ERK5 signalling rescues intestinal epithelial turnover and tumour cell proliferation upon ERK1/2 abrogation

The ERK1/2 MAPK signalling module integrates extracellular cues that induce proliferation and differentiation of epithelial lineages, and is an established oncogenic driver, particularly in the intestine. However, the interrelation of the ERK1/2 module relative to other signalling pathways in intestinal epithelial cells and colorectal cancer (CRC) is unclear. Here we show that loss of Erk1/2 in intestinal epithelial cells results in defects in nutrient absorption, epithelial cell migration and secretory cell differentiation. However, intestinal epithelial cell proliferation is not impeded, implying compensatory mechanisms. Genetic deletion of Erk1/2 or pharmacological targeting of MEK1/2 results in supraphysiological activity of the ERK5 pathway. Furthermore, targeting both pathways causes a more effective suppression of cell proliferation in murine intestinal organoids and human CRC lines. These results suggest that ERK5 provides a common bypass route in intestinal epithelial cells, which rescues cell proliferation upon abrogation of ERK1/2 signalling, with therapeutic implications in CRC.

T he extracellular signal-regulated kinases 1 and 2 (ERK1/2) are part of the classical family of mammalian mitogenactivated protein kinases (MAPKs), which also include three c-Jun amino-terminal kinases (JNK1/2/3), four p38 isoforms and its lesser-known counterpart, ERK5. The serine/threonine kinases ERK1 (MAPK3, also known as p44 MAPK) and ERK2 (MAPK1, also known as p42 MAPK) show 83% amino acid identity, are ubiquitously expressed and typically activated by growth factors and phorbol esters, whereas the p38 and JNK pathways are mainly activated by inflammatory cytokines and stress 1 . MAPKs are involved in regulation of mitosis, gene expression, cell metabolism, cell motility and apoptosis. ERK1/2 are activated by MEK1 and MEK2, which themselves are activated by Raf-1, A-Raf or B-Raf 1,2 . Ras proteins (K-Ras, H-Ras or N-Ras) are small GTPases that can be activated by receptor tyrosine kinases (RTKs) or G-protein coupled receptors (GPCRs), which recruit Raf proteins to the plasma membrane where they are activated. Together, these modules constitute the Ras-Raf-MEK-ERK pathway 3 .
The activation of ERK1/2 results in their nuclear translocation where they can phosphorylate a variety of nuclear targets such as Elk-1, c-Fos and c-Myc 1 , in addition to p90 ribosomal S6 kinases (p90RSKs) and mitogen-and stress-activated protein kinases, MSK1/2. The full repertoire of substrates for ERK1/2 consists of at least 160 cellular proteins 4 . These proteins are typically involved in the regulation of cell proliferation-more specifically, G1/S-phase cell cycle progression-and differentiation. However, their cellular effects are context-dependent and determined by the spatial and temporal dynamics of ERK1/2 activity 5 , which are highly regulated by scaffolding proteins and phosphatases 3,6,7 .
Despite vast literature on the role of ERK1/2 in cell proliferation, the absolute requirement of this signalling module in rapidly dividing tissues relative to other signalling pathways is unknown. The small intestinal epithelium is particularly suitable to address this question given the short (4-8 days) and dynamic life cycle of intestinal epithelial cells (IECs). Lgr5 þ intestinal stem cells at the intestinal crypt base produce transit-amplifying cells, which then undergo a number of proliferative cycles before terminal differentiation into absorptive enterocytes at the crypt-villus border. Enterocytes then migrate to the villus tip where they undergo anoikis and are shed into the gut lumen 8 . All of these cellular events are tightly coordinated by the Wnt, Notch, bone morphogenetic protein (BMP) and Hedgehog pathways 9 , whereas the roles of ERK1/2 remain to be charted. In the intestines, the ERK1/2 pathway is likely activated by autocrine and paracrine factors downstream of RTKs, such as epidermal growth factor receptor (EGFR) 10 , and by exogenous microbialderived substrates that signal through the Toll-like receptor (TLR)/MyD88 pathway 11 .
To study the effects of ERK1/2 in the adult intestinal epithelium, we generated mice with a conditional (IEC-specific) and tamoxifen-inducible deletion of Erk2 on the Erk1 À / À background, which completely abrogates this pathway. We show that the ERK1/2 signalling module, surprisingly, is dispensable for IEC proliferation. Genetic deletion of Erk1/2 in primary IEC or treatment of colorectal cancer (CRC) cell lines with MEK1/2 inhibitors results in compensatory activation of the ERK5 pathway. Moreover, the treatment of human CRC lines with a combination of MEK1/2 and ERK5 inhibitors is more efficacious in the inhibition of cancer cell growth. Thus, compensatory signalling by ERK5 suggests a potential rescue pathway that has clinical implications for targeted therapy in colorectal cancer.
ERK1/2 DIEC causes wasting and enterocyte dysfunction. After tamoxifen treatment, DIEC mice developed a moribund phenotype between 8 and 10 days of follow-up, whereas their fl/fl littermates remained healthy ( Supplementary Fig. 2a). This coincided with the clinical signs of malabsorption, including progressive body weight loss and 100% lethality of DIEC mice within 2 weeks of follow-up (Fig. 1b,c). Necropsy was performed at day 10 of follow-up of tamoxifen-treated DIEC mice and their fl/fl littermates to evaluate the underlying pathology. DIEC mice showed luminal distension of the small intestine and loss of abdominal fat, as well as a trend for shortening of the small intestine and colon (Fig. 1d,e and Supplementary Fig. 2b,c). Serum analysis at day 10 revealed that DIEC mice suffered from hypoalbuminaemia, hypoglycaemia, hypocalcaemia and decreased alkaline phosphatase levels when compared with fl/fl littermates ( Fig. 1f-h and Supplementary Fig. 2d,e). Together with steatorrhoea ( Supplementary Fig. 2f), these findings are indicative of malabsorption. DIEC mice also showed lymphocytopenia and increased levels of faecal albumin ( Supplementary Fig. 2g,h), consistent with a protein-losing enteropathy. These observations suggest a loss of absorptive surface in the small intestine and/or severe intestinal pathology. Of note, normal serum levels of critical electrolytes, such as Na þ , K þ and PO 4 3 À , were maintained in DIEC mice ( Supplementary Fig. 2i). In addition, the serum levels of lipids (triglycerides, cholesterol) were normal or even increased in DIEC mice ( Supplementary Fig. 2j). This hyperlipidaemia may be associated with loss of gut barrier function 16 , which also occurred in DIEC mice ( Supplementary Fig. 2h,k), leading to lipid accumulation in the intestinal lamina propria ( Supplementary  Fig. 3a). Since the Vil-Cre ERT2 model may lead to recombinase activity in the kidney 14 , which could explain some of the observed metabolic aberrations, we analysed kidney function in fl/fl and DIEC mice. We did not observe any microscopic differences in the parenchymal tissue indicative of kidney failure, nor increased blood urea nitrogen levels in DIEC mice ( Supplementary Fig. 2l,m), suggesting the observed phenotype was primarily caused by intestinal pathology.
Histological analysis showed that the crypt-villus morphology was severely perturbed in DIEC mice. This included signs of crypt elongation and villus shortening, more severely in the ileum compared with the jejunum and duodenum, respectively (Fig. 1i). Notably, at the most severely affected sites in the ileum, there was dilation of lacteals in villi, also known as intestinal lymphangiectasia ( Fig. 1i and Supplementary Fig. 3b), consistent with proteinlosing enteropathy. Even though alkaline phosphatase (AP) Δ Figure 1 | Wasting disease associated with malabsorption in Erk1/2 DIEC mice. (a) Immunofluorescent staining of paraffin-embedded sections from small intestine or colon taken from Erk1 À / À ;Erk2 fl/fl ('fl/fl') and Erk1 À / À ;Erk2 fl/fl ;Vil-Cre ERT2 ('DIEC') mice for an IEC marker, Claudin-1 (green) and ERK1/2 (red). Nuclear counter staining with Hoechst (blue). (b) Body weight changes of fl/fl and DIEC mice (n ¼ 6 per group) after five consecutive treatments with tamoxifen (1 mg day À 1 ). *Po0.05, **Po0.001 by analysis of variance. (c) Survival of fl/fl (n ¼ 10) and DIEC mice (n ¼ 9) after tamoxifen treatment (Kaplan-Meier). ARTICLE staining did not show differences by confocal fluorescence microscopy ( Fig. 1j and Supplementary Fig. 3c), detailed analysis of the brush border on enterocytes by electron microscopy revealed that microvilli of DIEC mice were short, thick and nonuniform (Fig. 1k), a hallmark of defective enterocyte differentiation. Together, these observations indicate a malabsorption and protein-losing enteropathy phenotype, resulting in rapid lethality of DIEC mice. These data suggest an essential role for the ERK1/2 signalling module in the orchestration of normal crypt-villus architecture, maintenance of the gut barrier and absorptive enterocyte maturation in the small intestine.
ERK1/2 DIEC has no major effects on colon homeostasis. Although efficient knockout was achieved in the colon by tamoxifen treatment in Erk1 À / À ;Erk2 fl/fl ;Vil-Cre ERT2 mice ( Fig. 1a and Supplementary Fig. 1b,c), no gross alterations were observed in the colon crypt architecture resulting from DIEC ( Supplementary Fig. 3d). More detailed analysis showed a loss of the luminal, flat epithelial sheet in the colon (Supplementary Fig. 3d-g), analogous to the villous changes in the small intestine. However, Erk1/2 deletion in the colon did not result in perturbation of goblet cell differentiation or an increased rate of apoptosis ( Supplementary Fig. 3h-k). Given the lack of a spontaneous phenotype in the colon in DIEC mice, we focused on the changes in physiology in the small intestine hereafter.
ERK1/2 DIEC leads to migration and differentiation defects. We hypothesized that the architectural derangement in the small intestine of DIEC mice might be explained by a migration defect on loss of epithelial ERK1/2 signalling. Consistent with this, we observed retention of AP-positive IECs in small intestinal crypts in DIEC mice, whereas AP immunoreactivity is normally not observed distal to the crypt-villus junction ( Fig. 1j and Supplementary Fig. 4a). BrdU pulse-chase experiments demonstrated a dramatic reduction in the migratory rate of epithelial cells in DIEC mice compared with fl/fl littermates (Fig. 2a,b). This defect appeared to involve mostly migration in the villous compartment, which was observed 424 h after BrdU labelling. Differentiation and maturation of epithelial cells from the secretory lineage-most prominently, goblet and Paneth cells-are directly correlated with their position along the crypt-villus axis. The long-lived Paneth cells make antimicrobial peptides and provide a stem cell niche at the crypt base; goblet cells produce mucus 17 . Less abundant secretory cell types include enteroendocrine cells that release a variety of hormones, and Tuft cells that produce eicosanoids 17 . These secretory cells, with the exception of Paneth cells, follow the migration pattern of absorptive enterocytes and share their fate at the villus tip. Indeed, we observed reduced numbers and an irregular distribution of Alcian blue-positive staining goblet cells in the small intestine of DIEC mice when compared with fl/fl littermates (Fig. 2c, Supplementary Fig. 4b-d). Paneth cells, which can be identified by positive staining for the granule protein matrilysin (also known as matrix metalloproteinase 7, MMP7), displayed more dramatic changes. Paneth cells in DIEC mice were mislocalized along the cryptvillus axis, showed reduced granule contents and had smaller secretory granules that were heterogeneous in size and found outside the apical cytoplasm ( Fig. 2d-f). MMP7 þ cells in the small intestine of DIEC mice also showed a more rounded morphology rather than their regular wedge-like shape in fl/fl control littermates, even though they expressed the prototypical Paneth cell marker Cryptdin (Fig. 2f,g). Alcian blue/MMP7 double staining showed an increased number of cells expressing the Paneth cell marker, with a relative loss of mucin positive cells ( Supplementary Fig. 4c,d). We hypothesized that these observations might be explained by a differentiation block in the secretory lineage, resulting in the accumulation of immature 'intermediate' cells-a precursor cell type that is normally rarely found in the lateral wall of the crypt base 18 . We did not observe any intermediate cells in the ileal sections from fl/fl mice, which showed the typical electron microscopic features of mature goblet and Paneth cells, respectively (Fig. 2h,i and Supplementary  Fig. 4e). In contrast, intermediate cells were evident in the ileal sections from DIEC littermates (Fig. 2h,i), with their goblet cells showing larger electrodense cores, and conversely, Paneth cells displaying smaller cores and increased amounts of free ribosomes in the cytoplasm ( Supplementary Fig. 4f). These data suggest that ERK1/2 signalling is critical for the migration of IEC along the crypt-villus axis, in addition to the full differentiation and maturation of IEC from the secretory cell lineage.
Abrogation of ERK1/2 results in enhanced ERK5 signalling. Despite the well-established role of MAPK ERK1/2 signalling in cell cycle progression, there were no signs of epithelial cell hypoproliferation in DIEC mice. In fact, we observed crypt hyperplasia and an increased number of Ki67 þ cells in the ileum (Fig. 3a,b and Supplementary Fig. 5a) and to a lesser degree in the colon ( Supplementary Fig. 5b,c) of DIEC mice compared with fl/fl littermates. This was associated with increased expression of the proliferation marker PCNA and cell cycle-associated protein cyclin D1 (Fig. 3c). Upregulation of cyclin D1 promotes progression through the G 1 /S-phase, thus indicating that on genetic ablation of ERK1/2, cell cycling in the intestinal epithelium remains intact. There was no direct evidence for upregulated Wnt/b-catenin signalling in IEC after ERK1/2 deletion in vivo ( Supplementary Fig. 5d), or after siRNA-mediated knockdown in TOPflash reporter cell lines in vitro (Supplementary Fig. 5e-g). We hypothesized that an alternative MAPK pathway may compensate for the loss of ERK1/2 signalling to drive the hyperproliferative phenotype of DIEC mice. We did not observe compensatory upregulation of p38 or JNK phosphorylation upon ERK1/2 abrogation (Fig. 3d). In contrast, constitutive levels of phospho-ERK5 were undetectable in fl/fl mice, whereas ERK1/2 abrogation resulted in a marked increase of ERK5 phosphorylation (Fig. 3d). Loss of constitutive phosphorylation of p90RSK confirmed sustained abrogation of signalling downstream of the ERK1/2 module in DIEC mice, despite the concomitant ERK5 activation (Fig. 3d).
To further study the interactions between the ERK1/2 and ERK5 pathways, we used human colorectal cancer (CRC) cell lines HCT116 and DLD-1 that harbour heterozygous KRAS G13D (gain-of-function) mutations. Treatment of HCT116 cells with the specific MEK1/2 inhibitor, PD0325901, resulted in the upregulation of p-ERK5 levels in a time-and dose-dependent manner (Fig. 3e,f), suggesting compensatory ERK5 activation upon inhibition of the ERK1/2 pathway. Specificity of p-ERK5 immunoblotting was demonstrated by reversal of the signal by co-treatment with an ERK5 inhibitor, XMD8-92 (Fig. 3f). Similar results were achieved with the DLD-1 cell line (Fig. 3g,h), as well as Caco2 cells (Fig. 3i,j). The latter cell line is KRAS wildtype, which suggests that upregulation of ERK5 signalling occurs independently of KRAS mutation status. These data are consistent with the upregulated p-ERK5 levels we observed in DIEC mice that do not harbour oncogenic mutations. Since abrogation of the Raf-MEK-ERK module can result in feedback activation of EGFR 19 , which could then be responsible for supraphysiological activity of the ERK5 pathway 20,21 , we tested whether EGFR inhibition prevented ERK5 upregulation. However, treatment of DLD-1 or Caco2 cells with MEK1/2 inhibitor in the presence of the EGFR inhibitor, erlotinib, still resulted in enhanced p-ERK5 levels ( Supplementary Fig. 5h). These data suggest that the compensatory upregulation of the ERK5 pathway upon abrogation of ERK1/2 signalling takes place through an EGFR-independent mechanism. ERK1/2 and ERK5 are interconnected signalling pathways. We addressed the role of the ERK1/2 versus ERK5 signalling modules on cell proliferation in primary IEC cultures. To exclude any indirect effects by stromal, hematopoietic or neuronal cells on  IECs, we used the small intestinal organoid culture system 22 .
Intestinal organoids were generated from untreated DIEC mice and, after purification and several passages, treated with tamoxifen or its vehicle (ethanol) for two consecutive days in vitro. This protocol resulted in rapid and robust (498%) knockout of Erk2 transcripts in tamoxifen-treated organoids (Fig. 4a,b and Supplementary Fig. 6a-c). Genetic ablation of Erk1/2 did not interfere with crypt budding or cell proliferation in intestinal organoids, analogous to our observations in vivo, but resulted in morphological changes ( Supplementary Fig. 6a,d), possibly related to migration and differentiation defects. We evaluated the requirement for ERK5 signalling in Erk1/2 DIEC organoids for cell proliferation. Treatment with tamoxifen or with ERK5 inhibitor, XMD8-92, alone did not prevent organoid growth, whereas co-treatment resulted in the disappearance of budding crypts and a marked loss of expression of proliferation  marker Ki67 (Fig. 4c,d and Supplementary Fig. 6f). Genetic deletion of Erk2 was confirmed by quantitative PCR (Q-PCR) analysis ( Supplementary Fig. 6e). Expression of the ERK1/2 target gene c-Fos was abrogated by tamoxifen treatment but increased upon ERK5 inhibition, suggesting compensatory upregulation; this was abolished upon co-treatment with XMD8-92 (Fig. 4e). Furthermore, targeting ERK1/2 and ERK5 together inhibited expression of the intestinal stem cell marker, Lgr5, and the Paneth cell marker, Mmp7 (Fig. 4f). Both the cell types are crucial for the regenerative potential of the intestinal crypt. These data suggest ARTICLE that ERK1/2 and ERK5 act as interconnected signalling pathways that maintain the self-renewal capacity of the intestinal epithelium.
To confirm these findings in IEC with genetic deletion of Erk5, intestinal organoids were generated from Erk5 fl/fl mice 23,24 . Established organoid lines were then transduced with retroviral Puro-Cre ERT2 or lentiviral Puro-Cre, followed by puromycin selection. Tamoxifen treatment of these transduced intestinal organoids resulted in B50% reduction of Erk5 mRNA levels (Fig. 5c). The deletion of Erk5 did not result in enhanced levels of p-ERK1/2 in intestinal organoids cultures (Supplementary Fig. 6g). In accordance with the aforementioned data, we found that tamoxifeninduced reduction of Erk5 expression in combination with PD0325901 treatment was more efficient in abrogating cell proliferation when compared with either condition alone (Fig. 5a-d). Targeting ERK1/2 and ERK5 inhibits tumour cell proliferation. Given the functional interconnection between the ERK1/2 and ERK5 pathways in IEC, we next evaluated the effects of targeted inhibition of both signalling modules on tumour cell proliferation.
In a two-dimensional growth model, we found that treatment of the colorectal cancer cell line, HCT116, with the MEK1/2 inhibitor, PD0325901, was more effective in suppressing cell proliferation in combination with ERK5 inhibitor, XMD8-92, compared with either treatment alone (Fig. 6a,b). Similar results were achieved by measuring cells in S/G2/M-phase by propidium iodide labelling after treatment with DMSO, MEK1/2 inhibitor alone, ERK5 inhibitor alone or the combination of both inhibitors, respectively ( Supplementary Fig. 7a,b). These results were confirmed in the other CRC cell lines, including DLD-1, HT-29 and SW480 ( Supplementary Fig. 7c,d). Of note, siRNA-mediated knockdown of ERK1/2 and/or ERK5 validated these MAPKs as drug targets in CRC, although cell lines were relatively more susceptible to ERK1/2 knockdown ( Supplementary Fig. 7e,f). We then used a threedimensional growth model with intestinal organoids generated from Apc fl/fl mice 25 , that were transduced with Adeno-Cre in vitro (referred to hereafter as 'Apc À / À '). Intestinal organoids with a somatic deletion of tumour suppressor APC display upregulated Wnt signalling activity, which results in a rounded, cyst-like morphology, distinct from wild-type organoids 26 ( Supplementary  Fig. 7g). We have previously shown that ERK1/2 signalling is required for Wnt-driven adenoma formation in vivo 11 . Furthermore, intestinal organoids that are APC deficient showed upregulated levels of p-ERK1/2 and p-ERK5 compared with WT organoids (Fig. 6c). In line with this, PD0325901 treatment inhibited organoid growth and the expression of proliferation marker Mki67 and oncogenes c-Myc and Fra1, which was also observed with XMD8-92 treatment, with the combination treatment being superior to either treatment alone (Fig. 6d,e). Synergistic, specific or antagonistic effects of the MAPK inhibitors were observed on the expression of stem cell markers Lgr5 and Ascl2, Paneth cell marker Mmp7 and immediate-early gene Egr1, respectively ( Supplementary Fig. 7h). Confirmation of on-target effects of the inhibitors was confirmed by the analysis of c-Fos expression and immunoblotting for MAPK proteins (Fig. 6e, Supplementary Fig. 7i). Notably, we observed that ERK5 protein levels were increased on long-term MEK1/2 inhibitor treatment, whereas ERK5 inhibitor treatment resulted in reduced ERK5 protein levels ( Supplementary Fig. 7i), suggesting antagonistic interactions between the two MAPK modules. Suppression of total ERK5 levels by long-term treatment with XMD8-92 was also observed in multiple CRC lines ( Supplementary Fig. 7j). Together, these data suggest that ERK1/2 and ERK5 are major, interrelated transducers of extracellular mitogenic signals. Next, we assessed the effects of MAPK inhibitors on the growth of Apc À / À organoids that express mutant KRAS G12V by lentiviral transduction. These organoids showed marked morphological differences compared with both wild-type and Apc À / À organoids ( Supplementary Fig. 7g), suggesting abnormal cellular homeostasis, reminiscent of oncogenic KRAS signalling in human CRC. Mutant KRAS expression appeared to preferentially activate ERK1/2 compared with ERK5 in organoids on the Apc À / À background (Fig. 6c). Treatment with MAPK inhibitors showed that only the combination of MEK1/2 and ERK5 inhibitors reversed the typical morphological changes induced by oncogenic KRAS on the Apc À / À background (Fig. 6f). Interestingly, treatment with ERK5 inhibitor was superior over MEK1/2 inhibitor with regard to suppression of proliferation and expression of the oncogene c-Myc, stem cell markers Lgr5 and Ascl2 and Paneth cell marker Mmp7 (Fig. 6g, Supplementary  Fig. 7k), suggesting differential effects of abrogating MEK1/2 versus ERK5 signalling depending on KRAS activity in threedimensional growth. The direct effects of MEK1/2 and ERK5 inhibition were validated by the evaluation of c-Fos expression and immunoblotting for MAPK proteins (Fig. 6g, Supplementary  Fig. 7l). Thus, these data suggest that pharmacological inhibitors of ERK5 can be exploited to target Wnt-driven intestinal tumour cell proliferation, and may be superior over MEK1/2 inhibitors in CRC tumours depending on KRAS mutation status.

Discussion
Here we show that ERK1/2 signalling in mouse intestinal epithelium is dispensable for cell proliferation, while it resulted in abnormal differentiation of enterocytes, wasting disease and ultimately lethality (Fig. 1). Consistent with these findings, ERK1/2 MAPKs were shown to be associated with the enterocyte brush border and activated upon RTK stimulation or feeding 27 or electrical field stimulation in polarized epithelium 28 . This seems at odds with literature that suggest that maintained ERK1/2 signalling precludes enterocyte differentiation 29,30 . A possible explanation for this discrepancy could be that cycling IEC in the transit amplifying zone of the crypt require relatively high levels of active ERK1/2, which is readily blocked by pharmacological intervention, whereas a transition to low level ERK1/2 activity in IEC migrating into the villus compartment promotes the absorptive enterocyte differentiation program that is only perturbed upon complete genetic deletion of Erk1/2. Little is known about the role of ERK1/2 signalling in the life cycle of secretory cells in the gut. A recent report by Heuberger et al. 15 described that IEC-specific deletion of non-receptor tyrosine phosphatase, Shp2, resulted in the loss of p-ERK1/2 levels in the small intestine. This coincided with an increased number of Paneth cells at the expense of goblet cells in the small intestine, as well as shortening of villi. They also observed the strongest staining for epithelial p-ERK1/2 in the TA zone. This p-ERK1/ 2 þ staining pattern and the architectural organization of the TA zone were lost in Shp2 knockout mice. Interestingly, the deleterious effects of Shp2 deficiency were rescued by expression of constitutively active MEK1. A model was proposed in which the balance between Wnt/b-catenin and MAPK signalling determines Paneth cell versus goblet cell differentiation, respectively 15 . This proposed crucial role for ERK1/2 MAPK signalling in intestinal secretory cell differentiation is consistent with our observations in ERK1/ 2 DIEC mice. Migration and differentiation are functionally intertwined in the intestines, as demonstrated by the immature phenotype of mislocalized Paneth cells observed in DIEC mice (Fig. 2). Critical to epithelial cell migration is proper cytoskeleton reorganization mediated by the small GTPases of the Rho family, cell polarization regulated by Cdc42 and dynamic adhesion through cell-matrix and cell-cell interaction via integrin/FAK/Src signalling 31 . The ERK1/2 module is used as a downstream effector of many of these pathways in the intestine, including Rho GTPases 32 , FAK 33 and Src 34 , and has been suggested to promote cell motility 33,35 . RTK signalling also contributes to cell migration, for example, Eph-Ephrin receptor interactions are crucial for correct positioning of Paneth cells 36 . Ephrin receptorinduced epithelial cell migration has been shown to be mediated by Src and ERK1/2 activation 37,38 , which may explain the Paneth cell mislocalization observed in DIEC mice. In summary, the ERK1/2 module is indispensable for full maturation of both absorptive enterocytes and the secretory lineage (Fig. 7a), confirming its crucial role in the integration of cellular cues required for determination of epithelial cell fate.  Figure 7 | Roles of ERK1/2 and ERK5 in intestinal homeostasis and tumorigenesis. (a) When the ERK1/2 pathway is intact, extracellular cues that are transduced via RTKs or GPCRs activate Ras under physiological conditions, or alternatively, Ras is constitutively active in colorectal cancer (RasD*), which preferentially activates the Raf-MEK1/2-ERK1/2 module. The nuclear and transcriptional targets of ERK1/2 are crucial for enterocyte and secretory cell differentiation, IEC migration, as well as cell proliferation under homeostatic and oncogenic conditions. Importantly, ERK1/2 activation also results in the activation of negative feedback mechanisms that suppress its upstream kinases (for example, RTKs, son of sevenless, Raf) and activate dual specificity phosphatases (DUSPs), resulting in the silencing of the ERK5 module. (b) Upon abrogation of MEK1/2 or genetic knockout of Erk1/2, the lack of negative feedback mechanisms (that is, feedback activation) results in upregulation of the Ras-Raf-MEK5-ERK5 module, which maintains cell proliferation under physiological conditions, or results in continued tumour cell proliferation in colorectal cancer, respectively. However, the lack of activation of ERK1/2-specific targets results in differentiation and migration defects of intestinal epithelial cells culminating in malabsorption, wasting disease and mortality. Compensatory upregulation of the ERK5 pathway in CRC can be reversed by targeted treatment with its specific inhibitor, XMD8-92. NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11551 ARTICLE An unexpected finding was the redundancy of ERK1/2 in the gut with regard to cell proliferation. Erk1/2 deletion was compensated by upregulated ERK5 signalling. Genetic targeting of ERK1/2 in vitro previously showed that Erk2 knockdown is more effective than Erk1 knockdown in suppressing cell proliferation, although this may be related to higher expression levels of the former 39 . The effect of gene dosage was demonstrated in vivo by the observations that while Erk1 À / À mice are viable 12 and Erk2 À / À mice die in utero 13 , Erk2 þ / À mice are only viable when at least one copy of Erk1 is present. However, mice heterozygous ( þ / À ) for both Erk1 and Erk2 alleles were born at lower than Mendelian ratio 39 . More recently, it was reported that transgenic expression of ERK1 can compensate for Erk2 deletion 40 , demonstrating functional redundancy between both family members. Deletion of Erk1/2 in adult skin tissue resulted in hypoplasia, which was associated with G2/M cell cycle arrest, without notable differentiation defects of keratinocytes 41 . These data differ from our observations in the intestines, which might be explained by incomplete and transient siRNA-mediated knockdown of ERK1/2 in primary keratinocyte cultures 41 , compared with more efficient genomic deletion of Erk1 and Erk2 that is typically achieved by the Villin-Cre-ERT2 system 14 , possibly resulting in different outcomes.
Both ERK1/2 and ERK5 have been described to promote cell cycle progression, although they have different upstream signalling partners, MEK1/2 and MEK5, respectively 1 . Furthermore, ERK2 and ERK5 proteins share only about 66% sequence identity, and MEK5 is phosphorylated by MEKK2/3, which can also activate the p38 and JNK pathways 42 . The ERK5 pathway is classically activated by stress stimuli, in addition to mitogens; thus, it shares features of both the ERK1/2, and p38 and JNK pathways, respectively 43 . ERK5 induces expression of cyclin D1 (refs 44,45), and suppresses expression of cyclin dependent kinase inhibitors 46 , thereby promoting G1/S-phase cell cycle progression. Importantly, the role of ERK5 in IEC differentiation and intestinal homeostasis is currently unknown. Knockout of Erk1/2 in IEC induced activity of ERK5, which was not detectable in naive mice (Fig. 3). These data suggest that the ERK1/2 and ERK5 modules may share proximal signalling components. Although EGFR is a likely candidate in this context 19,20 , we found that abrogation of EGFR signalling did not prevent enhanced ERK5 activity upon MEK1/2 inhibition. Although it was originally suggested that ERK5 signalling is independent of Ras 20 , other groups established that Ras, either through physiological signalling 47 , or by its oncogenic activity 48,49 , activates the MEK5-ERK5 signalling axis. Thus, rewiring of signalling networks downstream of Ras could explain the supraphysiological activity of ERK5 upon conditional deletion of Erk1/2 in the intestines. In fact, it has been shown that ERK1/2 signalling mediates negative feedback on ERK5 activity 50 , possibly through transcriptional activation of dual specificity phosphatases (DUSPs) 51 . Alternatively, ERK1/2-induced FOS-like antigen 1 (Fra-1) may negatively regulate MEK5 (ref. 52). These data suggest that ERK5 is a default bypass route downstream of RTK-Ras and activated upon loss of ERK1/2-mediated repression, thereby ensuring the transduction of mitogenic signals to the nucleus (Fig. 7b). Consistent with this concept, we found that ERK5 inhibition induces atrophy of DIEC intestinal organoids (Fig. 4). In addition, important downstream transcriptional targets of ERK5 and ERK1/2 overlap, such as immediate-early gene Fra1 and oncogene c-Myc, whereas c-Fos and Egr1 were specifically induced by ERK1/2 ( Fig. 6 and Supplementary Fig. 7). Specificity of ERK1/2 over ERK5 and other MAPK family members for the activation of c-Fos has been previously described 53 , demonstrating their differential biological output despite the shared ability to transduce potent mitogenic signals.
Our findings may be relevant for the use of MAPK inhibitors in the treatment of colorectal cancer. Although there was only a mild phenotype in the colons of DIEC mice under homeostatic conditions, the Ras-RAF-MEK-ERK pathway is generally upregulated in malignant cells including CRC 54 . Targeted therapy typically results in feedback activation of upstream players of the targeted kinase, which are then able to reactivate the same pathway or utilize bypass signalling routes 55 . For example, on activation, ERK1/2 phosphorylates EGFR, son of sevenless 56 , and Raf 57 , thereby terminating upstream signalling activity. Knockout of Erk1/2 eliminates this negative feedback. Our data suggest that ERK5 is a putative resistance pathway in the context of targeted treatment with MEK1/2 or ERK1/2 inhibitors (Fig. 7b). Different classes of MEK1/2 inhibitors display various modes of resistance to therapy (innate, adaptive and acquired) 58 . Since we have only used one MEK1/2 inhibitor (PD0325901) in our studies, it will be necessary to evaluate other classes of inhibitors in combination with ERK5 inhibitors. Importantly, while treatment with either the MEK1/2 or ERK5 inhibitor suppressed tumour growth in murine Apc À / À organoids, only the latter was able to inhibit the proliferation of Apc À / À ;KRAS G12V organoids (Fig. 6), which are more representative of human CRC. In line with this, suppression of ERK5 expression by forced expression of miR-143/145 inhibited intestinal adenoma formation in the Apc Min/ þ model 59 , and activated MEK5 correlated with more invasive CRC in human 60 . ERK5 has been previously reported to mediate resistance to cytotoxic chemotherapy-induced apoptosis 61 . The highly specific and bioavailable ERK5 inhibitor, XMD8-92, has shown antitumour effects in multiple preclinical cancer models by inhibiting tumour angiogenesis, metastasis and chemoresistance 62 . Furthermore, ERK5 inhibition does not induce feedback activation of upstream or parallel signalling pathways 62 .
In conclusion, the ERK1/2 and ERK5 MAPK modules display a high degree of signalling plasticity in the intestinal epithelium, which has implications for targeted treatment of colorectal cancer.
Cell culture and siRNA treatments. HCT116, DLD-1, Caco2, HT-29 and SW480 cell lines were obtained from ATCC and cultured in high-glucose DMEM (Mediatech), supplemented with 4 mM glutamine, 50 U ml À 1 penicillin, 50 mg ml À 1 streptomycin (all from Invitrogen) and 10% fetal calf serum. The cell lines were routinely tested for mycoplasma contamination with the MycoAlert Mycoplasma Detection Kit (Lonza). For cell line authentication, short tandem repeat (STR) analysis was performed on isolated genomic DNA with the GenePrint 10 System from Promega, and peaks were analysed using GeneMarker HID from Softgenetics. Allele calls were searched against STR databases maintained by ATCC (www.atcc.org), DSMZ (www.dsmz.de) and Texas Tech University Children's Oncology Group (cogcell.org). For some experiments, the cells were pre-treated with PD0325901, XMD8-92 or DMSO (control) before western blotting or cell viability assays. For siRNA-mediated knockdown, the cells were plated at 5 Â 10 3 cells per well in 96-well plates and transfected with 40 nM siRNA with optimized Dharmacon transfection reagents (Thermo). Oligonucleotides used: siGENOME non-targeting Control siRNA #5, siCONTROL TOX, ON-TARGETplus SMART-POOL human CTNNB1 (all from Thermo), SignalSilence Erk1/2 (6560) and SignalSilence ERK5 siRNA I (7301) (Cell Signaling Technologies).
Cell viability and reporter assays. XTT assays were performed according to the manufacturer's instructions (Biotium). In short, HCT116, DLD-1, HT-29 or SW480 cells were plated at 2-4 Â 10 3 cells per well (n ¼ 8 per condition) in 96-well plates (Corning) and XTT assays were performed on five consecutive days. For TOPflash luciferase reporter assays, HCT116 or SW480 cells were co-transfected with the TCF Reporter Plasmid (Millipore) and pcDNA3.1( þ ) vector (Thermo Scientific) that contains a Geneticin-selectable marker. Stable lines were selected with 800 mg ml À 1 Geneticin (Thermo). The cells were lysed in total cell lysis buffer and Luciferase substrate (Promega) was added, followed by measurement of bioluminescence.
Mice. Eight-to-12-week-old mice on the C57BL/6 background and bred in the animal facility of the Stein Clinical Research building at the University of California, San Diego were used for all the experiments, except for Erk5 fl/fl mice that were solely used for the generation of ex vivo intestinal organoids. Age-matched male and female mice were used for experiments. To generate the Erk1 À / À ;Erk2 fl/fl ;Vil-Cre ERT2 mouse strain, Erk1 À / À ;Erk2 fl/fl mice (a gift of Dr S. Hedrick, UCSD) 13 were intercrossed with Vil-Cre ERT2 mice that express the CRE transgene under control of the IEC-specific Villin promoter, which is inducible by tamoxifen 14 . Genotyping of the Erk1 and Erk2 genes, as well as the CRE transgene, with gDNA obtained from intestinal epithelial cells, skin and spleen tissue was used to confirm the genotype of the mice. Erk1 À / À ;Erk2 fl/fl ;Vil-Cre ERT2 mice were fertile and viable, and did not shown any overt phenotype without tamoxifen treatment. To induce recombination of the Erk2 gene (deletion of exon 3) in IEC, the mice were treated with tamoxifen (1 mg day À 1 ) dissolved in safflower oil for five consecutive days by oral gavage. The Erk5 fl/fl mice were generated by Dr C. Tournier (University of Manchester, Manchester, UK) 23,24 , and transferred from Dr Z. Xia's laboratory (University of Washington, Seattle, WA, USA). Gut barrier assays were performed with FITC-Dextran or by faecal albumin measurements with the Mouse Albumin ELISA kit (Bethyl Laboratories), as previously described 63 . All the experimental procedures were conducted in accordance with the University of California, San Diego institutional guidelines for Animal Care and Use (IACUC).
Mouse blood analysis. Analysis of blood chemistry from blood plasma (albumin, globulin, total protein, liver enzymes, glucose, electrolytes, ALP, lipid spectrum, blood urea nitrogen) and haematological parameters (lymphocyte counts, haemoglobin) from EDTA blood were performed by the STAT Veterinary Lab, San Diego, CA, USA.
IEC isolation. IEC isolation was performed as previously described 64 . Briefly, the intestines were washed three times in PBS with 1 mM dithiothreitol, followed by incubation in HBSS with 5 mM EDTA and 0.5 mM dithiothreitol and harvest of epithelial cells.
RNA extraction and Q-PCR. For the preparation of RNA samples from murine IEC, the cell lysates were prepared with RLT buffer after IEC isolation procedures and stored at À 80°C until extraction. For RNA isolation from small intestinal organoid cultures, purified IEC were recovered from Matrigel using the BD Cell Recovery Solution following the manufacturer's instructions and cell lysates were prepared with RLT buffer. RNA extraction was performed with the RNeasy Mini Kit (Qiagen), followed by cDNA synthesis with the qScript cDNA superMix kit (Quanta Biosciences). Q-PCR was performed on the AB7300 Real-Time PCR System (Applied Biosystems) using PerfeCta SYBR Green FastMix (Quanta Biosciences). Oligonucleotide sequences for Q-PCR were custom designed with NCBI Primer-BLAST and synthesized by IDT Technologies (see Supplementary  Table 1). For selected experiments, PCR products were run in TAE buffer on 2% agarose gels and visualized with SYBR Safe DNA (Invitrogen).
Immunohistochemistry and in situ hybridization. Immunohistochemistry was performed using standard staining procedures 64 . Briefly, 4-6-mm paraffin sections from murine small or large intestinal tissues were incubated with 1:200 dilutions of anti-Ki67 or anti-MMP7 antibodies (overnight, 4°C), followed by detection with biotinylated secondary antibodies (1 h at room temperature; 1:500 dilution) and horseradish peroxidase-streptavidin conjugates (Jackson; 30 min at room temperature), followed by the addition of 3,3 0 -diaminobenzidine (DAB; Vector Laboratories). Counterstaining was performed with haematoxylin 560 (Surgipath). Alcian Blue (American MasterTech) and alkaline phosphatase (Vector Laboratories) stainings were performed according to the manufacturer's instructions. Oil red O stainings were performed by treating frozen sections with 60% isopropanol, followed by staining with Oil red O for 15 min, followed by rinsing with 60% isopropanol and haematoxylin counterstaining. For BrdU pulsechase experiments, the mice were injected with BrdU (2 mg intraperitoneally) at 2, 24 or 48 h before analysis. The BrdU In-Situ Detection Kit (BD) was used for the detection of BrdU-labelled cells according to the manufacturer's instructions. Quantification of BrdU þ cells along the crypt-villus axis was performed following published methods 11,68 . In situ hybridization was performed to detect Cryptdin-1 expression to visualize Paneth cells in small intestinal sections, as previously described 69 .
TUNEL assay. Performed as previously described 11 following the manufacturer's instructions (BD Biosciences).
Flow cytometry. Flow cytometry analysis of intestinal epithelial cells was performed using standard protocols 11,70 . Propidium iodide (PI) stainings were performed with the FxCycle PI/RNase Staining Solution (Molecular Probes) following the manufacturer's instructions. Briefly, the cells were fixed with 4% paraformaldehyde and prepared at 1 Â 10 6 cells per condition. The cells were pelleted, followed by the addition of 0.5 ml of PI/RNase Staining Solution. After a 15-min incubation period, the samples were run on a BD Accuri C6 flow cytometer.
Transmission electron microscopy. Transmission electron microscopy (TEM) was performed by using a Zeiss EM10C electron microscope with a Gatan 785 11 megapixel digital camera at the Microimaging Core Facility (VA San Diego Healthcare System). Mouse small intestinal samples were fixed with 2% paraformaldehyde with 2.5% glutaraldehyde, embedded in epoxy 812 and sectioned with a Reichert Ultracut ultramicrotome for placement on TEM grids.
Statistical analysis. The data are represented as mean ± s.e.m. or as stated in the figure legends. The P values are stated in the figure legends; a P value o0.05 was considered significant. Unpaired Student's t-test was used for statistical analyses to compare two data sets with normal distribution, Mann-Whitney U-test was used for nonparametric data, analysis of variance was used to compare multiple data sets, log-rank analysis was applied for survival curves. All the statistical tests were performed with Prism 6.0 (GraphPad, La Jolla, CA, USA).