Wnt5a Potentiates TGF-β Signaling to Promote Colonic Crypt Regeneration After Tissue Injury

See allHide authors and affiliations

Science  05 Oct 2012:
Vol. 338, Issue 6103, pp. 108-113
DOI: 10.1126/science.1223821

Gut, Heal Thyself

Foods, drugs, and pathogens all represent possible threats to our guts on a daily basis. Fortunately, the gut is quite good at repairing itself—but how? Working in mice, Miyoshi et al. (p. 108, published online 6 September; see the Perspective by Barrett) selectively injured intestinal crypts containing intestinal stem cells and observed therepair process. The noncanonical Wnt ligand, Wnt5a, was required for crypt regeneration. Wnt5a inhibited intestinal stem cell proliferation, which paradoxically promoted regeneration of crypt tissue.


Reestablishing homeostasis after tissue damage depends on the proper organization of stem cells and their progeny, though the repair mechanisms are unclear. The mammalian intestinal epithelium is well suited to approach this problem, as it is composed of well-delineated units called crypts of Lieberkühn. We found that Wnt5a, a noncanonical Wnt ligand, was required for crypt regeneration after injury in mice. Unlike controls, Wnt5a-deficient mice maintained an expanded population of proliferative epithelial cells in the wound. We used an in vitro system to enrich for intestinal epithelial stem cells to discover that Wnt5a inhibited proliferation of these cells. Surprisingly, the effects of Wnt5a were mediated by activation of transforming growth factor–β (TGF-β) signaling. These findings suggest a Wnt5a-dependent mechanism for forming new crypt units to reestablish homeostasis.

Tissue regeneration requires proper spatial allocation and organization of stem cells for efficient return to homeostasis (1, 2). Crypts of Lieberkühn are subunits that house intestinal stem cells and are lost in response to a variety of insults, including ischemia, infection, irradiation, and inflammatory bowel disease (3). Although individual crypts undergo fission to replicate during homeostasis (fig. S1A) (4, 5), the mechanism of their regeneration is unknown. Thus, crypt regeneration is a proxy for proper stem cell organization and provides an excellent system to uncover the principles underlying stem cell replacement and/or organization in vivo.

To model crypt/epithelial stem cell loss, we previously developed an injury system to focally excise crypts from the absorptive inner lining of the mouse colon (6). In response to the excision of ~1-mm2 areas from the inner lining of mouse colons (~250 to 300 crypts), a reproducible program of epithelial alteration occurred. During the first phase (0 to 4 days postinjury), a flattened layer of nonproliferative epithelial cells [wound-associated epithelial (WAE) cells] emanated from crypts adjacent to the wound and migrated over the wound-bed surface. During the second phase (4 to 8 days postinjury), crypts adjacent to the wound formed lateral, open extensions toward the center of the wound bed, forming an array of channel-like structures (fig. S1, B and C). Histologic cross sections of wound channels at day 6 postinjury showed that they resembled crypts (Fig. 1A) but were distinguished by a predominately proliferative, undifferentiated cell population (Fig. 1B and fig. S1, D and E) (7).

Fig. 1

Colonic crypts regenerate from existing crypts during injury repair. (A) A hematoxylin and eosin (H&E)–stained section of a wound at day 6 postinjury. The asterisk indicates the center of the wound bed; the arrowhead indicates a wound channel that consists of immature epithelial cells (boxed region, see inset). Crypts distant from the wound site are represented in the panel labeled “Uninjured area.” Scale bars, 200 μm. (B) Sections stained for Ki-67 (brown) to label proliferative cells at various time points after biopsy injury. An arrowhead labels the wound channel at day 6 postinjury. Scale bars, 100 μm. n = 3 wounds per time point for (A) and (B). (C) Migration of clonal cell populations from labeled crypts after wounding in Vil-CreERT: Rosa26R mice. Cells expressing LacZ were visualized by the staining with 5-bromo-4-chloro-3-indolyl-β-d-galactoside. Dotted lines outline the original wound area. Scale bars, 200 μm. (D) A H&E-stained section of a wound at day 8 postinjury in a Vil-CreERT: Rosa26R mouse. LacZ-positive cells (blue) were present in a wound channel (arrowhead; see inset). The wound bed is delineated by a bracket. Scale bar, 200 μm. (E) A H&E-stained section of a wound at day 28 postinjury in a Vil-CreERT: Rosa26R mouse shows clusters of LacZ-positive crypts in the wound bed (shown by the bracket). Scale bar, 200 μm. n = 6 wounds per time point for (C), (D), and (E).

We hypothesized that new crypts within wounds arose from existing crypts adjacent to the wound bed (6). Therefore, we used Vil-CreERT: Gtrosa26tm1Sor (Rosa26R) mice to perform lineage-tracing experiments (see the supplementary materials and methods) (8, 9). To mark a subset of crypts before injury, we activated Cre recombinase in the intestinal epithelium with a single tamoxifen injection, followed by a 1-week delay. Four days postinjury in these mice, coherent columns of LacZ-positive WAE cells emanated from adjacent crypts toward the wound center (Fig. 1C and fig. S2A) (6). At day 8 postinjury, distinctive LacZ-positive epithelial channels emanated from adjacent crypts toward the center of the wound (Fig. 1, C and D, and fig. S2A). At later time points (14 and 28 days postinjury), clusters of LacZ-positive crypts were located within the original wound-bed site (Fig. 1, C and E). Sections of these LacZ-positive crypts showed a return to homeostasis; differentiated epithelial cells (i.e., goblet cells) were present in the upper crypt, similar to crypts in nonwounded areas (fig. S2B). These experiments showed that new colonic crypts can originate from crypts adjacent to the wound site and suggest that epithelial stem cells migrated through distinctive channels to enter the wound bed (fig. S2C).

At day 6 postinjury, we readily observed multiple invaginations of wound channels, suggesting that multiple fission events occurred to produce new crypts (Fig. 1A). Comparison of transcript abundance (10) of budding wound channels (day 6 postinjury) to crypts in uninjured areas showed that the wound-channel epithelium contained highly proliferative cells (fig. S3A) enriched in Wnt signaling transcripts (canonical and noncanonical combined) (fig. S3, A and B). As canonical Wnt signaling is critical for intestinal stem cell homeostasis (1114) and noncanonical Wnt signaling plays a role in intestinal development (15), we screened expression of 19 Wnt ligands and 4 R-spondin coactivators by reverse transcription polymerase chain reaction (RT-PCR) (fig. S3C). Interestingly, mRNAs encoding the noncanonical Wnt ligand Wnt5a were significantly enriched in the wound bed compared with the adjacent uninjured mucosa (Fig. 2, A and B, and fig. S3, D and E).

Fig. 2

Wnt5a-positive mesenchymal cells stimulate crypt regeneration after injury. (A) RT-PCR analysis of Wnt5a from RNAs isolated from an embryonic day 13.5 embryo and its placenta tissue (controls), the wound bed, and adjacent uninjured mucosa. Gapdh, glyceraldehyde-3-phosphate dehydrogenase. (B) Plots of mean (+SD; error bar) relative Wnt5a mRNA expression levels as determined by quantitative RT-PCR analysis of wound beds and adjacent uninjured mucosa at day 4 postinjury. Data were analyzed using the Student’s t test (n = 4 wounds per group). (C) Mouse colon sections at day 6 postinjury (uninjured area and wound) stained by in situ hybridization for Wnt5a (purple) and hyaluronic acid (basement membrane, brown). Dotted lines outline the apical epithelial surface. Scale bars, 100 μm. (D) Serial sections of uninjured area and a colonic wound at day 6 postinjury stained for Wnt5a and Axin2 mRNA, respectively. Arrowheads indicate Wnt5a-positive cells associated with wound-channel clefts (insets). Methyl green labeled nuclei are shown. Scale bars, 100 μm. (E) Serial sections of a colonic wound channel at day 6 postinjury stained for Wnt5a mRNA (top) and Ki-67 (bottom). Wnt5a-positive cells (arrowheads) were localized near quiescent epithelial cells. Sections were counterstained with nuclear fast red (top) and hematoxylin (bottom). Scale bars, 100 μm. n = 3 wounds per assay. (F) Sections from a CAGGCreERTM:Wnt5a+/+ (Wnt5a+/+) mouse and a CAGGCreERTM:Wnt5aflox/flox (Wnt5ako/ko) mouse at day 6 postinjury stained by H&E. Arrowheads indicate wound-channel invaginations. The arrow indicates an immature wound channel without invaginations. Scale bars, 200 μm. (G) Graph of the average distance between wound-channel invaginations (±SD) (n = 7 wounds per group). Each dot represents the average distance for an individual wound channel. Data were analyzed using Student’s t test. (H) H&E-stained sections from a Ubc-Cre-ERT2:Wnt5aflox/+ (Wnt5ako/+) and Ubc-Cre-ERT2:Wnt5aflox/flox (Wnt5ako/ko) mouse at day 8 postinjury (n = 3 mice analyzed per group). Arrowheads indicate the space between cryptlike structures that developed from wound channels. Arrows indicate abnormal immature wound channels with no cryptlike structures. Scale bars, 500 μm. (I) Schematic diagrams of defects in crypt regeneration in Wnt5ako/ko mice.

Noncanonical Wnts expressed in undifferentiated mesenchymal cells affect tissue regeneration in hydra and zebrafish (16, 17). Similarly, in situ hybridization showed an expanded population of Wnt5a-positive stromal cells in the colonic wound bed (Fig. 2C and fig. S4) compared with uninjured mucosa (Fig. 2C) (18). Importantly, a subpopulation of Wnt5a-positive cells localized near wound channels. Similar to other injury models, the Wnt5a-positive stromal cells in the colonic wound bed did not express markers of differentiation including α-smooth muscle actin (myofibroblasts) or β-catenin (endothelial cells) (fig. S5, A and B). We noted an additional subpopulation of Wnt5a-positive cells beneath the mucosal wound bed in the serosal area (fig. S6A). During development, one potential source of undifferentiated stromal progenitor cells in the mouse intestine is mesothelial cells that form the serosa (the outer surface of the intestinal tube) (19). We performed a genetic lineage tracing experiment for mesothelial cells using WT1CreERT2: Rosa26R mice (20). Tamoxifen injection activates Cre recombinase in mesothelial cells that, in turn, marks these cells by LacZ. A population of Wnt5a-positive cells was derived from these WT1-marked cells at day 6 postinjury (fig. S6, B to E), suggesting that Wnt5a-positive cells are partly derived from mesothelial cells.

We next examined the association of Wnt5a-positive cells with wound-channel epithelial cells. During injury repair, wound channels were composed of an expanded population of proliferative (Ki-67-positive), canonical Wnt-active (Axin2-positive) (21) epithelial cells. Notably, Wnt5a-positive cells localized to clefts at the base of wound channels, suggesting areas of nascent crypt formation (Fig. 2D). In addition, Wnt5a-positive cells were also located adjacent to nonproliferative wound-channel epithelial cells (Fig. 2E). This was specific for wound channels, as in uninjured areas, Wnt5a-positive mesenchymal cells were not associated with crypt bases that contain canonical Wnt-active epithelial cells (Fig. 2D). These results suggested that Wnt5a-positive mesenchymal cells may induce crypt formation by locally inhibiting proliferation of the stem/progenitor cell population within the wound channel.

To test this hypothesis in vivo, we generated a Wnt5a conditional knockout (ko) allele (fig. S7, A to D). We crossed Wnt5aflox mice with CAGGCreERTM (22) or Ubc-Cre-ERT2 (23) transgenic mice for global cellular targeting and generated CreERT-expressing Wnt5aflox/flox mice (Wnt5ako/ko), as well as CreERT-expressing Wnt5a flox/+ and Wnt5a+/+ mice (controls). After serial tamoxifen injections to activate Cre recombinase, we created colonic mucosal injuries and analyzed repair. Wnt5ako/ko mice generated by both methods contained normally appearing WAE cells (fig. S7E), but also contained abnormal wound channels at days 6 and 8 postinjury. Compared with uninjured controls, at day 6 postinjury, Wnt5ako/ko wound channels contained significantly fewer invaginations (Fig. 2, F and G), and at day 8 postinjury, they did not develop into new cryptlike structures (Fig. 2H). These results showed that Wnt5a has a crucial role in the proper formation and eventual subdivision of wound channels into crypts (Fig. 2I).

To test the direct effects of Wnt5a on wound-channel epithelium, we established an in vitro culture system that mimicked wound channels. Conditioned media from an L cell line expressing Wnt3a, R-spondin3, and noggin (L-WRN) (11, 13, 14) maintained highly proliferative, epithelial colonic stem/progenitor cells (Lgr5-positive) (24) as organoid spheres in culture (fig. S8 and tables S2 and S3); these properties were similar to in vivo wound-channel epithelial cells. Interestingly, Wnt5a-soaked beads (fig. S9) induced clefts within colonic organoids more frequently than control beads (Fig. 3, A and B). Epithelial proliferation in areas of contact with Wnt5a-soaked beads was suppressed compared with areas contacting control beads (Fig. 3C). Furthermore, addition of Wnt5a to the organoid culture medium suppressed colonic sphere formation stimulated by Wnt3a and R-spondin1 (Fig. 3D). Wnt5a decreased Ki-67 and Lgr5 expression in a dose-dependent manner (fig. S10A). However, Axin2 expression was not suppressed by Wnt5a (fig. S10A), in contrast to previous findings suggesting that Wnt5a antagonizes canonical Wnt signaling (2528). The connection between noncanonical Wnt signaling and the cell cycle is unclear. Increased expression of cyclin-dependent kinase (Cdk) inhibitors can inhibit epithelial proliferation. We found that Wnt5a increased expression of multiple Cdk inhibitors, most notably p15INK4B (Cdkn2b) in treated colonic epithelial organoids (fig. S10B). These results show that a focal source of noncanonical Wnt initiates a critical intermediary step to reestablish homeostasis of the colonic epithelium.

Fig. 3

Wnt5a inhibits proliferation of colonic epithelial stem cells. (A) Focal Wnt5a-induced clefts in colonic epithelial organoids. A control (left) or a Wnt5a-soaked bead (right) was placed adjacent to different colonic organoids. Scale bars, 200 μm. (B) Plot of the mean cleft incidence of colonic organoids (+SD; error bars) attached to control or Wnt5a-soaked beads (n = 3 experiments). A Student’s t test was used to determine significance. (C) Colonic organoids attached to either control (left) or Wnt5a-soaked beads (right) were stained for Ki-67 (green). Yellow dotted lines outline the bead attachment area. Nuclei were counterstained with bis-benzimide (blue). Representative images from three samples (per group) are shown. Scale bars, 200 μm. (D) Representative images of colonic epithelial organoids cultured for 48 hours in Wnt3a/Rspo1 and indicated amounts of Wnt5a. The rightmost panel is a control without Wnt3a/Rspo1 (n = 3 experiments). Scale bars, 500 μm.

Because p15INK4B expression can be induced by transforming growth factor–β (TGF-β) (29), a classical inhibitor of epithelial proliferation (30), we tested whether Wnt5a activates the TGF-β signaling pathway. In colonic organoids, recombinant TGF-β1 suppressed proliferation and induced Serpine1 (PAI-1) expression, indicating Smad transactivation (31) downstream of the TGF-β type 1 receptor (fig. S11, A and B). Surprisingly, Wnt5a also stimulated Serpine1 expression in a dose-dependent manner (Fig. 4A) and enhanced Smad3 phosphorylation/nuclear localization (Fig. 4B). In injured Wnt5ako/ko animals compared with control mice, the loss of Wnt5a expression was associated with diminished phosphorylation of Smad3 in the stem/progenitor population at the base of wound channels (Fig. 4, C and D, and fig. S11, C and D). The action of Wnt5a required kinase activity of the TGF-β receptor as SB431542, a kinase inhibitor for TGF-β type 1 receptor, suppressed all Wnt5a effects on cell growth, as well as Ki-67 and Serpine1 expression (Fig. 4E and fig S11E). Thus, Wnt5a expression is required to potentiate TGF-β signaling at the base of wound channels and mimics the effects of Wnt5a in vitro.

Fig. 4

Wnt5a activates the TGF-β signaling pathway. (A) Colonic organoids were cultured for 24 hours with recombinant Wnt5a. Plots of mean (+SD; error bars) relative mRNA expression levels of Serpine1 and Mki67 were determined by quantitative RT-PCR analysis (n = 3 samples per group). Data were analyzed using one-way analysis of variance followed by Tukey’s test (P < 0.0001). Asterisks indicate differences, compared with the baseline condition (0 ng/ml), that were significant (P < 0.05) in the posttest. (B) Nuclear localization of p-Smad3 protein. Colonic epithelial cells grown on Matrigel-coated chambers (Matrigel, BD, Franklin Lakes, NJ) were incubated without ligands (control) and with either Wnt5a (400 ng/ml) or TGF-β1 (1 ng/ml) for 2 hours and then fixed and stained for p-Smad3 (representative images from three experiments). Scale bars, 10 μm. (C) Distribution of p-Smad3 in the wound channels (right) and uninjured crypt units (left). Colonic sections from Ubc-Cre-ERT2: Wnt5a+/+ (Wnt5a+/+) and Ubc-Cre-ERT2: Wnt5aflox/flox (Wnt5ako/ko) wounds at day 6 postinjury were stained for p-Smad3 (bottom). Cell nuclei were visualized with bis-benzimide (top). Arrowheads and arrows indicate the base of wound channels. Scale bars, 50 μm. (D) Quantification of p-Smad3 in wound channels. Plots of the mean ratio of signal intensity (+SD) in epithelial cells located in the base of wound channels compared with the base of crypts in unwounded areas (from the same tissue section) were determined as described in the supplemental materials and methods (n = 4 wounds per group). (E) Colonic organoids were cultured for 24 hours without ligands (control) and with Wnt5a (400 ng/ml), SB-431542 (10 μM), or both together. Representative bright-field pictures were shown (n = 3 experiments). DMSO, dimethyl sulfoxide. Scale bars, 500 μm. (F) Colonic organoids were cultured for 24 hours without ligands (control) and with Wnt5a (400 ng/ml) and TGF-β1 (1 ng/ml). Two pairs of populations expressing shRNA for independent target sequences and their controls (SHC002) were examined. Plots of mean (+SD) relative mRNA expression levels were determined by quantitative RT-PCR analysis (n = 3 samples per group). The asterisks in (D) and (F) indicate differences that were significant in the Student’s t test: **P < 0.01, ***P < 0.001, ****P < 0.0001.

Despite sharing several similar components, the canonical and noncanonical Wnt signaling pathways use distinct co-receptors, LRP5/6 and ROR1/2, respectively (32). Intestinal organoids exclusively expressed Ror2 (fig. S12), and Ror2−/− and Wnt5a−/− mice similarly display defective gut elongation (15, 33). Using colonic organoids with short hairpin RNA (shRNA) knockdown of Ror2 (fig. S13, A to C), we found that Wnt5a and TGF-β1 showed diminished Serpine1 induction (Fig. 4F), indicating that the activation of the TGF-β signaling pathway by Wnt5a was mediated through Ror2. Ror2 knockdown did not affect growth inhibition of Wnt5a and TGF-β1 (fig. S14), suggesting that TGF-β inhibited the cell cycle in a transcription-independent manner, as previously described (34), and that the signaling through Ror1/2 enhanced transcriptional activity of the Smad protein complex. Although the core components of the TGF-β signaling pathway consisting of type I and II receptors and Smad proteins are relatively simple, regulatory mechanisms (including protein modification and translocation) are highly complex (35). This experimental system will help to identify more precise molecular mechanisms of cell cycle inhibition by noncanonical Wnt and TGF-β.

Here, we show that wound channels are a critical intermediate structure during colonic epithelial wound repair. These channels undergo Wnt5a-mediated subdivision via a previously unknown mechanism to potentiate TGF-β signaling. The effects of Wnt5a are primarily mediated by focal inhibition of proliferation, though we cannot rule out effects on cell polarity/asymmetric cell division. The process of crypt regeneration appears to reuse elements of a fetal developmental program (15, 36). Even though a colonic mesenchymal niche is not yet defined, our findings suggest that Wnt5a-positive cells are a critical part of this niche during injury repair, as they affect the organization of the regenerating epithelium. We propose that such epithelial-mesenchymal interactions during repair will occur in other tissues and organisms (1), such as hair follicles, that elevate specific Wnt family members during regeneration (37).

Supplementary Materials

Materials and Methods

Figs. S1 to S14

Tables S1 to S3

References (3853)

References and Notes

  1. Acknowledgments: We thank R. Kopan for comments on the manuscript, W. Pu for reagents, and R. Head and C. Storer for the microarray analysis. This work was funded by the NIH (grant DK90251); the Pew Scholars Program in the Biomedical Sciences; grant 5T35DK074375 (Trans- National Institute of Diabetes and Digestive and Kidney Diseases Short-Term Training for Medical Students); the Washington Univ. Digestive Disease Research Core (NIH grant P30-DK52574); and the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research. Microarray data are available on Array Express (accession no. E-MTAB-1175). Materials transfer agreements will be required for the acquisition of the Wnt5af/f mice from the National Cancer Institute and the L-WRN cell line from the Washington Univ. Medical School. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials.
View Abstract

Navigate This Article