Tài liệu Báo cáo khoa học: Solution properties of full-length integrin aIIbb3 refined models suggest environment-dependent induction of alternative bent ⁄extended resting states doc

Solution properties of full-length integrin
a
IIb
b
3
refined
models suggest environment-dependent induction of
alternative bent
⁄
extended resting states
Camillo Rosano
1
and Mattia Rocco
2
1 Nanobiotecnologie, Istituto Nazionale per la Ricerca sul Cancro (IST), Genova, Italy
2 Biopolimeri e Proteomica, IST, Genova, Italy
Introduction
Integrins are heterodimeric transmembrane (TM) cellu-
lar receptors involved in mechanical anchoring and
two-way signaling [1]. Each a and b subunit has a
modular structure with a large extracellular portion,
a single TM region and a cytoplasmic domain [1–3].
The integrin activation mechanism is regulated by
conformational changes, the details of which have not
yet been fully elucidated [2,3]. X-ray crystallography
Keywords
blood coagulation; hydrodynamics;
modeling; modular proteins; protein
structure
Correspondence
M. Rocco, Biopolimeri e Proteomica, IST

c ⁄ o CBA, Largo R. Benzi 10, I-16132
Genova, Italy
Fax: +39-0105737-325
Tel: +39-0105737-310
E-mail:
(Received 12 November 2009, revised 1
May 2010, accepted 29 May 2010)
doi:10.1111/j.1742-4658.2010.07724.x
The recently published novel integrin a
IIb
b
3
ectodomain crystallographic
structure and NMR structures of its transmembrane ⁄ cytoplasmic segments
were employed to refine previously developed molecular models. Alterna-
tive complete a
IIb
b
3
models were built and evaluated, and their shape was
compared with EM maps and their computed hydrodynamic ⁄ conforma-
tional properties were compared with the available experimental data. A
partially extended ⁄ closed model, or a mixture of bent ⁄ closed and exten-
ded ⁄ closed conformations, are both compatible with the results of a recent
small-angle neutron scattering study of Triton X-100-solubilized resting
a
IIb
b
3
, while new electron microscopy evidence of nanodiscs-embedded

a
IIb
b
3
supports the bent ⁄ closed resting form. However, only an extended ⁄ -
closed model matches well the hydrodynamics of either octyl-glucoside-sol-
ubilized or nanodiscs-embedded resting a
IIb
b
3
, suggesting that different
solubilization strategies and substrate interactions might operate a confor-
mational selection between alternative, stable states. Furthermore, extende-
d ⁄ open models are required to match the electron tomography map and
the hydrodynamics following the priming-induced b
3
hybrid domain swing-
out, but without immediate full tail separation. Importantly, both extension
and opening transitions can occur by pivoting at the recently identified b
3
hinge point, which does not appear to be freely flexible. The structure and
mechanism of action of integrins thus seem to depend on discrete transi-
tions and to be more tightly coupled to the local environment than previ-
ously thought.
Abbreviations
bc, bent ⁄ closed; DMPC, 1,2-dimyristoyl-sn-glycero-3-phosphocholine; DMPG, 1,2-dimyristoyl-sn-glycero-3-phospho-(1¢-rac-glycerol); DPPC,
1,2-dipalmitoyl-sn-glycero-3-phosphocholine; ec, extended ⁄ closed; EGF, epidermal growth factor; EM, electron microscopy; eotc,
extended ⁄ open ⁄ tails crossed; eots, extended ⁄ open ⁄ tail separated; ET, electron tomography; NMA, Normal Modes Analysis; OG, octyl-
glucoside; pec, partially extended ⁄ closed; PSI, plexin ⁄ semaphorin ⁄ integrin; SANS, small-angle neutron scattering; TEM, transmission
electron microscopy; TM, transmembrane.

3190 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
has revealed similar bent shapes for resting and
primed extracellular region constructs (‘ectodomains’)
[4,5], while ligand binding-induced large structural
re-arrangements in smaller constructs suggested exten-
sion, ‘opening’ and tails separation [3,6]. Lower-resolu-
tion structural data, such as transmission electron
microscopy (TEM) [7], cryo-electron microscopy
(cryo-EM) [8] and electron tomography (ET) [9] have
provided maps ranging from supporting the crystallo-
graphic models of the ectodomains, to being in accor-
dance with the extended conformation after priming of
full-length, detergent-solubilized samples.
Previously, we presented a multiresolution study of
integrins a
v
b
3
(ectodomain) and a
IIb
b
3
(full-length), in
which crystallography ⁄ NMR-based models were fit in
the EM ⁄ ET maps and their hydrodynamic parameters
were then computed and compared with the experi-
mental data [10]. Our modeling work [10] suggested
that full-length integrins might be already extended,
with the transition to the open form (involving the
swing-out of the b

3
hybrid domain [6]) taking place
without the requirement of full separation of the TM
helices. Since then, crystallographic structures of the
a
IIb
b
3
, a
v
b
3
and a
x
b
2
ectodomains [11–13] and two
NMR-based structures of the a
IIb
b
3
TM helices – one
embedded in a small bicelle [14] and the other solubi-
lized in a mixed solvent and including the cytoplasmic
domains [15] – have been published. Furthermore, new
low-resolution data of full-length a
IIb
b
3
have very

recently appeared: a small-angle neutron scattering
(SANS) study after solubilization in Triton X-100 [16];
and an EM ⁄ analytical ultracentrifugation study after
reconstitution in phospholipid nanodiscs [17]. Impor-
tantly, the new a
IIb
b
3
and a
v
b
3
ectodomain crystallo-
graphic structures reported the previously unresolved
structure of the b
3
I-epidermal growth factor (EGF)1-2
modules, revealing the potential hinge points for sub-
unit extension [11,12]. Because our previous a
IIb
b
3
models [10] were developed by homology modeling of
the a
IIb
subunit on the a
v
template [4,5], we felt that
the new data warranted a significant revision of the
models. This included inserting the recently published

NMR-based structures of the TM helices ⁄ cytoplasmic
domains for the resting integrin state, while for the
primed state we employed the computer models [18,19]
utilized in our previous work [10]. Particular care was
exerted when modeling the major, still-unresolved,
loop in the calf-2 module of a
IIb
, and we resorted to
using ab initio modeling procedures. Extended ⁄ closed
and extended ⁄ open models were derived from the fully
bent ⁄ closed crystallographic model by fitting to the ET
map, and two series of intermediate models were
obtained by morphing between these initial and final
conformations. The models were then assessed by com-
paring their hydrodynamic and conformational param-
eters (computed using the new UltraScan SOlution
MOdeler (US-SOMO) bead-modeling implementation
[20–22]) with experimental data [10,16,17]. While the
fully bent ⁄ closed crystallographic model was incompat-
ible with all the available solution data for resting
a
IIb
b
3
, differences remained between a partially exten-
ded ⁄ closed SANS-complying model and the extended ⁄ -
closed ET ⁄ hydrodynamics-based model. However, the
SANS data could also be interpreted as deriving from
a mixture of bent and extended conformations. More-
over, only an extended ⁄ open model without full tail

separation was in accordance with both the ET and
hydrodynamic data of primed a
IIb
b
3
. Our revised mod-
els further support an alternative view of the confor-
mational states and mechanism of action of integrins,
and suggest a more tightly coupled synergy than previ-
ously thought with the extracellular, TM and cytoplas-
mic environments.
Results
Model building
The first refined model, A2bB3-bent ⁄ closed (bc)), con-
sisted of the new a
IIb
b
3
ectodomain structure, 3FCS
[11], to which the new TM ⁄ cytoplasmic domain NMR
model, taken from the 2KNC structure [15], was
attached. The TM helices, which were nicely superim-
posable with those of the recent 2K9J NMR structure
[14], were embedded in 50 octyl-glucoside (OG) mole-
cules to mimic a solubilizing micelle [10]. When recon-
necting their N-terminal part to the C-terminal ends of
the ectodomain, particular care was exerted to allow
for reasonable mechanical coupling. The intermediate-
length loop in a
IIb

(E764-D774) was modeled using
ModLoop [23], whereas the G75-S78 and D477-Q482
loops in b
3
were taken from the new a
v
b
3
ectodomain
structure, 3IJE [12]. Given their probable flexible nat-
ure, no attempt was made to fully optimize the confor-
mation of these segments. As for the still-unresolved
long loop at the end of the a
IIb
calf-2 module (G840-
Q873), alignment of all human integrin a subunits
using ClustalW [24] showed that those with a proven
or putative cleavage site in the calf-2 module have
longer loops between conserved cysteine residues than
uncleaved integrins (Supporting information Fig. S1).
Two structured stretches (mainly a-helical, some
extended) were consistently predicted by ab initio mod-
eling using Robetta [25] within this region, mostly
preserved by the cleavage sites (see Fig. S1; a gallery
of predicted structures is shown in Fig. S2). While a
C. Rosano and M. Rocco Refined a
IIb
b
3
models

FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3191
single, robust conformation clearly cannot be obtained
using these methods alone, we find interesting this pre-
diction of a structured region in all integrins examined.
Therefore, and because a complete structure is
required for reliable hydrodynamic computations, a
best-fitting model generated, using Robetta, for the
G840-Q873 segment was selected and inserted into the
a
IIb
calf-2 module. Finally, models of the carbohydrate
chains were then attached at the previously defined
N- and O-glycosylation points [10], with the only mod-
ifications that, on the basis of potential structural
clashes and a new prediction of the O-glycosylation
sites (see the Materials and methods), S218 and T615
were selected in place of T259 and T619, respectively,
for two O-linked carbohydrates. An overview of this
model can be seen in Fig. 1A,F.
A fully extended, closed conformation (A2bB3-ec;
Fig. 1C,H) was then modeled starting from our previ-
ous A2b-5 ⁄ ET model [10] and manually replacing by
structural superimposition each module in a
IIb
with
the corresponding ones from the A2bB3-bc model. A
slightly lower opening angle (134° versus 143°) between
the thigh and calf-1 domains was employed to better
accommodate the a
IIb

subunit in the ET map [9]. The
a
IIb
b-propeller and the b
3
bA ⁄ hybrid ⁄ plexin ⁄ sema-
phorin ⁄ integrin (PSI) domains of the 3FCS structure
[11] were inserted as a single block to preserve the
a
IIb
⁄ b
3
interface. The b
3
I-EGF1-4 and bTD domains
were then added in the bent conformation, superim-
posing the latter on its counterpart taken from the
A2b-5 ⁄ ET original model. While a span of $ 20° was
very recently found for the interdomain angle between
AB
CDE
FG
HIJ
Fig. 1. An overview of the new, refined models. Models A2bB3–bc (panels A and F), A2bB3-pec (panels B and G), A2bB3-ec (panels C
and H), A2bB3-eotc (panels D and I) and A2bB3-eots (panels E and J), are shown as ribbons (protein only, panels A–E) and as surface
(protein) and space-filling (carbohydrates and OG moieties) representations (panels F–J). The a
IIb
b
3
modules are indicated in panel E, and

color-coded in the same way in all models (in addition, carbohydrates are yellow, and OG molecules are orange). In panels H–J, a mesh
representation (purple) of the ET map [9] is superimposed on the models.
Refined a
IIb
b
3
models C. Rosano and M. Rocco
3192 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
the I-EGF1-2 modules during the examination of 10
independent molecules in a
x
b
2
crystals [13], Normal
Modes Analysis (NMA) [26] of this region did not
reveal any freely extensible joint but only a sideways
oscillation, as previously observed with the a
v
b
3
-
derived models [10]. Mutating in silico the hinge cyste-
ine residues C473 and C503 to alanines did not sub-
stantially change the situation. The extension of this
segment was then performed by aligning the b
3
C473-
C503 residues on the z-axis and manually performing
an opening of $ 40° between the I-EGF1-2 modules
along this hinge. To reconnect this part of the chain

with the bA ⁄ hybrid ⁄ PSI head domain, we first gener-
ated a pathway of 12 structures from the fully closed
to the complete swung-out hybrid domain (taken from
the 3FCU structure [6]), using the Yale morph server
[27]. Aligning all these structures on the bA domain in
the A2bB3-extended ⁄ closed (ec) model framework
revealed a conformation that could be linked to the
extended I-EGF1-4 ⁄ bTD segment. A swing-out, of
$ 6°, of the hybrid domain was present in this confor-
mation, consistent with that of $ 10° observed in the
a
v
b
3
ectodomain fitted in the TEM map [7]. Finally,
after introduction of the mature a
IIb
cut at R856 [28],
Robetta [25] was again employed to remodel the
G847-R856 and D857-C879 stretches.
While an initial version of this work was undergoing
evaluation, a SANS study of native, full-length a
IIb
b
3
solubilized in Triton X-100 was published [16]. In that
study, low-resolution ($ 20 A
˚
) shape reconstruction
from the I(q) versus q profiles using dammin [29] and

related programs [16] produced dummy atom bead
models that were judged to be compatible with the
bent ⁄ closed conformation [16]. A pairwise distance dis-
tribution function p(r) versus r curve was also derived
from the experimental data (Fig. 1b of ref. [16]), allow-
ing a direct comparison with the p(r) versus r that
could be calculated from our models using the new
US-SOMO implementation [22]. Because the SANS
data were collected at a content of 16% D
2
O, con-
trast-matching the Triton X-100 detergent, the OG
moieties were removed from our models before the
p(r) versus r computation was performed. As can seen
in Fig. 2, neither the A2bB3-bc (blue line) nor the
A2bB3-ec (magenta line) p(r) versus r curves matched
the SANS-derived curve (black line-connected circles).
A series of 30 intermediate conformations were then
generated by morphing between the A2bB3-bc and the
A2bB3-ec models, without the carbohydrates attached,
using the Yale morph server [27], and their p(r) versus
r curves were computed and compared with the
SANS-derived curve. The carbohydrates were then
re-attached to a restricted number of models having
the closest fit, and a best-matching, partially extended ⁄
closed model, A2bB3-partially extended ⁄ closed (pec),
was then individuated (Fig. 2, orange line), having a
angle of $ 80° between the thigh and calf-1 domains
(Fig. 1B,G). However, the less-than-optimal fit in the
high r region suggested that the experimental data

could also derive from a mixture of at least two well-
defined conformations. This was confirmed by the
green curve in Fig. 2, which was obtained by averag-
ing, in a 1 : 1 ratio, the p(r) versus r of models
A2bB3-bc and A2bB3-ec. Although we did not
attempt any further refinement, it is conceivable that a
closer match could be obtained by mixing differently
bent and extended models in appropriate ratios.
Finally, to allow a comparison to be made with the
hydrodynamics of the other models, the OG moieties
were re-attached to the A2bB3-pec model.
The transition to the extended-open forms was then
achieved, starting from the A2bB3-ec model, by first
restoring to 143° the angle between the thigh and calf-
1 domains, and then superimposing the a
IIb
b-propeller
and the b
3
bA ⁄ hybrid ⁄ PSI domains of the 3FCU
structure [6]. An initial extended ⁄ open ⁄ tail separated
model [A2bB3-extended ⁄ open ⁄ tail separated (eots);
Fig. 1E,J] was then completed by allowing the b
3
I-EGF1-4 and bTD domains to follow the swing-out,
and repositioning the b
3
TM helix on the same plane
of its a
IIb

counterpart; each helix was surrounded by
its own 50-molecule OG micelle. We did not use a
homology model of the b
2
fully extended PSI ⁄
hybrid ⁄ I-EGF1-3 2P28 crystal structure [30] because it
Fig. 2. Comparison between the SANS- and the models-derived
pairwise distance distribution function p(r) versus r. The SANS-
derived p(r) versus r function, digitized from Fig. 1b of ref. [16], is
shown as black circles connected with a black line. Also shown are
the calculated p(r) versus r functions for models A2bB3-bc (blue
line), A2bB3-pec (orange line), A2bB3-ec (magenta line), and for a
1 : 1 mixture of models A2bB3-bc and A2bB3-ec (green line). OG
moieties were removed from the models before computation of
the p(r) versus r function (see the text for details).
C. Rosano and M. Rocco Refined a
IIb
b
3
models
FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3193
would have pushed the b
3
tail outside the hypothetical
membrane plane [10]. The other extended ⁄ open model,
[A2bB3-extended ⁄ open ⁄ tails crossed (eotc) (Fig. 1D,I)],
was then made by first inserting the computer-based
models of the TM helices [18,19] in the ‘open’ state,
superimposing the two a
IIb

segments and re-attaching
the NMR-based cytoplasmic domains. The b
3
subunit
was then reconnected by first pivoting back the
I-EGF3-4 and bTD domains by $ 29° along the C473-
C503 hinge, and then performing a slight rotation of
the bTD module at the main chain of D606 to dock it
against the calf-2 domain with its C-terminus in prox-
imity of the N-terminus of the b
3
‘open’ TM helix.
Overall evaluation of the models
The five new a
IIb
b
3
models are shown in ribbon repre-
sentation in Fig. 1A–E, with their color-coded mod-
ules indicated in Fig. 1E. Surface and space-filling
representations are then proposed in Fig. 1F–J, where
the ET map [9] is superimposed to the fully extended
models. Note how the a
IIb
extracellular modules, and
the b
3
bA module in the extended ⁄ closed conforma-
tion, fit extremely well in the map, with the b
3

hybrid ⁄ PSI domains nicely occupying the remaining
electron density blob in the extended ⁄ open conforma-
tions. Furthermore, note how the left-side inferior lobe
of the ET map is larger and could easily accommodate
at least part of the b
3
I-EFG4 ⁄ bTD domains in both
the closed and open ⁄ tails-crossed models, again sug-
gesting that in solubilized, primed a
IIb
b
3
, full tail sepa-
ration is not the most populated state. Close-up views
of the b
3
pivot region, highlighting the potential hinge
movement, are shown in Fig. 3A–D. The C-terminal
region of the models’ ectodomains, including the mod-
eled E764-D774 and G840-Q873 loops, the attachment
of the ectodomain to the TM helices, and the inter-
faces between the calf-2 and bTD modules, are also
shown in Fig. 3E–G. Note the partial loss of the pre-
dicted P851-I861 helix within the G840-Q873 loop
after introduction of the mature R856-D857 proteo-
lytic cut.
In Table 1 we present the computed sedimentation
and diffusion coefficients for the five new models, com-
pared with the experimental values (<s
0

ð20;wÞ
>
w
and
<D
0
tð20;wÞ
>
z
) for resting and primed a
IIb
b
3
[10]. As seen,
the fully bent and partially extended models of the
resting integrin are incompatible with the experimental
data, while the extended ⁄ closed model is in excellent
agreement, as previously noted for the less-refined
A2b-5 ⁄ ET model [10]. As for the extended ⁄ open struc-
tures, maintaining the TM helices in contact leads to
hydrodynamic values in much better agreement (well
within the experimental range) than those of the full
swing-out model.
While this work was being prepared for resubmis-
sion, another work concerning the a
IIb
b
3
structure
appeared [17]. In this study, purified full-length a

IIb
b
3
was inserted into phospholipid nanodiscs [31], chroma-
tographed and studied using EM and analytical ultra-
centrifugation. In order to compare our models with
this new evidence, we inserted them into atomic-scale
models of the nanodiscs, which were prepared starting
from a model containing a MSP1D1 membrane pro-
tein scaffold and 2 · 80 1,2-dipalmitoyl-sn-glycero-
3-phosphocholine (DPPC) phospholipid molecules in
the bilayer [31] (kindly provided by A. Shih and
S. G. Sligar, University of Illinois, IL, USA). The
DPPC phospholipids were mutated in silico to 1,2-di-
myristoyl-sn-glycero-3-phosphocholine (DMPC) by
removing the last two C atoms from the two hydro-
phobic tails, followed by a translation, of $ 2A
˚
,of
each layer in a direction perpendicular to the disc’s
plane, thereby providing an approximate restoration of
the bilayer interface, without any further optimization.
The total number of phospholipids was not changed,
as MSP1 ⁄ DMPC nanodiscs were found to contain
about the same number of units, 154 ± 4 [32]. Then,
every second DMPC phosphocholine head was
replaced with a phospho-(1¢-rac-glycerol) moiety to
convert them into 1,2-dimyristoyl-sn-glycero-3-phos-
pho-(1¢-rac-glycerol) (DMPG) phospholipids, assuming
a 1 : 1 ratio of DMPC : DMPG in the nanodiscs

used experimentally [17]. The His-tag and TEV prote-
ase sequence (MGHHHHHHDYDIPTTENLYFQG),
which was absent in the atomic model but present in
the original MSP1D1 construct and not removed
by Ye et al. [17], was modeled by a combination of
ab initio modeling using Robetta [25] and insertion
from the 3GZH.pdb structure (YDIPTTENLYFQG).
It was then grafted at the two N-termini of the
MSP1D1-DMPC ⁄ DMPG nanodisc structure, again
without any further optimization. The computed
molecular mass for this nanodisc model,
156 853 gÆmol
)1
, is in good agreement with the experi-
mental value, of 181 500, measured by Ye et al. [17],
which was recalculated after correcting for the com-
puted

v of the nanodisc, 0.875 cm
3
Æg
)1
(F. Ye and
A. Bobkov, personal communication). No further opti-
mization of the DMPC : DMPG ratio was attempted,
as it would have only slightly affected the computed
hydrodynamic parameters. About 18 DMPC : DMPG
molecules were then removed to make space for the
TM helices of the integrins, either from the center of
the nanodisc or close to the protein belt, to take into

account possible effects of the high gravitational field
Refined a
IIb
b
3
models C. Rosano and M. Rocco
3194 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
present during ultracentrifugation. In the lateral con-
figuration, two different orientations of the integrins
were considered (‘internal’ and ‘external’, see Fig. 4),
thus taking into account potential asymmetry effects.
These integrin ⁄ nanodisc models allowed computation
of the sedimentation coefficients and their comparison
with the published value [17], corrected to standard
conditions. Four representative models can be seen in
Fig. 4, and a comparison between experimental and
computed s values is reported in Table 2. Similarly to
the OG-solubilized integrins, and in contrast with the
EM images of Ye et al. [17], the data in Table 2 show
AB
C
EFG
D
Fig. 3. Close-up views of selected regions in the new models. Panels A-D: close-up views of the hinge region in four of the new models
(A2bB3-bc, panel A; A2bB3-ec, panel B; A2bB3-eotc, panel C; A2Bb3-eots, panel D). The modules are aligned on the I-EGF3 module and
color-coded as indicated in panel A, and the D477-Q482 loop (dark green) and the C473-C503 hinge disulfide (cyan) are clearly visible. The
blue sticks are the other disulfides present in the I-EGF1 ⁄ 2 modules. Panels E–F: close-up views of the bottom region of the ectodomain,
connected to the TM helices, in models A2bB3-bc (panel E), A2bB3-ec (panel F) and A2bB3-eotc (panel G). The modules are color-coded as
indicated in panel G. The G840-Q873 modeled loop is clearly visible (orange-red), together with the post-translational cut at R856 and the
new N-terminus at D857 (grey). The E764-D774 modeled loop is colored magenta.

C. Rosano and M. Rocco Refined a
IIb
b
3
models
FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3195
that the fully bent integrin-conformation is at odds
(+12 ) 14%) with the solution data, whereas the
extended ⁄ closed conformation is in very good agree-
ment with them (+1 ) 3%).
Discussion
The revised a
IIb
b
3
models, which are now based on
more complete structural evidence, allow a better eval-
uation of proposed alternative conformational states
and the supposed transitions between them. To begin
with, the fully bent, crystallographic state observed for
the a
v
b
3
, a
IIb
b
3
and a
x

b
2
ectodomains has now been
found to be at odds with data on full-length, solubi-
lized samples, either obtained by ‘bulk’ technologies,
such as analytical ultracentrifugation [10,17], dynamic
light scattering [10] and the recent SANS study [16], or
by single particle averaging following ‘gentler’ EM
methods, such as cryo-EM [8] and ET [9]. Interest-
ingly, from the hydrodynamic data, a strong consensus
on the degree of extension between a
IIb
b
3
, solubilized
either in OG [10] or in nanodiscs [17], is emerging.
Both data sets support a predominance of nearly fully
extended structures in the resting state, fitting very well
inside the ET map [9] (Tables 1 and 2; Fig. 1C,H;
Fig. 4B,D). However, the SANS data [16] are compati-
ble with either an intermediate extension (Fig. 1B,G),
or, more likely, in our opinion, with an approximately
equimolecular mixture of bent and extended structures
(see Fig. 2). Possible reasons for this discrepancy might
derive from the different purification and solubilization
procedures used in the SANS study, which involved a
freeze-drying step and utilized Triton X-100 as the
detergent [16]. In particular, the micelles formed by
Triton X-100 are much larger than those formed by
OG (radii of 4.7 versus 2.3 nm, respectively [33,34]),

and should clearly engulf also the cytoplasmic
domains. This could potentially deregulate any confor-
mational control exerted at the inner membrane inter-
face, which should instead be preserved in the
OG- and nanodiscs-solubilized samples.
The new a
IIb
b
3
and a
x
b
2
ectodomain structures have
been used to reinforce the scheme calling for bent rest-
ing integrins with a transition to fully extended, open,
tails separated structures following activation [11,13].
The recently published negative-staining EM images,
derived either from ectodomain constructs [13] or from
full-length, nanodisc-solubilized samples [17], also
show a predominance of bent forms in the resting
state. Regarding the ectodomain studies, an important
question can be raised of whether it is more ‘physio-
logical’, for example, a truncated construct in which
the TM and cytoplasmic regions are absent, immobi-
lized either in a crystal lattice or on an EM grid, or is
a full-length, native molecule isolated in a small
micelles-forming mild detergent or in a confined lipid
bilayer. While good arguments could be made either
way, is interesting to note that previous TEM work [7]

has shown that integrin ectodomains can bind macro-
molecular ligands without relevant structural changes,
suggesting that extension is not intrinsically needed for
activation. This was further supported by recent fluo-
rescent lifetime imaging microscopy on full-length a
v
b
3
in live cells that did not detect any change in height
following activation [12]. However, it could instead be
that the lack of the TM and cytoplasmic regions
and ⁄ or the environmental conditions specifically dereg-
ulate fundamental conformational transitions in other
integrins such as a
IIb
b
3
. The probable co-existence of
bent and extended forms in Triton X-100, as suggested
by our analysis of the recent SANS data [16] (see
Fig. 2), further reinforce the hypothesis that the cyto-
plasmic and TM regions are involved in this event.
However, it cannot be excluded that the purification
procedures and the presence of detergent favor the
extended conformation, although no signs of extra OG
binding, besides presumably around the TM segments,
were found in our solution studies [10], whereas the
Table 1. Comparison between experimental and computed hydrodynamic parameters for full-length, OG-solubilized a
IIb
b

3
and the refined
models.
Experimental ⁄ models <s
0
ð20;wÞ
>
w
(S)
Percentage ± SEM or
percentage difference <D
0
tð20;wÞ
>
z
(F)
Percentage ± SEM or
percentage difference
a
IIb
b
3
resting 8.18 ± 0.07
a
± 0.9 3.07 ± 0.08
a
± 2.6
A2bB3-bc (bent ⁄ closed) 9.08 + 11.0 3.31 + 7.8
A2bB3-pec (partially extended ⁄ closed) 8.99 + 9.9 3.27 + 6.5
A2bB3-ec (extended ⁄ closed) 8.20 + 0.2 2.98 )2.9

a
IIb
b
3
primed 7.65 Ä 7.97
b
)6.5 Ä )2.6
c
2.64 Ä 2.99
b
)14 Ä )2.5
c
A2bB3-eots (extended ⁄ open ⁄ tails separated) 7.41 )9.4
c
2.62 )14.7
c
A2bB3-eotc (extended ⁄ open ⁄ tails crossed) 7.81 )4.5
c
2.84 )7.5
c
a
From a previous publication [10].
b
Range of experimental values as presented previously [10].
c
Percentage difference from the resting
experimental values.
Refined a
IIb
b

3
models C. Rosano and M. Rocco
3196 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
phospholipid moieties are confined within the scaffolds
of the nanodiscs. Furthermore, the extended ⁄ closed
state appears to be a stable conformation in purified,
solubilized full-length a
IIb
b
3
, which is fully able to
make the transition to the extended ⁄ open state upon
priming and to revert back to the extended ⁄ closed
state when the priming agent is removed [35]. In addi-
tion, the transition to the open state can take place on
the cell membrane with changes in the integrin height
that vary widely from system to system (e.g. [36]; see
also [37,38], and references therein). Instead, the recent
studies with nanodiscs-embedded full-length a
IIb
b
3
[17]
provide apparently contradictory results: while the
hydrodynamic data are fully consistent with an
extended resting integrin, the EM images show a pre-
dominance of bent forms. Importantly, the shape dis-
tribution appears to be rather bimodal, with the
molecules assuming either a bent or an extended form,
and an absence of well-defined intermediate conforma-

tions. A similar bimodal distribution was also recently
AB
CD
Fig. 4. Nanodiscs-embedded a
IIb
b
3
bent ⁄
closed and extended ⁄ closed models.
MSP1D1-DMPC ⁄ DMPG-nanodiscs center-
embedded a
IIb
b
3
models A2bB3-bc-ndc
(panel A) and A2bB3-ec-ndc (panel B) and
laterally embedded A2bB3-bc-ndle (panel C,
‘external’ orientation) and A2bB3-ec-ndli
(panel D, ‘internal’ orientation). The a
IIb
b
3
and MSP1D1 protein regions are shown
in ribbon representation inside a
semitransparent surface. For a
IIb
b
3
, the
color coding is as in Fig. 1, while the two

MSP1D1 identical subunits are orange and
red, respectively. The carbohydrates are
shown as gold sticks, and the DMPC and
DMPG lipids are shown in space-filling
mode (light grey, DMPC; slate grey,
DMPG).
C. Rosano and M. Rocco Refined a
IIb
b
3
models
FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3197
observed in the EM images of the a
x
b
2
ectodomain
[13]. Thus, true to the previously proposed ‘switch-
blade’ extension mechanism [39], the conformational
change indeed appears to be snap-like, albeit not
directly linked to priming or activation. In this light, it
could be hypothesized that interactions with the EM
grid favor a snap-back to the bent conformation, thus
offering an explanation for the apparent contradiction
between the EM and the hydrodynamic data [17].
Interestingly, the recent a
IIb
b
3
-nanodiscs EM study

shows an increase of extended conformations upon
binding of a talin head domain to the cytoplasmic tails
[17], suggesting that this interaction indeed stabilizes
the extended form, at least partially preventing a snap-
back to the bent form upon deposition on the EM
grid.
As for the full tail separation, while it cannot be
excluded that the constraints imposed by the solubiliz-
ing micelle could somewhat oppose it, it should be
noted that in this system, after prolonged incubation
with priming agents, the formation of dimers, trimers
and oligomers seems not to be impaired [40], implying
at least a re-arrangement at the TM level. In addition,
we would like to point out how the new EM images of
talin head domain-bound full-length a
IIb
b
3
integrin
embedded in nanodiscs [17] are fully consistent with
our previously developed model of extended ⁄ open
integrins without tail separation [10], which has been
further refined in the present study (Fig. 1D,I;
Table 1). The phospholipids environment and the rela-
tively large radius of the nanodiscs should provide
ample space to permit the integrin TM domains to
fully separate on activation, and yet no clear-cut evi-
dence of such an event emerges from the new EM
images [17]. Moreover, it is interesting to note that the
pivot points probably utilized on extension [11] could

as well act as a ‘universal’ joint, coupling the swing-
out to a simple conformational change in the TM heli-
ces. In this respect, recent work has claimed that the
b
3
S527F mutation in a
IIb
b
3
induces the high-affinity
state by hindering the adoption of the bent conforma-
tion [41]. However, the closest residue in the ‘head’
region, S401 of a
IIb
, is more than 20 A
˚
away in the
bent conformation, while the S527F mutation perturbs
a cluster of polar residues (R671 and N675 in a
IIb
,
R498, D524, S527, R530, D546 and Y556 in b
3
) both
in the bent and in the extended-closed conformations.
Furthermore, the S527-containing C523-C544 loop
appears to be only slightly perturbed by the pivoting
movement. Therefore, it seems more likely that this
mutation affects only the swing-out, thus favoring the
high-affinity state, without any implication for the

bent-to-extended transition. Still in this respect, it has
been proposed that integrins are inherently flexible,
continuously exploring the conformations from bent to
extended (e.g. [3]), but NMA analysis does not support
a fully flexible hinge at the I-EGF1-2 interface. In
addition, a Kratky plot of the SANS data has pro-
vided no evidence of intrinsic flexibility in Triton
X-100-solubilized full-length a
IIb
b
3
(see Fig. S3 in
[16]). While the extension could clearly take place
around the b
3
C473-C503 disulfide bond, this move-
ment appears to require an ‘external’ driving force.
This could result from conformational changes at the
cytoplasmic face, for instance following interactions
with cytoskeleton components, or at the ectodomain
ligand-binding region, such as in small molecules-
induced priming events. However, given the complex
disulfide pattern in this region (see Fig. 3A–D), it
could also be hypothesized that the extension is
initially brought about under enzymatic control, as
previously suggested (see [42], and references therein),
perhaps following integrin insertion in the outer cell
membrane.
Strong support for the bent-to-extended transition
taking place on activation has instead been claimed

from several antibody-binding studies (see [3], and
references therein), recently revisited on the basis of
the new a
x
b
2
structure [13]. We have mapped the
Table 2. Comparison between experimental and computed sedi-
mentation coefficients for full-length a
IIb
b
3
embedded in the
MSP1D1-DMPC ⁄ DMPG nanodiscs and the new models.
Experimental ⁄ models <s
ð20;wÞ
>
w
(S)
Percentage
difference
a
IIb
b
3
resting 9.02
a
na
A2bB3-bc-ndc (bent ⁄ closed,
nanodisc, centered)

10.3 + 14.2
A2bB3-bc-ndli (bent ⁄ closed,
nanodisc, lateral ⁄ internal
orientation)
10.3 + 14.2
A2bB3-bc-ndle (bent ⁄ closed,
nanodisc, lateral ⁄ external
orientation)
10.1 + 12.0
A2bB3-pec-ndc (partially
extended ⁄ closed, nanodisc,
centered)
10.1 + 12.0
A2bB3-ec-ndc (extended ⁄
closed, nanodisc, centered)
9.31 + 3.2
A2bB3-ec-ndli (extended ⁄
closed, nanodisc, lateral ⁄
internal orientation)
9.26 + 2.7
A2bB3-ec-ndle (extended ⁄
closed, nanodisc, lateral ⁄
external orientation)
9.13 + 1.2
a
From a previous publication [17], after correction to standard
conditions.
Refined a
IIb
b

3
models C. Rosano and M. Rocco
3198 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
antibody-binding sites identified in a
L
b
2
as markers of
activation [43–45] on our refined models A2bB3-bc,
A2bB3-ec and A2bB3-eotc (see Supporting informa-
tion Fig. S3). While the patches corresponding to the
NKI-L16 and AO3 anti-a
L
antibody-binding sites seem
to be more exposed in the extended ⁄ closed and exten-
ded ⁄ open conformations (Fig. S3, panels B–C), they
do not appear to be buried in the bent ⁄ closed confor-
mation (Fig. S3, panel A). Noting that the majority of
these epitope residues in the thigh module are close to
the junction with the b-propeller module, it is possible
that the differences in antibody binding stem from
changes in the relative orientation of these two mod-
ules. Indeed, a variation of $ 18° is observed between
the thigh ⁄ b-propeller modules in 10 independent mole-
cules in a
x
b
2
crystals [13], suggesting a degree of flexi-
bility similar to that present at the I-EGF1 ⁄ 2 interface.

These changes could be induced by binding events at
the b-propeller ⁄ bI interface, or be a long-range propa-
gation of the hybrid domain swing-out. As for the
CRB-LFA-1 ⁄ 2 and KIM127 epitopes mapped on the
b
3
subunit, they are clearly buried in the bent confor-
mation (Fig. S3, panel A). However, antibody binding
might be somewhat restricted also in the extended ⁄
closed conformation model (Fig. S3, panel B), while
all sites appear to be fully exposed in the extended ⁄
open conformation model (Fig. S3, panel C). Finally,
the activating mutations mapped to the E534 and
M535 residues in b
3
(steel blue spheres in Fig. S3, pan-
els B and C) are more difficult to reconcile with the
extension being uncoupled from activation, as they are
not in contact with other residues in the A2bB3-ec and
A2bB3-eotc models. We note, however, that in the
extended ⁄ open model, the I-EGF3 module (where
these residues reside) has undergone a clockwise rota-
tion that would bring them facing, and possibly con-
tacting, the a
IIb
subunit if this orientation was present
also in the extended ⁄ closed state. Our extended ⁄ closed
model was conservatively developed, introducing the
fewest possible changes not supported by experimental
evidence. However, it cannot be excluded that a rota-

tion of the lower b
3
leg already accompanies extension,
bringing the activating residues into contact with other
residues in the a
IIb
subunit. In addition, some caution
should be exerted in transferring the results of anti-
body-binding and mutational studies, carried out on
different integrins, to the a
IIb
b
3
integrin, which might
be controlled by a different conformational regulatory
system. Given the complexity of the integrin activa-
tion ⁄ signaling network (e.g. see [46] and references
therein) further experimental work is clearly needed to
fully resolve these issues.
The putative structured segment in the calf-2 loop
that was consistently generated by ab initio modeling
in all integrin a subunits that undergo post-transla-
tional cleavage could be physiologically relevant, but
caution should be exerted to avoid overinterpretations
being made. A more extensive study, for instance
involving repeating the ab initio generation after
scrambling the loop sequences, should be performed to
strengthen the prediction, but this is outside the scope
of the present study. In any case, that this loop was
not resolved in any ectodomain crystal structure indi-

cates its inherent flexibility, which should be enhanced
by the post-translational cleavage and is probably
accompanied by loss of the putative secondary struc-
ture in a
IIb
b
3
, where the cut is located in the middle of
the predicted helix. Indeed, the two loop fragments are
outside the ET map in the models shown in Fig. 1.
Apart from studies demonstrating the necessity of this
cleavage for the proper function of a
v
b
5
[47], little is
known of its role in integrins lacking an a-I ⁄ A domain
(see also [1]). With the above-mentioned caveats, it is
tempting to speculate that the putative structure in this
region is involved in interactions with other membrane
proteins, perhaps helping integrins to assume or modu-
late their resting conformation. NMR structural stud-
ies of this module, with and without proteolytic
cleavage, could help to clarify this issue.
To conclude, our work further suggests that all
evidence should be accounted for when proposing inte-
grin activation mechanisms. In particular, the bent-to-
extended transition has been repeatedly implied as a key
step in integrin activation [3,11,13], but the solution evi-
dence on full-length integrins seems to indicate other-

wise. Although the thermodynamic cost of such a huge
conformational change could still be manageable, its
physiological need is unclear, and alternative explana-
tions for the role of the bent conformation, such as the
‘packaging for transport’ hypothesis that we have
already proposed [10], should be investigated on a case-
by-case basis. The refined models presented here, which
have been deposited in the public Protein Model Data-
Base (PMDB; should also
aid the design of mutational studies aimed to fully clar-
ify these important issues. While we recognize that they
clearly cannot provide definitive atomic details of unre-
solved, modeled segments, such as the calf-2 long loop
and the relative orientation of the ectodomain ⁄ TM
regions, or of the changes induced by extension not seen
in crystal studies, we exerted great care and chose con-
servative assumptions during modeling. Our refined
models should also allow more realistic steered mole-
cular dynamics simulations of the conformational
C. Rosano and M. Rocco Refined a
IIb
b
3
models
FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3199
transitions of integrins (e.g. [11]) to be performed, for
instance by the use of nanodiscs-embedded models.
Materials and methods
Molecular models were built and refined mostly as previ-
ously described [10]. The Modeller ModLoop utility [23]

( was used for
intermediate-length loop modeling, while for the longer,
non-resolved loop in a
IIb
, the Robetta ab initio approach
[25] was employed on a dedicated webserver (http://robetta.
bakerlab.org/) with default parameters. The hydrodynamic
parameters were computed using the SOMO bead-modeling
approach [20] with the US-SOMO implementation [21,22]
( The US-SOMO set-
tings were: accessible surface area cut-offs of 20 A
˚
and
50% for the atomic structures and for the beads in the
models, respectively; synchronous overlap removal in all
steps, with outward translation of the exposed side-chain
beads; bead fusion thresholds of 70% between exposed
beads; exclusion of the buried beads from the hydro-
dynamic computations and volume corrections; and
stick boundary conditions and computation referred to the
diffusion center. The partial specific volume

v of the single-
micelle models was 0.723 cm
3
Æg
)1
, while that of the two-
micelles model was 0.731 cm
3

Æg
)1
[10]. The

v values for
DMPC and DMPG were 0.973 cm
3
Æg
)1
(adjusted at 20 °C
from the experimental value at 30 °C [48]) and 0.925
cm
3
Æg
)1
(adjusted at 20 °C from the calculated [49] value at
25 °C), respectively. The calculated

v value at 20 °C for the
MSP1D1 nanodisc scaffold protein with the initial methio-
nine, the His-tag and the TEV protease site, was 0.731 cm
3
Æ
g
)1
. For the nanodiscs-embedded integrins, the calculated

v
at 20 °C thus was 0.776 cm
3

Æg
)1
, assuming a 1 : 1 ratio of
DMPC : DMPG and 142 total lipid molecules (160 ) 18
that were removed to make space for the TM regions of
the integrins). The pairwise-distance distribution function
utility of the new SAXS ⁄ SANS simulation module within
US-SOMO [22] was used to compute the p(r) versus r for
the models. The experimental p(r) versus r data were
digitized from the data in Fig. 1b of ref. [16] using the
DigitizeIt 1.5 shareware program (itizeit.
de/), and the p(r) was then normalized to 1.0. The YinOY-
ang 1.2 [50] ( />was employed for O-glycosylation predictions. NMA was
performed utilizing the elNemo server [26] (.
cnrs-mrs.fr/elnemo/), as previously reported [10]. Morphing
between conformations was carried out with the Yale
morph server [27] ( using
CNS (adiabatic mapping) interpolation. Alignments were
performed using ClustalW [24] on the UniProt server
( with default settings. With the
exception of Fig. S2, molecular graphics images were
produced using the UCSF Chimera package [51] (alpha
version 1.3, build 2577) from the Resource for Biocomput-
ing, Visualization, and Informatics at the University of
California, San Francisco (supported by NIH P41 RR-
01081; which was also
used for fitting models inside the ET map. Paint Shop
Pro 5 (Jasc Sofware, Corel Inc., Mountain View, CA,
USA; ) was used to assemble all
figures.

Acknowledgements
We thank R.R. Hantgan (Wake Forest University,
NC, USA) for comments. We are grateful to M.A. Ar-
naout (Harvard Medical School, Charlestown, MA,
USA) for very kindly providing us with the coordinates
of the new a
v
b
3
ectodomain structure before public
release; to F. Ye, M. Ginsberg (UCSD, CA, USA) and
A. Bobkov (The Burnham Institute, CA, USA) for pro-
viding important details of their experimental work and
feedback on our calculations of the nanodiscs proper-
ties; and to A. Shih and S.G. Sligar (University of Illi-
nois at Urbana-Champaign, IL, USA) for providing a
nanodisc atomic model and related information. We
are indebted to E. Brookes (UTHSCSA, San Antonio,
TX, USA) for his constant and timely improvements to
the US-SOMO program. This work was partially sup-
ported by the Italy-USA project ‘Farmacogenomics
oncology – Oncoproteomics’ (Grant 527B ⁄ 2A ⁄ 3) to
MR. The models accession codes in the PMDB
database are PM0076386 (A2bB3-bc), PM0076372
(A2bB3-pec), PM0076362 (A2bB3-ec), PM0076363
(A2bB3-eotc), and PM0076371 (A2bB3-eots).
References
1 Hynes RO (2002) Integrins: bidirectional, allosteric sig-
naling machines. Cell 110, 673–687.
2 Arnaout MA, Goodman SL & Xiong J-P (2007) Struc-

ture and mechanics of integrin-based cell adhesion. Curr
Opin Cell Biol 19, 495–507.
3 Luo BH, Carman CV & Springer TA (2007) Structural
basis of integrin regulation and signaling. Annu Rev
Immunol 25, 619–647.
4 Xiong JP, Stehle T, Diefenbach B, Zhang R, Dunker
R, Scott DL, Joachimiak A, Goodman SL & Arnaout
MA (2001) Crystal structure of the extracellular seg-
ment of integrin a
v
b
3
. Science 294, 339–345.
5 Xiong JP, Stehle T, Zhang R, Joachimiak A, Frech M,
Goodman SL & Arnaout MA (2002) Crystal structure
of the extracellular segment of integrin a
v
b
3
in complex
with an Arg-Gly-Asp ligand. Science 296, 151–155.
6 Xiao T, Takagi J, Coller BS, Wang JH & Springer TA
(2004) Structural basis for allostery in integrins and
binding to fibrinogen-mimetic therapeutics. Nature 432,
59–67.
Refined a
IIb
b
3
models C. Rosano and M. Rocco

3200 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS
7 Adair BD, Xiong JP, Maddock C, Goodman SL,
Arnaout MA & Yeager M (2005) Three-dimensional
EM structure of the ectodomain of integrin aVb3in
a complex with fibronectin. J Cell Biol 168, 1109–
1118.
8 Adair BD & Yeager M (2002) Three-dimensional model
of the human platelet integrin a
IIb
b
3
based on electron
cryomicroscopy and x-ray crystallography. Proc Natl
Acad Sci USA 99, 14059–14064.
9 Iwasaki K, Mitsuoka K, Fujiyoshi Y, Fujisawa Y,
Kikuchi M, Sekiguchi K & Yamada T (2005) Electron
tomography reveals diverse conformations of integrin
aIIbb3 in the active state. J Struct Biol 150, 259–267.
10 Rocco M, Rosano C, Weisel JW, Horita DA &
Hantgan RR (2008) Integrin conformational regulation:
uncoupling extension ⁄ tail separation from changes in
the head region by a multiresolution approach.
Structure 16, 954–964.
11 Zhu J, Luo B-H, Xiao T, Zhang C, Nishida N &
Springer TA (2008) Structure of a complete integrin
ectodomain in a physiologic resting state and activation
and deactivation by applied forces. Mol Cell 32,
849–861.
12 Xiong JP, Mahalingham B, Alonso JL, Borrelli LA,
Rui X, Anand S, Hyman BT, Rysiok T, Mu

¨
ller-Pom-
palla D, Goodman SL et al. (2009) Crystal structure of
the complete integrin aVb3 ectodomain plus an a ⁄ b
transmembrane fragment. J Cell Biol 186, 589–600.
13 Xie C, Zhu J, Chen X, Mi L, Nishida N & Springer
TA (2010) Structure of an integrin with an aI domain,
complement receptor type 4. EMBO J 29, 666–679.
14 Lau T-L, Kim C, Ginsberg MH & Ulmer TS (2009)
The structure of the integrin aIIbb3 transmembrane
complex explains integrin transmembrane signaling.
EMBO J 28, 1351–1361.
15 Yang J, Ma YQ, Page RC, Misra S, Plow EF & Qin J
(2009) Structure of an integrin aIIbb3 transmembrane-
cytoplasmic heterocomplex provides insight into integrin
activation. Proc Natl Acad Sci USA 106, 17729–17734.
16 Nogales A, Garcı
´
aC,Pe
´
rez J, Callow P, Ezquerra TA
& Gonza
´
lez-Rodrı
´
guez J (2010) Three-dimensional
model of human platelet integrin aIIbb3 in solution
obtained by small angle neutron scattering. J Biol Chem
285, 1023–1031.
17 Ye F, Hu G, Taylor D, Ratnikov B, Bobkov AA,

McLean MA, Sligar SG, Taylor KA & Ginsberg
MH (2010) Recreation of the terminal events in
physiological integrin activation. J Cell Biol 188,
157–173.
18 Gottschalk KE, Adams PD, Brunger AT & Kessler H
(2002) Transmembrane signal transduction of the a
IIb
b
3
integrin. Protein Sci 11, 1800–1812.
19 Gottschalk KE (2005) A coiled-coil structure of the
aIIbb3 integrin transmembrane and cytoplasmic
domains in its resting state. Structure 13, 703–712.
20 Rai N, No
¨
llmann M, Spotorno B, Tassara G, Byron O
& Rocco M (2005) SOMO (SOlution MOdeler): differ-
ences between X-ray and NMR-derived bead models
suggest a role for side chain flexibility in protein hydro-
dynamics. Structure 13, 723–734.
21 Brookes E, Demeler B, Rosano C & Rocco M (2010)
The implementation of SOMO (SOlution MOdeller) in
the UltraScan analytical ultracentrifugation data analy-
sis suite: enhanced capabilities allow the reliable hydro-
dynamic modeling of virtually any kind of
biomacromolecule. Eur Biophys J Biophys Lett 39,
423–435.
22 Brookes E, Demeler B & Rocco M (2010) Develop-
ments in the US-SOMO bead modeling suite: new
features in the direct residue-to-bead method, improved

grid routines, and influence of accessible surface area
screening. Macromol Biosci, in press. doi:0.1002/mabi.
200900474.
23 Fiser A & Sali A (2003) ModLoop: automated model-
ing of loops in protein structures. Bioinformatics 19,
2500–2501.
24 Thompson JD, Higgins DG & Gibson TJ (1994)
CLUSTAL W: improving the sensitivity of progressive
multiple sequence alignment through sequence weight-
ing, position-specific gap penalties and weight matrix
choice. Nucleic Acids Res 22 , 4673–4680.
25 Bonneau R, Strauss CE, Rohl CA, Chivian D, Bradley
P, Malmstro
¨
m L, Robertson T & Baker D (2002) De
novo prediction of three-dimensional structures for
major protein families. J Mol Biol 322, 65–78.
26 Suhre K & Sanejouand YH (2004) ElNemo: a normal
mode web-server for protein movement analysis and the
generation of templates for molecular replacement.
Nucleic Acids Res 32, W610–W614.
27 Krebs WG & Gerstein M (2000) The morph server:
a standardized system for analyzing and visualizing
macromolecular motions in a database framework.
Nucleic Acids Res 28, 1665–1675.
28 Calvete JJ, Scha
¨
fer W, Henschen A & Gonza
´
lez-Rodrı

´
-
guez J (1990) Characterization of the b-chain N-termi-
nus heterogeneity and the a-chain C-terminus of human
platelet GPIIb: posttranslational cleavage sites. FEBS
Lett 272, 37–40.
29 Svergun DI (1999) Restoring low resolution struc-
ture of biological macromolecules from solution
scattering using simulated annealing. Biophys J 76,
2879–2886.
30 Shi M, Foo SY, Tan SM, Mitchell EP, Law SKA &
Lescar J (2007) A structural hypothesis for the transi-
tion between bent and extended conformations of the
leukocyte b2 integrins. J Biol Chem 282, 30198–30206.
31 Denisov IG, Grinkova YV, Lazarides AA & Sligar SG
(2004) Directed self-assembly of monodisperse phospho-
lipid bilayer Nanodiscs with controlled size. J Am Chem
Soc 126, 3477–3487.
C. Rosano and M. Rocco Refined a
IIb
b
3
models
FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS 3201
32 Bayburt TH, Grinkova YV & Sligar SG (2006) Assem-
bly of single bacteriorhodopsin trimers in bilayer nano-
discs. Arch Biochem Biophys 450, 215–222.
33 Hantgan RR, Braaten JV & Rocco M (1993) Dynamic
light scattering studies of a
IIb

b
3
solution conformation.
Biochemistry 32, 3935–3941.
34 Rocco M, Spotorno B & Hantgan RR (1993) Modeling
the a
IIb
b
3
integrin solution conformation. Protein Sci 2,
2154–2166.
35 Hantgan RR, Rocco M, Nagaswami C & Weisel JW
(2001) Binding of a fibrinogen mimetic stabilizes
integrin aIIbb3¢s open conformation. Protein Sci 10,
1614–1626.
36 Ye F, Liu J, Winkler H & Taylor KA (2008) Integrin
a
IIb
b
3
in a membrane environment remains the same
height after Mn
2+
activation when observed by cryo-
electron tomography. J Mol Biol 378, 976–986.
37 Askari JA, Buckley PA, Mould AP & Humphries MJ
(2009) Linking integrin conformation to function. J Cell
Sci 122, 165–170.
38 Askari JA, Tynan CJ, Webb SED, Martin-Fernandez
ML, Ballestrem C & Humphries MJ (2010) Focal adhe-

sions are sites of integrin extension. J Cell Biol 188,
891–903.
39 Beglova N, Blacklow SC, Takagi J & Springer TA
(2002) Cysteine-rich module structure reveals a fulcrum
for integrin rearrangement upon activation. Nat Struct
Biol 9, 282–287.
40 Hantgan RR, Paumi C, Rocco M & Weisel JW
(1999) Effects of ligand-mimetic peptides Arg-Gly-
Asp-X (X = Phe, Trp, Ser) on a
IIb
b
3
integrin
conformation and oligomerization. Biochemistry 38,
14461–14474.
41 Vanhoorelbeke K, De Meyer SF, Pareyn I, Melchior C,
Planc¸ on S, Margue C, Pradier O, Fondu P, Kieffer N,
Springer TA et al. (2009) The novel S527F mutation in
the integrin b3 chain induces a high affinity aIIbb3
receptor by hindering adoption of the bent conforma-
tion. J Biol Chem 284, 14914–14920.
42 Essex DW & Li M (2006) Redox modification of plate-
let glycoproteins. Curr Drug Targets 7, 1233–1241.
43 Lu C, Ferzly M, Takagi J & Springer TA (2001) Epi-
tope mapping of antibodies to the C-terminal region of
the integrin b
2
subunit reveals regions that become
exposed upon receptor activation. J Immunol 166,
5629–5637.

44 Zang Q & Springer TA (2001) Amino acid residues in
the PSI domain and cysteine-rich repeats of the integrin
b2 subunit that restrain activation of the integrin a
X
b
2
.
J Biol Chem 276 , 6922–6929.
45 Xie C, Shimaoka M, Xiao T, Schwab P, Klickstein LB
& Springer TA (2004) The integrin a subunit leg
extends at a Ca
2+
-dependent epitope in the thigh ⁄ genu
interface upon activation. Proc Natl Acad Sci USA 101,
15422–15427.
46 Gahmberg CG, Fagerholm SC, Nurmi SM, Chavakis
T, Marchesan S & Gro
¨
nholm M (2009) Regulation of
integrin activity and signalling. Biochim Biophys Acta
1790, 431–444.
47 Berthet V, Rigot V, Nejjari M, Marvaldi J & Luis J
(2004) The endoproteolytic processing of a
v
b
5
integrin
is involved in cytoskeleton remodelling and cell migra-
tion. FEBS Lett 557, 159–163.
48 Greenwood AI, Tristram-Nagle S & Nagle JF (2006)

Partial molecular volumes of lipids and cholesterol.
Chem Phys Lipids 143, 1–10.
49 Durchschlag H & Zipper P (1994) Calculation of the
partial volume of organic compounds and polymers.
Prog Colloid Polym Sci 94, 20–39.
50 Gupta R & Brunak S (2002) Prediction of glycosylation
across the human proteome and the correlation to pro-
tein function. Pac Symp Biocomput 7, 310–322.
51 Pettersen EF, Goddard TD, Huang CC, Couch GS,
Greenblatt DM, Meng EC & Ferrin TE (2004) UCSF
Chimera – a visualization system for exploratory
research and analysis. J Comput Chem 25, 1605–1612.
Supporting information
The following supplementary material is available:
Fig. S1. Alignment of the long loop region in the calf-
2 module of integrin’s a subunits.
Fig. S2. A gallery of structures predicted by Robetta
[20] for the long loop in the calf-2 region of integrins’
a subunits without the insertion domain.
Fig. S3. Mapping the a
L
and b
2
activation-sensitive
antibodies binding sites on the a
IIb
b3 models A2bB3-
bc (panel A), A2bB3-ec (panel B) and A2bB3-eotc
(panel C), shown in surface representation mode with
the relevant residues highlighted in space-filling mode.

This supplementary material can be found in the
online version of this article.
Please note: As a service to our authors and readers,
this journal provides supporting information supplied
by the authors. Such materials are peer-reviewed and
may be re-organized for online delivery, but are not
copy-edited or typeset. Technical support issues arising
from supporting information (other than missing files)
should be addressed to the authors.
Refined a
IIb
b
3
models C. Rosano and M. Rocco
3202 FEBS Journal 277 (2010) 3190–3202 ª 2010 The Authors Journal compilation ª 2010 FEBS