HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

J. Biol. Chem., Vol. 276, Issue 38, 35854-35866, September 21, 2001

This Article Abstract Full Text (PDF) All Versions of this Article: 276/38/35854    most recent M103639200v1 Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Coe, A. P. F. Articles by Humphries, M. J. Search for Related Content PubMed PubMed Citation Articles by Coe, A. P. F. Articles by Humphries, M. J. Social Bookmarking                      What's this?

Generation of a Minimal ₅₁ Integrin-Fc Fragment^*

Alexander P. F. Coe, Janet A. Askari, Adam D. Kline, Martyn K. Robinson^§, Hishani Kirby^§, Paul E. Stephens^§, and Martin J. Humphries^¶

From the Wellcome Trust Centre for Cell-Matrix Research, School of Biological Sciences, University of Manchester, 2.205 Stopford Building, Oxford Road, Manchester, M13 9PT and ^§ Celltech Group plc., 216 Bath Road, Slough SL1 4EN, United Kingdom

Received for publication, April 24, 2001, and in revised form, May 31, 2001

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The tertiary structure of the integrin heterodimer is currently unknown, although several predictive models have been generated. Detailed structural studies of integrins have been consistently hampered for several reasons, including the small amounts of purified protein available, the large size and conformational flexibility of integrins, and the presence of transmembrane domains and N-linked glycosylation sites in both receptor subunits. As a first step toward obtaining crystals of an integrin receptor, we have expressed a minimized dimer. By using the Fc dimerization and mammalian cell expression system designed and optimized by Stephens et al. (Stephens, P. E., Ortlepp, S., Perkins, V. C., Robinson, M. K., and Kirby, H. (2000) Cell. Adhes. Commun. 7, 377-390), a series of recombinant soluble human ₅₁ integrin truncations have been expressed as Fc fusion proteins. These proteins were examined for their ligand-binding properties and for their expression of anti-integrin antibody epitopes. The shortest functional ₅-subunit truncation contained the N-terminal 613 residues, whereas the shortest ₁-subunit was a fragment containing residues 121-455. Each of these minimally truncated integrins displayed the antibody binding characteristics of ₅₁ purified from human placenta and bound ligand with the same apparent affinity as the native receptor.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Integrins are , heterodimeric transmembrane receptors that play central roles in cell adhesion, migration, differentiation, and survival (1). Several lines of evidence indicate that integrins also contribute to the progression of a wide variety of diseases, including inflammatory, thrombotic, and neoplastic conditions (2-4), and that the integrin families are valid therapeutic targets. The rational design of integrin antagonists based on ligand peptide motifs such as RGD and LDV is currently well advanced. Although the tertiary structure of the integrin heterodimer is unknown, this information would aid the process of drug development, and it represents one of the most important outstanding questions in the field.

The overall shape and dimensions of the IIb₃ and ₅₁ integrin heterodimers have been revealed by rotary shadowing electron microscopy (5-7). Both receptors consisted of an N-terminal globular head of 8-12 nm with two extended tails of 18-20 nm that corresponded to the C termini (7). Similarly, a soluble IIb₃ integrin, generated by removal of the IIb and ₃ transmembrane and cytoplasmic domains, and the ₃ cysteine-rich repeats also contained a globular head, but its tails were 4-6 nm shorter (8).

In the absence of a tertiary structure for the integrin heterodimer, several predictive models have been generated (9-12), and these have subsequently been supported by biochemical analyses (13-16). Whereas the - and -subunits are unrelated in primary sequence, they share common structural features including an N-terminal globular ligand-binding domain, C-terminal stalk regions, transmembrane domains, and short cytoplasmic domains (17, 18). The N-terminal portion of -subunits contains seven homologous repeats, each 60-70 amino acid residues in length. These repeats are quite similar in sequence, and repeat four in some integrins and repeats five to seven in all receptors contain EF-hand-like divalent cation-binding motifs. The seven repeats have been predicted to fold cooperatively, forming an all- structure known as a -propeller fold (9) (Fig. 1). One-third of -subunits contain a von Willebrand factor A-domain of about 200 amino acid residues in length inserted between repeats two and three. In all integrins examined to date, this domain contains the major elements involved in ligand binding. The tertiary structures of several A-domains have been solved by x-ray crystallography and shown to adopt a Rossmann fold (16, 19-22). Ligand interaction has been localized to one face of this domain via a cation coordinated at a so-called metal ion-dependent adhesion site (MIDAS)¹ (23-26). The C-terminal stalk region of -subunits has been predicted to be mainly -strand and to form 4-6 -barrels or -sandwiches (27). The region of the -subunit that contains ligand-binding residues has also been predicted to form an A-domain (the A-domain) that interacts with ligands via a MIDAS motif (Refs. 10, 20, 27-33; Fig. 1). This domain is preceded by a recently described plexin-semaphorin-integrin (PSI) domain (34) and is followed by four cysteine-rich EGF-like domains.

Integrin ₅₁ is a widely distributed and well studied cell-surface fibronectin receptor. Analysis of the cell binding function of fibronectin has identified a central cell-binding domain composed of catenated modules known as type III repeats. An RGD peptide motif contained in the 10th type III repeat has been shown to be the major binding site for ₅₁ (35, 36), although a requirement for a "synergy" region contained in the 9th type III repeat has also been demonstrated (37-39). The fibronectin crystal structure has shown that both the RGD motif and the synergy region are on the same face of the fibronectin molecule (40, 41). As both integrin subunits contribute to ligand binding, the fibronectin structure imposes constraints on the tertiary structure of the integrin heterodimer; specifically the - and -subunits are likely to be arranged side-by-side in order to form the ligand-binding pocket (Fig. 1 (19)).

Specific regions and individual amino acid residues in the ₅-subunit that contribute to the ₅₁ ligand-binding pocket have been identified recently (42, 43). Ligand-binding residues lie in loop regions within the first three repeats of the -propeller model, as do the epitopes of all anti-₅ function-blocking mAbs (44-50). Therefore, the C terminus of the ₅-subunit (which lies outside of the N-terminal repeats and prior to the transmembrane region; Fig. 1) does not directly contribute to ligand binding although it may contain regulatory elements (27). Analogously, the ₃- and ₂-subunit C termini have been shown to be unnecessary for ligand binding, since removal of these domains by truncation followed by co-expression with the relevant full-length -subunit generated functional integrins (8, 51). Interestingly, both of these -subunit truncations maintained conserved cysteine residues and thereby retained a disulfide knot between the N terminus and mid-section of the -subunit, which was proposed to aid structural integrity (52).

Several groups have reported the production of soluble variants of integrins in which the transmembrane and cytoplasmic regions have been removed as follows: M₂ (53, 54), IIb₃ (8, 55-58), _L2 (59), ₃₁ (60), ₈₁ (61), _V3 (62), _V5 (63), and ₄₁ (64, 65). Although all of the soluble integrins were produced by deletion of the transmembrane and cytoplasmic domains, some reports have implicated these regions in heterodimer formation and maintenance (66-69). More recently soluble integrins have been produced as fusions in which the transmembrane and cytoplasmic domains were replaced with sequences to aid heterodimerization (60, 63, 65). As described in Ref. 65, we have expressed ₄₁ as an Fc fusion in mammalian cells. The Fc domain was utilized to drive heterodimer formation, and the C_H3 domains were modified to contain specific mutations that enhanced this process.

With the exception of Ref. 8, all previous soluble integrins have represented the entire ectodomains of each subunit. The apparent lack of success in generating soluble truncated -subunits may be due to poor heterodimer formation. We therefore reasoned that expression of the ₅- and ₁-ectodomain as a human ₁ Fc fusion would facilitate heterodimer formation in these truncated subunits. A series of integrin-Fc truncations were generated and tested for ligand- and mAb-binding properties in comparison to native ₅₁, purified from human placenta. The results identify a minimal integrin that retains full ligand binding functionality.

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Monoclonal Antibodies (mAbs)-- Rat mAbs 16 and 11, recognizing the human ₅-subunit, and mAb 13, recognizing the human ₁-subunit, were gifts from K. Yamada (NIDCR, National Institutes of Health, Bethesda). Mouse anti-human ₅ mAb JBS5 was from Serotec (Oxford, UK). Mouse anti-human ₁ mAbs 8E3 and 12G10 were developed in-house (77, 78). Mouse anti-human ₁ mAb TS2/16 was a gift from F. Sánchez-Madrid (Hospital de la Princesa, Madrid, Spain). Rat anti-human ₁ mAb 9EG7 was a gift from D. Vestweber (University of Munster, Germany). Goat anti-human ₁ Fc mAb was from Jackson Immunochemicals (Stratech Scientific, Luton, UK). All antibodies were used as purified IgG. Oligonucleotides were purchased from MWG Biotech (Southampton, UK), Oswel (Southampton, UK), or PE-Applied Biosystems (Warrington, UK). Restriction enzymes were from New England Biolabs (Hitchin, UK) or Roche Molecular Biochemicals. Integrin ₅₁ was purified from human placenta as described previously (78). A recombinant fragment of the cell-binding domain of fibronectin (III-(6-10)) was produced as before (80) and purified using DEAE-Sephacel (Amersham Pharmacia Biotech) and hydroxylapatite (Bio-Rad) chromatography, as described previously (18). GRGDS and GRDGS peptides were synthesized using Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry on an Applied Biosystems 431A peptide synthesizer and purified as described previously (81).

Mutagenesis of Integrin Subunit cDNAs and Expression Vector Construction-- The full-length cDNA construct encoding the human ₅-subunit (in pcDNA3) was a gift from K. Yamada (NIDCR, National Institutes of Health, Bethesda (82)). The cDNA construct encoding the human ₁-subunit (Asp⁷⁰⁸) ectodomain-Fc (in pV.16hFc) was generated as in Ref. 65. By using oligonucleotide-directed PCR mutagenesis, a unique HindIII restriction site was engineered into the 5' end of the ₅ cDNA immediately prior to, and in-frame with, the ATG start codon. A unique SalI site was introduced at the junction between the extracellular domain and the transmembrane region, in-frame with and after residue Glu⁹⁵¹ of ₅ (79) by the same method. The ₅ cDNA was subcloned into a derivative of pEE12 (83), pEE12.2hFc. pEE12.2hFc had the vector SalI site removed and replaced by a NotI site. The human ₁ Fc domain, encoding 14 residues of the upper hinge, the hinge, and the constant domain regions, C_H2 and C_H3, was inserted as a SalI/EcoRI genomic fragment (84). Mutagenesis of the ₅-subunit cDNA thus enabled variants of the extracellular domain of the ₅ gene to be cloned upstream of the Fc domain as in-frame fusions.

Constructs containing C-terminal truncated cDNAs were generated via a similar PCR method that incorporated a 3' SalI site immediately after, and in-frame with, the codon specifying the residue chosen as the truncation point. Therefore, the ₅ and ₁ truncated ectodomains were also produced as in-frame Fc fusions. Constructs containing N-terminal truncated cDNAs were also generated via PCR, by using a large oligonucleotide that contained a 5' HindIII site immediately prior to, and in-frame with, a murine antibody leader sequence (85) upstream and in-frame with the codon specifying the residue chosen as the truncation point.

To increase the likelihood of heterodimerization between the ₅ and ₁ chains, specific mutations were introduced into the C_H3 domains of the chimeric proteins by oligonucleotide-directed PCR mutagenesis. Briefly, in the Fc DNA contained in pV.16hFc, residue 366 was changed from a threonine to a tyrosine (creating a "knob"). Conversely, in the Fc DNA contained in pEE12.2hFc, residue 407 was changed from a tyrosine to a threonine (creating a "hole") as described (86).

Lipofection of CHOL761h Cells-- CHOL671h cells (28) were transiently transfected using the LipofectAMINE^TM method (Life Technologies, Inc.), according to the manufacturer's instructions. Cells were seeded into 75-cm³ flasks and grown to confluence. 20 µg of the ₅ and ₁ vectors were mixed with 1 ml of Dulbecco's modified Eagle's medium, and a second mixture of 60 µl of LipofectAMINE^TM reagent in 1 ml of Dulbecco's modified Eagle's medium was prepared. The two mixtures were combined before incubating for 30 min at room temperature. The lipofection mixture was added onto the cells and incubated at 37 °C for 4 h before replacement with culture medium. 24 h post-transfection the culture medium was replaced, and cells were incubated at 37 °C in a humidified atmosphere of 5% (v/v) CO₂ for 96 h.

Purification of ₅₁-Fc Integrins-- Culture supernatant was harvested from transfected CHOL761h cells by centrifugation at 1000 × g. Soluble integrin was purified via the Fc domain after mixing with protein A-Sepharose (1 ml) at 4 °C for 16 h. The Sepharose beads were collected into a 1.6-ml column and washed with 10 column volumes of Tris-buffered saline containing 1 mM MgCl₂ and 1 mM CaCl₂. The soluble integrin was eluted using 0.1 mM citric acid, pH 3, and neutralized in 1 M Tris-HCl, pH 8. Protein containing fractions were identified by SDS-PAGE.

Solid Phase Ligand Binding Assay-- Purified integrin was tested for ligand binding activity using a solid phase protein-protein interaction assay. Soluble integrin was coated into half-volume wells of a 96-well ELISA plate (Costar, High Wycombe, UK) in PBS+ for 16 h at room temperature. The coating solution was removed and replaced with 5% (w/v) BSA (Calbiochem) in Tris-buffered saline (200 µl per well) to block nonspecific binding and incubated for 2 h at room temperature. A biotinylated tryptic fibronectin fragment, FnIII-(6-10) (80), containing the ₅₁ central cell-binding domain was used as ligand. FnIII-(6-10) was diluted in Tris-buffered saline, 1 mM MnCl₂, 1 mg/ml BSA (buffer A) to the required concentration, and 50 µl was added to each well. The plate was incubated for 1 h at room temperature. At this stage, when required, GRGDS peptide or GRDGS peptide, diluted to the required concentration in buffer A, was co-incubated with FnIII-(6-10) (1 µg/ml) as a specificity test. The plate was washed three times with buffer A (200 µl per well) before addition of 50 µl of ExtrAvidin®-peroxidase (1:500 dilution in buffer A; Sigma) and incubated for 20 min at room temperature. The plate was washed four times with buffer A (200 µl per well), and 50 µl of 0.1% (w/v) 2,2'-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) (ABTS) in 0.1 M sodium acetate, 0.05 M NaH₂PO₄, pH 5.0, 0.01% H₂O₂ (v/v) was added. Absorbance readings were measured at 405 nm on a Dynatech MR4000 plate reader. Each sample was assayed in quadruplicate, and attachment to BSA (<5% of the total) was subtracted from all measurements. Each experiment shown is representative of at least three separate experiments.

Sandwich ELISA for Epitope Expression-- Purified integrins were tested for correct folding by detection of antibody epitopes. Goat anti-human ₁ Fc antibody (diluted to 2.6 µg/ml in PBS+; Jackson Immunochemicals, Stratech Scientific Ltd., Luton, Bedfordshire, UK) was coated into wells of a 96-well Immulon4 ELISA plate (Dynatech, Chantilly, VA) for 16 h at 4 °C. The coating solution was removed and replaced with purified proteins diluted in PBS+ to 25 µg/ml, and the plate was incubated for 1 h at room temperature. The plate was washed three times with buffer A (200 µl per well) before addition of anti-₅ and anti-₁ monoclonal antibodies (diluted to 10 µg/ml in buffer A) to the relevant wells. The plate was incubated for 1 h at room temperature before washing three times with buffer A (200 µl per well). 100 µl of appropriate peroxidase-conjugated secondary antibodies (1:2000 dilution in buffer A; Jackson Immunochemicals, Stratech Scientific Ltd., Luton, Bedfordshire, UK) were added to each well and incubated for 1 h at room temperature. The plate was washed four times with buffer A (200 µl per well), and 100 µl of ABTS substrate was added. Absorbance readings were measured at 405 nm on a Dynatech MR4000 plate reader. Each sample was assayed in quadruplicate, and attachment to BSA (<5% of the total) was subtracted from all measurements. Each experiment shown is representative of at least three separate experiments.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Generation of Recombinant Soluble ₅(Glu-951)₁(Asp-708)-Fc

Recombinant soluble human ₅₁ integrin was produced as an Fc chimera following fusion of the cDNAs encoding the ectodomains of each subunit to genomic DNA encoding the Fc of human ₁ IgG. Separate vectors encoding the ectodomains of the ₅-subunit and ₁-subunit fused to specially modified human ₁ Fc domains (Fig. 1) were transiently expressed in CHOL761h cells. Integrin was purified using protein A affinity chromatography and analyzed by non-reducing SDS-PAGE. The major product was a protein of 300 kDa corresponding to that expected for an intact integrin heterodimer maintained by disulfide bonds in the immunoglobulin hinge (Fig. 2A). A protein of 360 kDa, expressed at a greatly reduced level, was also observed. This protein was shown to correspond to ₅(E951)₅(E951)-Fc homodimer using Western blotting with subunit-specific mAbs (data not shown). Supernatant was also examined by Western blotting under reducing conditions using an anti-human Fc antibody. The results revealed a band of 180 kDa, which corresponded to the ₅(E951)-Fc protein, and a band of 140 kDa, which corresponded to the ₁(D708)-Fc protein (Fig. 2b).

View larger version (31K): [in this window] [in a new window]   Fig. 1.   Modular diagram of the recombinant soluble human 51-Fc integrin. The N terminus of the 5-subunit is predicted to form a -propeller, represented here by a seven-sided disc, repeat W1 is green; W2 is turquoise, etc., until the yellow W7, which contains a cation shown by an orange sphere coordinated at the EF hand-like motif. Note the cations within repeats 5 and 6 are obscured. The C terminus of the 5-subunit is also predicted to be an all--strand structure, forming -barrels or sandwiches (27). The transmembrane and cytoplasmic domains are replaced with the Fc "hole" domain, represented by orange ovals. The putative 1-subunit A-domain is shown by a green oval, with a cation, shown by an orange sphere, coordinated at the MIDAS motif. Preceding the A-domain is the PSI domain (blue sphere), which is disulfide-bonded to the inter-domain region, shown by a kink, that follows the domain of no known homology (pink oval). C-terminal to these domains are four cysteine-rich EGF-like domains, shown by yellow spheres. Following these is the Fc knob domain, represented by orange ovals, with specific mutations shown by red triangles. The upper hinge and disulfide bonds are shown in gray. The ligand binding pocket is formed from elements within 5 N-terminal repeats 2 and 3 and the 1-subunit putative MIDAS motif. The C termini are shown extended for clarity.

View larger version (26K): [in this window] [in a new window]   Fig. 2.   Characterization of purified 51-Fc. A, the 51-Fc was purified by protein A affinity chromatography and analyzed by SDS-PAGE. The 5(E951)5 (E951)-Fc homodimer and 5(E951)1 (D708)-Fc heterodimer are marked. B, the 51-Fc integrin was resolved by reducing SDS-PAGE, prior to transfer to nitrocellulose and Western blotting analysis. The 51-Fc fusion was detected using a peroxidase-conjugated anti-human 1 Fc antibody (0.5 µg/ml). The 5(E951)-Fc and 1(D708) proteins are marked. C, the 5(E951)1(D708)-Fc integrin was dispensed into wells of a 96-well plate pre-coated with anti-human 1 Fc antibody. Anti-5 (mAb16, mAb11, JBS5) and anti-1 (12G10, TS2/16, mAb13) antibodies (1 µg/ml) were detected using appropriate HRP-conjugated secondary antibodies (1 µg/ml; hatched bars). Placenta-derived native human 51 integrin served as positive control (diluted 1:100 or ~0.5 µg/ml; open bars). D, varying concentrations of biotinylated FnIII-(6-10) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(D708)-Fc, and bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis, the apparent affinity of FnIII-(6-10) binding was 2.26 ± 0.14 nM. E, varying concentrations of GRGDS or GRDGS peptide together with biotinylated FnIII-(6-10) (1 µg/ml) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(D708)-Fc. Bound ligand was detected using ExtrAvidin®-peroxidase. By nonlinear regression analysis the concentration of GRGDS peptide required for 50% inhibition (IC50) of FnIII-(6-10) binding was 2.22 ± 0.49 µg/ml. F, the 5(E951)1(D708)-Fc integrin was dispensed into wells of a 96-well plate pre-coated with anti-human 1 Fc antibody (2.5 µg/ml). Varying concentrations of divalent cations or EDTA were included in the biotinylated FnIII-(6-10) (1 µg/ml) soluble phase. Bound ligand was detected using ExtrAvidin®-peroxidase.

The ability of protein A-purified ₅(E951)₁(D708)-Fc to bind to a panel of conformation-dependent anti-₅ and anti-₁ mAbs was examined to assess whether the protein was correctly folded. A sandwich ELISA technique was employed in which the integrin-Fc fusion was captured via its Fc domain using an anti-human ₁ Fc antibody. The anti-human ₁ Fc antibody and placenta-derived human ₅₁, which served as a positive control, were coated directly on to the ELISA plate. The reactivity of ₅(E951)₁(D708)-Fc to both anti-₅ mAbs (16, 11, JBS5) and anti-₁ mAbs (12G10, TS2/16, mAb13) was very similar to native receptor (Fig. 2c). mAb reactivity varied depending on the mAb used; however, this was consistent with the variation seen with purified native ₅₁ integrin, suggesting that ₅(E951)₁(D708)-Fc adopted a native fold. The anti-₅ mAb11 recognizes a linear epitope and serves as a marker for the expression of the ₅(E951)₁(D708)-Fc. As can be seen in Fig. 2C, the level of mAb11 binding to ₅(E951)₁(D708)-Fc was greater than to placenta-derived ₅₁; this observation may be explained by the binding of mAb11 to the ₅(E951)₅(E951)-Fc homodimer.

Purified ₅(E951)₁(D708)-Fc was assayed for its ability to bind ligand, a recombinant fibronectin fragment containing type III repeats 6-10 (FnIII-(6-10)). Integrin-Fc fusion was directly coated onto ELISA plates, and bound FnIII-(6-10) was detected by streptavidin-HRP. The results (Fig. 2D) showed that ₅(E951)₁(D708)-Fc was able to bind ligand. The apparent affinity of this interaction was calculated by non-linear regression analysis (18). A K_d value of 2.26 ± 0.14 nM was obtained, which was slightly higher than the published data for placenta-derived ₅₁ (1.1 nM (18)). To determine whether the integrin-fibronectin interaction was specific, competition by GRGDS peptide was tested. The results (Fig. 2e) showed that FnIII-(6-10) binding was inhibited by GRGDS, but not by GRDGS, suggesting that the interaction was indeed specific and that fibronectin binding was RGD-dependent. The IC₅₀ value, or concentration of GRGDS required for half-maximal inhibition, was calculated as 2.22 ± 0.49 µg/ml, but GRDGS peptide had little effect, even at 100 µg/ml.

The divalent cation requirement for FnIII-(6-10) binding to purified ₅(E951)₁(D708)-Fc was assessed using a FnIII-(6-10)-binding ELISA and varying concentrations of divalent cations. The results (Fig. 2f) demonstrated that no ligand binding was observed in the absence of divalent cations or in the presence of EDTA or Ca²⁺. Conversely in the presence of Mn²⁺ or Mg²⁺ ligand binding was readily detectable. The degree of ligand binding was greater with Mn²⁺ than with Mg²⁺. This indicated that Mn²⁺ was a more efficient activator of ₅(E951)₁(D708)-Fc than Mg²⁺. Ligand binding increased with increasing concentrations of either Mn²⁺ or Mg²⁺; however, whereas ligand binding peaked with 1 mM Mn²⁺, it continued to increase even at 3 mM Mg²⁺. These data are in good agreement with the published data for placenta-derived human ₅₁ (70), showing that ₅(E951)₁(D708)-Fc interacts with FnIII-(6-10) in a divalent cation-dependent manner.

Thus, the Fc system enabled the production of soluble full-length ₅₁-Fc integrin. Incorporation of the Fc domains did not compromise folding of the integrin or the relative positioning of the subunits, since the ligand-binding pocket and regulatory elements involved in the divalent cation response of the integrin were intact.

Truncation of the ₁-Subunit

To identify the domain(s) of the ₁-Fc subunit, which when co-expressed with ₅(E951)-Fc, were necessary and sufficient for ligand recognition, various truncations were generated.

C-terminal Truncation-- The cDNA encoding the ₁(D708) subunit was C-terminally truncated to ₁(P455) and fused in-frame to the human ₁ Fc genomic DNA. This truncation point was chosen because it lies just C-terminal to the conserved cysteine residues that form long range structural disulfide bonds and would therefore not be expected to perturb these structural features (Table I). Expression of the ₁(P455)-Fc construct would yield a C-terminally truncated ₁-subunit, lacking the four cysteine-rich repeats and encompassing the ligand-binding domain from Gln¹ through Pro⁴⁵⁵, reminiscent of previous studies (8, 51).

View this table: [in this window] [in a new window]   Table I The 1 truncation series Truncation points were chosen in accordance with previous  subunit truncation studies (8, 51). Truncation Pro455 was confined to a predicted loop region to minimize structural disturbance and maintained proposed disulfide bonds (52), shown after alignment of 1 and 3. All other truncations disrupted the proposed 1 Cys7-Cys444 bond. Truncation name numbering refers to the position of the amino acid residue within the mature protein.

The construct encoding the ₁(P455)-Fc subunit was co-expressed with ₅(E951)-Fc in Chinese hamster ovary cells and integrin Fc-purified using protein A affinity chromatography. Purified protein was examined by Western blotting under reducing conditions using an anti-human Fc antibody. The results revealed a band of 180 kDa, which corresponded to the ₅(E951)-Fc protein and a band of 110 kDa that corresponded to the ₁(P455)-Fc protein (Fig. 3A, lane 2). For comparison, the ₅(E951)₁(D708)-Fc integrin is shown in lane 1, revealing the 140-kDa ₁(708)-Fc subunit.

View larger version (33K): [in this window] [in a new window]   Fig. 3.   Characterization of 5(E951)1(P455)-Fc. A, 51-Fc (lane 1) and 5(E951)1(P455)-Fc (lane 2) were resolved by reducing SDS-PAGE, prior to transfer to nitrocellulose and Western blotting analysis. The integrin-Fc fusions were detected using peroxidase-conjugated anti-human 1 Fc antibody (0.5 µg/ml). The 5(E951)-Fc, 1(D708)-Fc, and 1(P455)-Fc proteins are marked. B, 5(E951)1(P455)-Fc integrin was dispensed into wells of a 96-well plate pre-coated with anti-human 1 Fc antibody (2.5 µg/ml). Anti-5 (mAb16, mAb11, JBS5) and anti-1 (12G10, TS2/16, 9EG7, mAb13) antibodies (1 µg/ml) were detected using appropriate HRP-conjugated secondary antibodies (1 µg/ml; hatched bars). The full-length 5(E951)1(D708)-Fc integrin served as positive control (open bars). C, varying concentrations of biotinylated FnIII-(6-10) (3-0.01 µg/ml; X axis) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(P455)-Fc. Bound ligand was detected using ExtrAvidin®-peroxidase. Absorbance values were the average of four wells, and error bars show the standard deviation of the four readings. By non-linear regression analysis, the apparent affinity of FnIII-(6-10) binding was 1.89 ± 0.11 nM. This value represents the 50% or the half-maximal value of ligand binding. D, varying concentrations of GRGDS or GRDGS peptide together with biotinylated FnIII-(6-10) (1 µg/ml) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(P455)-Fc, and bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis the concentration of peptide required for 50% inhibition (IC50) of FnIII-(6-10) binding was 4.64 ± 0.64 µg/ml.

The effects of removal of the ₁ C terminus on the folding of ₅(E951)₁(P455)-Fc was assessed by sandwich ELISA (Fig. 3b). As above, mAb reactivity varied depending on the mAb used; however, this was consistent with ₅(E951)₁(D708)-Fc, suggesting that ₅(E951)₁(P455)-Fc adopted a native fold. The binding of the anti-₁ mAb 9EG7 was abolished, which is consistent with its epitope lying within EGF-like repeats 2-4 (71). Most important, all other anti-₁ mAbs reacted with the ₁(P455)-Fc-truncated subunit, demonstrating that the N-terminal ligand-binding domain does not require the C-terminal domains for folding. Similarly, reactivity of the anti-₅ mAbs demonstrated that the ₁ N-terminal domains was sufficient for folding of the ₅-subunit.

To establish whether ₅(E951)₁(P455)-Fc was able to bind FnIII-(6-10), protein A-purified protein was examined by a FnIII-(6-10) interaction assay. ₅(E951)₁(P455)-Fc was functional and exhibited an apparent K_d of 1.89 ± 0.2 nM (calculated by non-linear regression analysis (18); Fig. 3c). This was in good agreement with the calculated value for ₅(E951)₁(D708)-Fc (2.26 ± 0.14 nM). The ₅(E951)₁(P455)-Fc-FnIII-(6-10) interaction was also RGD-dependent (Fig. 3d).

Further C-terminal Truncation and Involvement of Conserved Cysteine Residues-- The results shown in Fig. 3, b and c, are summarized in Table II. Table II also shows the results of additional ₁ truncations upstream of Pro⁴⁵⁵ to Val⁴³⁶ and Asn⁴¹³. These truncations were generated to assess the involvement of conserved cysteine residues (52) in structural maintenance and to delimit the C terminus of a minimal ₁-subunit (Table I). As shown in Table II, co-expression of these constructs with ₅(E951)-Fc yielded integrins that displayed no reactivity to conformation-dependent mAbs or to FnIII-(6-10). The anti-₅ mAb11 (which recognizes a linear epitope in the ₅ light chain), the anti-₁ mAb 8E3, and an anti-human ₁ Fc antibody displayed reactivity suggesting the ₅(E951)₁(V436)-Fc and ₅(E951)₁(N413) proteins were expressed (but in a misfolded state). Expression of both subunits was subsequently confirmed by Western blotting (data not shown).

View this table: [in this window] [in a new window]   Table II mAb reactivity of 5(E951)1(TR)-Fc proteins Integrin-Fc fusions included the three C-terminal 1 truncations (1(P455)-Fc, 1(V436)-Fc, and 1(N413)-Fc) and the three cysteine-serine 1(P455)-Fc mutants (C7S, C444S, and C442S,C444S,C446S). These six 1 constructs were co-expressed with 5(E951)-Fc. Soluble integrins were captured by Fc, and the binding of various anti-5 and anti-1 mAbs was assessed by ELISA. mAbs that reacted with the mutated and/or truncated integrins are indicated by a +; those that did not react with the mutated and/or truncated integrins are indicated by a . Anti-5 mAbs included mAb16 and JBS5, whereas anti-1 mAbs included 12G10, TS2/16, and mAb13; Fc is an anti-human 1 Fc mAb. The binding of FnIII-(6-10) was also tested, both alone and in the presence of the stimulatory anti-1 mAb 12G10 (1 µg/ml); + indicates binding occurred;  indicates absence of binding.

These findings suggested that conserved cysteine residues (52) may be involved in maintaining the tertiary structure of the ₁-subunit. However, as shown in Table II, mutation of four cysteine residues (Cys⁷, Cys⁴⁴², Cys⁴⁴⁴, and Cys⁴⁴⁶) to serine yielded proteins that retained mAb and FnIII-(6-10) reactivity. This suggested that whereas these conserved cysteine residues may be involved in structural maintenance, other elements exist, including hydrophobic amino acids between Val⁴³⁶ and Pro⁴⁵⁵ that have a greater involvement. It is possible that truncation to Val⁴³⁶ and Asn⁴¹³ exposes these hydrophobic amino acid residues, leading to protein misfolding.

N-terminal Truncation-- To minimize the ₁-subunit further, the requirement of the N-terminal PSI domain for folding and function was assessed. An N-terminal truncation to ₁(Y121), in the ₁(P455)-Fc construct, generated a ₁-subunit encompassing the ligand-binding domain from Tyr¹²¹ through Pro⁴⁵⁵ (Table I). The ₁(Y121-P455) truncation point was two amino acid residues upstream of the start of the putative ₁-subunit A-domain (10).

₁(Y121-P455)-Fc was co-expressed with ₅(E951)-Fc in CHOL761h cells. Integrin-Fc was purified using protein A affinity chromatography and examined by Western blotting under reducing conditions using an anti-human Fc antibody. The results revealed a band of 180 kDa, which corresponded to the ₅(E951)-Fc protein and a band of 90 kDa that corresponded to the ₁(Y121-P455)-Fc protein (Fig. 4A, lane 3).

View larger version (26K): [in this window] [in a new window]   Fig. 4.   Characterization of 5(E951)1(Y121-P455)-Fc. A, 51-Fc (lane 1), 5(E951)1(P455)-Fc (lane 2), and 5(E951)1(Y121-P455)-Fc (lane 3) were resolved by reducing SDS-PAGE, prior to transfer to nitrocellulose and Western blotting analysis. The integrin-Fc fusions were detected using peroxidase-conjugated anti-human 1 Fc antibody (0.5 µg/ml). The 5(E951)-Fc, 1(D708)-Fc, 1(P455)-Fc and 1(Y121P455)-Fc proteins are marked. B, the 5(E951)1(Y121-P455)-Fc integrin was dispensed into wells of a 96-well plate pre-coated with anti-human 1 Fc antibody. Anti-5 (mAb16, JBS5) and anti-1 (12G10, TS2/16, 8E3, mAb13) antibodies (1 µg/ml) were detected using appropriate HRP-conjugated secondary antibodies (1 µg/ml; hatched bars). 5(E951)1(P455)-Fc served as positive control (open bars). C, varying concentrations of biotinylated FnIII-(6-10) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(Y121-P455)-Fc, and bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis, the apparent affinity of FnIII-(6-10) binding was 0.98 ± 0.07 nM. D, varying concentrations of GRGDS or GRDGS peptide together with biotinylated FnIII-(6-10) (1 µg/ml) were dispensed into wells of a 96-well plate pre-coated with purified 5(E951)1(Y121-P455)-Fc. Bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis the concentration of peptide required for 50% inhibition (IC50) of FnIII-(6-10) binding was 1.51 ± 0.62 µg/ml.

The effects of removal of the ₁ N terminus on folding were assessed by sandwich ELISA. The results showed that ₅(E951)₁(Y121-P455)-Fc integrin displayed reactivity to both anti-₅ and anti-₁ mAbs, demonstrating that the PSI domain was not required for folding (Fig. 4B). The binding of the non-anti-functional anti-₁ mAb 8E3 was abolished, as was binding to another ₁ truncation (Lys⁸⁷-Pro⁴⁵⁵; not shown), suggesting that this epitope is localized to the PSI domain. Although the pattern of mAb reactivity was the same as larger fragments, the level of mAb binding to ₅(E951)₁(Y121-P455)-Fc integrin was reduced as compared with ₅(E951)₁(P455) protein. This may be due to increased instability of the product caused by removal of the N terminus of the subunit.

To establish whether ₅(E951)₁(Y121-P455)-Fc was able to bind FnIII-(6-10), purified protein was examined by a FnIII-(6-10) interaction assay. Again the truncated integrin retained ligand binding activity and displayed an apparent K_d of 0.98 ± 0.07 nM (Fig. 4C). This was in good agreement with the calculated values for other soluble ₅₁-Fc integrins (above). Binding was again RGD-dependent (Fig. 4D).

Truncation of the ₅-Subunit

To identify the domain(s) of the ₅-Fc subunit that, when co-expressed with the minimized ₁-subunit, were necessary and sufficient for ligand recognition, various ₅ truncations were generated (Tables III and IV). The secondary structure of the -subunit has been proposed to be exclusively -strand through the presence of an N-terminal -propeller and C-terminal -barrels/sandwiches (9, 27). The ₅ primary sequence was therefore analyzed for secondary structure elements using the PHD algorithm (data not shown). A truncation series that encompassed Phe¹ through to Ala⁸⁵⁵ (representing the ₅ heavy chain, see Ref. 8) and Phe¹ through Ser⁴⁰⁹ was generated by PCR (Table III). Truncation points were confined to putative loop regions and maintained the predicted disulfide bonding pattern, elucidated via alignment with IIb (72) (data not shown), to minimize structural disturbance. Since accuracy cannot be guaranteed with the secondary structure prediction, truncations were made in proposed domain boundaries and also half-way through proposed domains. These constructs were subsequently co-expressed with the ₁(P455)-Fc subunit in mammalian cells.

View this table: [in this window] [in a new window]   Table III The 5 truncation series Truncation points were chosen after analyzing the 5 primary sequence for secondary structure elements using PHD. Truncation points were confined to predicted loop regions to minimize structural disturbance and maintained proposed disulfide bonds (72), shown after alignment of 5 and IIb. Truncation name numbering refers to amino acid position within the mature protein. The 5(R470)-Fc subunit did not form a folded and functional heterodimer with the 1(Y121-P455)-Fc subunit.

View this table: [in this window] [in a new window]   Table IV mAb reactivity of 5(TR)1(P455)-Fc proteins Integrin-Fc fusions included the entire 5 C-terminal truncation series constructs individually co-expressed with the 1(P455)-Fc construct. Soluble integrins were captured by Fc, and the binding of various anti-5 and anti-1 mAbs was assessed by ELISA. mAbs that reacted with the truncated integrins are indicated by a +; those that did not react with the truncation are indicated by a . Anti-5 mAbs included mAb16 and JBS5, whereas anti-1 mAbs included 12G10, TS2/16, 8E3, and mAb13; Fc is an anti-human 1 Fc mAb. The binding of FnIII-(6-10) was also tested both alone and in the presence of the stimulatory anti-1 mAb 12G10 (1 µg/ml); + indicates binding occurred;  indicates absence of binding. In all cases, saturating amounts of protein were captured by the anti-Fc mAb, but it should be noted that the least well folded constructs also tended to be expressed at lower levels as might be expected for unstable proteins.

The effects of C-terminal truncation of the ₅-subunit on the folding of ₅(TR)₁(P455)-Fc were assessed by sandwich ELISA. The results (Table IV) showed that C-terminal truncation of the ₅-subunit in-between, and including, ₅(A855)-Fc and ₅(D613)-Fc formed heterodimers with the ₁(P455)-Fc subunit and expressed the epitopes of both anti-₅ and anti-₁ mAbs. The ₅(R470)₁(P455)-Fc integrin displayed a similar mAb reactivity profile, but the levels of binding were reduced.

₅(A855-D613)₁(P455)-Fc heterodimers bound FnIII-(6-10), and the interaction was increased in the presence of the stimulatory anti-₁ mAb 12G10 (Table IV). Consistent with the mAb data, FnIII-(6-10) binding to the ₅(R470)₁(P455)-Fc integrin was greatly reduced. The stimulatory anti-₁ mAb 12G10 was required to promote FnIII-(6-10) binding, suggesting that this integrin was locked in an inactive conformation.

The ₅ truncations Ser⁵⁹², Ser⁵⁷⁹, Ala⁵²⁷, Gly⁴²⁷, and Ser⁴⁰⁹ did not react with conformation-dependent anti-₅ mAbs or interact with FnIII-(6-10), even in the presence of 12G10; however, slight reactivity with anti-₁ mAbs was observed (Table IV). The ₅(S409) truncation demonstrated significantly greater reactivity to the anti-₁ mAbs than the Ser⁵⁹², Ser⁵⁷⁹, Ala⁵²⁷, and Gly⁴²⁷ truncations. Reactivity with the anti-human ₁ Fc antibody was observed for all integrins, demonstrating that they were expressed. These findings suggested that while the ₁(P455)-Fc subunit was correctly folded, the ₅-subunits were misfolded.

As a control, each of the ₅(TR)-Fc constructs and the ₁(D708)-Fc and ₁(P455)-Fc constructs were expressed alone. The results (Table V) showed that the ₅(TR)-Fc single chain proteins displayed reactivity to the anti-human ₁ Fc mAb, demonstrating that they were expressed; however, mAb reactivity was abrogated. Interestingly, expression of the ₁(D708)-Fc or ₁(P455)-Fc subunit alone generated a protein that bound all anti-₁ mAbs, but not anti-₅ mAbs, suggesting that the protein was correctly folded. It is therefore surprising that those ₅ truncations that did not show mAb reactivity did not react with the anti-₁ mAbs to a greater extent. Only the ₅(S409)₁(P455)-Fc integrin showed appreciable levels of anti-₁ mAb binding. This suggested the misfolded ₅ Ser⁵⁹², Ser⁵⁷⁹, Ala⁵²⁷, and Gly⁴²⁷ subunits obscured the anti-₁ mAb epitopes, whereas the ₅(S409) truncation did not obscure the ₁(P455) epitopes.

View this table: [in this window] [in a new window]   Table V mAb reactivity of single chain 5(TR)-Fc, 1(D708)-Fc, and 1(P455)-Fc proteins Integrin single chain Fc fusions included the entire 5 C-terminal truncation series constructs and the 1(D708)-Fc and 1(P455)-Fc subunit constructs expressed individually. Soluble single chain-Fc integrins were captured by Fc, and the binding of various anti-5 and anti-1 mAbs was assessed by ELISA. mAbs that reacted with the single chain Fc integrins are indicated by a +; those that did not react with the truncation are indicated by a . Anti-5 mAbs included mAb16 and JBS5, whereas anti-1 mAbs included 12G10, TS2/16, and mAb13; Fc is an anti-human 1 Fc mAb. The binding of FnIII-(6-10) was also tested, both alone and in the presence of the stimulatory anti-1 mAb 12G10 (1 µg/ml), + indicates binding occurred;  indicates the absence of binding.

A representative sample of the correctly folded and ligand-binding-competent ₅(TR)₁(P455)-Fc integrins were co-expressed and purified by protein A affinity chromatography. Purified proteins were resolved by reducing SDS-PAGE prior to transfer to nitrocellulose and analysis by Western blotting using an anti-human Fc antibody. The results (Fig. 5A) showed a band of 110 kDa common to all samples that corresponded to the ₁(P455)-Fc subunit. The size of the ₅-subunit decreased consistent with the truncation point (Table III). The ₅(R470)-Fc subunit was 110 kDa and therefore co-migrated with the ₁(P455)-Fc subunit. The expression of the ₅(R470)₁(P455)-Fc integrin was reduced, which may account for the decreased mAb and FnIII-(6-10) reactivity of this integrin. Most -subunit bands were quite broad, which appears to reflect heterogeneity in glycosylation. The ₅(D613) construct contains nine potential N-linked glycosylation sites, and the difference in apparent maximum molecular weight of the Fc-linked subunit estimated either by SDS-PAGE (~120 kDa; Fig. 5A) or primary sequence (~95 kDa) suggests that most sites can be derivatized. The heterogeneity was more apparent after removal of the Fc tag following Tev cleavage; here a number of distinct bands were observed that each differed in molecular mass by ~2.5 kDa, a value similar to that expected for a complex oligosaccharide side chain (data not shown).

View larger version (32K): [in this window] [in a new window]   Fig. 5.   Characterization of 5(TR)1 (P455)-Fc integrins. A, various 5 C-terminal truncations were all purified by protein A-Sepharose affinity chromatography and resolved by reducing SDS-PAGE prior to transfer to nitrocellulose and analysis by Western blotting. The 51-Fc integrins were detected using a peroxidase-conjugated anti-human 1 Fc antibody (0.5 µg/ml). Lane 1, 5(E951)-Fc; lane 2, 5(G795)-Fc; lane 3, 5(N694)-Fc; lane 4, 5(D613)-Fc; lane 5, 5(R470)-Fc were co-expressed with 1(P455)-Fc. B, the 5(D613)1(P455)-Fc integrin was dispensed into wells of a 96-well plate pre-coated with anti-human 1 Fc antibody (2.5 µg/ml). Anti-5 (mAb16, mAb11, JBS5) and anti-1 (12G10, TS2/16, 9EG7, mAb13) antibodies (1 µg/ml) were detected using appropriate HRP-conjugated secondary antibodies (1 µg/ml; hatched bars). The truncated 5(E951)1(P455)-Fc served as positive control (open bars). C, varying concentrations of biotinylated FnIII-(6-10) were dispensed into wells of a 96-well plate pre-coated with purified 5(D613)1(P455)-Fc. Bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis, the apparent affinity of FnIII-(6-10) binding was 1.81 ± 0.17 nM. d, varying concentrations of GRGDS or GRDGS peptide together with biotinylated FnIII-(6-10) (1 µg/ml) were dispensed into wells of a 96-well plate pre-coated with purified 5(D613)1(P455)-Fc. Bound ligand was detected using ExtrAvidin®-peroxidase. By non-linear regression analysis the concentration of peptide required for 50% inhibition (IC50) of FnIII-(6-10) binding was 3.59 ± 0.7 µg/ml.

To determine the apparent affinity of the ₅(D613)₁(P455)-Fc-FnIII-(6-10) interaction, protein A-purified ₅(D613)₁ (P455)-Fc was examined using varying concentrations of FnIII-(6-10) by ELISA.