Research ArticleCell Biology

Mitotic cell responses to substrate topological cues are independent of the molecular nature of adhesion

See allHide authors and affiliations

Science Signaling  25 Feb 2020:
Vol. 13, Issue 620, eaax9940
DOI: 10.1126/scisignal.aax9940

Adhesion signaling not required

Adherent cells position their spindle poles parallel to the plane of the substrate and orient the spindle within that plane depending on cell shape–dependent mechanical forces. Both integrins, which mediate adhesion to the extracellular matrix (ECM), and cadherins, which mediate cell-cell interactions, are implicated in establishing spindle position. Anastasiou et al. found that both cadherin- and integrin-mediated adhesion supported proper spindle positioning in cultured cells. Although adhesion-dependent integrin activation was not required to position spindles parallel to the plane of the substrate, force-dependent integrin activation at the cell cortex was required to orient the spindle in response to mechanical cues generated by adhesion topology. Thus, spindle positioning does not depend on the molecular nature of adhesion and does not specifically require integrin-based adhesion, but integrin signaling is required to transduce mechanical cues that orient the spindle.

Abstract

Correct selection of the cell division axis is important for cell differentiation, tissue and organ morphogenesis, and homeostasis. Both integrins, which mediate interactions with extracellular matrix (ECM) components such as fibronectin, and cadherins, which mediate interactions between cells, are implicated in the determination of spindle orientation. We found that both cadherin- and integrin-based adhesion resulted in cell divisions parallel to the attachment plane and elicited identical spindle responses to spatial adhesive cues. This suggests that adhesion topology provides purely mechanical spatial cues that are independent of the molecular nature of the interaction or signaling from adhesion complexes. We also demonstrated that cortical integrin activation was indispensable for correct spindle orientation on both cadherin and fibronectin substrates. These data suggest that spindle orientation responses to adhesion topology are primarily a result of force anisotropy on the cell cortex and show that integrins play a central role in this process that is distinct from their role in cell-ECM interactions.

INTRODUCTION

Regulation of spindle orientation is critical in multicellular organisms and important in both symmetrically and asymmetrically dividing cells. During asymmetric divisions, the spindle is oriented along a polarity axis so that cell fate determinants are asymmetrically inherited, thus influencing the fate of the daughter cells. During symmetric divisions, the spindle is often oriented parallel to the plane of the tissue or the plane of attachment, or both, guiding tissue morphogenesis and maintaining epithelial integrity (1, 2). Both the positioning and orientation of the mitotic spindle are achieved through the capture of astral microtubules (MTs) at discrete regions on the cell cortex by a conserved tripartite complex. This complex is composed of the heterotrimeric guanosine 5′-triphosphate (GTP)–binding protein (G protein) regulator leucine-glycine-asparagine repeat protein (LGN), the Gαi subunits of G proteins, and the nuclear mitotic apparatus protein (NuMA) (3). The motor proteins dynein and dynactin are anchored to the cortex through interactions with this complex and exert forces on astral MTs to position the spindle between two anchoring points (49). Thus, spindle orientation is determined by the polar distribution of LGN-NuMA on the cell cortex. Factors shown to influence this polarity include cell-cell adhesion, integrin-based adhesion, cell shape, and mechanical forces (7, 1012). In polarized cells, additional restrictions that function in conjunction with the above are also imposed. In many epithelial tissues, for example, LGN is excluded from the apical cortex by apically localized atypical protein kinase C, thereby restricting divisions to the plane of the epithelium (13, 14).

The role of forces exerted on the cortex of mitotic cells has emerged as a major regulator of spindle orientation, and it is clear that the spindle can rapidly respond to external mechanical stimuli. Specifically, nonpolar adherent cells sense forces transmitted through actin-rich retraction fibers (RFs) and can dynamically reorient their spindles along force vectors (15). In addition, work in Danio rerio and Xenopus laevis shows that the same holds true in polarized cells like embryonic epithelia, where forces are present because of tissue-level tension and transmitted through adherens and tight junctions (1619). Evidence from both Xenopus and cultured Madin-Darby canine kidney cells suggests that cortical force distribution is the major factor determining spindle orientation in polarized epithelia rather than cell shape (20, 21).

The role of the extracellular matrix (ECM) and integrin-based adhesion in directing spindle orientation is well documented both in vitro (22) and in vivo (23). Blocking antibodies that interfere with the interactions of cells with the fibronectin (FN)–rich ECM on the blastocoel roof of Xenopus embryos causes randomization of spindle orientation in the deep cells that are in contact with the matrix and normally divide parallel to the plane of the epithelium (24). Similarly, integrin β1 knockout mice display spindle orientation defects during skin stratification that appear to be a consequence of defects in apicobasal polarity, leading to the randomization of the LGN crescent on the cell cortex. In this context, loss of α-catenin also impairs the apical localization of protein kinase C ζ and leads to defects in spindle orientation (12). Last, one of the central transducers of integrin signaling, focal adhesion kinase (FAK), also plays a role in spindle orientation in the context of Xenopus epiboly, in which disrupting FAK function elicits similar defects to what was reported initially in experiments disrupting the ECM (21, 25).

In vitro, there is evidence that integrin-mediated adhesion is responsible for the clear bias of nonpolar cells to divide parallel to the plane of adhesion (11). Cells cultured on integrin ligands orient their spindles parallel to the substrate, unlike cells grown on poly-l-lysine (PLL) (11). In addition, disruption of integrin β1 function elicits spindle orientation defects (11, 18), as does loss of any of several focal adhesion (FA) proteins, including FAK, paxillin, and breast cancer anti-estrogen resistance protein 1 (also known as p130Cas) (18). Moreover, it is suggested that a direct interaction between integrin-linked kinase and dynactin-2 (p50) links integrins to the dynein complex and controls the position of force generators (8). The precise mechanism of spindle orientation with respect to the plane of attachment has not been definitively determined; however, shape anisotropy is unlikely to be the major determinant but, rather, RF-derived forces. RFs are restricted to the lateral aspect of the mitotic cortex and absent from both the part of the cell that is in contact with the substrate and the top of the cell, effectively restricting applied forces and the spindle within the plane of attachment. The clearest evidence of this comes from experiments carried out in HeLa cells sandwiched between two glass coverslips coated with FN. In this case, cells are prevented from fully rounding up and display a clear long axis parallel to the plane of attachment. This might be predicted to favor spindle orientation parallel to the long axis; however, the presence of RFs on both the dorsal and ventral cell surfaces in this situation leads the spindle to orient vertical to the plane of attachment (21). Integrin-dependent adhesion is also responsible for guiding spindle orientation within the plane of attachment. Specifically, asymmetric adhesion geometry leads to asymmetric force distribution on the cortex of metaphase cells through the RFs. This asymmetric force distribution dictates spindle orientation, with cells aligning along the axis receiving the greatest force (15, 26). In addition, interphase adhesion geometry is, at least in part, transmitted to the mitotic cell through caveolin-1 in a process requiring integrin activation (27). Last, integrin αvβ5 adhesions are involved in maintaining cell-ECM attachment during mitotic rounding and division and are proposed to be required for spatial memory transmission (28).

Despite extensive evidence implicating integrins and the ECM in directing spindle orientation, the mechanism through which they are involved in this process is not clear. We have previously suggested that integrin-based complexes play a role in spindle orientation that is distinct from their role in cell adhesion. Specifically, integrin β1 is involved in the sensing of forces applied to the cortex of mitotic cells, both in vivo and in vitro (18). Here, using cadherin substrates, we demonstrated that cell-ECM adhesion had no direct role in spindle orientation responses of cultured cells to planar substrates or adhesion geometry and that ECM-free substrates could guide spindle orientation, both with respect to the plane of attachment and in response to substrate topological cues. Under ECM adhesion–independent conditions, integrins and FA proteins were still required for correct spindle orientation, effectively decoupling the role of integrins in this process from their role in cell-ECM interactions. The identical responses of cells to cadherin and FN substrates suggested that neither cadherin-based adhesion nor integrin-based adhesion plays a direct role in the regulation of spindle orientation, but they simply provide mechanical anchoring and spatial cues to the dividing cell.

RESULTS

Planar substrates guide spindle orientation parallel to the plane of attachment irrespective of the molecular nature of adhesion

Adherent, nonpolarized cells cultured on ECM-coated glass orient their mitotic spindles parallel to the plane of the substrate (11). The spindle fails to orient properly in cells attached to PLL-coated substrates, which do not spread. The spindles of cells attached on FN-coated substrates are misoriented when the cells are treated with arginylglycylaspartic acid peptide, which blocks attachment by binding to integrins, or inhibitory antibodies against integrin β1, leading to the conclusion that integrin-mediated adhesion is indispensable for spindle orientation (11, 18). To examine whether integrin-dependent cell adhesion is required for spindle orientation parallel to the plane of adhesion, we compared HeLa cells attached to glass substrates coated with FN to cells attached to glass substrates coated with a chimeric protein composed of the extracellular domain of N-cadherin fused to its C-terminal with the Fc domain of human immunoglobulin G1 (IgG1) (N-cadherin Fc) in the absence of serum. Most of the cells attached on N-cadherin Fc were fully spread within 30 min and formed linear adherens junctions (AJs) as previously described (29, 30). Unlike cells on FN, cells on Ν-cadherin Fc displayed no detectable FA formation or integrin activation (Fig. 1, A and B). To preclude the possibility that integrin-based adhesion was contributing to spindle orientation responses to cadherin substrates, we ensured that cells were attaching only through AJs and were devoid of FAs by allowing cells to attach for 30 min or less under serum-free conditions. During mitotic cell rounding, adhesive complexes were disassembled on both FN and N-cadherin Fc substrates, and the receptors (integrin β1 and N-cadherin, respectively) became redistributed on the cell cortex with minimal enrichment at the ventral surface (Fig. 1C). HeLa cells attached on N-cadherin Fc and FN substrates were evaluated with respect to spindle orientation (Fig. 1, C and D). N-cadherin–mediated adhesion led to a spindle orientation that was indistinguishable from that of cells adhering on FN, suggesting that integrin-dependent adhesion was dispensable for spindle orientation parallel to the substrate plane (Fig. 1, C and D). To ensure that this phenomenon was not cell type specific, we repeated these experiments using U2OS cells, which also produce N-cadherin. These cells behaved similarly to HeLa cells on N-cadherin substrates and formed AJs, which were devoid of detectable FAs (fig. S1A). Quantification of spindle angles from cells attached on N-cadherin Fc or FN revealed that U2OS cells displayed oriented divisions on both substrates (Fig. 1, E and F).

Fig. 1 Cadherin- and integrin-based adhesions elicit divisions parallel to the substrate plane.

(A and B) Confocal images of interphase HeLa cells on coverslips coated with fibronectin (FN) or N-cadherin Fc (N-Fc). Cells were stained for β-catenin (which colocalizes with cadherins), active integrin β1, and either FAK [N = 220 (A)] or paxillin [N = 256 (B)]. N, number of interphase cells across all conditions from three independent experiments. (C) Representative Z-stacks at the plane of the spindle poles and side projections (xz) of metaphase HeLa cells on FN and N-Fc substrates. Cells were stained for α-tubulin and β-catenin. N = 180 total number of metaphase cells across all conditions from four independent experiments. (D) Distribution of spindle-to-substrate angles in (C). Means ± SEM: HeLa cells on FN, 7.235° ± 0.4565° (N = 60); HeLa cells on N-Fc, 6.428° ± 0.5170° (N = 60); HeLa on PLL, 32.87° ± 1.961° (N = 60). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. (E) Z-stacks at the plane of the spindle poles and side projections of metaphase U2OS cells on FN and N-Fc substrates. Cells were stained for α-tubulin and β-catenin. N = 100 total number of metaphase cells across all conditions from four independent experiments. (F) Distribution of spindle-to-substrate angles in (E). Means ± SEM: U2OS cells on FN, 6.566° ± 0.4883° (N = 50); U2OS cells on N-Fc, 6.702° ± 0.4522° (N = 50). Data were analyzed by the Mann-Whitney test. N, number of metaphase cells, from each condition, from three independent experiments. Scale bars, 10 μm. n.s., not significant.

To ensure that cell spreading and spindle orientation on N-cadherin substrates depended on AJs, we treated HeLa cells with EGTA, which selectively chelates Ca2+ ions. Given the absolute requirement of Ca2+ ions for AJ formation and stabilization, EGTA would presumably lead to AJ disassembly without affecting integrin-based adhesion, which does not require Ca2+ ions (31, 32). EGTA treatment led to rounding of cells on N-cadherin substrates but had no effect on cells attached on FN, confirming that cells seeded on cadherin substrates attached and spread specifically through the formation of AJs, as expected (fig. S1B). To examine the effects of AJ disassembly on spindle orientation, cells were monitored after EGTA addition and fixed when cell rounding was observed (20 min) but before cell detachment. EGTA treatment disrupted spindle orientation on N-cadherin substrates, whereas it had no effect on FN substrates, showing that cadherin-based adhesion guides spindle orientation parallel to the plane of attachment (Fig. 2, A and B). Last, to eliminate the possibility that ECM ligands secreted by the cells within the 30-min period of attachment played a role in this context, we treated cells with brefeldin-A, a well-characterized inhibitor of the secretory pathway (33, 34). Cells were treated before seeding on N-cadherin substrates and were allowed to spread in the presence of the inhibitor for 30 min. Brefeldin-A treatment led to disassembly of the Golgi complex (fig. S1C) yet had no effect on spindle orientation in cells attached on N-cadherin substrates (Fig. 2, C and D). These results show that integrin-based adhesion is dispensable for planar spindle orientation on N-cadherin substrates.

Fig. 2 Cadherin engagement is necessary for correct spindle orientation on cadherin substrates.

(A) Optical sections and side projections (xz) of control and EGTA-treated metaphase HeLa cells on FN or N-Fc substrates. Cells were stained for β-catenin and α-tubulin, and images were acquired at the plane of the spindle poles. N = 240 total number of metaphase cells across all conditions from three independent experiments. (B) Distribution of spindle-to-substrate angles in (A). Means ± SEM: control on FN, 7.235° ± 0.4565° (N = 60); EGTA-treated on FN, 5.210° ± 0.5494° (N = 60); control on N-Fc, 6.428° ± 0.5170° (N = 60); EGTA-treated on N-Fc, 24.42° ± 1.514° (N = 60). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. (C) Distribution of spindle-to-substrate angles in metaphase control and brefeldin-A–treated HeLa cells on N-Fc. Control on N-Fc, 8.910° ± 0.6321° (N = 80); brefeldin-A–treated on N-Fc, 9.809° ± 0.7211° (N = 80). Data were analyzed by the Mann-Whitney test. N, number of metaphase cells, from each condition, from three independent experiments. n.s., not significant. (D) Representative optical sections and side projections of representative control and brefeldin-A–treated metaphase HeLa cells on N-Fc. Cells were stained for β-catenin and α-tubulin. N = 160 total number of metaphase cells across all conditions from three independent experiments. Scale bars, 10 μm.

LGN and NuMA are recruited to the cortex and become polarized in the absence of cell-ECM interactions

The evolutionarily conserved molecular complex composed of Gαi, LGN, and NuMA is essential for correct spindle orientation and positioning in different tissues, both in invertebrate and vertebrate systems (7, 10, 3538). During mitosis, this complex becomes localized at the cell cortex and guides the recruitment of the MT motor dynein (7, 39). The spatially restricted placement of LGN and NuMA on the cortex leads to the defined anchoring of astral MTs at specific regions of the cortex and generates pulling forces on the spindle poles, leading to the orientation and/or positioning of the spindle (10, 40, 41). LGN and NuMA both become polarized in mitotic cells adherent on FN, displaying clear enrichment at the lateral cortex and the spindle capture sites (10, 18, 40). To examine whether cadherin-based adhesion elicits a similar polarization, cells were allowed to attach on FN or N-cadherin Fc and then stained using antibodies specific for NuMA and LGN. In both FN-attached and N-cadherin Fc–attached cells, LGN and NuMA were enriched at the lateral cortex and absent from the dorsal (apical) and ventral (basal, adherent) domains (Fig. 3A). In addition, both LGN and NuMA displayed correct polarity irrespective of the substrate onto which cells were attached and were clearly enriched at the spindle capture regions (Fig. 3A). These results suggest that integrin-based adhesion does not play a role in either the cortical recruitment or the polarization of the capture complex in this context. They also suggest that N-cadherin does not recruit LGN to the cortex, in contrast to what was previously suggested for E-cadherin (16, 20, 42). If, in fact, N-cadherin recruited LGN, then this would not only promote LGN localization to the ventral region of the cell but also presumably lead to cell divisions oriented perpendicular to the plane of the substrate, neither of which was observed (Fig. 3A). Nevertheless, to ensure that N-cadherin ligation did not lead to LGN recruitment, we generated micropatterned N-cadherin substrates, with stripes of N-cadherin Fc. Confocal imaging of the basal area of these cells failed to reveal a polarized distribution or enrichment of LGN on the N-cadherin Fc stripes or on the linear AJs forming on the stripes, confirming that N-cadherin–based AJs did not recruit LGN to the cell cortex (Fig. 3B). Overall, these data show that the Gαi-LGN-NuMA capture complex can be recruited to and become polarized on the lateral cortex of mitotic cells in the absence of cell-ECM interactions, in agreement with the ability of cadherin substrates to guide planar spindle orientation. They also suggest that, unlike what has been reported for E-cadherin, N-cadherin adhesions do not recruit LGN to the cortex but simply provide mechanical anchoring during cell division.

Fig. 3 The spatial distribution of LGN and NuMA does not depend on the molecular nature of adhesion.

(A) Representative optical sections and side projections (xz) of HeLa cells seeded on FN and N-Fc. Images were acquired at the plane of the spindle poles. Cells were stained for α-tubulin, LGN, and NuMA as indicated. N = 215 total number of metaphase cells across all conditions from three independent experiments. (B) Representative confocal image of interphase HeLa cells expressing LGN-GFP on N-Fc linear micropatterns and labeled for LGN, β-catenin, and N-Fc. Images were acquired at the plane of attachment. N = 112 total number across all conditions of interphase cells from three independent experiments. (C) Distribution of spindle-to-substrate angles in metaphase control and E-cadherin–GFP–expressing HeLa cells on FN or E-cadherin Fc (E-Fc). Means ± SEM: control on FN, 9.769° ± 0.4638° (N = 74); control on E-cadherin, 34.17° ± 1.305° (N = 71); E-cadherin–GFP on FN, 9.315° ± 0.4771° (N = 65); E-cadherin–GFP on E-cadherin Fc, 9.184° ± 0.4588° (N = 67). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells from three independent experiments. (D) Representative optical sections and side projections of HeLa cells expressing E-cadherin–GFP on E-cadherin Fc substrate. Cells were stained for α-tubulin, LGN, and NuMA as indicated. Images were acquired at the plane of the spindle poles. N = 277 total number of metaphase cells across all conditions from three independent experiments. (E) Distribution of spindle-to-substrate angles in control and LGN-C′ mCherry–expressing metaphase HeLa cells on FN or N-Fc. Means ± SEM: LGN-C′ mCherry on FN, 3.248° ± 0.8769° (N = 74); LGN-C′ mCherry on N-Fc, 7.509° ± 0.3213° (N = 71); control on FN, 3.189° ± 0.8956° (N = 65); control on N-Fc, 6.878° ± 0.2764° (N = 67). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. Scale bars, 10 μm.

To explore the discrepancy between published work on E-cadherin and our findings using N-cadherin, we transiently expressed E-cadherin fused to green fluorescent protein (GFP) in HeLa cells and seeded them on E-cadherin Fc substrates. E-cadherin–expressing HeLa cells attached and spread on E-cadherin Fc as expected (93.5% of cells spread) (38), whereas control cells failed to spread and remained round (6.5% of cells spread). Quantification of spindle angles of cells attached on E-cadherin Fc or FN revealed that although both E-cadherin–expressing and control cells oriented their spindles parallel to the plane of attachment on FN, only E-cadherin–expressing cells oriented their spindles parallel to the substrate plane on E-cadherin Fc (Fig. 3C and fig. S1D). These data show that E-cadherin enables HeLa cell attachment and spreading on E-cadherin Fc and consequently allows these cells to orient their spindle parallel to the plane of attachment. Thus, E-cadherin planar substrates guide spindle orientation parallel to the plane of attachment as effectively as N-cadherin Fc and FN, confirming that cell-ECM interactions are dispensable in this process. These data also suggest that, like N-cadherin–based AJs, E-cadherin AJs fail to recruit LGN to the ventral region of the cell because such a recruitment would result in a spindle orientation that is perpendicular to the plane of attachment.

We went on to examine the localization of LGN and NuMA in E-cadherin–expressing HeLa cells dividing on E-cadherin Fc substrates. Both LGN and NuMA display correct polarization on the lateral cortex of E-cadherin–expressing cells on E-cadherin Fc substrates, whereas no enrichment was observed in the ventral area (Fig. 3D). To assess the contribution of Gαi-LGN-NuMA on spindle orientation under cadherin-based adhesion conditions, we eliminated the interaction between NuMA and LGN using a portion of the C terminus of LGN that acts as a dominant negative. This construct lacks the ability to bind NuMA and effectively abolishes cortical MT anchoring by competing with endogenous LGN for Gαi (3, 14, 38, 43). HeLa cells were transiently transfected with the LGN–C terminus construct and seeded on FN and N-cadherin Fc. Quantification of spindle angles of cells attached on N-cadherin Fc or FN revealed that LGN–C terminus expression elicited spindle misorientation on both substrates (Fig. 3E). This shows that the Gαi-LGN-NuMA complex is required for correct spindle orientation irrespective of the substrate.

Overall, the above data show that LGN and NuMA can be recruited to the lateral cortex and polarize correctly in the absence of cell-ECM interactions on both N- and E-cadherin substrates and that integrin signaling is not required for the elimination of the complex from the ventral cell membrane. At the same time, they suggest that any interactions taking place between cadherins and LGN may act as secondary cues for refining the localization of LGN but do not play a central role in the overall spatial distribution of the protein on the cell cortex during mitosis.

The spatial distribution of cadherin adhesion governs spindle orientation through asymmetric RF-derived force application on the cell cortex

Micropatterns of defined geometry, which dictate a specific cell shape and adhesion geometry, induce specific spindle orientation within the plane of attachment. This orientation depends on the distribution of the RFs, which exert forces on the cortex of a mitotic cell (15, 18, 26, 44). Adhesion geometries that lead to the generation of an asymmetric distribution of RFs, and consequently force anisotropy on the cell cortex, can direct spindle orientation in a predictable manner, with spindles aligning along the greatest force vector. In effect, RFs provide a memory of a cell’s interphase adhesion geometry by generating anisotropic force distribution on the cell cortex (15, 22, 26). Spindle orientation parallel to the plane of attachment can be a consequence of the absence of force on the ventral and dorsal regions of adherent mitotic cells, due to the lack of RFs terminating in those regions (18). Taking this possibility into consideration, we postulated that because adhesion on N- and E-cadherin Fc substrates guided spindle orientation parallel to the plane of the substrate, it would also effectively guide orientation within the XY plane in response to a defined adhesion geometry.

We initially examined whether mitotic cells on N-cadherin Fc substrates were able to generate RFs, a prerequisite for correct spatial memory in a mitotic cell. To do that, we allowed HeLa cells to adhere on N-cadherin Fc and FN and compared their ability to generate RFs during cell division (Fig. 4A). RFs formed on both substrates; however, on N-cadherin Fc, these were negative for active integrin β1 and displayed strong β-catenin staining, whereas the opposite was true for RFs on FN, which displayed active integrin β1 staining and weak β-catenin staining. We next generated both N-cadherin Fc and FN micropatterns and examined the ability of the two substrates to guide spindle orientation within the attachment plane. We adapted published techniques to establish a protocol for generating L-shaped and linear micropatterns of the adhesion substrates on glass coverslips (45). We initially seeded cells on L-shaped micropatterns coated with N-cadherin Fc or FN to examine the ability of cells to spread under these conditions. Cells on FN micropatterns adhered, spread, and displayed clear FA formation and activation of integrin β1 at the cell periphery, whereas cells on N-cadherin micropatterns displayed linear AJ formation (Fig. 4B). We went on to evaluate and compare the ability of cells seeded on N-cadherin Fc and FN patterned substrates to orient their spindles along either the long axis of linear patterns or the hypotenuse of L-shaped micropatterns. Both FN and N-cadherin Fc patterns were equally effective in directing spindle orientation within the plane of attachment (Fig. 4, C and D). These data show that AJ formation can effectively guide spindle orientation based on adhesion geometry, through asymmetric RF distribution and the generation of force anisotropy on the cortex, in a similar manner to what has been described for FN and FAs (15). In conjunction with the above data showing that spindle orientation parallel to the plane of attachment can also be effectively guided by cadherin-based adhesion (Fig. 1), it becomes clear that adhesion provides cues to the cell during division that are purely mechanical in nature, and these are independent of the molecular nature of the adhesion complex.

Fig. 4 Spindle responses to adhesion geometry are independent of the molecular nature of adhesion.

(A) Optical sections and side projections (xz) of mitotic HeLa cells on FN or N-Fc. Cells were stained for active integrin β1 and β-catenin. Images were acquired at the plane of spindle poles. N = 197 total number of metaphase cells across all conditions from three independent experiments. (B) Representative images of interphase HeLa cells on FN or N-Fc L-shaped micropatterned substrates. Cells were stained for active integrin β1, β-catenin, β-tubulin, FN, and N-Fc as indicated. Images were acquired at the plane of attachment. N = 152 total number of interphase cells across all conditions from three independent experiments. (C) Metaphase HeLa cells on FN or N-Fc L-shaped micropatterned surfaces. Cells were stained for paxillin, β-catenin, β-tubulin, FN, and N-Fc as indicated. The images were acquired at the plane of mitotic spindle. N = 82 total number of metaphase cells across all conditions from three independent experiments. (D) Distribution of XY spindle angles in metaphase HeLa cells on FN or N-Fc L-shaped and linear micropatterns. Means ± SEM: FN bar patterns, 6.399° ± 0.4586° (N = 44); FN L-shaped patterns, 5.829° ± 0.4129° (N = 31); N-Fc linear patterns, 6.235° ± 0.2674° (N = 85); N-Fc L-shaped patterns, 6.402° ± 0.2679° (N = 82). Data were analyzed by the Mann-Whitney test. N, number of metaphase cells, from each condition, from three independent experiments. Scale bars, 10 μm. n.s., not significant.

Integrin β1 activation is necessary for spindle responses to planar substrates irrespective of the molecular nature of adhesion

There is evidence suggesting that integrin activation and FA proteins are required for spindle orientation independently of their role in cell adhesion (18, 21). Integrin β1 becomes activated on the lateral cortex of mitotic cells, leading to the formation of a cortical mechanosensory complex (CMC; which is composed of FA proteins including FAK, paxillin, p130Cas, and Src), and governs spindle responses to mechanical force (18). As shown above, cells seeded on N-cadherin Fc did not form FAs, suggesting that integrins are not required for cell spreading under these conditions. We thus wanted to examine a possible role for integrin β1 in spindle orientation, in an integrin adhesion–independent context. We initially compared the state of integrin β1 during interphase and mitosis on both N-cadherin Fc and FN substrates. Cells on FN displayed active integrin β1 at FAs during interphase and at the lateral cortex during mitosis, whereas cells on N-cadherin Fc did not display integrin β1 activation during interphase but did have active integrin β1 on the lateral cortex during mitosis (Fig. 5, A and B). These results suggest that integrin β1 activation on the cortex is independent of the mode of adhesion.

Fig. 5 Integrin activation is required for spindle responses to planar cadherin substrates.

(A) Optical sections of interphase HeLa cells on FN or N-Fc. Cells were stained for β-catenin and active integrin β1, and images were acquired at the cell-substrate plane. N = 357 total number of interphase cells across all conditions from three independent experiments. (B) Optical sections and side projections (xz) of control HeLa cells on FN or N-Fc at the plane of the spindle pole. Cells were stained for β-tubulin and active integrin β1. N = 205 total number of metaphase cells across all conditions from three independent experiments. (C) Optical sections and side projections of control and AIIB2-treated metaphase HeLa cells on FN or N-Fc. Cells were stained for β-tubulin and β-catenin. Images were acquired at the plane of each spindle pole. N = 342 total number of metaphase cells across all conditions from three independent experiments. (D) Distribution of spindle-to-substrate angles in control and AIIB2-treated metaphase HeLa cells on FN or N-Fc. Means ± SEM: Control on FN, 7.358° ± 0.3919° (N = 85); AIIB2 treated on FN, 31.63° ± 1.110° (N = 85); control on N-Fc, 6.665° ± 0.3849° (N = 86); AIIB2 treated on N-Fc, 30.82° ± 1.104° (N = 86). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. (E) Distribution of the cell spread of control and AIIB2-treated interphase HeLa cells on FN or N-Fc. Means ± SEM: Control on FN, 19.89 ± 0.5394 μm (N = 180); AIIB2 treated on FN, 12.88 ± 0.3790 μm (N = 181); control on N-Fc, 21.67 ± 0.4622 μm (N = 176); AIIB2 treated on N-Fc, 19.56 ± 0.5394 μm (N = 181). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of interphase cell diameter, from each condition, from four independent experiments. Scale bars, 10 μm.

Next, we addressed a possible role for cortical integrin activation during mitosis in spindle orientation in the absence of ECM-based adhesion. We allowed cells to attach and spread on N-cadherin Fc and FN for 20 min and then treated them with AIIB2, a well-characterized allosteric integrin β1 inhibitory antibody before evaluating spindle orientation (4648). Despite having no effect on cell spreading on N-cadherin Fc, inhibition of integrin β1 elicited spindle misorientation on both substrates (Fig. 5, C to E). To verify these results, we went on to use two additional well-characterized integrin β1 inhibitory antibodies, P4C10 and P5D2. Both antibodies elicited spindle misorientation on both substrates, confirming a role of integrin β1 activation in the selection of the division axis irrespective of the nature of adhesion (fig. S2, A and B).

The CMC proteins p130Cas and c-Src (Src) are necessary for spindle orientation in the absence of cell-ECM interactions

The above data showed that integrin β1 has a central role in spindle orientation, independent of its role in cell-ECM interactions. Previous work revealed that the tension-driven activation of integrin β1 on the lateral cortex of mitotic cells leads to the recruitment of several FA proteins, including FAK, p130Cas, and Src, thus establishing the CMC. This complex, through a yet-undefined mechanism, appears to spatially bias astral MT capture, guiding spindle responses to external mechanical stimuli. Given the fact that the CMC is established on activated integrin β1 and that integrin β1 is involved in determining spindle orientation irrespective of the nature of adhesion, we hypothesized that CMC components known to be involved in spindle orientation on ECM-based substrates would also play a role in this process on cadherin substrates (18). We initially examined whether CMC components were also localized at the lateral cortex of mitotic cells on N-cadherin Fc substrates. HeLa cells were seeded and allowed to attach on N-cadherin Fc for 30 min under serum-free conditions and before staining with antibodies for the major CMC members (FAK, p130Cas, and Src). These proteins, just like active integrin β1, became localized to the lateral cortex of mitotic cells, supporting a possible involvement in spindle orientation responses to cadherin substrates (Fig. 6A). We next directly addressed the role of p130Cas in spindle orientation on N-cadherin Fc by taking advantage of p130Cas-null mouse embryonic fibroblasts (49). When adhering to FN, these cells display spindle orientation defects that are rescued by the expression of wild-type p130Cas (18). When seeded on N-cadherin Fc, p130Cas-null fibroblasts attached and spread, forming linear AJs without forming detectable FAs (fig. S2C). Evaluation of spindle orientation revealed that p130Cas-null cells displayed spindle misorientation both on N-cadherin Fc and FN substrates (Fig. 6, B and C). This defect was rescued on both substrates by the expression of wild-type p130Cas, using a vector that expresses GFP from an internal ribosome entry sequence to track transfected cells (Fig. 6, B and C) (50). These data effectively uncouple the role of p130Cas in cell-ECM interactions from its role in spindle orientation.

Fig. 6 The CMC proteins p130Cas and Src are required for spindle responses to planar cadherin substrates.

(A) Optical sections of representative metaphase HeLa cells on N-Fc) at the spindle plane, costained for β-tubulin plus FAK, p130Cas, or Src. (B) Distribution of spindle-to-substrate angles in cells in p130Cas−/− and p130Cas-reconstituted metaphase cells seeded on FN and N-Fc. Means ± SEM: p130Cas−/− on FN, 33.03° ± 2.280° (N = 60); p130Cas−/− on N-Fc, 32.01° ± 2.167° (N = 60); p130Cas−/− + WTp130Cas on FN, 8.756° ± 0.5179° (N = 60); p130Cas−/− + WTp130Cas on N-Fc, 8.462° ± 0.5809° (N = 60). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. (C) Representative optical sections and side projections (xz) of p130Cas−/− and p130Cas-reconstituted metaphase cells seeded on FN and N-Fc. Cells were stained for β-tubulin and β-catenin, and the images were acquired at the spindle poles plane. N = 240 total number of metaphase cells across all conditions from three independent experiments. (D) Distribution of spindle angles in control and PP2-treated metaphase HeLa cells on FN and N-Fc. Means ± SEM: control on FN, 7.402° ± 0.5001° (N = 60); control on N-Fc, 6.428° ± 0.5170° (N = 60); PP2 treated on FN, 16.075° ± 1.304° (N = 60); PP2 treated on N-Fc, 13.84° ± 0.9846° (N = 60). Data were analyzed by the Mann-Whitney test (****P < 0.0001). N, number of metaphase cells, from each condition, from three independent experiments. (E) Representative optical sections at the spindle poles and side projections of control and PP2-treated metaphase HeLa cells on FN and N-Fc acquired at the pane of spindle poles. N = 240 total number of metaphase cells across all conditions from three independent experiments. Scale bars, 10 μm.

We went on to address the role of another CMC member, Src. Src is required for spindle responses to both planar substrates and micropatterns (18, 22). We seeded HeLa cells on N-cadherin Fc and FN and treated them with the well-characterized Src inhibitor 4-amino-5-(4-chlorophenyl)-7-(dimethylethyl)pyrazolo[3,4-d]pyrimidine (PP2) (51). Cells treated with the Src inhibitor displayed spindle misorientation on both substrates (Fig. 6, D and E), suggesting that Src’s role in this process is also independent of its role in cell adhesion. These data, combined with data revealing a similar function for integrin β1, show that the role of the CMC in spindle orientation responses to planar substrates is independent of the roles of its member proteins in cell-ECM interactions.

DISCUSSION

Precise regulation of mitotic spindle orientation is essential for a broad spectrum of processes in multicellular organisms, including cell fate decisions, tissue morphogenesis and homeostasis, and epithelial integrity (1, 2). The evolutionarily conserved Gαi-LGN-NuMA complex plays a central role in this process (10, 12, 35). Specifically, this complex is established at the cortex of mitotic cells and becomes polarized, leading to the creation of specific anchoring regions for astral MTs, through NuMA’s interactions with dynein (4, 10, 39). The dynein motor complex subsequently exerts pulling forces on astral MTs to position the spindle (4, 6, 52). The polarization of the capture complex can be a consequence of polarizing molecular cues or asymmetric mechanical cues (10, 12, 35, 40). This is the case in HeLa cells, which display LGN enrichment initially throughout the cortex when plated on FN, with the distribution becoming subsequently restricted, corresponding to the final orientation of the spindle. However, when HeLa cells are plated on micropatterns, which generate an asymmetric distribution of force on the cortex, then LGN displays polarity before spindle orientation is established (53).

The role of integrins in spindle orientation is well established, and spindle orientation parallel to the plane of adhesion, as well as in response to micropatterned substrates, is believed to rely on integrin signaling (11, 22). Our findings indicate that integrin-based cell-ECM adhesion is dispensable for the generation of spindle orientation cues provided by cell adhesion. This is true for both orientation parallel to the plane of adhesion and spatial cues that dictate spindle orientation within the plane of attachment. We demonstrated that cells attaching on N- and E-cadherin Fc substrates, under serum-free conditions and in the absence of integrin ligands, can orient their spindles as effectively as cells attached on FN. This suggests that cell adhesion and adhesion geometry only provide mechanical cues to the dividing cell and that there is no molecular link or signaling from the adhesive complexes that is required for spindle orientation during division. Thus, any mechanical anchoring to the substrate that would lead to the formation of RFs upon cell rounding would be expected to effectively guide spindle orientation in space. Previous work revealed that orientation within the plane of attachment of epithelial cells can be influenced by the presence of E-cadherin junctions, with the spindle aligning with such complexes. This was attributed to a direct interaction of LGN with E-cadherin (16, 20, 42). Although we were unable to detect any ventral (basal) enrichment of LGN in interphase cells attached on cadherin-patterned substrates, we cannot preclude the possibility that such an interaction does take place. However, the fact that during mitosis, both LGN and NuMA displayed exclusively lateral localization with no ventral enrichment on N- and E-cadherin substrates shows that the spatial distribution of LGN, at least during mitosis, is not primarily determined by the distribution of the cadherin receptors. Thus, although it is possible that AJ formation on the lateral aspect of a cell can refine or bias LGN’s spatial distribution, either directly through binding to clustered cadherins or indirectly through force exertion at specific regions of the cortex, it is clearly not the major determinant for the localization of LGN. This is consistent with the observation that only a minor fraction of LGN is at the cortex in interphase cells, with cortical recruitment increasing during mitosis, a time during which LGN cannot bind cadherins because binding to NuMA and cadherins is mutually exclusive (20, 42, 54).

The spatial cues provided by adhesion of a cell to its substrate, irrespective of whether this is integrin based (ECM) or not (cadherin based), define interphase cell shape and determine the force distribution on the cortex during mitosis. How these cues are sensed and transmitted to the spindle is unclear; however, the CMC appears to play a central role in this process. Integrin β1 becomes activated on the lateral cortex of mitotic cells, at areas away from the cell-ECM interface, and this activation leads to the recruitment of several FA proteins, including FAK, paxillin, Src, and p130Cas, forming the CMC (18). Signaling from this complex is required for spindle responses to force both in vitro and in vivo (18). Differentiating between ligand-dependent and ligand-independent roles of integrins is not trivial; however, evidence from previous work suggested that the CMC is established in the absence of an integrin ligand. Our findings support the notion that the CMC is required for spindle orientation responses in an ECM adhesion–free context, effectively uncoupling the role of integrins and a subset of FA proteins in cell adhesion from their role in spindle orientation. Specifically, our work provides clear evidence of an adhesion-independent, and possibly a ligand-independent, role for integrins in the sensing of mechanical cues that guide spindle orientation. Both integrin activation and CMC proteins, like p130Cas and Src, are necessary for responses of the spindle when cells are attached on N-cadherin Fc, despite having no role in cell adhesion and spreading on this substrate. These data are in agreement with data from Xenopus showing that inhibition of CMC proteins in the animal cap elicits spindle orientation defects in both the deep layer and the outer epithelial cell layer of gastrula-stage embryos, whereas elimination of FN elicits defects that are restricted to the deep layer, which is in contact with the ECM (18, 24, 25, 55). Thus, although the outer cells can divide correctly in the absence of ECM components, they cannot divide correctly in the absence of CMC signaling, suggesting that CMC member proteins transduce orientation-based signals in cells that are not in contact with the basement membrane and do not require ECM components for correct spindle orientation (18). The mechanism by which the CMC influences spindle capture on the cortex still remains unclear. Neither the cortical recruitment nor the eventual polarization of LGN or NuMA is affected when integrin β1 is inhibited or in the absence of proteins like FAK or p130Cas (18). Intrinsic signals, similar to the previously described Ras-related nuclear protein GTP gradient, eventually lead to asymmetric distribution of the complex irrespective of extrinsic cues provided by mechanical stimuli (40, 56). This is supported by the finding that spindle misorientation under CMC inhibition is accompanied by polarized LGN and NuMA, corresponding to the final spindle position (18).

In conclusion, we propose that cell division on planar substrates is independent of cell-ECM interactions and that previously described ligand-dependent involvement of integrins in this process stems from a purely mechanical role in cell attachment, cell spreading, and the determination of adhesion geometry on such substrates. Attachment of cells on N-cadherin Fc provides precisely the same information to the cell, both with respect to division parallel to the plane of attachment and with respect to cortical force distribution–dependent orientation within that plane. Last, integrins and other CMC-associated proteins retain a central role in spindle orientation in ECM-free conditions, strengthening the notion that activation of integrins at the cell cortex by mechanical stimuli is a central force-sensing mechanism and likely involved in additional morphogenetic processes.

MATERIALS AND METHODS

Cell culture, transfection, and plating

HeLa and U2OS cell lines (American Type Culture Collection) were cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum (FBS). p130Cas−/− and p130Cas-reconstituted cells (provided by S. Cabodi) were cultured in DMEM with 10% FBS and 1% nonessential amino acids. All cell lines were transfected with the indicated plasmids using Lipofectamine 2000, according to the manufacturer’s instructions. For all experiments, cells were mechanically detached in MACS buffer [containing 1× phosphate-buffered saline (PBS) (pH 7.4), 2 mM EDTA, and 0.5% bovine serum albumin] and seeded in DMEM for 30 min before fixation.

Surface coating and cell adhesion on substrates

All cell adhesion experiments were performed using previously described protocols with modifications (30, 57). Briefly, coverslips were charged using piranha solution for 1 hour at 25°C, washed three times with water, and dried at 100°C for about 10 min. Then, coverslips were treated with 20% (3-aminopropyl) triethoxysilane (Sigma-Aldrich) in isopropanol for 90 min, washed three times with isopropanol, and dried at 100°C for about 30 min. For FN coating, silanized coverslips were incubated with bovine plasma FN (10 μg/ml; Invitrogen) in 1× PBS for 60 min at 37°C. For N-cadherin and E-cadherin substrate generation, silanized coverslips were initially incubated with goat anti-human IgG (10 μg/ml; Sino Biological) in 1× PBS for 1 hour at 37°C. Coverslips were subsequently incubated in human N-cadherin Fc (Sino Biological) or E-cadherin Fc (Sino Biological) in 1× PBS at a concentration of 10 μg/ml for 60 min at 37°C.

Micropatterned substrate generation

For the generation of N-cadherin Fc and FN micropatterns, circular glass coverslips were sonicated for 15 min with heat, washed with distilled water and isopropanol, and dried at room temperature. Coverslips were charged using piranha solution for 30 min at 25°C, washed three times with distilled water, and dried at 100°C for about 10 min. Then, coverslips were exposed to ultraviolet (UV)/ozone (185 and 254 nm) for 10 min, followed by incubation in PLL(20)-g[3.5]-PEG(2) (100 μg/ml; SuSoS Surface Technology) for 30 min at 37°C. L- and linear-shaped patterns were generated by exposure of coverslips to UV/ozone (E511, Ossila) (185 and 254 nm) for 10 min, using a custom photolithography mask purchased from JD Photo Data. Coverslips were washed with 1× PBS, dried at room temperature, and silanized using vapor (3-aminopropyl) triethoxysilane (Sigma-Aldrich) for 5 s, followed by washes with distilled water. Coverslips were then incubated with goat anti-human IgG-Fc antibody (10 μg/ml; Sino Biological) or FN for 60 min at 37°C. Anti-human IgG-Fc coverslips were subsequently incubated with human N-cadherin Fc (10 μg/ml; Sino Biological) in 1× PBS for 60 min at 37°C.

DNA constructs

The pLZRS-ires-p130Cas wild-type GFP construct was provided by S. Hanks (50, 58), and the pCDNA3.1 E-cadherin–GFP wild-type construct and the LGN-C′ membrane cherry construct (pTK38_mCherry-LGN-C) were purchased from Addgene (nos. 28009 and 46346).

Immunostaining

Cells on coverslips were washed with 1× PBS and fixed with 4% paraformaldehyde in PBS for 12 min. Fixation was followed by the addition of 50 mM glycine in PBS (pH 8), and then, cells were permeabilized using 0.2% Triton X-100 in PBS for 15 min. Cells were blocked using 10% normal donkey serum in PBS for 30 min and incubated with primary antibodies for 90 min at 25°C. For active integrin β1 staining, cells were first blocked in 10% normal donkey serum, incubated with integrin β1 antibodies for 90 min at 25°C before Triton treatment. Cells were then permeabilized with 0.03% Triton X-100 for 6 min, blocked for 15 min, and stained with primary antibodies for 90 min. For cortical staining of CMC member proteins, cells were permeabilized with 0.03% Triton X-100 for 6 min. The primary antibodies used are as follows: active integrin β1 9EG7 (1:700; 550531, BD Pharmigen), β-tubulin (1:200; E7, Developmental Studies Hybridoma Bank), α-tubulin (1:500; sc-53030, Santa Cruz Biotechnology), β-catenin (1:700; sc-7199, Santa Cruz Biotechnology), β-catenin (1:500; 11279-H20B, Sino Biological), NuMA (1:500; ab36999, Abcam), paxillin (1:1000; 610569, BD Biosciences), paxillin (1:500; 10029-1-Ig, Proteintech), LGN (1:1000; ABT174, Millipore), FAK (1:500; 66254-I-Ig, Proteintech), p130Cas (1:500; sc365200, Santa Cruz Biotechnology), and Src (1:500; sc8056, Santa Cruz Biotechnology). Cells were washed several times with PBS and then incubated with secondary antibodies (1:500; Invitrogen, Jackson) for 45 min at 25°C. Mounting was performed using ProLong diamond antifade (Invitrogen).

Drug and antibody treatments

HeLa cells were treated with 1.5 mM EGTA for 20 min at 37°C before fixation. For integrin inhibition, cells were incubated with the AIIB2, P4C10, and P5D2 integrin β1 inhibitory antibodies (1:100; Hybridoma Bank) for 20 min at 37°C before fixation. For Src inhibition, cells were allowed to attach and spread on substrates for 20 min and then treated with 4 μM PP2 inhibitor (Sigma-Aldrich) for 30 min before fixation. For secretion blockage, cells were treated with 20 mM brefeldin-A (Santa Cruz Biotechnology) for 2 hours, mechanically disrupted with MACS buffer containing 20 mM brefeldin-A, and seeded in the presence of brefeldin-A for 30 min.

Imaging

Imaging was performed on a laser scanning confocal microscope (LSM 710, Zeiss) using ZEN10, Zeiss AxioImager Z1 microscope using a Zeiss AxioCam MR3 and AxioVision 4.8, and Motic AE31 Elite inverted phase contrast microscope.

Quantification and statistical analysis

For all experiments, at least three independent replicates were performed (defining as replicate each coverslip with attached cells). Quantification of spindle orientation was performed by determining the orientation of the division axis of metaphase cells. Measurements were performed with ImageJ or ZEN10, using the Z-projections from images of the cells. The angle between the line connecting the two spindle poles of a cell and the line parallel to the plane of the substrate was measured. Statistical analyses were performed on collected data using the nonparametric Mann-Whitney test, as judged by the D’Agostino-Pearson normality test, using GraphPad Prism.

SUPPLEMENTARY MATERIALS

stke.sciencemag.org/cgi/content/full/13/620/eaax9940/DC1

Fig. S1. Planar cadherin substrates drive cell divisions parallel to the plane of adhesion.

Fig. S2. Integrin activation is necessary for mitotic cell responses to planar substrates.

REFERENCES AND NOTES

Acknowledgments: We thank S. Hanks and S. Cabodi for providing the p130Cas constructs and cell lines. We also thank V. J. Promponas for the evaluation of the statistical analysis of this work and M. Papacharalampous for establishing silane-based micropatterning. Funding: This work was cofunded by the European Regional Development Fund and the Republic of Cyprus through the Research Promotion Foundation (project no. INFRASTRUCTURE/1216/0060). Author contributions: O.A. and R.H. performed the experiments. O.A., R.H., and P.A.S. designed the experiments, analyzed the data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or in the Supplementary Materials.

Stay Connected to Science Signaling

Navigate This Article