Subepithelial Myofibroblasts Are Critical Regulators of Intestinal Epithelial Restoration

Intestinal epithelial migration is supposed to occur in a passive manner driven by the mitotic pressure exerted either by the cryptal stem cells, under physiological conditions, or the newly formed epithelium, upon damage. However, whether interactions between different neighboring cell types and with the matrix contribute to epithelial movement remain elusive. Here, we developed a novel three-dimensional in vitro intestinal mucosa model of gap closure, that includes both the epithelium and the basement membrane cellular compartments in a spatially relevant manner, to show that intestinal subepithelial myofibroblasts (ISEMFs) play a crucial role in epithelial restoration. ISEMF-derived biochemical cues boost epithelial proliferation and maintain epithelial barrier integrity. While, at the wounded area, ISEMF actively migrate to the epithelial front where they activate and generate ɑ-SMA contractile stress fibers along the direction of the epithelial cells promoting epithelial migration. Furthermore, ISEMF also deposit collagen paths that act as “guide rails” for directing IEC migration. Thus, the mere physical presence of ISEMFs can greatly accelerate restoration in wound healing, which suggests that ISEMFs should be recognized as new potential therapeutic targets.


Introduction
The small intestinal epithelium is formed by a monolayer of tightly packed and polarized epithelial cells arranged into crypt-villus units, in which one finger-like protrusion (villus) is surrounded by several epithelial invaginations (crypts) (Clevers, 2013;Leushacke and Barker, 2014). The intestinal epithelium is one of the fastest renewing tissues in mammals. This fast turn-over is fueled by the rapid division of the cryptal Lgr5 + intestinal stem cells (ISCs) which generate committed progeny that mature into different adult intestinal epithelial cell (IEC) types while following a rapid migratory path towards the villus tip where they undergo apoptosis and are spilled into the lumen (Barker et al., 2007;Barker, 2013). One of the main functions of the intestinal epithelium is to act as a physical protective barrier that prevents external insults from coming into contact with the immune system of the lamina propria. Any dysregulation of the normal epithelial homeostasis might cause the loss of intestinal epithelial barrier integrity. In such situations, an efficient restoration of the epithelium, including epithelial proliferation and migration, is critical to avoid the development of pathological inflammatory conditions that can lead to inflammatory bowel diseases (IBDs) or cancer development (Terzić et al., 2010;Karin and Clevers, 2016). Thus, intestinal epithelial migration is a crucial cellular process to maintain intestinal tissue homeostasis in both physiology and disease.
Not much is known about the contribution of mesenchymal cells in intestinal epithelial migration.
In the adult intestine, IECs sit directly above the basement membrane, which is composed of a reticular collagen fibrils sheet and intestinal subepithelial fibroblasts, including myofibroblastshenceforth referred to as intestinal subepithelial myofibroblasts (ISEMFs) (Roulis and Flavell, 2016), and telocytes (Aoki et al., 2016). Based on morphology, ultrastructure, location and molecular markers, intestinal telocytes are in retrospect counterparts of myofibroblasts and therefore considered, by many authors, equivalent populations of subepithelial mesenchymal cells (McCarthy, Kraiczy and Shivdasani, 2020). ISEMFs form a syncytium that surround the crypts and villi of the intestine (Richman et al., 1987;Federici and Boulis, 1999;Powell et al., 2005;Mifflin et al., 2019). They are characterized by the expression of intracellular cytoskeletal ɑ-smooth muscle actin (ɑ-SMA), bundles of myofilaments and are interconnected via gap junctions that allow them to contract similarly to smooth muscle cells (Hinz et al., 2001;Eyden, 2008;Eyden, Curry and Wang, 2011;Powell et al., 2011;Roulis and Flavell, 2016). It is known that resident ISEMFs possess secretion capacities and, together with other subepithelial mesenchymal cells, act as ISC niche components by producing molecular gradients that regulate Wnt and BMP signaling along the crypt villus axis. (Powell et al., 2011;Roulis and Flavell, 2016;Shoshkes-Carmel et al., 2018;McCarthy et al., 2020). They have also been shown, in vitro, to support intestinal epithelial cell growth and differentiation through paracrine signals (Lahar et al., 2011;Lei et al., 2014). Resident myofibroblasts are also found in other epithelia such as in neonatal lungs (McGowan et al., 2008;Branchfield et al., 2016), where they contribute to the alveologenesis (Branchfield et al., 2016) and provide the lungs with elasticity and tensile strength (Toshima, Ohtani and Ohtani, 2004). However, the function of resident myofibroblasts in the physiological migration and the restoration of the homeostasis upon tissue damage in epithelial tissues is not very well understood.
Analyses carried out in epithelia such as the skin (Gabbiani, Ryan and Majno, 1971), cornea (Gorono et al., 1992;Jester, Petroll and Cavanagh, 1999) or cleaved palate (Chitturi et al., 2015), indicate that fibroblasts and myofibroblasts could contribute to epithelial restoration (Tomasek et al., 2002;Hinz, 2010). The wound healing process in these tissues is associated with the progressive activation of fibroblasts to myofibroblasts in response to mechanical stress and transforming growth factor β1 (TGF-β1) signalling (Desmouliere et al., 1993;Vaughan, Howard and Tomasek, 2000;Hinz et al., 2001). Activated myofibroblasts express ɑ-SMA and develop stress fibers that exert contractile forces. These forces are directly transmitted to the ECM and induce the closure of the wound (Tomasek et al., 2002). In addition, activated myofibroblasts deposit collagen and express metalloproteases further contributing to the tissue matrix remodeling. However, it remains unclear whether these mechanisms occur in the intestine, and which is the specific role of the resident ISEMFs in promoting intestinal epithelial migration and restoration under both physiological and pathological conditions.
Here, we employ a three-dimensional (3D) intestinal mucosal model of gap closure to analyze the contribution of biochemical and mechanical ISEMF-derived signals to intestinal epithelial migration and restoration. Our 3D model includes primary IECs, primary ISEMFs, and the basement membrane matrix components, secreted by the ISEMFs, in a spatially relevant manner that mimics the tissue compartments in vivo. Our results indicate that ISEMFs are essential for a fast and efficient epithelial reconstitution. We show that ISEMs are important paracrine inducers of epithelial growth in our system. Furthermore, we prove that the physical presence of ISEMFs in the culture favors a directed epithelial migration. We demonstrate a coordinated migration with IECs and ISEMFs moving in a cooperative and guided manner, through their physical interaction, to facilitate the wound closure. Thus, we suggest resident ISEMFs as key regulators of active epithelial restoration of the intestinal epithelium through both paracrine and physical cues.

Intestinal epithelial migration is impaired in the absence of a basement membrane cellular compartment
To study intestinal epithelial migration in a gap closure process using a physiologically relevant setup, we employed a method previously developed in our lab to grow intestinal epithelial cell monolayers (Altay et al., 2019). Briefly, we seeded intestinal organoid-derived single cells on hard tissue plates coated with a thin film of Matrigel ® ( Figure 1A). Cells adhered to the matrix and grew into a cobblestone-shaped epithelial monolayer that recapitulated the in vivo-like cell type and composition. They organized into proliferating crypt-like domains and differentiated villus-like regions (Altay et al., 2019) ( Figure 1B). Furthermore, expression of the filamentous actin (F-actin) in the apical cell side and of the adherent junction protein β-catenin at the basolateral cell side demonstrate that the newly formed epithelial monolayers were mature and polarized ( Figure 1B, low right panels). To perform gap closure studies, we attached elastomeric barriers to the substrate before seeding with cells to spatially control the growth of the epithelial monolayer. By removing the barrier this creates a cell-free gap in the middle of the tissue culture plate ( Figure   1A). To track intestinal epithelial movement via live imaging we employed intestinal organoids, henceforth referred to as Lgr5-EGFP/RCL-tdT, which express high levels of tdTomato fluorescence in all cells (Figure 1 -Figure Supplement 1). With these tools we obtained live images to track intestinal epithelial movement for a period of 60h after barrier removal (Video 1).
For a successful gap closure, epithelial tissues need to migrate toward the gap maintaining tissue integrity until closure is achieved (Begnaud et al., 2016). In contrast, with our experimental setup the intestinal epithelial cells failed to close the gap. Specifically, the intestinal epithelial cells did not move in a cohesive manner and the migration front did not form properly. Instead, groups of cells appeared at the boundary between the tissue and the gap forming discrete protrusions ( Figure 1C, t = 0h, white arrows). Initially, these protrusions were similar to the finger-like structures observed in other types of epithelia (Reffay et al., 2011). However, epithelial cells in between protrusions did not advance forward, and the finger-like structures became more elongated and started to lose their directional movement towards the gap ( Figure 1C, t = 30h and t = 60h, white arrows). These abnormal advances of the leading edge led to the formation of holes in the sheet. Consequently, cells lost directionality and started to migrate back in the direction of these newly formed gaps (Figures 1C,D,white asterisks). Immunostaining analysis for β-catenin at different times after barrier removal confirmed the progressive loss of the epithelial tissue structure ( Figure 1D). Thus, the intestinal epithelial gap closure is impaired when only epithelial cells are present in the culture.
A recent study proposed the existence of an actin-related protein 2/3 complex-dependent active migratory force that would explain the physiological epithelial migration at the villi in vivo (Krndija et al., 2019). The appearance of protruding multicellular structures towards the gap in our results from an in vitro model of only intestinal epithelial cells would initially seem to support for this hypothesis ( Figure 1). However, the failure in completing the closure of the gap indicates that the mere active migration together with the mitotic pressure arising from the crypt-like domains cannot account for the restoration of epithelial homeostasis. This suggests that other non-epithelial factors might be involved in this active epithelial restoration. Given the anatomical proximity of ISEMFs to IECs within the intestinal epithelium and their phenotypical and functional characteristics we decided to evaluate the potential of ISEMFs as epithelial migration regulators.

Phenotypical and functional characterization of primary ISEMFs
We adapted our epithelial crypt isolation protocol to further digest the basal lamina and obtain subepithelial fibroblasts. The obtained cells were then seeded onto tissue plates coated with a thin film of Matrigel ® and their phenotype was analyzed using immunofluorescence. Cells stained positive for α-SMA and vimentin and negative for desmin, confirming their ISEMF identity and differentiating them from other subepithelial mesenchymal cells types (Figure 2A, left panels) (Lahar et al., 2011;Lei et al., 2014;McCarthy, Kraiczy and Shivdasani, 2020). Furthermore, primary isolated ISEMFs produced and expressed fibronectin, laminin, and collagen IV, thus confirming their ECM-protein secretory capacities (Figure 2A, right panels). This protein expression profile was maintained when cells were seeded onto non-coated plastic tissue plates, demonstrating the cell-specific expression of the cited proteins (Figure 2 - Figure Supplement 1 A and B). One of the most relevant characteristics of myofibroblasts is the development of ɑ-SMA stress fibers that connect with the ECM at focal adhesion sites and between cells via adherent junctions (Hinz et al., 2007). Higher magnification images clearly demonstrated the expression of ɑ-SMA in stress fibers ( Figure 2B). These have been correlated with the production of isometric tension (Harris, Stopak and Wild, 1981), and provide myofibroblasts with stronger contractile capacities compared to regular fibroblasts (Vaughan, Howard and Tomasek, 2000;Hinz et al., 2001;Hinz and Gabbiani, 2003). To analyze the ability of our ISEMFs to contract, we seeded them on floating collagen discs and checked for their contraction, comparing ISEMFs to non ɑ-SMA-expressing fibroblasts (NIH-3T3) and epithelial cells (Caco-2), which served as negative controls. Primary ISEMFs reduced the area of the collagen discs by about 80% after one day in culture while NIH-3T3 and Caco-2 only achieved a reduction of 20%. While this difference was slightly reduced after 2 days in culture, it still showed a more than twofold stronger contractile capacity of the ISEMFs compared to regular fibroblasts ( Figure 2C), as previously shown (Hinz et al., 2001). Overall, this characterization confirms the ISMEFs-typical cell morphology, markers expression, protein secretion and contractility capacities of our primary isolated intestinal subepithelial cells.

ISEMFs boost epithelial proliferation through paracrine signals
We then evaluated the impact of ISEMFs on intestinal epithelial growth by performing 3D Matrigel ® drop co-cultures of crypts and ISEMFs and observing the evolution of the crypt's growth and morphology for 4 consecutive days. To disentangle the ISEMF-derived paracrine signals from the crypt-ISEMF physical interactions on intestinal epithelial cell growth, we compared crypts cultured with basic medium, cultured with ISEMF-derived condition medium (ISEMF_CM), and crypts co-cultured with ISEMFs in direct physical contact. For comparison we also added cultures in which the ISEMFs were present but physically separated from the Matrigel ® drop containing the crypts by using a Transwell ® insert ( Figure 3A, schematics). Our results show that when crypts are cultured with medium conditioned by the ISEMFs (either with ISEMF_CM or with the ISEMFs seeded in Transwell ® inserts), they form epithelial cysts. This contrasts with the normal budding and differentiated organoids formed when crypts are cultured with the basic medium ( Figure 3A, first three columns). Those intestinal cysts are hollow epithelial spheres mainly composed of Ki67 + proliferative cells ( Figure 3B, C). They closely resemble crypts with hyperactivation of the Wnt pathway (Sato et al., 2011;Drost et al., 2015;Jardé et al., 2018). Indeed, it has been shown that ISEMFs produce Wnt molecules (Gregorieff et al., 2005;Farin, Van Es and Clevers, 2012;Kabiri et al., 2014). Thus, ISEMFs might boost epithelial proliferation and cyst formation through the paracrine activation of the β-catenin signaling pathway. Interestingly, intestinal crypts grown in direct contact with ISEFMs also form cysts which grow through a series of contraction-relaxation pulsatile cycles (Video 2). Those cysts are bigger and not as spherical as those formed with only ISEMFs paracrine signals. ( Figure 3A, last column). In addition, we observed that ISEMFs physically interact with the intestinal cysts, pulling them closer together until eventually causing their contact and fusion to create bigger structures ( Figure 3D and Video 3). Over time, those fused cysts evolve into flattened structures corresponding to rudimentary epithelial layers (Figure3D, right panel). ɑ-SMA immunostaining revealed the presence of ISEMFs at the protruding edges of those epithelial monolayers ( Figure 3E). ISEMFs appear to be pulling on the epithelial cells causing them to stretch and move ( Figure 3E). Overall, these results suggest that primary ISEMFs boost intestinal epithelial proliferation through paracrine signals and reveal a key contribution of ISEMFs in regulating epithelial movement through epithelial-myofibroblast direct physical interactions.

The physical presence of ISEMFs is crucial for an efficient and directed epithelial migration
To further analyze the paracrine and physical effects of ISEMFs on intestinal epithelial movement,  Figure 4D). ISEMF-derived paracrine signaling induced a progressive increase of the cell net displacements for cells closer to the migration front (gray stripes in the graphs of Figure 4D). However, this increase was enhanced when ISEMFs were physically present, retrieving the typical velocity profiles of in vitro epithelial expansion (Lee et al., 2017). wound (peaked around 0º) was dependent on whether epithelial cells were exposed to paracrine signals of ISEMFs or to their physical presence. While 35% of trajectories on the control sample displayed -20º < a < 20º, this increases up to 47% for samples cultured with ISEMF-derived paracrine signals and up to 70% (twofold with respect to the control) by the physical presence of the ISEMFs underneath the epithelia. As for the net displacement, we analyzed the alignment of individual cell trajectories with the gap direction as a function of the cell distance to the migrating front. To ease the comparison, we defined and alignment index (cos(2·a)) that equals 1 when the net displacement is in the direction of the gap and -1 when it is perpendicular. We found that the physical presence of ISEMFs increases dramatically such alignment within the first 250 µm away from the migrating front ( Figure 4E and ISEMFs that regulate epithelial migration and make it more efficient.

ISEMFs accumulate and align with IECs at the epithelial migration front for an active and collaborative epithelial migration toward the gap.
To better characterize the physical IEC-ISEMF interaction during epithelial migration, we analyzed by immunofluorescence the progression of the co-culture from the moment of the barrier removal until the closure of the gap. We utilized ɑ-SMA antibodies to specifically visualize ISEMFs' morphology and actin stress fibers, and β-catenin antibodies to visualize the structure of the epithelial monolayer. We found that if ISEMFs were seeded on a thin layer of Matrigel ® suggests an active and collaborative process between the two cell types to achieve a more efficient epithelial migration.

ISEMFs and IECs orient with the direction of the gap and migrate following ISEMF-secreted collagen paths
In order to shed some light on how this active and collaborative migration might occur, we first analyzed the onset of gap closure (first 9 hours after removing the stencil) by particle image velocimetry (PIV). This allowed us to obtain velocity fields from the phase contrast movies of the migration fronts, which comprise both the IECs and the ISEMFs covering the gap. Strikingly, two clearly differentiated regions appeared on the maps of the x-component of the mean velocities (vx) ( Figure 6A). The region at the left of the migration front (corresponding to IECs) exhibited on average vx > 0, which corresponds to cells migrating toward the gap. In contrast, the region at the right of the migration front (corresponding to ISEMFs) exhibited on average vx < 0, which correspond to cells migrating from the gap toward the epithelial front. When looking at the temporal evolution of the x-component of the instantaneous velocity ( Figure 6B), we observed the expansion of the vx > 0 region toward the gap, which matches with the advancing of the epithelia.
Furthermore, the averaged profiles of the x-component (vx) and the y-component (vy) of the velocities across the migration fronts and the adjacent gap regions ( Figure 6C) showed that migration occurred mainly in the direction perpendicular to the front for both cell types. In the case of the epithelial region, vx was positive and vy was zero in agreement with the results obtained by manually tracking individual cells ( Figure 4B). In the case of the region associated to the ISEMFs (the gap region), vx was negative and vy was zero, which suggests that the myofibroblasts actively migrated toward the front of the migrating epithelia. Then, we analyzed the average cell orientation of both ISEMFs and IECs relative to the direction of the gap. On Day 4, the ISEMFs tended to be oriented perpendicular to the gap. Just one day later, most ISEMFs exhibited an orientation parallel to the gap ( Figure 6D, lower graph). This became more random by Day 6, especially far from the migration front. In contrast, the IECs maintain their mean orientation slightly parallel with the gap throughout ( Figure 6D, upper graph), while the number of cells oriented perpendicular with the gap is reduced along time. Altogether, these suggest that the ISEMFs may be an active cell component of the culture in the gap closure process. On the one hand, by actively migrating toward the front of the epithelia and, on the other hand, by modifying their orientation to be aligned with the IECs and parallel with the gap to make the epithelial migration more efficient.
We then reasoned that ISEMFs might contribute to the epithelial migration by secreting ECMproteins. We found that the secretion of collagen IV and its accumulation increased with time ( Figure 6E). Interestingly, we observed ordered collagen-path depositions accumulating between the ISEMFs layer and the epithelium ( Figure 6F, 6G), in a similar fashion to the observed ISEMFs preferential orientations. Taken together, these results suggest that when the epithelial migration starts, ISEMFs expressing ɑ-SMA + stress fibers actively align with the IECs and orient with the gap at the epithelial migration front. These collagen-paths deposited by ISEMFs below the IECs might serve as tracks for the epithelial movement. In combination with the increased directionality of epithelial migration close to the front, our data indicate that epithelial movement toward the gap closure is more effective and directionally guided by the physical presence of the underneath ISEMFs.

Discussion
We employed a 3D in vitro intestinal model of gap closure to analyze the effect of the resident population of ISEMFs on epithelial restoration. We could show that ISEMF-derived paracrine signals regulate epithelial proliferation. According to literature, this might happen through the secretion of non-epithelial Wnt pathway activators including Wnt2, Wnt2b, Wnt4, Wnt5a, Wnt9b and Wnt11 (Gregorieff et al., 2005;Farin, Van Es and Clevers, 2012;Kabiri et al., 2014), and the enhancers of the Wnt pathway, R-spondin 2 and R-spondin 3 (Kim et al., 2008;Kabiri et al., 2014;Lei et al., 2014). All these have shown to be factors secreted by fibroblasts that potentiate intestinal stem cell activity by paracrine action, and our results indicate that they also improved epithelial migration. However, we show that only ISEMFs physical presence allows for the complete epithelial restoration in a process that involves ISEMFs active migration, expression of ɑ-SMA + stress fibers and reorientation with the gap as soon as the migration process starts. This suggest that resident ISEMFs might act as a platform (similar to a conveyor belt), which through their contractile properties, and capacity to generate forces they could mechanically stimulate epithelial cells to move in a certain direction. In addition, we could show that ISEMFs deposit collagen paths which could act as "guide rails" for migrating intestinal epithelial cells. This phenomenon seems reminiscent of how cancer associated fibroblasts (CAFs) promote tumor invasion by creating ECM "highways" that tumor cells use to escape from the tumor bulk (Gaggioli et al., 2007;Attieh et al., 2017;Gopal et al., 2017). Whether CAF-derived mechanical signals also contribute to tumor cell mobility is still unclear, although it seems possible considering that CAFs, similarly to ISEMFs, are also characterized by an activated, highly contractile phenotype (Turley, Cremasco and Astarita, 2015).
Intestinal epithelial migration, both under physiological and pathological conditions is thought to happen passively pushed by the stem cell division at the bottom of the crypts or by the proliferation of the newly formed epithelium upon damage (Parker et al., 2017). Our results present a counterpoint to this idea by introducing the concept of ISEMF-derived physical signals facilitating epithelial active migration. Despite not having been proved in vitro before, this is not the first time that passive intestinal epithelial migration, is being challenged. In a recent publication, Krndija et al. elegantly demonstrated that physiological epithelial migration along the intestinal crypt-villus axis cannot be explained by only the mitotic pressure exerted by stem cells at the bottom of the crypts. To do that, they combined in vivo genetic mouse models with biophysical modeling. They proposed that, while mitotic pressure action is constrained mainly to the crypts, epithelial migration in the villus could be an active migration process, determined by an actin-related protein 2/3 complex (Krndija et al., 2019). Given our results, it would be interesting to analyze the contribution of ISEMF-derived physical cues for regulating the active physiological epithelial migration along the intestinal crypt-villus axis. A pioneering work in the field that dates back more than 50 years analyzed the colonic ISEMF dynamics in vivo via 3 H-timidine auto-radiographic examination. Authors proved that ISEMFs actually migrate along the walls of the crypts, and reach their tops in an equivalent time that the employed by the epithelial cells, thus indicating an intimate relationship between both cell types and a possible synchronous migration (Pascal, Kaye and Lane, M.D, 1968), analogous to the conveyor belt principle. This work also revealed localized arrangements of collagen fibrils underneath the epithelium in a fairly organized arrangement, compared to the looser collagen network present within the rest of the lamina propria (Pascal, Kaye and Lane, M.D, 1968). This would favor the notion, presented here, that collagen paths built by the ISEMFs guide for epithelial movements. In this context, a 3D intestinal scaffold harboring primary IECs, a basement membrane and ISEMFs grown in a spatial physiologically relevant manner would represent the perfect in vitro tool to study the effects of cell-to-cell and cell-to-matrix interactions on regulating physiological intestinal epithelial migration.
In addition to be a major organ involved in digestion, the intestinal epithelium is a protective physical barrier that prevents external insults to come into contact with the lamina propria immune system. Specific genetic conditions, external pathogens, and physical or chemical agents might induce a loss of integrity of this epithelial barrier. Once it occurs, a fast and efficient repair of the epithelial damage is crucial for maintaining intestinal homeostasis and preventing uncontrolled inflammatory responses (Terzić et al., 2010;Karin and Clevers, 2016). Thus, understanding the mechanisms that regulate intestinal epithelial repair upon intestinal damage is critical to design new avenues of inflammatory bowel diseases and cancer therapies. Here, we proposed a new functionality for ISEMFs, acting as important players in intestinal epithelial restoration by regulating both epithelial proliferation and migration, which should prompt their recognition as new potential therapeutic targets.

Mouse models
All experimental protocols involving mice were approved by the Animal care and Use Committee of Barcelona Science Park (CEEA-PCB) and the Catalan government and performed in accordance with their relevant guidelines and regulations. Lgr5-EGFP-IRES-creERT2 mice have been previously described (Barker et al., 2007)
The first 4 days of culture the Rho kinase inhibitor Y-27632 (Sigma) was added to the culture.
Outgrowing crypts were passaged once a week and organoid stocks were maintained for up to 4 months.
Treated intestinal organoids were mechanically and enzymatically digested as described above and cell sorted to obtain pure GFP + and tdTomato + cell populations. Sorted cells were cultured in Matrigel ® drops with ENR_CV-medium plus the Rho Kinase inhibitor Y-27632 (Sigma) to obtain a new in vitro line of intestinal organoids named Lgr5-EGFP/RCL-tdT, which expresses a robust tdTomato fluorescence in all cells and GFP fluorescence in Lgr5 + stem cells. Outgrowing crypts were passaged once a week and organoid stocks were maintained for up to 4 months.

To obtain organoid-derived intestinal epithelial cells (IECs), organoids were first obtained from
Lgr5-EGFP-IRES-creERT2/RCL-tdT mice. Then, fully-grown organoids were subjected to a digestion protocol. Briefly, Matrigel ® drops containing organoids were disrupted by pipetting with TrypLE Express1X (Gibco) and transferred to a Falcon tube at 4°C, where mechanical disruption was applied using a syringe with a 23 G 1" needle (BD Microlance 3), was applied. Next, disrupted organoids were further digested by incubating them for 5 to 7 minutes at 37ºC with vigorous handshaking every minute. Successful digestion to single cells was confirmed via inspection under the microscope.

Intestinal subepithelial myofibroblast isolation and culture
Intestinal subepithelial myofibroblasts (ISEMFs) were isolated from mouse intestine using a modified version of a previously reported protocol (Khalil et al., 2013). Briefly, left over tissue from the crypt isolation procedure (see above) was further digested by incubating at 37°C for 60 min at 200 rpm with 2000 U of collagenase (Sigma). The digested tissue was then pelleted and lysed at 47ºC for 5 minutes using ACK lysis buffer (Gibco). The pellet, containing lamina propria cells, was resuspended in Dulbecco's modified medium (DMEM) (Life Technologies) containing 10% foetal bovine serum (FBS) (Gibco), 1% penicillin/streptomycin (Sigma), and 1% minimum essential medium non-essential amino acids (MEM-NEAA) (Gibco) and cultured in tissue culture plates (25 mm 2 ). After 1 week in culture, only lamina propria fibroblasts, mainly myofibroblasts, remained attached. Plates reached confluence after approximately 20 days in culture. Cell division of primary myofibroblasts is limited and after 6 -8 passages the cells become senescent.
Detection by immunofluorescence of α-smooth muscle actin (α -SMA), vimentin, and reduced desmin expression was used to assess the purity of the myofibroblast cultures.

Setup of the 3D intestinal mucosa model to study epithelial migration in vitro (gap closure model)
Ibidi µ-Slides 8 well (Ibidi GmbH) were coated with Matrigel ® to form thin (< 20 µm) films by spreading 10 µL of Matrigel ® at 3 mg ml -1 throughout each well, followed by incubation at 37°C for 1 hour as previously described (Altay et al., 2019(Altay et al., , 2020. 2% Pluronic ® F-68 (Sigma) treated elastomeric barriers (Polydimethylsiloxane (PDMS) stencils) of approximately 2,5 x 10 mm were used to spatially confine cell growth and generate an epithelial migration gap in the middle of the well. Different experimental conditions were analyzed: control (only IECs were present in the culture), paracrine (IECs were culture with ISEMF_CM (see SI Methods)) and physical (IECs and ISEMFs were co-cultured in physical contact). For the control samples, first elastomeric barriers were placed (using PDMS grease, Corning) on the Matrigel ® coated wells, horizontally dividing the well into two parts. Then, 3·x 10 5 IECs cm -2 were seeded and cultured with ENR_CVmedium containing Y-27632 for 2 days. For the paracrine samples, we employed the same seeding protocol than before, but using ISEMF_CM/ ENR_CV-medium containing Y-27632 for 2 days. For the physical samples, first 10 5 ISEMFs cm -2 were seeded on Matrigel ® coated wells and were cultured with complete DMEM medium for 1 day. Then, the elastomeric barriers were placed onto the ISEMF monolayers and 3·x 10 5 IECs cm -2 were seeded on top and cultured with ENR_CV-medium containing Y-27632 for 2 days more. For physical samples (discontinuous ISEMFs layer), first PDMS Stencils were placed on Matrigel ® coated wells, next 10 5 ISEMFs cm -2 were seeded and cultured with complete DMEM medium for 1 day; then 3·x 10 5 IECs cm -2 were seeded on top and cultured with ENR_CV-medium for 2 additional days. After the indicated tissue culture times, the elastomeric barriers were carefully removed using self-closing tweezers to avoid the detachment of the underlying cellular or Matrigel ® layer. Epithelial growth and migration to reepithelize the gap were monitored using time-lapse microscopy, conducted at 10x magnification with an Axio Observer 7 epifluorescence inverted microscope (Zeiss) employing temperature (37°C), relative humidity (95%), and CO2 (5%) regulation. Phase contrast and Alexa 546 channels were used. Images were acquired every 10 min up to 48 h of culture. The time-lapse videos and fluorescence images were processed with Fiji (http://rsb.info.nih.gov/ij, NIH, USA). For an optimal tracking of individual Lgr5-EGFP/RCL-tdT IECs (tdTomato + ) those were mixed and seeded with non-fluorescent IECs in a 1:0.5 ratio.

Collagen contraction gel assay
Collagen type I from rat tail (Corning) was neutralized with NaOH, made isotonic with 10X PBS and diluted to a final concentration of 3 mg ml -1 on ice. For these experiments we employed: primary intestinal subepithelial myofibroblasts (ISEMFs), NIH-3T3 fibroblasts (ATCC® CRL-1658 TM ) and Caco-2 cells (ATCC® HTB-37 TM ). Cells were trypsinized following standard methods, counted, and mixed with the neutralized collagen I solution on ice. Approximately 200 µl of collagen-cell mixture was pipetted into pre-warmed 6-well plates to form one disc per well of approximately 2 cm in diameter. Plates were then kept at 37º C for 30 minutes for collagen I gelation. Next, collagen-containing cell discs were detached from the plate base employing a sterile 1000 µl plastic tip. 2 ml of complete DMEM medium was added to each well. Macroscopical pictures of the collagen discs were taken every 24 h to monitor their contraction. Calculation of the disc areas was done by analyzing the corresponding pictures with Fiji (http://rsb.info.nih.gov/ij, NIH, USA).

Cell migration analysis
The centroid trajectories of Lgr5-EGFP/RCL-tdT cells were tracked using the Manual Tracking Plug-in in Fiji (http://rsb.info.nih.gov/ij, NIH, USA). Data analysis was performed using a custommade code in Matlab (Mathworks, USA). Cell centroid positions during the experiment were defined as ri = r(iDt), being Dt the time between consecutive images and r a vector. The vector difference between the initial (t = t0) and the final point (t = tf) is defined as the displacement vector d and its module || d || as the net displacement. The alignment index of the trajectories is defined as cos(2·a), being a the angle between the displacement vector d and the gap direction, defined as the direction perpendicular to the epithelial edge. This index equals 1 when the trajectory is parallel to the direction of the gap and -1 when it is perpendicular. For the cell migration analysis, ncontrol = 136 cells, nparacrine= 164 cells and nphysical = 204 cells randomly distributed within the epithelia from N = 3 independent experiments were analyzed. PIVlab 2.37 (Thielicke and Stamhuis, 2014) was used to quantify the displacements of IECs and ISEMFs on phase contrast time-lapse movies. Briefly, an interrogation window of 23 µm with a step of 11.5 µm was used to perform Particle Image Velocimetry (PIV) either by FFT window deformation or by ensemble correlation. Image sequencing was set as time-resolved and the resulting velocities were filtered using a standard deviation filter (8*STD) and a local median filter (threshold = 3). For the PIV analysis, n = 4 migration fronts from N = 2 independent experiments were used.

Cell orientation analysis
Orientation fields were obtained using the OrientationJ (Püspöki et al., 2016) plug-in in Fiji. First, images were pre-processed by subtracting the background (40 pixels rolling ball), then a bandpass filter was applied (limits 3 and 40 pixels), and the background noise was subtracted again (40 pixels rolling ball). The resulting images were processed with OrientationJ using a local window of 40 pixels to obtain a structure tensor on a grid of 40 pixels x 40 pixels. With vector fields and their associeated values provided by the plug-in (dominant direction w, energy E and coherence C) we used a custom-made Matlab code to determine the correlation maps. Briefly, we selected the angles with associated energy E > 0.1 and coherency C > 0.05 for each cell type (IEC or ISEMFs) and discarded the rest. These sets of angles were used to calculate the mean cell orientation and the correlation index. To do so, we collected pairs of dominant angles and obtained the difference between them q = wIEC -wISEMF for each x, y position within the image (note that we only accounted for x, y positions when the dominant angles for both cell types satisfied the conditions for E and C). Then, we defined the correlation index between the two cell types as cos(2·q), which equals 1 when both cells are oriented parallel to each other and -1 when they are perpendicularly oriented. For this analysis, we used immunostained samples from N = 2 independent experiments.

Statistics
No statistical methods were used to predetermine sample size. Measurements were performed on experimental replicates (n) obtained in different independent experiments (N). Data presentation (as Mean value ± standard deviation (SD) or as Mean value ± standard error of the mean (SE)) is defined at the corresponding figure caption. D'Agostino normality test, t-test, and Kruskal-Wallis test were performed using GraphPad Prism 9. Specific values are noted at the corresponding figure captions. or ISEMFs 0, 1, and 2 days after embedding them in the collagen discs. For easier identification the collagen discs have been highlighted using a white dotted line. Scale bar: 500 µm. Lower panel: bar graph showing the mean fold decrease in collagen disc area for each cell type after 0, 1, and 2 days (N = 3). The data is presented as mean ± standard deviation (statistics performed through pairwise Student's t-test; significance level set at * p < 0.05 and ** p < 0.01).    Statistically significant differences were assessed by Kruskal-Wallis test with *** p < 0.001 and ns p > 0.05.