CD2AP links actin to PI3 kinase activity to extend epithelial cell height and constrain cell area

Epithelial cells line all organs, body cavities, lumens, and ducts. They mediate the selective transport of materials from one side of the epithelial barrier to the other. To perform these functions, epithelial cells must build three distinct membranes, apical, lateral, and basal, each of which performs different functions. The composition of each membrane domain is tightly regulated and varies between different types of epithelial cells to meet physiological demands (Caceres et al., 2017). The size of each membrane domain also varies between cell types in accordance with the underlying physiology, and while epithelia can selectively control the size of each membrane domain, it is the height of the lateral membrane that is used to categorize epithelia into squamous versus cuboidal versus columnar morphologies (Lowe and Anderson, 2015).

The height of the lateral membrane is connected to cell function. For example, type 1 alveolar epithelial cells in lung are very thin (squamous) to facilitate gas exchange (Bertalanffy and Leblond, 1955; Guillot et al., 2013). Such cells build a short lateral membrane. Other cell types, like the transporting epithelial cells in the kidney, build taller lateral membranes in order to increase the number of transport proteins in the lateral membrane to increase transcellular flux of specific solute molecules through the epithelial barrier (Larsson et al., 1983; Zhai et al., 2003, 2006). While the height of lateral membrane is closely connected to cell function, little is known about what controls it. This is bound to be a complicated problem involving specific transcription factors, adhesion molecules, polarized membrane trafficking, cytoskeletal organization, and the signaling feedback loops that control them (Tang, 2017).

Earlier work identified phosphoinositide 3-kinase (PI3K) and its product, Pins(3,4,5)P₃, as important determinants of lateral membrane height (Gassama-Diagne et al., 2006; Jeanes et al., 2009), but how PInsP₃ production leads to extension of the lateral membrane is not known. Modulation of the actin cytoskeleton is a strong possibility. Certain actin binding proteins, such as ankyrin, spectrin (He et al., 2014; Jenkins et al., 2015; Kizhatil et al., 2007), and tropomodulin (Weber et al., 2007), are important for maintaining the height of lateral membranes. Actin assembly factors, including EVL, CRMP1, Arp2/3, WAVE2, and myosin 1c, are also important for extension of the lateral membrane (Kannan and Tang, 2015, 2018; Yu-Kemp et al., 2017). Finally, rho and p120 catenin, which helps control rho activity (Noren et al., 2000), are both implicated in lateral membrane extension (Yu et al., 2016). Since PI3K plays an important role in building the actin cytoskeleton in different cell types (Cain and Ridley, 2009), it is possible that PI3K’s effect on cell height is due, at least in part, to its effects on actin.

The connection between PI3K and actin is best understood in amoeboid cells, where PI3K activation triggers Arp2/3-dependent actin polymerization to generate a protruding leading edge as part of directional cell migration toward a chemotactic signal (Cain and Ridley, 2009; Funamoto et al., 2001, 2002; Hannigan et al., 2002; Weiger and Parent, 2012). Far less is known about whether PI3K plays a role in actin assembly in normal epithelial cells in which cell motility is largely suppressed. Nevertheless, nontransformed epithelial cells in culture maintain fast actin assembly/disassembly turnover dynamics despite the fact that cells are not moving (Tang and Brieher, 2012). Much of the actin assembly occurring at cell–cell junctions and within the actin cortex is Arp2/3 dependent (Tang and Brieher, 2012, 2013; Van Itallie et al., 2015; Yu-Kemp et al., 2017), and loss of these actin networks often leads to decreased cell height and conversion from a cuboidal to squamous morphology (Tang and Brieher, 2012; Yu-Kemp et al., 2017).

Previously, we identified CD2AP as another protein necessary for assembling the actin cytoskeleton at apical cell–cell junctions as well as the apical actin cortex (Tang and Brieher, 2013). CD2AP’s positive effect on actin assembly in cells is paradoxical, because CD2AP itself suppresses actin polymerization in vitro, but this could be an epiphenomenon (Tang and Brieher, 2013). How CD2AP promotes actin assembly in cells is complicated, because CD2AP is modular protein that binds to several other partners through CD2AP’s three SH3 domains, its proline-rich domain, and other small motifs and regions (Dustin et al., 1998). Some of CD2AP’s binding partners include cortactin (Lynch et al., 2003; Zhao et al., 2013), anillin (Monzo et al., 2005), and capping protein (Bruck et al., 2006; Hernandez-Valladares et al., 2010; Hutchings et al., 2003). These interactions might explain how CD2AP promotes actin assembly in cells. Alternatively, CD2AP’s potent effect on actin assembly in cells might be due to interactions with PI3K. CD2AP binds to the p85α regulatory subunit of PI3K, and it helps recruit PI3K to the plasma membrane in podocytes (Huber et al., 2003; Ren and You Yu, 2016). Given the importance of CD2AP in junctional actin assembly and epithelial barrier function (Cochran et al., 2015; Tang and Brieher, 2013), inherited kidney disease (Kim et al., 2003; Shih et al., 1999), and possibly Alzheimer’s disease (Cochran et al., 2015 ,Hollingworth et al., 2011; Naj et al., 2011), we reinvestigated the role of CD2AP in junctional actin assembly. That line of investigation led us to discover an important role for CD2AP in phosphatidylinositol (3,4,5)-trisphosphate (PI(3,4,5)P₃) synthesis and epithelial cell height.

Previous work showed that disrupting actin assembly along lateral membranes and within the apical actin cortex by depleting MDCK cells of CRMP1, EVL, WAVE2, or Arp2/3 results in large changes in cell area with a reduction in cell height (Yu-Kemp et al., 2017). CD2AP is also responsible for F-actin accumulation at cell–cell junctions, along lateral membranes, and in the apical actin cortex (Tang and Brieher, 2013). To test if CD2AP is also involved in epithelial cell morphology regulation, shCD2AP-MDCK cells were generated. The morphology of CD2AP depleted cells was compared with that of control cells using fluorescently tagged phalloidin to stain for F-actin, which also allowed us to compare actin polymer levels as well as view the overall size and shape of the cells. WT-MDCK cells showed brighter F-actin staining at cell–cell borders and within the apical membrane cortex than CD2AP-depleted cells. (Fig. 1 A). Immunofluorescence imaging of CD2AP showed reduced CD2AP content in the shRNA-treated cells (Fig. 1 A, middle). Reduction of CD2AP protein levels by the shRNA was further confirmed by Western blotting, which showed an overall ∼50% reduction in CD2AP relative to control cells (Fig. 1 B). The large drop in actin and CD2AP fluorescence intensity made it difficult to see the outline of the cells and the organization of the actin cytoskeleton that remains following CD2AP knockdown. Therefore, we rescaled the images of the CD2AP-knockdown cells to make them brighter and easier to see the large change in cell area that accompanies loss of F-actin (Fig. 1 A, bottom row). Quantification of F-actin levels from phalloidin-stained cells showed a strong correlation between the level of CD2AP and the level of F-actin both in the apical cell cortex (Fig. 1 C) and at cell–cell boundaries (Fig. 1 D), with the reduced levels of CD2AP in the knockdown cells producing further reductions in F-actin.

Confocal microcopy was then used to assess the morphological changes upon CD2AP knockdown. Cell height and spreading area were measured from xz and xy projections, respectively, of WT-MDCK and shCD2AP cells using F-actin to mark the cell borders. xz projections showed that CD2AP depletion decreased cell height from an average of 3.8 ± 0.6 µm in WT cells to an average of 2.2 ± 0.3 µm in CD2AP-depleted cells corresponding to an ∼40% reduction in cell height (Fig. 1, E and F). The reduction in lateral membrane height was accompanied by an increase in cell spreading area by approximately threefold from an average of 161 ± 69 µm² in control cells to 470 ± 112 µm² in CD2AP-depleted cells (Fig. 1 G). These data show that CD2AP is required for maintaining F-actin and the proportional relationship between apical–basal surfaces relative to the lateral membrane.

The change in cell height could be a downstream consequence of there being fewer cells in the monolayer. To test if changes in cell height are due to compressive forces from cell packing density, we compared cell area and cell height in cultures of WT versus shCD2AP cells plated at extremely low seeding density. An ∼5.6-fold increase in cell expansion (xy axis; Fig. 2, A, B, and D) and 1.5-fold decrease in cell height (Fig. 2, C and E) was detected upon CD2AP knockdown in the absence of cell crowding. Thus, CD2AP regulates cell shape independent of cell density.

The flattened cells are reminiscent of PI3K inhibition phenotypes (Gassama-Diagne et al., 2006; Jeanes et al., 2009), suggesting that CD2AP’s effect on lateral membrane height might be linked to PI3K activity and PI(3,4,5)P₃ synthesis. To further investigate the possible connection between CD2AP and PI3K, we used immunofluorescence to examine the distribution of CD2AP and the p85α subunit of PI3K in confluent sheets of MDCK cells. Both CD2AP and p85α were organized into puncta that colocalized along cell–cell contacts in mature MDCK cells (Fig. 3 A). Line scans along the cell–cell boundary confirmed the overlap of CD2AP with PI3K p85α (Fig. 3 B).

Previous results used coimmunoprecipitation and pull-down assays to demonstrate an interaction between CD2AP and PI3K p85α (Huber et al., 2003). We performed microscale thermophoresis (MST) measurements to confirm the association between CD2AP and PI3K p85α and to obtain a dissociation constant. These experiments showed that the N-terminal portion of CD2AP (residues 1–475) bound to PI3K p85α (residues 1–315) with K_d = 0.94 ± 0.41 µM (n = 3; Fig. 3, C and D). Thus, CD2AP physically associates with PI3K p85α and colocalizes with it at cell–cell boundaries in epithelial cells.

We used immunofluorescence to test if CD2AP was necessary for recruiting PI3K to membranes and cell–cell boundaries. As shown in Fig. 4 A, CD2AP knockdown diminished the PI3K p85α signal both underlying the apical cortex and along cell–cell borders. Quantification confirmed the strong correlation between CD2AP and PI3K p85α (Fig. 4, B and C). Similarly, CD2AP knockdown also decreased PI3K p110γ accumulation at cell borders (Fig. 4, D and E). As shown in Fig. 4 F, the p110γ catalytic subunit of PI3K coimmunoprecipitates with CD2AP. Western blots showed the total levels of p85α and 110γ were unaffected by CD2AP depletion (Fig. S1). The dim PI3K signal in CD2AP-knockdown cells is therefore due to its diffuse distribution over a larger area as opposed to a decrease in expression. Thus, CD2AP is necessary for PI3K p85α and p110γ membrane recruitment in MDCK cells.

Loss of PI3K at membranes was accompanied by a decrease in the amount of PI(3,4,5)P₃ at lateral membranes as shown by the loss of PH-GFP fluorescent signal at cell boundaries (Fig. 4 G). Line scans across cell boundaries showed a peak of CD2AP fluorescence in register with the PH-GFP peak in WT cells (Fig. 4 H). In contrast, both CD2AP and PH-GFP signals were diminished in CD2AP-depleted cells, and fluorescence intensity across cell boundaries was flat (Fig. 4 H). Population analysis revealed that PH-GFP signals peak in ∼70% of WT-MDCK cells and only ∼13% of shCD2AP-MDCK cells (Fig. 4 I). Fig. 4 J shows aggregate data from different cells plotting the ratio of the peak PH-GFP intensity at the cell–cell boundary to the average intensity of PH-GFP in the middle of the cell. Together, the results show that CD2AP is necessary for recruiting PI3K and synthesizing PI(3,4,5)P₃ at cell–cell boundaries.

Previous results have shown that the height of the lateral membrane and constraint of the apical and basal surfaces depends on actin nucleation and elongation factors (Yu-Kemp et al., 2017). PI3K is an important activator for actin assembly in many cell types, but its role in maintaining actin assembly in mature epithelial cell sheets is not known. To test this, we treated cells with the PI3K inhibitor LY294002 and used fluorescence microscopy to image actin, CD2AP, and PI3K p85α (Fig. 5 A). Z stacks were collected to measure cell height (Fig. 5 B). The results showed that cell area increased (Fig. 5 C) and cell height decreased (Fig. 5 D) in the presence of the inhibitor, as expected from previous results (Gassama-Diagne et al., 2006; Jeanes et al., 2009). Quantification of fluorescence intensities showed that inhibiting PI3K caused a large drop in both F-actin and PI3K p85α levels at cell boundaries, but the level of CD2AP was unaffected (Fig. 5 E). These results showed that sustained levels of F-actin at cell–cell boundaries in mature epithelial sheets depends on continuous PI3K activity and suggested that PI3K maintains lateral membrane height by activating actin polymerization.

Actin assembly promotes PI3K activity in other systems (Gassama-Diagne et al., 2006; Hawkins et al., 1995; Benard et al., 1999; Weiner et al., 2002; Srinivasan et al., 2003; Welch et al., 2003). We hypothesized that a similar, reciprocal relationship between actin and PI3K activity might exist in epithelial cells. To test this, we treated WT-MDCK cells with latrunculin B (LatB) for 2 h before staining the cells for actin, CD2AP, and the P85α subunit of PI3K (Fig. 6 A). Quantification showed that actin, CD2AP, and PI3K levels significantly dropped upon LatB treatment (Fig. 6, B and C), and this was accompanied by expansion of the cell-spreading area (Fig. 6 D), indicating that actin polymerization can be both upstream and downstream of CD2AP and PI3K (Fig. 6 E).

To better understand the relationships among CD2AP, actin, PI3K, and cell proportions, we performed a structure–function analysis on CD2AP. CD2AP is a modular protein consisting of three SH3 domains at the N-terminus followed by a proline-rich segment, binding sites for capping protein, and finally a coiled-coil domain near the C-terminus (Fig. 7 A). We made various deletions in CD2AP and then tested the extent to which these deletions could rescue CD2AP-knockdown phenotypes. A schematic of the deletion mutants is shown in Fig. 7 A. Representative images of the actin cytoskeleton in rescued MDCK cells are provided in Fig. 7 B. For quantification, we plotted the actin fluorescence intensity as a function of the fluorescence intensity of the transfected CD2AP construct (Fig. 7, C–H) and also the cell area for each of the deletion constructs (Fig. 7 I). First, we demonstrated that full-length CD2AP rescued both actin assembly and cell area in CD2AP-depleted cells (Fig. 7, B, C, and I). Transfected cells show cytosolic staining of the epitope tag and its accumulation at cell boundaries. Similarly, CD2AP 1–475, which contains only the SH3 domains and the proline-rich domain, rescued actin and appropriate cell area (Fig. 7, B, F, and I). CD2AP constructs lacking either the capping protein or cortactin-binding sites also rescued actin and cell area (Fig. 7, B, D, E, and I). The levels of CD2AP 336–605 were equal to endogenous levels of CD2AP in WT cells. The other constructs we overexpressed ∼24–53% relative to the normal level of endogenous CD2AP (Fig. 7, C–F). Linear regression analysis showed that the correlations between the expression levels of these CD2AP constructs with F-actin levels were the same as endogenous CD2AP with F-actin levels in WT-MDCK cells. This suggests that the rescuing constructs restore actin assembly using a mechanism similar to the endogenous CD2AP protein. The strong correlations were seen for every CD2AP construct capable of rescuing actin assembly in CD2AP-knockdown cells with the exception of the CD2AP mutant (Δ378–397/ΔPR2). In this case, the regression relationship suggested that the CD2AP mutant (Δ378–397/ΔPR2), which omitted the cortactin-binding site (Lynch et al., 2003), was less efficient in rescuing F-actin (Fig. 7 E), implying that the CD2AP–cortactin interaction contributes to F-actin accumulation in vivo.

We also identified two informative deletion constructs that failed to rescue cell area. The SH3 domains alone failed to rescue actin assembly or cell area (Fig. 7, B, H, and I). Interestingly, the SH3 domains alone also failed to target to the plasma membrane, whereas all the other constructs that managed to rescue the CD2AP-knockdown phenotype localized to cell–cell boundaries (Fig. S2). In addition, a C-terminal construct lacking the SH3 domains but containing the proline-rich segment and capping protein–binding site rescued actin assembly but was comparatively inefficient in rescuing cell area (Fig. 7, B, G, and I).

We compared the extent of rescue of CD2AP 1–475 to that of CD2AP 336–605, because the N-terminal domain fully rescues the knockdown phenotype while the C-terminal construct exhibits defect in controlling cell dimensions. Fig. 7 J shows fluorescent images of cells from these two rescue experiments side by side. To quantify the relationship between the expression of CD2AP 1–475 or 336–605, we plotted the fluorescent intensity of the deletion constructs versus the actin intensity or cell area. As shown in Fig. 7 K, comparable levels of 1–475 or 336–605 yield comparable levels of F-actin. However, as shown in Fig. 7 L, comparable levels of the two deletion constructs yield different cell areas with 336–605–expressing cells being approximately twofold larger in area than cells expressing the same amount of 1–475. CD2AP 1–475 retains the SH3 domains, which bind to PI3K p85α (Huber et al., 2003), while the 336–605 construct lacks the SH3 domains (Fig. 7 M). As expected, CD2AP 1–475 concentrated PH-GFP fluorescence signals better along cell borders than CD2AP336-605 in shCD2AP-MDCK cells (Figs. 7 Nand 5 O). Our data further support the idea that maintaining proper epithelial cell architecture requires CD2AP to bind to and recruit PI3K. Intriguingly, it appears that F-actin assembly per se is not sufficient to support lateral membrane extension, and there might be multiple redundant mechanisms through which CD2AP maintains the actin cytoskeletal network that does or does not involve binding to cortactin.

Next, we tested if CD2AP binding to and recruiting PI3K to cell membranes is sufficient to restore normal cell-spreading area in CD2AP-deficient cells. We first confirmed that the three SH3 domains of CD2AP (residues 1–335) interact with PI3K p85α with a K_d = 0.35 ±0.31 µM (n = 3; Fig. S3, A and B). Because the CD2AP 1–335 aa fragment did not localize properly to the membrane (Fig. S3 C), we fused it at the C-terminus to the EGFP-CAAX-box motif and expressed it in shCD2AP-MDCK cells (Fig. S3 D). Our results showed that targeting the SH3 domains of CD2AP to the plasma membrane restored lateral F-actin, PI3K levels, and cell dimensions (Fig. 8, A–C; and Fig. S4). Quantification showed that CD2AP-1-329-EGFP-CAAX expression fully restored F-actin (Fig. 8 D) and PI3K p110γ (Fig. 8 E) intensity along cell borders, and induced 4.9-fold reduction of cell spreading area (from 1,100 ± 447 µm² to 226 ± 80 µm²) in CD2AP-knockdown MDCK cells (Fig. 8 F). To test if recruitment of PI3K to the plasma membrane is sufficient to rescue phenotypes resulting from CD2AP depletion, we targeted the catalytic subunit of PI3K to the plasma membrane using a C-terminal CAAX sequence (p110γ-CAAX; McFall et al., 2001). Cells expressing this construct were difficult to maintain in cell monolayers; therefore, we examined sparsely plated cells. Under such conditions, p110γ-CAAX rescued cell area (Fig. 8, G and H) and F-actin levels (Fig. 8 I) in CD2AP-depleted cells. These results show that a major function of CD2AP in epithelial cell morphology is to recruit PI3K to the membrane, but with the caveat that these membrane-targeting constructs somehow disrupt normal formation of epithelial sheets.

Previous results showed that CD2AP suppresses the rate of actin assembly in vitro (Tang and Brieher, 2013), but whether this activity is important for CD2AP function in cells is not known. To better understand if there is any relationship between CD2AP-mediated suppression of actin polymerization and CD2AP function in cells, we tested the effect of the various CD2AP deletion mutants on actin assembly in vitro. Wide-field fluorescence imaging was used to measure actin polymerization rates in the absence (Fig. 9 A) or presence of CD2AP 1–335 (Fig. 9 B), full-length CD2AP (Fig. 9 C), or CD2AP 1–475 (Fig. 9 E). Rates of actin polymerization were determined from kymographs and plotted (Fig. 9, D and F), which showed that all of these CD2AP constructs inhibited actin assembly. Using MST, we found that CD2AP does not bind to actin monomers at a concentration below 30 µM (Fig. 9, G and H). As a positive control, MST showed that profilin binds to actin monomer with dissociation constant of ∼0.15 µM (Fig. 9, I and J) in agreement with earlier measurements (Vinson et al., 1998). The MST data show that CD2AP is not an actin monomer–sequestering protein, in agreement with previous results (Tang and Brieher, 2013). Pyrene actin assembly was used to confirm that one of the CD2AP constructs, CD2AP 1–475, inhibited actin polymerization in a dose-dependent manner (Fig. 9 K). From the pyrene assays, CD2AP 1–475 was found to inhibit actin assembly with a half maximal inhibitor concentration of ∼5 µM (Fig. 9 L), which is in agreement with results from single-filament imaging (Fig. 9 F). We also confirmed that the C-terminal half of CD2AP (336–605) also inhibits actin assembly (Fig. 9 M). From these results, we can conclude that the ability of CD2AP to suppress actin polymerization is not sufficient to account for the effect of CD2AP on membrane proportions, because 1–335 failed to rescue knockdown phenotypes (Fig. 7, B, H, and I), despite the fact that it inhibited actin filament (+) end assembly. Determining whether CD2AP-mediated suppression of actin filament assembly is necessary for controlling membrane proportions will require more detailed mutagenesis of the three SH3 domains and other sequences within the N-terminal half of the protein.

All epithelial cells have to build three different membrane domains, apical, basal, and lateral, each of which performs different functions. While much is known about how each membrane domain is built, little is known about how epithelial cells control the ratios of these three membrane domains. We have identified CD2AP as an important factor in controlling membrane proportions, and Fig. 6 E shows a speculative yet reasonable model for how CD2AP builds up the lateral membrane and restricts the area of the apical membrane. We place CD2AP above PI3K, because targeting the SH3 domains of CD2AP to the membrane rescued PI(3,4,5)P₃ accumulation, cell height, and actin assembly in CD2AP-depleted cells (Fig. 8, A–F). We envision the actin cytoskeleton as the downstream effector that converts the PI3K signal into the mechanical work necessary to build up the lateral membrane and constrain cell area. We draw a positive feedback arrow from actin to CD2AP that would yield more PI3K signaling and more actin assembly. Some form of positive feedback from actin to PI3K is justified by our observation that drugging actin assembly blocks PI3K activity at cell–cell boundaries. Actin-dependent activation of PI3K has been seen in several other systems (Gassama-Diagne et al., 2006; Hawkins et al., 1995; Benard et al., 1999; Weiner et al., 2002; Srinivasan et al., 2003; Welch et al., 2003). However, the molecular mechanisms connecting actin to PI3K activation are first being uncovered (Nguyen et al., 2016; Graziano et al., 2017). Drawing the feedback arrow from actin to CD2AP is speculative, but it is supported by the facts that CD2AP binds to both F-actin and PI3K, so more actin polymer mass would create more binding sites for CD2AP and PI3K near the plasma membrane. However, this mechanism is incomplete, because it does not explain why the three SH3 domains of CD2AP alone are not sufficient to rescue all the phenotypes. Rather, the three SH3 domains could rescue actin, PI3K signaling, and cell proportions only when they were experimentally targeted to the membrane. Therefore, additional sequences beyond the SH3 domains are required to properly target CD2AP to lateral membranes. An additional, obvious problem with the model is that it never comes to the steady state. Actin depolymerization, which tends to be proportional to the amount of actin polymer present (Theriot et al., 1992; Kueh et al., 2010), would at least put a brake on runaway lateral membrane production, but ultimately, we expect there to be more layers of signaling loops that control the central machinery shown in Fig. 6 E to establish the appropriate proportions of the three membrane domains. PTEN is a strong candidate for antagonizing the positive feedback loop, because it appears to limit cell height and expand apical area in certain epithelial cells in the prostate (Wang et al., 2006).

The actin cytoskeleton is emerging as a key effector system for building up the lateral membrane and constraining cell area in the plane of the apical and basal membranes. Previous results showed that the actin assembly factors CRMP1, EVL, WAVE2, and Arp2/3 are all necessary for maintaining cell height and constraining cell area (Yu-Kemp et al., 2017), so these proteins are the likely effectors linking CD2AP and PI3K signaling to actin polymerization and cell geometry. All of these actin assembly factors are associated with the formation of branched actin networks that tend to generate protrusive force (Yu-Kemp and Brieher, 2016; Yu-Kemp et al., 2017). Perhaps cells use the energy of actin polymerization to push the lateral membrane up, as hypothesized previously (Tang, 2017), in which case actin along the lateral membrane drives cell height. Alternatively, the actin in the apical membrane, which also depends on CD2AP and the actin assembly factors mentioned above, generates the force to lengthen the lateral membrane. If myosin-dependent contractile forces in the apical actin network were to be exerted on apical cell junctions, those forces could pull the lateral membrane up and pull the apical membrane in to constrain cell area (Tang, 2017). The apical contraction idea is attractive, because it helps explain the role of rho (Yu et al., 2016) and myosin (Kannan and Tang, 2018) in cell height control, and there is strong evidence for tensile forces acting on cell–cell junctions in MDCK cells (Kannan and Tang, 2015; Choi et al., 2016). On the other hand, the basal to apical actin flux seen in certain epithelial cells (Kametani and Takeichi, 2007) is consistent with the pushing idea, and pushing the lateral membrane up might explain why cell-height phenotypes are seen in individual cells and not just confluent sheets. Other mechanisms are possible, including actin-dependent membrane trafficking (Zeng et al., 2017). Whatever the precise mechanism, we are at least developing a list of candidate proteins and hypotheses to better explore squamous versus cuboidal versus columnar epithelial morphologies.

Antibodies to CD2AP were raised against a synthetic Cys-tagged peptide, EKKYFPIKPEEKDEKSVLEQKPSK, mapping within the C-terminus of canine CD2AP and affinity purified by Pacific Immunology. Antibodies to PI3K p85α (catalog no. 1637, mouse), p110γ (catalog no. 166365, mouse), and SBP (catalog no.101595, mouse) were purchased from Santa Cruz Biotechnology. GAPDH (catalog no. 60004-1-lg, mouse) was purchased from Proteintech. Secondary antibodies were obtained from Bio-Rad Laboratories (HRP anti-rabbit and HRP anti-mouse) and Invitrogen (Alexa Fluor 568 anti-rabbit, Alexa Fluor 647 anti-mouse, and Alexa Fluor 647 phalloidin). FITC-phalloidin, TRITC-phalloidin, and LatB were purchased from Sigma-Aldrich. Protein concentrations were determined by the Bradford or DC Protein Assay (Bio-Rad Laboratories).

pRS vectors expressing shRNAs targeting N-terminal region of CD2AP were generated as previously described (Tang and Brieher, 2013). pBabe puro pI3K p110 CAAX was obtained from Addgene (catalog no. 13339; McFall et al., 2001). pEGFP-Akt-PH was obtained from Addgene (catalog no. 39533; Watton and Downward, 1999). The coding sequences of human PI3K p85α (catalog no. OHu21143D; GenScript) corresponding to 1–315 aa and various human CD2AP constructs (FL-CD2AP 1–639 aa; CD2AP 1–335 aa, 1–475 aa, 336–605 aa; GFP-1-475 aa; and GFP-1-335 aa) were amplified by PCR (iProof) and cloned in-frame into the bacterial expression vector pET30a+ (EMD) with 6-His tag at the N-terminus for expression in BL21 cells (NEB). CD2AP residues 1–377 and residues 398–639 were fused together to make the FL-CD2AP ΔPR2 cortactin-binding mutant; CD2AP residues 1–474 and 513–639 were fused together to generate the FL-CD2AP Δ474-513 CP-binding mutant. FL-CD2AP (1–639 aa); CD2AP 1–335 aa, 1–329 aa, 1–475 aa, and 336–605 aa; FL-ΔPR2; and FL-Δ474-513 were subcloned into the G418-selectable mammalian expression vector pNTAPB (Agilent Technologies) with a streptavidin-binding peptide tag at the N-terminus for expression in CD2AP-knockdown MDCK cells. EGFP-CaaX was amplified using pEGFP-N1-CIBN-EGFP-Caax (a gift from Chandra Tucker, University of Colorado, Denver, CO) as template and cloned into pNTAPB-CD2AP-1-329 at the C-terminus. CD2AP 336–605 aa, 1–475 aa, and 1–335 aa was subcloned into puromycin-selectable GFP-pLenti-III-HA vector for transient expression in CD2AP-knockdown MDCK cells. All cloning products were verified by ACGT sequencing services.

MDCK cells (MDCK II) were maintained in MEM (Cell Media Facility, the School of Chemical Sciences, University of Illinois, Urbana-Champaign) supplemented with 5% FBS (Gemini Bio Products). WT and CD2AP-deficient MDCK cells were stably or transiently transfected with plasmids using PolyJet In Vitro DNA Transfection Reagent (SignaGen Laboratories) according to the manufacturer’s instructions. Cells were under puromycin (3 µg/ml) or puromycin (3 µg/ml) + G418 (6 µg/ml) selection 24 h after transfection for 14 d. For Western blots, MDCK cells were washed twice and lysed in lysis buffer containing 1% Triton X-100, 1 mM EDTA, 1 mM PMSF, and 1 mM AEBSF. After homogenization, cell lysates were centrifuged at 14,000 ×g for 15 min at 4°C, supernatant was collected, protein concentration was determined using the DC Protein Assay, and equal amounts of protein were separated by SDS-10% PAGE. The whole-cell lysates (10–20 µg each) were fractionated in 10% SDS–polyacrylamide gels, transferred to nitrocellulose membranes (GE Healthcare), and blotted with antibodies according to the established procedures. HRP-conjugated IgG antibodies (Bio-Rad) were used as the second antibody. The blots were either exposed to autoradiography film and visualized by chemiluminescence or visualized with iBright Imaging Systems (ThermoFisher Scientific). The amount of α-tubulin or GAPDH detected in the blots was used as an internal control to ensure equal loading of the samples. Three independent Western blots were performed, and representative data are shown.

For expression of recombinant truncation mutants of CD2AP, BL21 cells (NEB) were induced with 500 µM isopropyl β-D-1-thiogalactopyanoside for 7 h at 25°C. Cells were centrifuged at 5,000 rpm for 15 min and resuspended in 20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM EDTA, 40 µM AEBSF, and 10 mg Lysozyme. After a freeze/thaw cycle, lysed cells were centrifuged at 17,000 rpm for 1 h. The supernantant was loaded onto a nickel column (QIAGEN). The column was washed with 10-bed columns of 10 mM Tris, pH 8.0, and 500 mM NaCl; 10-bed columns of 10 mM Tris and 150 mM NaCl; and 20-bed columns of 10 mM Tris, 500 mM NaCl, and 20 mM imidazole, pH 8.0; and eluted with 10-bed volumes of 10 mM Tris, pH 8.0, 150 mM NaCl, and 200 mM imidazole. Protein elutions were concentrated using Centricon (Millipore) and purified by chromatography through a gel filtration column Superdex-200 (GE Healthcare) in 150 mM NaCl, 20 mM Hepes, and 10 mM β-mercaptoethanol. Recombinant actin (Brieher et al., 2004) and filamin (Brieher et al., 2004) were purified as previously described. Actin was labeled with fluorophores as previously described (Brieher et al., 2004).

MDCK cells grown on coverslips were rinsed twice in PBS and then fixed and permeabilized as previously described (Tang and Brieher, 2013). For Figs. 1 Eand 5 B, images were collected with LSM700 confocal microscope (Zeiss) with a NeoFluar 20×/0.5 NA or 63×/1.4 oil differential interference contrast lenses. The Z-section step size was 0.5 µm. Cell height and spreading area were measured randomly throughout the samples while viewing only the channel showing F-actin. Cell height was measured in ImageJ Orthogonal Views xz or yz sections. For other images, fixed cell fluorescence wide-field images were collected with an Axio Imager using AxioVision Release 4.7 with the Colibri illumination system (Zeiss) using an oil 63× (NA 1.4) or 20× (NA 0.8) objective. Images were acquired using a 1,000 × 1,000 charge-coupled device camera (ORCA-ER; Hamamatsu Photonics). Cell spreading area was measured based on the apical section circumference of individual cells. Composite images were generated using ImageJ software. Images were constructed in Adobe Illustrator. Quantifications and line scans were performed with ImageJ, and OriginLab was used for statistical analyses.

0.625 µM Alexa Fluor 647–G-actin (20% labeled) was mixed with FL-CD2AP, CD2AP 1–335, and CD2AP 1–475 at indicated concentrations, and 10 mg/ml bovine casein was allowed to polymerize in photo buffer (Nadkarni and Brieher, 2014; 50 mM KCl, 2 mM MgCl₂,1 mM EGTA, 5 mM Hepes, pH 7.2, 2 mM ATP, 0.2 mM 6-hydroxy-2,5,7,8-teramethylchroman-2-carboxylic acid [Trolox], 20 mM imidazole, 4.5 mg/ml glucose, and 0.2 mg/ml glucose oxidase) underneath glasses that were coated with actin-bundling protein α-actinin4. Images were collected every 5 s for 10 min at room temperature with a light microscope (Axio Imager) using the image system described above with the 63× objective lens (NA 1.4). Polymerization rates were determined from kymographs using ImageJ.

For Fig. 9 M, pyrene-actin polymerization was monitored in 96-well plates by fluorescence dequenching of pyrene (excitation, 365 nm; emission, 410 nm) upon initiation of actin polymerization in the absence or presence of 5 µM CT-CD2AP 336–605 aa with a SpectraMax M2 plate reader as previously described (Tang and Brieher, 2013).

For Fig. 9 K, the capping rates of the barbed ends of actin seeds were measured as previously described (Huang et al., 2005). Briefly, 2.5 µM unlabeled G-actin was prepolymerized in actin assembly buffer (50 mM KCl, 1 mM EGTA, 2 mM MgCl₂, 5 mM Hepes, pH 7.2, 40 mM imidazole, pH 7.0, and 1 mM ATP) to generate actin seeds. Actin seeds (2.5 µM) were mixed with CD2AP 1–475 aa at concentrations indicated in the figure legends before being diluted 10-fold into actin assembly buffer along with 2.5 µM 25% pyrene-labeled G-actin (total volume, 200 µl). The initial rates of polymerization were determined from the increase in pyrene fluorescence E_x365/E_m410 using ImageJ. Readings were normalized using OriginLab software.

MST was performed using a Monolith NT.115 instrument (Nanotemper Technologies). MST power was 40%, and excitation power was set to either 20% or 60%. For Fig. 3, C and D, a solution of PI3K p85α 1–335 aa was serially diluted from 68.5 µM to 2 nM in the presence of 20 nM GFP-CD2AP 1–475 aa. For Fig. S3, A and B, PI3K p85α 1–315 aa (from 3.05 nM to 200 µM) was titrated into GFP-CD2AP-1-335 aa (40 nM). For Fig. 9, G and H, CD2AP 1–475 was serially diluted from 38 µM to 1.16 nM in the presence of 100 nM Cy5-G-actin. For Fig. 9, I and J, profilin was serially diluted from 103 µM to 3.13 nM in the presence of 100 nM Cy5-G-actin. The samples were loaded into Monolith NT.115 Standard Treat Capillaries (catalog no. K002; Nanotemper Technologies) and measured in 20 mM Hepes, pH 7.5, with 50 mM NaCl and 0.05% Tween 20. Data were analyzed using Nanotemper Analysis software, v.1.2.101. Error bars represent SD from three independent experiments.

WT-MDCK cells were washed with ice-cold PBS twice, harvested, and hypo-osmotically shocked with 2 mM imidazole, pH 7.4; cells were then lysed in 1% Triton X-100 lysis buffer for 10 min and homogenized through a 25-guage needle on ice. After centrifugation at 14,000 ×g (15 min; 4°C), equal amounts of cell lysates were incubated for 1 h at 4°C with CD2AP antibody or preimmune serum (under rotary agitation), followed by incubation with 100 µl protein G-Sepharose beads for 2 h (under rotary agitation). The beads were washed three times with lysis buffer to remove nonspecific binding, and protein complexes were then eluted from the beads by boiling samples in loading buffer with denaturant SDS. Equal amounts of samples were then analyzed by Western blot.

All experiments were repeated at least three times. Data distribution was assumed to be normal, but this was not formally tested. One-way ANOVA was performed in OriginLab, P values are indicated in the figure legends, and differences were considered significant at P ≤ 0.001 (**) or P ≤ 0.05 (**). Coefficient of determination R² and fitted curves were calculated and generated by Excel and provided in the graphs.

Fig. S1 shows that CD2AP knockdown does not affect the total amount of PI3K p85α and PI3K p110γ in shCD2AP-MDCK cells. Fig. S2 shows that GFP-CD2AP-336–605 and GFP-CD2AP-1–475 localize at the cell boundaries. Fig. S3 shows GFP-CD2AP-1–335 binds to PI3K p85α 1–315. CD2AP-1–335 does not localize at cell borders, but CD2AP-1–329-EGFP-CAAX does. Fig. S4 shows that targeting the SH3 domains of CD2AP to the membrane rescues F-actin and PI3K p110γ intensity at cell borders and restores cell morphology.

Research reported in this publication was supported by the National Institute of General Medical Sciences of the National Institutes of Health under award number R01GM106106 to W.M. Brieher.

The authors declare no competing financial interests.

Author contributions Y. Wang contributed to methodology, validation, formal analysis, investigation, writing (original draft), and visualization; and W. Brieher contributed to conceptualization, methodology, investigation, writing (review and editing), supervision, project administration, and funding acquisition.