APC-driven actin nucleation powers collective cell dynamics in colorectal cancer cells

Summary Cell remodeling relies on dynamic rearrangements of cell contacts powered by the actin cytoskeleton. The tumor suppressor adenomatous polyposis coli (APC) nucleate actin filaments (F-actin) and localizes at cell junctions. Whether APC-driven actin nucleation acts in cell junction remodeling remains unknown. By combining bioimaging and genetic tools with artificial intelligence algorithms applied to colorectal cancer cell, we found that the APC-dependent actin pool contributes to sustaining levels of F-actin, as well as E-cadherin and occludin protein levels at cell junctions. Moreover, this activity preserved cell junction length and angle, as well as vertex motion and integrity. Loss of this F-actin pool led to larger cells with slow and random cell movement within a sheet. Our findings suggest that APC-driven actin nucleation promotes cell junction integrity and dynamics to facilitate collective cell remodeling and motility. This offers a new perspective to explore the relevance of APC-driven cytoskeletal function in gut morphogenesis.


INTRODUCTION
In the gut, cell-cell junction remodeling is key to safeguarding the integrity of the epithelial barrier and to provide structural support for the epithelial monolayers. [1][2][3][4][5][6][7][8] In this context, the actin cytoskeleton controls the abundance of adhesive molecules at cell-cell junctions, restricts the mobility of the cell adhesion components at the membrane, and supports the integrity of cell junctions. [8][9][10][11] Moreover, a mesh of actin is assembled under the cytoplasmic membrane that is critical for contractility of cells and their resulting architecture. 12,13 Furthermore, actin-rich protrusions generate force for cells to migrate out of the crypts toward the villi tips where they are shed off. 14,15 The tumor suppressor adenomatous polyposis coli (APC) is considered the master regulator of gut homeostasis. [16][17][18] Mutations in APC, most of them sporadic, have been found in > 80% of colorectal cancer cases, 16,19 with many leading to alterations in gut organization including the presence of (precancerous) polyps. Loss of heterozygosity via somatic mutation drives tumorigenesis. For instance, patients with inherited syndrome familial adenomatous polyposis (FAP) present hundreds of polyps formed in early adulthood with a very high risk to progress to cancer by the age of forty. 20,21 Much of our current knowledge on the roles of APC has been inferred from studies using gene silencing, knock-outs, or expression of truncated proteins lacking the C-terminal part of APC. Yet, APC is a multi-functional 2843-amino acid protein with multiple interacting partners. As such, the full biological significance of APC's broad range of activities is far from clear. Historically, APC's cytoskeletal roles have been overlooked, 22 although they might play significant functions in organizing the gut epithelium, independent of APC's control of cell proliferation via Wnt signaling. 23 We have previously generated a specific ''separation-of-function'' APC mutant-herein named APC-m4. 24 APC-m4 only changes two amino acids-L2539A I2541A-out of the 2843 residues of the entire protein ( Figure 1A; 24 ). APC-m4 is abrogated in F-actin nucleation but capable of binding actin monomers. 24 Transient transfection of APC-m4 in U2OS osteosarcoma and MDA-MB231 breast cancer cells caused a delay in focal adhesion turnover and migration of single cells. [24][25][26] Evidence showed that full-length APC is required for proper localization of tight and adherens components at cell junctions, an association dependent on an ''active'' actin cytoskeleton. 27-30 SW480 colorectal cancer cell line expresses a truncated version of APC (1-1388 amino acids), and therefore lacks the C-terminal domain of the protein, where the cytoskeleton activity resides. Reintroduction of full-length APC protein in SW480 rescues the localization of adhesive components and F-actin at adherens junctions, 1  The lines below the schematic represent the two cell lines used in this study, and the corresponding length of the APC protein that they express. These are SW480 cell line, which expresses a truncated product lacking both signaling and cytoskeletal functions (1-1388 amino acid), and HCT116 cell line that expresses full-length APC (1-2843 amino acid). (B) Qualitative graph showing the percentage of cells that belong to the ''normal'', ''discontinuous'', and ''round'' category assigned in function of the appearance of the F-actin label at the cell junctions (see STAR Methods). two-way ANOVA Sidak's multiple comparisons test was performed to find the statistical differences. ''****'' is p < 0.0001, ''***'' is p < 0.001, and ''ns'' is not significant. N = 3 independent repeats for each condition; n(APC-WT) = 137 cells in total, n(APC-m4) = 95 cells in total. These precedents raise the question of whether APC-driven actin nucleation activity may contribute to any collective cell remodeling event at cell junctions and consequent motility-associated event(s) critical for gut homeostasis. Here, we have implemented a combination of genetics, cell imaging tools, and artificial intelligence algorithms applied to the study of colorectal cancer cell monolayers to investigate the impact of APC-driven actin nucleation on a) the levels and dynamics of cell adhesion components, b) cell size, and c) the directionality of cell migration within a sheet. Our data establish a role for APC in coordinating actin with cell adhesion dynamics to control collective cell remodeling and directed cell motility in colorectal cancer.

RESULTS
Actin nucleation mediated by APC is required for maintaining proper levels and dynamics of adhesive proteins at cell junctions We have explored whether APC-driven actin nucleation activity may contribute to any collective cell remodeling event at cell junctions. To that end, we have generated colorectal (adeno) carcinoma-derived epithelial cellular lines (SW480 and HCT116) expressing stably wild-type APC (APC-WT) or the mutant APC-m4 (Figures 1A and S1A-S1D; see STAR Methods). As mentioned, SW480 cell line expresses a truncated version of APC that lacks the C-terminal part of protein, where the cytoskeleton activity resides. 32 In contrast, HCT116 cell line expresses full-length APC. 32 Despite effects of full length APC-m4 mutant appear to be dominant, 24 similar to the cancer-linked C-terminal truncations of APC, 33 we used SW480 as a control for specificity of effects related to C-terminal activities. We used these newly generated stable cell lines expressing ectopic (full length) APC-WT or APC-m4 to assess F-actin at cell-cell junctions. To do so, we first grew SW480 expressing stably APC-WT and APC-m4 epithelial monolayers, fixed and stained cells with Alexa 488-phalloidin. Confocal laser scanning microscopy showed that in SW480 APC-WT cells F-actin localized at cell junctions, as expected ( Figures 1B and 1C). In contrast, in APC-m4 cells, F-actin was more diffused at cell junctions ( Figures 1B and 1C). In addition, the label at APC-WT cell junctions appeared smooth and continuous compared to puncta marking cell junctions in APC-m4 cells. Based on the appearance of F-actin label at cell junctions that we correlated with their healthiness and morphology of the cell, we divided cells into three groups: ''normal'', ''discontinuous'', and ''round''. From this qualitative analysis, we found that 67.49% of APC-WT cells displayed ''normal'' F-actin at the junctions compared to 18.24% of APC-m4 cells, 26.16% of APC-WT cells presented ''discontinuous label'' compared to 59.20% of APC-m4 cells, and no significant differences were found for the ''round'' category between APC-WT and APC-m4 cells ( Figure 1B). A similar distribution of F-actin label, and therefore distribution of ''normal'', ''discontinuous'', and ''round'' cells was observed in HCT116 APC-WT and APC-m4 monolayers (Figure S1E). Together, these findings suggest that the loss of APC-driven actin nucleation significantly compromised F-actin localization at cell junctions in monolayers.
Subsequently, we asked whether the F-actin pool generated by APC had any effect on the levels of F-actin at cell junctions, and/or levels and dynamics of adhesive components at cell junctions. To answer those questions, we stained SW480 APC-WT or APC-m4 monolayers to visualize phalloidin (F-actin marker) along with antibodies to visualize occludin (a classic tight junction marker) by immunofluorescence. Dual-color imaging showed that F-actin and occludin localized at cell junctions in APC-WT cells as expected (Figure 1C). However, both labels were spread into the cytoplasm in APC-m4 cells ( Figure 1C). Quantitative analysis showed a significant decrease in the fluorescence intensity of F-actin and occludin; specifically, at the cell-cell junctions of APC-m4 cells compared to the same labels in APC-WT cells ( Figures 1D and  1E). Our results support a role for F-actin nucleation by APC in stabilizing a pool of actin and junctional proteins at cell junctions.
Next, we sought to further investigate whether APC-mediated actin nucleation affected the ultrastructure of cell junctions using scanning electron microscopy (SEM). We found that APC-WT cell attachments were iScience Article well-formed and presented interdigitated junctions ( Figure 1F). Recent studies have argued that the formation of membrane protrusions engaging in the form of a ''handshake'' or ''tethering nanotubes'' (known as TENT) is critical to maintain cohesive junctions. 7,34 Accordingly, we observed prominent TENT structures in APC-WT cell junctions ( Figure 1F). In sharp contrast, no interdigitation between cell junctions and barely any TENTs were observed in the APC-m4 SEM images ( Figure 1F). These results confirmed that the structural connections between cells were strongly perturbed in the mutant ( Figure 1F).
To examine whether the loss of actin nucleated by APC had any effect on cell junction dynamics, we transiently transfected SW480 APC-WT and APC-m4 with E-cadherin protein fused to a GFP-tag-a classic marker for adherens junctions. Using live imaging in a confocal laser scanner microscope, we analyzed various junctional parameters in monolayers ( Figure 2, Video S1). APC-WT cells had straighter junctions that stayed almost unchanged in length over the 20-min observation window (Figures 2A-2C). In contrast, a significant variation in both junction length and angle were observed in APC-m4 cells (Figures 2A-2C). Additionally, we quantified the movement of each cell junction over time by drawing a perpendicular line to each junction using this region of interest to acquire kymographs ( Figure 2D). Junction velocity was determined from the kymograph slope. We found that APC-m4 junctions had a higher velocity than APC-WT ($3.7 x10 À3 vs. $4.3 x10 À3 mm/s, Figure 2D). Together these data indicated that the loss of F-actin nucleation by APC led to greater plasticity of cell junctions.
From our time lapses, we noticed that cell junctions, but especially the tricellular adhesive sites (herein named vertexes) at cell junctions exhibited less E-cadherin in APC-m4 cells relative to APC-WT cells (Figure 2A). Given that recent evidence showed that changes in E-cadherin levels at the vertexes of cell junctions affect their motion and generate an asymmetry of vertex contraction, 35 we investigated various vertex parameters in APC-m4. We observed that vertexes were more disrupted in the mutant cells over time. Analysis of vertex integrity revealed that $26% of APC-m4 junctions had intact junction vertexes in contrast to $85% of APC-WT ( Figure 2E). Moreover, instances of single or both vertexes being disrupted were $34% and $40%, respectively, in APC-m4, compared to only $10% and $5%, respectively, in APC-WT ( Figure 2E). In addition, vertex velocities in APC-m4 junctions were slightly higher than in APC-WT junctions ($4.1 x10 À3 vs. $4.6 x10 À3 mm/s; Figure 2F). Furthermore, both APC-WT vertexes in a junction had similar velocities ( Figure 2G). However, APC-m4 vertexes in a junction moved at significantly different velocities ($6.6 x10 À4 mm/s difference; Figure 2G). Together, these data demonstrate that cells expressing APC-m4 presented lower levels of adhesive components, developed less stable and robust junctions with vertexes displaying different strength and motility compared to the junctions and corresponding vertexes of cells expressing APC-WT as control.
These data suggest that APC-driven actin nucleation is the key to preserving robust integrity and dynamics of the components at cell junctions, as well as facilitating tight structural bonds between cells.

Loss of APC-driven actin nucleation results in larger cells
The actin cytoskeleton is central to cell shape. 3,36-39 From our images, it seemed that some APC-m4 cellsimpaired in actin nucleation-were larger than APC-WT cells. We therefore analyzed the size of individual APC-WT and APC-m4 cells grown as an adherent monolayer ( Figure 3A). We used a customized version of a (G) Cartoon representing how the difference between the velocities of two vertexes at the same cell junction was analyzed and violin plot showing those differences in vertex velocities within a junction. N = 3 independent repeats for each condition; n(APC-WT) = 44 junctions in total, n(APC-m4) = 52 junctions in total. Data in B and D-F are displayed as ''superplots'' showing the mean of the different replicates (circles) and the distribution of ''n'' junctions or vertexes analyzed, as stated in each graph, (color-coded dots) was superimposed as violin plot. Black lines are the mean and standard deviation; paired two-tailed t test was performed to find the statistical differences, except for B that ratio paired two-tailed t test was performed, using N = 3 independent replicates for all graphs. ''*'' is p < 0.05, ''ns'' is not significant. See also Video S1. iScience Article deep-learning-based segmentation method called cellpose. 40 Quantification of HCT116 APC-WT and APC-m4 cells showed differences in size (187.6 mm 2 and 203.8 mm 2 , respectively; Figures 3B and 3C). To further validate our results, we also measured the cell size in the SW480 colorectal cancer cell line in which the endogenous APC lacks the C-terminal region (where the actin-related function resides). A similar trend was found between SW480 APC-WT and APC-m4, though more pronounced than in the HCT116 APC-WT and APC-m4 cells ( Figures 3D and 3E versus Figures 3B and 3C). Together, these data demonstrated that the change in cell size could be attributed to the lack of actin nucleation activity by APC.
APC also synergizes with additional nucleation proteins including formins to further promote actin filament polymerization in vitro. 24, [41][42][43] To determine whether formin-dependent actin polymerization contributes to cell size changes, we grew SW480 APC-WT and APC-m4 monolayers and treated them with the pan-formin small-molecule inhibitor SMIFH2 for 1 h. 44,45 All cell monolayers were fixed, and individual cell sizes were calculated. Double inhibition of formin and APC-dependent actin activity did not result in any significant difference in cell-spread area compared to single inhibition in the mutant ( Figure 3F).
To rule out that the defects observed in APC-m4 cell size were caused by changes in cell division, we performed flow cytometry analysis to determine the percentage of SW480 APC-WT or APC-m4 cells at different stages of the cell cycle (G1, M, and G2). No significant differences were found at any cell stage ( Figures 3G and 3H). Overall, these results support that cell size maintenance requires proper functioning of APC-driven actin nucleation-independent of formin function and cell proliferation.

APC-mediated actin nucleation governs collective cell migration
Collective cell migration requires a constant modification of cell shape, which is intrinsically coupled with dynamic remodeling of the actin cytoskeleton. 46-48 Therefore, we explored whether collective cell migration was compromised in APC-m4 colorectal cancer cells. To address this, HCT116 cells expressing APC-WT or APC-m4 were grown as monolayers, scratched to create a wound and imaged during healing ( Figure 4A, Video S2). We analyzed the trajectories of individual cells migrating at wounds over time and quantified directionality, migration speed, accumulated traveled, and Euclidean distances (the shortest distance between two points, see STAR Methods). We observed that the APC-m4 cell trajectories were more erratic (cells moved significantly slower and covered shorter distances than the APC-WT cells) indicating random cell migration ( Figures 4B-4F). In addition, we observed that mutant cells exhibited a high incidence of cell detachment ( Figure 4A, Video S2), also contributing to the loss of coherent directionality. Similar results were observed in SW480 colorectal cell monolayers ( Figure S2).
Since APC regulates centrosome reorientation to establish polarity of cell migration, [49][50][51] it could be that APC-dependent actin nucleation effects on migration were due to cell polarity defects. To investigate that, APC-WT and APC-m4 epithelial monolayers were grown, scratched, and after 16 h from wounding, migrating cells were fixed. All cells were co-stained with pericentrin antibodies (a marker for centrosome reorientation) and DAPI (nuclear marker) ( Figure 4G). Analysis of centrosome orientation toward the cell front did not show major differences between APC-WT and APC-m4 cells, suggesting that the front-rear axis is not altered by loss of APC-driven actin nucleation ( Figure 4H). These results implicate APC-driven actin nucleation in promoting proper directionality and speed of colorectal cells moving as a collection, independently of its cell polarity roles.

DISCUSSION
In this study, we have exploited the power of a specific ''separation-of-function'' APC mutant named APC-m4 that specifically abrogates actin nucleation function 24 to probe APC's involvement in cell-cell junction integrity via the actin cytoskeleton. Our study offers crucial mechanistic insight to bridge the gap between two functional links previously proposed: APC and cell junctions, and F-actin and cell junctions. Specifically, our findings integrate these two links by highlighting the requirement of APC-driven actin nucleation activity to sustain cell junction integrity and strength, cell size, and consequent directionality of cell migration.  It has been shown that cells expressing truncated APC proteins lacking the C-terminus exhibit reduced levels of E-cadherin molecules at cell junctions. 31,52 In turn, the expression of full-length APC restores cadherin protein levels, suggesting a relevant role of the C-terminal domain of APC. 31 Others proposed that a mesh of F-actin surrounds E-cadherin clusters to act as an ''insulator'' or ''corral'' to delimit their location, fusion, and mobility. 11,53 Indeed, inhibition of formin activity reduces both F-actin and E-cadherin levels, with concomitant perturbation of E-cadherin stability at cell junctions. 54 It is believed that rearrangements at cell junctions may be governed by forces yet to be defined in mechanistic terms. For example, actomyosin-based contractility might induce conformational changes in adhesive proteins, in turn, affecting their stability and/or binding to partners. 55-58 Interestingly, modulation of E-cadherin levels at the vertexes of cell junctions affects their motion and generates asymmetry of vertex (and therefore cell junction) contraction. 35 Here, we show that expression of APC-m4 in colorectal cancer cells led to a reduction in F-actin at cell junctions relative to the levels observed in cells expressing APC-WT. Accordingly, APC-m4 cells presented cell junctions with reduced levels of adhesive components, as well as perturbed dynamics, including a failure to maintain cell junction length and angle, an increase in the incidence of vertex disruption and a perturbed correlation of velocities between the two vertexes in a junction that shifted from symmetric to asymmetric vertex contraction. In support of the precedents outlined above, our data suggest that the pool of actin nucleated by APC might act as a fence that confines adhesive components within cell junctions and, at the same time, confer force to enable cell junction rearrangements while preserving their integrity and strength.
Dynamic cell shape, subject to remodeling of the actin cytoskeleton, is critical to proper cell migration. 3,36,47,59 Interestingly, we found that APC-m4 cells were larger and moved more erratically and slowly than APC-WT cells, despite being able to polarize centrosomes toward the leading edge. The larger size of mutant cells might stem from the reduction in E-cadherin, prompting the translocation to the nucleus of the yes-associated protein 1 (YAP) transcription factor. 60 Alternatively, but perhaps along with any potential alteration of YAP, it could be due to the reduction in actin polymerization that in turn affects cell stiffness area and volume. 59 The fact that APC-m4 migrating cells did not present major problems in repositioning centrosomes toward the leading edge compared to APC-WT cells goes in line with previous reports. On one hand, APC's clustering at microtubule plus-ends and consequent stabilization of microtubules at the leading edge seems to be required for the establishment of normal epithelial polarity. 23, [49][50][51]61,62 On the other hand, focal adhesion assembly or disassembly rates between focal adhesions located at the leading versus trailing edges in APC-m4 individual cells did not show any statistical differences compared to APC-WT cells. 25 It is worth mentioning that APC-m4 mutant cells are capable to bind to G-actin monomers, 24 and somehow APC-actin interactions could facilitate interplay between both cytoskeletons to reposition centrosomes toward the leading edge front in the mutant cells. These observations suggest that independently of APC's roles in cell polarity and/or microtubule stability, APC-driven actin nucleation activity in colorectal cancer cells may be required for proper shape and cell-cell attachments during cell migration to sustain their directionality and speed.

Conclusion
We propose that actin nucleated by APC acts as a ''force buffer'' to support the build-up of adhesive molecules allowing cells to undergo the required rearrangements, underlying correct cell junction dynamics. As a result, this actin pool might contribute to sustaining the structure of the epithelial monolayer, lining the gut, and facilitate its deformation to ultimately render the typical crypts-villi landscape. Then, APC-driven actin nucleation might be involved in the ''epithelial morphology'' pathway to control gut organization independently of the microtubule dynamics and Wnt signaling pathways. 23 In addition, the effects of APC-driven actin nucleation on directed cell migration in wounded monolayers, suggest that APC alone, Figure 4. Continued ranging from 0 to 180 grades (in degrees) that correspond to less oriented (yellow) to more oriented (purple), respectively, toward the leading edge; as represented in the cartoon. Orientation in grades was calculated as the angle from pericentrin in relation to the nucleus and to the front of the wounding migrating cells. Mann-Whitney U test was performed to find statistical differences. ''ns'' is p < 0.05. Data are from three independent repeats; n(APC-WT) = 31 angles (one per cell), n(APC-m4) = 33 angles (one per cell). See also Figure S2.

OPEN ACCESS
iScience 26, 106583, May 19, 2023 9 iScience Article or perhaps in coordination with Arp2/3 and/or formins, might be critical in orchestrating directed cell motility along the crypt-villus axis.
Our results provide a further connection between the well-established relevance of actin polymerization activity for the integrity of the gut barrier and the architecture of the gut monolayer on one hand, and APC's standing as the master regulator of gut homeostasis on the other. In our view, APC-mediated actin nucleation activity at cell junctions would not only support the integrity of the epithelial barrier that is critical to avert infection or inflammatory disorders, such as inflammatory bowel disease but also to withstand epithelial cell deformation that could otherwise drive the path to polyps and ultimately to colorectal cancer. Thus, our findings offer a new perspective to explore the relevance of APC-driven cytoskeletal functions in gut morphogenesis and human disease.

Limitations of the study
The major limitation of this study is that as a model system, we have used cell monolayers. In such set ups, cells are not in their physiological environment and interactions with the extracellular matrix that can influence the results can be missed. Supporting assays and analysis in intestinal organoids expressing endogenous APC wild-type or mutant-which would mimic patient response in a more physiological context-would validate the effects of APC-driven actin nucleation in collective cell dynamics. However, work on organoids, especially expressing full-length APC or point mutations at endogenous levels, as well as live imaging studies to monitor cell remodeling and migration using organoids is currently very challenging. Development of those experimental models and powerful technologies would provide invaluable insights to determine the role of APC-mediated cytoskeletal activities in gut tumorigenesis.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
For wound healing assays, 80,000 cells -HCT116 or SW480 APC-WT or APC-m4 -were seeded on 8-well chambers with collagen IV (Ibidi, Germany) containing DMEM medium containing high glucose buffer (Gibco, Life Technologies), 10% FBS, 20 mM L-glutamine, 1 mM sodium pyruvate, and 5 ml of Penicillin-Streptomycin antibiotics and incubated at 37 C, 5% CO 2 , and 95% humidity. Once cells were confluent, the medium was removed and replaced with fresh medium but without FBS for 6 -7 hours. After that time, axial wounds were performed using a standard pipette tip. Then, serum-free medium was removed, and cells were washed twice with PBS (#20012027, ThermoFisher). Next, the chambers were replenished with DMEM medium containing 10% FBS.
To monitor wound closure, SW480 and HCT116 cells were imaged at 30 min intervals for 24 hr or 48 hr, respectively, maintaining cells at 37 C with 5% CO 2 and 95% humidity using an Okolab stage top incubator H301-T-UNIT-BL-PLUS, Okolab CO2-UNIT-BL and HM-ACTIVE Humidity Controllers connected to a H301-K-FRAME holder adapted and an inverted Leica DMi8 microscope (Leica microsystems). Images were captured using an HC PL FLUOTAR 10x/0.30 air objective, and a Leica DFC9000GT camera using a LAS X Life Science Microscope Software (version 3.5.5.19976). To assess repositioning of centrosomes after wounding, cells were allowed to migrate for 16 hours maintained at 37 C, 5% CO 2 and 95% humidity in an incubator, then fixed with warm 4% paraformaldehyde in PBS and proceed for immunofluorescence to co-stain with pericentrin antibodies (centrosome marker) and DAPI (nuclear marker).

Fluorescence intensity measurements
To measure APC protein levels in cells, a custom macro in Fiji was used to generate and save sum slices z-projections for all images. Next, cell boundaries were manually drawn on these images and saved in a separate folder. Another custom macro in Fiji was then used to measure integrated density, mean grey value, and cell area. Finally, corrected total cell fluorescence (CTCF) was calculated using the following formula: iScience Article To measure F-Actin and Occludin protein levels at the cell junctions, immunofluorescence images were opened in Fiji and cell boundaries were manually traced from the F-Actin channel. Regions of interest (ROIs) from F-Actin channel were saved and over imposed with cells/junctions in the Occludin channel.
Mean grey values were obtained in ImageJ for both channels and saved as .CSV. Data was then analyzed and plotted in GraphPad Prism (version 9.0; GraphPad Software, La Jolla, CA) as Superplots. 66,67 Superplots displayed the mean of the different replicates (N = 4 in these cases, in circles) and the distribution of 'n' as level of F-Actin or Occludin at the cell junctions (as color-code dots) was superimposed to the means as violin plot. Mean was shown as a line; paired two-tailed t test was performed to find the statistical differences between replicates (N = 4). This way of representing data linked paired measurements together and conveyed the repeatability of the work, eliminating the need to normalize data to directly compare different experimental replicas.

Qualitative assessment of cell junction appearance
To qualitatively assess the effect on cell-cell junctions of APC-driven actin nucleation impairment, a criterion based on both the healthiness and morphology of the cell junction of each cell was established. We divided cells in three different categories: 'normal', 'discontinuous', and 'round'. Normal cells were defined as cells with all, or at least three cell junctions labelled with F-actin throughout the cell junction (i.e. continuous signal), whereas discontinuous cells were those cells with two or more disrupted cell junctions, therefore not all cell junctions were properly labeled. 'Round' cells were defined as cells that have lost their typical morphology and acquired roughly a circular shape. Mitotic cells were excluded from the measurements. Qualitative count was performed in a blind way by several members of the team for three different sets of images. The percentage of cells corresponding to each category was calculated using Microsoft Excel, and then plotted with GraphPad Prism (version 9.0; GraphPad Software, La Jolla, CA).

Cell size quantification
For cell size quantification of individual cells in monolayers, we first generated TIFF files of the middle frame of F-actin immunofluorescence images using Fiji. Then, individual cell size was quantified using a deeplearning-based segmentation method called CellPose ( 40 , RRID:SCR_021716). Briefly, parameters to segment cells were chosen in function of the cell diameter, which we estimated using Fiji. Images were loaded into CellPose and ''CP'' was selected as model to segment cells. The outlines of the cells were saved as .txt. Original TIFF files were opened in Fiji, the macro called imagej_roi_converter.py was run and outlines were over imposed on the monolayer image. For accurateness of the results, cells from the edges of the images were manually scrutinized and removed from the counts. We measured cell areas which were saved as .CSV. Data was then analyzed and plotted with GraphPad Prism (version 9.0; GraphPad Software, La Jolla, CA).

Flow cytometry DNA content analysis
Initially, cell cycle phase distribution was obtained in a CytoFLEXâ flow cytometer (Beckman Coulter; 405 nm laser). Next, cells were gated by size and granularity on forward scatter (FSC)/side scatter (SSC) plot and cell debris was excluded from the analysis. Mean fluorescence intensity from 10,000 cells/events was scored for each sample. Further quantitative data analyses to gate out debris were performed using the FlowJoä software version 10.8.1 (BD Life Sciences), a software application with an integrated environment for viewing and analyzing flow cytometric data. Plots were generated in GraphPad Prism.

Cell migration analysis
Individual cells migrating at the edge of the wound were tracked using the manual tracking from ImageJ, whose coordinates were saved as a .txt file. Those files were then exported into Chemotaxis and Migration Tool version 2.0 software (Ibidi, Germany RRID: SCR_022708) to obtain migratory related parameters such as the directionality, speed [mm/min], Euclidean Distance [mm], and Accumulated Distance [mm] of each cell tracked Data were combined from time-lapse image series collected from at least three independent experimental days, then plotted with GraphPad Prism (version 9.0; GraphPad Software, La Jolla, CA).

Cell junction analysis
To measure cell junction length, a line was manually drawn on the cell junction using the ''segmented line'' tool in Fiji and line length was measured using the ''measure'' tool. To calculate the junction angle, a ll OPEN ACCESS iScience Article statistical differences between replicates. This way of representing data linked paired measurements together and conveyed the repeatability of the work, eliminating the need to normalize data to directly compare different experimental replicas. Differences were considered significant if p-value was <0.05 (*), <0.01(**), <0.001(***), or < 0.0001 (****), as indicated in each figure legend.