miR-200c regulates FGFR-dependent epithelial proliferation via Vldlr during submandibular gland branching morphogenesis

MicroRNAs (miRNAs) are small non-coding RNAs that target multiple messenger RNAs (mRNAs) and function as post-transcriptional regulators of gene expression. They have emerged as key regulators of proliferation (Wang, S. et al., 2008), apoptosis (Fujita et al., 2009), maintenance of progenitor cell identity (Marson et al., 2008) and embryonic stem cell fate (Tiscornia and Izpisua Belmonte, 2010). During development, miRNAs regulate skin (Yi et al., 2006), blood vessel (Wang, S. et al., 2008) and neuronal (Yoo et al., 2009) morphogenesis, and the expression of multiple miRNAs in the developing lung (Wang et al., 2007) and submandibular gland (SMG) (Jevnaker and Osmundsen, 2008; Hayashi et al., 2011) suggests that they regulate gene expression associated with development; however, the functional characterization of miRNAs during epithelial branching morphogenesis remains to be elucidated.

miR-200c is among the miRNAs implicated in disease, and a reduction in miR-200c expression is associated with multiple cancers that have invasive cell phenotypes (Burk et al., 2008). A direct target of miR-200c is Zeb1, a transcriptional repressor of E-cadherin in epithelial cells (Gregory et al., 2008). A decrease in miR-200c expression increases Zeb1, which downregulates E-cadherin, thus increasing epithelial-to-mesenchyme transition (EMT) during tumor transformation (Gregory et al., 2008). Conversely, decreasing Zeb1 expression upregulates E-cadherin and inhibits EMT (Park et al., 2008). Zeb1 also promotes tumor formation by reducing the expression of miR-200c, which cooperates to suppress stem cell factors in cancer. Additionally, miR-200c strongly suppresses cell proliferation of human breast cancer stem cells during tumor initiation in vivo (Shimono et al., 2009). Therefore, an inverse correlation between miR-200c expression and cell proliferation occurs during cancer formation, but it is not known whether this also occurs during normal organ morphogenesis.

A connection between miR-200c and low density lipoprotein receptors (LDLRs) has not been previously described. LDLRs bind to apolipoproteins that regulate vasculogenesis (Oganesian et al., 2008) and lipid metabolism in adipocytes (Yin et al., 2008) and hepatocytes (Sanderson et al., 2010). Among the LDLRs, very low density lipoprotein receptor (Vldlr) and apolipoprotein-E receptor 2 (ApoER2; also known as Lrp8) bind to the extracellular matrix protein reelin (Reln), and induce neuronal migration and layering of the cerebral cortex (Trommsdorff et al., 1999; Cariboni et al., 2005) via PI3K (Jossin and Goffinet, 2007) and ERK1/2 phosphorylation (Webb et al., 1999). Knockout mice for Vldlr, Reln or Dab1, which encodes the intracellular adaptor protein that mediates reelin/Vldlr signaling, are viable with similar neurological phenotypes, and the deletion of Reln or Dab1 was recently reported to decrease terminal branching, duct elongation and luminal organization during mammary morphogenesis (Khialeeva et al., 2011). However, the mechanisms that regulate Vldlr function during epithelial morphogenesis remain to be elucidated.

Epithelial organs such as SMGs develop by branching morphogenesis, which requires coordinated proliferation and differentiation of the epithelium in order to form a complex branched structure (Patel et al., 2006; Walker et al., 2008). During this process, proliferation is localized predominantly in the epithelial end buds and is induced by multiple signaling pathways, including muscarinic receptors (Knox et al., 2010), FGFRs (Hoffman et al., 2002), EGFRs (Steinberg et al., 2005; Rebustini et al., 2009) and PDGFRs (Yamamoto et al., 2008). By contrast, less is known about the mechanisms that restrict or decrease epithelial proliferation at discrete locations and stages of development so that normal branching morphogenesis can occur.

We have discovered that miR-200c decreases FGFR-dependent epithelial proliferation via regulation of Vldlr function during SMG branching morphogenesis. Our observations provide mechanistic insight into how miR-200c counterbalances proliferation during normal epithelial morphogenesis, and unveil an unexpected role for its target genes, Vldlr and Reln, in regulating FGFR-dependent epithelial proliferation during branching morphogenesis.

SMGs were harvested from embryonic day (E) 12 or E13 ICR mice (Harlan, Frederick, MD, USA) and the epithelia were separated from the mesenchyme and cultured as previously described (Rebustini and Hoffman, 2009). The epithelia were cultured with 400 ng/ml Fgf10 and 100-500 ng/ml reelin (both from R&D Systems, Minneapolis, MN, USA), function-blocking Vldlr antibody (R&D Systems, 0.5 μM) (D’Arcangelo et al., 1999; Sinagra et al., 2005) and 0.5 μg/ml heparan sulfate (Sigma, St Louis, MO, USA). Dab1^exKIneo mice have been described previously (Khialeeva et al., 2011).

Total RNA and the Megaplex Pools protocol and reagents (Applied Biosystems, Foster City, CA, USA) were used to detect miRNA expression. Briefly, cDNA was synthesized using total RNA and Megaplex RT, followed by a pre-amplification step and PCR detection using TaqMan low density array (TLDA, miRNA mouse pool A, Applied Biosystems) containing primers for 377 mouse miRNAs and the housekeeping gene snRNA-U6. The data were analyzed using SDS Database System 2.3 (Applied Biosystems), and RealTime StatMiner software (Integromics, Philadelphia, PA, USA). Relative fold change in miRNA expression was calculated comparing the cycle threshold (CT) numbers of each detected miRNA with snRNA-U6, and comparing differences in CTs between the E13 SMG epithelium and mesenchyme, and the E13 SMG epithelial end buds and main duct. A fold change of 2.0 was defined as a threshold for significant change in expression, and miRNAs detected with CT values above 32 were defined as undetectable.

The miRCURY LNA-based control, the miR-34a, miR-200c, miR-204 and miR-135a antagomir inhibitors (Exiqon, Vedbaek, Denmark), and the miScript miR-200c mimic (Qiagen, Valencia, CA, USA) were transfected at 400 nM in E12 SMG organ cultures and in E13 SMG epithelia using RNAiFect (Qiagen), as previously described (Rebustini et al., 2009). DNase-free total RNA was prepared using micro-RNAqueous and mirVana miRNA isolation kits (Ambion, Austin, TX, USA).

Total RNA samples were DNase treated (Ambion) prior to cDNA synthesis using iScript reagents (BioRad, Hercules, CA, USA). SYBR-Green qPCR was performed using 1 ng cDNA and primers designed using Beacon Designer software (supplementary material Table S1). Melt-curve and primer efficiency analyses were performed as previously described (Hoffman et al., 2002). Gene expression was normalized to Rps29 and to the corresponding experimental control, and reactions were run in triplicate and independently repeated three times. For miRNA-200c expression analysis, the Universal RT miRNA-cDNA synthesis kit, LNA-modified oligonucleotides and the miRCURY microRNA PCR System (Exiqon) were used. Gene expression was normalized to the housekeeping gene snRNA-U6 and to the corresponding experimental group control. For analysis of the miR-200 family the mirVana miRNA isolation kit was used and TaqMan microRNA assays (Applied Biosystems) were used for miR-200a, miR-200b, miR-200c, miR-141 and miR-429. Expression was normalized to the housekeeping gene snRNA-U6 and the experimental control, and another housekeeping gene, snoRNA202, was included.

Cell proliferation was detected using Click-iT EdU Imaging Systems (Invitrogen, Carlsbad, CA, USA). Briefly, SMGs were incubated with EdU for 30 minutes, fixed in 1:1 acetone:methanol or 4% paraformaldehyde (PFA), and incubated with a Click-iT EdU reaction cocktail for 30 minutes. After EdU detection, the tissue was washed and blocked for 1 hour in 10% heat-inactivated donkey serum, 1% BSA, MOM IgG blocking reagent (Vector Laboratories, Burlingame, CA, USA), prior to overnight incubation at 4°C with a rabbit primary anti-rat E-cadherin antibody (Sigma). The SMGs were washed in PBS-T (PBS containing 0.2% Tween 20) and the antibodies were detected using Cy3- or Cy5-conjugated secondary Fab fragment antibodies (Jackson Laboratories, Bar Harbor, ME, USA). Fluorescence was analyzed using a Zeiss 510 laser-scanning microscope (Zeiss, Thornwood, NY, USA) and quantified and normalized to epithelial area as previously described (Rebustini et al., 2007).

SMGs were cultured for 2 hours, fixed in 4% PFA in PBS for 20 minutes, and used for in situ hybridization (ISH) as previously described (Kloosterman et al., 2006). Briefly, the embryonic organs were bleached in 0.6% H₂O₂ for 30 minutes, permeabilized with 10 μg/ml proteinase K for 15 minutes, post-fixed with 4% PFA and 0.3% glutaraldehyde for 15 minutes, pre-incubated with hybridization buffer for 3 hours, incubated with 7.5 pmol DIG-labeled LNA-modified miR-200c probe (Exiqon) and diluted in 0.7 ml hybridization buffer. Positive and negative probes to detect U6-snRNA and miR-159 (LNA-modified and DIG-labeled, Exiqon), respectively, were also used. Hybridization was performed at 55°C for 14 hours in an automated incubator (Intavis Bioanalytical Instruments, Köln, Germany), and followed by post-hybridization washes and DIG detection using sheep anti-DIG antibody and NBT/BCP ready-to-use tablets (both from Roche Diagnostics, Indianapolis, IN, USA).

Alternatively, fluorescent in situ hybridization (FISH) was performed for miR-200c detection, as previously described (Silahtaroglu et al., 2007), with modifications. Briefly, after the hybridization and post-hybridization steps, the glands were washed in TN buffer (0.1 M Tris-HCl pH 7.5, 0.15 M NaCl) and blocked for 1 hour (1% blocking reagent, 10% sheep serum, 1% BSA). An anti-DIG-POD (Roche Diagnostics, Indianapolis, IN, USA) and anti-E-cadherin (previously described for EdU analysis) antibodies were added for incubation for 12 hours at 4°C, followed by washes with TNT buffer (TN buffer containing 0.3% Triton X-100) and detection of E-cadherin antibody as previously described. Finally, a Cy3-tyramide reagent diluted 1:100 in amplification solution (Perkin Elmer Life Sciences, Boston, MA, USA) was incubated with the glands for 5 minutes at room temperature in the dark, followed by extensive washings in TNT buffer and whole-mount laser-scanning microscopy analysis.

MAPK and AKT activation were detected by western blot using RIPA protein lysates after recombinant reelin and Vldlr function-blocking antibody treatments. Briefly, 1-5 μg of total protein from each treatment were loaded onto NuPAGE 4-12% polyacrylamide gels (Invitrogen, Carlsbad, CA, USA) and the proteins were separated and transferred to PVDF membranes, which were blocked in 5% BSA in TBS-T and incubated with rabbit anti-mouse antibodies to detect total and phosphorylated MAPK and PI3K (Cell Signaling Technology, Boston, MA, USA). The detection was performed using an anti-rabbit IgG HRP-linked antibody (Cell Signaling Technology) and West Dura Supersignal substrate (Thermo-Pierce, Rockford, IL, USA).

The miRNAs detected in SMGs using TLDA analysis were analyzed in the miRBase microRNA database (http://www.mirbase.org) (Griffiths-Jones et al., 2008). Target prediction was performed for miR-200c using three databases: TargetScan (http://www.targetscan.org) (Friedman, R. C. et al., 2009), Microcosm (EMBL-EBI Enright group, http://www.ebi.ac.uk/ enright-srv/microcosm) and PicTar (http://www.pictar.mdc-berlin.de) (Krek et al., 2005). After generating an aggregated list of miR-200c predicted target genes, their presence was further confirmed in the SMG by Agilent whole mouse genome microarray analysis (Agilent, Santa Clara, CA, USA). Epithelia from E13 SMGs were mechanically separated into end buds and ducts and three independent pools of RNA were analyzed. Triplicate arrays were analyzed with GeneSpring Software (Agilent, Santa Clara, CA, USA), comparing the relative fold change in expression between end buds and ducts. The array data are available in the SGMAP database (Salivary Gland Molecular Anatomy Project, http://sgmap.nidcr.nih.gov/sgmap/sgexp.html).

Branching morphogenesis was measured by counting the number of end buds, and the morphogenic index in epithelial cultures was calculated by multiplying the number of end buds by the length of the ducts, as previously reported (Patel et al., 2007). All SMG data were obtained using at least five SMGs per group, and each experiment was repeated three times. Data were analyzed for statistical significance using Student’s t-test to compare two different groups, and by one-way ANOVA to compare more than two experimental groups.

We screened the expression of miRNAs using TaqMan low density arrays in mouse E13 SMG epithelium, mesenchyme, epithelial end buds and the main duct using total RNA isolated from these tissues (Fig. 1A; supplementary material Fig. S1A-E). We detected 35 miRNAs that were more abundant in the epithelium than in the mesenchyme (Fig. 1B; supplementary material Table S2), 116 miRNAs that were more abundant in the mesenchyme than in the epithelium (supplementary material Fig. S1C and Table S2), and 113 miRNAs that were ubiquitously expressed in both tissues (supplementary material Table S2). The corresponding gene ontology analysis of the 20 most abundant miRNAs present in the epithelium and mesenchyme is included (supplementary material Tables S3, S4). The SMG mesenchyme contains both endothelial and neuronal cells, and some of the miRNAs detected in the mesenchyme have previously been reported to be endothelial cell- and neuronal cell-specific (supplementary material Fig. S1D,E). Among the epithelial miRNAs, eight were more highly expressed in the end buds, including miR-34a and miR-200c, and six were more highly expressed in the main ducts, including miR-204 and miR-135a (Fig. 1C,D; supplementary material Table S2). Other miR-200 family members, such as miR-141, miR-200a, miR-429 and miR-200b, are ubiquitously expressed throughout the epithelium (Fig. 1E; supplementary material Table S2).

We hypothesized that these miRNAs might regulate gene expression in discrete regions of the epithelium during branching morphogenesis, and focused on the two most highly expressed miRNAs in the end buds (miR-34a and miR-200c) and in the main duct (miR-135a and miR-204). We used miRNA antagomirs in a loss-of-function approach to investigate whether these miRNAs influence branching morphogenesis. Intact E12 SMGs were cultured in the presence of a control off-target (miR-control) or miR-200c, miR-34a, miR-204 and miR-135a antagomirs, and branching morphogenesis was analyzed by counting the number of epithelial end buds after 42 hours (Fig. 1F). Loss-of-function of miR-200c and miR-34a increased the number of end buds, whereas targeting miR-204 or miR-135a decreased branching morphogenesis. Both miR-204 and miR-135a are more highly expressed in ducts, which contain progenitor cells (Knox et al., 2010), and these miRNAs may positively regulate progenitor cell growth. Since knockdown of miR-200c resulted in the greatest increase in epithelial branching, we focused on identifying its role during SMG morphogenesis.

We analyzed miR-200c expression throughout SMG development by qPCR, and normalized its expression to a housekeeping gene (snRNA-U6) and to the initial stage of gland development (E12) before branching morphogenesis began. The expression of miR-200c increased from E14 to E15, when the SMG epithelium is rapidly proliferating, and then further increased by 84-fold in the adult tissue (Fig. 2A). We confirmed the localization of miR-200c expression in E13 epithelial end buds using fluorescent in situ hybridization (FISH) and E-cadherin staining to highlight the cell-cell junctions (Fig. 2B). Our results demonstrate that miR-200c expression is highest within the epithelial end buds.

We also measured an increase in proliferation 48 hours after miR-200c loss-of-function using EdU incorporation (Fig. 2C). Since an antagomir could inhibit miRNA families with high sequence homology in their seeding or targeting sequences, we investigated whether the effect of the miR-200c antagomir was specific. We measured the expression of other miR-200 family members (Peter, 2009) using TaqMan microRNA assays 18 hours after miR-200c antagomir treatment (Fig. 2D). Whereas the expression of miR-200c was significantly reduced, the expression of other miR-200 family members did not significantly change.

Since miR-200c regulates E-cadherin (Cdh1), a marker of epithelial identity, by directly targeting its transcriptional repressor Zeb1 (Gregory et al., 2008), we also analyzed these genes as positive controls to monitor the efficiency of miR-200c loss-of-function in our experiments. As expected, the direct target Zeb1 was upregulated and Cdh1 was downregulated after miR-200c loss-of-function (Fig. 2E). We also analyzed the expression of the FGF signaling components Fgfr1b, Ffgr2b, Fgf1 and Etv4, which are involved in epithelial proliferation (Rebustini et al., 2007; Rebustini et al., 2009), and assessed the differentiation status of the epithelium by measuring the expression of the end bud marker Aqp5 (Wei et al., 2007; Patel et al., 2008) and the progenitor cell marker Krt5 (Knox et al., 2010). The expression of Fgfr1b, Ffgr2b, Fgf1, Etv4 and Aqp5 increased, whereas Krt5 decreased (Fig. 2E). Our experiments using intact SMGs showed that miR-200c loss-of-function increases epithelial proliferation and the expression of FGFR signaling components and Etv4, which is a direct transcriptional target of Fgf10/Fgfr2b signaling in the SMG (Lombaert and Hoffman, 2010). The decrease in Krt5 and the increase in Aqp5 expression suggest that cell differentiation is also taking place. To confirm that miR-200c regulates Fgf10-dependent epithelial proliferation, we analyzed miR-200c function in isolated SMG epithelium.

We cultured isolated SMG epithelium in 3D laminin matrix in the presence of Fgf10 (Rebustini and Hoffman, 2009) with either control or miR-200c antagomirs. E-cadherin and cell proliferation were detected by immunostaining and EdU incorporation, respectively. After miR-200c loss-of-function, both epithelial morphogenesis and end bud proliferation increased, whereas E-cadherin decreased (Fig. 3A,B). We measured the expression of other miR-200c family members by TaqMan PCR after miR-200c antagomir treatment (Fig. 3C). The expression of miR-200c was significantly reduced, whereas the expression of other miR-200 family members did not significantly change. We used qPCR to measure gene expression associated with epithelial proliferation, FGFR signaling and differentiation (Fig. 3D), which recapitulated our findings from intact SMG cultures (Fig. 2E). The expression of Fgfr1b, Ffgr2b, Fgf1, Etv4 and Aqp5 increased, whereas Krt5 and Cdh1 decreased. The increase in Etv4, which is downstream of Fgfr2b signaling, suggests that FGFR signaling is influenced by mir-200c antagomir treatment. We also detected an increase in the amount of epithelial morphogenesis after miR-200c antagomir treatment, and the addition of SU5402, an inhibitor of FGFR signaling, reduced the miR-200c-dependent increase in morphogenesis (Fig. 3E). Taken together, our findings confirmed that loss-of-function of miR-200c in the epithelium increases Fgf10-dependent proliferation and demonstrates that FGFR signaling is required for the miR-200c-dependent increase in morphogenesis. The reduction of Krt5 and the increase in Aqp5 suggested that the epithelium differentiates when miR-200c function is reduced. However, none of the genes analyzed is predicted to be a direct target of miR-200c. Therefore, in order to identify the mechanism connecting miR-200c function with FGFR-dependent proliferation and Krt5 or Aqp5 expression, we needed to identify and confirm potential miR-200c target genes in the SMG that might control these processes.

We identified potential miR-200c target genes using three prediction databases: MicroCosm, TargetScan and PicTar (supplementary material Fig. S2), and generated a list of 36 genes (supplementary material Table S5) predicted by these three programs. We analyzed the expression of these genes throughout SMG development using a public microarray database, the Salivary Gland Molecular Anatomy Project (http://sgmap.nidcr.nih.gov/sgmap/sgexp.html), and confirmed the expression of 31 of the target genes by qPCR. We also compared the relative expression of these 31 target genes in E13 SMG epithelial end buds compared with the main duct by microarray analysis (Fig. 4A). Two of the predicted targets, Vldlr and Hs3st1, were more highly expressed in the epithelial end buds than in the duct (Fig. 4A). In order to confirm that the 31 genes were potential targets of miR-200c in the SMG epithelium, we analyzed their expression by qPCR after miR-200c loss-of-function by antagomir treatment of isolated SMG epithelium (Fig. 4B). We found a significant increase in expression (fold change) of Vldlr (4.2), Hs3st1 (4.0), Asap1 (4.6), Rlf (4.8), Dusp1 (7.0) and Reln (18) after 18 hours of antagomir treatment. Zeb1 (3.7), which is a known target for miR-200c, was used as a positive control to check for miR-200c loss-of-function. Conversely, we also detected a reduction in expression of the same target genes in gain-of-function experiments using a miR-200c mimic in epithelial cultures (supplementary material Fig. S3A). In addition, we detected an increase in the expression of these target genes after 18 hours of miR-200c loss-of-function using antagomirs in the intact SMG (Fig. 4B, inset), and a decrease in expression using the corresponding miR-200c mimic (supplementary material Fig. S3B). Thus, we confirmed that miR-200c targets the expression of Vldlr and Hs3st1, which are expressed in the epithelial end buds, and identified other target genes such as Dusp1, Asap1 and Rlf; these other target genes are involved in the regulation of MAPK signaling (Calvisi et al., 2008; Lin et al., 2008), proliferation during prostate cancer (Lin et al., 2008), and inhibition of embryonic stem cell differentiation (MacLean-Hunter et al., 1994), respectively. In addition, we confirmed that the VLDR ligand Reln is also a miR-200c target gene. However, the functions of Vldlr in salivary glands have not been reported; therefore, we confirmed the protein localization in the SMG.

Whole-mount immunostaining confirmed the presence and distribution of Vldlr in the epithelial end buds of E13 SMGs and showed punctate localization within epithelial cells and in the cell membranes along with E-cadherin (Fig. 5A). Since miR-200c loss-of-function increased both Vldlr and reelin, we hypothesized that Vldlr function regulates the expression of Fgfr2b signaling components and influences epithelial proliferation. To test our hypothesis, we used both gain- and loss-of-function of Vldlr. We directly stimulated Vldlr with recombinant reelin (Sinagra et al., 2005) and used a Vldlr function-blocking antibody (Oganesian et al., 2008) to inhibit Vldlr activity (Fig. 5B). Proliferation was measured by EdU incorporation in E12 SMGs treated with reelin and the Vldlr function-blocking antibody. Both branching morphogenesis and epithelial proliferation increased with reelin treatment and both decreased with Vldlr function-blocking antibody treatment (Fig. 5B). Analysis of gene expression by qPCR showed that Fgfr1b, Fgfr2b, Etv4 and Aqp5 increased with reelin treatment and decreased with Vldlr function-blocking antibodies (Fig. 5C). We found that gain- and loss-of-function of Vldlr influences proliferation and FGFR gene expression in SMGs, and investigated whether the mechanism involved miR-200c function directly in the epithelium.

To directly test the effects of gain- and loss-of-function of Vldlr we used isolated SMG epithelia cultured with Fgf10. We analyzed epithelial proliferation by EdU detection after antagomir treatment, comparing miR-200c with a miR-control antagomir (Fig. 6A-F). Epithelia treated with reelin exhibited a dose-dependent increase in proliferation (Fig. 6B,C). Importantly, the Vldlr function-blocking antibody inhibited both epithelial morphogenesis and proliferation induced by the miR-200c antagomir (Fig. 6E,F), and the epithelia resembled the control treatment (Fig. 6A). We also measured the endogenous expression of miR-200c after reelin, miR-200c antagomir and anti-Vldlr antibody treatment (Fig. 6G). As expected, the antagomir treatment downregulated miR-200c expression. Surprisingly, reelin treatment reduced miR-200c expression and the Vldlr-function blocking antibodies increased miR-200c expression ∼2-fold above control levels. These data suggest that Vldlr function also regulates miR-200c expression.

The morphology of the epithelia treated with recombinant reelin at 100 ng/ml resembled that of epithelia treated with the miR-200c antagomir (Fig. 6B,D), and we predicted that the downstream gene expression after reelin treatment would be similar to that following miR-200c antagomir treatment (Fig. 6H). Accordingly, reelin treatment increased the expression (fold change) of Fgfr1b (2.1), Fgfr2b (1.9), Etv4 (2.1) and Aqp5 (4.0), and decreased Cdh1 (0.5) and Krt5 (0.4) expression, as assessed by qPCR. These results suggest that exogenous reelin treatment mimics miR-200c loss-of-function, and supports our hypothesis that the downstream effect of miR-200c on Fgfr2b-mediated proliferation is mediated by Vldlr function.

In addition, MAPK signaling is downstream of Fgf10/Fgfr2b signaling during SMG epithelial proliferation (Steinberg et al., 2005; Rebustini et al., 2009); therefore, we investigated whether PI3K and MAPK were affected after gain- and loss-of-function of Vldlr. SMG epithelia were treated with recombinant reelin or Vldlr function-blocking antibodies and analyzed for MAPK and AKT phosphorylation. As expected from our previous work on FGFR signaling in the SMG epithelium (Steinberg et al., 2005), MAPK but not AKT phosphorylation was reduced by the Vldlr function-blocking antibody and stimulated by recombinant reelin at 24 hours (Fig. 6I). Taken together, these data demonstrate that the increase in both FGFR gene expression and epithelial proliferation are Vldlr dependent, and that feedback from Vldlr influences miR-200c expression. We conclude that a miR-200c-regulated Vldlr function plays a role by regulating FGFR signaling and proliferation in the intact SMG during branching morphogenesis.

We have discovered a functional role for miR-200c in controlling epithelial end bud proliferation during branching morphogenesis. In our working model (Fig. 7), miR-200c regulates genes that alter Vldlr function, which affects the expression of downstream FGFR-dependent genes and proliferation. In addition, miR-200c targets Zeb1 and Cdh1 expression, both involved in epithelial differentiation. In our model, we show E-cadherin separate to reelin/Vldlr signaling; however, a potential link between these two signaling systems would likely be β-catenin, which is downstream of E-cadherin and is also reported to be downstream to reelin/Vldlr signaling (He et al., 2010). We focused our studies on miR-200c based on its high expression in discrete regions of the epithelium (Fig. 1B,C, Fig. 2B), its relative abundance throughout SMG development (Fig. 2A), and on the striking induction of morphogenesis and proliferation caused by its loss-of-function (Fig. 1F). We also demonstrated that Vldlr and Hs3st1 are targeted by miR-200c in the end buds (Fig. 4). Furthermore, we determined that inhibiting Vldlr function increases, and stimulating Vldlr with reelin decreases, miR-200c expression (Fig. 6). These data suggest a feedback loop between miR-200c regulation of Vldlr and Vldlr function regulating miR-200c expression. We propose that miR-200c provides a mechanism to negatively regulate proliferation and premature differentiation in the SMG end buds as the organ enlarges during branching morphogenesis.

Our finding that miR-200c decreases proliferation during epithelial morphogenesis corroborates previous observations on the role of miR-200c and Vldlr function during cancer formation. miR-200c is downregulated or silenced in epithelial tumors (Burk et al., 2008; Shimono et al., 2009), increased Vldlr expression is also associated with cancer metastasis (He et al., 2010), and Reln is abnormally overexpressed in epithelial tumors (Perrone et al., 2007). We report that miR-200c reduces Vldlr and Reln expression in the epithelial end buds and consequently reduces FGFR-dependent proliferation, thus counterbalancing the proliferative signals from FGFR signaling during development. We also demonstrate that stimulation of Vldlr decreases miR-200c expression, but whether a similar feedback occurs during epithelial tumor formation is not known. Interestingly, the promoters for both Mir200c and Reln are epigenetically regulated by methylation in cancer (Dohi et al., 2010; Vrba et al., 2010), although it remains to be investigated whether Mir200c is subject to epigenetic regulation during organ development.

Among the 31 predicted miR-200c target genes present in the SMG (Fig. 4A), we validated seven genes using a miR-200c loss-of-function approach (supplementary material Fig. S3) followed by qPCR analysis (Fig. 4B). This result is in accordance with the relatively low success rate for the prediction of target genes using bioinformatics databases (Doran and Strauss, 2007). In addition, the use of an aggregated list containing only target genes detected by all three prediction databases (TargetScan, PicTar and MicroCosm; supplementary material Fig. S2) proved to be highly stringent. Interestingly, Zeb1, Sox2 and Klf4, which are targeted by miR-200c (Wellner et al., 2009) and increased in expression with antagomir treatment in SMG epithelium (supplementary material Fig. S4), were only predicted by two out of three databases (supplementary material Table S5). However, the presence of the binding sites for the miR-200c seeding region at the 3′UTR in Zeb1, Vldlr, Hs3st1 and Reln mRNAs (supplementary material Fig. S5) suggests that these might be miR-200c targets. Their expression increased after miR-200c antagomir treatment and decreased with the miR-200c mimic treatments in both epithelial culture and intact SMGs (supplementary material Fig. S3). We also detected an increase in the expression of the pro-proliferative targets Dusp1 (Calvisi et al., 2008; Lin et al., 2008), Asap1 (Lin et al., 2008) and Rlf (MacLean-Hunter et al., 1994) after miR-200c loss-of-function in isolated epithelia (Fig. 4B) but not in intact SMGs (Fig. 4B, inset). Expression of these genes is more abundant in the mesenchyme and might be controlled by other mesenchymal factors, rather than by miR-200c expressed in the epithelium.

Our data suggest a role for reelin, which is highly expressed in neuronal tissues (Herz et al., 2006), in modulating neuronal-epithelial crosstalk during SMG branching morphogenesis. We previously demonstrated that the innervation of the developing SMG maintains the Krt5-positive epithelial progenitor cells during branching morphogenesis (Knox et al., 2010). Thus, reelin binding to Vldlr could influence the differentiation status of the epithelium. We found that Krt5, Krt15 and Aqp3 expression associated with basal progenitor cells decreased with miR-200c loss-of-function, whereas the epithelial differentiation marker Aqp5 increased (Figs 4, 5; supplementary material Fig. S4), suggesting that miR-200c might be involved in epithelial progenitor cell differentiation in the end buds. In addition, others have demonstrated that miR-200c is highly expressed in primitive embryonic stem cells (Spruce et al., 2010) and targets Sox2, Nanog and Klf4 (Wellner et al., 2009). We also demonstrated that after miR-200c loss-of-function Sox2 and Klf4 are upregulated (supplementary material Fig. S4). Since mammary gland ductal morphogenesis is reduced in both Reln^–/– and Dab1^–/– mice (Khialeeva et al., 2011), we analyzed the SMGs in Dab1^–/– mice, which are 36% smaller than those of wild-type controls and also have reduced ductal morphogenesis (supplementary material Fig. S6). In the mammary gland there was an increase in Krt14-positive basal cells, demonstrating that differentiation was disrupted (Khialeeva et al., 2011). Accordingly, we detected a decrease in Krt14 after miR-200c loss-of-function when Vldlr expression increased (supplementary material Fig. S4). However, it remains to be determined whether miR-200c influences the proliferation of specific SMG progenitor cells or whether progenitor cells in the SMGs of Dab1^–/– mice are disrupted.

Here, we confirmed that epithelial Hs3st1 expression is targeted by miR-200c (Fig. 4B). Hs3st1, which encodes a sulfotransferase that modifies heparan sulfate (HS), is a potential miR-200c target (Hurteau et al., 2006) that is expressed in the SMG epithelial end buds (Fig. 4A). Previous studies revealed that sulfate modifications of HS influence Fgfr2b signaling and epithelial proliferation during SMG morphogenesis (Patel et al., 2008). We speculate that the effects of miR-200c on Hs3st1 expression also influence FGFR signaling, although further investigation is required to confirm this.

We used recombinant reelin and a Vldlr function-blocking antibody as tools to investigate Vldlr function during SMG morphogenesis. Importantly, we established that Vldlr regulates the expression of the components of the FGFR signaling pathway (Fig. 5C, Fig. 6H) via MAPK (Fig. 6I), which also affected morphogenesis (Fig. 6A-E) and downstream gene expression (Fig. 6H). Although previous studies showed that reelin binds to α3β1 integrin (Dulabon et al., 2000) and that both α3- and α6-integrins are crucial for SMG morphogenesis and signal via AKT and MAPK (Rebustini et al., 2007), we did not detect changes in AKT phosphorylation using recombinant reelin or Vldlr function-blocking antibody (Fig. 6I). Since reelin is not normally abundant in the SMG epithelium, further investigations are needed to identify other ligands that may influence Vldlr, such as apolipoproteins.

In summary, we propose that miR-200c counterbalances epithelial end bud FGFR-dependent proliferation by targeting Vldlr function, and that this regulation ensures appropriate epithelial outgrowth and differentiation during SMG branching morphogenesis. We also speculate that cancer formation and the dysregulation of the miR-200c/Vldlr regulatory pathway during normal development share common mechanisms that warrant investigation in the future.

We thank Vaishali Patel, Isabelle Lombaert, Sarah Knox and Shelagh Powers for helpful discussions and critical reading of the manuscript.