Erk and MAPK signaling is essential for intestinal development through Wnt pathway modulation

Adult stem cells are crucial for cell proliferation and regeneration of differentiated cells during tissue homeostasis. The intestinal epithelium undergoes rapid cell turnover, and well-controlled cell differentiation and maturation, making it an excellent system for studying adult stem cell function and self-renewal. The maintenance of epithelial self-renewal is supported by two populations of intestinal stem cells (ISCs): Lgr5-positive fast-cycling crypt base columnar (CBC) stem cells wedged between Paneth cells (Barker et al., 2007); and quiescent +4 stem cells located just above the Paneth cells (Sangiorgi and Capecchi, 2008). Recently, increasing evidence suggests that +4 stem cells are very likely secretory progenitors (van Es et al., 2012) and should be considered a ‘facultative stem cell’ (Visvader and Clevers, 2016). More recently a novel damage-induced quiescent cell type called the revival stem cell has been identified; it is probably not a regular ISC because it is rarely found under homoeostatic conditions (Ayyaz et al., 2019). ISCs constantly generate transit-amplifying (TA) cells that further differentiate into postmitotic absorptive lineage (enterocytes) and secretory lineage (goblet cells, Paneth cells, enteroendocrine cells, Tuft cells, etc.) cells in the small intestine. Stem cell hierarchies in the small intestine are rigorously controlled by the orchestration of reciprocal niche signals from neighboring Paneth cells and underlying mesenchymal cells (Sato et al., 2011; Clevers, 2013).

The maintenance of ISCs and tissue homeostasis is tightly orchestrated by several crucial signaling pathways. Wnt signaling is the essential driving force supporting ISCs and maintaining the crypts as well as inducing Paneth cell maturation (van Es et al., 2005a; Liu et al., 2013). In cooperation with Wnt signaling, the Notch (Fre et al., 2009; van Es et al., 2005b) and EGF (Wong et al., 2012; Miettinen et al., 1995) pathways promote crypt expansion and regulate cell differentiation. In contrast, the BMP (He et al., 2004; Haramis et al., 2004) and hedgehog (Kosinski et al., 2010; Akiyoshi et al., 2006) pathways negatively regulate epithelial Wnt signaling to promote cell differentiation. Moreover, epithelial cells secrete hedgehog, mainly Indian hedgehog (Ihh), to induce mesenchymal BMP production, which is crucial for epithelial-mesenchymal interactions and crypt formation (Ramalho-Santos et al., 2000). However, it is still obscure how the Ihh pathway is regulated by upstream signals in the intestinal epithelium.

The mitogen-activated protein kinase (MAPK) pathway, a downstream signaling target of EGF and other mitogens, plays pivotal roles in regulating intestinal epithelial cell (IEC) growth, Paneth versus goblet cell choice and the ISC population; however, the underlying mechanism is still disputed (Haigis et al., 2008; Trobridge et al., 2009; Heuberger et al., 2014; de Jong et al., 2016). Intestinal-specific deletion of the MAPK pathway upstream activator Shp2 promotes Paneth cell differentiation and expands the ISC pool through activation of the Wnt pathway (Heuberger et al., 2014). Expression of the constitutively active G12 V mutant Kras in mouse intestines promotes goblet cell differentiation in a Notch pathway-dependent manner, but does not affect stem cells and the Wnt signaling pathway (Feng et al., 2011). It suggests that the role of MAPK signaling in intestinal homeostasis is complicated and individual components of this cascade may have distinct physiological functions in intestinal homeostasis.

As one of the major direct targets of the Ras-Raf-MEK cascade, the extracellular signal-regulated kinases 1 and 2 (Erk1/2) are crucially involved in multiple physiological and pathological functions in various tissues, including embryonic development, reproduction, T cell development, immunoregulation and tumorigenesis (Yao et al., 2003; Fan et al., 2009; Fischer et al., 2005). Erk1 and Erk2 are homologous genes with distinct functions during embryonic development, but in the majority of the tissues they mainly play redundant roles. Recently, it was reported that IEC-specific depletion of Erk1/2 in adult mice results in dysregulated epithelial cell migration, secretory cell differentiation and nutrient absorption with aberrant Erk5 activation (de Jong et al., 2016). Although Erk5 activation could partially explain the Erk1/2 depletion-induced cell proliferation and migration phenotype, it is still not clear how Erk1/2 regulates the cell differentiation and morphology changes that are crucial for intestinal homeostasis. To further understand the physiological function of Erk1/2 at earlier stages, we deleted Erk1/2 in IECs during embryonic stages using Villin-Cre, and found dramatic expansion of the ISC population, impaired secretory lineage differentiation and enterocyte maturation leading to postnatal lethality. Unexpectedly, Erk1/2 deficiency also stimulated hyperproliferation and migration of epithelial cells, and caused profound morphological changes to polyp-like structures in the small intestines in postnatal mice. Most of the defects in Erk1/2 mutant intestines were attributed to strongly activated Wnt signaling through decreased Ihh production and Ras/Akt/mTOR signaling activation, due to loss of Erk1/2-mediated feedback inhibition. Our study demonstrates that Erk1/2 plays a crucial role during intestinal development through modulation of the Wnt signaling pathway through both cell-autonomous and non-autonomous mechanisms.

Previous studies have shown that active Erk1/2 is localized in the TA zone of the crypt (de Jong et al., 2016). Through immunofluorescence staining, we found that, in addition to TA cells, phospho-Erk (p-Erk)-positive cells also colocalized with Lgr5-positive ISCs, which expressed eGFP in intestinal tissue of adult Lgr5-eGFP-IRES-creERT2 mice (Barker et al., 2007) (Fig. 1A). In the embryonic stages, Erk is activated uniformly throughout the IECs (Fig. 1B) as ISCs and progenitors occupy the entire intestinal epithelium during embryonic development. These data suggest that Erk/MAPK signaling is activated in highly proliferating IECs during development and adulthood. To analyze the function of Erk1/2 signaling in intestinal development, compound genetically modified mice were generated by crossing Erk1^−/−;Erk2^fl/fl (Fischer et al., 2005; Pages et al., 1999), as we previously described (Fan et al., 2009), with Villin-Cre transgenic mice (Madison et al., 2002) (referred to as Erk^ΔIEC hereafter). Tissue-specific depletion of Erk1/2 was confirmed by immunofluorescence and immunohistochemistry (Fig. S1A).

The body size of newborn Erk^ΔIEC mice was similar to littermate controls (data not shown) but the subsequent weight gain was dramatically reduced in Erk^ΔIEC pups, probably owing to poor nutrition status because Erk1/2 was specifically deleted in intestinal epithelial cells (Fig. 1C). Although the time of lethality varied, all of the Erk^ΔIEC pups died within 30 days, demonstrating an essential role for Erk signaling in IECs, as previously reported for deletion of Erk1/Erk2 in adult mice (de Jong et al., 2016). Through histological analysis of tissues from P14 mice, we found that the morphology of the entire intestine was severely disrupted in Erk^ΔIEC mice (Fig. S1B). In some particular sites, branched villi were observed in the small intestine of Erk^ΔIEC mice (Fig. 1D), and the morphology of the colon was also disorganized with polyp-like structures (Fig. 1E). As determined by scanning electron microscope (SEM), irregular projections of fused giant villi were formed in Erk^ΔIEC small intestines but not in controls (Fig. 1F). Further analysis showed that no obvious morphological defects were observed until P7, and deletion of Erk1 in mice did not affect intestinal development (Fig. S1C,D). These data suggest that Erk signaling plays a pivotal role in regulation of intestine morphology during development.

Numerous studies have shown that the Erk/MAPK pathway is crucial for cell proliferation; therefore, we assumed that ablation of Erk would inhibit intestinal cell growth. To our surprise, the number of Ki67⁺ proliferating cells was dramatically increased in both the small intestine and colon of Erk^ΔIEC mice, and the crypts were swollen with increased size in Erk^ΔIEC mice compared with controls (Fig. 2A). Increased cell proliferation in Erk^ΔIEC intestines was first observed at postnatal day 7, but not in embryonic stages and P0 (Fig. S2A). Progenitors in the crypts proliferate quickly and migrate toward the villi. To determine whether the expanded number of proliferating cells in crypts is due to a reduction in cell migration rate, BrdU pulse-chase experiments were performed. We found that the cells in Erk^ΔIEC mice exhibited a much faster migration rate from the crypt to villus in both small intestine and colon tissues (Fig. 2B; Fig. S2B). To determine whether this phenotype is a cell-autonomous function of Erk signaling in epithelial cells, the small intestinal organoid culture system was employed (Sato et al., 2009). Isolated Erk^ΔIEC organoids were much larger than controls, consistent with the above observation of expanded crypts in Erk1/2^ΔIEC mice (Fig. 2C). Through consecutive monitoring of single live organoids, the growth rate and the number of budding protrusions were dramatically increased in Erk^ΔIEC organoids (Fig. 2C), which was further confirmed by EdU labeling (Fig. 2D). These data demonstrate that Erk1/2 has a cell-autonomous role in limiting IEC cell proliferation and migration.

Next, we analyzed the role of Erk1/2 in intestinal epithelial cell differentiation. Consistent with previously reports (de Jong et al., 2016), ablation of Erk1/2 in IECs led to impaired secretory lineage differentiation and maturation. A dramatically increased number of Paneth cells with disordered localization was observed in P14 Erk^ΔIEC intestines (Fig. 3A). Moreover, premature Paneth cells were present in P6 Erk1/2^ΔIEC mice but not in the controls (Fig. S3A). In Erk^ΔIEC intestines, the number of goblet cells, but not enteroendocrine cells, was significant decreased (Fig. 3B; Fig. S3B,C). The number of Tuft cells, a rare secretory lineage cell type initiating anti-parasite type 2 immunity in the intestines (Gerbe et al., 2016), was dramatically reduced, as determined by immunostaining for the specific maker Dclk1 and the mRNA expression levels of other Tuft cell markers, such as Chat, Trpm5 and Ptgs1 (Haber et al., 2017) (Fig. S3D-F). The number of alkaline phosphatase (AP)-positive enterocytes was reduced and AP activity declined in Erk^ΔIEC mice compared with controls. It demonstrates that the differentiation and maturation of absorptive enterocytes of the Erk^ΔIEC mice is inhibited (Fig. S3G). Moreover, through SEM analysis, the enterocyte brush border microvilli were sparse, short and thin in Erk^ΔIEC mice (Fig. S3H), suggesting an impaired absorption function, which is probably a major reason why mutant mice were lean and malnourished.

As active Erk1/2 was observed in ISCs (Fig. 1A), we investigated whether ISCs were affected in Erk1/2^ΔIEC intestines. Olfm4-positive cells were clearly expanded in the crypts of Erk^ΔIEC intestines compared with controls both at P7 and P14 (Fig. 3C and Fig. S3I). Similar results were also observed in Erk1/2 mutant Lgr5-eGFP-IRES-creERT2 mice, as determined by staining of Lgr5-GFP positive cells (data not shown). Expansion of ISCs was further supported by real-time qRT-PCRs, which showed that levels of common ISC markers such as Ascl2 (van der Flier et al., 2009), Aqp1 and Smoc2 (Munoz et al., 2012), were significantly increased (Fig. 3D and Fig. S3J). These data suggest that Erk signaling is crucial for intestinal epithelial cell differentiation, including the choice between enterocyte and secretory cell lineages, as well as for controlling terminal differentiation in the secretory lineage. Moreover, Erk is also important for maintaining the ISC population.

As shown in Fig. 1, deletion of Erk1/2 in epithelial cells leads to disruption of the villus structure, which is controlled by the interaction of both the epithelium and the mesenchyme underneath. We investigated the α-smooth muscle actin (α-SMA)-positive muscularis mucosa (MM) and found that the layer of α-SMA-positive MM beneath the crypts was absent in some parts of Erk1/2^ΔIEC small intestine but not in control mice (Fig. 3E). Further analysis showed that α-SMA-positive MM was smaller at postnatal day 7, but not at embryonic stages and P0 (Fig. S4A). Although α-SMA-positive MM constitutes a consecutive layer in colon tissues, the thickness of the layer was reduced in Erk mutant tissue (Fig. 3F). Mesenchymal cells in the villus core and pericryptal region underwent rapid proliferation postnatally. Both the number and the size of proliferating pericryptal cells were greatly reduced in Erk^ΔIEC mice compared with the controls at P6 (Fig. 3G,H). These data suggest that epithelial Erk signaling is crucial for intestinal subepithelial cell proliferation and maturation.

The perturbation of intestinal cell differentiation and ISC expansion observed in Erk^ΔIEC mice share a similar phenotype with intestinal-specific deletion of the Apc gene, which results in hyperactivation of Wnt/β-catenin signaling (Sansom et al., 2004). As shown in Fig. 4A, we found that the spectrum of cells expressing direct Wnt/β-catenin target genes, such as Myc and Sox9, was broader in Erk^ΔIEC intestines than in control intestines. Moreover, the expression level of another Wnt target, cyclin D1, was dramatically elevated in Erk^ΔIEC intestinal tissues compared with controls (Fig. 4B). Consistently, more cells with a nuclear localized active form of β-catenin in Erk^ΔIEC crypts were observed (Fig. 4C). RNA-sequencing was performed to uncover the transcriptome profile of Erk^ΔIEC intestines versus controls. After analysis of the geneset of canonical Wnt target genes in the intestine (Van der Flier et al., 2007), we found that most of the Wnt/β-catenin-dependent genes were significantly upregulated in Erk^ΔIEC mice compared with controls (Fig. 4D), suggesting a global activation of Wnt downstream signaling.

It is well accepted that Wnt signaling activates the Notch pathway through transcriptional activation of Notch ligands, such as Dll1 and jagged 1, in development and colorectal cancer (Galceran et al., 2004; Rodilla et al., 2009). We found that the expression of both Notch ligands and downstream genes was dramatically elevated in Erk^ΔIEC intestines compared with controls (Fig. 4E). This suggests that the loss of goblet cells in Erk^ΔIEC is likely due to Wnt-induced Notch activation. To further confirm that Notch hyperactivation in Erk^ΔIEC intestines is due to elevated ligand expression, we used DBZ, a γ-secretase inhibitor, which attenuates Notch activation at the receptor level, to treat the mice. As a positive control, DBZ treatment increased the goblet cell number in the intestines of control mice. DBZ also restored the differentiation of goblet cells in Erk^ΔIEC mice to a level similar to DBZ-treated controls (Fig. 4F). These data suggest that Erk deficiency induces Wnt signaling activation, which increases Notch ligand expression and Notch pathway activation to regulate secretory cell lineage differentiation.

The phenotype of polyp-like morphology observed in Erk^ΔIEC intestines is similar to that in hamartomatous polyposis syndromes, which are usually caused by genetic mutations leading to activation of Akt/mTor signaling pathway in the gastrointestinal tract (Zbuk and Eng, 2007). To test whether Erk depletion induced Akt/mTor signaling, we examined the activity of this pathway in the intestines. The activity of Akt is not very strong in wild-type intestines, as previously described (He et al., 2007). In contrast, Erk^ΔIEC crypts contained many more crypt cells with stronger p-Akt staining signals than controls (Fig. 5A). As expected, elevated Akt activity also induced hyperactivation of mTor signaling, as shown by the p-S6 and p-4E-BP1 level in intestinal tissue. mTor activity is normally restricted to the crypt and lower part of the villus, but in Erk^ΔIEC mice, cells residing in the upper part of the villi exhibited stronger p-S6 and p-4E-BP1 staining as early as P7 (Fig. 5B,C and Fig. S5A). Further studies using western blotting also confirmed that the Akt/mTor pathway was strongly activated in Erk^ΔIEC intestines (Fig. 5D). It has been reported that Akt activates β-catenin through direct phosphorylation of Ser552, which promotes crypt expansion in intestines (He et al., 2007). Activated Akt in Erk-deficient intestines also elevated the level of p-β-catenin Ser552 (Fig. 5D), suggesting that aberrant Akt activation is one of the factors driving Wnt hyperactivation in the Erk^ΔIEC intestine. To understand the reason why Erk deficiency induces Akt/mTor activation, we examined the activity of MAPK components upstream of Erk1/2, because many studies have demonstrated negative-feedback regulation of the MAPK pathway by Erk in cancer cells (Lake et al., 2016). Inhibition of Erk or Mek can activate the PI3K/Akt pathway via distinct mechanisms in many types of cancer (Lake et al., 2016; Zmajkovicova et al., 2013; Prahallad et al., 2012). We found that active Ras, Raf and Mek1/2 levels, but not Egfr levels, were dramatically elevated in Erk^ΔIEC intestines (Fig. 5E). This suggests that deletion of Erk1/2 results in activation of MAPK, probably downstream of the receptor. Immunohistochemistry studies also confirmed that Mek1/2 was activated in Erk^ΔIEC intestines, especially in tissue with malformed villi and crypts (Fig. 5F). However, ERK^ΔIEC organoids still required EGF in culture medium (data not shown), suggesting not fully functional activation of MAPK upon ERK depletion. These data suggest that hyperactivation of MAPK components, such as Ras and Raf, transactivates Akt signaling, as previously reported (Blalock et al., 2003; Lake et al., 2016).

Deletion of Erk1/2 specifically in epithelial cells also impaired mesenchymal cell proliferation and maturation (Fig. 3), suggesting that Erk regulates factors secreted by epithelial cells that function in the mesenchyme. We focused on hedgehog, as it is secreted from IECs and plays essential roles in intestinal development and villus patterning via its receptors in mesenchymal cells (Kosinski et al., 2010; Madison et al., 2005). Through analysis of the RNA-Seq data, we found that hedgehog signaling was dramatically impaired (Fig. 6A). A qRT-PCR assay demonstrated that mRNA levels of Ihh, as well as hedgehog pathway downstream targets, such as Hhip, Ptch1 and Gli1, were significantly reduced in Erk^ΔIEC intestines (Fig. 6B).

Immunohistochemistry studies revealed that the expression of Ihh was reduced in Erk^ΔIEC epithelial cells compared with controls (Fig. 6C). To confirm that Ihh expression is controlled by the Erk pathway in epithelial cells, organoids from both control and Erk^ΔIEC mice were cultured. Downregulation of Ihh expression was confirmed in mutant organoids using western blotting (Fig. 6D). Ptch1, the receptor and direct downstream target of the hedgehog pathway, was downregulated in Erk^ΔIEC mesenchymal cells (Fig. 6E). In the intestines, Bmp4 and its signaling are the major downstream effectors of the Ihh pathway responsible for intestinal morphology and function (Roberts et al., 1995; van Dop et al., 2009). We found that downregulation of Bmp4 protein was correlated with the Ihh level in Erk^ΔIEC intestinal tissues (Fig. 6F). Immunohistochemistry analysis demonstrated that phospho-Smad1/5/8 protein levels were greatly reduced in Erk^ΔIEC intestines, suggesting downregulation of BMP signaling (Fig. 6G). These data suggest that, in IECs, Erk proteins regulate the expression of hedgehog ligands, which further modulates mesenchymal cells through Bmp signaling.

As rapamycin treatment efficiently suppresses hamartomas in a mouse model of Peutz-Jeghers polyposis (Wei et al., 2008), we investigated whether suppression of mTOR activity could rescue the defects observed in mice lacking Erk in the intestines. Erk^ΔIEC pups were treated with rapamycin from P7 to 4 weeks of age. Rapamycin treatment significantly prolonged the lifespan and increased the survival rate of Erk^ΔIEC mice. All Erk^ΔIEC pups died within 4 weeks; however, after rapamycin treatment, 40% of them survived for the 8 weeks of observation (Fig. 7A). Through histological analysis of intestinal tissues, we found that rapamycin treatment restored the crypt-villus structure and prevented polyp-like defective tissue formation (Fig. 7B), although the villi in rapamycin-treated Erk^ΔIEC mice were still dilated compared with controls. Additionally, Ki67-positive proliferating cells were present only in the crypts of rapamycin-treated Erk^ΔIEC mice and in controls (Fig. 7C). Rapamycin treatment also balanced secretory lineage differentiation in Erk^ΔIEC intestines, as the number of Paneth cells and goblet cells was restored to normal levels (Fig. 7D,E). Ectopic localized Paneth cells were not observed in rapamycin-treated Erk^ΔIEC mice (Fig. 7D). These data suggest that rapamycin treatment prolonged the lifespan of Erk^ΔIEC mice and restored epithelial cell proliferation and differentiation.

To further confirm the effect of rapamycin at the molecular level, we examined the p-4E-BP1 levels in rapamycin-treated Erk^ΔIEC intestines and found they were comparable with control mice (Fig. 7F). Hyperactivated Wnt/β-catenin signaling in Erk^ΔIEC intestinal tissues was suppressed by rapamycin treatment, as determined by the level of Wnt downstream targets, such as cyclin D1, Myc and Sox9 (Fig. 7G-I). In addition, rapamycin treatment also increased the protein level of p-Smad1/5/8 in villi of Erk^ΔIEC intestines, suggesting the restoration of the BMP/Smad pathway (Fig. 7J). Consistent with the restoration of BMP signaling, hedgehog downstream targets (but not hedgehog ligands) were also recovered (Fig. 7K), as further supported by the reappearance of α-SMA positive ISEMFs in Erk^ΔIEC mice after rapamycin treatment (Fig. 7L). As Akt is the upstream signal of mTor, an Akt inhibitor should also partially rescue the phenotypes of Erk1/2 deletion in IECs. We tested the effect of an Akt inhibitor (MK-2206) on inhibition of signaling cascades in organoid cultures. MK-2206 inhibited Akt and 4EBP phosphorylation (Fig. S6A). Akt inhibitor treatment significantly decreased the number of Ki67-positive cells compared with control Erk^ΔIEC organoids (Fig. S6B). Moreover, Akt inhibition also partially restored cell differentiation defects in Erk^ΔIEC organoids, such as decreased Paneth cell number and increased goblet cell number (Fig. S6C-G). These data demonstrate that suppression of a hyperactivated Akt-mTor pathway partially rescues Erk deficiency-induced intestinal defects at the histological and molecular level.

In this study, we have investigated the role of the MAPK/Erk pathway in IECs during postnatal development. Erk1 and Erk2 are highly activated in stem/progenitor populations and depletion of Erk1/2 in IECs during prenatal stages surprisingly increases epithelial cell proliferation, crypt dilation, and impaired cell differentiation. More importantly, we have shown for the first time that ablation of Erk1/2 in IECs disrupts intestine morphology, leading to polyp-like structures in a hedgehog-BMP-dependent non-cell-autonomous manner. Erk1/2 deficiency increased the Wnt and Akt-mTor pathways through multiple mechanisms, and suppression of mTor by rapamycin partially rescued cellular and molecular defects, significantly increasing the survival rate of Erk^ΔIEC mice.

Multiple lines of evidence demonstrate that the MAPK pathway plays a crucial role in determining secretory lineage fate in the intestines. Consistent with its upstream regulators, such as Shp2 (Heuberger et al., 2014; Yamashita et al., 2014) and Ras (Feng et al., 2011), Erk also promoted goblet cell and inhibited Paneth cell differentiation, suggesting that this shift is a conserved function of the MAPK pathway. As the most recently defined secretory cell type, Tuft cells do not share common progenitors with goblet and Paneth cells (Gerbe et al., 2011), and the molecular mechanisms for Tuft cell differentiation are largely unknown; however, a recent study has shown that the differentiation of Tuft cells is promoted by Wnt signaling (Ladang et al., 2015). Here, we demonstrated that ERK deficiency induced Wnt hyperactivation but reduced the Tuft cell population, suggesting that Erk signaling is crucial for Tuft cell differentiation downstream of the Wnt pathway.

Hedgehog ligands are secreted by epithelial cells in the intestines to regulate mesenchymal cell development and function, which are crucial for crypt-villus patterning and ISC expansion. We found that Erk^ΔIEC mice phenocopy the mice with a conditional knockout of Ihh, the key hedgehog ligand in intestine development (Kosinski et al., 2010; van Dop et al., 2010). Shh has been shown to play mild roles in IEC proliferation and secretory lineage differentiation, but specific deletion of Shh in mouse IECs does not affect intestinal morphology (Gagne-Sansfacon et al., 2014). Thus, we believe Erk mainly affected Ihh signaling. More importantly, downregulation of the Bmp4-Smad pathway and activation of Wnt signaling were observed in Erk^ΔIEC mice, which is consistent with intestinal-specific deletion of Ihh in mice (van Dop et al., 2010). We found that the expression of Ihh was downregulated in Erk1/2-deficient tissue and organoids, suggesting the regulation is through a cell-autonomous manner. Our study demonstrates a previously uncovered connection between MAPK/ERK signaling and the hedgehog pathway in intestinal development.

The crosstalk between Wnt and MAPK signaling is complicated under both physiological and pathological conditions. It is well accepted that Wnt and MAPK/ERK are mutually activating pathways in colorectal cancer. For example, Wnt signaling induces MAPK/ERK pathway through stabilization of Ras protein (Jeong et al., 2012) and MAPK/ERK activates Wnt signaling through phosphorylation of LRP6 and other mechanisms (Lemieux et al., 2015; Ji et al., 2009; Cervenka et al., 2011). However, growing evidence has demonstrated negative regulation mechanisms between these two pathways, especially during homeostasis. Using IECs as a model, a study suggests that Shp2/MAPK signaling inhibits the Wnt pathway potentially due to increased stability of a shorter TCF4 (TCF4M/S) splice variant, which produces a dominant-negative truncation (Heuberger et al., 2014). Using PORCN inhibitor to acutely suppress Wnt signaling in vivo, it has been reported that the Wnt pathway suppresses the MAPK cascade to maintain a stable pool of ISC (Kabiri et al., 2018). Here, we also found that the Erk pathway restrains Wnt signaling activity to maintain cell migration and differentiation in the intestines, through more-complicated mechanisms in both cell-autonomous and non-autonomous manners. First, epithelial Erk1/2 stimulated the Ihh level, activating the Hh pathway to enhance BMP production and ultimately restrain Wnt/β-catenin. In addition, we found that deletion of Erk1/2 resulted in activation of the PI3 K/Akt pathway, which has been shown to activate Wnt/β-catenin through inhibition of Gsk-3β and direct phosphorylation of β-catenin at Ser552 to activate β-catenin (Lee et al., 2010; He et al., 2007; Nusse and Clevers, 2017). However, depletion of Erk1/2 in adult IECs did not show upregulation of Wnt/β-catenin signaling (de Jong et al., 2016), further suggesting that the MAPK/Erk pathway regulates Wnt signaling in a developmental stage-dependent manner. Our study is consistent with a recent report demonstrating that pharmacological inhibition of MEK/Erk activates the Wnt pathway and expands the stem cell population in colorectal cancer through inhibition of Egr1-mediated Axin1 expression (Zhan et al., 2019). These studies, including ours, suggest that inhibition of MAPK/ERK signaling eventually activates Wnt signaling in intestinal development and cancer through multiple mechanisms.

Deletion of Erk1/2 activated the Ras/Raf/Mek cascade owing to loss of Erk1/2-dependent negative feedback regulation, leading to transactivation of Akt/mTor signaling in the intestines. Activation of mTor inhibited secretory cell lineage differentiation due to increased Notch activity (Zhou et al., 2015). Using rapamycin to treat Erk IEC mice rescued secretory lineage differentiation. In addition, rapamycin treatment also restored intestinal architecture and significantly increased the survival rate of Erk^ΔIEC mice. Although we found the expression level of Ihh target genes was restored, Ihh expression was not recovered in Erk^ΔIEC mice after rapamycin treatment, suggesting that rapamycin activates the Hh pathway in mesenchymal cells through an unrevealed mechanism.

In summary, we demonstrate that Erk1/2 signaling is crucial for intestinal development and ISC homeostasis. Through our current study, we provide evidence that the ERK/MAPK pathway crosstalks with the Hh and Wnt pathways, suggesting that ERK1/2 are crucial factors that integrate cell fate decision signaling during intestinal development. Given that inhibition of MAPK in cancers usually activates multiple oncogenic pathways (Lake et al., 2016), our findings also demonstrate the complicated feedback loop operating during normal tissue development. Further elucidation of the exact molecular mechanisms of this feedback regulation will improve the understanding of intestinal homeostasis and tumorigenesis.

All mice were bred and maintained in specific pathogen-free conditions in the animal facilities of East China Normal University, Shanghai. Erk1^−/− and Erk2^flox/flox mice were a gift from Dr Hengyu Fan (Life Sciences Institute, Zhejiang University, China). Villin-Cre mice were obtained from the National Resource Center of Mutant Mice (NRCMM), and Lgr5-eGFP-IRES-CreERT2 mice were obtained from the Jackson Laboratory. For rapamycin treatment, 7-day-old mice were treated by intraperitoneal injection of 2 mg/kg/day for 28 days. All experiments conformed to the regulations drafted by the Association for Assessment and Accreditation of Laboratory Animal Care in Shanghai and were approved by the East China Normal University Centre for Animal Research (protocol number AR2013/12008).

The intestines were fixed in 4% paraformaldehyde, dehydrated and embedded in paraffin. For immunohistochemistry staining, after antigen retrieval, the sections were blocked with 1% BSA and incubated with primary antibody overnight at 4°C. After incubation with secondary antibody, sections were developed with DAB reagent, counterstained with Hematoxylin, dehydrated and mounted in neutral resin. Primary antibodies and their dilutions were as follows: p-Erk1/2 (1:100, Cell Signaling Technology, 4370); GFP (1:100, Santa Cruz Biotechnology, sc-101525); Ki67 (1:2000, NeoMarker, rm-9106); lysozyme (1:2000; Dako, A009902); Muc2 (1:50, Santa Cruz Biotechnology, sc-15334); chromogranin A (1:2000, Immunostar, 20085); Dclk1(1:100, Abcam, ab31704); BrdU (1:500, Abcam, ab8152); β-catenin (1:100, BD Transduction Laboratories, 610154); c-Myc (1:100, Santa Cruz Biotechnology, sc-40); Sox9 (1:500, Abcam, ab185966); p-4EBP1 (1:400, Cell Signaling Technology, 2855); p-S6 (1:400, Cell Signaling Technology, 4858); p-Akt (1:100, Cell Signaling Technology, 3787); Ihh (1:50, Santa Cruz Biotechnology, sc-13088); Ptch1 (1:100; Santa Cruz Biotechnology, sc-518102); p-Smad1/5/8 (1:100, Cell Signaling Technology, 13820); and α-SMA (1:50, Abcam, ab5694). For quantitative analysis, cell counting was performed in at least ten sections from three or more mice for each genotype.

A small piece of proximal duodenum was rinsed with ice-cold PBS and lysed in ice-cold lysis buffer. Protein extracts were quantified by Micro BCA Protein Assay Kit (Thermo Scientific) and resolved by SDS-PAGE. Primary antibodies were as follows: β-actin (1:5000, Sigma-Aldrich, A5441); Ihh (1:500, Santa Cruz Biotechnology, sc-13088); Bmp4 (1:500, Abcam, ab39973); Erk1/2, p-Erk1/2, Akt, p-Akt, S6K1, p-S6K1, 4EBP1, p-4EBP1, Egfr, p-Egfr, Ras, Ras-GTP, Raf, p-Raf, Mek1/2, p-Mek1(s217/221) and cyclin D1 (1:1000, Cell Signaling Technology, 4695, 4370, 4685, 3787, 2708, 9206, 9644, 2855, 4267, 3777, 3339, 8821, 9422, 9421, 9122, 9121, 2978, respectively).

Crypts were released from murine small intestine and mixed with 50 µl of Matrigel (BD Bioscience) in 96-well plates. 100 μl of basic culture medium containing growth factors (50 ng/ml EGF, 500 ng/ml R-spondin1 and 100 ng/ml Noggin, all from Peprotech) was added. Organoid growth was quantified by scoring the number of buds and Edu⁺ cells on microscopic images from at least 90 organoids from three or more mice for each genotype. For small-molecule treatment, the following inhibitors were used: MK2206 (500 nM; SelleckChem, HY-108232). Paraffin sections were generated and analyzed by immunofluorescence staining.

Total RNA was extracted from a small piece of proximal duodenum tissue using Trizol reagent (Invitrogen). DNA was synthesized with the Superscript RNase H2Reverse Transcriptase Kit (Takara) using 2.5 mM random hexamers (Takara). Real-time PCR was performed using the SYBR Green PCR Kit (Takara) on a Mx3005P thermal cycler (Stratagene) and actin was used as an internal control.

Tissue slides were deparaffinized, rehydrated and washed in PBS. After pretreatment with 0.2 N HCl and proteinase K, slides were post-fixed in 4% paraformaldehyde. Slides were washed and placed in fresh acetic anhydride solution then washed with PBS and rinsed in 5×SSC. Slides were placed in a box humidified with 5×SSC (pH 7.5)/50% formamide, then hybridization solution was added and the slides incubated in a box at 65°C for at least 1 h. The slides were supplied with hybridization solution containing 500 ng/ml of digoxigenin-labeled probe and incubated at 72°C for 48 h. The slides were then washed in SSC buffers and incubated with anti-Dig antibody overnight at 4°C. Sections were subsequently washed and colored with NCBI/NBT (Roche) for 12 h at room temperature in the darkness, then dehydrated and mounted with neutral resins.

Small intestinal tissue samples were fixed with 2.5% glutaraldehyde and post-fixed in 1% osmium tetroxide. For scanning EM, tissue was dehydrated and dried using the CO₂ critical point method. The dried samples were bonded to a metal sample stage and then sputtered onto a layer of metallic film in a vacuum evaporator. The villus structure was assessed by scanning EM.

Small intestinal samples were fixed with 2% paraformaldehyde with 2.5% glutaraldehyde and post-fixed in 1% osmium tetroxide in 100 mM phosphate buffer. Tissue was dehydrated, embedded in epoxy resin and sectioned for placement on TEM grids. Images were captured with a JEOL transmission electron microscope at 120 kV (model JEM-1210).

Small intestines of 14-day-old control and Erk^ΔIEC mice were removed, flash frozen on dry ice and RNA was harvested using Trizol reagent. Illumina TruSeq RNA Sample Prep Kit was used with 1 μg of total RNA for the construction of sequencing libraries. The RNA libraries were then sequenced on Illumina Hiseq 2000 platform and 125 bp paired-end reads were generated. RNA-seq data were separately aligned to the mouse reference genome mm10 using Tophat2 (version 2.1.0) and the expression levels of all genes were estimated by the program ‘htseq-count’ in HTSeq software (version 0.6.0). GSEA analyses presented in this article were performed by Java GSEA implementation (version 3.0). RNA-seq raw data and processed data of this project have been submitted to GEO under accession number GSE134249.

Statistical analyses were performed with Prism GraphPad software. Data are expressed as mean±s.e.m. Unpaired two-tailed Student's t-test or one-way ANOVA was used for statistical comparison of means, with P<0.05 defined as statistically significant. Unless indicated in the figure legends, all the experiments were performed at least three times with similar results.

We thank Dr Lingheng Li for providing us with the p-β-catenin ser552 antibody. We thank Dr Stefan Siwko in the Texas A&M University Health Science Center for scientific editing and comments. This work was supported by ECNU Public Platform for innovation (011).