A study of the structure, function and distribution of β₅ integrins using novel anti-β₅ monoclonal antibodies

Cell-surface adhesion receptors in the integrin family play a critical role in conveying signals from the extracellular environment that can lead to specific alterations in cell behavior. The integrin family of cell-surface adhesion receptors is composed of heterodimeric noncovalently asso-ciated α and β subunits, which are both transmembrane glycoproteins. The variety of different integrin heterodimers (20 so far reported) facilitate cell adhesion during diverse biological processes such as development, inflammation, wound repair, and hemostasis (Hemler, 1990; Hynes, 1992; Larson and Springer, 1990; Plow and Ginsberg, 1989).

The integrin β₅ subunit is one of several (including β₁, β₃, β₆ and β₈) that associate with the β_v subunit to form heterodimers, which may have different functional activi-ties. However, precise elucidation of the functional activity of α_v β₅ and other α_v-containing heterodimers has been difficult, primarily due to the simultaneous expression of multiple subunits on cells. For example, a structure termed α_v β_X was described, which appeared to mediate cell adhe-sion to both vitronectin and fibronectin (Cheresh et al., 1989), but subsequent reports suggested that α_v β_X could be a mixture of α_v β₅, a vitronectin receptor (Busk et al., 1992; Smith et al., 1990); and α_v β₆, a fibronectin receptor (Busk et al., 1992). Functional comparisons of α_v β₃ and α_v β₅ indicated that, although these two integrins both bind to vit-ronectin, only the former integrin was able to localize to focal adhesions and mediate cell spreading and migration on vitronectin (Leavesley et al., 1992; Wayner et al., 1991). The β₅ subunit has been cloned through the use of oligonucleotide probes derived from conserved regions of the β₁, β₂ and β₃ sequences (McLean et al., 1990; Ramaswamy and Hemler, 1990; Suzuki et al., 1990), and its primary sequence was found to most resemble (with 43–55% similarity) those of α₁, β₃ and β₆ (Argraves et al., 1987; Fitzgerald et al., 1987; McLean et al., 1990; Ramaswamy and Hemler, 1990; Rosa et al., 1988; Sheppard et al., 1990; Suzuki et al., 1990).

The α_v β₅ heterodimer was initially identified at the protein level, by use of a monoclonal antibody (mAb) to β_v (Cheresh et al., 1989; Smith et al., 1990), or polyclonal anti-serum to the unique cytoplasmic domain of β₅ (Ramaswamy and Hemler, 1990). More recently, mono-clonal antibodies directed against the α_v β₅ complex have been utilized for further biochemical and functional characterization of that integrin (Leavesley et al., 1992; Wayner et al., 1991). However, mAbs directly recognizing the β₅ subunit alone have not been described, thus leaving untested the possibility that β₅ function or distribution could extend beyond the α_v β₅ complex.

In the present study, we have generated a panel of mon-oclonal antibodies directly recognizing the integrin β₅ subunit, including mAbs that stimulate, inhibit or have no effect on adhesion function. These have allowed us to carry out a detailed analysis of the structure, function and distri-bution of β₅.

Human B cell lines were grown and maintained in RPMI with 10% fetal bovine serum (FBS). Human lung carcinoma cell line A549, and other adherent cell lines, listed in Table 1 (obtained from the American Type Culture Collection, ATCC), were also grown as monolayers in RPMI with 10% FBS, and were routinely passaged using PBS containing 2.5 mM EDTA. Human umbili-cal vein endothelial cells (HUVECS) were a gift from Dr Paul Anderson’s laboratory (Dana Farber Cancer Institute, Boston, MA, USA) and were also kept in RPMI with 10% FCS. CHO dhfr⁻ cells were grown in MEM-α⁺ with 10% FBS, and switched to MEM-α⁻ with 10% dialyzed FBS (Hazleton) after transfection. Methotrexate (Aldrich) was used for selective amplification, as described (Giancotti and Ruoslahti, 1990). Hybridomas were kept in RPMI with 10% CPSR-3 (Sigma, St. Louis, MO, USA) plus Hybrimax enhancing supplement (Sigma). Cloning was performed in the presence of MRC-5 human lung fibroblast feeder layers (ATCC). For transfection, CHO cells were washed once in PBS/EDTA, twice in MEM-α⁺ and resuspended in serum-free MEM-α⁺ media with purified plasmid DNA (1 μ,g p901 dhfr⁺ and 10 μ,g ₅-pECE). The dhfr⁺ p901 vector was kindly provided by Dr M. Rosa (Biogen, Cambridge, MA, USA). The PECE vector (Giancotti and Ruoslahti, 1990) was prepared by digestion with EcoRI to remove the β₁ insert present in the original construct provided by Dr E. Ruoslahti (Center for Cancer Research, La Jolla, CA, USA). Full-length β₅ cDNA was ligated directly into PECE using EcoRI sites. After pulsing with a Bio-Rad Gene Pulser (set at 280 V and 960 μF), cells were resuspended in MEM-α⁺ medium, with 10% FCS. After 5 days, the growth medium was switched to MEM-α⁻. After 8–10 more days, cells were sorted, based on staining with the B5-IIID8 monoclonal antibody against β₅, and the 5% most positive population was grown in the presence of methotrexate to amplify ₅ expression. Several additional rounds of amplification and sorting were then carried out until human β₅ expression was at least as high as that of hamster β₅.

RBF/DnJ hyperimmune mice (Jackson Labs) were immunized 3 times at 7-day intervals using 10⁸ A549 cells per immunization. Cells were injected into five sites, including axillas and the intraperitoneal cavity. Serum collected from tail veins was tested by immunoprecipitaton of surface ¹²⁵I-labeled A549 cells. All animal work was performed using the anesthetic Metofane (Pitman Moore, IL, USA). Splenocytes were fused with the myeloma partner P3X (Kearney et al., 1979) at a 1:1 ratio, using PEG 1500 (Boehringer Mannheim). Clones were observed after growth in HAT medium for 5 days.

Monoclonal antibodies utilized were: anti-β₁; A1A5 (Hemler et al., 1983); anti-α_v; LM142 (Cheresh and Harper, 1987); P3G8 (Wayner et al., 1991) and 13C2 (Davies et al., 1989); anti-α_v β ₅ complex; P3G2 (Wayner et al., 1991). Anti-β ₅ cytoplasmic domain rabbit polyclonal serum was obtained as described (Ramaswamy and Hemler, 1990). Both fibronectin (Fn) and vit-ronectin (Vn) were purchased from Telios Pharmaceuticals (La Jolla, CA, USA).

Cells were incubated in PBS with 1% BSA (Sigma) and 5% human serum (HS, Gibco Co.) for 30 minutes on ice, then washed 3 times in the same solution. Subsequently, 2×10⁵ to 3×10⁵ cell samples were incubated individually with antibodies (undiluted hybridoma supernatants or ascites diluted 1:100). Cells were then washed 3 times in PBS with 1% BSA and 1% HS, and treated with goat anti-mouse IgG coupled to fluorescein (Cappel) for 45 min on ice. After 3 washes they were resuspended in PBS with 1% BSA and analyzed utilizing a FACScan flow-cytometer (Becton Dickinson).

Cells were surface-labeled with ¹²⁵I using lactoperoxidase and lysed in the presence of 1% NP-40 (Calbiochem) and protease inhibitors (PMSF, aprotinin, leupeptin; Sigma). Immunoprecipita-tions were performed using Protein A-Sepharose beads (Pharmacia), previously blocked with unlabeled extracts to prevent non-specific binding of radiolabeled material. Labeled extracts were pre-cleared by incubation with normal mouse or rabbit serum, followed by Staphylococcus aureus Cowan 1 (Pansorbin, Cal-biochem). Immunoprecipitations were performed using mono-clonal antibodies diluted 1:100 (ascites) or undiluted hybridoma culture supernatants, with 3×10⁶ c.p.m. of radiolabeled cell extract per sample. Rat anti-mouse IgG mAb 187.1 (Yelton et al., 1981) was added to optimize binding to Protein A-Sepharose. Incuba-tions lasted from 30 minutes to overnight, then samples were washed 3 times in 1% NP-40 and analyzed by SDS-PAGE using 6–8% acrylamide gels, under non-reducing conditions unless otherwise stated. Based on molecular weight markers, the β₅ subunit routinely migrated as a band of approx. 85,000–90,000 (non-reduced) or 90,000-95,000(reduced), and α_v was approx. 150,000 (non-reduced) or approx. 120,000 (reduced).

Cells were washed twice with cold TBS (137 mM NaCl, 20 mM Tris, pH 8.0) and samples were prepared for western blots as described (Towbin et al., 1979). Transfer of protein to nitrocellu-lose was confirmed by Ponceau S staining (Sigma), and then filters containing fractionated extracts were probed with anti-₅ hybridoma supernatants, diluted 1:2.

Normal thymus was obtained from children (aged from 6 months to 11 years) undergoing cardiac surgery. Skin and kidney were kindly provided by Dr Atul K. Bhan (Department of Pathology, Massachusetts General Hospital). Cryostatic sections (4 μm) of human thymus, skin and kidney were air dried and kept overnight at room temperature before staining. Sections were first fixed with acetone for 10 min then immediately blocked with 1% horse serum for 30 min. The staining steps were performed in a ‘moist’ cham-ber at room temperature. For the streptavidin-biotin immunoper-oxidase staining assay, sections were sequentially stained with anti-β₅ mAb for 60 min, then with a biotinylated horse anti-mouse antibody (1:240 dilution) (Vector Laboratories, Burlingame, CA, USA) for 35 min, and finally with avidin-biotinylated horserad-ish peroxidase complex (1:100 dilution; Vector Laboratories) for 45 min. After staining, the sections were washed 3 times with PBS. Visualization of the peroxidase activity was performed by a 10 min incubation with 3-amino-9-ethylcarbazole (Sigma) in sodium acetate buffer, pH 5.0. Sections were counterstained with Gill no.1 hematoxylin (Sigma).

Protein ligands were coated onto 96-well microtiter plates (Flow Labs) at 1 μg/ml. Following overnight incubation at 4°C, 1% BSA was added to block non-specific adhesion. Cells were labeled by incubation with the fluorescent dye BCECF for 20 min. Then, after washes with PBS, 5×10⁴ of the labeled cells were added in triplicate to plates, in the presence or absence of antibodies (1 μg/ml or 50 μl of hybridoma supernatant) and incubated for 20 minutes at 37°C, followed by 3 washes with RPMI/0.1% BSA. Cells remaining attached to the plate were analyzed using a Flu-orescence Concentration Analyzer (IDEXX Co, Portland, ME, USA). Results were reported as the mean ± s.d.

Hybridomas were prepared using splenocytes from a mouse immunized with the A549 human lung carcinoma cell line and screened by flow cytometry and immunoprecipitation. Hybridomas were first selected for positive surface staining of A549 cells, and then further selected for immuno-precipitation of complexes resembling α_v β₅ from surface-labeled A549 extracts. Positive surface staining of the JY cell line (β_v+; β₃+; β₅−) was utilized to select against anti-α _v or anti-β ₃ antibodies (β ₃ and β₅ display similar mobility in SDS-PAGE). By these criteria we obtained six hybridomas (B5-IA9, B5-IIID8, B5-IVA4, B5-IVF2, B5-VID4 and B5-IG8).

To confirm that the β ₅ subunit was recognized by the B5 series of monoclonal antibodies, three of the anti-β₅ mon-oclonal antibodies were shown to be specifically reactive with CHO cells transfected with a cDNA encoding the human β₅ subunit. The CHO-β₅ transfectants were sorted with B5-IIID8 and stained with B5-IA9 (Fig. 1) and also B5-VID4 (not shown). No reactivity with these antibodies was detected in CHO cells transfected with vector alone. The CHO cell line did express hamster β₅ as detected by mAb P3G2 (which crossreacts with human and hamster α_v β₅ complex). To assess further the reactivity of the dif-ferent anti-β₅ monoclonal antibodies, western blot analysis was carried out using A549 total cell extracts. These exper-iments revealed that B5-IA9, B5-IVF2 and B5-VID4 directly recognize a single approx. 95,000 M_r band (Fig. 2, lanes a, b and e) corresponding to the reduced form of the β₅ subunit. Other anti-β₅ mAb (B5-IVA4, B5-IIID8 and B5-IG8) showed no reactivity on western blots, suggesting that they recognize epitopes unstable to SDS-PAGE treatment.

After immunodepletion, using polyclonal serum to the C terminus of β₅, none of the B5 series of mAbs immuno-precipitated any detectable β₅ or associated subunits (Fig. 3A). In a control experiment (Fig. 3B), β₁ and its associated subunits were fully depleted (lanes b,f) without removing β₅ (lanes g,h).

Quantitative analysis of the level of cell surface staining of β₅ compared to α_v revealed a surprising result. The level of β₅ present on A549 cells (stained with mAb B5-IA9, Fig. 4A) was apparently 1.5-to 3-fold greater than the level of α_v, as defined by mAb LM142. This difference was not due to a discrepancy in the reactivity of the antibodies them-selves, since on BT-20 cells (Fig. 4B) the level of β₅ was only slightly greater than α_v, and β ₅ and α_v were present at identical levels on MG-63 cells (Fig. 4C). Additional anti-α_v antibodies (13C2, P3G8) and an anti-v β ₅ mAb (P3G2) were tested and these also showed 1.5-to 3-fold lower reactivity than β ₅ antibodies on A549 cells, but nearly identical reactivity on MG-63 cells (not shown). Also, results with all six anti-β ₅ antibodies confirmed that ₅ was present at greater levels than α_v on A549 cells (Fig. 5). These results with anti-α_v, α_v β ₅ and β ₅ antibodies were confirmed using mAb culture supernatants, ascites fluid and purified IgG. Together, these results clearly indicate a cell-type-specific difference in the relative amounts of α_v and β ₅ on A549 cells but not on MG-63 cells. Immunoprecipitation studies were carried out to characterize further ₅ heterodimers present on A549 cells. As shown in Fig. 6A, anti-₅ polyclonal antibody, the B5 series of monoclonals (B5-IA9, B5-IIID8, B5-IVA4, B5-IVF2, B5-VID4, B5-IG8), and monoclonal antibodies against α _v β ₅ (P3G2) and anti-α_v (LM142), all immunoprecipitated similar complexes of α _v (140,000 M_r) and β ₅ (90,000 M_r) from surface 125I labeled A549 cell extracts. In some of the lanes (e.g. f and i) an additional band of approximately 115,000 M_r was seen in anti-β ₅ mAb immunoprecipitates. However, we do not think that this corresponds to an addi-tional ₅-associated subunit because it is also sometimes seen in anti-α _v immunoprecipitations, and this 115,000 Da band migrated in about the same position as reduced α _v (Fig. 6B). These results suggest that the 115,000 Da band could be a trace of partially reduced α _v that shows up together with non-reduced α _v.

In another experiment, immunodepletion of all α_v pro-tein with anti-α_v antibodies (P3G8 and 13C2) resulted in removal of all detectable anti-β₅ reactivity (not shown). Thus, this experiment also failed to provide any evidence for an additional β₅-associated subunit.

To test if anti-β ₅ monoclonal antibodies could interfere with cell attachment to extracellular matrix proteins, adhe-sion assays were performed using A549 and K562 cell lines. In the absence of added antibody, A549 cells attached rather well to fibronectin (Fn) and vitronectin (Vn) (Fig. 7A), whereas K562, a cell that expresses low levels of β ₅ and β ₃, displayed low binding to Fn and Vn (Fig. 7B). Adhesion of A549 cells to both substrata was sub-stantially blocked by both mAb B5-IVA4 and B5-VID4 (Fig. 7A). In contrast, mAb B5-IVF2 and B5-IG8 stimulated cell adhesion to both Vn and Fn, and mAb B5-IA9 stimulated A549 adhesion to Fn (Fig. 7A). The low level of adhesion by K562 cells to Vn was not markedly influ-enced by anti-₅ mAb, but adhesion to Fn was again stim-ulated by mAb B5-IVF2 and B5-IG8 (Fig. 7B). Similar results were obtained when a 120 kDa chymotryptic frag-ment of Fn was used for K562 adhesion (Fig. 7C). These results, in addition to western blotting and ELISA assays performed using anti-Fn and anti-Vn polyclonal sera, rule out the possibility of Vn contamination of the commercial preparation of Fn used in the adhesion assays. Unlike the results with A549 cells, attachment to Fn by K562 cells was not stimulated by mAb B5-IA9 (Fig. 7B). The basal adhesion of K562 cells to FN-120 was substantially blocked by two different anti-β ₁ antibodies (mAb16 and mAb13) as shown in Fig. 7C. However, in the presence of stimulatory anti-β ₅ antibody IG8, the anti-β₁ antibodies were no longer very effective, even when added together. Anti-β ₅ antibodies also had little effect on this β ₅-dependent adhesion of K562 cells to fibronectin (not shown). Thus the adhesion observed in the presence of stimulatory anti-β ₅ antibody appears not to be due to indirect effects mediated through β ₁ integrins.

In control experiments, the same cell lines showed moderate to weak adhesion to laminin, and weak or minimal attachment to collagen and fibrinogen, respectively, but none of this was either stimulated or inhibited by any of the anti-β ₅ antibodies (not shown). β ₅-dependent adhesion by β ₅-CHO cells could not be specifically analyzed in adhesion assays due to the high level of functional hamster β ₅.

As shown above (Fig. 4), β₅ was detected on three differ-ent adherent cell lines (A549, BT-20, MG-63). Analysis of a larger panel of cell lines (Table 1) demonstrated that β₅ was most highly expressed on carcinoma cell lines, but was also expressed on all other adherent cell lines tested. Notably, β₅ was absent from all T and B lymphoid cell lines tested, but weakly present on some myeloid cell lines (K562, U937). Whereas similar results were obtained regardless of which anti-β₅ antibody was utilized for the cell lines shown in Table 1, these antibodies showed dif-ferential staining of a human umbilical vein endothelial cell (HUVEC) line as indicated in Table 2. Notably, mAb VID4 stained more avidly, IA9 stained weakly, and the others gave intermediate results. Also, on CHO cells, three mAb gave strong staining (IVA4, IVF2, IG8), indicating a cross-reactivity with hamster seen with the other three antibodies β ₅, not seen with the other three antibodies (IA9, IIID8, IVF2).

The tissue distribution of β₅ was examined in immunohis-tochemical staining experiments by utilizing five anti-β₅ mAb (IVF2, IA9, IG8, IVA4 and VID4). In the human thymic cryostatic sections, heterogeneous staining patterns were found for the five anti-β₅ mAb. B5-IVF2 gave a strong staining on basement membranes, blood vessels, and on both the cortical and medullary stromal cell network (Fig. 8A). B5-IA9 and B5-IG8 showed similar staining patterns. High intensity staining of both mAb were detected on the capillary endothelial cells, especially in the medulla area (Fig. 8B and D). They both stained cortical and medullary stromal cells, Hassall’s corpuscles (Fig. 8B) and basement membranes, with slightly higher intensity staining for B5-IA9. Although B5-IVA4 weakly stained the basement mem-brane, very strong staining was found in the blood vessels, capillaries, both cortical and medullary stromal cells, and Hassall’s corpuscles (Fig. 8C). Furthermore, high expression of B5-VID4 was only observed on the capillary endothelial cells, blood vessels, and cortical stromal cells (data not shown).

Human skin and kidney sections were also examined for their expression of β₅. In the skin sections shown in Fig. 9A, the basal layer of epidermis, and perhaps also strati-fied squamous epithelium, were found to be positive for all five anti-β₅ mAb with only a slight difference in the inten-sity. Hair follicles also appeared to be weakly positive. Positive staining was also associated with the vascular basement membrane and inflammatory cells, possibly macrophages. Human kidney sections showed specific cytoplasmic and basement membrane staining positivity in the glomerular endothelial cells (Fig. 9B), juxta glomerular apparatus (Fig. 9C), proximal convoluted tubule cells (Fig. 9C), and cells comprising the collecting tubules (Fig. 9B) for all five anti-β₅ mAb. The tissue staining results are sum-marized in Table 2. As seen, particularly with the thymus staining results, there was marked variability in the stain-ing patterns observed with the different antibodies.

Whereas no antibodies directly recognizing the ₅ subunit had previously been reported, we describe here the utiliza-tion of six new anti-β₅ antibodies to carry out a detailed analysis of the structure, function and distribution of the integrin β₅ subunit. We know that these antibodies recognize β₅, because they each immunoprecipitated a complex closely resembling ₅, and the reactivity of each was immunodepleted by pre-treatment of cell extracts with anti-β₅ polyclonal serum. Furthermore, three of these antibod-ies specifically bound to β₅-transfected CHO cells, and three of these antibodies directly immunoblotted denatured β₅ subunit.

The data summarized in Table 3 indicate that each of these antibodies may recognize distinct β₅ epitopes. The mAbs B5-IVF2 and IVA4 appear most similar, only differing in their ability to recognize β₅ in western blots, whereas all of the other antibodies differ from each other in at least two of the five parameters listed in Table 3. The distinctive staining pattern listed in Table 2 is also consistent with each of these mAb recognizing different β ₅ epi-topes. Because our studies of β ₅ structure, function and distribution have made use of all six of these unique antibodies, the chance for erroneous conclusions due to spurious mAb crossreactivity is greatly minimized.

Our results clearly suggest that the integrin β ₅ subunit can play a role in cell adhesion to both vitronectin and fibronectin. Not only did two different anti-β ₅ antibodies block adhesion to both of the ECM proteins, but two to three other anti-β ₅ antibodies stimulated adhesion to those proteins. Our results firmly support the role of β ₅ in adhe-sion to fibronectin (independent of β ₁), and in fact on K562 cells, the stimulatory effect of anti-β ₅ antibodies on adhe-sion was more obvious on fibronectin than vitronectin. Thus, these results should help clarify some past confusion as to whether β ₅ is a receptor for vitronectin (Busk et al., 1992; Smith et al., 1990), or for vitronectin plus fibronectin (Cheresh et al., 1989). We speculate that perhaps β ₅ may have been functionally inactive in those previous studies where adhesion to fibronectin was not observed. With the identification of β ₅ as another fibronectin receptor, it is not immediately obvious why cells would need nine different fibronectin receptors (including also α_v β₁, α_v β₃, α_IIb β₃, α _v β₆, α₃ β₁, α₄ β₁, α₄ β₇, α₅ β₁). Possibly these diverse receptors not only provide the capability to adhere to dif-ferent regions of fibronectin, and with different affinities, but it also appears that different receptors may utilize their unique cytoplasmic domains to mediate diverse post-ligand-binding events (Chan et al., 1992; LaFlamme et al., 1992; Wayner et al., 1991).

The finding that β₅ integrin activity can be stimulated has important implications, suggesting that this integrin can exist in a state that is less than fully active, but is con-vertible to a more fully active functional state. In this regard, β₅ appears to resemble integrins in the β₁ (Arroyo et al., 1992; Chan and Hemler, 1993; Masumoto and Hemler, 1993; Kovach et al., 1992; Neugebauer and Reichardt, 1991; van de Wiel-van Kemenade et al., 1992), β₂ (Keizer et al., 1988) and β₃ (O’Toole et al., 1990) families, all of which can be stimulated by anti--chain mAbs to achieve higher activity. The availability of stimulatory anti-β₅ antibodies will not only allow assessment of the relative activation state of β₅ integrins, but the integrins should have practical value by specifically enhancing the role of β₅ during adhesion assays and during ₅ integrin purification.

In most cases the level of β₅ seen on cell surfaces was less than or equal to the level of αv, consistent with all of the β₅ being associated with αv, and with some v also being associated with other subunits such as β ₁, β ₃, β ₆ or β ₈. However, at present we have no good explanation for why the level of ₅ should be slightly greater than αv on a few cell lines such as A549. Because the apparent excess of β₅ was seen with all six distinct anti-β ₅ mAb, and because v was detected using four different antibodies, including the β ₅ complex specific antibody, it is unlikely that the difference is caused by a spurious cross-reactivity, or an unusual epitope seen only on certain cell lines. However, we have not obtained evidence for any additional β ₅-associated subunits other than αv. No other potential subunits were observed in anti-β ₅ immunopre-cipitations, and immunodepletion with anti-αv mAb did not leave any remaining β ₅ or β ₅-associated protein (not shown). Also, the inclusion of Mn2+ during and after cell lysis, or the use of cross-linking reagents, did not cause the immunoprecipitation of anything resembling a possible additional β ₅-associated subunit (not shown). Thus, while we cannot conclusively rule out the possibility of β ₅ being associated with an additional subunit (besides αv), at present we have no evidence for such an α subunit.

The weak 115,000 M_r band often seen in β₅ immuno-precipitations is not an additional β₅-associated subunit because it also appears in β₃ and v immunoprecipitations, suggesting that it is related to, or associated with, αv. One likely possibility is that the 115,000 M_r band represents a trace of partially reduced αv. Consistent with this, upon complete reduction of αv, only a single reduced αv band was seen in addition to the β₅ band.

Although the tissue distribution of several different integrins has been analyzed in human thymus (Zutter, 1991), kidney (Korhonen et al., 1990) and skin (Peltonen et al., 1989) sections, the distribution of β₅ has not been studied in detail in those or any other tissues. A few studies (e.g. see Damjanovich et al., 1992) have shown staining of several cell types in normal tissues with anti-αv antibodies, but reagents were not available to determine which subunit was present. We have utilized our newly prepared anti-β ₅ mAb to show moderately diverse expression of β ₅ on different cell types in thymus, skin and kidney. In the cortex and medulla of the thymus, we assume that β ₅ is staining epithelial cells, rather than thymocytes, since fresh thymo-cytes in suspension are completely negative for β ₅ (not shown). The heterogeneous distribution of β ₅ on different cell types could be functionally relevant. In the thymus high expression of β ₅ on capillary endothelial cells (consistent with this, cultured umbilical vein endothelial cells were also β ₅ positive) suggests that β ₅ could be involved in mediat-ing the migration and localization of thymocytes, perhaps by regulating the process of precursor T-cell entry to the thymus and the release of mature T-cells to the peripheral blood circulation. Diverse expression of β ₅ on thymic stro-mal cells, particularly cortical and medullary epithelial cells (data not shown), might also suggest a role of β ₅ in T-cell maturation in the thymus. The pronounced staining of epithelium in skin and kidney sections is consistent with the relatively high expression of β ₅ on cultured epithelial cell lines. The widespread expression of β ₅ on all adherent cell lines tested, but absence from cultured lymphoid cells, is reminiscent of β ₃ expression (Fradet et al., 1984). One possibility is that in vitro growth of adherent cell lines may select for β ₅ expression.

Consistent with the anti-β ₅ panel of monoclonal anti-bodies recognizing different epitopes, some of these anti-bodies showed markedly different tissue staining patterns, especially in the thymus. This result emphasizes that it is risky to base conclusions on data obtained with a single mAb, and also suggests that β ₅ may have either different conformations or be surrounded by different molecules in its different cellular locations.

In summary, our panel of newly developed anti-β ₅ mAb reagents greatly extends the available information about structure, function and distribution of β ₅, and these mAb should prove to be valuable tools in future studies.

We thank Dr Atul K. Bhan (Department of Pathology, Massa-chusetts General Hospital, Harvard Medical School, Boston) for providing skin and kidney tissues, Dr Joel L. Schwartz (Depart-ment of Immunopathology, Dana-Farber Cancer Institute) for crit-ical comments in discussing skin and kidney staining patterns, and Drs D. Cheresh and E. Wayner for gifts of monoclonal antibod-ies. This work was supported by NIH grant CA42368 (to MEH). Song Ye was supported by NIH grant AI15669 (to J. L. Stro-minger). R. Pasqualini was the recipient of post-doctoral fellow-ships from IARC (International Agency for Research on Cancer), World Health Organization; FAPESP (Foundation for Research Support of Sao Paulo), and is now supported by CNPq (National Council for Research Support, Brazil).