Bending gradients: How the intestinal stem cell gets its home

Stem cells are restricted late in chick endodermal morphogenesis as zigzags become compact and begin to morph into previllus bulges

As Lgr5 expression is difficult to detect in the developing chick midgut, we utilized single molecule fluorescent in situ hybridization to locate Lgr5-expressing cells across chick intestinal development. Lgr5 is expressed uniformly throughout the early embryonic intestinal epithelium and continues as such through the early stages of epithelial morphogenesis into ridges and zigzags (Figure 2A, S2A). However, by E15, as the zigzags attain their maximal compaction just before they begin to morph into the bulges that will give rise to villi, Lgr5 expression is diminished in the tip of the epithelial fold. By hatching, expression is predominantly limited to the intervillous space (Figure 2A and S2A). Similarly, Sox9 is expressed uniformly in the early chick intestinal epithelium and is lost at the tips of the folds by E15 (Figure 2B). Thus, both the localization of putative stem cells and of the Wnt signaling that supports them, become restricted just before the pre-villus bulges start to emerge. We have previously noted that this transition from zigzags to bulges also correlates with and, indeed, depends upon a progressive restriction of proliferation from the tips of the folded luminal surface E15 (Shyer et al., 2013).

Open in a separate window

Figure 2

Restriction of progenitor identity is observed during the slower progression of villus formation in chick

(A) Quantification of single LGR5 mRNA molecules per unit length across the base, middle, and tip of epithelial folds over time (quantifications were done on at least 3 gut samples for each stage). Data are represented as mean +/− one standard deviation See also Figure S2. (C) Immunofluorescence for Sox9 in the chick intestine across development, from E13 when expression is uniform in the epithelium, through E15 when Sox9 is restricted from the tips of the folds, and at hatch when Sox9 is expressed predominantly in the intervillous space. Scale bars, 50 μm.

A signaling center correlating with the timing and localization of stem cells in the forming gut

A potential clue for how epithelial proliferation and stem cell identity might be regulated comes from the mouse, where lack of proliferation at the villus tip has previously been correlated with the presence of a signaling center in the distal mesenchyme of the nascent villi, called the “villus cluster” (Karlsson et al., 2000), which expresses PDGFRα, Gli1, Ptc1, Bmp2 and Bmp4 (Karlsson et al., 2000, Walton et al., 2012). In the chick, we find the same suite of genes are expressed at a high level in the equivalent location at the tip of the highly folded epithelium at E15, although the same genes are expressed at a lower level at earlier time points in a narrow band directly under the entire epithelium (Figure 3A). The time when the villus cluster genes are up-regulated is the same stage as when the overlying distal epithelium loses stem cell marker expression and as when proliferation decreases in the distal domain of the epithelium (Shyer et al., 2013). The chick villus cluster includes cluster-specific expression of Foxf1, a transcription factor implicated in villi formation (Ormestad et al., 2005) but not previously observed in the cluster, as well as PDGFRα, Ptc1 and Bmp4, (Figure 3A). We also examined phospho-SMAD staining, as a reporter of Bmp activity, during chick gut morphogenesis. Phospho-SMAD reactivity is identified with a timing that correlates with the onset of high level Bmp expression in the villus cluster and negatively correlates with the localization of Lgr5 expression (Figure 3B).

Open in a separate window

Figure 3

As the proto-villi form from E13 to E15 the villus cluster signaling center forms in the mesenchyme at the distal tip

(A) Luminal views of the zigzag topography from E13 to E15, expression of cluster genes goes from uniform under the wide folds of the epithelium at E13 (left) to predominantly localized to the mesenchyme under the forming villi at E15 (right). (B) PhophoSMAD staining demonstrates high BMP activity in the villus cluster and the adjacent epithelium. Close-up views (below) of a single fold at E15 highlight epithelial staining (arrowhead), which is less intense than staining in the mesenchymal cluster. Scale bars, 50 μm.

It has recently been shown that, in mouse, the villus cluster expression of Bmp4 and the general Shh target Ptc1 are downstream of hedgehog signaling (Walton et al., 2012, Ormestad et al., 2005). Moreover, it has long been known that at earlier stages in chick gut formation, Sonic hedgehog (Shh) is responsible for inducing expression of Bmp4 in the underlying mesenchynme (Roberts et al., 1995). Accordingly, we find that in the chick, villus cluster-specific expression of Ptc1, BMP4, as well as Foxf1 is lost upon inhibition of Hedgehog signaling by cyclopamine and expanded in response to additional Shh protein (Figure 4A). As expected, the decrease or increase of Bmp4 expression, in response to cyclopamine or Shh, respectively, is reflected by a concomitant respective loss of or broadening of phospho-SMAD reactivity (Figure 4B).

Open in a separate window

Figure 4

ISC localization is regulated by BMP signaling from the underlying mesenchymal villus cluster signaling center

(A) In situ hybridizations of E14 chick intestines cultured for 36 hours without (control) or with cyclopamine or recombinant Shh ligand. (B) PhosphoSMAD staining of cultured samples demonstrates the impact of compounds and recombinant proteins on BMP activity. (C) Edu labeling of E14 chick intestines cultured for 36 hours with the listed compounds and recombinant proteins. Below - quantification of percent Edu-positive cells across the sub-regions of epithelial folds, at least three folds on each of three samples were counted. (D) Sox9 staining of cultured samples demonstrates the effect of compounds and recombinant proteins on Wnt activity. (E) Quantification of single molecule FISH for LGR5 performed on sections from at least 3 E14 chick intestines cultured for 36 hours without (control) or with cyclopamine. See also Figure S2. Data are represented as mean +/− one standard deviation. Scale bars, 50 μm.

A feedback loop from the villus cluster to the epithelium localizes stem cells to the base of the forming villi

To test whether signals from the villus cluster, in fact, direct the fate of cells in the neighboring epithelium, we excised a small segment of intestine from an E14 chick embryo, when progenitors are uniformly distributed and before the villus cluster has formed, and manipulated cluster signals in vitro during 36 hours of culture. Control cultures display strong Edu labeling, indicative of proliferation exclusively at the base of the fold, just like their E15.5 in vivo counterparts (Figure 4C). However, culturing in the presence of the hedgehog inhibitor cyclopmaine or the Bmp inhibitor Noggin results in expansion of proliferation throughout the endoderm, including the villus tips (Figure 4C). Conversely, in explants cultured in the presence Shh or Bmp4, proliferation is absent not just from the tips of the villi, but from the entire endodermal layer. As shown above, Shh activity is responsible for inducing Bmp4 expression in the underlying mesenchyme. To confirm this epistatic relationship in this context, we simultaneously treated cultures with both Shh and Noggin. Application of both Shh and Noggin to gut segments in culture mimics the effects of Noggin alone, maintaining proliferation throughout the endoderm (Figure 4C). Thus, as expected, endodermally-derived Shh activity is upstream of mesenchymal Bmp4 expression, and Bmp4 activity represses endodermal proliferation.

Wnt signaling is an important niche signal for maintaining ISCs in the mature intestine. Moreover, mouse mutants with loss of villus cluster signals show an expansion of Wnt expression (Madison et al., 2005, Ormestad et al., 2005), suggesting that the presence of Bmp signaling at the tips of the villi may lead to the observed loss of ISCs in the overlying epithelium by reducing Wnt activity. Blocking Shh or BMP signaling resulted in uniform staining of the Wnt target Sox9 throughout the gut epithelium, while control gut tissue only showed Sox9 expression in the lower half of the villi (Figure 4D). Conversely, in explants cultured in the presence of Shh or Bmp4, Sox9 is absent in the endodermal layer (Figure 4D).

To directly verify that this signaling cascade regulates ISC restriction, we assessed the expression of the ISC marker Lgr5 in the presence of repressed Shh activity. As anticipated, when cyclopamine is added, abolishing villus cluster gene expression, the resulting intestine segments maintain expression of Lgr5 throughout the folded epithelium, whereas expression is lost at the tip in control segments (Figure 4E, S2B). Together, these results support a model where Shh activity in the gut endoderm induces villus cluster gene expression in the subadjacent mesenchyme at the tips of the villi. This signal center then produces Bmp4 which reciprocally feeds-back on the endoderm to block Wnt activity and hence repress ISC identity and cell proliferation at the distal end of the growing villi.

Physical changes in the morphology of the lining of the gut creates local maxima of signaling activity to induce the villus cluster

There is, however, an obvious problem with this model: We have shown that Shh is expressed uniformly throughout the gut endoderm at the stages of development under consideration (Figure 3B, ​,4A),4A), yet the putatively Shh-dependent villus cluster genes are only induced at the distal tips of the villi. A plausible model explaining this localized, elevated response to Shh takes note of the fact that a uniformly secreted protein will be at a higher concentration in locations where the target tissue is surrounded by morphogen-producing tissue (e.g. at the curved tip of the highly folded epithelium) than where it is only adjacent to the source of the morphogen on one side (e.g., at the base of the folds). This is supported by computational modeling, which shows that a highly folded epithelium, or finger-like pocket indeed results in both an increase in a morphogen concentration gradient and a greater depth of high level signaling below the endoderm, relative to a similarly scaled, wider fold (Figure S3 and Supplementary methods). The slow, stepwise nature of villi formation in chick allows for a detailed investigation of this hypothesis for how the villus cluster arises. During the stages where the lumen takes on an increasingly compact zigzag topography (E13, E14, E15) we find that the cross-sectional shape of these structures changes in concert (low peak, narrow peak, rounded tip, respectively) (Figure 5A, 5B). This would be predicted to lead to increasingly concentrated gradients of endodermally derived signaling at the tip (schematized in Figure 5B).

Open in a separate window

Figure 5

Non-uniform mesenchymal signals are downstream of uniform epithelial signal

(A) Luminal views of the chick intestine from E13 to E15, as progenitor identity is lost from the tips of the folds (also shown in Figure 3A). Dotted lines represent the plane of section for transverse views in B, C, D. (B) Schematic of diffusion of signal from an epithelium of the particular shape at each stage, darker color represents more signal. Note the increasing signal overlap in the underlying mesenchyme as the fold narrows. See also Figure S3. (C) In situ hybridization for Bmp4 (above) PDGFRα (below) expression from E13 to E15 matches the predicted pattern in B (also shown in Figure 3A). (D) Distribution of Shh protein in folded tips of the chick intestine at E13 and E15 (left). Antibody staining intensity across the 100μm region boxed on the left was quantified using the Plot Profile function in Fiji (right). Brightness values were normalized to background levels for each image. A comparison of Shh staining intensity in E13 (graphed in blue) versus E15 (graphed in red) shows increased Shh staining in the E15 mesenchyme (dotted line denotes epithelial-mesenchymal border). The staining intensities across the E13 and E15 epithelia are not significantly different (p<0.08). Three different z-slices from each of three samples were averaged for each stage. Below, the staining intensity found in a 5μm by 5μm region, a 5μm distance from the E15 tip epithelium (pink) is significantly brighter than the in the same sized region 5μm from the E15 base epithelium (yellow) (p<0.001). Measurements from two different z-slices from each of three samples were averaged for each E15 region. Data are represented as mean +/− one standard deviation. Scale bars, 25 μm.

To directly test this idea, we examined the distribution of Shh with an antibody directed against this protein. Anti-Shh staining intensity was plotted along a line from the tip of the folded epithelium and orthogonal to it (Figure 5D). Prior to E15, anti-Shh reactivity is identified in the epithelium and the mesenchyme just subjacent to the endoderm. However, at the transition from zigzags to bulges, the mesenchyme in the distal domain of the folded tissue showed significantly elevated Shh protein accumulation. In addition, the shape of the gradient tapers off much more slowly within the highly folded epithelium of the E15 gut than within the broader fold seen at E13. This is consistent with expectations, since – in addition to Shh protein diffusing from the tip, the mesenchyme within the narrowly folded E15 epithelium is exposed to Shh secreted from the epithelium lateral to it, augmenting the gradient. At both stages, the highest level of Shh staining is observed within the epithelium itself, which is to be expected as the antibody will detect both extracellular and intracellular protein in the tissue producing the Shh. Importantly, however, the level of Shh produced by the epithelium, averaged for the 9 sections assayed at each time point, is equivalent at E13 and E15. To further verify that the architecture of the tissue affects Shh protein accumulation, we compared the concentration of Shh protein 5 microns below the tip of the epithelial fold at E15 versus the concentration present at the same distance below the base of the fold. As expected, the intensity of staining is much higher within the fold, providing a mechanism explaining localized high-level Shh signaling at the epithelial tips.

If the mesenchyme responds to Shh by activating villus cluster genes at a high threshold concentration, this would explain the observed localization of high level villus cluster gene expression. Indeed, examination of the expression pattern of villus cluster markers such as PDGFα and Bmp4 gives results consistent with this model (Figure 5C). Consistent with epithelial morphogenesis acting upstream of increased Shh signaling and hence villus cluster gene activity, and not vice versa, after treating with cyclopamine to block hedgehog signaling we observed no alteration in the global structure of the epithelium or in individual epithelial or mesenchymal cell shape, using membrane-bound β-catenin to outline cell contours (Figure S4).

To test whether the bending of the epithelium into more tightly curved domains, with consequent high levels of localized signaling, is indeed responsible for the up-regulation of villus cluster genes in the tips of these structures, we undertook a simple experimental manipulation designed to “open” the normally tightly folded epithelium. Ringlets of embryonic intestine were excised at E14 and placed into culture in vitro. The folds in the epithelium arise due to constraint on the proliferating inner layers by subadjacent differentiated smooth muscle (Shyer et al., 2013). To alter this physical constraint, half of the rings were turned inside out, putting the endoderm and mesenchyme outside of the rings of smooth muscle, allowing the epithelium more length to take on a less folded form (Figure 6A). Following 36 hours of culture, the inside out ringlets indeed had a broader contour than their right side out counterparts. After culture, the ringlets were sectioned and processed for in situ hybridization with various villus cluster probes (Ptc1, Bmp4, PDGFRα, Foxf1). Each of these was strongly expressed at the fold tips of the control ringlets, but all were expressed uniformly at a lower level under the epithelium in the inside out ringlets (Figure 6B). Moreover, phospho-SMAD staining, indicative of Bmp up-regulation in villus clusters, is also greatly diminished in the inside out ringlets relative to control cultures (Figure 6B). These results suggest that the villus cluster forms in the mesenchyme at the tip of the fold because those cells are almost completely encapsulated by Shh-expressing epithelium allowing high threshold responses to be activated.

Open in a separate window

Figure 6

Epithelial shape directs cluster formation

(A) Experimental schematic: a ring of E14 intestine (left) is cultured for 36 hours either as a control segment, or after first being flipped inside out (right). (B) After 36 hours in culture, the cluster signal arises in the control rings (top), similar to what would be found in an E15 intestine. The rings that were flipped inside out before culture have an epithelial shape similar to E13 intestine and concomitantly an in situ pattern and phosphoSMAD staining that matches expression at E13. Proliferation (quantified as in Figure 4), Sox9 expression, and Lgr5 expression are all lost from the tips of folds that form in the control rings. See also Figure S2. (C) Experimental schematic: a slab of E10 intestine (left) is cultured for 36 hours either as a control segment (where wide ridges will be maintained), or under a fine grid which induces many small villi-like bumps (right). (D) After 36 hours in culture, the cluster gene expression and phosphoSMAD staining in control segments is nearly uniform under the epithelium. However, samples grown under the grid form villi-like bumps and display non-uniform expression of cluster genes and BMP activity with highest expression in areas of highest curvature. Proliferation and Sox9 expression are uniform in the control epithelium but in the samples cultured under the grid, proliferation and Sox9 expression are lost from the tips of folds that form particularly in areas where the curvature is highest, and where clusters of mesenchymal expression arise. Data are represented as mean +/− one standard deviation. Scale bars, 50 μm.

Preventing villlus cluster formation by flipping the intestines inside out results in an absence of the localized Bmp signal we demonstrated is responsible for restricting ISC localization within the gut epithelium. Thus, the inside-out ringlets of guts would be expected to maintain stem cell properties and proliferation throughout their epithelium. Indeed, such manipulations lead to maintenance of uniform proliferation throughout the epithelium, whereas proliferation is lost in the epithelium surrounding the cluster that forms in control rings (Figure 6B). Similarly, uniform expression of the Wnt target Sox9 and the ISC marker Lgr5 is maintained in the inside-out guts lacking villus cluster gene expression, whereas it is restricted from the folded tips in controls (Figure 6B and S2C).

As a second way of preventing late stages of epithelial morphogenesis, we took advantage of a drug, FK506, that has been shown to block smooth muscle differentiation (Fukuda et al. 1998). As we previously showed (Shyer et al., 2013), differentiation of smooth muscle layers is necessary for generation of the compressive forces that buckle the endoderm into ridges, zig-zags, and then villi. We cultured guts in vitro from ridge stage to late zig-zag stage, with or without the presence of FK506. Consistent with the results described above, without longitudinal muscle differentiation, and hence without progressing beyond parallel ridges, the entire endoderm remains proliferative and villus cluster genes are never up-regulated. As in vivo, control cultures display restricted distal proliferation and activation of villus cluster gene expression (Figure S5).

These results indicate that the three dimensional folding of the epithelium is necessary to locally increase Shh signaling (as seen in Ptc1 expression) and induce the villus cluster genes. To see if it is also sufficient, we sought to create villus-like structures at a stage when the epithelium is normally not as tightly folded. Slabs of embryonic gut were excised at E10, when the gut is folded into several wide ridges, and placed into culture in vitro. Half of the slabs were placed under a fine grid, causing the luminal surface to fold, with continued growth, into many small villus-like bumps, long before endogenous villus formation takes place (Figure 6C). Slabs were cultured for 36 hours and then processed for in situ hybridization with various villus cluster probes (Ptc1, Bmp4, Foxf1, PDGFα). After 36 hours in culture, there is no change in expression of Shh itself, which continues to be expressed uniformly in the epithelium under these conditions (Figure 6D). However, while villus cluster gene expression in control segments is nearly uniform at a low level under the epithelium, samples grown under the grid display elevated expression in the mesenchyme under areas of highest epithelial curvature. PhosphoSMAD staining is observed in the same locations reflecting the change in BMP pathway activity (Figure 6D). Therefore, simply morphing the tissue into the necessary shape can induce villus cluster-like local maxima of Shh responsive genes. Further, while proliferation and Sox9 expression are uniform in the control epithelium, in the samples cultured under the grid proliferation and Sox9 expression are lost from the tips of the folds, surrounding the areas where mesenchymal expression of cluster genes is highest (Figure 6D).

Taken together, these results demonstrate that the villus cluster genes are induced at local maxima of Shh activity, resulting from the additive effect of signaling that is compounded through the folding of the overlying epithelium.

Localization of ISCs in mice

Intriguingly, although the intestinal lining of both birds and eutherian mammals are characterized by the presence of long finger-like villi, this morphology appears to have evolved convergently, as the gut morphology of lower animals including fish (Walker et al., 2004), amphibians (McAvoy et al., 1998), reptiles (Ferri et al., 1976 and Kotzé et al., 1995), and even monotremes (Krause et al., 1975) include various forms of ridges and folds to increase the surface area of the lining of the gut, but not individual villi. The tight packing and long projections of individual villi represent an optimized solution for increasing surface area (hence allowing maximal absorption of nutrients) may have been selected for independently in the two most highly metabolic lineages, mammals and birds. The step-wise progression of mucosal folds from ridges to zigzags to villi has been well described in the chick (Coulombre and Coulombre, 1958). A similar series of transitions, involving segementation of previllus ridges to form villi, has been described for several mammals including cattle (Winkler and Willie, 1998) and humans (Hilton,1902, Lacroix et al., 1984). In striking contrast, the villi of the murine intestine form directly from the floor of a smooth epithelium (Sbarbati et al., 1982). The process of villus formation in the mouse does, nonetheless, share at least some mechanistic aspects with the chick and other guts where villi form via segmentation. In both chick and mouse embryonic gut, villus formation is prevented by blocking differentiation of the smooth muscle (which at least in chick acts as a barrier to expansion of the epithelium, thereby causing mucosal buckling). Moreover, modeling of the physical properties of the embryonic mouse intestine indicate that compressive mechanical forces induced by constrained growth are sufficient to explain the emergence of villi in mice as in chick (Shyer et al., 2013).

Consistent with this, we found that concomitant with smooth muscle differentiation, the mouse epithelium buckles into small alcoves that could, in principle, lead to local elevated concentrations of Shh protein prior to the onset of villus cluster gene expression. As in the chick, stem cell markers and Wnt-responsive genes are expressed uniformly throughout the gut epithelium prior to this point, and are down-regulated at the tips of the forming villi as the villus cluster genes are expressed. Also as in the chick, blocking the Shh pathway, and thus downstream BMP signaling, is sufficient to expand proliferation and the expression of Lgr5, suggesting that the presence of Shh signaling normally acts to restrict them from the villus tips. Finally, creating villus-like structures prematurely results in the up-regulation of a marker of the villus cluster, through geometric constraint. While the central features involved in gut stem cell localization during villus formation, thus, appears to be the same in mice and chicks, there is some evidence that there may be differences as well. For example, formation of the villus cluster in the mouse appears to involve cell aggregation (Walton et al., 2012) as well as induction of gene expression, a feature we have not observed in the chick. Further work will be required to gain a fuller picture of how villus formation and stem cell location are achieved in mice, and to integrate other findings with the results described here.

Bmp antagonism of Wnt activity in restricting proliferation and stem cell activity

Our data show that the net result of the Shh-Bmp signaling cascade is a restriction of proliferation as well as a decrease in expression of Wnt-dependent stem cell markers at the tips of the developing epithelial folds. We did not, in the context of this study, explore how this is achieved. However, a similar Bmp antagonism of Wnt activity has previously been described in the context of the adult intestinal stem cell niche. As we observed embryonically, Bmp ligands are also strongly produced by the intervillus mesenchyme near the tips of the adult villi, with a decreasing gradient towards the crypts (He, 2004; Hardwick et. al., 2004; Haramis et al., 2004; Batts et al., 2006). Moreover this Bmp activity in the adult villus acts to suppress Wnt signaling to control the balance of stem cell renewal and differentiation (He, 2004). In this context, the Bmp and Wnt pathways are integrated intracellularly at the level of a PTEN/Akt-dependent mechanism (Tian et al., 2005). It seems likely that this same or a similar mechanism is employed downstream of Bmp activity at the earlier stage investigated here.

Mechanically-based induction of gene expression

The physical reshaping of morphogenic gradients represents an intriguing paradigm in the integration of mechanics and developmental signaling. Of course, in addition to this mechanism, many instances have been described wherein forces impact gene expression through mechanosensory signal transduction. In a formal sense, it is certainly possible that mechansenory signaling also contributes to the activation of target gene expression during gut epithelial morphogenesis. However, we emphasize that ectopic action of Shh is sufficient to induce villus cluster gene expression and to restrict the location of stem cells and proliferation, while blocking Shh activity is sufficient to result in a loss of villus cluster gene expression and expansion of proliferation and stem cell localization. Moreover, addition of cyclopamine has no effect on the contour of the epithelium or the shape of individual epithelial cells (Figure S4). As the epithelium is bent equivalently under conditions with or without cyclopamine, the cells should be seeing equivalent strains and stresses, and hence similar mechanosensory signaling. Yet the cultures with cyclopamine lose villus cluster gene expression, while control cultures do not, clearly indicating that there is at least a major part of the process that is independent of mechanosensory transduction.

Immunohistochemistry and Edu Staining

Small intestines were collected from embryos at desired stages and fixed in 4% paraformaldehyde in PBS and embedded in OCT, allowing for 14μm transverse sections of the gut tube. CD44 immunohistochemistry was performed with rat anti-CD44 (v6) (1:100 Biosciences) and detected using the Anti-Rat HRP-DAB Cell & Tissue Staining Kit (R & D Systems). The following antibodies were used for immunofluorescence staining at the listed concentrations: Sox9 (1:100, R & D Systems), β-catenin (1:100, Sigma), PDGFα (1:100 in chick, 1:300 in mouse, Santa Cruz), FITC-conjugated Smooth muscle actin (1:100, Abcam), phospho-SMAD 1/5 (1:300, Cell Signaling), Shh (5E1, 1:20). Sections were incubated with primary antibody overnight at 4 degrees and then incubated with Alexa secondary antibodies used at 1:300 for 2 hours at room temperature. DAPI (molecular probes) was used as a nuclear counter stain. 100μM Edu (Invitrogen) was added to guts in culture and samples were harvested after 4 hours of Edu incubation. Edu was detected in sectioned tissue using the Click-iT Edu system (Invitrogen).

In situ hybridization and single molecule fluorescent in situ hybridization (FISH)

Tissue samples for section in situ hybridization were fixed overnight in 4% PFA. After fixation, the tissue was rinsed in PBS and incubated in 30% sucrose overnight at 4°C before being embedded in OCT. 14um thick cryosections were collected for DIG-labeled RNA in situs and 10um thick sections were collected for single molecule FISH. DIG-labeled in situs were performed as described previously (Brent et al. 2003). Single molecule FISH experiments were performed according to Raj et al. 2008 and Itzkovitz et al. 2011.

We thank Alan Rodrigues and Sandy Klemm for assistance with single molecule FISH probe design and experiments, Jessica Lehoczky for help with mice, Maxwell Heiman for the use of his microscope, and Jinhyun Choo for help with the simulation. We also thank D. Gumucio and K. Walton for discussions and for sharing data prior to publication. This work was supported by grants from the NIH (HD047360) to CJT and the MacArthur Foundation to L.M.