Structure of the insulin receptor ectodomain reveals
a folded-over conformation
Neil M. McKern1, Michael C. Lawrence1, Victor A. Streltsov1, Mei-Zhen Lou1, Timothy E. Adams1,
George O. Lovrecz1, Thomas C. Elleman1, Kim M. Richards1, John D. Bentley1, Patricia A. Pilling1,
Peter A. Hoyne1, Kellie A. Cartledge1, Tam M. Pham1, Jennifer L. Lewis1, Sonia E. Sankovich1,
Violet Stoichevska1, Elizabeth Da Silva1, Christine P. Robinson1, Maurice J. Frenkel1, Lindsay G. Sparrow1,
Ross T. Fernley1, V. Chandana Epa1& Colin W. Ward1
The insulin receptor is a phylogenetically ancient tyrosine kinase
receptor found in organisms as primitive as cnidarians and
insects. In higher organisms it is essential for glucose homeosta-
(IGF-1R) is involved in normal growth and development2. The
insulin receptor is expressed in two isoforms, IR-A and IR-B; the
former also functions as a high-affinity receptor for IGF-II and is
implicated, along with IGF-1R, in malignant transformation3.
Here we present the crystal structure at 3.8A˚resolution of the
IR-A ectodomain dimer, complexed with four Fabs from
the monoclonal antibodies 83-7 and 83-14 (ref. 4), grown in the
presence of a fragment of an insulin mimetic peptide5. The
structure reveals the domain arrangement in the disulphide-
linked ectodomain dimer, showing that the insulin receptor
adopts a folded-over conformation that places the ligand-binding
regions in juxtaposition. This arrangement is very different from
previous models6. It shows that the two L1 domains are on
opposite sides of the dimer, too far apart to allow insulin to
bind both L1 domains simultaneously as previously proposed7.
Instead, the structure implicates the carboxy-terminal surface of
the first fibronectin type III domain as the second binding site
involved in high-affinity binding.
The structural organization of the insulin receptor (IR; Sup-
plementary Fig. S1) is as follows. Each ectodomain monomer
contains a leucine-rich repeat (L1) domain, a cysteine-rich (CR)
region and a second leucine-rich repeat (L2) domain, followed by
three fibronectin type III domains (FnIII-1 to FnIII-3). FnIII-2
contains an insert domain (ID) of 120 residues, within which lies
the a–b cleavage site. The disulphide bond between each a- and
b-chain involves the cysteines C647 and C860 (ref. 8). In addition,
there are a–a disulphide bonds at C524 in the FnIII-1 domains and
between the triplet C682, C683 and C685 in the ID8.
The crystallographic asymmetric unit of our structure comprises
one ectodomain a–b chain pair (that is, the IRDb monomer), one
83-7 Fab and one 83-14 Fab. The IRDb dimer structure is shown in
Fig. 1 and the comparative sequences and secondary structure
assignments are shown in Supplementary Fig. S2. Each a–b mono-
mer has an inverted ‘V’ layout with respect to the cell membrane
(Fig. 1 and Supplementary Fig. S3). One leg of the Vis formed by the
L1, CR and L2 domains; the other is formed by an extended linear
the inverted V lies the connection between L2 and FnIII-1; these
domains arenot in extensivecontact in the monomer (516A˚2buried
molecular surface per interacting pair of L2 and FnIII-1 domains).
The arrangement of the L1, CR and L2 domains is similar, although
not identical, to that observed for the human IR10and IGF1-R11
L1-CR-L2 fragment structures.
The IRDb monomer is paired in the unit cell to form the
disulphide-bonded IRectodomain homodimer (Fig. 1), which arises
from a two-fold rotation of the monomer about an axis running
V. The L2 domain of the first monomer contacts the FnIII-1 domain
of the second monomer at the apex of the inverted V, whereas the L1
domain of the first monomer contacts the FnIII-2 domain of the
second at the approximate midpoint of one of the two legs of the
inverted V. The latter intermonomer interaction (878A˚2buried
molecular surface per interacting pair of L1 and FnIII-2 domains),
Figure 1 | The IRDb ectodomain homodimer, showing the juxtaposition of
domains between the monomers. One monomer is shown in tube
representation, the other in atomic sphere representation. Individual
domains are coloured as follows: L1, brown; CR, yellow; L2, green; FnIII-1,
magenta; FnIII-2, cyan; and FnIII-3, blue. The locations of potential
N-linked glycosylation sites are shown in black.
1CSIRO Molecular & Health Technologies, 343 Royal Parade, Parkville, Victoria 3052, Australia.
Vol 443|14 September 2006|doi:10.1038/nature05106
© 2006 Nature Publishing Group
although not extensive, is relatively well packed and contains two
potential salt bridges (E124–R780 and R804–E153 or R804–E154).
The other intermonomer interaction (L2 and FnIII-1, burying
1,344A˚2of molecular surface per interacting pair of domains) is
more extensive, but less well-packed, and contains two potential salt
bridges (R454–D496 and E394–R498). Both pairs of salt bridges are
potentially conserved in IGF-1R(Supplementary Fig. S2). Located at
the apex of the IRDb dimer, in close proximity to each other, is the
pair of K460 residues in the last leucine-rich repeat in L2 (Fig. 2).
These residues have been implicated both in regulating cooperative
interactions between the two monomers in the IR dimer and as the
sites for chemical crosslinking with disuccinimidyl suberate12.
V structure where, in the intact IR, they would extend through the
cell membrane (Supplementary Fig. S1). The N-linked glycosylation
sites in the L1-CR-L2 fragment are distributed over its outer surface
within the homodimer (Fig. 1), away from the surfaces involved in
monomer–monomer interactions or in ligand binding (see below).
By contrast, the glycosylated asparagines in the FnIII domains are
towards the inner surface of the FnIII domain leg (Fig. 1 and
Supplementary Fig. S3).
The structure provides experimental confirmation of three FnIII
domains in the IR (see Supplementary Figs S1 and S2). Although
previouspredictions oftheFnIIIboundaries andstrandlocations are
seen to be largely correct2, none positioned the C
or FnIII-2 correctly or placed the ID within the CC
and only one of the predictions in each case was correct for the E
strands of FnIII-1 (ref. 13) or FnIII-3 (ref. 14).
The inter-a-chain disulphide bond involving the pair of C524
0strands of FnIII-1
0loop of FnIII-2,
residues8lies in a large, 28-residue loop (508–535) linking the C and
C647–C860 can be observed: C647 is in the ID in a long extension of
on the inside of the inverted V, almost to the bottom of FnIII-3, with
the pair of S655 residues at the end of the visible electron density. Its
partner, C860, is located close to the C terminus of the C
FnIII-3. This additional connection between FnIII-2 and FnIII-3
would reduce the flexibility at the junction of these two domains.
Such rigidity may be important in signal transduction because
removal of this a–b disulphide bond by the C647S mutation
abolishes basal and insulin-stimulated phosphorylation without
affecting insulin binding15. An internal disulphide bridge links
C786 and C795 at the base of the FG loop in FnIII-2. The unpaired
C872 has its side chain partially buried in FnIII-3 and has no
counterpart in the IGF-1R (Supplementary Fig. S2) or the IR-related
Low-resolution simulated projection images generated from the
IRDb structure, both with and without the 83-7 and 83-14 Fabs
attached, can be compared with experimental negative stain electron
micrograph images obtained previously16(Fig. 3). The correspon-
dence of the projected structure with these images is marked,
suggesting that the quaternary structure of the ectodomain observed
microscopy studies16. The epitope for 83-7 Fab, which does not
inhibit insulin binding4, is located in the segment 233–281 of the CR
domain, whereas the monoclonal antibody (Mab)83-14, which does
inhibit insulin binding to IR on cells4, binds only to FnIII-1 (Fig. 2).
The four Fabs protrude as extended ‘handles’ from the IR ecto-
reconstruction from scanning transmission electron microscopy
dark-field images7. In that model, the L1 domains protrude at the
membrane distal end of the molecule and the FnIII-2 and FnIII-3
domains lie coplanar to the membrane7. By contrast, our crystal
structure shows that the L1 domains lie towards the centre of the
molecule at the start of one leg of the inverted V monomer structure
with the FnIII domains arranged in a linear manner forming the
other leg (Fig. 1 and Supplementary Fig. S3). The ligand-binding
faces of the two L1 domains are thus located on opposite sides of the
0strands of the FnIII-1 domain. The a–b disulphide involving
Figure 2 | The IRDb ectodomain homodimer, showing the attached pairs of
Fabs. The IRDb homodimer is depicted with one molecule in tube
representation and the other in atomic sphere representation. Domains are
coloured as in Fig. 1 and the view is down the two-fold axis from the
membrane distal end. The Fab fragments are represented as thin tubes and
coloured as follows: 83-7 heavychain, dark blue; 83-7 light chain, light-blue;
83-14 heavy chain, red; and 83-14 light chain, pink. Glycosylation sites are
shownin black. The inset showsthe location of the pairsof chargedresidues
Fabs 83-7 and 83-14 are given in Supplementary Figs S5 and S6.
Figure 3 | Comparison of the crystal structure of the Fab-complexed IRDb
ectodomain with single-particle electron microscope images. a, Simulated
projection images at ,20A˚resolution obtained from the crystal structure
presented here (see Supplementary Methods). b, Proposed projected
views of previously observed particles (reprinted, with permission, from
ref. 16). The correspondence between a and b is marked. (The z-projection
in b of IRDb complexed with 83-7 Fab does not correspond to its partner
view ina and isprobably avariantof thex-projection of the samecomplex.)
NATURE|Vol 443|14 September 2006
© 2006 Nature Publishing Group
dimer ,65A˚from each other, a distance that is too great to allow
insulin to contact both L1 domains simultaneously, as previously
suggested for the high-affinity state of the insulin–IR complex7.
These differences may be a consequence of structural transitions
on insulin binding, similar to those seen in the epidermal growth
factor receptor family17. Such extensive structural transitions17seem
unlikely, however, given that ligand binding causes the IR to become
more compact rather than more extended18,19and that electron
microscopy imagesofIR intheabsenceofinsulin arestill interpreted
to have the L1 domains protruding at the distal end of the receptor20.
In our maps, a segment of electron density lies across the highly
conservedhydrophobicpatch onthebinding faceofL1(Supplemen-
corresponding to the 20-residue peptide S519N20 included in the
crystallized receptor complex protein solution (see Supplementary
Methods), subsequent analysis of crystals grown in the presence of
alternative binding peptides5showed that this density remained
unchanged (data not shown), suggesting that it is not related to
the S519N20 peptide. It could correspond to the so-called CT
peptide21, a region at the end of the a-chain that is known to be in
close juxtaposition to the L1 domain10. Unfortunately, the relatively
poor quality of the density precludes definitive identification.
The details of the low-affinity insulin-binding site (which involves
the central b-sheet of the L1 domain, the CT peptide21–23and the
central modules of the CR region), are discussed in the Supplemen-
tary Information and in a report of the three-dimensional structure
of the L1-CR-L2 fragment of IR10. The latter report presents a model
for insulin binding to the L1 domain10, in which the main hydro-
phobic surface of insulin interacts with the hydrophobic patch on
the second b-sheet of L1. Such an arrangement has the residues
corresponding to the hexamer face of insulin, the so-called second
site24–26, fully exposed and free to interact with other domains in the
Our structure suggests that this second region of insulin binds to
one or more of the three loops (AB, CC
0and EF) in FnIII-1 of
the other monomer to generate the high-affinity signalling complex
(Fig. 4). Evidence supporting the requirement of these portions of
FnIII-1 for high-affinity binding comes from, first, an analysis of
fragment (which starts at residue 390) with a photoreactive PheB1
insulin derivative29; andthird,the differential binding properties ofa
series of truncated IRectodomains21,23. The only differences between
the IR593.CT fragment21and the IR mini-receptor23(IR462.CT),
FnIII-1 and the assembly of IR593.CT into a disulphide-linked
The formation of the high-affinity state is associated with a
rearrangement of IR domains18,19,24, which in turn leads to trans-
phosphorylation of the intracellular kinase and regulatory regions of
the receptor. The details of these domain movements remain to be
established because soluble IRectodomain does not undergo ligand-
the extensive interaction between the L1 and CR domains in each
monomer and the expected rigidity of the FnIII-2–FnIII-3 connec-
tions as a result of the a–b disulphide bonds, the points at which
these domain movements potentially occur in the monomer are the
CR-L2, L2–FnIII-1 and FnIII-1–FnIII-2 junctions. The interface
between the L1 domain of one monomer and the FnIII-2 domain
of the other suggests that any closure of the L1-CR unit of one
monomer towards the C-terminal end of the FnIII-1 domain of the
other would also drag the FnIII-2–FnIII-3 unit with it, thereby
changing the relative position of the intracellular kinase domains
in the IR dimer and potentially inducing signalling (see Supplemen-
tary Information). Such arguments remain speculative and high-
resolution structures of insulin–IR complexes will be required to
establish the details of ligand–receptor binding and the domain
rearrangements that accompany signalling.
Protein and peptide preparation and characterization. Expression and puri-
fication of IRDb protein, production of Fabs 83-7 and 83-14, cloning and
sequencing of Mab variable region complementary DNA, and synthesis of
S519N20 peptide are described in the Supplementary Methods. The sequences
of the Fabs areshown in Supplementary Figs S5 and S6. The IRDb protein binds
insulin with similar affinity (dissociation constant, 1nM) as intact ectodomain
(data not shown).
Crystallization and X-ray data collection. Crystals were produced by vapour
diffusion using 1ml of complex in 10mM HEPES (pH7.5), 0.02% sodium azide
and 10% D-trehalose solution and 1ml of well solution containing 0.24M
ammonium tartrate and 15% PEG 3350 at either 48C or 208C. A di-m-iodobis
(ethylenediamine) diplatinum (II) nitrate (PIP)-derivatized crystal was generated
by soaking the above crystals in well solution containing 2mM PIP. Details of
data collection and processing are given in the Supplementary Methods and
Table S1. The space group was identified as C2221with unit-cell dimensions
a ¼ 123.0A˚, b ¼ 319.7A˚, c ¼ 204.9A˚.
Phasing, model building and refinement. Experimental details, references and
final refinement statistics are given in the Supplementary Methods and Sup-
plementary Table S1. The locations of the L1, CR, L2 and Fab domains were
found by molecular replacement using PHASER with native data set 1. Search
models were the L1-CR and the L2 domains from the L1-CR-L2 fragment10and
the Hy-HEL5 antibody (PDB entry 1BQL) set at 58 hinge angle increments.
Phase improvement was obtained with RESOLVE and BUSTER-TNT, and the
resultant electron density map showed the likely location of the three FnIII like
domains. The molecular replacement solution was confirmed using native data
set 2 including, in the process, phase information derived from the PIP-
derivative data set (Supplementary Table S1), processed with SHARP and
subjected to solvent-flattening with DM and/or SOLOMON. Model building
of the FnIIIdomains using domains 7, 8 and 9 of the fibronectin structure (PDB
code 1FNF) as a guide, crystallographic refinement using iterative cycles of
BUSTER-TNTand/or REFMAC5, and manual model building using XtalView/
Xfit are described in the Supplementary Methods. Within BUSTER-TNT,
scattering from the missing atoms was modelled with a low-resolution homo-
graphic exponential distribution, and maximum entropy density completion
was used at the end of each round of refinement to recover the density for
missing parts of the structure.
Figure 4 | Insulin bridging two monomers in the IRDb homodimer. The L1,
CR and L2 domains (atomic sphere representation) are from one monomer
in the IR homodimer, whereas the FnIII domains (tube representation) are
from the other. Domains are represented and coloured as in Fig. 1. Insulin
(grey atomic sphere representation) is positioned on L1 according to the
model of insulin bound to the L1-CR-L2 fragment10with no further
NATURE|Vol 443|14 September 2006
© 2006 Nature Publishing Group
Received 19 April; accepted 21 July 2006. Download full-text
Published online 6 September 2006.
1. Kitamura, T., Kahn, C. R. & Accili, D. Insulin receptor knockout mice. Annu. Rev.
Physiol. 65, 313– -332 (2003).
Adams, T. E., Epa, V. C., Garrett, T. P. J. & Ward, C. W. Structure and function
of the type 1 insulin-like growth factor receptor. Cell. Mol. Life Sci. 57,
1050– -1093 (2000).
Denley, A., Wallace, J. C., Cosgrove, L. J. & Forbes, B. E. The insulin receptor
isoform exon 112(IR-A) in cancer and other diseases: a review. Horm. Metab.
Res. 35, 778– -785 (2003).
Soos, M. A. et al. Monoclonal antibodies reacting with multiple epitopes on the
human insulin receptor. Biochem. J. 235, 199– -208 (1986).
Schaffer, L. et al. Assembly of high-affinity insulin receptor agonists and
antagonists from peptide building blocks. Proc. Natl Acad. Sci. USA 100,
4435– -4439 (2003).
De Meyts, P. & Whittaker, J. Structural biology of insulin and IGF1 receptors:
implications for drug design. Nature Rev. Drug Discov. 1, 769– -783 (2002).
Ottensmeyer, F. P., Beniac, D. R., Luo, R. Z. T. & Yip, C. C. Mechanism of
transmembrane signaling: insulin binding and the insulin receptor. Biochemistry
39, 12103– -12112 (2000). Corrigendum. Ibid. 40, 6988.
Sparrow, L. G. et al. The disulfide bonds in the C-terminal domains of the
human insulin receptor ectodomain. J. Biol. Chem. 272, 29460– -29467 (1997).
Peng, K. et al. Optimizing long intrinsic disorder predictors with protein
evolutionary information. J. Bioinform. Comput. Biol. 3, 35– -60 (2005).
10. Lou, M. et al. Crystal structure of the first three domains of the human insulin
receptor reveals major differences from the IGF-1 receptor in the regions
governing ligand specificity. Proc. Natl Acad. Sci. USA 103, 12429– -12434
11. Garrett, T. P. J. et al. Crystal structure of the first three domains of the type-1
insulin-like growth factor receptor. Nature 394, 395– -399 (1998).
12. Kadowaki, H. et al. Mutagenesis of lysine 460 in the human insulin receptor:
effects upon receptor recycling and cooperative interactions among binding
sites. J. Biol. Chem. 265, 21285– -21296 (1990).
13. Marino-Buslje, C., Martin-Martinez, M., Mizuguchi, K., Siddle, K. & Blundell,
T. L. The insulin receptor: from protein sequence to structure. Biochem. Soc.
Trans. 27, 715– -726 (1999).
14. Schaefer, E. M., Erickson, H. P., Federwisch, M., Wollmer, A. & Ellis, L.
Structural organization of the human insulin receptor ectodomain. J. Biol. Chem.
267, 23393– -23402 (1992).
15. Cheatham, B. & Kahn, C. R. Cysteine 647 in the insulin receptor is required for
normal covalent interaction between a- and b-subunits and signal
transduction. J. Biol. Chem. 267, 7108– -7115 (1992).
16. Tulloch, P. A. et al. Single-molecule imaging of human insulin receptor
ectodomain and its Fab complexes. J. Struct. Biol. 125, 11– -18 (1999).
17. Burgess, A. W. et al. An open-and-shut case? Recent insights into the
activation of EGF/ErbB receptors. Mol. Cell 12, 541– -552 (2003).
18. Lee, J., Pilch, P. F., Shoelson, S. E. & Scarlata, S. F. Conformational changes of
the insulin receptor upon insulin binding and activation as monitored by
fluorescence spectroscopy. Biochemistry 36, 2701– -2708 (1997).
19. Florke, R. R. et al. Hormone-triggered conformational changes within the
insulin-receptor ectodomain: requirement for transmembrane anchors.
Biochem. J. 360, 189– -198 (2001).
20. Yip, C. C. & Ottensmeyer, P. Three-dimensional structural interactions of
insulin and its receptor. J. Biochem. 278, 27329– -27332 (2003).
21. Surinya, K. H. et al. Role of insulin receptor dimerization domains in ligand
binding, cooperativity, and modulation by anti-receptor antibodies. J. Biol.
Chem. 277, 16718– -16725 (2002).
22. Kurose, T. et al. Cross-linking of a B25 azidophenylalanine insulin derivative to
the carboxy-terminal region of the a-subunit of the insulin receptor.
Identification of a new insulin-binding domain in the insulin receptor. J. Biol.
Chem. 269, 29190– -29197 (1994).
23. Kristensen, C., Wiberg, F. C. & Andersen, A. S. Specificity of insulin and insulin-
like growth factor I receptors investigated using chimeric mini-receptors—role
of C-terminal of receptor a subunit. J. Biol. Chem. 274, 37351– -37356 (1999).
24. Schaffer, L. A model for insulin binding to the insulin receptor. Eur. J. Biochem.
221, 1127– -1132 (1994).
25. De Meyts, P. The structural basis of insulin and IGF-1 receptor binding and
negative co-operativity, and its relevance to mitogenic versus metabolic
signaling. Diabetologia 37 (suppl. 2), S135– -S148 (1994).
26. De Meyts, P. Insulin and its receptor: structure, function and evolution.
BioEssays 26, 1351– -1362 (2004).
27. Zhang, B. & Roth, R. A. A region of the insulin receptor important for ligand
binding (residues 450– -601) is recognized by patient’s autoimmune antibodies
and inhibitory monoclonal antibodies. Proc. Natl Acad. Sci. USA 88, 9858– -9862
28. Schumacher, R. et al. Signaling-competent receptor chimeras allow mapping of
major insulin receptor binding domain determinants. J. Biol. Chem. 268,
1087– -1094 (1993).
29. Fabry, M. et al. Detection of a new hormone contact site within the insulin
receptor ectodomain by the use of a novel photoreactive insulin. J. Biol. Chem.
267, 8950– -8956 (1992).
30. Hoyne, P. A. et al. High affinity insulin binding by soluble insulin receptor
extracellular domain fused to a leucine zipper. FEBS Lett. 479, 15– -18 (2000).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank K. Siddle for the 83-7 and 83-14 monoclonal
antibody cell lines; O. Dolezal for advice on cloning and sequencing the
monoclonal antibody variable region cDNAs; B. Van Donkelaar for technical
crystallography inputs; and L. Lu, L. Cheong and T. Phan for contributions to
protein production. We acknowledge the help provided by beamline staff at
both the Advanced Photon Source and the Photon Factory. This work was
supported by the Australian Synchrotron Research Program, which is funded by
the Commonwealth of Australia under the Major National Research Facilities
Program. Additional financial support was provided under the Generic
Technology component of the Industry Research and Development Act 1986,
from Biota Diabetes Research Pty Ltd.
Author Information Atomic coordinates of the IR-Fab complex have been
deposited in the Protein Data Bank with accession number 2DTG. Reprints and
permissions information is available at www.nature.com/reprints. The authors
declare no competing financial interests. Correspondence and requests for
materials should be addressed to C.W.W. (Colin.Ward@csiro.au) or M.C.L.
NATURE|Vol 443|14 September 2006
© 2006 Nature Publishing Group