Nature Structural & Molecular Biology homepage
Advanced search
 Journal home > Archive > Table of Contents > Letter > Full text
Journal home
Advance online publication
Current issue
Archive
Press releases
Supplements
Focus
Guide to authors
Online submissionOnline submission
Permissions
For referees
Free online issue
Contact the journal
Subscribe
Advertising
work@npg
naturereprints
About this site
For librarians
 
NPG Resources
Nature
Nature Cell Biology
Nature Reviews Molecular Cell Biology
The EMBO Journal
Nature Reports Stem Cells
Nature Reports Avian Flu
NPG Subject areas
Biotechnology
Cancer
Chemistry
Clinical Medicine
Dentistry
Development
Drug Discovery
Earth Sciences
Evolution & Ecology
Genetics
Immunology
Materials Science
Medical Research
Microbiology
Molecular Cell Biology
Neuroscience
Pharmacology
Physics
Browse all publications
Letter
Nature Structural Biology  9, 282 - 287 (2002)
Published online: 18 March 2002; | doi:10.1038/nsb779

Cysteine-rich module structure reveals a fulcrum for integrin rearrangement upon activation

Natalia Beglova1, 2, Stephen C. Blacklow1, 2, Junichi Takagi1, 3 & Timothy A. Springer1, 3

1 Department of Pathology, Harvard Medical School, 75 Francis St., Boston, Masschusetts 02115, USA.

2 Brigham and Women's Hospital, Harvard Medical School, 75 Francis St., Boston, Masschusetts 02115, USA. These authors contributed equally to the work.

3 The Center for Blood Research, 200 Longwood Avenue, Boston, Massachusetts 02115, USA.

Correspondence should be addressed to Timothy A. Springer SpringerOffice@cbr.med.harvard.edu or Stephen C. Blacklow sblacklow@rics.bwh.harvard.edu
Cysteine-rich repeats in the integrin beta subunit stalk region relay activation signals to the ligand-binding headpiece. The NMR solution structure and disulfide bond connectivity of Cys-rich module-3 of the integrin beta2 subunit reveal a nosecone-shaped variant of the EGF fold, termed an integrin-EGF (I-EGF) domain. Interdomain contacts between I-EGF domains 2 and 3 observed by NMR support a model in which the modules are related by an approximate two-fold screw axis in an extended arrangement. Our findings complement a 3.1 Å crystal structure of the extracellular portion of integrin alphaVbeta3, which lacks an atomic model for I-EGF2 and a portion of I-EGF3. The disulfide connectivity of I-EGF3 chemically assigned here differs from the pairings suggested in the alphaVbeta3 structure. Epitopes that become exposed upon integrin activation and residues that restrain activation are defined in beta2 I-EGF domains 2 and 3. Superposition on the alphaVbeta3 structure reveals that they are buried. This observation suggests that the highly bent alphaVbeta3 structure represents the inactive conformation and that release of contacts with I-EGF modules 2 and 3 triggers a switchblade-like opening motion extending the integrin into its active conformation.

Signaling cascades that impinge on integrin cytoplasmic domains, as well as truncation and mutation of these domains, activate ligand binding in the integrin headpiece1. Recent evidence shows that movement apart or unclasping of the juxtamembrane segments of the integrin alpha and beta subunits is sufficient for activation2. Close association between membrane proximal segments restrains activation3. The Cys-rich modules of the C-terminal portion of the integrin beta subunit extracellular domain are particularly important in signaling; most monoclonal antibodies (mAbs) to integrin activation epitopes map to this region, including mAbs that either directly activate ligand binding or selectively bind either when integrins are activated or have bound ligand1. In the integrin beta2 subunit, activation epitopes have been mapped to specific mouse/human amino acid substitutions in Cys-rich repeats 2 and 3 (Fig. 1b)4. Xenogeneic amino acid substitutions that activate ligand binding by the integrin alphaXbeta2 also map to these two Cys-rich repeats, suggesting that contacts between the second and third beta subunit Cys-rich repeats and the beta subunit stalk restrain integrins in the inactive conformation5 (Fig. 1b). Consistent with opening of interfaces upon activation that are concealed in the resting state, activation epitopes are fully exposed on isolated beta subunits4, 6.

Figure 1. Integrin domain organization and the Cys-rich repeats.
Figure 1 thumbnail

a, Diagram of integrin architecture based on the structure of alphaVbeta3 (ref. 8). Domains are beta-propeller (beta-P), plexin-semaphorin-integrin (PSI), hybrid (H), EGF (E) and beta-terminal domain (betaTD). A probable position for the I domain is shown. Domains not resolved or only partially resolved in alphaVbeta3 are shown with dashed lines. b, Structural alignment of I-EGF3 to EGF module 1 of coagulation factor IX12, I-EGF4 of beta3 and alignment by sequence to other I-EGF modules. beta-strands defined here in beta2, beta3 (ref. 8) and factor IX12 are underlined in red; loops at the C-terminal end of the I-EGF3 module are underlined in black. Residues not defined in the beta3 I-EGF3 (beta3_3) structure are italicized, as are residues that were defined in the beta2 I-EGF3 (beta2_3) structure but not formally included within the module boundary defined here. Residues in activation epitopes are marked with red symbols above them: filled circles are residues common to the MEM48 and CBR LFA-1/2 epitopes; plus signs are additional residues in the MEM48 epitope and diamonds are residues in the KIM127 epitope4. Residues that restrain activation of alphaXbeta2 are marked with red squares5. Asterisks below the sequences indicate the best conserved portions of the structures. Cys residues that are disulfide bonded to one another in beta2 I-EGF3, beta3 I-EGF4 (beta3_4) and factor IX are highlighted in the same color; a different disulfide bond arrangement was reported for beta3 I-EGF3, as discussed in the text.



Full FigureFull Figure and legend (54K)
In the integrin headpiece, the structure of the inserted (I) domain is known and has been shown to undergo significant conformational movement7. The recent structure of the alphaVbeta3 integrin extracellular region8 specifies the relative positions of 8 of the 12 integrin domains in a compact global architecture. However, the PSI (plexin-semaphorin-integrin) domain, Cys-rich modules 1 and 2, and the first 10 residues of Cys-rich module 3 could not be resolved in the reported alphaVbeta3 structure (Fig. 1a). The second and third Cys-rich modules of the beta subunit are wedged in the bend between the integrin stalk and headpiece domains (Fig. 1a), poised to communicate structural rearrangements between the cytoplasmic domains and the ligand-binding headpiece. To understand why exposure of antibody epitopes on these modules accompanies integrin activation, we have determined the structure of Cys-rich repeat 3 (module 3) of the integrin beta2 subunit, identified sites of contact between modules 2 and 3, and modeled the mode of interaction of activation epitopes with other integrin domains.

NMR structure and disulfide bond determination
Recombinant Cys-rich repeats 2, 3 and 2+3 (the repeat two and three pair) from the integrin beta2 subunit were constructed using newly defined module boundaries9, 10. The calculated NMR structure of repeat 3 (Table 1) reveals a variant of the EGF fold, termed the integrin- or I-EGF fold (Figs 1b, 2a−c). Disulfide bonds were independently determined by protease digestion and resolution of disulfide-bonded peptides by mass spectrometry (Table 1). In I-EGF3, the structure is stabilized by four disulfide bonds between the first and fifth, second and fourth, third and sixth, and seventh and eighth Cys residues (Fig. 1b). The structure-based sequence alignment (Fig. 1b) shows that three disulfide bonds are shared with classical EGF domains. The disulfide unique to I-EGF domains links the N-terminus to the turn between the two beta-strands (Figs 1b, 2c).

Figure 2. Solution structure of I-EGF3.
Figure 2 thumbnail

a, Stereo view of the 15 lowest energy structures of I-EGF3. The best-fit superposition of the polypeptide backbone is shown. Side chains of the Cys residues engaged in the four disulfide bonds are illustrated in orange. b, Stereo view of the minimized mean structure of I-EGF3. The side chains are in ball-and-stick representation on a ribbon trace of the module backbone. c, Best-fit superposition of I-EGF3 with the crystal structures of EGF module 1 of coagulation factor IX12 and I-EGF4 of alphaVbeta3 (ref. 8). The Calpha trace and Cys side chain atoms are green for I-EGF3, red for the factor IX EGF module and blue for I-EGF4 of alphaVbeta3. Cys side chain bonds are yellow. The Ca2+ coordinated to the factor IX EGF module is depicted as a gold sphere. Figure generated using Insight 2000 (MSI) in (a, b) and Ribbons39 in (c).



Full FigureFull Figure and legend (87K)
Table 1. Summary of NMR constraints, structural statistics and disulfide assignments
Table 1 thumbnail

Full TableFull Table
The disulfide bond connectivity of I-EGF3 assigned here using chemical methods and further verified by NMR distance restraint data corresponds to the arrangement predicted by structure-based alignment and tested by expression of truncated beta2 subunits in mammalian cells9, 10. However, this connectivity differs from the assignments reported in the recent 3.1 Å crystal structure of alphaVbeta3 (ref. 8). The first 10 residues of I-EGF3 could not be resolved in the alphaVbeta3 structure because of poor electron density; thus, Cys 1 and 2 are missing from the beta3 I-EGF3 module. The remaining Cys residues were reported to be linked in a 3−5, 4−6 and 7−8 arrangement, in contrast to the 1−5, 2−4, 3−6 and 7−8 arrangement found here. The most plausible explanation for the discrepancy is that authors of the alphaVbeta3 paper incorrectly traced the backbone in the N-terminal portion of I-EGF3 and could not properly assign the disulfides with Cys 4 and 5 because their disulfide-bonded partners were unresolved in the electron density. This misinterpretation is explicable given the 3.1 Å resolution of their native X-ray data and the lack of good density in the unresolved region of I-EGF3 and in the contiguous I-EGF modules 1 and 2. Sequence alignment predicts that in all integrin beta subunits, the disulfide bond connectivity for I-EGF modules 2 and 4 will match that of I-EGF3 (Fig. 1b)9, 10. Indeed, the disulfide connectivity of I-EGF4 reported in the alphaVbeta3 crystal structure8 is identical to the pattern reported here for I-EGF3 of the beta2 subunit. Furthermore, despite three insertions and deletions, our beta2 I-EGF3 structure superimposes well with the beta3 I-EGF4 structure, with a root mean square (r.m.s.) deviation of 2.6 Å for all 37 alignable positions (Figs 1b, 2c).

Structure comparison
Compared to classical EGF modules with three disulfide bonds11, the I-EGF3 module is less elongated, with a nosecone-like shape (Fig. 2c). The antiparallel sheet between the two beta-strands is shortened because four highly conserved residues among classical EGF-like modules are deleted in I-EGF modules (Fig. 1b). The loop at the C-terminal end of the module between Cys 2 and 3 is unusually long compared to EGF-like domains and contributes to the flare at the C-terminus that helps dock the nose of the next module. The nose at the N-terminal end differs markedly in conformation from classical EGF modules (Figs 1b, 2c). The N-terminal tip is bent because of the additional disulfide bond in the hairpin turn between the shortened beta-strands. This extra disulfide bond is structurally analogous to calcium coordination in the calcium-binding subset of EGF-like modules11, 12, 13, in which calcium coordination also bridges the turn connecting the two beta-strands to the N-terminal end of the module (Fig. 2c).

Relationship between modules 2 and 3
NMR chemical shift perturbation studies, in which heteronuclear fingerprint spectra of individual modules 2 and 3 are compared with the spectrum of the module 2+3 pair, demonstrate that intimate contacts are present between modules 2 and 3 in the module pair (Fig. 3a,b). Covalent connection of the modules leads to perturbations in chemical shift and linewidth for residues in the C-terminal half of module 2, at the N-terminal end of module 3 and in the turn between beta-strands 1 and 2, which is connected to the N-terminal end of the module by the disulfide connecting the first and fifth Cys residues. Residues with the largest amount of disorder in the structure of module 3 (Fig. 2a) comprise most of those exhibiting perturbation upon covalent connection to module 2.

Figure 3. Interdomain interactions between I-EGF2 and I-EGF3.
Figure 3 thumbnail

a, Comparison of assigned 15N-1H heteronuclear single quantum coherence spectra of I-EGF2 (green), I-EGF3 (blue) and the I-EGF2+3 domain pair (red) in an overlay plot. b, Sum of the normalized HN- and 15N-chemical shift differences between resonances of I-EGF2+3 and the corresponding resonances of I-EGF2 and I-EGF3. The maximum chemical shift perturbation in the proton dimension is 0.63 p.p.m., and the maximum chemical shift perturbation in the nitrogen dimension is 5.07 p.p.m. The amino acid sequence of the I-EGF2+3 domain pair is at the bottom, color-coded by the extent of chemical shift perturbation at each site. Resonances of residues with a white background are not significantly perturbed in the spectrum of the domain pair. Residues showing perturbations >2 standard deviations (s.d.) from the mean of the unperturbed residues are on a green background. A yellow background indicates a perturbation of >4 s.d. Resonances of residues displayed on an orange background undergo line broadening and are not detected in the spectrum of the module pair but are observed in the spectra of individual modules (asterisked in the plot). c, Model for the structure of the I-EGF2+3 domain pair. A surface representation of the structure has been superimposed on a ribbon trace of the module backbone with side chains represented as sticks. The surface has been colored according to the perturbation scheme in (b). Intermodule NOEs used to estimate the approximate interdomain orientation are illustrated by dotted lines. Calpha atoms of residues in the CBR LFA-1/2 (Leu 534, Phe 536 and His 543) and MEM48 (also includes Arg 541 and Phe 546) epitopes are light blue spheres; Calpha atoms in the KIM127 (residues Gly 504, Leu 506 and Tyr 508) epitope are purple spheres; and Calpha atoms of residues that restrain activation of human alphaXbeta2 (Gln 525 and Val 526) are black spheres5. Figure was prepared with GRASP40.



Full FigureFull Figure and legend (156K)
To build a model of the module 2+3 pair, we used three unambiguous NOEs between modules to define their relative orientation. The resulting model predicts that modules 2 and 3 are related by an approximate two-fold screw axis and are positioned in an extended arrangement with respect to each other (Fig. 3c). This interdomain orientation is similar to the orientation between I-EGF modules 3 and 4 in alphaVbeta3 (ref. 8) and that between tandem calcium-binding EGF-like modules in which coordination of calcium sets the interdomain orientation13.

Structure and exposure of integrin activation epitopes
The structure of I-EGF3 and the model of the module 2+3 pair reveal for the first time the position of residues in integrin activation epitopes. In I-EGF3, residues Leu 534, Phe 536 and His 543 are shared in the epitopes for the activating antibodies CBR LFA-1/2 and MEM48 (Figs 1b, 3c). The epitope for MEM48 extends farther to include Arg 541 and Phe 546 (Figs 1b, 3c)4. Gln 525 and Val 526, located on the other side of the domain, participate in constraining alphaXbeta2 in its resting conformation when paired with an alphaX subunit from the same species, suggesting that they lie at an important species-specific alphabeta contact site in the latent heterodimer5. In the epitope in I-EGF2 for the activating antibody KIM127, residues Gly 504, Leu 506 and Tyr 508 are located in the long loop protruding from the C-terminal end of the module (Figs 1b, 3c).

Several studies suggest that the mechanism of integrin activation is general. For example, activation epitopes also become exposed on I-EGF modules of beta1- and beta3-integrins14, 15, 16, 17, 18, although they have not been mapped to specific residues as for beta2-integrins. To understand the molecular rearrangements that occur in an intact integrin during activation, we superimposed the model of the beta2 I-EGF2+3 domain pair on the partial I-EGF3 module of alphaVbeta3 (see Methods). The alphaVbeta3 extracellular domain unexpectedly exhibits an overall V-shape in which the ligand-binding headpiece bends back toward the base of the stalk region and is oriented toward the cell membrane8 (Figs 1a, 4a). In the superposition, I-EGF modules 2 and 3 are wedged into the extreme bend or 'genu' between the integrin headpiece and stalk (Fig. 4a). I-EGF3 sits between the beta subunit hybrid domain and the alpha subunit calf-1 domain, whereas the C-terminal portion of I-EGF2 nestles between the alpha subunit thigh and calf-1 domains, close to the fulcrum at the genu (Fig. 4a). In this bent integrin conformation, residues Gln 525 and Val 526 of I-EGF3, which are in an alphabeta interface that restrains integrins in the inactive state5, are buried and face calf-1 domain of the alpha subunit (Fig. 4a). Three other segments of I-EGF2 and I-EGF3 that restrain activation5 are also masked. Furthermore, activation epitopes are buried in the bent integrin conformation. The alpha subunit thigh domain and, in particular, the beta subunit hybrid domain mask the face of I-EGF3 presenting the CBR LFA-1/2 and MEM48 activating epitopes (Fig. 4a). Moreover, the KIM127 epitope of I-EGF2 is deeply buried in the crevice formed by the bend between the alpha subunit thigh and calf-1 domains. A surface representation of a manually docked antibody Fab fragment illustrates that the epitopes are inaccessible to its antigen-binding site (Fig. 4a). To objectively measure antigenic burial, we calculated the antibody-accessible surface using the approach to calculate solvent-accessible surface, except that we rolled over the surface a 20 Å diameter sphere to generously estimate the number of residues that might be accessible to a 40 Å diameter Fab fragment (Fig. 4a). All of the three identified residues in the KIM127 epitope are accessible to a 20 Å diameter sphere in an extended, active conformation (Fig. 4b), whereas none are accessible in the bent alphaVbeta3 conformation. Furthermore, three of five residues in the CBR LFA-1/2 and MEM48 epitopes in I-EGF3 are accessible in the extended conformation, whereas one is accessible in the bent conformation and only 14% as much as in the extended conformation.

Figure 4. Integrin activation.
Figure 4 thumbnail

a, Superposition of I-EGF2+3 from beta2 upon the C-terminal end of alphaVbeta3 (see Methods). A backbone ribbon trace of beta2 I-EGF2+3 and alphaVbeta3 is shown. Polypeptide backbones are red for beta2 I-EGF2+3, green for alphaV and yellow for beta3. Calpha atoms of residues in the KIM127 epitope (Gly 504, Leu 506 and Tyr 508) are purple, and cyan in the CBR LFA-1/2 (Leu 534, Phe 536 and His 543) and MEM48 (also includes Arg 541 and Phe 546) epitopes4. Calpha atoms of residues that restrain activation of human alphaXbeta2 (Gln 525 and Val 526) are black5. The surface of a representative Fab antibody fragment (gray; PDB code 1A3R), with its antigen-binding site oriented toward and at the closest approach to the KIM127 and CBR LFA1/2 epitopes, is also shown to illustrate that the KIM127 and CBR LFA-1/2 epitopes are inaccessible to its antigen-binding site (see text). A gray sphere 20 Å in diameter is included at the bottom left as a size reference. Figure prepared with GRASP40. b, Model for integrin activation. The model of the resting integrin resembles the form of alphaVbeta3 observed in the crystal structure8. The active integrin is depicted in an extended conformation as seen in electron micrographs of alphaIIbbeta3 bound to its ligand fibrinogen18. The triangular interaction site in I-EGF3 dovetails with the alpha subunit in the resting state and not in the active state.



Full FigureFull Figure and legend (106K)
alphaVbeta3/beta2 models the inactive conformation
Given the strong conservation of structure−function relationships among integrins, the masking of activation epitopes and the positioning in alphabeta contacts for residues that restrain integrins in the inactive state, our composite alphaVbeta3/beta2 model forces us to suggest that the genuflected conformation represents a latent, inactive structure. The genuflected conformation has been suggested to be unlikely to occur on the cell surface8. However, extensive contacts between the headpiece and stalk regions stabilize the bent conformation in which 1,400 Å2 of solvent-accessible surface area are buried19, 20. This value is in the range of physiological interaction surfaces and out of the range of crystal packing artifacts (P < 0.02)21. The observation that antigenic and interaction sites that are inaccessible or buried in an inactive integrin are also buried in the composite model suggests that the bent conformation is physiologically relevant and represents the inactive state. Experience from previous crystal structures of integrin I domains, which adopt the inactive state when there is no ligand7, also suggests that alphaVbeta3 represents the latent conformation because it was crystallized in the absence of a physiological ligand. Although the headpiece in the alphaVbeta3 structure was suggested to be in an active conformation8, the lines of evidence in that paper actually support the opposite conclusion. First, alphaVbeta3 was crystallized with Ca2+ and without exposure to Mg2+ or Mn2+ (ref. 8), yet Ca2+ is known to stabilize integrins in the inactive conformation, whereas Mg2+ and especially Mn2+ are activators. Although a similar alphaVbeta3 preparation has been shown to be active in the presence of Ca2+ in a solid phase binding assay when pretreated during coating on plastic with 10 muM Mn2+ (ref. 22), alphaVbeta3 does not bind ligand in the presence of Ca2+ in the absence of such pretreatment23. Therefore, the alphaVbeta3 that was crystallized and the alphaVbeta3 used in ligand binding assays seem to represent different conformational states. Integrins that are released from the cell surface usually are active in binding ligand, suggesting release of a restraint that keeps them inactive on the cell surface. However, studies of both cell surface and soluble integrins suggests that the latent conformation is maintained by complementary interactions holding the integrin alpha and beta subunit juxtamembrane segments in close association2, 3, 24. The C-termini of the alphaV and beta3 extracellular domains are very close together in the alphaVbeta3 structure8, mimicking the close association in the juxtamembrane region that maintains integrins in the inactive state. Finally, the I-like domain of alphaVbeta3 was suggested to resemble the open, ligand-binding conformation of integrin I domains rather than the closed, inactive conformation8. However, automatic superposition25 with either the closed or open conformations of the integrin alpha2 and alphaM I domains shows a significantly better fit of the I-like domain to the closed conformation than the open conformation, with 11 more residues within 3 Å of the common framework. These residues include five in the beta6−alpha7 loop, which dramatically rearranges between the two conformations.

A general model for integrin activation
The flexibility at the genu "may be linked to integrin regulation"8. To accommodate the model of the I-EGF2+3 structure into a mechanism for integrin activation, we propose that activation triggers a switchblade-like opening of the interface between the headpiece and stalk, which extends the ligand-binding headpiece of the integrin heterodimer away from the plasma membrane (Fig. 4b). This extended form, which is commonly seen in electron micrographs of integrins, including integrins simultaneously bound to ligand and activating antibody18, exposes not only the epitopes studied here, but also a much broader surface. This correlates with (i) selective exposure of epitopes on active and ligand-bound integrins, but not on inactive integrins, and (ii) the predominant localization of activation epitopes to the beta subunit, because 70% of the surface buried in the headpiece−stalk interface is on the beta subunit19, 20. Conformational changes of this magnitude are well precedented; influenza hemagglutinin, which undergoes a conformational change to induce fusion of viral and cellular membranes, rearranges to reposition its hydrophobic fusion peptide by 100 Å toward the target cell membrane26, 27.

We propose that in resting integrins, some hinging motion occurs about the fulcrum at the genu, and the accessibility of antibodies increases as the distance of the epitope from the fulcrum becomes greater. This would explain why antibodies that bind in I-EGF2, such as KIM127, bind only to pre-activated integrins, whereas antibodies that bind further away from the fulcrum, such as CBR LFA-1/2 in I-EGF3 and KIM185 in I-EGF4 (refs 4,28), can bind to integrins in the inactive state and activate them by acting like a wedge in the headpiece−stalk interface. We envision that in integrin activation by biological inside-out signaling, movement apart of the juxtamembrane domains2, 3 results in separation of the alpha and beta subunit stalks and destabilizes their interface with the headpiece, thereby triggering the switchblade-like opening and linked conformational changes within ligand-binding domains. Our data suggest that I-EGF modules 2 and 3, which reveal antibody-reactive epitopes during integrin activation, play an important role in these rearrangements.

Methods
Sample preparation.
Integrin beta2 subunit repeat two (I-EGF2, residues 460−513), repeat three (I-EGF3, residues 513−552) and the repeat 2+3 pair (I-EGF2+3, residues 460−552), all containing an extra C-terminal Ala residue, were cloned into the T7-based29 pMM expression vector30. The I-EGF2, I-EGF3 and I-EGF2+3 polypeptides were expressed in Escherichia coli and purified in reduced form from inclusion bodies as described30. For NMR spectroscopy, 15N-labeled protein was produced by growth in M9 minimal media supplemented with 15NH4Cl as the sole nitrogen source. Refolding of each polypeptide was performed by dialysis for 48 h at 4 °C in a redox buffer containing reduced (2.5 mM) and oxidized (0.5 mM) glutathione (10 mM Tris, pH 8.2) to permit disulfide bond rearrangement, essentially as described31. The refolded modules were then purified by reversed-phase HPLC, and their identities were verified by MALDI-TOF mass spectrometry. The presence of a unique disulfide-bonded species among many possibilities, as judged by HPLC (not shown), provided convincing evidence that each individual repeat and the repeat pair are properly folded.

NMR spectroscopy.
NMR spectra were recorded at 298 K, unless otherwise specified, on a Bruker 600 MHz spectrometer using standard procedures32, and processed with GIFA v4.3 (ref. 33). I-EGF2, I-EGF3 and I-EGF2+3 were at 1 mM in 5 mM NaPO4, pH 6.2. Sequential resonance assignments were obtained primarily from a combination of 3D-NOESY and 3D-TOCSY experiments: 15N-HSQC-NOESY (taumix = 150 ms), 15N-HSQC-TOCSY (taumix = 60 ms), 15N-NOESY-HSQC (taumix = 150 ms) and 15N-TOCSY-HSQC (taumix = 60ms). These data were supplemented with information from homonuclear 1H−1H 2D experiments, including NOESY (taumix = 100 ms and taumix = 150 ms), TOCSY (taumix = 60 ms) and DQF-COSY performed on unlabeled I-EGF3 samples in H2O and D2O. Coupling constants for I-EGF3 were derived from an 15N-HMQC-J spectrum. Temperature coefficients for the backbone amide protons of I-EGF3 were determined from 15N-HSQC spectra taken at 284 K, 293 K, 298 K, 303 K and 313 K. To identify slowly exchanging amide protons, an 15N-labeled I-EGF3 sample was lyophilized and redissolved in D2O. The decay of amide proton signal intensity was monitored for 6 h in a series of 15N-HSQC experiments at 298 K. The first spectrum was taken 5 min after D2O was added to the sample. Amide proton exchange rates for random coil peptides were calculated with SPHERE34. In the chemical shift perturbation experiment (Fig. 3), each single module 15N-HSQC spectrum was compared with the 15N-HSQC spectrum of I-EGF2+3.

Structure determination.
NOE-based distance constraints derived from the spectra listed above were calibrated to intraresidue Hbeta−Hbeta crosspeaks and divided into three categories: strong (<3.0 Å), medium (<4.0 Å) and weak (<6.0 Å), with appropriate pseudo atom corrections introduced as necessary. Phi-torsion angle restraints of -120° plusminus 35° and -60° plusminus 35° were used for residues with measured 3JHNHalpha coupling constants of >8 Hz and <4 Hz, respectively. Three distance restraints were incorporated to describe each disulfide bond. The distance between the two sulfur atoms was set to 2.02 plusminus 0.05 Å, and the Sgamma−Cbeta distances were constrained to 2.99 plusminus 0.05 Å. Structures were calculated by the standard hybrid distance geometry simulated annealing protocol as implemented in CNS35. Inclusion of hydrogen bond constraints for three hydrogen bonds between Arg 541 O and Leu 534 N, Lys 539 O and Phe 536 N, and Arg 532 O and His 543 N at the final stage of structure calculations is based on the following observations. (i) The vast majority (74%) of initial calculated structures with appropriate covalent geometry satisfied accepted hydrogen bond criteria (O...N distance 3.6 Å and O...H-N angle 120°). (ii) Amide protons of residues Leu 534, Phe 536 and His 543 show protection factors of 620, 700 and 1,630, respectively, on the basis of hydrogen exchange measurements. (iii) The temperature coefficient for these residues, determined by measuring the change in amide proton chemical shift with temperature, is <5 (−p.p.b. K-1). (iv) The Halpha chemical shift index predicts two beta-strands spanning residues 533−536 and 539−543, encompassing all three identified hydrogen bonds. Hydrogen bond distance restraints, introduced during the last stage of calculations, were rHN...O = 1.5−2.3 Å and rN...O = 2.4−3.3 Å (ref. 36).

Modeling and superposition.
The minimized average structure of I-EGF3 was used as a template to model I-EGF2 in the I-EGF2+3 pair using MODELLER-4 (ref. 37). The interface between the two repeats in the pair was modeled by including distance restraints corresponding to NOEs observed between the Leu 506 side chain and the backbone and side chain of Asn 518, and between the Ile 507 side chain and the backbone of Cys 514. Best-fit superposition of the factor IX EGF module12 and beta3 integrin I-EGF4 (ref. 8) upon beta2 I-EGF3 was performed using MALIGN-3D of MODELLER-4. To create the superposition of beta2 I-EGF2+3 on beta3, we used the Calpha coordinates of beta2-residues Lys 539−Glu 547 and Ala 550−Gln 552 at the C-terminal end of I-EGF3 where the two structures are in good agreement, with an r.m.s. deviation of 1.3 Å.

Coordinates.
Coordinates have been deposited in the Protein Data Bank (accession code 1L3Y).

 Top
Received 8 February 2002; Accepted 14 February 2002; Published online: 18 March 2002.

REFERENCES
  1. Humphries, M.J. Biochem. Soc. Trans. 28, 311−339 (2000). | PubMed | ISI | ChemPort |
  2. Takagi, J., Erickson, H.P. & Springer, T.A. Nature Struct. Biol. 8, 412−416 (2001). | Article | PubMed | ISI | ChemPort |
  3. Lu, C., Takagi, J. & Springer, T.A. J. Biol. Chem. 276, 14642−14648 (2001). | Article | PubMed | ISI | ChemPort |
  4. Lu, C., Ferzly, M., Takagi, J. & Springer, T.A. J. Immunol. 166, 5629−5637 (2001). | PubMed | ISI | ChemPort |
  5. Zang, Q. & Springer, T.A. J. Biol. Chem. 276, 6922−6929 (2001). | Article | PubMed | ISI | ChemPort |
  6. Huang, C., Lu, C. & Springer, T.A. Proc. Natl. Acad. Sci. USA 94, 3156−3161 (1997). | Article | PubMed | ChemPort |
  7. Emsley, J., Knight, C.G., Farndale, R.W., Barnes, M.J. & Liddington, R.C. Cell 101, 47−56 (2000). | Article | PubMed | ISI | ChemPort |
  8. Xiong, J.-P. et al. Science 294, 339−345 (2001). | Article | PubMed | ISI | ChemPort |
  9. Takagi, J., Beglova, N., Yalamanchili, P., Blacklow, S.C. & Springer, T.A. Proc. Natl. Acad. Sci. USA 98, 11175−11180 (2001). | Article | PubMed | ChemPort |
  10. Tan, S.-M. et al. FEBS Lett. 505, 27−30 (2001). | Article | PubMed | ISI | ChemPort |
  11. Bork, P., Downing, A.K., Kieffer, B. & Campbell, I.D. Q. Rev. Biophys. 29, 119−167 (1996). | PubMed | ISI | ChemPort |
  12. Rao, Z. et al. Cell 82, 131−141 (1995). | Article | PubMed | ISI | ChemPort |
  13. Downing, A.K. et al. Cell 85, 597−605 (1996). | Article | PubMed | ISI | ChemPort |
  14. Shih, D.T., Edelman, J.M., Horwitz, A.F., Grunwald, G.B. & Buck, C.A. J. Cell Biol. 122, 1361−1371 (1993). | Article | PubMed | ISI | ChemPort |
  15. Faull, R.J. et al. J. Biol. Chem. 271, 25099−25106 (1996). | Article | PubMed | ISI | ChemPort |
  16. Bazzoni, G., Shih, D.-T., Buck, C.A. & Hemler, M.A. J. Biol. Chem. 270, 25570−25577 (1995). | Article | PubMed | ISI | ChemPort |
  17. Takagi, J., Isobe, T., Takada, Y. & Saito, Y. J. Biochem. (Tokyo) 121, 914−921 (1997). | PubMed | ChemPort |
  18. Du, X. et al. J. Biol. Chem. 268, 23087−23092 (1993). | PubMed | ISI | ChemPort |
  19. Lee, B. & Richards, F.M. J. Mol. Biol. 55, 379−400 (1971). | Article | PubMed | ISI | ChemPort |
  20. Bailey, S. Acta Crystallogr. D 50, 760−763 (1994). | Article | PubMed | ISI |
  21. Janin, J. Nature. Struct. Biol. 4, 973−974 (1997). | Article | PubMed | ISI | ChemPort |
  22. Mehta, R.J. et al. Biochem. J. 330, 861−869 (1998). | PubMed | ISI | ChemPort |
  23. Smith, J.W., Piotrowicz, R.S. & Mathis, D. J. Biol. Chem. 269, 960−967 (1994). | PubMed | ISI | ChemPort |
  24. Hughes, P.E., O'Toole, T.E., Ylanne, J., Shattil, S.J. & Ginsberg, M.H. J. Biol. Chem. 270, 12411−12417 (1995). | Article | PubMed | ISI | ChemPort |
  25. Huang, C., Zang, Q., Takagi, J. & Springer, T.A. J. Biol. Chem. 275, 21514−21524 (2000). | Article | PubMed | ISI | ChemPort |
  26. Carr, C.M. & Kim, P.S. Cell 73, 823−832 (1993). | Article | PubMed | ISI | ChemPort |
  27. Bullough, P.A., Hughson, F.M., Skehel, J.J. & Wiley, D.C. Nature 371, 37−43 (1994). | Article | PubMed | ISI | ChemPort |
  28. Andrew, D. et al. Eur. J. Immunol. 23, 2217−2222 (1993). | PubMed | ISI | ChemPort |
  29. Studier, F.W. & Moffatt, B.A. J. Mol. Biol. 189, 113−130 (1986). | Article | PubMed | ISI | ChemPort |
  30. Blacklow, S.C. & Kim, P.S. Nature Struct. Biol. 3, 758−762 (1996). | Article | PubMed | ISI | ChemPort |
  31. North, C.L. & Blacklow, S.C. Biochemistry 38, 3926−3935 (1999). | Article | PubMed | ISI | ChemPort |
  32. Cavanagh, J., Palmer, I.A.G., Fairbrother, W. & Skelton, N.J. Protein NMR Spectroscopy: Principles and Practice. (Academic Press, San Diego; 1996).
  33. Pons, J.-L. Malliavin, T.E. & Delsuc, M.A. J. Biomol. NMR 8, 445−452 (1996). | Article | ISI | ChemPort |
  34. Bai, Y., Milne, J.S., Mayne, L. & Englander, S.W. Proteins 17, 75−86 (1993). | PubMed | ISI | ChemPort |
  35. Brunger, A.T. et al. Acta Crystallgr. D 54, 905−921 (1998). | Article |
  36. Stickle, D.F., Presta, L.G., Dill, K.A. & Rose, G.D. J. Mol. Biol. 226, 1143−1159 (1992). |