In colon cancer cells fascin1 regulates adherens junction remodeling

Adherens junctions (AJs) are a defining feature of all epithelial cells. They regulate epithelial tissue architecture and integrity, and their dysregulation is a key step in tumor metastasis. AJ remodeling is crucial for cancer progression, and it plays a key role in tumor cell survival, growth, and dissemination. Few studies have examined AJ remodeling in cancer cells consequently, it remains poorly understood and unleveraged in the treatment of metastatic carcinomas. Fascin1 is an actin‐bundling protein that is absent from the normal epithelium but its expression in colon cancer is linked to metastasis and increased mortality. Here, we provide the molecular mechanism of AJ remodeling in colon cancer cells and identify for the first time, fascin1's function in AJ remodeling. We show that in colon cancer cells fascin1 remodels junctional actin and actomyosin contractility which makes AJs less stable but more dynamic. By remodeling AJs fascin1 drives mechanoactivation of WNT/β‐catenin signaling and generates “collective plasticity” which influences the behavior of cells during cell migration. The impact of mechanical inputs on WNT/β‐catenin activation in cancer cells remains poorly understood. Our findings highlight the role of AJ remodeling and mechanosensitive WNT/β‐catenin signaling in the growth and dissemination of colorectal carcinomas.

is an actin-bundling protein that is absent from the normal epithelium but its expression in colon cancer is linked to metastasis and increased mortality. Here, we provide the molecular mechanism of AJ remodeling in colon cancer cells and identify for the first time, fascin1's function in AJ remodeling. We show that in colon cancer cells fascin1 remodels junctional actin and actomyosin contractility which makes AJs less stable but more dynamic. By remodeling AJs fascin1 drives mechanoactivation of WNT/β-catenin signaling and generates "collective plasticity" which influences the behavior of cells during cell migration. The impact of mechanical inputs on WNT/β-catenin activation in cancer cells remains poorly

| INTRODUCTION
Adherens junctions (AJs) maintain the physical integrity of the mature epithelium, but they also allow cells to respond to mechanical and biochemical signals by facilitating cell movement, intracellular signaling, and gene transcription. 1,2 The crosstalk between the E-cadherin-βcatenin complex (CCC) and the contractile forces generated by the actomyosin cytoskeleton controls AJ assembly and stability. 3,4 Notwithstanding their relative stability in the mature epithelium, AJs are continuously remodeled during morphogenesis, tissue regeneration, and cancer progression. [4][5][6] While much is known about AJ assembly in epithelial cells, AJ remodeling is less well understood. Dysregulation of AJs plays a critical role in malignant transformation and metastasis. 5,6 Increased AJ plasticity also regulates collective cell migration. Collective cell migration is advantageous for tumor cells because it eliminates the need for all cells to detect external signals for migration and it couples the mechanical forces among cells for maximum plasticity making them more responsive to motogenic cues. [7][8][9] Circulating tumor cells that migrate collectively have a 50-fold higher metastatic potential compared with single circulating tumor cells. [10][11][12] Accordingly, tumor cells that acquire AJ plasticity are more aggressive, with higher metastatic risk and poorer clinical outcome. [12][13][14] The mechanisms guiding singlecell migration are reasonably well understood, however, cell-intrinsic mechanisms that steer collectively migrating cells are not fully resolved. For instance, collective cell migration involves the integration of guidance cues between functionally and morphologically distinct cell populations (e.g., leader-follower cells) that are not fully defined. 15 During morphogenesis and during cancer invasion the organization of the AJs varies considerably and atypical patterns of collective cell migration appear when sheets or strands of cells or detached clusters of cells that maintain remodeled cell-cell adhesions migrate as loosely cohesive groups of cells. [15][16][17] Such "collective plasticity" occurs when tumor cells infiltrate local tissue. [18][19][20] It is generally agreed that "collective plasticity" allows tumor cells to adapt as they navigate their way through confined spaces. 21 Hybrid epithelial-mesenchymal transition (EMT) is implicated in AJ remodeling and it contributes to invasion of loosely connected tumor cells. 18,19 Identification of the molecular mechanisms responsible for "collective plasticity" is of great importance for the prevention and treatment of metastatic cancers. We note that despite the implications that pathological remodeling of AJ proteins is integral to malignant transformation and tumor invasion, very few studies have examined AJ remodeling in cancer cells.
The CCC also regulates signaling via βcatenin that functions as a transcriptional co-activator of canonical WNT signaling. [22][23][24] Independent of WNT ligand and Frizzled receptor, mechanoactivation of WNT signaling is a wellconserved gene regulatory mechanism. 25 Additionally, direct mechanical activation of the WNT receptor Frizzled leads to WNT/β-catenin activation. 26 It has also been proposed that the mechanical stretching of the E-cadherin-βcatenin binding site controls mechanoactivation of WNT signaling. 27 In vivo studies have shown that imposing uniaxial deformation to Drosophila embryos is sufficient to induce nuclear localization of βcatenin and the expression of WNT target genes. 28 Mechanical strain-induced nuclear accumulation of βcatenin also occurs in gastrulating zebrafish embryos. 29 Mechanoactivation of WNT signaling is also well-documented in osteoblasts, human mesenchymal stem cells and endothelial cells exposed to laminar shear stress. [30][31][32][33] Mechanotransduction in epithelial cells was identified following the discovery that longterm mechanical pressure exerted by colon tumor growth triggers proliferation of surrounding non-tumorigenic cells via WNT/β-catenin activation. 34 Additionally, mechanical strain applied to quiescent epithelial cells causes their rapid reentry into cell cycle mediated by the nuclear accumulation of βcatenin. 35 Changes in AJ actomyosin contractility, matrix stiffness, Rho-associated protein kinase, and phosphorylation of βcatenin are implicated in the mechanoactivation of WNT/β-catenin signaling. 27,[35][36][37][38] In spite of that, the molecular mechanisms regulating WNT mechanoactivation in epithelial cells remain poorly defined. understood. Our findings highlight the role of AJ remodeling and mechanosensitive WNT/β-catenin signaling in the growth and dissemination of colorectal carcinomas.

K E Y W O R D S
adherens junctions, bidirectional migration, carcinomas, collective migration, Fascin1, mechanotransduction, WNT signaling Fascin1 is an actin-bundling protein expressed throughout the fetal gut but its expression is lost in the adult intestinal epithelium. 39,40 However, fascin1 is reexpressed in colorectal carcinomas (CRCs), where its expression correlates with a clinically aggressive disease with a higher incidence of metastasis and poorer outcome. [41][42][43][44] Fascin1 is both a prognostic marker and driver of colorectal carcinogenesis. In CRC patients fas-cin1 is an independent negative factor for survival and increases the risk of both disease recurrence and death by sevenfold. 41,45,46 Animal studies have shown that overexpression of fascin1 in a CRC cell line increases metastatic lesions by tenfold. 44 Moreover, constitutive overexpression of fascin1 in the intestinal epithelium of an Apc-mutated mouse accelerates tumor initiation and progression. 47 The general assumption is that fascin1's function in filopodia and invadopodia assembly propels colorectal tumor cell migration and invasion. 48,49 However, the role of fascin1 in the etiology of CRCs has not been directly investigated.
In the present study, we investigated the molecular mechanism of AJ remodeling in colon cancer cells and identified for the first time, fascin1's function in AJ remodeling. We show that by rearranging the AJ actin cytoskeleton fascin1 drives mechanoactivation of WNT/β-catenin signaling. Additionally, we find that fascin1 assembles ultra-long, filamentous cell-cell junctions that have their origins in AJs. These filamentous junctions (FJs) are important for "collective plasticity" and regulate migration of loosely cohesive groups of cells. More importantly, we discovered that fascin1 regulates bidirectional cell migration. This is, to the best of our knowledge, the first identification of an actin-binding protein that regulates cell-intrinsic reverse and bidirectional cell migration. Our findings provide evidence that in vivo these functions of fascin1 play a critical role in tumor growth and metastasis. Since fascin1 is expressed in a wide range of carcinomas, our findings could have wider implications for understanding AJ plasticity and its role in carcinogenesis. 43

| Tumor spheroid culture
Tumor spheroids were generated as described previously. 53 Briefly, cells were grown on standard tissue culture dishes to 80% confluence and were replated 24 h prior to initiating tumor spheroids. Cells were detached with 0.25% trypsin containing 0.2 g/L EDTA and plated on agarose-coated 10cm dishes at a concentration of 1 × 10 6 cells/mL in DMEM with 10% fetal bovine serum (FBS). Spheroids were morphologically evaluated with a Nikon Eclipse TE2000-U inverted microscope equipped with a CoolSnap ES chargecoupled device (CCD) camera. For hanging drop cultures, 2000 cells in 20 μL drops were plated on the inside lids of 10-cm dishes, and the lids were inverted prior to incubation, allowing for the drops to be maintained in suspension.

| Immunoblotting
Standard protocols were followed for Western blotting. 54 Cell lysates were prepared as described previously. 54 Briefly, proteins were separated by SDS-PAGE on 8%-12% polyacrylamide gels. Proteins were transferred to 0.45-μm nitrocellulose membranes. Membranes were blocked with milk, incubated with primary antibodies and corresponding HRP-conjugated secondary antibodies, and then developed by ECL. All Western blots are representative of at least three independent experiments with similar findings.

| MRC-5 conditioned media
preparation MRC-5 conditioned media was prepared as described previously. 55 MRC-5 cells were seeded into 10-cm dishes at 1 × 10 4 cells/cm 2 and grown to confluence in DMEM low glucose medium with 10% FBS and 1% penicillin-streptomycin. Once cells reached confluence, fresh medium was added, and cells were incubated for 2 days. The conditioned medium was collected and passed through 0.45-μm filter to remove floating cells and debris and stored at −20°C.

| Cell migration assay
Migration assay was performed as described previously. 52 Briefly, cells were plated on 6-well tissue culture dishes and allowed to grow to confluence. Prior to the experiment, cells were serum-starved for 24 h. To initiate wound healing, the monolayers were wounded with a sterile blade. Individual wells were rinsed with phosphate-buffered solution (PBS), and cells were treated with MRC-5 conditioned media (diluted in 3 volumes of DMEM with 10% FBS). 250 μL of 5 μg/ mL of mitomycin C was added to inhibit cell proliferation. Cells were treated for 20 h prior to imaging.

| Scratch-wound image analysis
For image analysis, cells were identified through a series of filters applied to individual images using Python, which produced binary images. Adjacent frames were shifted to maximize the total overlap in the binary images, and cells were assigned to tracks based on maximum overlap. Cells unaccounted for by this method were assigned tracks by minimizing the centroid displacement between remaining cells. Edge cells were identified as being within 75 μm of the migratory front. Loose and bulk cells were identified by their position relative to the migratory front and edge cells. Instantaneous centroid velocities were used for further analysis. Cell tracks were detected in a minimum of 10 frames and analyzed.

| Mature AJ remodeling assay
For mature AJ remodeling assays, 5 × 10 4 cells were plated in 6-well tissue culture dishes and allowed to form small colonies (<40 cells). The cells were treated with MRC-5 conditioned media (diluted in 3 volumes of DMEM with 10% FBS) and imaged every 5 min. Images were captured with a Nikon Eclipse TE2000-U inverted microscope and CoolSnap ES CCD camera.

| Invasion assay
For cell-invasion assays, tumor spheroids formed by the hanging drop method were isolated 96 h after suspension culture and embedded in a Collagen I matrix (2 mg/mL). The Collagen I matrix was allowed to solidify for 90 min prior to being overlaid with DMEM containing 10% FBS and 50 ng/mL hepatocyte growth factor (HGF). The invasion process was allowed to occur over a period of 8 days and data collected at the end of Day 8.

| Soft agar colony formation assay
A layer of 0.5% agarose-containing DMEM was solidified at the bottom of 6-well dishes to prevent cells from attaching. Cells were mixed into liquid 0.33% agarose at the desired cell numbers and allowed to solidify on top of the 0.5% agarose layer. For HT-29/19A-Scr and -shF cells 500 cells were used that were grown in DMEM with 10% FBS. Colonies greater than 0.05 μm 2 in area were counted.

| Mice
NOD-Scid IL2Rgamma null mice (NSG) were purchased from Jackson Laboratory, (Bar Harbor, ME). For xenograft experiments, mice were injected subcutaneously in the right flank with 2 × 10 6 HT-29/19A-Scr and -shF cells stably expressing Luciferase in a 1:1 mixture of Matrigel dissolved in PBS. Three weeks post-injection mice were imaged using an IVIS Imaging System (PerkinElmer), euthanized, and the weight of the xenografts measured. For secondary tumor growth, mice were injected intra-splenic with 1 × 10 6 HT-29/19A-Scr and -shF cells in a 1:1 mixture of Matrigel dissolved in PBS. Three weeks post-injection mice were euthanized, and secondary tumor growth was measured by the ratio of liver to total body weight.
Approximately 300-500 labeled cells (100 cells/mL) were microinjected into the yolk sac of 2-day-old embryos using a pressure injector (Harvard Apparatus, Holliston, MA) and manipulator (MM33-Righ, Märzhäuser Wetzlar, Germany). Glass needles (1 mm in diameter, Sutter Instrument Company, Novato, CA) were used for the microinjection. Following the microinjection, zebrafish were incubated at 34°C for 2-3 h, washed in E3 media, followed by removal of embryos that had been injected into the vasculature. Successfully injected embryos with cells in the yolk sac were placed in 96-well plates and incubated at 34°C. Embryos were imaged for tumor cell dissemination at 24, 48, and 72 h post-injection. Images were captured using an OLYMPUS IX51 fluorescence microscope equipped with an OLYMPUS XM10 camera and cellSens Dimension software (Olympus, Center Valley, PA, USA).

| Immunofluorescence and live cell fluorescence microscopy
All fluorescence images were acquired using the Olympus FV2000 inverted confocal microscope. For immunofluorescence analysis, cells were fixed and permeabilized for 30 min with 3.7% paraformaldehyde and 0.3% Triton X-100 dissolved in PBS. Unless otherwise indicated all data were obtained by performing three independent experiments with n = 10 fields/group and >300 cells/group analyzed each time. Nuclei were counterstained with DAPI. Tumor spheroids were studied in three independent experiments with 10 spheroids per group in each experiment. Plot profiles were generated by measuring the distribution of proteins in regions of interest (ROI) highlighted by a white line, 10 μm long.

| Chromatin immunoprecipitation (ChIP)
ChIP was carried out as described previously. 56 Briefly, cells were plated in 10-cm dishes and fixed with 1.5% paraformaldehyde after 24 h to chemically cross-link protein and DNA complexes. Cells were lysed, nuclei were purified, and then chromatin was sheared by sonification. Immunoprecipitation was carried out with βcatenin monoclonal antibody and Protein G magnetic beads. For PCR, the following primers were used:

| Quantitative RT-PCR
HT-29/19A-Scr and -shF cells were grown to 80% confluence in 6-well dishes. Total cellular RNA was extracted using TRIzol Reagent. cDNA was synthesized using iScript cDNA synthesis kit according to the manufacturer's instructions. Quantitative RT-PCR was carried out with 10 ng of cDNA in a total volume of 10 μL using SsoAdvanced universal SYBR Green Supermix and assayed on BioRad CFX 96 real-time PCR detection system. The following cycle parameters were used: 95°C for 30 s, 40 cycles of 95°C for 10 s and 62°C for 30 s. The CFX manager software provided with the thermocycler (Bio-Rad, Hercules, CA) was used to determine Ct values. The following primer sets were used for quantitative RT-PCR: 2.2.14 | Reporter assay HT-29/19A-Scr and -shF cells were transfected in 24-well dishes using Lipofectamine 2000 with βcatenin reporter Super 8X TOPFlash and pSV-β-Galacosidase control vector. All cells were lysed approximately 24 h post-transfection, and luciferase activity was measured with Luciferase Assay kit and βgalactosidase activity was measured with chlorophenol red-β-D-galactopyranoside (CPRG). Luciferase reporter activity was normalized to βgalactosidase activity. Luminescence was quantified with the VICTOR Multilabel Plate Reader (PerkinElmer, Waltham, MA, USA).

| Fascin1 and β-catenin immunohistochemistry scoring
Tumor tissue microarray from Stage IV CRC patient cohort were immunostained for fascin1 and total βcatenin. Fascin1 was scored using the method of Hashimoto et al. 41 Fascin1 and βcatenin expression in these tumor samples was evaluated by immunohistochemistry performed in paired tumor panels from the same tumor.

| Statistical analysis
Data are expressed as mean ± standard deviation (SD) or ± standard error of mean (SEM). Statistical significance was determined by Student's t-test, one-way analysis of variance (ANOVA), Mann-Whitney U test, Kuiper's test, or Watson U2 test. p < .05 was considered to be statistically significant.

| Study approval
All animal studies were performed in accordance with animal protocols approved by the University of Houston Institutional Animal Care and Use Committee.
All authors had access to the study data and have reviewed and approved the final manuscript.

| Fascin1 is upregulated in a significantly large number of CRCs
There is no consensus on the frequency of fascin1 expression in CRCs either because of the small size of patient cohort or because these studies have been performed in colorectal cancers at different stages of carcinogenesis. We elected to analyze fascin1 expression in a large cohort of 299 stage IV CRCs, from patients who participated in the Phase III MAX clinical trial. 57 Stage IV identifies cancers that have spread from the colon to distant organs and tissues; and cancers that have a relative 5-year survival rate of 5%-10%. 58 We analyzed these tumors for fascin1 protein expression using the Hashimoto et al scoring criteria. 41 In this cohort, we identified low to high (score 1-3) fascin1 expression in 34.4% of cases, and moderate to high (score 2-3) fascin1 protein levels in 22.4% of the cases ( Figure 1A). Additionally, we determined that fascin1 mRNA is more highly expressed in high grade/poorly differentiated CRCs (G1 > G2 > G3; Figure 1B) confirming fascin1's association with more aggressive colorectal disease. These findings indicate that understanding the underlying molecular mechanism of fascin1 could improve treatment and increase survival for at least a third of patients with advanced colorectal cancer.

| Fascin1 remodels adherens junctions and assembles ultra-long, filamentous cell-cell junctions
To identify the function of fascin1 in the pathogenesis of colorectal cancer, we performed loss-of-function studies  Figure S1A). 59 We found that spheroids of HT-29/19A parental cells and cells expressing scrambled shRNA (HT-29/19A-Scr) grew as loosely adherent cluster of cells with a "grape-like morphology." Comparatively, fascin1 KD cells (HT-29/19A-shF) formed tightly organized, compact spheres ( Figure 1C), implicating a role for fascin1 in the remodeling of cell-cell adhesions. In sub-confluent cell cultures, HT-29/19A-Scr cells formed nascent cellcell adhesions that were ultra-long, relatively large, and resembled filopodia with distal ends that morphed into lamellipodia-like structures as two cells made cell-cell contact ( Figure 1D). In contrast, HT-29/19A-shF cells extended much smaller lamellipodial protrusions as two cells approached each other, similar to nascent adherens junction (nAJ) assembly by normal epithelial cells. 60 Notably, HT-29/19A-Scr cells formed these protrusions even in the absence of direct cell-cell contact highlighting fascin1's autonomous role in their assembly ( Figure S1B). Alternatively, HT-29/19A-Scr cells formed ultra-long intercellular conduits that lacked the lamellipodia-like distal ends ( Figure 1E). These remodeled intercellular junctions contained fascin1, F-actin and tubulin, as well as E-cadherin implying these protrusions, are nAJs and that fascin1 appropriates the AJ assembly machinery to construct these ultra-long filamentous cell-cell junctions ( Figure 1D-F). By measuring the rate of their assembly, we determined that fascin1, F-actin, and tubulin are all required to assemble these nAJs ( Figure S1C,D). To identify fascin1's role in AJ remodeling independent of oncogenic transformation, we performed gain-of-function studies by overexpressing EGFP-tagged fascin1 in the normal epithelial MDCK cells ( Figure S1E). Compared with HT-29/19A cells, morphologically similar, ultra-long and large cellcell junctions were also identified in MDCK EGFP-Fascin1 cells (Figure 2A). These filamentous junctions in EGFP-Fascin1 cells were remarkably large and included within them functional organelles like mitochondria and lysosomes ( Figure 2B). Extraordinarily these ultra-long cell-cell junctions were also seen in confluent cultures of EGFP-Fascin1 cells often connecting two significantly distant cells ( Figure 2C, cell 1 connected to cell 6). These elongated intercellular junctions were not attached to the bottom but hovered over the monolayer facilitating distant cell-cell connections ( Figure 2D). To further analyze fascin1's role in AJ remodeling, we treated small clusters of EGFP and EGFP-Fascin1 cells with conditioned media from MRC-5 cells (as a source of hepatocyte growth factor (HGF)) at concentrations that did not induce epithelial cell scatter. 61 The EGFP cells maintained stable cellcell adhesions with few short-lived changes (Video S1, Figure 2E). In contrast, the EGFP-Fascin1 cells continuously remodeled their cell-cell adhesions, dissolving, and reforming them resulting in loosely connected cell clusters linked by ultra-long, filamentous junctions (Video S2, Figure 2E). These dynamically remodeled cell-cell adhesions in EGFP-Fascin1 cells differed in length and lifespan both of which were positively correlated ( Figure 2F). The smaller changes in cell-cell adhesions of EGFP cells displayed a similar correlation between size and lifespan ( Figure 2F). However, FJs assembled by EGFP-Fascin1 cells were much longer and had a much longer lifespan compared with the small changes in cell-cell adhesions seen in EGFP cells ( Figures 2F and S1F). These data indicate that in the presence of fascin1, epithelial AJs are more receptive to extracellular cues and respond by undergoing dynamic remodeling. To the best of our knowledge, these data identify for the first time ultra-long, filamentous cellcell junctions that have their origins in AJs. Based on that we have elected to call these cell-cell adhesions filamentous junctions (FJs).

| Fascin1 transforms adherens junction mechanotransduction
Normal epithelial cell-cell junctions have two spatially distinct actin pools: the junctional actin and peripheral actomyosin bundles also called circumferential bundles ( Figure 3A). 62 The dynamic junctional actin pool is located beneath the plasma membrane and is bound to CCC. 62 During AJ assembly, newly polymerized actin is incorporated into this pool. The circumferential bundles are less dynamic and are the major contractile actin pool at the AJs that are aligned parallel to the cell-cell junctions. 62 They consist of αactinin bundled actin filaments associated with non-muscle myosin II (NMII). The alignment of the circumferential bundles parallel to the cell-cell junctions, requires the actin-bundling protein EPLIN. 63 In the mature epithelium, these two pools of actin become indistinguishable and appear as cortical actin at the cell-cell boundaries. 64 Our data show that the fascin1 KD HT-29/19A-shF cells display F-actin arranged parallel to the cell-cell boundaries and E-cadherin that co-localizes with the F-actin, similar to normal epithelial cells ( Figure 3B). However, in the fascin1 expressing HT-29/19A-parental and -Scr cells, the AJ F-actin was aligned perpendicular to the cell-cell boundaries and appeared as intercellular stress fibers with the E-cadherin co-localized to the ends of these stress fibers ( Figure 3A,B, Top; Figure S2A). The  Figures 3A,B and S2A). The discontinuous AJs in parental and HT-29/19A-Scr cells formed as a consequence of stress fiber-induced tension acting on both sides of the cell-cell junctions ( Figure 3A,B, Top, Figure S2A) rather than changes in the overall protein levels of E-cadherin between HT-29/19A-Scr and -shF cells ( Figure S2B). This remodeling was limited to the AJs, and there was no discernable difference in the lateral junctions between HT-29/19A-Scr and -shF cells ( Figure 3B, Middle). Fascin1's AJ remodeling function was reconstituted in the rescue experiments performed in the fascin1 KD HT-29/19A-shF cells by overexpressing EGFP-Fascin1 ( Figures S1A and 3C). We performed gain-of-function studies in HT-29 cells, which lack endogenous fascin1 protein, by overexpressing EGFP-Fascin1 ( Figure S2C). HT-29 EGFP-Fascin1 cells recapitulated the effects of endogenous fascin1, confirming fascin1's direct role in AJ remodeling ( Figure 3D). This property of fascin1 was also replicated in MDCK EGFP-Fascin1 cells ( Figures S1E and S2D). These data demonstrate that fascin1's AJ remodeling function is independent of oncogenic transformation. In the colon cancer cells, fascin1 was associated with the junctional actin pool and co-localized with E-cadherin in the discontinuous AJs ( Figure 3E). In contrast, αactinin a major actin-bundling protein of circumferential bundles localized to the stress fibers that formed end-on attachments with the E-cadherin complex ( Figure 3F). These data indicate that in colon cancer cells fascin1 and αactinin segregate into distinct AJ-associated actin networks. In vitro an intrinsic behavior of fascin1 and αactinin is to mutually exclude each other and segregate into separate actin bundles. 65 We reasoned that fascin1 re-expression in CRC cells could competitively displace AJ-associated actin-bundling proteins like αactinin and potentially EPLIN. We speculated that this could be the molecular mechanism of fascin1 responsible for AJ remodeling in colon cancer cells. Displacement of EPLIN would have significant implications because EPLIN stabilizes AJs by inhibiting actin turnover and EPLIN is required for the parallel alignment of circumferential bundles at the cell-cell boundaries. 63 Moreover, both αactinin and EPLIN regulate AJ mechanosensitivity, which would implicate fascin1 in the remodeling of AJ mechanotransduction. 66,67 As expected, in HT-29/19A-Scr cells we identified a displacement of αactinin and EPLIN away from the AJs to the cytoplasm ( Figure 4A,B). Fascin1 also displaced αactinin from focal adhesions (FAs) which implies that this intrinsic behavior of fascin1 and αactinin also remodels FAs in colon cancer cells ( Figure S2E). Notably, FAs in HT-29/19A-Scr cells were associated with fewer stress fibers compared with the shF cells. There was no significant difference in total αactinin protein levels between control and fascin1 KD HT-29/19A cells ( Figure S2F). To quantitatively measure changes in F-actin at the AJs, we transfected HT-29/19A-Scr and -shF cells with mCherry-Lifeact ( Figure 4C). In these cells,  Figure 2E).
In epithelial cells, the E-cadherin complex integrates the effects of mechanical signals by modifying the AJ recruitment of αcatenin and vinculin, the two major mechanosensors. In response to intercellular actomyosin tension, monomeric αcatenin (α M ) couples the Ecadherin-βcatenin complex to the junctional actin and recruits vinculin which further strengthens the AJs by Actin binding to both βcatenin and the junctional actin. [68][69][70] In this manner, vinculin and αcatenin cooperatively couple the cadherin-catenin dynamics with mechanical signaling. Our findings with αactinin and EPLIN implicate a role for fascin1 in the remodeling of AJ mechanotransduction.
To validate these findings, we examined the AJ recruitment of vinculin and the assembly of the vinculin-αcatenin complex in HT-29/19A-Scr and -shF cells. Our data show that in fascin1 expressing colon cancer cells vinculin is displaced from the AJs (apical) and FAs (basal) ( Figures 4D and S2E). Not surprising then is our finding that fascin1 KD in HT-29/19A-shF cells restored vinculin localization to the cell adhesion sites similar to vinculin's distribution in normal epithelial cells (Figures 4D and  S2E). 71 There was no significant difference in total vinculin protein levels between HT-29/19A-Scr and -shF cells ( Figure S2F). Quantitative co-immunoprecipitation (IP) assay with an antibody that specifically immunoprecipitates monomeric αcatenin revealed that fascin1 knockdown results in higher levels of vinculin in complex with αcatenin ( Figure 4E). The αcatenin that dissociates from the cadherin-catenin complex forms a homodimer (α D ) as seen in input sample ( Figure 4E). Together these findings demonstrate that fascin1 expression in CRC cells is associated with the redistribution of key AJ mechanosensors, that is, αactinin, EPLIN, αcatenin, and vinculin.

| Fascin1 activates mechanosensitive WNT/β-catenin signaling
We reasoned that by remodeling AJ mechanotransduction fascin1 could influence gene transcription in CRC cells, for example, by mechanoactivation of WNT/β-catenin signaling. Indeed, subcellular fractionation revealed that fascin1 KD significantly inhibits the nuclear accumulation of βcatenin (60.2%, n = 3, p < .001) and causes a redistribution of βcatenin from the nucleus and cytoplasm to the AJs without impacting total βcatenin or E-cadherin protein levels ( Figures 5A and S2B). Most notably, reexpression of EGFP-Fascin1 in the HT-29/19A-shF cells reconstituted nuclear accumulation of βcatenin validating fascin1's direct role in βcatenin signaling ( Figures 5A  and S1A). Nuclear accumulation of βcatenin was also documented in actively invading tumor spheroids of HT-29/19A-Scr cells ( Figure 5B). These data imply that fas-cin1's WNT/β-catenin activation function also influences cancer cell invasion. Nuclear accumulation of βcatenin was validated in MDCK cells overexpressing EGFP-Fascin1 further confirming fascin1's unequivocal role in βcatenin signaling ( Figures S1E and S3A). In the CRC patient samples we noted that tumors with high fas-cin1 protein expression (score 2-3) also displayed widespread accumulation of nuclear βcatenin ( Figure 5C). Fascin1 KD in CRC cells reduced the expression of Wnt target genes including the marker for stemness, CD44; the ECM remodeling proteins, matrix metalloproteinase, MMP7, and MMP9; and the proto-oncogene C-MYC which regulates differentiation, cell cycle progression, self-renewal, and chemoresistance ( Figure 5D). 72,73 To further define the mechanism of Wnt target gene activation in HT-29/19A cells we performed chromatin immunoprecipitation (ChIP) assay with βcatenin antibody. The goal was to determine if fascin1 expression in colon cancer cells increases βcatenin binding to promoters of Wnt target genes. Our data show that fascin1 knockdown in HT-29/19A-shF cells significantly decreases βcatenin binding to the C-MYC, MMP9 and CCND1 promoters, whereas there was no change with MMP9 negative control (NC) ( Figure S3B). Fascin1's function as a driver of Wnt activation was validated with the Super8x TOP-Flash luciferase TCF-LEF reporter assay ( Figure S3C). Loss of WNT activation in the HT-29/19A-shF cells correlated with reduced anchorage-independent growth ( Figure 5E), restoration of tight junctions ( Figure 5F, left panel), and re-establishment of contact inhibition ( Figure 5F, right panel). In contrast, HT-29/19A parental and Scr cells displayed multilayering of cells (Figure 5F right panel; Figure S4A). Rescue experiments performed in HT-29/19A-shF cells overexpressing EGFP-Fascin1 also revealed loss of contact inhibition and multilayering of cells ( Figure S4B). Fascin1 overexpression in MDCK cells also results in multilayering of cells ( Figure S4C). At the same time, fascin1 KD in other CRC cell lines, for example, F I G U R E 4 Fascin1 Displaces adherens junction associated mechanosensors. (A) Confocal images of HT-29/19A-Scr and -shF cells show the distribution of E-cadherin (green) and αactinin (red). Nuclei were counterstained with DAPI. Plot profiles were generated by measuring the distribution of E-cadherin and αactinin in a ROI identified by a white line, 10 μm long. Scale bar, 10 μm. (B) Confocal image of HT-29/19A-Scr and -shF cells shows the distribution of E-cadherin (red), and EPLIN (green). Nuclei were counterstained with DAPI. Plot profiles were generated by measuring the distribution of EPLIN and actin in a ROI identified by a white line, 10 μm long. Scale bar, 5 μm. (C) Live cell imaging of HT-29/19A-Scr and -shF cells stably expressing mCherry-Lifeact. Nuclei were counterstained with Hoechst. Scale bar, 10 μm. Kymograph analysis of mCherry-Lifeact at AJs of HT-29/19A-Scr and -shF cells measured at two ROI (labeled 1 and 2; 5 μm each). Quantitative measurement of speed (μm/min) of migration and distance (μm) moved by AJ-associated F-actin and cross-linked F-actin filaments is shown on the right. Data shown are mean ± SEM. Statistical analysis was done using Student's t-test. Asterisk (**) denotes p < .01, n = 6. (D) Confocal images of HT-29/19A-Scr and -shF cells shows xz distribution of vinculin (green) and actin (red). Nuclei were counterstained with DAPI. Scale bar, 10 μm. Accumulation of cytoplasmic vinculin in Scr and shF cells was measured from three independent experiments. Data shown are mean ± SEM. Statistical analysis was done using Student's t-test. Asterisk (***) denotes p < .001, n = 3. (E) Total cell lysates from HT-29/19A-Scr and -shF cells were immunoprecipitated with αcatenin antibody and immunoblotted with vinculin antibody. Western blot shows co-immunoprecipitation of monomeric (α M ) αcatenin with vinculin. Homodimeric αcatenin (α D ) that does not associate with AJs can be seen in the Input samples. Data shown are mean ± SEM. Statistical analysis was done using Student's t-test. Asterisks (*) denote, ****p < .0001, n = 3.  HCT-116 recapitulate fascin1's effect on contact inhibition ( Figure S4D). These data suggest that fascin1's function in βcatenin signaling is conserved in CRC cells. Together, these studies predict that by activating mechanosensitive WNT target genes fascin1 could play a direct role in colorectal carcinogenesis.

| Fascin1 activates mechanosensitive WNT/β-catenin signaling to accelerate primary and secondary tumor growth
We postulated that mechanoactivation of WNT signaling is the molecular mechanism of fascin1 that drives colorectal carcinogenesis. To test this idea, we assessed the effects of fascin1 KD on primary and secondary tumor growth of CRC cells in vivo. Fascin1 KD significantly reduced primary tumor weight compared with control cells ( Figure 6A; p < .01, n = 6), and it reduced the number of actively proliferating tumor cells validating fascin1's role in colorectal tumor growth ( Figure S4E). More importantly, fascin1 KD induced a re-localization of total and active non-phospho (NP) βcatenin from the nucleus and cytoplasm to the cell membrane ( Figure 6B,C, respectively). The active (NP) βcatenin antibody was used to demonstrate that cytoplasmic βcatenin in the tumor cells was functionally active. Remarkably, in the primary tumors that express fascin1, we identified ultra-long FJs that resembled morphologically the FJs we identified in 2D cultures of HT-29/19A-Scr cells ( Figure 6D). These data indicate a role for FJs in colorectal tumorigenesis in vivo. Fascin1 KD also significantly reduced secondary tumor growth in mice ( Figure 6E; p < .01, n = 8) as well as prevented the cytoplasmic and nuclear accumulation of total and active βcatenin in the secondary tumors ( Figure 6E).
Similarly, fascin1 KD also inhibited metastasis in a zebrafish embryo metastasis assay ( Figure S4F) validating fascin1's conserved function in colorectal carcinogenesis.

| In CRCs fascin1 expression is linked to hybrid EMT
Since fascin1 remodels AJs we postulated that fascin1 re-expression in CRCs could influence signals regulating EMT. Additionally, fascin1 overexpression in epithelial ovarian cancers is linked to increased SNAI1 levels implicating a direct role for fascin1 in EMT. 74 We analyzed RNA-Seq data from CRCs profiled by the TCGA consortium, which revealed a significant inverse correlation between mRNA expression of fascin1 and the epithelial phenotype stability factors (PSFs) OVOL1 and GRHL2 ( Figure 7A) and conversely, a significant positive correlation with the metastable markers SNAI1/2/3. 75 The modest inverse correlation of fascin1 with OVOL and GRHL2 suggests dampening of epithelial characteristics but not complete EMT and the positive correlation with SNAI suggests stabilization of the hybrid epithelial-mesenchymal (E/M) state. We also identified a significant positive correlation between fascin1 and ZEB1/2 expression. Since stromal cells (e.g., fibroblasts) also express fascin1 and ZEB1/2 proteins, we also performed these analyses in RNA-Seq data obtained from 61 CRC cell lines, which revealed similar findings confirming these changes were tumor intrinsic ( Figure 7B; Table 1). Finally, we examined the expression of these TFs in the normal mucosa and intestinal tumors derived from mutant Apc mice. In the mutant Apc mice, there is significant upregulation of fascin1 and Snai3 mRNA in intestinal tumors but no change in the levels of Ovol1, Grhl2, or Zeb1/2 ( Figure 7C).

| Fascin1 regulates "collective plasticity" and bidirectional cell migration
Since collective cell migration is a hallmark of hybrid EMT we assessed the function of fascin1 in a wound-closure assay performed in MDCK EGFP and EGFP-Fascin1 cells. As expected, EGFP cells migrated as a sheet of tightly adherent cells a pattern associated with normal epithelial cell migration (Video S3; Figure 7D-F). Compared to the control cells, fascin1 overexpression significantly increased cell migration rates which is consistent with fascin1's canonical function in cell motility ( Figure 7D). 76 Unlike control cells, fascin1 expressing cells migrated as loosely coupled groups of cells resembling cell migration patterns previously associated with "collective plasticity" (Video S4; Figure 7D-F). 15 We identified three groups of EGFP-Fascin1 cells migrating at different speeds and with different patterns and we refer to them as loose, edge, and bulk cells (Video S4; Tables S2 and S3; Figure 7D-F). Closer to the wound edge, the loose and edge cells used FJs of variable lengths to connect the back of a leader cell (L) with the front of a follower cell (F) (Video S4 and Figure S5A). Compared to the edge and bulk cells, the migration pattern of loose cells was also unique and resembled more closely chain migration. By measuring the area of migrating bulk cells, it is clear that migrating bulk cells also remodel cell-cell adhesions (Video S4 and Figure S5B). Compared with EGFP cells, migrating EGFP-Fascin1 cells displayed large gaps within the migrating cell clusters, (Video S4 and Figure S5C). These gaps are indicative of increased tension at cell-cell adhesion sites and reveal continuous AJ remodeling during EGFP-Fascin1 cell migration. In the EGFP-Fascin1 cells we made the most extraordinary finding that at the leading edge some cells display significantly more backward migration events in the direction opposite from the wound edge which is quantitatively measured by the shape metrics (Video S4; Figures 7D-F and S5D, Tables S2 and S3).
Based on the analysis of X and XY (radial) coordinates, the backward migration of EGFP-Fascin1 loose cells was significantly different from the consistent forward migration of edge and bulk cells, and it was different from the forward migration of EGFP cells. Notably, the backward migration of a loose cell occurred concomitantly with the elongation of the FJ connecting the follower cells to the leader cell (Video S4). The backward migrating loose cells eventually become a part of the edge cell cluster and reverse their direction this time moving forward with the edge cells (Video S4 and Figure S5D). In some instances, the backward moving follower cells also brought back with them the leader cells (Video S4) These findings suggest that the backward migration may serve to limit the number of leader cells. Like fascin1, villin1 is an actinbundling protein that assembles filopodia and increases epithelial cell migration rates. 52,77 However, migrating MDCK cells overexpressing EGFP-Villin1 do not display "collective plasticity" or bidirectional migration underscoring the unique function of fascin1 in collective cell migration (Video S5). To the best of our knowledge, this is the first report of an actin-binding protein that regulates bidirectional cell migration.

| DISCUSSION
Using a large cohort of CRC patients, we show that fas-cin1 is upregulated in approximately a third of late-stage, aggressive, and metastatic CRCs which underscores the critical need to understand fascin1's function in colorectal carcinogenesis. Our data show that the side-to-membrane association of circumferential actin bundles seen in normal epithelial cells is replaced in colon cancer cells with end-to-membrane association of stress fibers. Contraction of circumferential actin bundles, that are aligned parallel to the membrane, seals epithelial AJs. However, contraction of stress fibers that pull from both sides of the cell-cell F I G U R E 6 Fascin1's Function in WNT/β-catenin mechanoactivation drives primary and secondary tumor growth in mice. (A) IVIS imaging of NOD scid gamma (NSG) mice injected subcutaneously with HT-29/19A-Scr and -shF cells. Fascin1 knockdown (KD) significantly decreased primary tumor weight. Scale bar, 5 mm. Data shown are mean ± SEM. Statistical analysis was done using Student's t-test. Asterisk (**) denotes, p < .01, n = 6. (B) Confocal immunofluorescence shows distribution of total βcatenin (red) in primary tumors from mice injected with HT-29/19A-Scr and -shF cells. Nuclei were counterstained with DAPI. Scale bar, 10 μm. (C) Immunohistochemistry shows distribution of active non-phosphorylated (NP) βcatenin in primary tumors from mice injected with HT-29/19A-Scr and -shF cells. Yellow arrowheads identify nuclear accumulation of active βcatenin. Scale bars, 100 and 50 μm. Data shown are mean ± SEM. Statistical analysis was done using Student's t-test. Asterisk (****) denotes, p < .0001, n = 6. (D) Confocal image of primary tumor from mice injected with HT-29/19A-Scr cells shows distribution of fascin1 (green) and actin (red). Nuclei were counterstained with DAPI. Scale bar, 10 μm. Higher magnification of boxed area is shown on the right and identifies a FJ (identified by arrowhead) in the mouse primary tumor. Scale bars 10 and 2.5 μm. (E) Immunohistochemistry of secondary tumors from mice injected with HT-29/19A-Scr and -shF cells shows distribution of fascin1, total βcatenin and active (NP) βcatenin in tumor (T) and surrounding normal (N) tissue. Yellow arrowheads identify nuclear accumulation of βcatenin in the secondary tumors. Scale bars, 100 μm and for magnified images, 30 μm. White arrowheads identify fascin1 expression in the secondary tumors. Scale bars, 100 and 30 μm. Data shown are mean ± SEM. Statistical analysis was performed using the Student's t-test. Asterisk (****) denotes, p < .0001, n = 8.
junctions results in unstable and discontinuous AJs in colon cancer cells (Figure 3). We propose that in colon cancer cells the pushing force generated by the protrusion of FJs and the pulling force of the intercellular stress fibers makes AJs unstable but also more dynamic (Videos S1 and S2; Figure 4C). Using different cell lines and by performing gain-of-function, loss-of-function, and rescue experiments, we demonstrate that fascin1 is necessary and sufficient for  AJ remodeling in colon cancer cells. The association of fas-cin1 with the junctional actin pool predicts fascin1's role in bundling newly polymerized actin in FJs that make cellcell contacts during the dynamic remodeling of AJs. When mixed together in vitro actin-bundling proteins like fascin1 and αactinin, cause protein sorting based entirely on their inter-filament spacing properties. 65 An important difference between fascin1 and αactinin crosslinked actin bundles is the size of the inter-filament spacing (8 vs. 35 nm, respectively). 65 Our findings indicate that a similar sorting of actin-bundling proteins at the AJs of colon cancer cells results in AJ remodeling. We propose that this difference in inter-filament spacing elicits protein sorting displacing from the AJs not only αactinin but also the NMII that associates with αactinin. We propose that the combined loss of αactinin, EPLIN, and NMII from the AJs of colon cancer cells results in the loss of compaction and orientation of circumferential actin bundles in colon cancer cells, forming instead intercellular stress fibers ( Figure 3). Our model is supported by the finding that circumferential actin bundles that form at the cell equator during cytokines lose their compaction and orientation only when there is a combined loss of actin-bundling proteins and myosin. 78 It is known that in vitro both fascin1 and αactinin assemble stable cross-linked actin networks but when these two structures are mixed together αactinin crosslinks impose unfavorable and energetically costly defects in hexagonally packed structures formed by fascin1 and vice versa. 79 We propose that similar energetically costly defects occur when fascin1 is re-expressed in CRC cells contributing to the instability of the junctional actin pool and instability of the CCC. EPLIN inhibits actin turnover and displacement of EPLIN from AJs of colon cancer cells favors actin polymerization and the assembly of FJs. 63 In this study, we also identify fascin1's role in the assembly of ultra-long and filamentous intercellular junctions. Despite their morphological resemblance, FJs are not similar to either tunneling nanotubes (TNTs) or tumor microtubes (TMs) both of which have their origins in gap junctions but not AJs. 80,81 FJs are also distinct from digitation junctions, which are much smaller than FJs and have no association with either gap junctions or AJs. 82 FJs are also distinct from adhesion zippers, which are filopodial protrusions associated with nAJs. 82,83 FJs are much longer (~80 μm) and larger (contain mitochondria and lysosomes) than adhesion zippers. We speculate that FJs may be unique to carcinomas that express fascin1. We speculate that the mitochondria in FJs likely support the high demand for ATP during actin polymerization in order to assemble FJs. It has been proposed that TNTs correspond to invadopodia in vivo while TMs in vivo have been shown to regulate tumor cell migration, invasion, and proliferation. 81,84 Since primary mouse tumors display FJs, we hypothesize a similar role for FJs in colorectal tumorigenesis. However, it is possible that FJs have other functions, for example, as conduits for cell signaling allowing cells at the invasive front and the tumor center to communicate directly. Alternatively, they may function like cytonemes, to traffic morphogens, or to generate morphogen gradients between different tumor cells. Since stromal cells express fascin1, morphogen trafficking between the stromal and tumor cells via FJs could influence tumor heterogeneity. Tumor heterogeneity is responsible for the failure of most cancer therapies, and this may be the molecular basis for fascin1's strong association with treatment-refractory CRCs. 46 Several studies have suggested a role for mechanoactivation of WNT signaling in tumor growth. 34,35,85 However, the molecular drivers of mechanically driven tumor growth remain unidentified. We note that despite its well-described role in cell motility, fascin1 expression in many types of high-grade carcinomas is in fact associated with tumor growth. 48 In this study, we demonstrate fascin1's direct role in mechanoactivation of WNT/β-catenin signaling and colorectal tumor growth. It is noteworthy that conditional expression of fascin1 in mutant Apc mice promotes tumor initiation, increases tumor aggressiveness, and accelerates tumor progression. 47 It is known that mechanical cues initiate tumor-like gene expression patterns in pre-neoplastic  tissues of the mutant Apc mice. We speculate that fas-cin1's function in mechanoactivation of WNT/β-catenin signaling could explain why conditional expression of fascin1 in the mutant Apc mice initiates and accelerates colorectal tumorigenesis. 86 Even though fascin1 upregulation is associated with colorectal cancer, fascin1 is not an oncogene and overexpression of fascin1 in the intestinal epithelium does not induce tumor formation. 47 Notably, fascin1 re-expression also occurs in the colon of patients with ulcerative colitis and Crohn's disease suggesting instead a role for fascin1 in intestinal repair and regeneration. 87 We propose that fascin1 upregulation in the background of genetic or epigenetic changes regulates mechanoactivation of WNT/β-catenin signaling which drives the emergence and progression of neoplasia. We note that HCT-116 cells carry an activating mutation in βcatenin despite that fascin1 KD restores contact inhibition in these cells underscoring the key role of fascin1 in βcatenin signaling. Using a reporter plasmid, fascin1 has been identified as a target rather than a driver of WNT/β-catenin signaling. 44 While we did not see a similar effect of WNT signaling on fascin1 expression, a feedback mechanism between WNT and fascin1 is possible. Here, we also show mechanoactivation of βcatenin signaling in invasive CRC spheroids ( Figure 5B). We propose these findings are significant because they imply that fascin1 can regulate proliferation of cells at tumor invasive fronts. Proliferation of cells at the invasive front occurs during embryogenesis when neural crest cells migrate over long distances to colonize the entire length of the gastrointestinal tract. 88 During embryogenesis proliferation drives neural crest cell invasion. We propose that increased proliferation at the invasive front of fascin1 expressing tumors could similarly drive rapid metastasis to distant organs. The expression of TFs such as OVOL and ZEB and OVOL and SNAI induce the hybrid E/M metastable state. 89 We find that fascin1 expression in CRCs is linked to the hybrid E/M state, which is consistent with fascin1's function in AJ remodeling. Our study shows that in CRC cells fas-cin1 remodels both AJs and FAs and we propose that both those changes likely influence collective migration of cancer cells. Previously it was reported that fascin1 reorganizes FA-associated stress fibers to enhance cell migration rates. 90 While this study did not identify the molecular mechanism of fascin1 responsible for FA remodeling, our study shows that in colon cancer cells competitive displacement of αactinin and vinculin by fascin1 remodels FAs-and FAassociated stress fibers (Figures 4D and S2E). We propose that recognizing the intrinsic property of fascin1 to sort other actin-bundling proteins based on their inter-filament spacing will also provide a molecular mechanism for many other non-canonical functions of fascin1. 48,91-93 Our study identifies a novel mechanism for collective migration of epithelial cells. Here, we demonstrate that fascin1 expressing epithelial cells use FJs as guidance cues while steering a loosely cohesive network of migrating cells. Previously, filamentous cell-cell junctions have been shown to regulate neural crest cell migration during embryogenesis, and glioma cell migration in brain tumors. 17,81 Although FJs are much longer, we speculate that they also share some similarity with the "cadherin fingers" which provide guidance cues to collectively migrating endothelial cells. 94 We hypothesize that the plasticity of AJs and the adaptability of FJs enhance the navigation capabilities of collectively migrating colon cancer cells. 52 Fascin1 regulates the backward migration of cells at the leading edge but this does not affect wound closure over time because only a small number of cells display this behavior. The backward migration reduced the number of leader cells which suggests it could play a role in maintaining directionality. We speculate that the backward migrating cells could function as "steering cells" that maintain directionality of movement for the entire group of collectively migrating cells. These findings also mean that the relative position of cells at the leading edge is not stable which implies that all cells at the leading edge of an invasive colorectal tumor could have the capability to instill directionality. Such behavior would also enhance tumor dissemination. In the developing gut, bidirectional migration of neural crest cells regulates the organization of the enteric nervous system. 95 Based on that we interpret our findings as reactivation of embryonic cell migration patterns, which may be a recapitulation of fas-cin1's function in the fetal gut. Alternatively, bidirectional migration may be a recapitulation of fascin1's function in neural crest cells. 96

AUTHOR CONTRIBUTIONS
Eric Pham, Amin Esmaeilniakooshkghazi, Sudeep P. George, Afzal Ahrorov, Fabian R. Villagomez, Michael Byington, Srijita Mukhopadhyay, Srinivas Patnaik, Monali Naik and Saathvika Ravi, Niall Tebuttt, Jennifer Mooi, and Camilla M. Reehorst performed the experiments described in this study. Jacinta C. Conrad analyzed the migration data and provided important guidance for the experimental design of these studies. John M. Mariadason is a long-standing collaborator of the contributing author and provided the patient cohort data, analyzed the patient cohort data, and provided valuable suggestions during the writing of this manuscript.

ACKNOWLEDGMENTS
This study was supported by National Institutes of Diabetes and Digestive Kidney Diseases (grant DK-117476 to S.K.); and the School of Allied Health Sciences (grant DK-56338).