Active cell migration is critical for steady-state epithelial turnover in the gut

See allHide authors and affiliations

Science  16 Aug 2019:
Vol. 365, Issue 6454, pp. 705-710
DOI: 10.1126/science.aau3429

Active migration renews gut epithelia

Epithelial tissues are continuously renewed throughout adult life, and the gut epithelium is the fastest self-renewing tissue in mammals. Over 3 days or so, epithelial cells migrate from the crypts, where they are born, to the tips of the villi, where they die. It is commonly believed that migration is strictly passive, driven by mitotic pressure in crypts—as cells divide, they push their neighbors upward. Krndija et al. now challenge this concept and show that cells migrate actively, using actin-rich basal protrusions oriented in the direction of migration (see the Perspective by Jansen).

Science, this issue p. 705; see also p. 642


Steady-state turnover is a hallmark of epithelial tissues throughout adult life. Intestinal epithelial turnover is marked by continuous cell migration, which is assumed to be driven by mitotic pressure from the crypts. However, the balance of forces in renewal remains ill-defined. Combining biophysical modeling and quantitative three-dimensional tissue imaging with genetic and physical manipulations, we revealed the existence of an actin-related protein 2/3 complex–dependent active migratory force, which explains quantitatively the profiles of cell speed, density, and tissue tension along the villi. Cells migrate collectively with minimal rearrangements while displaying dual—apicobasal and front-back—polarity characterized by actin-rich basal protrusions oriented in the direction of migration. We propose that active migration is a critical component of gut epithelial turnover.

The gut epithelium is the largest mucosal surface of the body (1) and one of the most rapidly renewing tissues (3 to 5 days) in adult mammals (2). A single layer of columnar cells lines the villi (finger-like evaginations that project into the gut lumen) and the crypts (small invaginations into the connective tissue). Stem cells in the crypts give rise to transit-amplifying cells, which divide four to five times before terminal differentiation (2). After exiting the crypt, differentiated epithelial cells migrate for ~3 days, until they reach the top of the villus, where they get extruded (2). Continuous migration of cells between the spatially distant functional compartments of cell division and loss is a major component in gut epithelial renewal. The predominant theory is that cells migrate passively, driven by a pushing force resulting from cell division (i.e., mitotic pressure) (3, 4). However, irradiation and mitotic inhibitor treatment do not block gut epithelial migration, despite causing substantial cell loss in the crypts (5, 6), raising the possibility of active migration. In this study, we investigated the mechanism(s) of epithelial cell migration during homeostatic renewal in the small intestine.

To quantitatively test the contribution of mitotic pressure in epithelial cell migration along the villi, we performed 5-ethynyl-2′-deoxyuridine (EdU) pulse-chase assays in mice injected with hydroxyurea (HU), a specific S-phase inhibitor (7, 8). A low dose of HU (HUlo) efficiently inhibited mitosis, without affecting cell migration along the villi (Fig. 1, A to C, and figs. S1, A to D, and S2, A to C). This suggests that cell division per se is not the principal driving force for cell migration. High concentrations of HU (HUhi) were shown to inhibit cell migration but are proapoptotic (5). As expected, HUhi reduced crypt cellularity (fig. S1, A to D), but cell migration was decreased only in the lower villus region (Fig. 1, A to C, and fig. S2, A, B, and D). Together, these data suggest that the acting range of mitotic pressure is limited to crypts and the lower villus region.

Fig. 1 Epithelium continues migrating during mitotic block.

(A) Top: EdU and HU injection scheme with EdU chase time points. Bottom: Maximum Z projections (30 to 50 μm range). Dashed line in bottom panels indicates the crypt-villus interface. Scale bars, 50 μm. (B) Box-and-whisker plot corresponding to (A). ****P < 0.0001; ns, not significant; au, arbitrary units. (C) EdU fronts plotted as a function of time. (D) Theoretical model for epithelial renewal during gut homeostasis. (E) Predicted cell density profiles along the crypt-villus axis depending on the level of active migratory force. (F) Control villus—longitudinal section. Boxed regions are shown in high magnification. Scale bar, 40 μm. (G) Cell density profile of the villus shown in (F). E-cad, E-cadherin. (H) Model fit to average density profile. (I) HUhi treatment—longitudinal section. Boxed regions are shown in high magnification. Scale bar, 40 μm. (J) Cell density profile of the villus shown in (I). (K) Average density profiles for control (blue) and HUhi (red), with their theoretical predictions (lines).

We next investigated the alternative possibility that cells could migrate actively. Because of the potentially complex spatial interplay between passive and active migratory forces, we developed a biophysical model for gut epithelium renewal. This minimal continuum theory for migrating and proliferating cells (9, 10) (Fig. 1D, fig. S3, and theory note in the supplementary materials) predicted that proliferation forces alone would result in a gradual decrease in cell density toward the villus, whereas active migration would cause a density increase because of cell overcrowding near the villus top (Fig. 1E). We measured one-dimensional (1D) density profiles along the villus axis [as well as 2D density (fig. S4)] and observed that cell density first gradually decreased toward the middle of the villus and then increased in the top region of the villus (Fig. 1, F to H), in good qualitative agreement with the active migration model (Fig. 1E). We thus hypothesized that cell migration along the villus operates according to a two-tier migration system: proliferation forces dominate close to the crypt, whereas active migration forces become predominant in the upper part of the villus. By fitting the active migration force in our model, we obtained a good quantitative prediction for the nonmonotonous cell density profile along the villus (Fig. 1H and fig. S3G). Moreover, simulating a short-term mitotic pressure inhibition predicted a density decrease only at the villus bottom, which was quantitatively confirmed, with HUhi treatment decreasing cell density only at the bottom 10% of the villus (Fig. 1, I to K, and fig. S1, F and G). These data strongly suggested the existence of an active migratory mechanism, in addition to mitotic pressure, which contributes to cell migration along the villi.

The model also predicted an increase in cell speed along the villus axis, dependent on the strength of the active migration force (Fig. 2A). Using EdU pulse-chase assays, we found greater cell displacements in the middle villus compared with the villus bottom (Fig. 2B). To measure cell speed directly, we performed live imaging of gut explants derived from Villin:CreERT2/mTmG reporter mouse (1113), where membrane-targeted green fluorescent protein (GFP) is expressed mosaically in the epithelium (Fig. 2, C to E, and fig. S5A). Cells accelerated as they progressed along the villus axis (Fig. 2D and movie S1). Ex vivo speed profiles, together with in vivo EdU pulse-chase data, correlated well with model predictions for an active migratory mechanism (Fig. 2E).

Fig. 2 Epithelial cells migrate collectively with increasing velocity and create a gradient of tissue tension along the villi.

(A) Predicted cell velocity profiles along the crypt-villus axis depending on the level of active migratory force. (B) EdU pulse-chase assay (top); box-and-whisker plot (bottom left) and EdU fronts’ position (scheme; bottom middle); EdU fronts plotted as a function of time (bottom right). (C) Montage from a time-lapse movie of gut explants (Villin:CreERT2/mTmG; maximum projection of 50 μm); dashed line, villus. (Inset) A magnified region. Scale bar, 50 μm. (D) Top: Cell segmentation (cyan) and cell tracks (white) from the experiment shown in (C). Dashed line, villus. Scale bar, 50 μm. Bottom: Mean cell speeds plotted against relative cell position. (E) Average cell velocity plotted against villus axis, overlaid with model predictions. (F) Top: Skeletonized and tracked images from a time-lapse movie. Newly formed junctions (magenta), junctions formed after rearrangements (black), delamination (green), or division (none). Inset shows a magnified region. Scale bars, 20 μm. Bottom: Analysis of tissue deformation. Circles represent size change; bars represent pure shear (contraction-elongation). (G) Montage showing images before and after laser ablation in lower villus (left-hand panel) and upper villus (right-hand panel). Scale bars, 20 μm. Magnified regions: Postablation images (post-cut, red) overlaid with preablation images (pre-cut, green). Scale bars, 5 μm. (H) Left: Radar chart showing mean relaxation speed per four directions in lower (blue) and upper (purple) villus regions [as schematized in (G)]. Middle: Vertex separation ε(t) plot in function of time, for lower (purple) and upper (blue) villus region. Right: Initial recoil speed for lower and upper villus region. *P < 0.05.

We then characterized collective cell dynamics, with the tracked patch of cells retaining cohesiveness during migration and extending ~10% along the villus axis without simultaneously shrinking laterally (Fig. 2F, fig. S5B, and movie S4). This could be almost entirely accounted for by cell extension along the villus axis; contributions due to junction remodeling from cell rearrangements and delamination were limited, and, as expected, no divisions were observed (Fig. 2F). The observed cohesiveness, consistent with the clonal ribbons observed previously (14), could be well explained by a minimal 2D model of cells performing biased random walks toward the villus top, with cell-cell interactions minimizing cell dispersion and allowing collective migration (fig. S5, C to F).

Forces driving epithelial migration have been best characterized quantitatively in culture (1519), where active migration results in a buildup of tension in the colony center (16). Similarly, an active migration model predicts a buildup of mechanical tension in the villus bottom, driven by cell traction forces toward the top. Conversely, if proliferation forces acted alone, pressure should be highest (and tension lowest) close to the crypt, while dissipating along the crypt-villus axis as a result of friction with the basement membrane. Spatial differences in tissue tension can thus be used to infer how and where forces act during cell migration (16, 2023). We imaged mice with GFP-tagged endogenous nonmuscle myosin IIA heavy chain (24) to highlight the apical actomyosin belt associated with E-cadherin–containing adherens junctions (25) (fig. S5G). To map tensile forces, we made circular laser cuts of epithelial adherens junctions at the lower and upper villus regions (fig. S5H). The epithelium recoiled isotropically after the laser cut, indicating that it was always under tension (Fig. 2, G and H, and fig. S5H). The recoil speed was significantly higher in the lower villus region (Fig. 2H, fig. S5I, and movies S2 and S3). This further supported a model of active migration forces causing a crowding or increased pressure effect at the top and leaving the villus bottom under higher tension.

Actively migrating cells generate protrusive forces through lamellipodia or filopodia, which are driven by actin polymerization (26). Gut epithelial cells are columnar, apicobasally polarized, and not known to have lamellipodia. To assess if gut epithelial cells have cryptic protrusions, we generated a mouse that mosaically expresses LifeAct-mCherry in the gut epithelium, to follow F-actin in individual cells (Fig. 3A). We found that enterocytes have small, F-actin–rich basal feet in contact with the basement membrane, pointing in the direction of cell movement (Fig. 3, B to D). Membrane-targeted GFP also highlighted basal protrusions in Villin:CreERT2/mTmG reporter mice (fig. S6A). Radial distance map analysis (fig. S6, B and C) confirmed that these protrusions were front-back polarized, analogous to the front-back polarity described in migrating cells (Fig. 3, D to F, and fig. S6C). As actin-based protrusions typically require the actin-related protein 2/3 (Arp2/3) complex (27, 28), we tested whether a specific Arp2/3 inhibitor, CK-666 (29), disrupted basal protrusions and migration in vivo. Integrity of cell-cell adhesions, crypt cellularity, and mitosis were not affected during this short-term CK-666 treatment, excluding potential confounding effects on cell migration (fig. S7, B to D). In CK-666–treated mice, apicobasal polarity and the total protrusion area were not affected, but basal front-back polarity was lost (Fig. 3, D to G, and fig. S6D). Accordingly, CK-666 treatment inhibited cell migration along the villi (Fig. 4, A and B). Next, we generated an inducible, gut epithelium–specific Arp2/3 knockout (KO) mouse by crossing the Arpc4fl/fl mouse (30) with the Villin:CreERT2 line. Arpc4 depletion (fig. S8A) also decreased cell displacement along the villi (Fig. 4, C and D), without affecting crypt cell numbers, cell-cell junction integrity, or the apicobasal polarity (fig. S8, B to E). Inhibition of migration after CK-666 treatment was more pronounced in the upper than in the lower villus region (Fig. 4B). This agreed with the theoretical model, which predicted that inhibition of active migration would have a more pronounced effect in the upper villus region (fig. S7A).

Fig. 3 Epithelial cells display actin-rich basal protrusions oriented in the direction of migration.

(A) Mosaic expression of LifeAct-mCherry (red) in the intestinal epithelium. Maximum intensity projection (Z range, 30 μm). Scale bars, 50 μm. (B) Super-resolution 3D image (Z range, 2.4 μm) of a LifeAct-mCherry-expressing enterocyte. Scale bar, 4 μm. (C) Magnified 2D image of the boxed region shown in (B). Scale bar, 2 μm. (D) Maximum Z projections of basal part of enterocytes (Z range, 2.6 μm) expressing LifeAct-mCherry, control [dimethyl sulfoxide (DMSO)] or CK-666–treated. Scale bars, 2 μm. (E) Rose plots with average normalized radial distances for protrusions. n, number of cells analyzed. (F) Bar chart showing protrusion length with respect to front-back orientation. ****P < 0.0001. (G) Super-resolution images; apical marker ZO-1 (left) and basolateral markers α6 integrin (middle) and Na/K-ATPase (right). Boxed regions are shown in high magnification. Scale bars, 5 μm.

Fig. 4 Epithelial cell migration is driven by Arp2/3.

(A) EdU pulse-chase assays; maximum Z projections (30 to 50 μm). Dashed line indicates the crypt-villus interface. Scale bars, 100 μm. (B) Top: Box-and-whisker plot for EdU fronts. **P < 0.01, ****P < 0.0001. Bottom: EdU fronts plotted against time. Control, DMSO. (C) EdU pulse-chase assays; maximum Z projections (30 to 50 μm). Tam, tamoxifen. Scale bar, 50 μm. (D) Box-and-whisker plot for EdU fronts. ****P < 0.0001. (E) CK-666 treatment—longitudinal section. Boxed regions are shown in high magnification. Scale bar, 50 μm. (F) Cell density profile of the villus shown in (E). (G) Average density profiles. Control (blue, same as Fig. 1H); CK-666 (green); theoretical prediction (lines). (H) Left: Radar chart displaying relaxation speed in HUhi- or CK-666–treated mice, along four indicated directions. Middle: Vertex separation ε(t), in function of time. Right: Initial recoil speed, box-and-whisker plot. Control: DMSO or saline, for CK-666 and HU, respectively. *P < 0.05. (I) Mechanical model for epithelial migration in the small intestine.

Furthermore, the CK-666–treated tissue density profile matched our model prediction, abrogating the density increase at the villus top seen in controls (Fig. 4, E to G, and fig. S7, E and F). Arpc4-KO produced a similar cell density profile as CK-666 (fig. S8F), whereas the villus length was unaffected in both cases (figs. S7G and S8G). To validate the predicted roles of mitotic pressure and active migration in generating pressure and tension forces, respectively, we performed laser cut experiments. HUhi treatment led to an increase in tissue tension (i.e., decrease in pressure) on villi, whereas Arp2/3 inhibition led to a decrease in tissue tension (Fig. 4H). This supports our two-tier model of mitotic pressure and Arp2/3-dependent tensile migratory forces (Fig. 4I).

Maintenance of a functional intestinal barrier (31) requires tightly regulated epithelial renewal. Dysregulation of either proliferation, migration, or extrusion could lead to pathologies, such as inflammatory diseases and tumor formation (32). Here, we showed that Arp2/3-driven active migration occurs along the villus. We showed that intestinal epithelial cells exhibit dual polarity—apicobasal and front-back—characterized by actin-rich basal protrusions oriented in the direction of migration. We propose that active migratory forces are a key component in the homeostatic renewal of the adult gut epithelium.

Supplementary Materials

Materials and Methods

Theory Note

Figs. S1 to S8

Table S1

References (3351)

Movies S1 to S4

References and Notes

Acknowledgments: We thank J.-F. Joanny, D. Louvard, G. Montagnac, J. Rosenblatt, R. Galupa, and all members of the DMV lab for helpful discussions and critical reading of the manuscript. We thank A. M. Lennon-Dumenil for providing the Arpc4fl/fl mouse strain and R. Adelstein for the GFP-NMHCII-A strain. We thank O. Renaud for help with imaging and image processing. We acknowledge the Cell and Tissue Imaging facility (PICT-IBiSA) and Animal Facility, Institut Curie. Funding: This work was supported by ERC-2012-StG_20111109 grant STARLIN (D.M.V.) and ANR-16-CE13-0016-01 grant HOMEOGUT (D.M.V.). Author contributions: D.K. and D.M.V. conceived of the study and designed all experiments. Experiments were performed by D.K., S.R., and F.E.M. Data analysis was done by D.K., B.G., and E.H. Mouse colony management and transgenesis was done by F.E.M. Modeling was done by E.H. O.L. wrote a macro for protrusion orientation analysis. D.K., Y.B., E.H., and D.M.V. wrote the manuscript. Competing interests: The authors declare no competing interests. Data and materials availability: All data in the paper are presented in the main text or in the supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article