Hedgehog-responsive mesenchymal clusters direct patterning and emergence of intestinal villi

Hh-Responsive Mesenchymal Clusters Form Before Villus Emergence and Adhere Tightly to the Epithelium.

Because transgenic mice with reduced Hh signaling exhibit severely curtailed villus formation (11, 12), we sought to determine which cells within the mesenchyme are responsive to Hh signaling before and during villus formation. Using LacZ reporter alleles for Patched1 (15) and Gli1 (16), we followed Hh-responsive cells during villus development (E14.0 to E15.5). At E14.0, Hh-responsive cells are scattered throughout the mesenchyme, with a greater concentration closer to the epithelial Hh signal (Fig. 1 A and E) (17). Half a day later, still before villus emergence, distinct mesenchymal clusters appear, all of which are Ptc1⁺ and Gli1⁺ (Fig. 1 B and F). Shortly thereafter, villi begin to emerge directly above these mesenchymal clusters (Fig. 1 C and G). In no case was a villus seen to emerge without an associated Ptc1⁺ (or Gli1⁺) mesenchymal cluster. Clusters are tightly associated with the epithelial basement membrane at the tips of the growing villi (Fig. 1 D and H). Transmission electron microscopy (TEM) demonstrates the frequent and intimate association between cells of the mesenchymal clusters and overlying epithelium (Fig. 1 I and J).

Pdgfrα⁺ Mesenchymal Clusters Are Targets of Hh Signaling.

Because it was previously shown that clusters of mesenchymal cells express Pdgfrα, and that loss of PdgfA/Pdgfrα signaling reduces the number of clusters and villi (9), we examined the extent of overlap between Pdgfrα⁺ and Ptc1⁺ clusters. Mice engineered to express the H2B-EGFP fusion protein from the endogenous Pdgfrα locus (18) were mated with the Ptc1^LacZ/+ and Gli1^LacZ/+ Hh reporter lines and vibratome sections of intestines of mice carrying both marker alleles were analyzed by confocal microscopy (Fig. 1 K and L). Within the villus clusters, the majority of mesenchymal cells were positive for both Pdgfrα and the Hh reporters; no clusters were seen that were positive for only one of these signaling pathways. However, some mesenchymal cells outside of the clusters were individually responsive to Pdgf or Hh signaling, suggesting that the two pathways both play additional, unique roles during intestinal development.

Mesenchymal Clusters and Villi Form in Wave-Like Fashion, from Anterior to Posterior, Initiating on the Dorsal Side.

Previous studies have indicated that villi develop in a proximal-to-distal wave along the length of the intestine (19, 20). We therefore examined whether an anterior-to-posterior wave of cluster formation precedes the formation of villi. At E14.5, E15.5, and E16.5, intestines from Pdgfα^ΕGFP/+ mice were isolated and the presence or absence of clusters was assessed along the length of the intestine (Fig. 2A). At E14.5, clusters are present within the first 46% of the intestine (10.45/22.43 cm). By E15.5, clusters have formed in the proximal 71% of the intestine (18.71/26.2 cm), and by E16.5, clusters are present through the entire small intestine (47.45 cm). Representative images show the progression of this wave front of cluster formation from jejunum at E14.5 (Fig. S1 A and B), to proximal ileum at E15.5 (Fig. 2 B and D), to the ileocecal junction at E16.5 (Fig. 2C and Fig. S1C). Analysis of the leading front of cluster formation reveals that clusters first form on the dorsal side of the gut tube (marked by the connection to omentum) and spread distally and circumferentially (Fig. 2E). Examination of cluster sizes in all areas of the intestine at various stages demonstrates that clusters grow in size as development continues (Table S1). However, clusters associated with nascent villi at any stage in any region of the intestine consistently average 30 μm in diameter [30.15 ± 0.94 (SD) μm in E14.5 duodenum and 30.10 ± 0.80 μm in E16.5 ileum]. Cluster size distribution along the length of the intestine at E15.5 was assessed in opened, flattened intestines (Fig. S2A) using Imaris (Bitplane, v7.2). The average area of the 500 largest clusters was greatest in the duodenum and smallest in the ileum, consistent with the existence of a proximal to distal wave of cluster generation (Fig. 2F).

A proximal-to-distal wave of villus emergence follows cluster formation. To monitor the wave of villus growth with time, we measured the average length of the longest villi (top 5%) in 8–12 sections of duodenum, jejunum, and ileum at E15.5, E16.5, E17.5, and E18.5. Villi first emerge in the duodenum, then jejunum, and then ileum. Villi in all regions of the intestine continue to lengthen over developmental time (Fig. 2G). At all stages, villi are longest in the duodenum and shortest in the ileum and the rate of villus lengthening once villi emerge is similar in all regions up to late gestation. This finding is consistent with a wave of villus growth during fetal life. Although postnatal villus growth was not measured here, differential growth rates after birth could contribute to the fact that villi are shortest in the ileum in the adult (21–23).

Clusters Form in a Patterned Array.

Although mesenchymal clusters appear to be regularly distributed, we sought to statistically verify the existence of a patterned arrangement. Quadrat counting (used in point pattern analyses) was used to determine whether the field of clusters is arranged in a regular or random pattern (24, 25). This determination is accomplished by counting the number of clusters in regularly sized quadrats and then calculating the variance-to-mean ratio (VMR) of those counts. If clusters are randomly distributed, the counts and their variances are expected to follow a Poisson distribution, and the VMR will approach 1.0 (24). If the clusters are regularly patterned, the variance among quadrat counts will be small and the VMR will be <1.0.

Quadrat counting of clusters in the duodenum, jejunum, and ileum (three independent fields of each area) yielded VMRs of 0.53, 0.52, and 0.72, respectively, indicating that the clusters are regularly patterned. In a separate set of analyses, images of multiple fields of duodenum were stitched together using Leica LASAF 2.2.0 software to create large fields of clusters, in which over 900 clusters could be counted per field. In this instance, VMR was 0.41, P ≤ 0.001 compared with 1,000 Monte Carlo simulations. We conclude that, similar to hair, feather, and tooth patterning (13, 14, 26, 27), clusters of Hh-responsive cells are distributed in a patterned array that in turn dictates the later emergence of regularly patterned villi.

Multiple Rounds of Cluster Formation Give Rise to Approximately Four Rounds of Villus Emergence.

Sections of intestinal tissue at E15.5 and E16.5, regardless of their orientation, often reveal alternating short and long villi, suggestive of multiple rounds of villus formation (Fig. 3 A–C). We examined this result more carefully, from the standpoint of both cluster pattern and villus emergence, to document whether multiple rounds of villus emergence occur and to establish the approximate number of such rounds.

If new clusters form dynamically in the spaces between existing clusters, the nascent clusters are expected to initially be smaller. Thus, in the early patterning field, the nearest neighbor of a small cluster is expected to be a large cluster and vice versa. To quantitatively assess this, we developed a point pattern-analysis function: the neighbor-size correlation function (NSCF). NSCF gives the Pearson correlation of the sizes of neighboring clusters as a function of neighbor distance. The linear dependence between cluster size and neighboring cluster size gives a correlation value between +1 (all clusters are similar in size to their closest neighbors) and −1 (clusters and their nearest neighbors are dissimilar in size). Thus, if multiple rounds of cluster formation occur, we expect neighboring clusters to exhibit a negative NSCF. Indeed, the NSCF is negative up to a distance of nearly 80 μm, indicating that within this distance, large clusters are next to small clusters and vice versa (Fig. S2C and Table S2). Interestingly, this negative zone matches well with our direct measurements of average cluster-to-cluster distance of 70.9 ± 0.63 μm (SEM, n = 670). Fig. S2B shows the correlative data for the search interval that encompasses 40–60 μm. Taken together, these results provide a quantitative assessment of cluster patterning and support the conclusion that multiple rounds of cluster formation occur.

To estimate the number of rounds of villus formation, we first examined the average number of villi per intestinal cross section in 8–12 intestines at E15.5, E16.5, E17.5, and E18.5. At each time point, the number of villi is greatest in the duodenum, less in the jejunum, and fewest in the ileum (Fig. 3D). Moreover, for each area of the intestine, the number of villi per cross-section increases with time. These data confirm both the anterior-to-posterior wave of villus emergence and the presence of multiple rounds of villus formation. We next examined the length correlation between all adjacent villi in the duodenum. That value was negative (−0.12 ± 0.045 SEM, P = 0.009), as expected if neighboring villi are disparate in size because new villi arise in the space between existing ones. Because, at the earliest stages of villus development (E15.0), we frequently observe four areas of villus emergence per section (Fig. S3 A and A′), the emergence of a second, third, and fourth round of villi in the spaces between existing villi should produce the idealized pattern shown in Fig. S3 B′–D′, in which 8, 16, and 32 villi are seen per section. Experimentally, we count a mean ± SEM of 12.6 ± 0.13, 16.6 ± 0.12, and 30.5 ± 0.22 villi at E15.5, E16.5, and E18.5, respectively (Fig. S3 B–D). These findings are consistent with the existence of four sequential rounds of villus emergence before birth.

Hh Signaling Controls Cluster Formation and Size.

To probe the role of Hh signaling in cluster formation and villus emergence, we developed an organ-culture system based on a modification of the Jaskoll et al. protocol (28). Intestines were dissected from embryos between E13.5 and E14.5 (before the onset of villus formation), placed on transwell membranes, and cultured over media without serum supplement (Fig. S4A). Cultured intestines reproducibly developed clusters within 24 h and villi emerged within 48 h (Fig. S4 B, C, and G). Although intestinal viability (measured by cell refractility and peristaltic action) is maintained in this culture system for up to 10 d, the rate of intestinal villus formation slows with time. Thus, the studies here were limited to the first 48 h of culture.

To test whether Hh signals are required for villus formation in this explant system, E13.5 Pdgfrα^EGFP/+ intestines were treated with cyclopamine, a nontoxic analog, tomatidine, or vehicle and cultured for 48 h. Vehicle- or tomatidine-treated intestines formed clusters and developed villi (Fig. 4 A, B, E, and H). However, cyclopamine-treated intestines formed neither clusters nor villi. PDGFRα⁺ cells are still present in the mesenchyme (Fig. S5), although clusters cannot be detected by either PDGFRα (Fig. 4 C and I) or by DAPI staining of nuclei (Fig. 4F). Interestingly, when the same experiments were performed with E14.5 intestines, in which cluster formation had initiated in the proximal intestine before addition of cyclopamine, clusters associated with nascent villi were seen in the jejunum (Fig. 4 M and P, and Fig. S6 E and N). No clusters or villi were seen in the ileum where cluster initiation had not yet begun when treatment was started (Fig. 4R, and Fig. S6 F and O). The clusters seen in the jejunum were more widely spaced than those in control cultures (compare Fig. 4 M and P with K, L, and O, and Fig. S6 E with D). These results suggest that Hh signaling is required for the initiation of cluster formation, but not for the emergence of villi if clusters are already formed. The initial round of villus formation requires a Hh signal that is received between E13.5 and E14.5. Continued Hh signaling after E14.5 is required for initiation of subsequent rounds of cluster formation. Inhibition of Hh signaling with the pan-Hh blocking antibody, 5E1, produced similar results (Fig. S6 S and T).

In parallel experiments, the levels of Hh signaling were increased using the Smoothened agonist, SAG. This treatment resulted in larger, wider clusters that were associated with wider, bulbous villi (Fig. 4 D, G, and J). The average cluster area was significantly larger (36%, P < 0.0001) in SAG-treated intestines than in control intestines (Fig. 4S). These findings, together with the loss of clusters after inhibition of Hh signaling, indicate that Hh signaling not only initiates mesenchymal cluster formation, but also dictates the size of clusters, thus determining the shape of emerging villi.

Hh Promotes the Aggregation of Mesenchymal Clusters.

The above data are consistent with two alternative mechanisms for Hh action in cluster formation: Hh signals may promote the clustering of Ptc1⁺ cells, or Hh may control the localized proliferation of these cells. To investigate whether clusters form by localized proliferation or by aggregation, we collected 2-h BrdU-labeled intestines at E14.5, E15, E15.5, and E16 and compared the percentage of BrdU⁺ nuclei within clusters to the percentage of BrdU⁺ nuclei in scattered mesenchyme cells outside of clusters (Fig. 5 A–G). We expected that if clusters were generated by localized proliferation, the percentage of BrdU⁺ nuclei should be greater in forming clusters than in surrounding mesenchyme. In contrast, if clusters form by aggregation, the BrdU index should be similar in both compartments. At E14.5 when clusters are first beginning to form, this is indeed the case: the percentage of BrdU-labeled cells within clusters is similar to that in unclustered mesenchyme (33% and 36.6%). By E15.0, the BrdU-labeling index seen in clusters is reduced to half of the index seen in unclustered mesenchyme (17.5% vs. 38.5%, P = 0.0003). At E15.5, very few clusters could be found with BrdU-labeled cells: only 8 of 116 clusters (6.9%) had labeled cells. Within those labeled clusters, only 10.76% of clustered cells were BrdU⁺. The drop in the number of labeled clusters at later times is likely because of the fact that once clusters mature, cells become postmitotic [Fig. 5F and previously shown (9)]. Indeed, we found that many cells in the clusters possess primary cilia (Fig. 5 H and I), structures that require quiescence to form (29; and reviewed in refs. 30 and 31). Although we cannot rule out the possibility that limited localized proliferation may augment cluster formation, these data are most consistent with the idea that clusters form by aggregation, rather than by localized proliferation. Furthermore, given the loss of clusters after Hh signal reduction and the augmentation in their size with increased Hh signaling, we propose that the function of Hh in cluster formation and subsequent villus emergence is to promote the aggregation of Ptc1⁺ mesenchymal cells.

Mice.

All protocols for mouse experiments were approved by the University of Michigan Unit for Laboratory Animal Medicine. Gli1^LacZ/+, Ptc1^LacZ/+, and PDGFRα^EGFP/+ mice have been described previously (15, 16, 18) and are available through the Jackson Laboratory (strains 008211, 003081, and 007669, respectively).

Organ Culture.

Organ cultures were performed according to our modification of the method of Jaskoll et al. (28). Briefly, E13.5 or E14.5 intestines were dissected from the embryos in cold DPBS (Invitrogen 14040-133), connective tissue was separated, and intestines were placed on transwells (Costar 3428) in BGJb media (Invitrogen 12591-038) supplemented with 1% pen/strep (vol/vol) (Invitrogen 15140-122) and 0.1 mg/mL ascorbic acid. Intestines were cultured for 24–48 h (as indicated in the figure legends) at 37 °C with 5% CO₂ with media changes every 12 h.

Drugs.

Drugs were purchased as follows: Tomatidine (Calbiochem 614350), Cyclopamine (LC Laboratories C-8700), and SAG (Enzo Life Sciences ALX-270–426-M001).

TEM.

E15.5 intestines were fixed in 2.5% (wt/vol) glutaraldehyde in 0.1 M Sorensen’s buffer, pH 7.4, overnight at 4 °C. After postfixation in 1% (wt/vol) osmium tetroxide, tissues were rinsed and stained en bloc with aqueous 3% (wt/vol) uranyl acetate for 1 h, dehydrated in ethanol, rinsed in propolynoxide, and embedded in epoxy resin. Sections (70 nm) were stained with uranyl acetate and lead citrate and examined using a Philips CM100 electron microscope.

X-Gal Staining.

X-Gal staining of sectioned tissues was performed as described previously (17). Whole-mount tissues were fixed in 4% (wt/vol) PFA for 10 min, washed in PBS and X-Gal stained as previously described (17).

Vibratome Sectioning.

Fixed intestines were embedded in 7% (wt/vol) low-melt agarose (Sigma A9414) in PBS and vibratome sectioned at 100 μM.

Immunostaining.

Antibodies were: chicken anti–β-galactosidase 1:1,000 (AbCam ab9361), goat anti-GFP 1:1,000 (AbCam ab6662), mouse anti-Ecadherin 1:1,000 (BD 610182), rabbit anti-Arl13b 1:1,500 [kind gift from Tamara Caspary (Emory University, Atlanta, GA)], mouse anti-BrdU 1:400 (DSHB G3G4), or rat anti-BrdU 1:200 (Accurate OBT0030G).

Quadrat Counting and NCSF Analysis.

Clusters were counted in a set of n quadrats within an image of intestine. The number of clusters, x_i , were counted in the i th quadrat. The mean and sample variance were calculated as in ref. 46.

NCSF was implemented in a program written in R (47) to compute the Pearson correlation: Let S_V be an ordered list of the size of every cluster within a region of interest. Let r_i denote the i^th term in an ordered list of increasing radial distances (e.g., r₀ = 0 μm, r₁ = 10 μm, r₂ = 20 μm,…, r_n = 10 × n μm). For each cluster in S_V, let N_i be the average size of neighbors whose centers lie at a distance between r_i and r_i-1 from that cluster. Let S_i be an ordered list of the average neighbor sizes N_i for each cluster. Then the NCSF gives the correlation of S_v with S_i on the interval r_i-1 to r_i. See Table S2 for NSCF values, SEM and P values.

BrdU (50 mg/kg) was intraperitoneally injected into pregnant females at the stages indicated within the text and prepared as described previously (43).