The Structure of the β-Propeller Domain and C-terminal Region of the Integrin αM Subunit

The αM subunit of integrin Mac-1 contains several distinct regions in its extracellular segment. The N-terminal region has been predicted to fold into a β-propeller domain composed of seven β-sheets each about 60 amino acid residues long, with the I-domain inserted between β-sheets 2 and 3. The structure of the C-terminal region is unknown. We have used monoclonal antibodies (mAbs) as probes to study the dependence of the structure of different regions of the αM subunit on association with the β2 subunit in the αM/β2 heterodimer. All of the mAbs to the I-domain immunoprecipitated the unassociated αM precursor and reacted with the αM subunit expressed alone on the surface of COS cells. By contrast, four mAbs to the β-propeller domain did not react with the unassociated αM precursor nor with the uncomplexed αM subunit expressed on COS cell surface. The four mAbs were mapped to three subregions in three different β-sheets, making it unlikely that each recognized an interface between the α and β subunits. These results suggest that folding of different β-propeller subregions is coordinate and is dependent on association with the β2 subunit. The segment C-terminal to the β-propeller domain, residues 599–1092, was studied with nine mAbs. A subset of four mAbs that reacted with the αM/β2 complex but not with the unassociated αM subunit were mapped to one subregion, residues 718–759, and five other mAbs that recognized both the unassociated and the complexed αM subunit were localized to three other subregions, residues 599–679, 820–882, and 943–1047. This suggests that much of the region C-terminal to the β-propeller domain folds independently of association with the β2 subunit. Our data provide new insights into how different domains in the integrin α and β subunits may interact.

Although the overall structure of integrins is unknown, several structurally distinct domains in the extracellular portions of both ␣ and ␤ subunits have been predicted or identified. The N-terminal region of the integrin ␣ subunits contains seven repeats of about 60 amino acids each (8) and has recently been predicted to fold into a ␤-propeller domain that consists of seven ␤-sheets, with each ␤-sheet containing four anti-parallel ␤-strands (9). The leukocyte integrin ␣ subunits (10), the ␣1 (11) and ␣2 (12) subunits of the ␤1 subfamily, and the ␣E subunit (13) of the ␤7 subfamily contain an inserted domain or I-domain of about 200 amino acids that is predicted to be inserted between ␤-sheets 2 and 3 of the ␤-propeller domain (9). The three-dimensional structure of the I-domain from the Mac-1, LFA-1, and ␣2␤1 integrins has been solved and shows that it adopts the dinucleotide-binding fold with a unique divalent cation coordination site designated the metal ion-dependent adhesion site (14 -17). The integrin ␤ subunits contain a conserved domain of about 250 amino acids in the N-terminal portion. This domain has been predicted to have an "I-domainlike" fold (14,18,19). Very little is known about the structure of the C-terminal half of the extracellular portions of both ␣ and ␤ subunits. Electron microscopic images of integrins reveal that the N-terminal portions of the ␣ and ␤ subunits fold into a globular head that is connected to the membrane by two rodlike segments about 16 nm long corresponding to the C-terminal portions of the ␣ and ␤ extracellular domains (20 -22). This would suggest that the C-terminal portions of both subunits are quite extended.
Previous studies using mAbs 1 as probes have shown that the structure of specific domains in LFA-1 requires association of the ␣L and ␤2 subunits. mAbs to the ␤2 subunit conserved domain do not react with the unassociated ␤2 subunit, whereas mAbs to the regions preceding and following this domain do, indicating that the structure of the conserved domain is dependent on association with the ␣L subunit (23). mAbs to the I-domain react with the unassociated ␣L subunit (24). This finding together with the fact that the I-domain can be expressed as an isolated domain (14,16,25,26) show that the I-domain assumes a native structure independently of the ␤2 subunit. By contrast, two mAbs (S6F1 and TS2/4) mapped to the N-terminal region of the ␤-propeller domain, and one mAb (G-25.2) that maps to a region of 212 amino acids with 159 amino acids located in the ␤-propeller domain and the remainder in the C-terminal region, do not recognize the ␣L subunit in the absence of association with the ␤2 subunit (24). Another mAb (CBRLFA-1/1) that maps to a region overlapping the I-domain and ␤-propeller domain reacts weakly with the uncomplexed ␣L subunit. These results indicate that at least one region in the ␤-propeller domain is dependent on association with the ␤2 subunit for mAb reactivity, and it has been suggested that the most likely explanation is that folding of the ␤-propeller domain is not completed until after association with the ␤ subunit (24). Since mAbs specific for the region of the ␣L subunit C-terminal to the ␤-propeller domain have not been described, it is not known whether folding of this region is dependent on association with the ␤ subunit.
In this study, we have used mAb probes to study the structure of the Mac-1 ␣ subunit in the presence and absence of association with the ␤2 subunit. We have studied the ␤-propeller domain, the I-domain, and the extensive region C-terminal to the ␤-propeller domain. Compared with the previous studies on LFA-1, our studies on the ␤-propeller domain are more definitive, since mAb specificity is defined to individual amino acid substitutions between mouse and human, and mAb to epitopes that are widely separated in the predicted ␤-propeller structure all show a dependence on ␤ subunit association for reactivity. Furthermore, we employ a panel of mAbs that defines four different subregions within the C-terminal region of the ␣ subunit. The results show that epitopes in three of these regions have a native structure in the absence of ␤ subunit association, whereas a fourth epitope is dependent on association with the ␤ subunit. Thus, much of the C-terminal region of the ␣M subunit appears to assume a native fold independently of association with the ␤2 subunit.
DNA Constructs and Mutagenesis-The human wild-type ␣M subunit cDNA was subcloned in the expression vector pCDNA3.1ϩ (Invitrogen, Carlsbad, CA) as described. 3 For generating human-mouse ␣M chimeras, a SacII site was created immediately after the stop codon (nucleotides 3532-3534). By specifically primed reverse transcription of murine spleen mRNA (CLONTECH, Palo Alto, CA) from approximately 50 nucleotides downstream of the stop codon, the first strand of the mouse ␣M cDNA (33) was generated with Moloney murine leukemia virus reverse transcriptase (Stratagene, La Jolla, CA). By using this as a template for PCR, a 2-kilobase pair mouse ␣M cDNA fragment covering nucleotides from the SfiI site (nucleotide 1688) to the stop codon and having a SacII site immediately after the stop codon was made.
This mouse ␣M SfiI-SacII fragment was used to replace the corresponding human ␣M SfiI-SacII fragment to generate the initial chimeric ␣M cDNA encoding the N-terminal 529 residues of human sequence and the remaining C-terminal sequence from mouse. Using this initial chimeric construct as template, eight human-mouse ␣M chimeras with a variable mouse C-terminal portion were generated by overlap extension PCR (34,35). Briefly, outer primers for overlap PCR were just 5Ј to the SfiI site and 3Ј to the SacII site, and the first set of reactions was carried out using the human wild-type ␣M and the initial chimeric construct as templates. After the overlap extension reaction, the chimeric products were digested with SfiI and SacII, and the SfiI-SacII fragments were swapped into the human wild-type ␣M in vector pCDNA3.1ϩ. Human to mouse individual amino acid substitutions in the region from amino acids 718 -759 of human ␣M were made by overlap extension PCR (34,35). Briefly, the overlapping primers contained the desired mutations, and the outer primers were 5Ј to the SfiI site and 3Ј to the NdeI site, respectively. The overlap extension PCR products were digested with SfiI and NdeI and swapped into human wild-type ␣M in expression vector pEFpuro (36).
For mapping mAb epitopes in the ␤-propeller domain of the human ␣M subunit, 32 different chimeric ␣M subunits were made in which a short segment of mouse sequence comprising a predicted loop or a strand 4 was inserted in the human sequence. Mutagenesis was done by inverse PCR on plasmid pBluescript II containing Mac-1 ␣M cDNA fragments that included the NotI site 5Ј to the coding region and the BspEI site at amino acid residue 180 or included the BspEI-BbsI fragment from 180 to 672, as described elsewhere. 3 The mutated cDNA fragments were excised with NotI and BspEI or BspEI and BbsI and swapped into wild-type ␣M cDNA contained in plasmid pCDNA3.1ϩ. Mutants were named after the sheet (W) and the loop (L) or the strand (S) that was exchanged, e.g. hu(W7L3-4)mo has mouse sequence in the loop between strands 3 and 4 of W7, and hu(W1S4)mo contains mouse sequence in strand 4 of W1. In the following list, the amino acid segment or individual amino acid residue that was of murine origin is indicated for each mutant in the numbering system for the mature human ␣ subunit. These mutants are as follows: hu(W7L3-4)mo, 7-8; hu(W7L4 -1)mo, 16 All mutations were verified by DNA sequencing. At least two independent clones of each mutant were used for transfection, and identical results were obtained.
Transient Transfection-COS cells were transfected by the DEAEdextran method (36) with the ␣M cDNA alone or were co-transfected with the wild-type or chimeric ␣M and ␤2 cDNA. The wild-type and chimeric ␣M cDNA were in plasmid pCDNA3.1ϩ, and the ␤2 cDNA was contained in plasmid pEF-BOS (36). Three days after transfection, COS cells were detached with Hanks' balanced salt solution supplemented with 5 mM EDTA for flow cytometric analysis. 293 cells were transfected with the calcium phosphate method (37,38). Briefly, 7.5 g of wild-type or mutant ␣M cDNA in plasmid pEFpuro and 7.5 g of ␤2 cDNA in plasmid pEF-BOS were used to transfect one 6-cm plate of 70 -80% confluent cells. Two days after transfection, cells were detached with Hanks' balanced salt solution, 5 mM EDTA for flow cytometric analysis.
Flow Cytometry-COS cells and 293 cells were washed twice with L15 medium containing 2.5% FBS (L15/FBS) and resuspended to 1-2 ϫ 10 6 cells/ml in the same medium. 50 l of the cell suspension was incubated with an equal volume of the primary antibody (20 g/ml purified mAb, 1:100 dilution of mAb ascites, or 1:2 dilution of hybridoma supernatant in PBS) on ice for 30 min. Cells were then washed three times with L15/FBS and incubated with fluorescein isothiocyanate-conjugated goat anti-mouse IgG (heavy and light chain, Zymed Laboratories, San Francisco, CA) for 30 min on ice. For staining with mAb CBRM1/20 that requires Ca 2ϩ , 3 the primary and secondary antibodies were diluted in PBS supplemented with 1 mM Ca 2ϩ . After washing, cells were resuspended in cold PBS and analyzed on a FACScan (Becton Dickinson, San Jose, CA).
Radiolabeling, Immunoprecipitation, and Gel Electrophoresis-For metabolic labeling, U937 cells were plated in four 10-cm Petri dishes and induced with PMA for 3 days as described previously (39). Cells in each dish were washed twice with methionine-free RPMI 1640 medium and labeled with 0.625 mCi of [ 35 S]methionine in 5 ml of methioninefree RPMI 1640 containing 15% dialyzed FBS. After incubation at 37°C for 30 min, cells in two dishes were washed twice with cold PBS and lysed by addition of 3 ml of lysis buffer (1% Triton X-100, 20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 1 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, 0.24 TIU/ml aprotinin, and 10 g/ml each of pepstatin A, antipain, and leupeptin) and incubation for 30 min at 4°C with gentle agitation. For chase labeling, 5 ml of complete medium supplemented with 100 g/ml unlabeled methionine was added to each of the remaining dishes, and incubation at 37°C was continued for 16 h. The chase-labeled cells were lysed identically to pulse-labeled cells, and lysates were clarified by centrifugation at 12,000 rpm for 10 min at 4°C. For surface labeling, COS transfectants (2 ϫ 10 6 cells) were washed three times with PBS and resuspended in 1 ml of PBS. The cells were surface-labeled with 1 mCi of Na 125 I using two IODO-BEADS (Pierce) following the manufacturer's instructions. The labeled cells were washed three times with PBS containing 10% FBS and once with PBS and lysed as described above.
For immunoprecipitation, cell lysates were precleared by addition of 1/10 volume of recombinant protein G agarose (50% suspension in PBS) (Life Technologies, Inc.) and incubation at 4°C for 2-3 h with agitation. The precleared lysates were split into 250-l aliquots, and to each aliquot, 2.5 l of mAb ascites or 10 g of purified mAb or 250 l of mAb supernatant was added, and the final volume was adjusted to 500 l with lysis buffer. After incubation overnight at 4°C, followed by cen-trifugation at 12,000 rpm for 10 min at 4°C to remove protein aggregates, the antigen/antibody mixture was incubated with 50 l of protein G-agarose beads for 1.5-2 h at 4°C with agitation. Beads were washed three times with lysis buffer and once with lysis buffer without detergent. For immunoprecipitation with mAb CBRM1/20, lysis buffer and wash buffer were supplemented with 1 mM Ca 2ϩ . Bound proteins were eluted from beads with 50 l of Laemmli sample buffer by heating for 5 min at 100°C, and the immunoprecipitates were analyzed by 7.5% SDS-polyacrylamide gel electrophoresis (40). The gels were processed for fluorography for [ 35 S]methionine-labeled proteins or autoradiography for 125 I-labeled proteins.
Secondary Structure Prediction-The amino acid sequences between the ␤-propeller domain and the transmembrane segment of 36 integrin ␣ subunits (9) were aligned with ClustalW, and then the alignment was iteratively refined using default settings with PRRP and the Gonnet amino acid substitution matrix, and an evolutionary tree was prepared with PHYLP (41). The ␣M and ␣IIb subunits fall in different branches of this tree, each of which is well populated. One branch containing 11 subunits most closely related to human ␣M, i.e. murine ␣M, human ␣D, and ␣X, murine and human ␣L, human and rat ␣1, and bovine, human, and mouse ␣2, were realigned with one another using PRRP. They are 21-70%, x ϭ 34% identical to human ␣M. Another branch containing the 17 subunits most closely related to human ␣IIb, i.e. hamster, human, and mouse ␣3, human and Xenopus ␣5, chicken and human ␣6, mouse ␣7, chicken and human ␣8, and chicken, human, mouse, and Pleurodes ␣V, and YMA1 of Caenorhabditis elegans, were realigned in a separate group. They are 20 -38%, x ϭ 28% identical to human ␣IIb. The alignments in MSF format, with gaps in human ␣M and human ␣IIb removed to increase prediction accuracy, were separately submit- Immunoprecipitates were subjected to 7.5% SDS-polyacrylamide gel electrophoresis and fluorographed. The nonbinding mAb, X63, was used as negative control. CBRLFA-1/2 is specific for the ␤2 subunit, and all other mAbs are against the ␣M subunit (see Fig. 1). In lanes 20 -22, cell lysis and immunoprecipitation were carried out in the presence of 1 mM Ca 2ϩ .
ted for secondary structure prediction to PHD (42). 4 Smaller subgroups containing a higher degree of relationship to ␣M (6 ␣ subunits, with 27-70% identity to ␣M) or to ␣IIb (9 ␣ subunits, with 33-38% identity to ␣IIb) gave very similar predictions but with a slightly lower correlation between the ␣M and ␣IIb predictions.

mAbs to the ␤-Propeller Domain and a Subset of mAbs to the C-terminal Region Do Not React with the Unassociated ␣M
Subunit-To study whether folding of the ␣M subunit is dependent on association with the ␤2 subunit, we examined the expression of mAb epitopes on the unassociated ␣M subunit. Eighteen mAbs that have previously been mapped to different regions in the ␣M subunit were used (29) 2 (Fig. 1). Previous studies on leukocyte integrin biosynthesis have shown that the ␣ and ␤ subunit precursors are initially unassociated in the endoplasmic reticulum and that transport to the Golgi apparatus and processing from high mannose N-linked carbohydrates to complex carbohydrates are dependent on the formation of ␣ and ␤ complex (39,43,44). We therefore examined whether mAbs to the I-domain, to the ␤-propeller domain, and to the C-terminal region immunoprecipitated the unassociated ␣M precursor (␣ЈM). All mAbs immunoprecipitated the mature ␣M subunit with molecular size of about 170 kDa from the lysate of cells pulse-labeled with [ 35 S]methionine for 30 min and chased for 16 h (Fig. 2, lower panel). The ␣M subunit was complexed with the ␤2 subunit as shown by co-immunoprecipitation of the ␤2 subunit with the ␣M subunit. However, mAbs differentially precipitated the ␣ЈM precursor, which is slightly smaller than the mature ␣M subunit from the pulse-labeled cells (Fig. 2, upper panel). There was little or no ␣ЈM precursor associated with the ␤2 precursor (␤Ј2) in the pulse-labeled cells, since no detectable ␤Ј2 over background was co-precipitated by mAbs to the ␣M subunit, but ␤Ј2 was precipitated with mAb CBRLFA-1/2 to the ␤2 subunit (upper panel, lane 18). All mAbs to the I-domain precipitated ␣ЈM (upper panel, lanes 2-6). By contrast, three mAbs (CBRN1/6, CBRN3/4, and CBRM1/20) to the ␤-propeller domain did not precipitate ␣ЈM (upper panel, lanes 7, 8, and 21). mAb CBRM1/32 to the ␤-propeller domain did not precipitate the ␣M/␤2 complex or ␣ЈM from cell lysates (data not shown), suggesting that its epitope is sensitive to detergent extraction. Five mAbs (OKM1, CBRM1/10, CBRM1/23, CBRM1/25, and CBRM1/26) to the C-terminal region precipitated ␣ЈM (upper panel, lanes 9 and 10 and 14-16), whereas four other mAbs (CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30) precipitated no to very little ␣ЈM (upper panel, lanes 11-13 and 17). Thus, epitopes of mAbs to the I-domain are expressed on the unassociated ␣M precursor, whereas epitopes of ␤-propeller domain mAbs and a subset of mAbs to the C-terminal region are not.
To examine ␣ subunit structure independently of maturation events occurring during biosynthesis, we examined mAb reactivity with the unassociated ␣M subunit expressed on the surface of COS cells. COS cells were transfected with cDNA for ␣M alone or for both ␣M and ␤2 subunits, and mAb reactivity with the uncomplexed ␣M or the ␣M/␤2 complex expressed on the surface of COS transfectants was determined by immunofluorescent flow cytometry (Fig. 3). All mAbs to the I-domain (OKM9, TGM-65, CBRM1/1, CBRM1/2, and CBRM1/29) reacted with the unassociated ␣M subunit as well as with the The data obtained by immunofluorescent flow cytometry were confirmed by immunoprecipitation. COS transfectants expressing the ␣M subunit alone or the ␣M/␤2 complex were surface-iodinated, and the labeled proteins were immunoprecipitated from cell lysates. All mAbs precipitated the ␣M/␤2 complex from COS cells co-transfected with ␣M and the ␤2 (Fig. 4, lower panel). mAbs to the I-domain precipitated the ␣M subunit expressed alone on the COS cell surface (upper panel, lanes [3][4][5]. By contrast, ␤-propeller domain mAbs (upper panel, lanes 6, 7 and 15) and a subset of mAbs to the C-terminal region (upper panel, lanes 10, 11 and 14) failed to precipitate the uncomplexed ␣M subunit. Thus, expression of epitopes of the ␤-propeller domain mAbs and a subset of mAbs to the Cterminal region is dependent on ␣M and ␤2 heterodimer formation.
To test the possibility that the ␤2 subunit may directly contribute to the epitopes of the mAbs that did not react with ␣M in the absence of the ␤2 subunit, we expressed human ␣M in association with the mouse ␤2 subunit or the chicken ␤2 subunit on the surface of COS cells and human 293 cells. mAb reactivity with the transfectants was determined by immunofluorescent flow cytometry. mAbs CBRN1/6, CBRN3/4, CBRM1/20, CBRM1/32, CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30 that did not react with the unassociated ␣M subunit reacted with the human ␣M/mouse ␤2 and human ␣M/chicken ␤2 complexes as well as with the human ␣M/ human ␤2 complex (data not shown). These results suggest that the ␤2 subunit does not directly contribute to the epitopes of these mAbs.
Epitope Mapping of mAbs to the C-terminal Region and to the ␤-Propeller Domain of the ␣M Subunit-The finding that a subset of mAbs to the C-terminal region does not react with the unassociated ␣M subunit suggests that the structures of certain subregion(s) in this C-terminal 493-amino acid segment may be dependent on association with ␤2. To localize such subregion(s), as well as subregion(s) that fold independently of ␤2 association, epitopes of the nine mAbs to the C-terminal region were mapped using human-mouse ␣M chimeras. The chimeras were generated by progressively replacing the human sequences from the C terminus with the corresponding sequences from mouse ␣M (Fig. 5) and were co-expressed with human ␤2 in COS cells. mAb reactivity with chimeric ␣M/␤2 was determined by immunofluorescent flow cytometry (Table  I). All chimeras were expressed on the surface in association with the human ␤2 subunit, with levels of cell-surface chimeric ␣M/␤2 complex comparable with that of wild-type ␣M/␤2 complex. In addition, all chimeric ␣M/␤2 complexes were stained with mAbs to the ␤-propeller domain (Table I and Table I and Fig. 6. A 41-amino acid sequence (residues 718 -759) was required for epitopes of the four mAbs (CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30) that did not react with the unassociated ␣M subunit. The epitopes of five mAbs that reacted with the unassociated ␣M subunit were mapped to three other subregions as follows: OKM1 to a region immediately following the ␤-propeller domain (residues 599 -679); CBRM1/10, CBRM1/25, and CBRM1/26 to a region from residues 820 to 882; and CBRM1/23 to a region from residues 943 to 1047. Thus, mAb epitopes that map to one subregion (residues 718 -759) require association of ␣M with ␤2, whereas epitopes localized in three other subregions (residues 599 -679, 820 -882, and 943-1047) are independent of the ␤2 subunit.
The region from residues 718 to 759 contains eight amino acid differences between the human and mouse sequences (Fig.  7). To identify individual amino acid residues in this region that are required for epitopes of mAbs CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30, single or double amino acid residues in the human ␣M sequence were replaced with corresponding residues from mouse ␣M. The mutants were co-ex- pressed with the ␤2 subunit in 293 cells, and mAb reactivity was determined by immunofluorescent flow cytometry (Table  II). Substitution of Thr 725 to Glu (mutant T725E) completely abolished binding of all four mAbs, whereas substitution of Ser 728 and Ala 729 to Arg and Ser, respectively (mutant S728R/ A729S), completely abrogated binding of mAbs CBRM1/16, CBRM1/17, and CBRM1/30 and decreased binding of CBRM1/ 18. All other substitutions did not affect binding of all four mAbs. Thus, mAbs CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30 recognize overlapping epitopes.
To map mAb epitopes in the ␤-propeller domain, 32 different ␣M chimeras were constructed, in which a short segment of mouse sequence comprising a predicted loop or a strand was inserted in the human sequence, and mAbs were tested for reactivity with the ␣M chimeras co-expressed with the human ␤2 subunit in COS cells. Substitution of the 2-3 loop of W6 (mutant hu(W6L2-3)mo) or mutation of Arg 534 to Gln in this loop (mutant R534Q) completely abolished binding of mAb CBRM1/32, whereas binding of three other ␤-propeller domain mAbs (CBRN1/6, CBRN3/4, and CBRM1/20) and the I-domain mAb CBRM1/29 to these two mutants was not affected (Table  III). Arg 534 is predicted to be on the upper, outer edge of the ␤-propeller domain in W6 (Fig. 8). Substitution of the loop 3-4 of W4 (mutant hu(W4L3-4)mo) completely abrogated mAb CBRN1/6 binding and decreased CBRN3/4 binding. This sub-stitution did not affect CBRM1/32 and CBRM1/20 binding. The three human residues substituted in this mutant, Gln 421 , Thr 423 , and Met 425 , are predicted to be in the lower outer edge of the ␤-propeller domain in W4 (Fig. 8). All other substitutions in 30 different segments of the ␤-propeller domain had no effect on binding of mAbs CBRM1/32, CBRN1/6, and CBRN3/4. mAb CBRM1/20 was mapped to five amino acid residues in the 1-2 loop of W5 and the 3-4 loop of W6 3 on the bottom of the ␤-propeller domain (Fig. 8). DISCUSSION By using mAbs as probes, we have examined the structure of different regions in the Mac-1 ␣M subunit during biosynthesis and ␣M/␤2 heterodimer assembly and after expression on the cell surface. All five different mAbs to the I-domain reacted with the unassociated ␣M subunit, confirming that the folding of the I-domain does not require the ␤2 subunit. By contrast, four mAbs (CBRN1/6, CBRN3/4, CBRM1/20, and CBRM1/32) that map to three different subregions in the ␤-propeller domain did not react with the unassociated ␣M subunit (Fig. 6). CBRN1/6 and CBRN3/4 mapped to one or more of three residues in the 3-4 loop of W4 (residues 421-425) (Fig. 8). CBRM1/20 is specific for three amino acid residues in the 1-2 loop of W5 and two residues in the 3-4 loop of W6 3 (Fig. 8). The epitope for CBRM1/20 includes two residues, Asn 453 and The human (hu) wild-type or human-mouse chimeric ␣M subunit was expressed in association with human ␤2 on the surface of transfected COS cells. mAb reactivity was determined by immunofluorescence flow cytometry. ϩ, positive staining with mean fluorescence intensity comparable to human wild-type ␣M/␤2 stained with the same mAb; Ϫ, staining was not significantly different from mock-transfected cells stained with the same mAb; ND, not determined. Epitopes of CBRM1/1 and CBRM1/32 were previously mapped (29). Of note, TS1/18 and CBRM1/32 are specific for complexed ␤2 and ␣M, respectively.
FIG. 6. mAb reactivity with the ␣M and ␣L subunits in the absence of the ␤2 subunit. Schematic diagrams of the ␣M and ␣L subunits are shown. W1 to W7 are ␤-sheets 1-7 of the ␤-propeller domain. Numbers are positions of amino acid residues at the boundaries between domains and between subregions in the ␣M C-terminal region. mAbs, except for those to the I-domain, and their epitope localization are shown under ␣M and ␣L. ϩ indicates mAbs that react with the ␣M or ␣L in the absence of the ␤2 subunit; Ϫ indicates mAbs that do not react with the ␣M or ␣L in the absence of the ␤2 subunit; Ϫ/ϩ refers to weak reactivity with the unassociated ␣ subunit compared with mAbs to the I-domain. All tested mAbs to the I-domain react with the unassociated ␣ subunits. The original data on the ␣L subunit was reported elsewhere (24).
Asp 457 , that are predicted to coordinate with Ca 2ϩ in the 1-2 loop of W5, and binding of this mAb requires Ca 2ϩ with an EC 50 of 0.2 mM. 3 These mAbs did not immunoprecipitate the unassociated ␣M precursor or react with the ␣M subunit expressed alone on the surface of COS cells. mAb CBRM1/32 reacted with the ␣M/␤2 complex expressed on the cell surface but did not react with the ␣M subunit expressed alone on the cell surface. The epitope of CBRM1/32 requires residue Arg-534 in the 2-3 loop of W6 (Fig. 8). One possible interpretation of our results is that all three epitopes in the ␤-propeller domain require the presence of the ␤ subunit because the ␣ and ␤ subunits associate with one another in each of these regions, and each antibody binding site includes contacts with both the ␤ subunit and ␣ subunit. If so, the contacts with the ␤ subunit do not include any antigenic residues, because all mAb reacted equally well whether the human or murine ␤ subunit was associated with human ␣M. Furthermore, we tested the chicken ␤2 subunit, because 35% of the residues in the human and chicken ␤2 subunits differ, as opposed to only 18% between the human and the mouse (45). Amino acid differences between FIG. 7. Sequence alignment and secondary structure prediction of the C-terminal regions of representative integrin ␣ subunits. The sequence alignments are condensed from a master alignment of 36 integrin ␣ subunits ("Materials and Methods"). Cysteines known to be disulfide bonded in ␣IIb (48) are connected by solid lines; one disulfide that may differ in ␣M (see text) is shown with a dashed line. The arrow points at the main chymotryptic (CT) cleavage site (around Asn 570 ) in ␣IIb (47,48,50). The region preceding the chymotryptic cleavage site and following the ␤-propeller domain that contains the OKM1 epitope is predicted to fold into a domain (see text). The regions to which mAb epitopes localize are shown above the human ␣M sequence. Two different sequence alignments containing separate branches of the integrin ␣ subunit evolutionary tree were used to predict the secondary structure of human ␣M and ␣IIb with PHD (42). These predictions are independent of one another (see "Materials and Methods"). E, ␤-sheet; H, ␣-helix. species are preferentially found on the surface of proteins rather than buried. Although a substantial portion of surface residues are expected to differ on the chicken and human ␤2 subunits, whether the chicken or human ␤2 subunit was present did not affect mAb reactivity. The epitopes that were localized include some that are quite distant. The Arg-534 residue recognized by CBRM1/32 mAb is on the upper surface of the ␤-propeller, whereas residues recognized by the CBRM1/20 and the CBRN1/6 mAb are on the lower surface and point in opposite directions from one another. The C-␣ carbon of the Arg-534 residue is predicted to be 30 Ϯ 3 Å and 41 Ϯ 3 Å distant from residues recognized by the CBRM1/20 and the CBRN1/6 mAb, respectively, and the C-␣ carbons of residues recognized by the CBRM1/20 and the CBRN1/6 mAb are 23 Ϯ 7 Å distant from one another. The probability that three out of three different epitopes would include surfaces from both the ␣ and ␤ subunits, even though some epitopes are quite distant from one another, would appear to be low. Because of this, and the lack of effect of the species origin of the ␤ subunit on mAb reactivity, we favor the interpretation that association between the ␣ subunit and ␤ subunit is required for the ␤-propeller domain to assume its final three-dimensional structure, i.e. to assume the correct fold. Our data are consistent with the idea that there is an interface between the ␣ subunit ␤-propeller domain and the ␤ subunit, although we believe that the interface is not necessarily associated with any of the epitopes we have mapped. Conversely, a number of mAb to different epitopes in the conserved domain of the integrin ␤ subunit are not reactive in the absence of the ␣ subunit (23). Thus, the conserved domain of the ␤ subunit is a candidate for association with the putative ␤-propeller domain of the ␣ subunit. Analogously, the G-protein ␤ subunit ␤-propeller domain is not properly folded in the absence of association with the G-protein ␥ subunit (46).
A previous study on the LFA-1 ␤-propeller domain used two mAbs (S6F1 and TS2/4) that map to the ␣L subunit N-terminal 57 amino acids, i.e. to part of ␤-sheets W7 and W1, and one mAb (G-25.2) that maps to a 212-amino acid region spanning W5-7 of the ␤-propeller domain and part of the C-terminal TABLE II mAb reactivity with human ␣M subunit mutants carrying human-to-mouse substitutions in the region from residues 718 to 759 293 cells were transiently co-transfected with cDNAs for human ␤2 and the wild-type or mutated human ␣M subunit containing human-tomouse single or double amino acid residue substitution. mAb binding to the transfected cells was determined by immunofluorescence flow cytometry. ϩϩϩ, binding comparable to human wild type; ϩ, binding decreased to less than 30% of human wild type; and Ϫ, binding completely abolished.  III mAb reactivity with human ␣M subunit mutants containing human-to-mouse substitutions in the ␤-propeller domain 32 different chimeric ␣M subunits were made in which a short segment of mouse sequence comprising a predicted loop was inserted in the human sequence (see "Materials and Methods"). Mutants were named after the sheet (W) and the loop (L) that was exchanged, e.g. hu(W4L3-4)mo has mouse sequence in the loop between strands 3 and 4 of W4. The wild-type or mutated human ␣M subunit was transiently co-expressed in COS cells with human ␤2. mAb binding to the transfected cells was determined by immunofluorescence flow cytometry. ϩϩϩ, binding comparable to human wild type; ϩ, binding decreased to less than 30% of human wild type; and Ϫ, binding completely abolished. Only the mutants that affected binding of mAbs CBRM1/32, CBRN1/6 and CBRN3/4 were listed in the FIG. 8. mAb epitope localization in the ␤-propeller domain of the ␣M subunit. This stereoview of the side of the putative ␤-propeller domain with the upper surface on the top shows a C-␣ trace, with all atoms shown for residues involved in antigenic epitopes. They are R534 for CBRM1/32 (magenta), V450, N453, D457, G545, and G547 for CBRM1/20 (red), and Q421, T423, and M425 for CBRN1/6 and CBRN3/4 (purple). All of the latter three residues were substituted together, and the epitope may require only a subset of these three residues. The ␤-sheets (W) are shown in different colors, with W1 in cyan, W2 in orange, W3 in yellow, W4 in olive, W5 in green, W6 in aquamarine, and W7 in turquoise. Ca 2ϩ ions are gray spheres. The ␤-propeller domain of the Mac-1 ␣M subunit was modeled using the G-protein ␤ subunit ␤-propeller domain as template, as described. 3 This figure was made with Look of GeneMine TM (Molecular Applications Group, CA). region. These mAbs did not react with the unassociated ␣L subunit (24) (Fig. 6). Another mAb (CBRLFA-1/1) that overlaps the I-domain and W3 of the ␤-propeller domain showed weak reactivity in the absence of the ␤2 subunit. It is not known whether this mAb recognizes a boundary region between the I and ␤-propeller domains. Taken together, the findings on LFA-1 and Mac-1 demonstrate that multiple mAbs to different regions in the ␤-propeller domain do not react with the ␣ subunit in the absence of the ␤ subunit and suggest that the ␤-propeller domain folds as a unit and that this folding depends on association with the ␤ subunit.
mAbs to the C-terminal region of the ␣M extracellular domain differentially reacted with the unassociated ␣M subunit. Five mAbs (OKM1, CBRM1/10, CBRM1/25, CBRM1/26, and CBRM1/23) reacted with both the unassociated and the complexed ␣M subunit and were mapped to three subregions. OKM1 mapped to a subregion immediately following the ␤-propeller domain, residues 599 -679. CBRM1/10, CBRM1/25, and CBRM1/26 mapped to amino acids 820 -882, and CBRM1/23 mapped to residues 943-1047. Within each of these subregions, there are multiple differences between the mouse and human sequences (Fig. 7). Whether the multiple mAbs that react with residues 820 -882 recognize one or more epitopes within this subregion is not known. Minimally, these data show that three epitopes in three different subregions of the C-terminal segment are independent of the ␤2 subunit. By contrast, four other mAbs to the C-terminal region (CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30) only reacted with the ␣M/␤2 complex. These mAbs did not react with the unassociated ␣M precursor or with the uncomplexed ␣M subunit expressed on the COS cell surface. All four mAbs were mapped to residues Thr 725 and, additionally, Ser 728 and/or Ala 729 . Although CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30 did not react with the unassociated ␣M subunit, they reacted with the human ␣M/mouse ␤2 and human ␣M/chicken ␤2 complexes as well as with the human ␣M/human ␤2 complex (data not shown). Thus, association with the ␤2 subunit may be required for this region to assume its final structure. Although we believe that the interpretation that the ␣ and ␤ subunits both contribute to the antibody-binding site is less likely, either interpretation shows an important interaction with the ␤ subunit for the region of residues 725-729. Overall, the results show that three out of four epitopes in the C-terminal region of Mac-1 ␣M subunit are intact in the absence of association with the ␤2 subunit. If these results are representative of the Cterminal region as a whole, our data would suggest that much of this region folds independently of the ␤2 subunit. This is in marked contrast to the ␤-propeller domain.
To place our results on the C-terminal region within a structural framework, we predicted its secondary structure using the PHD program (42) (Fig. 7). By using a phylogenetic tree based on an iteratively refined alignment (41) of 36 ␣ subunit C-terminal region sequences, two subfamilies were identified. These subfamilies were large and contained members that were 1) sufficiently similar to one another to allow accurate alignment and to not be too divergent in tertiary structure, and 2) were sufficiently different from one another to contain a large amount of sequence information, and hence optimize prediction accuracy (42). An alignment of 11 subunits was used to predict the secondary structure of human ␣M, and an alignment of 17 other ␣ subunits was used to predict the structure of human ␣IIb (Fig. 7). Since no sequences were shared between the two alignments, and between the two groups there is only 16 -21% sequence identity, the predictions for ␣M and ␣IIb are largely independent of one another.
In the C-terminal segment, a total of 30 -34 ␤-strands were predicted. Of these, 22 were independently predicted in both ␣M and ␣IIb. Only 5 ␣-helices were predicted, and in each case these were predicted in only one of the two ␣ subunits. Thus, the C-terminal region is predicted to form domains of the all ␤ class. In this respect, it is similar to the ␤-propeller domain (9) and different from the I-domain which is of the ␣/␤ class (14,16). The disulfide bond topology of ␣IIb has been chemically determined (47,48). The conservation of cysteines suggest that 5 of 6 disulfide bonds are conserved in human ␣M, whereas one differs (Fig. 7). The first disulfide in this region, ␣IIb C473-C484, is confirmed by the sequence alignment of 36 integrin ␣ subunits, since these two cysteines are selectively absent in the chicken ␣6 subunit, and the cysteines and the loop in between them are absent in ␣2 subunits and ␣E subunits. The cysteines corresponding to the last disulfide bond in ␣IIb, Cys 885 -Cys 890 , are missing from ␣L subunits. Otherwise, there is only one predicted difference between disulfide bonds in ␣IIb and the leukocyte integrin ␣ subunits. The cysteine corresponding to ␣IIb Cys 484 is missing in all leukocyte integrin ␣ subunits, and all leukocyte integrin ␣ subunits contain a cysteine with no equivalent residue in ␣IIb, i.e. Cys 706 in ␣M. We predict that the cysteines at ␣IIb position 473, although aligned by sequence, are non-equivalent, i.e. that the cysteine in ␣M is involved in a different disulfide bond, to Cys 706 (dashed line in Fig. 7).
Folds of the all-␤ class as a general rule contain anti-parallel ␤-sheets (49). The vast majority but not all of the predicted ␤-strands in the ␣IIb and ␣M C-terminal regions are markedly amphipathic with alternating hydrophobic and hydrophilic residues. We therefore predict that the C-terminal region folds into 2-layer, anti-parallel ␤-sheet structures, i.e. ␤-sandwich or ␤-barrel domains of which the Ig fold is one of many representatives. The total length of the C-terminal region of about 500 residues, the number of predicted ␤-strands, and the overall number and location of disulfide bonds are appropriate for approximately four to six ␤-sandwich domains.
In ␣IIb, a main chymotryptic cleavage site is located around Asn 570 (47,48). Cleavage of cell-surface ␣IIb␤3 releases a ligand binding complex containing an N-terminal fragment of ␣II␤ of 55 kDa ending at approximately Asn 570 , and an 85-kDa N-terminal fragment of ␤3 (50). This suggests that the region around Asn 570 is well exposed and may represent a domain boundary region. It is interesting that four mAbs dependent on ␤ subunit association map to essentially the same site in ␣M (Fig. 7). The region preceding the chymotryptic cleavage site and following the ␤-propeller domain in ␣IIb contains one long range disulfide bond (Cys 490 -Cys 545 ), and six predicted ␤-strands. In ␣M, the corresponding region contains two predicted long range disulfide bonds (Cys 639 -Cys 696 , Cys 623 -Cys 706 ), and seven predicted ␤-strands. Based on these features, we predict that this region of about 120 amino acids following the ␤-propeller domain, residues 599 to about 718 for ␣M, and 450 to about 570 for ␣IIb, folds into a structurally independent domain. Consistent with this prediction, this region in ␣M appears to fold independently of association with the ␤ subunit, as shown with the OKM1 mAb. This contrasts with the flanking N-terminal ␤-propeller domain and the flanking C-terminal region from residues 725 to 729, to which mAbs CBRM1/16, CBRM1/17, CBRM1/18, and CBRM1/30 map.
Our results together with other recent studies provide new insight into how different domains in the integrin ␣ and ␤ subunits may associate. The I-domain is predicted to be connected to the upper surface of the ␤-propeller domain (9). The ␣ subunit ␤-propeller domain and the ␤ subunit conserved domain may associate, since both are dependent on ␣ and ␤ subunit association for folding (23,24) (this study). The predicted ␤-sandwich/␤-barrel domain that follows the ␤-propeller domain and contains the OKM1 epitope, residues 599 -718, is connected to the C terminus of strand 3 of W7 of the predicted ␤-propeller domain and hence to the bottom of the ␤-propeller domain. The following subregion of the ␣M subunit, from residues 725 to 729, may directly associate with the ␤2 subunit, or its structure may be indirectly dependent on associations elsewhere with the ␤ subunit. Other subregions in the C-terminal portions of ␣ and ␤ subunits might also participate in ␣ and ␤ subunit association as proposed for the ␣IIb␤ 3 integrin (32, 51), while retaining similar conformations in the unassociated and complexed forms.
In summary, the results from this study suggest that proper folding of the ␤-propeller domain of the integrin ␣M subunit requires association with the ␤2 subunit, whereas the I-domain folds independently of the ␤2 subunit. Much of the region C-terminal to the ␤-propeller domain folds prior to ␤ subunit association. Our results further advance the understanding of integrin structure and provide information that will be useful in guiding studies leading to the characterization of integrin three-dimensional structure.