Locking the b 3 integrin I-like domain into high and low affinity conformations with disulfides

Although integrin a subunit I domains exist in multiple conformations, it is controversial whether integrin b subunit I-like domains undergo structurally analogous movements of the a 7-helix that are linked to affinity for ligand. Disulfide bonds were introduced into the b 3 integrin I-like domain to lock its b 6- a 7 loop and a 7-helix in two distinct conformations. Soluble ligand binding, ligand mimetic mAb binding and cell adhesion studies showed that disulfide bonded receptor a IIb b 3T329C/A347C was locked in a low affinity state, and dithiothreitol treatment restored capability of being activated to high affinity binding; by contrast, disulfide bonded a II b b 3 V332C/M335C was locked in a high affinity state. The results suggest that activation of the b subunit I-like domain is analogous to that of the a subunit I domain, i.e. that axial movement in the C-terminal direction of the a 7-helix is linked to rearrangement of the I-like domain MIDAS into a high affinity conformation. this conformation was stabilized in crystals by ligand or ligand-like lattice contacts (18,19). The studies reported here on the b I-like domain show that disulfide bonds mutationally introduced into the b 6/ a 7 region lock integrins that lack I domains into two distinct affinity states. The data uniquely support the proposal that downward movement of the a 7 helix induces I-like domain activation, and demonstrate that a I and b I-like domains are activated by structurally analogous mechanisms.

the structurally homologous I domain inserted in some integrin a subunits undergoes a similar piston-like movement of its C-terminal a7-helix, which regulates the affinity of its MIDAS for ligand.
There is controversy concerning this proposed mechanism. Soaking of a ligandmimetic Arg-Gly-Asp (RGD) peptide into integrin a V b 3 crystals, in which a V b 3 was constrained in the bent conformation by lattice contacts, induced b6-a7 loop and a1helix movements, but not a7-helix displacement (5). It was therefore suggested that a I and b I-like domains are activated by distinct mechanisms. Demonstration of movement of an epitope in the a1-helix was used to support the hypothesis that the mechanism of Ilike domain activation differs from that of the I domain (10). On the other hand, conformational change at this region would not contradict C-terminal a7-helix movement, and the mutation L358A in the a7-helix of the b1 I-like domain causes activation, supporting some type of conformational change around the a7-helix upon ligand binding (11). Furthermore, solution x-ray scattering studies and exposure of epitopes on the inner side of the hybrid domain in the presence of ligand (11,12) support the direct observations of hybrid domain swing-out (4,8).
Here, we directly test the hypothesis that specific rearrangements occur in the b6-a7 loop and a7-helix of b I-like domains that are structurally analogous to those that occur in a I domains, and are linked to integrin activation. Disulfide bonds have previously been introduced into a I domains to constrain the b6-a7 loop and a7-helix.  (20) of GeneMine v3.5 using residues 108-333 and 347-353 of Protein Databank accession 1JV2 (6) as template, and aligning them with residues 108-333 and 340-346 of the model sequence, respectively. This corresponded to a 7-residue, 2-turn displacement of the a7helix along its helical axis; residues 334-339 were left non-templated.

Plasmid construction, transient transfection and immunoprecipitation. Plasmids
coding for full-length human a IIb and b 3 were subcloned into pEF/V5-HisA or pcDNA3.1/Myc-His(+) as described previously (4). Mutants were made using sitedirected mutagenesis with the Quikchange kit (Stratagene, La Jolla, CA) and DNA sequences were confirmed before being transfected into 293T cells using calcium phosphate precipitates (21). Transfected cells were metabolically labeled with [ 35 S]cysteine/methionine as described (4). Lysates in 20 mM Tris-buffered saline pH 7.4 (TBS) supplemented with 1 mM Ca 2+ / 1 mM Mg 2+ , 1% Triton X-100 and 0.1% NP-40 were immunoprecipitated with 1 mg of anti-b 3 mAb AP3 and protein G Sepharose at 4 o C for 1 hour, and subjected to non-reducing SDS 7.5% PAGE and fluorography (22). Expression and ligand binding activity of a IIb b 3 on CHO-K1 transfectants. The plasmids described above coding for a IIb and b 3 were introduced into CHO-K1 cells using calcium phosphate precipitates (21). Transfectants were selected with 5 mg/ml G418. Jose, CA) was measured as described (4).

Labeling of free cysteines and
Cell adhesion to immobilized fibrinogen. Cell adhesion was assayed as described (23).

Design of I-like domains locked in low affinity and high affinity conformations
We hypothesized that in both the unliganded and liganded a V b 3 structures (6) the C-terminal a7 helix of the b 3 I-like domain is in a position stabilizing a closed, low affinity conformation; therefore, these structures were used to design low affinity mutants. An open, high affinity conformation was modeled assuming that the a7-helix was displaced in the C-terminal axial direction by two a-helical turns (Methods). The distance between Cb atoms of T329 and A347 in the unliganded and RGD-liganded a V b 3 structures is 4.4 Å and 4.9 Å, respectively, whereas it is 9.6 Å in the hypothesized high affinity model. Therefore, the mutant b 3 T329C/A347C was expected to form a disulfide bond in the low but not the high affinity conformation, and to be stabilized in the low affinity, the specificity-determining, b2-b3 loop near the b 3 I-like MIDAS (24). As the single cysteine mutants b 3 V332C and b 3 M335C were recognized by 7E3 (Fig. 2B), the conformational change induced by the disulfide bond formed between V332C and M335C (see below) appears to diminish the 7E3 epitope. By contrast, mutant b 3 T329C/A347C was well recognized by 7E3 (data not shown, and Fig. 4A).
Non-reducing SDS-PAGE of 35 S-labeled, immunoprecipitated receptors showed that the wild type and mutant a IIb subunits migrated similarly ( showed almost no biotin labeling (Fig. 2D, lane 2), the a IIb b 3 single cysteine mutants V332C and M335C showed marked labeling (Fig. 2D, lanes 4 and 6). The cysteines introduced in the V332C/M335C and T329C/A347C mutants clearly formed disulfides, because labeling was at the same level as the wild type (Fig. 2D, lanes 5 and 7), whereas it would have been twice that of the single cysteine mutants if disulfides had not formed.
To estimate the number of free cysteines per b 3 subunit, the ratio of the intensity of avidin binding to that of anti-myc binding was determined. As an additional control, wild-type

Ligand binding properties of 293T transfectants with disulfide-locked receptors
Binding to soluble fibrinogen was first examined using two-color flow cytometry (4) in transiently transfected 293T cells, in which wild type a IIb b 3 basally has low affinity for ligand. Wild type a IIb b 3 bound fibrinogen when stimulated with the activating mAb PT25-2, but not basally in Ca 2+ (Fig. 3A). Each of the four single cysteine mutants behaved similarly to the wild type receptor (Fig. 3A). By contrast, the putative locked mutant was abolished by two blocking a IIb mAbs HA5 and 10E5, but neither blocked nor further activated by the activating b 3 mAb AP5 (Fig. 3B), confirming that the high affinity binding of the transfected cells was specific.

Functional properties of mutant receptors in CHO-K1 transfectants
To further examine the disulfide-locked receptors, stable CHO-K1 transfectants were established, and clones were selected that expressed similar quantities of wild-type exception of 7E3 mAb (Fig. 4A). Mutant a IIb b 3 V332C/M335C blunted, but did not completely abolish the binding of 7E3.
CHO-K1 transfectants expressing the wild type receptor did not bind soluble fibrinogen or PAC-1 in Ca 2+ , but bound when stimulated by activating mAb PT25-2 ( Fig   4B and 4C). Treatment with 5 mM DTT at 20° for 30 min slightly increased ligand binding to wild type a IIb b 3 in Ca 2+ , but this binding was much less than that seen with The affinity state of disulfide-bonded mutants was further tested in cell adhesion assays on immobilized fibrinogen. High affinity is required for binding to soluble ligand or ligand mimetic mAbs. In contrast, wild type a IIb b 3 can mediate cell adhesion to immobilized fibrinogen in the absence of activation, as long as high coating concentrations above 1 mg/ml of fibrinogen are used (Fig 4D), consistent with our previous report (9). DTT treatment slightly increased the avidity of the wild type , the exposure of LIBS1, LIBS6 and PMI-1 epitopes behaved similarly to wild type. The LIBS mAbs bound poorly to the mutant in Ca 2+ , but Mn 2+ /RGD fully exposed the epitopes. Therefore, the high affinity mutant is in an overall bent conformation. These findings suggest that the high affinity ligand binding of mutant respectively (16). Therefore, it is difficult to know if the change in the b6-a7 loop induced by disulfide formation between C332 and C335 will accurately mimic physiologic rearrangement of this loop. Nonetheless, the position of residue 332 is largely fixed by its position in the b6-strand and the backbone hydrogen bonds between the b5 and b6 strands. Therefore, the backbone rearrangement required to form the C332-C335 disulfide bond is almost certain to come from a downward displacement of the b6-a7 loop, bringing C335 into position to form the disulfide bond with C332 that was directly demonstrated here by chemical labeling studies. The a IIb b 3 V332/M335C mutant was constitutively active in soluble ligand binding assays, and appeared to be maximally activated. The mutant was also highly active in adhesion to fibrinogen. The activity of the a IIb b 3 V332C/M335C mutant was not reversed by reduction. It is likely that the C332-C335 disulfide bond is resistant to reduction, like most wild type b 3 disulfides; however, we cannot rule out the possibility that both the disulfide bond, and the combination of two free cysteines at positions 332 and 335, are activating, even though each single cysteine is not. A converse result was obtained in a similar study on a L I domains: a disulfide designed to stabilize the high affinity conformation was reversible by DTT, whereas a disulfide designed to stabilize the low affinity conformation was not reversible with DTT (13,14,17).
The high affinity a IIb b 3 V332C/M335C mutant did not constitutively express activation epitopes, but these were induced upon treatment with RGD peptide and Mn 2+ . An analogous result was obtained with an a L I domain locked in the high affinity state with a disulfide bond (14). Subsequent crystal structure studies on the isolated, high affinity a L I domain demonstrated that the C-terminal a7-helix had indeed been displaced downward by the disulfide bond introduced into the b6-a7 loop, although there was some deformation of the a7-helix by the mutation (16). The intact, high affinity a L b 2 heterodimer remained in the bent conformation, and extension was activated by Mn 2+ , as revealed by mAb to LIBS or activation epitopes. The interpretation for the a L I domains is that the a7-helix should not be viewed as a rigid rod but rather as a spring or a rope, in other words, some looping out may occur so that a downward movement of the a7-helix in the I domain is not necessarily transmitted to other integrin domains (14). Similarly, after introduction here of the C332-C335 disulfide between the b6-strand and the b6-a7 Integrins are important therapeutic targets in many inflammatory and vascular disorders. The rational design of mutations that allosterically stabilize high affinity or low affinity conformations of integrins demonstrates marked advances in our understanding of the molecular basis of affinity regulation. This progress also holds out the promise that drugs might be designed that stabilize the low affinity conformation of integrins, in contrast to the current generation of "ligand-mimetic" integrin antagonists that stabilize the high affinity conformation.