-
protein interaction directly.
Biotin was tagged with a short (152-basepair; 50-nm) DNA rod and incubated with streptavidin.
M
2
The resulting complexes were then imaged by AFM. The molecular volume of streptavidin
calculated from the dimensions of the protein particles (105+3 nm3) was in close agreement with the
value calculated from its molecular mass (114 nm3). Biotinylation increased the apparent size of
streptavidin (to 133+2 nm3), concomitant with an increase in the thermal stability of the tetramer.
3
Images of streptavidin with one to four molecules of DNA-biotin bound were obtained. When
two ligands were bound, the angle between the DNA rods was either acute or obtuse, as expected
from the relative orientations of the biotin binding sites. The ratio of acute : obtuse angles (1 : 3) was
lower than the expected value (1 : 2), indicating a degree of steric hindrance in the binding of the
DNA-biotin. The slight under-representation of higher occupancy states supported this idea.
4
Streptavidin with a single molecule of DNA-biotin bound was used to tag biotinylated
b-galactosidase, a model multimeric enzyme.
The ability to image directly the binding of a ligand to its protein target by AFM provides useful
5
information about the nature of the interaction, and about the e€ect of complex formation on the
structure of the protein. Furthermore, the use of DNA-biotin/streptavidin tags could potentially
shed light on the architecture of multi-subunit proteins.
British Journal of Pharmacology (2002) 135, 1943 ± 1950
Keywords: Biotin; streptavidin; protein complex; AFM; single-molecule imaging
Abbreviations: AFM, atomic force microscopy; PCR, polymerase chain reaction
Introduction
The interaction of receptors with small ligands is most
commonly studied using radioligand binding. This approach
is well-documented, and provides valuable quantitative
information, such as the anity of the binding interaction.
Radioligand binding cannot, however, shed any light on the
structure of the ligand-receptor complex, or on the e€ect of
ligand binding on receptor structure. These are important
issues, especially as it is becoming clear that the structure and
function of many proteins are modulated by ligand binding.
For instance, the binding of two molecules of acetylcholine to
the nicotinic receptor causes the ®ve-subunit assembly to
open a central cation channel (Karlin, 1987; Unwin, 1998).
We are developing methods for imaging proteins under near-
physiological conditions using atomic force microscopy
(AFM; see, for example, Ellis et al., 1999a,b; Berge et al.,
2000). In the present study we have used AFM to visualize
directly the binding of a small ligand to a multi-subunit
protein.
(Green, 1990). The complex of streptavidin with biotin has
been crystallized, and its structure solved at high
resolution (Weber et al., 1989; Hendrickson et al., 1989).
Furthermore, extensive mutational analysis has been
carried out to identify the requirements and consequences
of ligand binding (Freitag et al., 1998; 1999; Klumb et al.,
1998; Stayton et al., 1999). The protein is tetrameric and
has 222 point symmetry. The tetramer is composed of two
dimers, within which the monomers are in intimate
contact. In contrast, the dimers interact over a much less
extensive interface (Weber et al., 1989; Hendrickson et al.,
1989). The streptavidin tetramer binds four molecules of
biotin with extremely high anity (KD*10714 M; Green,
1990).
We have modi®ed the ligand, biotin, by addition of a short
(152-basepair; 50-nm) DNA rod, which enables its visualiza-
tion by AFM. We have then studied the binding of the DNA-
biotin to individual molecules of streptavidin. We report the
characteristics of the binding, and the consequences of the
binding for the structure of the protein. Finally, we
demonstrate the use of the DNA-biotin/streptavidin complex
to tag biotinylated b-galactosidase. We suggest that this
approach will be applicable to the study of the architecture of
multi-subunit protein assemblies.
The interaction between biotin and streptavidin provides
an ideal system for our purpose. Streptavidin is a 60-kDa
protein produced by the bacterium Streptomyces avidinii
*Author for correspondence; E-mail: jme1000@cam.ac.uk

1944
C.S. Neish et al
Ligand-protein interaction visualized by AFM
Methods
Sample preparation
where h is the particle height and r is the radius at half height
(Schneider et al., 1998). Molecular volume based on
molecular weight was calculated using the equation:
Vc ˆ ꢀM0=N0†ꢀV1 ‡ dV2†
where M0 is the molecular mass of the protein, N0 is
Avogadro's number, V1 and V2 are the partial speci®c
volumes of protein and water (0.74 cm3 g71 and 1 cm3 g71
respectively), and d is the extent of protein hydration
(0.4 mol water/mol protein) (Edstrom et al., 1990).
Values quoted are means+s.e.mean. The statistical sig-
ni®cance of di€erences between means was assessed using
Student's t-test for unpaired data.
ꢀ2†
Biotinylated 152-basepair DNA rods were produced by
polymerase chain reaction (PCR) ampli®cation, using Taq
polymerase (Promega, Southampton, UK), with pUB 110
plasmid DNA (Sigma, Poole) as template and the primers: 5'-
biotin-(15-carbon)-CATCATTCGGCAAATCCTTGAGCC-
3' (forward) and 5'-GGCATTCAGTCACAAAAGGTTGTT-
GC-3' (reverse) (Synthegen, Houston, TX, U.S.A.). The
DNA-biotin was puri®ed using a QIAquick PCR puri®cation
kit (Qiagen, Crawley, U.K.), and eluted in MilliQ ultrapure
water (Millipore Systems, Watford, U.K.) at a concentration
of 500 nM.
,
Probabilities of the various biotinylation states
Streptavidin was purchased from Sigma and used without
further puri®cation. Lyophilized streptavidin was reconsti-
tuted in 50 mM potassium phosphate bu€er pH 7.8, to a
concentration of 10 mM. For imaging, streptavidin was
The theoretical probabilities of the various biotinylation
states of streptavidin (i.e. unoccupied and occupied by 1 ± 4
molecules of biotin) was calculated using the binomial
distribution, assuming that all four binding sites are
equivalent. The probability of binding is given by:
diluted to
a ®nal concentration of 1 ± 10 nM. Where
appropriate, streptavidin and DNA-biotin were incubated
together at a known molar ratio in a total volume of 100 ml
at 228C for 1 h. All dilutions were carried out using MilliQ
water.
Biotinylated b-galactosidase from Escherichia coli, contain-
ing on average 2 ± 4 biotin molecules per enzyme tetramer,
was purchased from Sigma and used without further
puri®cation. Lyophilized b-galactosidase was reconstituted
by addition of water, to yield a 5 ± nM protein solution in
1 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM MgCl2 bu€er.
Where appropriate, b-galactosidase and DNA-biotin/strepta-
vidin were incubated together at a known molar ratio at 228C
for 1 h.
P ˆ NB=NS
ꢀ3†
where NB is the number of occupied sites and NS is the total
number of sites. The probabilities of the various occupation
states are then as follows:
4
Unoccupied:
One ligand bound:
Two ligands bound:
P0 ˆ ꢀ1 P†
P1 ˆ 4ꢀ1 P†3P
P2 ˆ 6ꢀ1 P†2P2
Three ligands bound: P3 ˆ 4ꢀ1 P†P3
Four ligands bound: P4 ˆ P4
Determination of the thermal stability of streptavidin
AFM imaging
Biotin (Sigma) was mixed with streptavidin (10 mM) at
biotin : streptavidin ratios of 0.25 : 1, 0.5 : 1, 1 : 1 or 2 : 1
(total volume 20 ml). These solutions were then incubated at
either 22 or 1008C for 30 min. Samples were analysed by
SDS-polyacrylamide gel electrophoresis, and proteins were
visualized by staining with Coomassie brilliant blue.
Samples were imaged on freshly cleaved, poly-L-lysine
(Sigma)-coated, mica coverslips (Goodfellow, Cambridge,
U.K.). Before AFM imaging 50 ml of the sample was allowed
to adsorb to the surface. After 10 min the sample was washed
with water and air-dried. Imaging was performed with a
Multimode atomic force microscope (Digital Instruments,
Santa Barbara, CA, U.S.A.). Samples were imaged in air,
using tapping mode with a root mean square amplitude of
0.7 V and a drive frequency of *300 kHz. Silicon cantilevers
with a speci®ed spring constant of 42 N m71 were used
(NCH Pointprobes; Nanosensors, Wetzlar-Blankenfeld, Ger-
many). Images were obtained using intermittent tip-sample
contact 50.1 nm apart, with an applied force of approxi-
mately 200 pN, which produces little lateral movement or
distortion of the sample.
Results
Images of the two principal reagents used in this study are
shown in Figure 1. Unoccupied streptavidin appeared as a
homogenous spread of globular particles (Figure 1A), and
DNA-biotin appeared as 50-nm rods (Figure 1B). The
observed length of the DNA rods is as expected for a 152-
basepair DNA duplex (Sambrook et al., 1989). The forward
primer used in the PCR ampli®cation of the DNA has a
biotin group at its 5' end, separated from the ®rst base by a
15-carbon (2.3-nm) spacer. The inclusion of the spacer was
found to be essential for ecient binding of the DNA-biotin
to streptavidin, presumably because the biotin binding cleft is
embedded in the protein structure (Weber et al., 1989;
Hendrickson et al., 1989). The biotin/15-carbon `tail' was not
visible in the images of the DNA rods.
Molecular volume calculation
The molecular volume of the protein particles was determined
from particle dimensions derived from AFM images. The
height and half-height diameters were measured from multi-
ple cross-sections of the same particle, and the molecular
volume of each particle was calculated using the following
equation, which treats the particle as a spherical cap:
The molecular dimensions of unoccupied streptavidin were
compared with the dimensions of the biotinylated protein.
Vm ˆ ꢀꢀh=6†ꢀ3r2 ‡ h2†
ꢀ1†
British Journal of Pharmacology vol 135 (8)

C.S. Neish et al
Ligand-protein interaction visualized by AFM
1945
The heights and diameters of a number of particles were
determined and used to calculate molecular volumes. Particle
diameter was measured at half the maximal height in order to
compensate for the tendency of AFM to overestimate this
parameter because of the geometry of the tip (Larmer et al.,
1997; Schneider et al., 1998; Ellis et al., 1999b; Berge et al.,
2000). Mean values for particle height, diameter and volume
are given in Table 1. A representative image of unoccupied
streptavidin is shown in Figure 2A. The distribution of
calculated molecular volumes (Figure 2B) is Gaussian, and
the mean volume (105+3 nm3; n=237) is very close to the
value of 114 nm3 calculated from the molecular mass of the
protein (60 kDa), indicating that the particles detected are
indeed streptavidin tetramers. When streptavidin was imaged
after incubation with biotin,
a homogenous spread of
particles was seen, as for unoccupied streptavidin (Figure
2C). Interestingly, the molecular volume calculated for the
biotinylated streptavidin (141+11 nm3; n=130; Figure 2D,
Table 1) was signi®cantly (P50.0001) greater than the value
obtained for unoccupied protein. When streptavidin was
incubated with DNA-biotin, and then imaged by AFM,
DNA rods of length 50 nm could be seen protruding from
the streptavidin molecules. At a molar ratio of DNA-biotin
to streptavidin of 1 : 1, the most common structure observed
was streptavidin occupied by a single DNA-biotin molecule,
giving the imaged particles a tadpole-like appearance; at a
ratio of 10 : 1 singly- and doubly-liganded particles were
equally common (Figure 2E). As for streptavidin complexed
with untagged biotin, the apparent molecular volume of
streptavidin occupied by DNA-biotin (133+2 nm3; n=210;
Figure 2F, Table 1) was signi®cantly (P50.0001) greater than
that of unoccupied streptavidin. In contrast, the molecular
volumes of streptavidin complexed with biotin and with
DNA-biotin were not di€erent from each other. As indicated
above, the standard error for the volume of the biotin/
streptavidin complex is larger than the errors for unoccupied
streptavidin and DNA-biotin/streptavidin. Two factors are
likely to contribute to this larger error: ®rst, the number of
samples analysed was smaller, and second, there is likely to
be
a
small population of unbiotinylated streptavidin
molecules in the `biotinylated' sample that cannot be
identi®ed. Since the anity of streptavidin for biotin is very
high, this population will be very small; however, it might
lead to a slight underestimation of the molecular volume of
the biotinylated streptavidin, and an increase in the standard
error.
The observed increase in the apparent molecular volume of
streptavidin on biotinylation was accompanied by an increase
in its thermal stability, which has been reported previously
(Gonzalez et al., 1997; 1999). When streptavidin was incubated
for 30 min at room temperature (228C) with biotin at various
biotin : streptavidin molar ratios and then analysed by SDS ±
PAGE under non-reducing conditions, the protein migrated
exclusively as a tetramer (apparent molecular mass 66 kDa) in
all cases (Figure 3A). In contrast, when the unoccupied protein
was incubated for 30 min at 1008C, it ran almost exclusively as
a monomer (apparent molecular mass 14 kDa; Figure 3B). As
the molar ratio of biotin : streptavidin was increased, however,
the proportion of the protein migrating as a tetramer became
progressively greater.
Figure 1 Streptavidin and DNA-biotin imaged by tapping-mode
AFM in air. (A) Streptavidin at a concentration of 5 nM was allowed
to adsorb to mica at 228C for 10 min. The mica was then washed
with water and air-dried. (B) DNA-biotin at a concentration of
50 nM was adsorbed to mica as in (A). Scale bars: 50 nm. A shade-
height scale is shown at the bottom.
The relative orientation of the biotin binding pockets in the
streptavidin tetramer provides two possible angles between
Table 1 Dimensions of unoccupied and biotinylated streptavidin molecules, as determined by tapping-mode AFM in air
Height (nm)
Diameter (nm)
Volume (nm3)
n
Unoccupied streptavidin
Streptavidin occupied by biotin
Streptavidin occupied by DNA-biotin
2.04+0.03
2.54+0.08
2.31+0.02
11.3+0.2
11.1+0.1
11.1+0.1
105+3
141+11
133+2
237
130
210
British Journal of Pharmacology vol 135 (8)

1946
C.S. Neish et al
Ligand-protein interaction visualized by AFM
Figure 2 E€ect of biotinylation on the structure of streptavidin. (A) Scan of unoccupied streptavidin (5 nM) imaged by tapping-
mode AFM in air. (B) Distribution of calculated molecular volumes of unoccupied streptavidin. Volumes were placed in 25-nm3
bins. (C) Scan of streptavidin complexed with biotin. Streptavidin (5 nM) was incubated with biotin (10 nM) for 1 h at 228C before
adsorption to mica. (D) Distribution of calculated molecular volumes of biotinylated streptavidin. (E) Scan of streptavidin
complexed with DNA-biotin. Streptavidin (5 nM) was incubated with DNA-biotin (50 nM) for 1 h at 228C before adsorption to
mica. (F) Distribution of calculated molecular volumes of streptavidin complexed with DNA-biotin. Scale bars: 50 nm. A shade-
height scale is shown at the bottom.
pairs of bound DNA-biotin molecules ± an acute angle
between ligands at adjacent monomers separated by a `weak'
interface, and obtuse angles between ligands either at
adjacent monomers separated by a `strong' interface or at
diagonally opposite monomers (Weber et al., 1989; Katz,
1997). There is no indication of co-operativity in the binding
of biotin to the streptavidin tetramer (Gonzalez et al., 1997),
and so all binding sites should be equivalent. Consequently,
British Journal of Pharmacology vol 135 (8)

C.S. Neish et al
Ligand-protein interaction visualized by AFM
1947
Figure 3 E€ect of biotinylation on the thermal stability of strepta-
vidin. Streptavidin, at a concentration of 10 mM, was incubated with
biotin at biotin : streptavidin molar ratios of 0 : 1, 0.25 : 1, 0.5 : 1, 1 : 1
or 2 : 1 for 30 min at either 228C (A) or 1008C (B). Samples were
analysed by SDS-polyacrylamide gel electrophoresis, and proteins
were visualized by staining with Coomassie brilliant blue. The
positions of molecular mass markers (kDa) are shown on the right.
one can predict the relative frequencies of acute and obtuse
angles between pairs of ligands. Once one ligand has bound,
there is only one site at which binding of the second can
produce an acute angle, but two sites at which binding can
produce an obtuse angle. Hence, for streptavidin bound by two
DNA-biotin molecules, the expected ratio of acute : obtuse
angles between the DNA rods is 1 : 2. The ability to visualize
individual DNA-biotin tags bound to streptavidin molecules
a€orded by AFM allowed us to test these predictions. The 50-
nm DNA rods are not completely rigid; nevertheless, it is a
relatively simple matter to distinguish acute and obtuse angles
between pairs of tags, and to place each particular angle into
one or other category. Figure 4 shows representative images of
unoccupied streptavidin and of streptavidin occupied by one,
two, three and four molecules of DNA-biotin. Figures 4C,D
show examples of streptavidin occupied by two DNA-biotin
molecules at an acute and an obtuse angle, respectively. Figure
4F shows an example of a streptavidin molecule tagged by four
DNA-biotin molecules, in which the expected acute and obtuse
angles between the DNA rods can be clearly seen. When a
population of streptavidin molecules (n=80) tagged by two
DNA-biotin molecules was analysed, it was found that 25% of
the angles between the rods were acute and 75% were obtuse.
Hence the ratio of acute : obtuse angles was 1 : 3 and not 1 : 2,
as expected. The simplest interpretation of this result is that
there is steric hindrance between the relatively large ligands,
such that binding at adjacent sites on the tetramer is not
favoured. An alternative possibility is that the under-
representation of acute angles is caused by mutual repulsion
Figure 4 AFM images of streptavidin complexed with multiple
DNA-biotin ligands. (A) Unoccupied streptavidin. (B) Streptavidin
occupied by one DNA-biotin molecule. (C) Streptavidin occupied by
two DNA-biotin molecules, with the DNA rods at an acute angle.
(D) Streptavidin occupied by two DNA-biotin molecules, with the
DNA rods at an obtuse angle. (E) Streptavidin occupied by three
DNA-biotin molecules. (F) Streptavidin occupied by four DNA-
biotin molecules, with two acute angles and two obtuse angles
between the DNA rods. Scale bar: 25 nm. A shade-height scale is
shown at the bottom.
between the DNA rods after the DNA-biotin has bound to
streptavidin, which might occur because the DNA carries a
signi®cant net negative charge. We believe this latter scenario to
be unlikely, because it would be expected to produce a
continuum of angles between acute and obtuse, as a result of
variation in the extent of repulsion between the DNA rods in
di€erent complexes, rather than the two distinct geometries
actually observed.
An alternative method of testing whether DNA-biotin
binding to streptavidin shows steric hindrance is to
determine the distribution of the di€erent occupancy states
of streptavidin (i.e., unoccupied, and occupied by one, two,
three and four molecules of DNA-biotin), and to compare
this with the theoretical distribution, calculated assuming
that all binding sites are equivalent. When streptavidin and
DNA-biotin were incubated together at the same concentra-
British Journal of Pharmacology vol 135 (8)

1948
C.S. Neish et al
Ligand-protein interaction visualized by AFM
Figure 5 Frequencies of the various forms of DNA-biotin liganding
at DNA-biotin : streptavidin molar ratios of 1 : 1 (A) and 10 : 1 (B),
compared with theoretical values. The concentration of streptavidin
was 5 nM in each case. The histograms show the actual (®lled bars)
and theoretical (open bars) percentages of the various occupation
states of streptavidin. Eight (A) or ®ve (B) separate samples,
containing a total of 1335 (A) or 1970 (B) streptavidin particles,
were analysed. Errors refer to the variation between samples. The
overall occupancies of the binding sites were 20+0.02% (A) and
34+0.01% (B).
Figure 6 Tagging of b-galactosidase by DNA-biotin/streptavidin.
(A) b-galactosidase was reconstituted to a concentration of 5 nM,
bound to mica and imaged immediately. (B) b-galactosidase was
reconstituted to a concentration of 5 nM and left at 228C for 6 h
before imaging. (C ± F) Streptavidin (5 nM) was incubated with
DNA-biotin (5 nM) for 1 h at 228C to produce complexes containing
predominantly 1 DNA rod. These complexes were incubated for a
further 1 h at 228C with b-galactosidase (0.5 nM), and then bound to
tion (5 nM) and the resulting complexes analysed by AFM,
it was found that the total occupancy of biotin binding sites
was 20+0.02% (n=8), close to the maximum possible
occupancy of 25%. The distribution of the di€erent
occupancy states, in comparison with the theoretical values,
is shown in Figure 5A. It can be seen that the percentage of
mica and imaged. Large arrows in
C indicate b-galactosidase
monomers tagged with single DNA-biotin/streptavidin complexes.
Small arrows indicate unattached DNA-biotin/streptavidin com-
plexes. The arrowhead in E indicates a `neck' on the protein complex
produced by the attachment of a DNA-biotin/streptavidin molecule.
Scale bars: 50 nm.
the streptavidin molecules having
a single DNA-biotin
bound was higher than predicted, and that the percentages
of protein molecules having larger numbers of ligands bound
(two, three and four) were all under-represented. This result
is in agreement with the previous observation regarding the
angles between the DNA rods, and supports the suggestion
that steric hindrance between the ligands a€ects the binding.
When the ratio of the concentrations of DNA-biotin to
streptavidin was increased to 10 : 1 by raising the DNA-
biotin concentration to 50 nM, the total occupancy of biotin
binding sites increased to 34+0.01% (n=5). In addition, the
percentages of streptavidin molecules having multiple
occupancy by DNA-biotin was shifted closer to the
predicted values (Figure 5B), indicating that the higher
concentration of DNA-biotin was tending to overcome the
interference between the ligands.
Following incubation of streptavidin with DNA-biotin at a
1 : 1 molar ratio, 58% of the streptavidin molecules became
occupied by a single ligand molecule. Further occupation of
the streptavidin was then found to occur preferentially at
sites on the opposite side of the molecule from the original
binding site. These binding properties lend themselves to the
use of the DNA-biotin/streptavidin complex as a reagent for
tagging biotinylated proteins. In order to investigate this
possibility, we examined the interaction between a biotiny-
lated target protein, the enzyme b-galactosidase, and
streptavidin occupied primarily by single DNA-biotin
molecules. b-galactosidase from Escherichia coli is a tetra-
British Journal of Pharmacology vol 135 (8)

C.S. Neish et al
Ligand-protein interaction visualized by AFM
1949
meric protein of molecular mass 540 kDa, composed of four
identical monomers (Cohen & Mire, 1971). As shown in Figure
6A, b-galactosidase imaged immediately after reconstitution in
aqueous bu€er appeared as an homogenous spread of large
particles, of molecular volume 1752+66 nm3 (n=60). As the
time after reconstitution of the b-galactosidase increased, the
large particles disappeared and were replaced by smaller
particles of molecular volume 483+32 nm3 (n=60), consistent
with the breakdown of the tetramers into monomers (Figure
6B). This instability of the b-galactosidase tetramer in EDTA-
containing bu€ers has been reported previously (Kaneshiro et
al., 1975). The theoretical molecular volume of the tetramer,
based on its molecular mass, is 1026 nm3. This is somewhat
smaller than the volume obtained from the AFM images, for
reasons that are at present unclear. Streptavidin tagged
predominantly by single molecules of DNA-biotin was
prepared as in Figure 5A. The DNA-biotin/streptavidin
complexes were then incubated with b-galactosidase for a
further 1 h at 228C and imaged by AFM. As shown in Figure
6C ± F, the b-galactosidase, now predominantly in its mono-
meric form, became tagged with the DNA-biotin/streptavidin
complexes. Free DNA-biotin/streptavidin was also visible
(Figure 6C). The streptavidin and b-galactosidase molecules
were not usually discernible, although in a few instances (see,
for example, Figure 6E), the streptavidin molecule with its
protruding DNA rod appeared as a `neck' on the image of the
complex. The mean molecular volume of the protein complex
(672+50 nm3, n=29) was close to the combined volumes of the
two proteins (133 nm3+483 nm3=616 nm3). As expected from
the extent of biotinylation of the b-galactosidase (on average
2 ± 4 moles of biotin per mole of tetramer), only single DNA
rods were observed attached to the complex. No DNA-tagged
complexes were seen when unbiotinylated b-galactosidase was
used (not shown).
was introduced recently in an attempt to minimize this error
(Schneider et al., 1998). Although the diameters (and therefore
volumes) calculated in this way must necessarily be regarded as
approximations, this method has actually been remarkably
successful. For example, Schneider et al. (1998) have shown a
very good correlation between predicted and calculated
molecular volumes, over a range of protein molecular masses
from 38 ± 900 kDa. It was also found that there was no
signi®cant di€erence between the molecular volumes obtained
in air and under ¯uid. In the present study also, the calculated
molecular volume of the streptavidin tetramer was in good
agreement with the predicted value, although the correlation
for b-galactosidase was less impressive.
The structure of strepatavidin has been solved to atomic
resolution. Biotin binds in clefts at the ends of each of the
four subunits that constitute the streptavidin tetramer (Weber
et al., 1989; Hendrickson et al., 1989). The clefts are lined
predominantly with aromatic and polar amino acid residues.
Biotin binding causes the displacement of bound water from
the cleft and involves numerous interactions between the
biotin molecule and the residues lining the cleft. Interestingly,
bound biotin interacts with tryptophan-120 of an adjacent
subunit, which strengthens intersubunit interactions at the
`weak' interface and is likely to contribute to the increase in
thermal stability of the tetramer (Sano & Cantor, 1995). In
addition, the bound biotin becomes partially buried through
the ordering and movement of a surface loop (amino acid
residues 45 ± 50), which is disordered in the unliganded
protein. This ordering of loop residues is likely to produce
a further enhancement of tetramer stability, this time at the
`strong' interface (Katz, 1997). These e€ects of biotinylation
result in a change in the quaternary structure of the protein.
For instance, there is a slight increase in the angle between
the subunits within each dimer, and a consequent rotation of
the dimers relative to the tetramer dyad axis (Weber et al.,
1989).
Discussion
In the present study, we have shown that the change in
quaternary structure of streptavidin on biotinylation results
in an increase in thermal stability, and is detected in the
AFM images as an increase in apparent molecular volume.
This increase in apparent volume is of interest, since the
structural changes described above would seem more likely to
cause a reduction in volume. Curiously, we found that
biotinylation had no signi®cant e€ect on the measured
diameter of the streptavidin particles, whereas the measured
height was signi®cantly increased, accounting for the overall
increase in the calculated volume (see Table 1). An intriguing
possibility arising out of these ®ndings is that at least part of
the increase in measured height might be due to a reduction
in the deformability of the protein (i.e. the scanning tip,
which presses downwards as it passes over the sample, might
be less able to `squash' the streptavidin tetramer when the
protein is biotinylated). This property would be entirely
consistent with the known increase in structural ordering of
the protein on biotinylation. We suggest, therefore, that
factors such as the `hardness' of the protein might contribute
to the molecular volumes determined by AFM.
We have used AFM to image proteins (streptavidin and
b-galactosidase) deposited at random on mica supports. To
minimize damage to the protein molecules by the AFM tip,
we used tapping-mode rather than contact-mode scanning
(Hansma et al., 1993). We found no evidence that the
proteins imaged had been adversely a€ected either in the
course of sample preparation or during scanning. For
instance, the spreads of protein molecules were remarkably
homogenous, and the standard errors on the calculated
molecular volumes were small (510% of the mean). In
addition, the streptavidin molecules were decorated by DNA-
biotin with the expected geometry.
We used molecular volume calculations for two purposes ±
to con®rm that the structures in the images had been correctly
identi®ed (e.g. as streptavidin tetramers, or b-galactosidase
tetramers or monomers), and as a means of detecting changes
in the structure of streptavidin upon biotinylation. It is widely
acknowledged that the determination of molecular volume
using AFM is potentially problematic, particularly because of
factors introduced by the geometry of the scanning tip (Larmer
et al., 1997). Speci®cally, when the diameter of the particle to be
imaged is of the same order as the diameter of the tip (i.e. in the
nanometre range), the particle diameter can be signi®cantly
over-estimated. The use of half-height diameter measurement
This study establishes DNA-biotin/streptavidin as an
e€ective protein tagging reagent. Furthermore, the clear
visibility of the DNA rods in AFM images should permit
the determination of the architecture of pharmacologically-
signi®cant multi-subunit proteins such as the nicotinic
British Journal of Pharmacology vol 135 (8)

1950
C.S. Neish et al
Ligand-protein interaction visualized by AFM
acetylcholine receptor, the GABAA receptor and the 5-HT3
receptor. These ligand-gated ion channels all consist of ®ve
homologous subunits arranged around a central pore. The
subunits present within all three of these receptors have been
well characterized (Karlin, 1987; Unwin, 1998; Farrar et al.,
1999; Davies et al., 1999), but the arrangement of the
subunits to form the functioning receptor is still unclear. By
biotinylating speci®c subunits, for example via introduced
cysteine residues (Noji et al., 1997), and then using
streptavidin tagged with single DNA-biotin molecules to
decorate the receptor, it should be possible to deduce the
receptor structure using AFM imaging. For example, if two
tagged subunits are adjacent, then the angle between the
DNA rods will be 728, whereas if they are separated by an
untagged subunit, the angle will be 1448. Our current aim is
to apply this analysis to the GABAA receptor. Finally, it may
be possible to couple receptor ligands directly to DNA, which
might permit the use of AFM to study the relative positions
of the binding sites for di€erent ligands on the same receptor
complex, through the use of di€erent-length DNA tags. The
GABAA receptor might also provide a suitable model system
for this approach, since it is the target for multiple ligands,
such as GABA itself, the benzodiazepines and barbiturates
(Sieghart, 1995).
This work was funded by Grant B12816 from the Biotechnology
and Biological Sciences Research Council (to R.M. Henderson and
J.M. Edwardson)
References
BERGE, T., ELLIS, D.J., DRYDEN, D.T.F., EDWARDSON, J.M.
&
HENDRICKSON, W.A., PAHLER, A., SMITH, J.L., SATOW, Y.,
MERRITT, E.A. & PHIZACKERLEY, R.P. (1989). Crystal structure
of core streptavidin determined from multiwavelength anom-
alous di€raction of synchotron radiation. Proc. Natl. Acad. Sci.
U.S.A., 86, 2190 ± 2194.
KANESHIRO, C., ENNS, C., HAHN, M., PETERSON, J. & REITHEL, F.
(1975). Evidence for an active dimer of Escherichia coli
b-galactosidase. Biochem. J., 151, 433 ± 434.
HENDERSON, R.M. (2000). Translocation-independent dimeriza-
tion of the EcoKI endonuclease visualized by atomic force
microscopy. Biophys. J., 79, 479 ± 484.
COHEN, R. & MIRE, M. (1971). Analytical-band centrifugation of an
active enzyme-substrate complex. Determination of active
subunits of various enzymes. Eur. J. Biochem., 23, 276 ± 281.
DAVIES, P.A., PISTIS, M., HANNA, M.C., PETERS, J.A., LAMBERT,
J.J., HALES, T.G. & KIRKNESS, E.F. (1999). The 5-HT3B subunit is
a major determinant of serotonin-receptor function. Nature, 397,
359 ± 363.
EDSTROM, R.D., MEINKE, M.H., YANG, R., YANG, X., ELINGS, V. &
EVANS, D.F. (1990). Direct visualization of phosphorylase-
phosphorylase kinase complexes by scanning tunneling and
atomic force microscopy. Biophys. J., 58, 1437 ± 1448.
ELLIS, D.J., BERGE, T., EDWARDSON, J.M. & HENDERSON, R.M.
(1999a). Investigation of protein partnerships using atomic force
microscopy. Microsc. Res. Tech., 44, 368 ± 377.
KARLIN, A. (1987). Going round in receptor circles. Nature, 329,
286 ± 287.
KATZ, B.A. (1997). Binding of biotin to streptavidin stabilizes
intersubunit salt bridges between Asp61 and His87 at low pH.
J. Mol. Biol., 274, 776 ± 800.
KLUMB, L.A., CHU, V. & STAYTON, P.S. (1998). Energetic roles of
hydrogen bonds at the ureido oxygen binding pocket in the
streptavidin-biotin complex. Biochemistry, 37, 7657 ± 7663.
LARMER, J., SCHNEIDER, S.W., DANKER, T., SCHWAB, A.
&
OBERLEITHNER, H. (1997). Imaging excised apical plasma
membrane patches of MDCK cells in physiological conditions
ELLIS, D.J., DRYDEN, D.T.F., BERGE, T., EDWARDSON, J.M.
&
HENDERSON, R.M. (1999b). Direct observation of DNA
translocation and cleavage by the EcoKI endonuclease using
atomic force microscopy. Nature Struct. Biol., 6, 15 ± 17.
FARRAR, S.J., WHITING, P.J., BONNERT, T.P. & MCKERNAN, R.M.
(1999). Stoichiometry of a ligand-gated ion channel determined
by ¯uorescence resonance energy transfer. J. Biol. Chem., 274,
10100 ± 10104.
with atomic force microscopy. P¯ugers Arch., 434, 254 ± 260.
È
NOJI, H., YASUDA, R., YOSHIDA, M. & KINOSHITA, JR K. (1997).
Direct observation of the rotation of F1-ATPase. Nature, 386,
299 ± 302.
SAMBROOK, J., FRITSCH, E.F. & MANIATIS, T. (1989). Molecular
Cloning: A Laboratory Manual. 2nd edn. New York: Cold Spring
Harbor Laboratory Press.
FREITAG, S., LE TRONG, I., CHILKOTI, A., KLUMB, L.A., STAYTON,
P.S. & STENKAMP, R.E. (1998). Structural studies of binding site
tryptophan mutants in the high-anity streptavidin-biotin
complex. J. Mol. Biol., 279, 211 ± 221.
FREITAG, S., LE TRONG, I., KLUMB, L.A., CHU, V., CHILKOTI, A.,
STAYTON & STENKAMP, R.E. (1999). X-ray crystallographic
studies of streptavidin mutants binding to biotin. Biomol. Eng.,
16, 13 ± 19.
SANO, T., & CANTOR, C.R. (1995). Intersubunit contacts made by
tryptophan 120 with biotin are essential for both strong biotin
binding and biotin-induced tighter subunit association of
streptavidin. Proc. Natl. Acad. Sci. U.S.A., 92, 3180 ± 3184.
SCHNEIDER, S.W., LARMER, J., HENDERSON, R.M. & OBERLEITH-
NER, H. (1998). Molecular weights of individual proteins
correlate with molecular volumes measured by atomic force
microscopy. P¯ugers Arch., 435, 362 ± 367.
È
GREEN, N.M. (1990). Avidin and streptavidin. Methods Enzymol.,
184, 51 ± 67.
SIEGHART, W. (1995). Structure and pharmacology of GABA(A)
receptor subtypes. Pharmacol. Rev., 47, 181 ± 234.
GONZALEZ, M., ARGARANA, C.E.
&
FIDELIO, G.D. (1999).
STAYTON, P.S., FREITAG, S., KLUMB, L.A., CHILKOTI, A., CHU, V.,
PENZOTTI, J.E., TO, R., HYRE, D., LE TRONG, I., LYBRAND, T.P.
Extremely high thermal stability of streptavidin and avidin upon
biotin binding. Biomol. Eng., 16, 67 ± 72.
GONZALEZ, M., BAGATOLLI, L.A., ECHABE, I., ARRONDO, J.L.R.,
& STENKAMP, R.E. (1999). Streptavidin-biotin binding ener-
getics. Biomol. Eng., 16, 39 ± 44.
ARGARANA, C.E., CANTOR, C.R.
&
FIDELIO, G.D. (1997).
UNWIN, N. (1998). The nicotinic acetylcholine receptor of the
Torpedo electric ray. J. Struct. Biol., 121, 181 ± 190.
WEBER, P.C., OHLENDORF, D.H., WENDOLOSKI, J.J. & SALEMME,
F.R. (1989). Structural origins of high-anity biotin binding to
streptavidin. Science, 243, 85 ± 88.
Interaction of biotin with streptavidin. Thermostability and
conformational changes upon binding. J. Biol. Chem., 272,
11288 ± 11294.
HANSMA, H.G., SINSHEIMER, R.L., GROPPE, J., BRUICE, T.C.,
ELINGS, V., GURLEY, G., BEZANILLA, M., MASTRANGELO,
I.A., HOUGH, P.V.C. & HANSMA, P.K. (1993). Recent advances in
atomic force microscopy of DNA. Scanning, 15, 296 ± 299.
(Received October 16, 2001
Revised January 1, 2002
Accepted February 11, 2002)
British Journal of Pharmacology vol 135 (8)