Here, we have used an α⁴ integrin cytoplasmic domain mutant to provide strong evidence for this hypothesis. Upon truncation of the α⁴ cytoplasmic domain, the α⁴β₁ integrin shows severe impairments in both constitutive and phorbol ester–induced static cell adhesion (9, 10), and also shows deficient adhesion strengthening under shear (11, 12). However, the reason for these defects was not previously understood. Because other integrin cytoplasmic domain mutations cause altered ligand binding (3, 13, 14), we closely examined binding of soluble vascular cell adhesion molecule (VCAM)-1¹ (15) to mutant and wild-type α⁴β₁ integrin. Not finding any alterations in ligand binding, we then examined receptor accumulation into cell surface clusters, and integrin lateral mobility. The results strongly support the hypothesis that integrin diffusion/clustering, independent of alterations in ligand binding, can play a major role in regulating integrin adhesive functions.

Ligand binding was carried out in 2 mM manganese, because calcium and magnesium (either alone, or together, at ~ 1–2 mM) fail to support binding of soluble VCAM (15, 26). To alleviate concern that manganese might mask differences in VCAM binding by inducing high affinity α⁴β₁ (15, 26, 27), manganese was titrated over a range of concentrations, whereas VCAM was held constant at 4 nM VCAM–Ig–AP (Fig. ​(Fig.22 C), or 500 nM VCAM-κ (Fig. ​(Fig.22 D). Manganese stimulated VCAM binding that was dose-dependent and α⁴-specific, but again no differences were apparent between K562–α4wt, K562–X4C2 and, K562– X4C0 cells (Fig. ​(Fig.2,2, C and D).

In CHO cells, compared with K562 cells, α⁴β₁ is constitutively more active with respect to mediating cell adhesion (9, 10). Nonetheless, in the CHO cellular environment, there were again no differences in direct VCAM binding to α⁴wt and X4C0 integrins at either optimal (Fig. ​(Fig.33 A; 4 nM VCAM–Ig–AP) or suboptimal (Fig. ​(Fig.33 B; 1 nM VCAM– Ig–AP) doses of ligand. Half-maximal direct VCAM–Ig– AP binding occurred at ~100 μM manganese for all transfectants examined, consistent with previously published manganese ED₅₀ values for VCAM–α⁴β₁ binding (15). In contrast with ligand binding, cell adhesion to immobilized VCAM was markedly diminished for CHO–X4C0 cells, compared with α⁴wt cells (Fig. ​(Fig.33 C). For example, adhesion at 0.1 and 1 μM Mn²⁺ was reduced by 88 and 69%, respectively.

To examine α⁴ tail deletion effects on very late antigen 4 conformation, we used the mAb 15/7, which recognizes a β₁ integrin conformation induced by ligand occupancy or manganese. When 15/7 epitope is induced by manganese, it correlates with increased ligand binding affinity (20). However, the 15/7 epitope also appears when the β₁ cytoplasmic domain is deleted, and ligand binding is diminished (28). Notably, 15/7 epitope expression is most readily induced on α⁴β₁, as compared with other β₁ integrins (Bazzoni, G., L. Ma, M.L. Blue, and M.E. Hemler, manuscript submitted for publication), and thus is an especially useful tool for evaluating altered α⁴β₁ conformations.

Negligible 15/7 epitope expression was seen for α⁴wt and mutant α⁴ integrins in K562 cells in the absence of stimulation (Table ​(Table1).1). However, the percentage of α⁴β₁ molecules expressing the 15/7 epitope increased dramatically upon addition of CS-1 peptide or manganese or both together to the K562–α⁴wt and K562–X4C2 cells. Importantly, stimulation with manganese and/or CS-1 peptide also resulted in comparably increased 15/7 epitope on K562–X4C0 cells (Table ​(Table1).1). No 15/7 epitope was detected in mock-transfected K562 cells (stimulated or unstimulated), demonstrating that 15/7 was specifically reporting α⁴β₁ conformational changes.

Having found that α⁴ tail deletion does not alter ligand binding or integrin conformation, we then sought alternative explanations for why tail deletion impairs cell adhesion. To this end, confocal laser microscopy was used to examine α⁴ tail deletion effects on accumulation of α⁴β₁ in clusters. As illustrated, wild-type α⁴ (Fig. ​(Fig.44 d) and X4C2 (Fig. ​(Fig.44 c) were detected in clusters on the surface of K562 cells after addition of recombinant soluble VCAM in the presence of manganese. In sharp contrast, X4C0 showed hardly any VCAM-induced accumulation in clusters (Fig. ​(Fig.44 b). The X4C0 subunit was present on the cell surface at levels comparable to α⁴wt and X4C2 (see Fig. ​Fig.1),1), suggesting that differences in signal strength reflect aggregated receptor and not differences in total receptor number. Cell surface staining was specific for α⁴, as shown by the lack of staining on mock-transfected K562 cells (Fig. ​(Fig.44 a). The distribution of α⁴ into clusters was dependent upon the addition of VCAM, because manganese alone (at 2 mM) did not induce clustering of α⁴ (data not shown).

Experiments were carried out at 4°C in the presence of sodium azide to prevent receptor internalization. Transverse sections of K562–α⁴wt cells (Fig. ​(Fig.44 e) showed peripheral, but not intracellular staining, consistent with cell surface clustering without receptor internalization. Also, transverse sections of K562–X4C0 cells showed no evidence for intracellular staining (data not shown). Furthermore, levels of cell surface α⁴ (X4C0) were unaltered after incubation with VCAM, as determined by flow cytometry (data not shown).

To extend our findings, we also examined α⁴ clustering induced by anti-α⁴ mAb, followed by polyclonal secondary antibody (Fig. ​(Fig.4,4, f–j). As indicated, clustering was again pronounced on K562–α⁴wt (Fig. ​(Fig.4,4, i and j) and K562– X4C2 (Fig. ​(Fig.44 h) cells, whereas minimal clustering was observed when the α⁴ tail was deleted (K562–X4C0 cells; Fig. ​Fig.44 g) or when no α⁴ was present (Fig. ​(Fig.44 f). Results in Fig. ​Fig.5,5, showing 9–15 cells/panel, confirm the single cell results shown in Fig. ​Fig.4.4. As indicated, nearly all of the K562– α⁴wt cells exhibit pronounced clustering, induced either by VCAM (Fig. ​(Fig.55 a) or by antibody (Fig. ​(Fig.55 c). In contrast, the X4C0 mutant was much less clustered (Fig. ​(Fig.5,5, b and d), despite being expressed on the cell surface at levels nearly equivalent to α⁴wt (see Fig. ​Fig.1).1).

The failure of truncated α⁴ to form cell surface clusters raises the possibility that increased or altered associations with the underlying cytoskeleton may impair the lateral mobility of truncated α⁴, restricting its redistribution into a cluster. Because restricted lateral movement of integrin receptors will likely be reflected by a lower integrin diffusion rate (22, 23), next we directly measured the diffusion coefficients of wild-type and truncated α⁴ in CHO transfectants at 37°C. The two-dimensional diffusivity of 40-nm gold particles, coated with anti-α⁴ mAb, was measured on the lamellipodia of CHO–α⁴wt and CHO–X4C0 cells spread on an α⁴-independent substrate, vitronectin. Movement of gold particles was viewed by high magnification, video-enhanced differential interference contrast microscopy and particles were tracked by computer with nanometer-level accuracy (23). A nonperturbing anti-α⁴ mAb, B5G10, was used because this mAb neither blocks nor stimulates α⁴β₁-mediated functions (29). It was shown elsewhere that nonperturbing antibodies coupled to 40-nm gold can report the random diffusion of integrins without stimulating the cross-linking and directed movement of these receptors (22).

As illustrated in Fig. ​Fig.6,6, A and C, gold particles bound to the lamella of CHO–α⁴wt cells diffused freely with a mean diffusion coefficient of 0.03 μm²/s (Fig. ​(Fig.66 E), consistent with the diffusion rate observed for other β₁ integrins (22), as well as other cell surface glycoproteins (30). However, truncation of the α⁴ cytoplasmic domain resulted in a significant decrease in the α⁴β₁ diffusion rate (P <0.01). Particles bound to CHO–X4C0 cells exhibited reduced lateral mobility (Fig. ​(Fig.6,6, B and D), with a diffusion coefficient that was sixfold lower (0.005 μm²/s) than wild-type α⁴β₁. No binding of gold particles was detected on mock-transfected CHO cells, demonstrating that the binding is α⁴β₁ specific (data not shown).

It is, perhaps, not surprising that replacement of the α⁴ tail with the α² tail had no effect on ligand binding or integrin conformation, because previously that mutation had no effect on cell adhesion, or tethering under flow (9, 12). Notably, replacement of the α^IIb cytoplasmic domain with that of α² did cause an increase in α^IIbβ₃ integrin ligand binding (14), suggesting that different rules may apply to regulation of the α^IIbβ₃ integrin.

The defect in cell adhesion seen for the X4C0 mutant is not due to altered ligand binding, but rather appears to arise from a reduced diffusion rate. Presumably, a lower rate of diffusion prevents the passive accumulation of integrin receptors into clusters. After initial cell contact with immobilized ligand, a dynamic, diffusion-dependent accumulation of clustered integrins may be needed to augment the overall cellular avidity for the ligand-coated surface. Notably, clustering deficiencies for the X4C0 integrin, directly measured here at 4°C, are consistent with an indirectly measured deficiency in X4C0 clustering seen previously at 37°C (18). In that case, X4C0 was defective in mediating antibody-redirected cell adhesion, a process dependent on mAb bridging between Fc receptors and clustered integrins (18).

How might α tail deletion cause decreased diffusivity leading to reduced clustering? We propose that the α chain cytoplasmic domain covers a negative site in the integrin β chain tail. Consistent with this model, it was previously shown that various integrin α chain tails can shield β chain tails from critical interactions with cytoskeletal proteins (32–34), whereas at the same time, α chains tails often make positive contributions to cell adhesion (9, 10, 35–37). Most likely, the unshielded and unregulated interactions of β tails with cytoskeletal proteins may lead to increased constitutive cytoskeletal anchoring, and thus diminished diffusion and clustering at adhesive sites. Supporting this notion, low doses of cytochalasin D markedly increased adhesion of truncated α⁴β₁, but not wild-type α⁴β₁. The range of cytochalasin D concentrations that promoted X4C0 adhesion (0.01–1 μg/ml) is consistent with previously published cytochalasin D concentrations that stimulated α^Lβ₂–mediated adhesion (7).

The overall importance of both diffusion and clustering to cell adhesion has been noted previously. For example, increases in the diffusion and lateral mobility of α^Lβ₂ (7) and LFA-3 (38) correlate with increases in cell adhesion and adhesion strengthening, respectively. Furthermore, integrin clustering is necessary for full integrin signaling (39), and clustering of α^Mβ₂ and α^Lβ₂ integrins promoted by phorbol ester or calcium also correlates with increased integrin-mediated adhesion (40, 41).

Regulation of integrin diffusion/clustering may be highly relevant towards the understanding of inside–out signaling mechanisms for β₁ and β₂ integrins, especially when affinity modulation is not involved. For example, stimulation of α⁴β₁–mediated adhesion with macrophage inflammatory protein-1β or with anti-CD3 or anti-CD31 antibodies did not detectably induce binding of soluble VCAM (26), and phorbol esters stimulated adhesion mediated by α⁵β₁, α^Mβ₂, and α^Lβ₂ without affecting soluble ligand binding (42–45). Notably, the effects of phorbol ester stimulation and integrin α⁴ tail deletion show a striking parallel. Like α⁴ tail deletion, phorbol esters also (a) fail to alter integrin affinity for ligand, (b) fail to alter α⁴β₁–dependent tethering under shear (11, 12), but (c) markedly regulate static cell adhesion and adhesion strengthening under hydrodynamic flow (11), and (d) regulate integrin diffusion rates (7). However, whereas α⁴ tail deletion leads to increased cytoskeletal restraints and diminished lateral diffusion, phorbol ester appears to release active cytoskeletal restraints, thereby increasing lateral diffusion of the α^Lβ₂ integrin (7). Together, these results emphasize that a diffusion/clustering mechanism may be of general importance for regulating adhesion, especially in the absence of changes in ligand binding (26). Also, impaired integrin diffusion/clustering may at least partly explain loss of cell adhesion observed upon the deletion of other integrin α chain cytoplasmic domains (35–37).