The solution structure and activation of visual arrestin studied
by small-angle X-ray scattering
Brian H. Shilton
1
, J. Hugh McDowell
2
, W. Clay Smith
2
and Paul A. Hargrave
2,3
1
Department of Biochemistry, University of Western Ontario, London, Ontario, Canada;
2
Departments of Ophthalmology and
3
Biochemistry and Molecular Biology University of Florida, Gainesville, Florida, USA
Visual arrestin is converted from a ÔbasalÕ state to an
ÔactivatedÕ state by interaction with the phosphorylated
C-terminus of photoactivated rhodopsin (R*), but the
conformational changes in arrestin that lead to activation
are unknown. Small-angle X-ray scattering (SAXS) was
used to investigate the solution structure of arrestin and
characterize changes attendant upon activation. Wild-type
arrestin forms dimers with a dissociation constant of
60 l
M
. Small conformational changes, consistent with
local movements of loops or the mobile N- or C-termini
of arrestin, were observed in the presence of a phospho-
peptide corresponding to the C-terminus of rhodopsin,
and with an R175Q mutant. Because both the phospho-

prevents arrestin from inappropriately associating with R*.
In fact, arrestin shows very little propensity to bind to R*
until the C-terminal region of R* becomes phosphorylated.
Thus, the C-terminal peptide appears to act as a ÔswitchÕ
that, once phosphorylated, converts arrestin into a state that
is able to bind to R*. The effects of rhodopsin’s phospho-
rylated C-terminal peptide can be mimicked by a synthetic
phosphopeptide or even certain point mutations: both
wild-type arrestin in the presence of the synthetic phospho-
peptide [7], and arrestin-R175Q on its own [8] are able to
bind to unphosphorylated R* and abrogate signalling to
transducin.
The crystal structure of arrestin is known [9,10], but it
is not clear how binding of the phosphorylated C-terminal
peptide of rhodopsin promotes tight complex formation
between arrestin and R*. One possibility is that binding of
phosphopeptide leads to a conformational change in
arrestin that increases its affinity for R*. Conformational
changes in arrestin can take place in solution, as
demonstrated by changes in the proteolytic digestion
pattern that result from phosphopeptide binding [7] or by
heparin binding [11], and changes in cysteine reactivity
due to phosphopeptide binding or the activating R175Q
mutation [12]. The nature and extent of the conforma-
tional change that leads to activation of arrestin is not
known. The situation is complicated by the fact that
visual arrestin participates in a monomer–dimer equilib-
rium [13,14]. It has been suggested that the arrestin dimer
may function as an inert storage form of the protein,
which can be recruited by dissociation to terminate the

interact with R*. This dimer could play an active role in
attenuation of R* signalling.
EXPERIMENTAL PROCEDURES
Protein expression and purification
Wild-type arrestin was prepared from bovine retina [7],
while arrestin R175Q was expressed from yeast cells and
purified as previously described [12]. In both cases, the
arrestin yielded a single band when analysed by SDS/
PAGE. Protein preparations were dialysed against 10 m
M
Hepes, 400 m
M
NaCl, pH 7.5, and concentrated to
approximately 0.13 m
M
by ultrafiltration; the ultrafiltrate
was retained and used for buffer subtraction during the
SAXS experiments. Following concentration, the protein
was flash frozen and maintained at )80 °C. When required,
additional concentration was carried out just prior to SAXS
measurements using 0.5 mL centrifugal ultrafilters (Milli-
pore Corp., Bedford, MA, USA).
Protein concentration
The concentration of BSA was measured using
A
280
(1%) ¼ 6.14, while the concentration of arrestin was
measured using A
278
(1%) ¼ 6.38 [15].

All measurements were made at the European Molecular
Biology Laboratory Outstation at the Deutsches Elektro-
nen-Synchrotron (Hamburg, Germany), beamline X33
[17], at 15 °C using radiation with a wavelength of
0.15 nm. Measurements were made with either a position-
sensitive linear detector or a Quadrant segment-shaped
multiwire detector [18,19]. Sample–detector distances of
1.2 m (high angle) and 3 m (low angle) were used to cover
therangeofmomentumtransfer(S ¼ 2sinh/k,where2h
is the scattering angle) from 0.02 to 0.8 nm
)1
. Fifteen
successive 1-min exposures were recorded for each sample;
there was no evidence of protein degradation over this
time interval. Recording of each protein sample was
preceded and followed by recording from the buffer alone;
these buffer measurements were compared and provided a
check on beam properties and the cleanliness of the cell
between readings of protein solutions. Averaging of
frames, corrections for detector response and beam
intensity, and buffer subtraction, were performed using
the programs
SAPOKO
(Svergun, D.I. & Koch, M.H.J.,
unpublished material) and
OTOKO
[20]. Phosphopeptide
was added to the protein samples and matching buffer
just prior to measurement.
Determination of binding constants from forward

Ið0Þ
M
ð1Þ
where f
M
and f
D
are the mass fractions of monomer and
dimer, respectively. Assuming that the monomer and dimer
are the only species present, the expression for total forward
scattering can be simplified as follows.
f
D
¼ 1 À f
M
Ið0Þ
Total
¼ð2 À f
M
ÞIð0Þ
M
ð2Þ
In the experiments, the total amount of protein is varied and
the forward scattering is measured. The dissociation con-
stant for dimerization, K
d
, is defined as:
K
d
¼

ð1 À f
M
Þ
ð4Þ
This expression for K
d
canbesolvedforthemassfractionof
monomer, f
M
, which can then be used in the expression for
I(0)
total
to yield the following equation:
3802 B. H. Shilton et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Forward scattering, I(0)
total
, was plotted as a function of the
mass concentration of arrestin, [arrestin]
total
, and the curve
was fit to Eqn (5) using nonlinear regression as implemented
in the program
KALEIDAGRAPH
(v3.08, Synergy Software),
setting the molecular mass of the monomer equal to
45.3 kDa and the forward scattering of the monomer,
I(0)
M
to the appropriate value, as determined using a BSA
standard. The only variable fitted was the value for K

dimer, respectively, and R
gM
and R
gD
are the radii of
gyration for the monomer and dimer, respectively.
Missing pieces of crystallographic models were built as
extended polypeptide using the program
O
[24]. Crystallo-
graphic models, which did not include crystallographic
water molecules, were fit to experimental SAXS data using
the program
CRYSOL
[25], with a solvent density of 0.36
electronsÆA
˚
)3
, using default limits for the contrast of the
hydration shell and average displaced solvent volume per
atomic group. The target function for
CRYSOL
is the v value,
which is a measure of the agreement between the theoretical
scattering from a model and the experimental data:
v
2
¼
1
N À 1

(S
j
) is the experimentally observed
scattering, and r(S
j
)istheerrorfortheexperimental
measurement.
RESULTS AND DISCUSSION
Arrestin participates in a monomer–dimer equilibrium
in solution
Dimers of arrestin have been detected by sedimentation
velocity [13] and sedimentation equilibrium [14]; in addition
to the dimer, an arrestin tetramer was taken to be the
predominant species when the protein was at high concen-
tration (220 l
M
or 10 mgÆmL
)1
), even though the tetramer
constituted only a minor component at 62 l
M
[14]. Our first
task in understanding the solution structure and activation
of visual arrestin was to ascertain its oligomeric state in our
preparations.
The X-ray solution scattering from particles at an angle
of 0° with respect to the direct beam, the Ôforward
scatteringÕ, is directly proportional to the molecular mass
of the scattering species. The forward scattering values for
various concentrations of wild-type arrestin were measured

D
W
M
q
4½arrestin
Total
0
@
1
A
ð5Þ
Ið0Þ
Total
¼ Ið0Þ
M
2 þ
K
D
W
M
À
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
K
2
D
W
2
M
þ 8½arrestin
Total

[20].(B)Identicalto(A),except
expandedintheregion0<S <0.09nm
)1
.
Ó FEBS 2002 Activation of visual arrestin (Eur. J. Biochem. 269) 3803
One preparation of wild-type arrestin, with a concentration
of 140 l
M
, was diluted with buffer to yield concentrations of
110, 86, 57 and 29 l
M
. These five points were plotted
(Fig. 2A, circles) and it was clear that there was a
concentration dependence for the I(0) values. The apparent
molecular masses at the different protein concentra-
tions were calculated based on a BSA standard, and ranged
from 50 to 80 kDa, consistent with a monomer–dimer
equilibrium.
We sought to characterize further the arrestin self-
association, and to this end measured SAXS from solutions
with concentrations of arrestin from 180 to 1300 l
M
(Fig. 2A, squares). Some of these data were recorded using
a 1.2 m sample–detector distance, and only the outer part of
the Guinier region (S values from 0.06 to 0.08 nm
)1
)was
available for the analysis. The increase of molecular mass in
response to increased protein concentration was much less
dramatic through this concentration range, progressing

experimental scattering (Fig. 2B), where it is quite clear that
the overall shape of the tetramer does not match what we
have observed in solution.
These data were fit to a simple model that incorporates a
monomer–dimer equilibrium, and relates the protein con-
centration to the forward scattering using the K
d
as the
single variable. Using this equation, it was found that the K
d
was 40 ± 20 l
M
. Other fits to the data were also tested. A
monomer–tetramer equilibrium was found to be incompat-
ible with the data because it requires a monomer molecular
mass of 30 kDa. Incorporation of a simple linear term in the
monomer–dimer model, with the slope of the line as a
second variable, results in a better overall fit to the data. In
this case, the K
d
of the monomer–dimer interaction
increases to 60 ± 25 l
M
. This linear term represents the
formation of high molecular mass, irreversible arrestin
aggregates, which have been observed in other studies
[14,26,27]. Because the forward scattering is directly pro-
portional to molecular mass, these high molecular mass
species comprise only a minor component of the mixture
and their effect is limited to the low angle Guinier region.

two protein concentrations, 140 and 1300 l
M
, were merged to provide
a representative scattering curve (dots) that covers a broad range of
momentum transfer values. The structure of the crystallographic
tetramer [9,10] was used to calculate the theoretical solution scattering
for this particle (solid curve), which was fit to the experimental data
using the program
CRYSOL
[25].
3804 B. H. Shilton et al. (Eur. J. Biochem. 269) Ó FEBS 2002
Effect of rhodopsin C-terminal phosphopeptide
on arrestin conformation
Interaction between arrestin and the C-terminal phospho-
rylated peptide of rhodopsin leads to tight binding of
arrestin so that it is able to stop R* signalling [28]. It has
been shown that the phosphopeptide does not have to be
covalently bound to rhodopsin: a short, soluble phospho-
peptide corresponding to the C-terminus of rhodopsin is
sufficient to ÔactivateÕ arrestin and allow it to bind to
photoactivated but not phosphorylated rhodopsin [7].
Given the structure of arrestin, it is conceivable that binding
of the phosphopeptide could drive large conformational
changes, such as a reorientation of the N- and C-terminal
domains, that produce the dramatically increased affinity
for R*; on the other hand, binding of the phosphopeptide
may simply alter the surface features and cause relatively
minor conformational changes. To distinguish between
these two possibilities, we investigated the changes in
arrestin structure and conformation produced by rhodop-

scattering curves for arrestin (110 l
M
) in the presence and
absence of phosphopeptide were compared (Fig. 3B), and it
can be seen that the scattering curves are identical out to an
S value of 0.2 nm
)1
. If the phosphopeptide caused a
reorientation of the N- and C-terminal domains, one would
expect to see a change in this scattering curve: the absence of
any such change indicates that the phosphopeptide has little
effect on the gross conformation of arrestin.
Because the effect of the phosphopeptide was not
detectable at low angles, additional measurements were
carried out at arrestin concentrations of 260 and 500 l
M
using a 1.2-m sample to detector distance to measure higher
angle scattering, which is sensitive to more subtle changes in
structure. The scattering in the presence and absence of
approximately 13 m
M
(33 mgÆmL
)1
) phosphopeptide was
Fig. 3. Effect of phosphopeptide on arrestin. (A) Dependence of I(0) on
protein concentration for wild type arrestin in the presence (squares) or
absence (circles) of phosphopeptide (1 m
M
). (B) Comparison of low-
angle scattering from arrestin (110 l

different shapes. At S values above 0.4 nm
)1
, the scattering
signal is not sufficiently strong to determine whether the
difference between the two curves is significant. The changes
in scattering at S values greater than 0.2 nm
)1
could be
produced by the presence of the phosphopeptide on the
surface of arrestin and/or by changes in ÔlocalÕ arrestin
structure. These results are consistent with a model where
binding of phosphopeptide causes a displacement of arres-
tin’s C-terminus [29] and/or changes in the conformation of
certain loops that facilitate R* binding [9].
The structure of arrestin R175Q resembles
that of wild-type arrestin
Replacement of arginine 175 with either glutamine or
glutamic acid produces a constitutively activated arrestin
molecule that binds photoactivated but unphosphorylated
rhodopsin [8,30]. We used SAXS to elucidate the structural
changes leading to activation of the R175Q mutant.
Increases in the concentration of arrestin R175Q produce
increases in forward scattering that are virtually identical to
those observed for wild-type arrestin (data not shown),
indicating that the R175Q mutation does not influence the
monomer–dimer equilibrium. Comparison of arrestin-
R175Q (110 l
M
) with the wild-type protein (130 l
M

common centre of mass. The R
g
canbeusedtoevaluatethe
overall shape of a scattering particle. The R
g
for a mixture of
scattering species depends on the R
g
for each species and
their mass fraction according to Eqn (7). The radius of
gyration of the monomer is 2.6 nm, determined using the
crystallographic coordinates [10]. The measured R
g
for wild-
type arrestin at 110 l
M
is 3.6 nm: at 110 l
M
the mass
fraction of monomer is 0.35, and therefore the R
g
for the
dimer is approximately 4.0 nm. Measurements at other
protein concentrations also indicated an R
g
for the dimer of
between 4.0 and 4.1. The dimer is therefore a highly
elongated molecule. The only way to make such an
elongated molecule from the monomeric species is to put
two monomers together such that their long axes are

of dimers. The tetramer may be composed of either AB and
CD dimers, or AD and CB dimers. The AB dimer has a
radius of gyration of 4.0 nm, but has an extremely limited
contact area (less than 1000 A
˚
2
of buried surface), making it
an unlikely candidate for the solution dimer, as noted
previously [14]. The AD dimer is the most compact in the
crystal, with an R
g
of 3.4 nm, yielding a relatively poor fit to
the scattering data (not shown). On purely theoretical
grounds, it is unlikely that either the AB or AD dimers exist
to a significant degree in solution for the following reason:
these dimers participate in Ôheterologous associationÕ [31].
That is, they do not have inherent two-fold symmetry, the
surface used for the dimer interface is different for each
protomer. Such an arrangement is unlikely to result in the
formation of stable dimers as observed in solution; rather,
the formation of such dimers in solution would be expected
to progress to larger and larger polymers.
In the C222
1
structure, there are two other dimers formed
by crystallographic symmetry operations: one of these
consists of two A chains interacting through their C-termi-
nal domains (the ÔAAÕ dimer), while the other consists of a B
chain and D chain interacting through their N-terminal
domains (the ÔBDÕ dimer). These dimers possess twofold

(Fig. 5A; v ¼ 3.2) but
only a slight improvement in the fit of the BD dimer model
(Fig. 5B; v ¼ 4.4). The agreement between the theoretical
scattering from the models and the experimental scattering
Fig. 5. Models of the arrestin dimer in solution. Models of the arrestin
dimer in solution. Structural models were constructed from dimers
present in the crystal structure of arrestin [10] using the macromolecular
modelling program
O
[24]. Approximately 3 kDa of polypeptide was
missing from the N-terminus, C-terminus and two internal loops of the
crystal structures; these missing pieces were modelled as extended
polypeptide to improve the agreement with experimental SAXS data
(dashed curves in both graphs). (A) Structure (ribbon diagram) and
predicted scattering (solid curve) of the ÔAAÕ dimer formed by interac-
tion between the C-terminal domains of arrestin monomers. (B)
Structure (ribbon diagram) and predicted scattering (solid curve) of the
ÔBDÕ dimer formed by interaction of N-terminal domains of arrestin
monomers. Ribbon diagrams were drawn using the Swiss PDB Viewer
[32]. (C) The agreement between theoretical and experimental scattering
curves is indicated by the square of the weighted residual (summed term
in Eqn 8) for each momentum transfer value; the solid line is for the AA
dimer, and the dashed line is for the BD dimer.
Ó FEBS 2002 Activation of visual arrestin (Eur. J. Biochem. 269) 3807
is indicated in Fig. 5(C), where the squared residuals are
plotted as a function of the momentum transfer: the AA
dimer has the best agreement in the lower angle region, up
to S ¼ 0.3 nm
)1
. Within the crystal structures of arrestin,

is likely the major species in vivo. It has been proposed that
the biological function of the arrestin dimer is to provide an
inert ÔstorageÕ form of the protein [14]; the implication is that
the dimer is not capable of binding to R*. However, the
ÔAAÕ dimer identified in this study has the intriguing
characteristic that the N-terminal domains are left open
and available for interaction with rhodopsin. In fact, the
structure of this dimer is such that both N-terminal domains
could conceivably interact with two rhodopsin molecules
simultaneously.
It may be that only monomeric arrestin is able to
effectively interact with R*P, and that the arrestin dimer is
a storage form of the protein [14]. An alternative possibility
raised by the present study is that dimeric arrestin has a
more active role in attenuation of rhodopsin signalling or
Fig. 6. Detection of structural changes in arrestin. To demonstrate
how changes in oligomeric structure and/or conformational would
affect scattering patterns, theoretical SAXS curves were calculated
from model structures derived from the high resolution crystal
structure of arrestin [10]. (A) The theoretical scattering from the
putative arrestin dimer (dotted curve; see Fig. 5A) is compared to that
of monomeric arrestin (solid curve). (B) The arrestin dimer (grey
backbone) is superimposed over a possible alternative conformation
(black backbone). In the alternative conformation, the N-terminal
domains of each monomer are rotated approximately 15° with respect
to the C-terminal domains; the overall effect is a slight ÔclosureÕ of the
arrestin dimer. (C) The theoretical scattering curves for the two
structures in (B) are compared: the dashed curve represents scattering
from the grey structure, while the solid curve represents scattering
from the black, structure. The difference between scattering from the

ism of quenching of phototransduction. Binding competition
between arrestin and transducin for phosphorhodopsin. J. Biol.
Chem. 272, 18125–18131.
5. Gibson, S.K., Parkes, J.H. & Liebman, P.A. (2000) Phosphor-
ylation modulates the affinity of light-activated rhodopsin for G
protein and arrestin. Biochemistry 39, 5738–5749.
6. Hamm,H.E.&Bownds,M.D.(1986)Proteincomplementofrod
outer segments of frog retina. Biochemistry 25, 4512–4523.
7. Puig, J., Arendt, A., Tomson, F.L., Abdulaeva, G., Miller, R.,
Hargrave, P.A. & McDowell, J.H. (1995) Synthetic phosphopep-
tide from rhodopsin sequence induces retinal arrestin binding to
photoactivated unphosphorylated rhodopsin. FEBS Lett. 362,
185–188.
8. Gray-Keller, M.P., Detwiler, P.B., Benovic, J.L. & Gurevich, V.V.
(1997) Arrestin with a single amino acid substitution quenches
light-activated rhodopsin in a phosphorylation-independent
fashion. Biochemistry 36, 7058–7063.
9. Granzin,J.,Wilden,U.,Choe,H.W.,Labahn,J.,Krafft,B.&
Bu
¨
ldt, G. (1998) X-ray crystal structure of arrestin from bovine
rod outer segments. Nature 391, 918–921.
10. Hirsch, J.A., Schubert, C., Gurevich, V.V. & Sigler, P.B. (1999)
The 2.8 A
˚
crystal structure of visual arrestin: a model for arrestin’s
regulation. Cell 97, 257–269.
11. Palczewski, K., Pulvermuller, A., Buczylko, J. & Hofmann, K.P.
(1991) Phosphorylated rhodopsin and heparin induce similar
conformational changes in arrestin. J. Biol. Chem. 266, 18649–

experiments: hardware and software. Nucl. Instrum. Methods
A249, 399–407.
21. Guinier, A. & Fournet, G. (1955) Small Angle Scattering of
X-rays. Wiley, New York.
22. Svergun, D.I., Semenyuk, A.V. & Feigin, L.A. (1988) Small-angle
scattering data treatment by the regularization method. Acta
Cryst. A44, 244–250.
23. Svergun, D.I. (1992) Determination of the regularization param-
eter in indirect-transform methods using perceptual criteria.
J. Appl. Cryst. 25, 495–503.
24. Jones, T.A., Bergdoll, M. & Kjeldgaard, M. (1990) O: a macro-
molecular modeling environment. In Crystallographic and Mod-
eling Methods in Molecular Design (Bugg,C.&Ealick,S.E.,eds),
pp. 189–195. Springer-Verlag, New York.
25. Svergun, D.I., Barberato, C. & Koch, M.H.J. (1995) CRYSOL – a
program to evaluate X-ray solution scattering of biological
macromolecules from atomic coordinates. J. App. Cryst. 28,
768–773.
26. Palczewski, K., Riazance-Lawrence, J.H. & Johnson, W.C. Jr
(1992) Structural properties of arrestin studied by chemical mod-
ification and circular dichroism. Biochemistry 31, 3902–3906.
27. Kotake, S., Hey, P., Mirmira, R.G. & Copeland, R.A. (1991)
Physicochemical characterization of bovine retinal arrestin. Arch.
Biochem. Biophys. 285, 126–133.
28. Wilden, U., Hall, S.W. & Ku
¨
hn, H. (1986) Phosphodiesterase
activation by photoexcited rhodopsin is quenched when rhodopsin
is phosphorylated and binds the intrinsic 48-kDa protein of rod
outer segments. Proc. Natl Acad. Sci. USA 83, 1174–1178.