First body of evidence suggesting a role of a tankyrase-binding motif (TBM) of vinculin (VCL) in epithelial cells

Cell culture

Vero cells (African green monkey kidney cells, ATCC CCL-81) were routinely cultured in MEM (PAA E15-888 or Capricorn MEM-STA) supplemented with 10% fetal bovine serum (FBS; PAA A15-151 or Capricorn), 100 U/mL penicillin and 100 μg/mL streptomycin (Sigma-Aldrich, St. Louis, MO, USA) at 37 °C and 5% CO₂. ATCC MCF-7 cells were received in passage number 6 (P6) as a kind gift from Archana Dhasarathy and Sergei Nechaev, University of North Dakota, USA. MCF-7 cells were cultured up to passage 14 (P14) in Capricorn DMEM-HPSTA with 10% FBS (GIBCO) in an incubator at 37 °C and 5% CO₂ or backed up frozen in 5% dimethylsulfoxide (DMSO) and 20% FBS in DMEM-HPSTA.

Protein extraction from subcellular fractions

Subcellular fractions were obtained from control cultures. The plasma membrane protein extraction kit (101Bio, #P503) was used according to the manufacturer’s instructions in order to extract proteins from nuclear (NF), cytoplasmic (CF), internal membrane (IMF) and plasma membrane (PMF) fractions of Vero cells (about 3 × 10⁷ cells). The fractions were then subjected to western blot to detect TNKS and PAR. Fraction controls included Lamin A/C (NF), CALR3 (CF), GRASP (IMF) and PMCA (PMF).

Whole cell protein extraction and affinity precipitation of PARylated proteins

Briefly, cells were grown in a T75 flask, washed, harvested by scraping and resuspended in lysis buffer (50 mM Tris, pH 8, 200 mM NaCl, 1 mM EDTA, 1% Triton X-100, 10% glycerol, 1 mM DTT, 0.1% SDS, and protease inhibitors). The extract was clarified (10 min, 15,000 g, 4 °C), and then subjected to affinity precipitation of PARylated proteins using WWE affinity resin (Tulip #2334) or Af1521 Macrodomain affinity resin (Tulip #2302), according to the manufacturer’s instructions. The resin was equilibrated in lysis buffer (10,000 g, 20 s) and total protein extract was added to the resin pellet and incubated ON in a cold chamber on a rotary shaker. The following day, the resin with the bound PARylated proteins was precipitated by centrifugation and the supernatant was saved as “flowthrough” (FT). The resin pellet was washed three times in lysis buffer. Affinity precipitated proteins (AP) were eluted by incubating in 50 µL protein loading buffer at 99 °C 10 min.

SDS-PAGE and western blot

Protein concentration was estimated by Bradford when appropriate. SDS-PAGE was done as described by Laemmli (1970). Semi-dry protein transfer onto nitrocellulose membranes was achieved using BiometraFastBlot (Analytik Jena). Membranes were blocked with 5% non-fat milk suspension in 0.05% TBS-Tween for 1 h and then incubated with the primary antibody ON and secondary antibody for 1 h. Western blots images were taken using the GeneGnome XRQ Chemiluminescence imaging system after developing with Western Lightning Plus ECL Substrate (Perkin Elmer, Waltham, MA, USA). When necessary, membrane stripping was achieved by incubation in mild Stripping buffer (1.5% Glycine, 0.1% SDS, 0.1% Tween-20, pH 2.2). The antibodies used and their correspondent concentrations are listed in Table 1.

Cell seeding in the presence of Tankyrase inhibitors to monitor epithelial belt assembly.

TNKS inhibitors were kindly provided by Dr. Lari Lehtiö (Biocenter Oulu, Faculty of Biochemistry and Molecular Medicine, University of Oulu, Finland). The reported potencies and IC50 are summarized in Table 2 (Haikarainen et al., 2013; Haikarainen, Krauss & Lehtio, 2014; Riffell, Lord & Ashworth, 2012; Lehtiö, Chi & Krauss, 2013; Mariotti, Pollock & Guettler, 2017; Thorsell et al., 2017; Haikarainen et al., 2016). Vero cells were treated with TNKS inhibitors (TNKSi) at the moment of seeding. Their effect on the PAR belt and on the accompanying cortical actin ring was initially evaluated qualitatively by immunocytofluorescence 5 h after seeding. Later, a quantitative evaluation was carried using the Cell Shape Index (see below). Alternatively, fewer cells were seeded and TNKSi was added on the monolayer and cells were fixed after reaching confluency.

Indirect immunocytofluorescence

Indirect immunocytofluorescence was carried as in Lafon-Hughes et al. (2014) on 12 mm round cover glasses placed in 24-well plates. Briefly, cells were fixed in 4% paraformaldehyde (PFA), permeabilized and subjected to blocking. Then, they were incubated with specific primary antibodies (see Table 1) diluted in blocking buffer (0.2% Tween, 1 % BSA in fPBS) for 2 h at 37 °C. After washing, cells were incubated (1 h, RT) with the correspondent anti-antibody mix in blocking buffer and/or cytopainter (ab176756). After washing, nuclei were counterstained with DAPI, coverslips were mounted in Vectashield (Vector 94010) or Prolong Gold (Molecular Probes P36930) and sealed. In order to check the specificity of the signals, controls without primary antibodies were carried in parallel.

Epifluorescence and confocal microscopy

Images were recorded with an epifluorescence microscope (40× objective) with DP controller software, an Olympus BX61/FV300 (Tokyo, Japan) with a Plan Apo 60 x/1.42 NA oil immersion objective or a ZEISS LSM800-Airyscan (Oberkochen, Germany) confocal microscope. Original images were taken in the same conditions as reference images of controls without primary antibodies at the same microscopy session. All images in each experimental series were taken with the same setting at the same confocal session. If modified, all were subjected to the same degree of brightness/contrast adjustment and Gaussian blur filtering, including the control without a primary antibody. The ImageJ free software was used for image processing.

VCL vs. PAR colocalization in epithelial belts: overlapping pixels highlight and quantification examples

Colocalization can be analyzed in terms of signals overlap or signals intensity correlation.

Relative expression of VCL in cells transfected with Tol2-GgVCL vs. Tol2-VCL/*TBM

Intensity quantification on confocal images was used to estimate the relative expression of VCL and GFP in transfected cells. Whole-cell contours and nuclear contours were used as ROIs to determine whole-cell GFP, whole-cell VCL, nuclear GFP and cytoplasmic VCL.

Cell shape index: combined roundness and circularity

Epithelial cells display tension in their adherens junctions belts and subcortical actin rings, giving rise to straight lines. Besides, even in the absence of neighbor cells, they have quite sharp borders. In contrast, mesenchymal or mesenchymal-like cells have serrated, wavy junctions and an irregular outline due to the presence of filopodia and lamellipodia (which can be decorated with ruffles) (Innocenti, 2018). To be able to quantify and compare morphological changes, it was required to express these morphologies in terms of numbers. ImageJ offers measurements of Roundness and Circularity. While Roundness accounts for gross changes (from circular to ovoid or triangular), it is insensitive to straight vs. serrated or undulated borders. On the other hand, Circularity is very sensitive to such edge characteristics. Thus, we defined a Cell Shape Index (CShin) combining Roundness and Circularity.

$$\mathbf{C}\mathbf{S}\mathbf{h}\mathbf{i}\mathbf{n}={\displaystyle \frac{\mathbf{R}\mathbf{o}\mathbf{u}\mathbf{n}\mathbf{d}\mathbf{n}\mathbf{e}\mathbf{s}\mathbf{s}}{\mathbf{C}\mathbf{i}\mathbf{r}\mathbf{c}\mathbf{u}\mathbf{l}\mathbf{a}\mathbf{r}\mathbf{i}\mathbf{t}\mathbf{y}}}$$

Figure 1A depicts values of CShin associated to some geometric figures (on the left). CShin is sensitive to radial symmetry, being higher for a circle than for an ellipse or higher for a square than a rectangle. If the clean edge of each geometric form is substituted by a serrated or undulated border, CShin values show a sharp increase. As can be seen in Fig. 1A (right), CShin is very sensitive to the edge characteristics of the shape. For example, CShin = 1 for a perfect circle and CShin = 1.81 for a form that conserves the circular proportions but has an undulated edge.

As a proof of concept, CShin was applied to a well-characterized cell model of an epithelial to mesenchymal transition (EMT). NMuMG (NAMRU Murine Mammary Gland) cells were measured untreated (when they have epithelial characteristics) or after treatment with TGF-β (when they have mesenchymal characteristics).

The Plotly Chart Studio online graph maker (https://chart-studio.plotly.com/create/box-plot/#/) was used to build box & whisker graphs, plus represent all the individual data points. Statistics were done using the free software available online at https://astatsa.com/, attributed to Vasavada (2016). As CShin statistical distribution is not normal, medians were compared using Mann-Whitney non-parametric test. Kruskall-Wallis non-parametric ANOVA followed by Conover test adjusted by Benjamini-Hochberg FDR method.

The shape change was evidenced as an increased CShin index (Fig. 1B). In this model, CShin mean and median were below 1 for epithelial cells and above 1 for mesenchymal cells, with a significant difference (p = 2.413887 × 10⁻⁸). Thus, the index proved to be useful to evidence the cell shape change from an epithelial to a mesenchymal cell.

TBM analysis in hVCL and homologs

Using SnapGene viewer, hVCL protein RefSeq sequence was searched for the c-TBM, RXXOXG, and the nc-TBM, RXXXOXG, where X is any aminoacid and O is Gly, Pro, Ala or Cys. After global lineal-sequence alignment was made using ClustalX 2.0 for Windows (Larkin et al., 2007), TBM conservation was checked among VCL homologs in different representative species. The Reference Sequences used were (as in Han, Van der Krogt & De Rooij, 2017): Human NP_003364.1, Mouse NP_033528.3, Chicken NP_990772.1, Xenopus NP_001090722.1, Zebrafish NP_001122153.1, Drosophila NP-476820.1 and C. elegans NP_501104.2). Comparison between hVCL and dVCL sequences were made using ExPASy alignment tool, SIM (Huang & Miller, 1991). Result indicated as Drosophila 1 corresponds to the comparison matrix PAM400, while result indicated as Drosophila 2 corresponds to the comparison matrix BLOSUM30, using equal parameters in both cases.

Drosophila VCL structure was predicted using CPHmodels-3.0 (Nielsen et al., 2010). Structural alignment and Figures were made using YASARA (Krieger, Koraimann & Vriend, 2002).

Generation of eukaryotic expression vectors with WT-GgVCL (Tol2-GgVCL/WT) or G454V-GgVCL (Tol2-GgVCL/*TBM)

To disrupt the association of VCL with TNKS, Gly454 (in TBM-II) was substituted for a Val residue. A ggc codon was substituted for gta by directed/circular mutagenesis using custom primers G454V (Table 3) and following the protocol published by Wang and Wilkinson (Wang & Wilkinson, 2000). pET15b/GgVcl 1-1066 (Addgene #46171) bearing the complete Gallus gallus VCL sequence was used as template. An AccI (BioLabs R0161S) (gtagac) site generated by the point mutations introduced was used as a diagnostic recognition site. The new plasmid, pET15/GgVcl/*TBM (Addgene #162781) is available at https://addgene.org/162781/.

The GgVCL and GgVCL/*TBM ORFs were then cloned into TOPO TA #450640 through the topo cloning reaction, following the manufacturer’s instructions, and then into a middle entry plasmid with multiple cloning sites (Tol2 Kit#237, pME-MCS), to obtain the pME-VCL vector.

The final mammalian expression vector was obtained through the use of Tol2 kit LR (Kwan et al., 2007; Yagita et al., 2010). The LR reaction involved the use of Gateway LR Clonase II Plus Enzyme Mix (Invitrogen 12538-120) to fuse 4 plasmids into a single final vector: the 5′entry plasmid with a β-actin promoter called p5E-actin (#299), the middle entry plasmid with the VCL insert pME-VCL (wt or mut), the 3′entry plasmid p3E-IRES-nls-GFP (#391) and the destination vector pDestTol2pA2 (#394). The resulting 14.5 KB eukaryotic expression vectors carried the WT or mutant VCL under a β-actin promoter and a GFP coding sequence driven by an internal ribosome entry site (IRES-GFP). Tol-2 GgVCL and Tol-2 GgVCL/*TBM constructs were amplified in NEB C3019I bacteria and corroborated by primer walking sequencing (Macrogen, Seoul, South Korea). Both Tol-2 plasmids (Addgene #162787 and #162790) are available at https://addgene.org/162787/ and https://addgene.org/162790/.

Transfection of MCF- 7 cells with Tol2-GgVCL or Tol2-GgVCL/*TBM

FUGENE transfection reaction: DNA (3:1) was used to transfect MCF-7 undisrupted monolayers. Cells were monitored under epifluorescence and phase contrast or subjected to ICF at the indicated times. 1 μg Pmax GFP plasmid (~3 KB) from Amaxa Mouse/Rat Hepatocyte Nucleofector kit was used as a positive control to evaluate transfection efficiency of a DNA plasmid.

VCL sequence harbors three TBMs conserved in vertebrates

All confirmed TNKS substrates to date contain one or more TBMs that are recognized via TNKS ARC domains (ARC1, ARC2, ARC4 and ARC5) (DaRosa, Klevit & Xu, 2018). Therefore, proteins harboring a TBM are putative TNKS PARylation targets. We examined the presence of TBMs in hVCL. Also, a comparative alignment was done with VCL sequences from model organisms including some vertebrates and two invertebrates: Drosophila melanogaster, whose adherens junctions are quite similar to those of vertebrates, and Caenorhabditis elegans, whose adherens junction proteins are not essential for general cell adhesion or for epithelial cell polarization.

Human VCL has three TBMs (Fig. 2A): TBM–I is RARGQG at positions 339-344 (exons 8-9), TBM–II is RRQGKG at 449-454 (exons 10-11) and TBM–III is RGLVAEG at positions 520-526 (exon 12). The first two sites are canonical (RXXOXG) and correspond to peripheral, easily accessible loops whereas the third is non-canonical (RXXXOXG) and is situated in a deeper position, on a helix.

All TBMs were conserved in other vertebrates such as Mus musculus, Gallus gallus, Xenopus laevis and zebrafish, while only one of them was also found in D. melanogaster. In contrast, none of them was present in C. elegans (Fig. 2A). It is interesting to note that the single TBM domain of Drosophila could align with both canonical TBMs (TBM I and II) of hVCL. Hereafter, structural comparison between hVCL and dVCL was done. The superimposition of hVCL with the predicted 3D-structure of dVCL indicated that Drosophila TBM was located in the same relative position in the structure as the human VCL TBM–II (Fig. 2B).

Endogenous PAR and TNKS were detected in the plasma membrane fraction of Vero cells

PAR had been previously detected associated to the epithelial belt in Vero cells by ICF using an anti-PAR antibody (Lafon-Hughes et al., 2014). Now we corroborated PAR presence in subcellular fractions whose purity was checked with specific markers (Fig. 3A). Instead of an antibody, the novel anti-PAR reagent MABE 1031 was used to evidence the presence of PAR associated to the plasma membrane fraction (PMF). MABE 1031 (AB_2665467) is a recombinant protein comprising the WWE domain from RNF146 and the Fc region of rabbit IgG. The latter is recognized by anti-rabbit secondary antibodies. Thus, it can be handled as an equivalent to a primary antibody in ICF, WB or affinity pull-down applications.

TNKS presence in the internal membrane fraction (IMF) was conspicuous and a slighter signal was also detected in nuclear fraction (NF), cytoplasmic fraction (CF) and plasma membrane fraction (PMF) (Fig. 3A). The cell fractionation results matched with ICF images (Figs. 3B–3E). In ICF, a TNKS ‘hat’ could be observed on one side of the nucleus, presumably corresponding to the Golgi complex. It was possible to see that a mild/slight TNKS signal spread towards the rest of the cell, reaching the plasma membrane or subcortical region sometimes (Figs. 3D, 3E). This TNKS distribution was robust, since it was observed in ICF experiments performed using two different anti-TNKS antibodies designed to recognize the Nt or the Ct (catalytic domain), as well as under two different fixation protocols (Fig. S1).

TNKS inhibitors disrupted the PAR belt synthesis and altered cell shape

A tight adherens junctions belt, indicative of tension, involves a straight PAR belt and a straight subcortical actin belt. In contrast, an incomplete lax belt is associated with punctuated (less abundant) or serrated PAR and F-actin signals.

Four TNKS inhibitors with different potency and specificity towards TNKS-1 and TNKS-2 (Table 2), were added to cell cultures at the moment of seeding (Fig. 4). Results were ordered according to TNKSi in vitro specificity, starting with TNKS-1-prone FLALL (Figs. 4B, 4G, 4L), then MN64, G007LK and the TNKS-2 selective OD35 in the last position (Figs. 4E, 4J, 4O). All of them hindered PAR belt synthesis. A punctuated or zipper-like discontinuous distribution of PAR was observed (Figs. 4B–4O). The PAR belt did not completely disappear, which could be explained by incomplete TNKS inhibition. Cortical F-actin changed accordingly (Figs. 4G–4J), reinforcing our previous results correlating PAR and F-actin distributions in epithelial cells and nerves (Lafon-Hughes et al., 2014; Lafon Hughes et al., 2017). In the presence of MN64, less cells remained attached after the PBS wash done previous to fixation. Such a cell density decrease has been previously documented in the presence of a TNKS and PARP-1/2/3 inhibitor called XAV939 (Lafon-Hughes et al., 2014).

With the aim of quantifying changes in cell shape, particularly those involving cell contours, we designed a Cell Shape Index or CShin (see “Materials and Methods”, section 9, Fig. 1).

Median CShin (Fig. 5) was below 1 in control cells, as expected for epithelial cells according to the index validation in NMuMG cells (Fig. 1B). No effect of the PARP-1/2/3 inhibitor Olaparib (OLA) was detected. In contrast, all TNKSi, namely FLALL, MN64, G007LK and OD35, induced a significative increase of CShin, indicating a measurable change in cell contours.

We did not find a correlation between TNKSi selectivity and the observed effects on PAR belt, the subcortical F-actin cytoskeleton and CShin. The potency and specificity of the inhibitors were determined in vitro while, in whole cells, factors as permeability, subcellular distribution, stability and metabolic alterations may affect the effective dose and outcome. Therefore, a straight-forward correlation was not necessarily expected.

As stated in the introduction, in epithelial cells the subcortical actin ring and the PAR belt are somehow associated. We are now arguing that PARylation of a target protein joining adherens junctions and the F-actin cytoskeleton by TNKS1/2 is at least one of the factors that could explain this association.

A VCL pool was a PARylation target

Using a WWE-affinity resin, covalently PARylated proteins were affinity precipitated. Next, cell junction or cytoskeletal proteins were tested by WB as putative PARylation targets (Fig. 6A). Interestingly, while β-catenin, E-cadherin, actin and α-tubulin were recovered only in the flow through (FT), a fraction of VCL was affinity precipitated (AP) by the WWE-affinity resin which consists of highly purified GST-RNF146(100-175) fusion protein bound to glutathione beads. Only anti- α-tubulin and anti-VCL antibodies were of mouse origin; thus, the result is clear-cut and cannot have been affected by successive membrane stripping, demonstrating the existence of a PARylated VCL pool. Given the AP was done in the absence of crosslinking and in the presence of high detergent and salt concentrations, non-covalent interactions were not expected to be maintained. This was reflected in the fact that other proteins of the adherens junctions complex were not affinity precipitated. Thus, this result indicated the existence of covalently PARylated VCL.

VCL was also affinity precipitated as a light band in experiments carried using the Af1521 Macrodomain resin, which recognizes mono or poly(ADP-ribosylated) proteins. PARP-1 was also checked as an AP control. While the specific TNKSi G007LK hindered only VCL affinity precipitation (Fig. S2A), XAV939, which inhibits PARP-1/2/3 & TNKS, hampered both VCL and PARP-1 affinity precipitation (Fig. S2B).

Altogether, these results support that TNKS is responsible for VCL PARylation.

To map the subcellular localization of PARylated VCL, we evaluated the co-occurrence of PAR and VCL signals in confocal ICF images. Figure 6B is a scheme representing the relative position of the subsequent confocal epithelial monolayer sections, running from basal to apical direction shown as a merged channels montage in Figs. 6C–6H. The correspondent masks and multiplied masks montages (Figs. 6I–6N and Figs. 6O–6T) are also displayed to highlight the overlapping points. VCL was present in all these sections but the PAR belt was distributed along a 1 to 1.5 µm in z (3 to 5 slices depending on voxel height). PAR and VCL signals overlap drew just a thin polygon (in this case, a pentagon) representing the epithelial belt. No PARylation was ever detected at VCL-rich focal contacts (neither in untreated or treated cell cultures of Vero, NMuMG or MCF-7 cells, nor in mice epithelial tissues).

To quantify these observations, colocalization was analyzed in terms of signals overlap or signal intensity correlation (see “Materials and Methods”).

We have included an example of intensity correlation analysis, taking a ROI from Fig. 6E. The ROI red and green channel images (Figs. 6U and 6V) were subjected to an intensity correlation analysis. Pearson Coefficient (Rr), which ranges from −1 to +1 (from anti-colocalization to complete colocalization), was 0.191. Li’s ICQ (which ranges from −0.5 to +0.5) was 0.235. Double positive PDMs (products of the differences from the means) were localized precisely in the belt region (Fig. 6W), just as the mask obtained through the correspondent masks multiplication (Fig. 6X). Therefore, the intensity correlation analysis and the overlapping analysis converged in the epithelial belt region. Thus, mask multiplication to highlight overlapping signals and the overlapping analysis was used hereon (see Figs. S3A–3SL). We have additionally explored PAR and VCL colocalization in an EMT model. Epithelial NMuMG cells display PAR and VCL colocalization (overlap) precisely at the epithelial belt region (Fig. S3M) which is lost upon treatment with TGF-β to induce EMT (Fig. S3N) (for further reference, see our previous work (Schacke et al., 2019)). We have also analyzed PAR-VCL in belt ROIs colocalization (overlap) focused on epithelial belt ROIs (Fig. 6Y). In this case, Mander’s coefficient M1 ranged from 0.262 to 0.660, indicating that ≈ 30% to 60% of belt VCL signal overlapped with PAR signal; M2 ranged from 0.255 to 0.882 indicating that ≈ 30% to 90% of belt PAR signal overlapped with VCL.

Partial colocalization of PAR and VCL (within 200 nm) had also been previously observed in Vero cells by ICF and confocal microscopy (Lafon-Hughes et al., 2014). FRET was carried after indirect ICF (Fig. S4), lowering the colocalization maximal distance to about 50 nm (estimated according to König et al 2006 (Konig et al., 2006)) as the addition of 10 nm FRET + 10 nm each primary antibody + 10 nm each secondary antibody. FRET results were not included in the main manuscript because this resolution did still not allow to distinguish among VCL and nearby PARylated proteins nor among covalently PARylated or PAR-bound proteins; thus, the affinity precipitation result constituted by far stronger evidence.

Not only not all VCL was PARylated (according to AP and ICF results), but also not all belt PARylation could be attributed to VCL; the band pattern detected after subcellular fractionation using the PAR-detecting reagent indicated there may be additional targets.

Mesenchymal-like shaped cells were observed under VCL/*TBM transient overexpression

We finally wanted to study whether VCL PARylation was related to its functions. Using the obtained information and given the importance of the single conserved c-TBM in Drosophila to preserve its function, we decided to mutate the equivalent motif (c-TBM-II) in vertebrate VCL.

Direct mutagenesis of VCL in pET15b/GgVCL was achieved using primer pair G454V and the circular mutagenesis protocol suggested by Wang and Wilkinson (2000) (Wang & Wilkinson, 2000) in the presence of DMSO. As depicted in Fig. 7A, a ggc codon (coding Glycine) was substituted by gTA (coding Valine), generating an AccI (BioLabs R0161S) diagnostic restriction recognition site (gTAgac). Then, after incorporating each VCL sequence to a middle entry plasmid (pME), pME-VCL and pME-VCL/*TBM were obtained. The recombination of a plasmid carrying a β-actin promoter, pME-VCL or pME-VCL/*TBM, a plasmid coding an IRES-GFP and a destination plasmid allowed to obtain the desired eukaryotic expression plasmids depicted in Fig. 7B.

Transfection of MCF-7 cells with Tol2-GgVCL (Figs. 8A, 8B, 8E–8J) or Tol2-GgVCL/*TBM (Figs. 8C, 8D, 8K–8O) was achieved using FUGENE reagent on cell monolayers. These Tol2 plasmids allow the identification of transfected cells that express the plasmid as GFP+ cells (Figs. 8A′–8O′, in green). Although carried in parallel and in the same conditions, the transfection with Tol2-GgVCL/*TBM rendered more GFP+ cells than the WT version, indicating that this transfection was more effective or less cytotoxic (low cell numbers did not allow carrying additional quantifications to distinguish among these two scenarios). Moreover, image intensity quantifications showed that GFP expression was similar in Tol2-GgVCL or Tol2-GgVCL/*TBM cells while VCL expression was doubled in the mutant with respect to the WT transfected cells. It is worth noting that cells transfected with Tol2-GgVCL/*TBM presented a distorted morphology without lineal tense cell junctions and with filopodia or lamellipodia, that looked mesenchymal-like (Figs. 8C, 8D, 8K–8O). Cell shape was measured and the results are in Fig. 9. While GgVCL transfected cells displayed a median CShin <1, cells transfected with GgVCL/*TBM had a significatively higher CShin. This result strongly suggested the involvement of TBM in the function of VCL to maintain cell shape.

To confirm this result, we aimed to use a KO-add back approach. We attempted to obtain VCL knocked-out MCF cells by a CRISPR/Cas9 strategy (Fig. S5A). Only a limited heterogeneous cell pool could be obtained as transfected cells grew extremely slowly and lost their colony forming ability under isolation (Figs. S5B–S5Q, Figs. S6A–S6H), precluding the obtainment of pure clones and the possibility of straight-forward VCL KO confirmation. The available cells, however, were transfected with the Tol2-GgVCL/*TBM plasmids (Figs. S7A–S7N) or Tol2-GgVCL (Figs. S7O–S7R). CShin was sensitive enough to detect changes in cell shape induced by the presumptive VCL partial “KO” which were reverted by the transfection with WT GgVCL but not with Gg VCL/*TBM (see contours examples in Figs. S7E, S7I, S7L & S7P, and the graph in Fig. S8). Unfortunately, the obstacles regarding the “KO” cells obtainment limited the conclusions that could be driven from these experiments and pointed out the need to implement an alternative approach to confirm these observations.