Regulation of mouse stomach development and Barx1 expression by specific microRNAs

Transient cellular interactions in embryos have lasting consequences that necessitate precise spatiotemporal control of gene expression. Multicellular organisms restrict gene expression domains in part by activating genes selectively and for defined periods in certain tissues. However, because many mRNAs are inherently stable, timed tissue-specific gene activation is insufficient to control the duration of gene expression and the decay of some transcripts might be as important in ensuring that their expression is suitably curtailed. MicroRNAs (miRNAs) represent a potent means for brisk downregulation of transcripts after their developmental requirements have lapsed. Indeed, miRNAs were first identified as regulators of C. elegans developmental chronology (Moss, 2007; Reinhart et al., 2000); moreover, in global analyses, transcripts show higher expression at developmental stages preceding the appearance of their cognate miRNAs and reduced expression as these miRNAs accumulate (Farh et al., 2005; Stark et al., 2005). Other studies implicate selected miRNAs in the development of vertebrate limbs (Hornstein et al., 2005), muscle (Chen et al., 2006; Cordes et al., 2009), heart (Zhao et al., 2007), neurons (Makeyev et al., 2007), blood (Li et al., 2007; Thai et al., 2007), retina (Decembrini et al., 2009) and intestine (Zeng et al., 2009). One recent study profiled miRNAs expressed in the mouse intestine and showed, through tissue-specific loss of all miRNAs, that they are necessary for epithelial barrier function (McKenna et al., 2010). More than 1000 miRNAs have been identified and some are known to sharpen spatial or temporal borders of gene expression (Stefani and Slack, 2008), but few miRNAs have well-established functions and defined target genes in tissue development and organogenesis.

Mesoderm-derived stromal tissue plays a prominent role in primitive gut tube patterning and digestive tract development (Haffen et al., 1987; Kedinger et al., 1998). The homeobox gene Barx1 is expressed in abundance but transiently in prospective stomach mesenchyme and is required to specify this organ (Kim et al., 2005; Kim et al., 2007). Barx1^−/− mouse embryos have a small, deformed stomach with a mucosal lining of the intestinal type, reflecting marked homeotic transformation (posteriorization) of the foregut. Expression of Wnt signaling antagonists is diminished in the absence of Barx1. Stomach endoderm normally shows an early burst of Wnt signaling, and Barx1-mediated attenuation of this signal seems necessary for proper gastric epithelial differentiation (Kim et al., 2005).

It is not known how Barx1 expression is itself controlled, nor, specifically, how Barx1 levels decline in late-gestation stomach mesenchyme. We postulated that temporal regulation might occur through specific miRNAs that target Barx1 mRNA. Here, we show that miRNAs influence mouse stomach organogenesis and the Barx1 expression domain. We identify miR-7a and miR-203 as key miRNAs in Barx1 regulation.

Animals were handled according to protocols approved by an institutional committee.

miRNA binding was estimated as a consensus from three different prediction algorithms: miRBase (http://microrna.sanger.ac.uk/targets/v4/) uses the miRanda algorithm to predict miRNA-mRNA pairs; the miRscan (http://genes.mit.edu/mirscan/) approach allows for G:U wobbles in seed matches; and PicTar (http://pictar.mdc-berlin.de/cgi-bin/PicTar_vertebrate.cgi) confirmed candidates predicted by the first two algorithms.

Tissue RNA was isolated using TRIzol (Invitrogen), polyadenylated using ATP and poly(A) polymerase (Ambion), and reverse-transcribed using oligo-dT adapters. miRNAs and 5.8S ribosomal RNA were reverse-primed with a universal 3′ adapter (FirstChoice RLM-RACE Kit, Ambion) and forward primers based on specific miRNA sequences (see Table S1 in the supplementary material). mRNA and miRNA levels were assessed by SYBR Green PCR in triplicate on an ABI Prism 7300 instrument (Applied Biosystems). Each experiment was repeated at least three times and PCR product sizes were verified by electrophoresis. In situ hybridization probes were generated using the mirVana Probe Construction Kit (Ambion) and hybridized overnight at 58°C. Signals were detected with alkaline phosphatase-conjugated sheep anti-digoxigenin antibody (1:2000; Roche).

We cloned Barx1 3′-UTR sequences from regions complementary to miR-7a or miR-203 into the XbaI site of pGL3-Control (Promega) and deleted the polylinker upstream of the SV40 promoter between the KpnI and XhoI sites (see Fig. 3A). Pre-miRNAs 7a and 203 were cloned between the EcoRV and EcoRI sites, upstream of an IRES-eGFP cassette, in pCIG (Megason and McMahon, 2002). To detect interaction between miRNAs and Barx1 mRNA in individual cells, we replaced luciferase cDNA in pGL3-ΔKX with DsRed1, amplified by PCR from the plasmid pDsRed1-N1 (Clontech), to generate the modified reporter pGL3-ΔKX_DsRed1. We digested pCIG and its derivative pre-miRNA constructs 7a and 203 with SalI and cloned these fragments into pGL3-ΔKX_DsRed1 in reverse orientation (see Fig. 3C).

Immortalized stomach mesenchymal cells (ISMCs) were generated by infecting E12 mouse stomach mesenchyme (Kim et al., 2005) with retrovirus expressing mutant p53 (Boehm et al., 2005). Dicer1 siRNAs, prepared in vitro using the Silencer siRNA Cocktail Kit (Ambion), were introduced into immortalized cells using siPORTAmine (Ambion) and into fresh fetal mesenchymal cells using Lipofectamine 2000 (Invitrogen). Recombinant endoderm-mesenchyme cultures were performed as described previously (Kim et al., 2005), with transfection of the pre-miR plasmids described above.

pGL3-Con_7a or pGL3-Con_203 were transfected transiently into ISMCs together with empty pCIG, pCIG-pmiR7a or pCIG-pmiR203 in equal mass ratios and pRL-CMV (Promega) at a 1:15 ratio, using Lipofectamine 2000. Luciferase was measured using the Dual Luciferase Reporter Assay Kit (Promega). Experiments were performed in triplicate at least three times for each reporter-pre-miRNA combination. Data are presented as normalized ratios of firefly:Renilla luciferase.

Barx1 mRNA is highly expressed in embryonic stomach mesenchyme, but not in intestinal mesenchyme, and the levels drop steadily after ~E12 until it is barely detectable in older embryos (Kim et al., 2005); stomach Barx1 expression is confined to the mesenchyme and mesogastrium and protein levels are also markedly reduced in late gestation (Fig. 1A). As miRNAs are known to function in post-transcriptional gene control, we postulated that miRNAs contribute to Barx1 downregulation. To test this hypothesis, we used two different strategies to deplete Dicer, the RNaseIII enzyme that generates small double-stranded RNAs (Bernstein et al., 2001); Dicer-deficient cells cannot process miRNA precursors and fail to produce mature miRNAs in vitro and in vivo (Carmell and Hannon, 2004).

We first used synthetic small interfering RNAs (siRNAs) to deplete Dicer from immortalized or cultured fresh mouse fetal stomach mesenchymal cells. In both cell types, depletion of Dicer increased steady-state levels of Barx1 mRNA (data not shown) and protein (Fig. 1B). To confirm a role for miRNAs in stomach mesenchyme in vivo, we crossed mice carrying the conditional floxed-null (Fl) Dicer1 allele Dcr1^lox(ex18-20) (Kanellopoulou et al., 2005) with mice that express Cre recombinase from the endogenous Nkx3-2 (Bapx1) locus (Verzi et al., 2009). As the latter strain expresses Cre throughout the full thickness of the intestinal and distal stomach mesenchyme and the spleen anlage (Verzi et al., 2009), floxed alleles recombine selectively in these tissues. At embryonic day (E) 17.5, Bapx1^+/Cre;Dcr^Fl/Fl embryos lacked a spleen and the stomach was consistently reduced to about half the normal size (n=5), with defective morphology confined to its distal two-thirds (Fig. 1C and see Fig. S1A in the supplementary material). Mesenchymal cell mass was preserved in proportion to organ dimensions and α-smooth muscle actin immunostaining revealed overtly normal smooth muscle differentiation (Fig. 1D). Mucosal maturation was overtly unaffected; cells organized into a columnar epithelium that resembled its control counterpart in morphology (Fig. 1D) and in the expression of molecular markers (Fig. 1E). Stratified squamous epithelium in the forestomach, a region lacking Bapx1 (Cre) expression, also appeared intact but intestinal epithelial maturation was markedly aberrant, with few and structurally deficient villi (see Fig. S1B in the supplementary material). The abnormalities in the intestinal mucosa, which lacks Bapx1 (Cre) expression (Verzi et al., 2009), probably represent an indirect consequence of Dicer1 deletion in the adjacent mesenchyme. Fully penetrant asplenia reveals a previously unappreciated requirement for miRNAs in genesis of the spleen.

Compared with littermate Dcr^Fl/Fl controls lacking Cre expression, Bapx1^+/Cre;Dcr^Fl/Fl embryos showed significantly increased Barx1 mRNA levels in the E17 stomach, where Barx1 is normally low or absent (Fig. 1E). Barx1 mRNA levels were also elevated in the duodenum, where it is normally undetectable (see Fig. S1C in the supplementary material), but we observed no expression in the Bapx1^+/Cre;Dcr^Fl/Fl ileum. Immunostaining confirmed persistent Barx1 expression in the mutant stomach (Fig. 1F) and ectopic expression in the proximal intestine (see Fig. S1D in the supplementary material); similar to the normal Barx1 distribution, these aberrant signals localized in the mesenchyme. The stomach morphogenesis defects almost certainly reflect the aggregate absence of many miRNAs and not merely the significant increase in Barx1 levels. For our purposes, the key conclusion from this experiment is that depletion of miRNAs in embryonic mouse gut mesenchyme interferes with Barx1 decay, indicating that miRNAs participate in the physiological regulation of Barx1 levels.

To identify specific miRNAs that might target Barx1 mRNA, we reasoned that stomach expression of candidate miRNAs would increase in late gestation, opposing the chronology of Barx1 mRNA decay. Because miRNAs target transcripts with imperfect complementarity, any miRNA can target several mRNAs and a given transcript may pair with none, one or many miRNAs (Stefani and Slack, 2008). Using three different databases derived from pairing rules for miRNA-mRNA interaction (Lewis et al., 2003; Lim et al., 2003a) – miRBase Targets (Griffiths-Jones et al., 2008), miRscan (Lim et al., 2003b) and PicTar (Krek et al., 2005) – we identified candidate miRNAs (see Table S2 in the supplementary material) and the probability that any of these might target the Barx1 3′ untranslated region (UTR), where most miRNAs bind their mRNA targets (Ambros, 2004). Quantitative (q) RT-PCR analysis of the ten leading candidates (Fig. 2A) verified that the levels of miR-7a and miR-203 increased substantially between E12 and E19 and in inverse proportion to the decline in endogenous Barx1 levels. Conventional RT-PCR confirmed these trends (Fig. 2B). miR-689, miR-710 and miR-764-3p showed similar trends to miR-7a and miR-203 by sensitive qRT-PCR but their expression was too low to detect by other means (Fig. 2B and data not shown); results for miR-135a-1 were inconsistent between the two RT-PCR methods and the other candidates lacked the predicted temporal profile. As the Barx1 3′-UTR regions that contain miR-7a and miR-203 recognition sites are also highly conserved across species (see Fig. S2 in the supplementary material), we focused on these candidates; Fig. 2C illustrates their complementarity with the 3′-UTR of mouse Barx1. In situ hybridization indicated an abundance of miR-7a and miR-203 in E19 mouse stomach; miR-7a, in particular, gave stronger signals in the mesenchyme (Fig. 2D,E), where Barx1 is expressed. Thus, different computational algorithms predict Barx1 mRNA as a target for miR-7a and miR-203, which are expressed in a tissue-specific and temporal pattern compatible with a role in Barx1 regulation.

To test whether miR-7a and miR-203 regulate Barx1 mRNA levels, we appended the mouse Barx1 3′-UTR, carrying either the miR-7a or miR-203 recognition sequence, to a firefly luciferase reporter gene (pGL3-Con_7a or pGL3-Con_203; Fig. 3A). Both regions were deleted in the control construct pGL3-ConΔKX. We introduced these constructs into immortalized stomach mesenchymal cells (ISMCs) together with plasmids encoding precursor (pre-miR) forms of miR-7a or miR-203. miR-7a significantly reduced luciferase levels when co-expressed with pGL3-Con_7a, but had no effect on pGL3-ConΔKX or pGL3-Con_203. Conversely, miR-203 significantly repressed luciferase levels in conjunction with pGL3-Con_203, but not with pGL3-ConΔKX or pGL3-Con_7a (Fig. 3A). Thus, each miRNA interacts specifically with its cognate site in the Barx1 3′-UTR to reduce expression of a linked reporter. To test whether miR-7a and miR-203 affect endogenous Barx1 levels in a native context, we forced pre-miR expression in ISMCs or cultured fresh E12 mouse stomach mesenchymal cells without luciferase reporter genes. Both miR-7a and miR-203 reduced Barx1 protein levels in these experiments, confirming miRNA targeting of Barx1 (Fig. 3B).

To determine whether these miRNA effects are cell-autonomous, we assessed them at single-cell resolution by marking miRNA-expressing cells, using the CMV promoter to express pre-miRNAs and nuclear-localized eGFP from a single transcript separated by an internal ribosome entry site. The same constructs carried, in reverse orientation, an SV40 promoter-driven DsRed reporter gene followed by the full Barx1 3′-UTR (Fig. 3C). Control cells lacking exogenous miRNA (carrying an empty pre-miRNA insertion site) expressed DsRed and eGFP in the same cells (Fig. 3C). By contrast, DsRed expression was abolished in eGFP-expressing stomach mesenchymal cells that carried pre-miR-7a or pre-miR-203 (Fig. 3D). These data directly implicate miR-7a and miR-203 in regulating Barx1 expression levels through the 3′-UTR.

To verify the functional consequences of miRNA-mediated Barx1 depletion, we cultured E12 mouse stomach mesenchyme and endoderm under conditions that ordinarily favor Barx1-dependent gastric epithelial differentiation and suppress intestinal differentiation (Kim et al., 2005). We transfected the mesenchymal cells with miR-7a, miR-203 or empty pre-miR plasmids before overlaying naïve endoderm and assessed expression of stomach and intestinal epithelial markers 1 week later (Fig. 4A). Forced expression of miR-7a or miR-203 induced intestinal markers (Fig. 4B), compatible with reduced Barx1 function in the targeted mesenchymal cells. Because limited pre-miR transfection efficiency would inevitably preserve some residual Barx1 function, we expected modest effects on gastric epithelial differentiation; nevertheless, stomach epithelial marker expression was consistently reduced (Fig. 4B). These results confirm that expression of miRNAs 7a and 203 impairs Barx1 function.

Dynamic tissue interactions are necessary for the proper patterning, differentiation and morphogenesis of the embryonic gut (Kedinger et al., 1998; Kim et al., 2005), mandating tight control of the distribution and duration of expression of crucial factors. Barx1 is necessary for proper stomach morphogenesis and epithelial specification, acting in part through the stage-specific antagonism of nearby Wnt signaling (Kim et al., 2005; Kim et al., 2007). The onset, duration and domain of Barx1 expression are therefore important determinants of foregut development. Cis-regulatory elements and transcription factors probably drive tissue-specific Barx1 expression but are presently unknown and likely to represent only one facet of spatiotemporal control. As an additional mechanism, the presence of distinct miRNA-complementary sequences in the short Barx1 3′-UTR suggests the possibility of concerted targeting by various miRNAs. Conditional Dicer depletion in vivo indicated that miRNAs are generally required for normal gut and spleen development, but more to our point, the experiment demonstrated that some miRNAs downregulate Barx1 levels in the stomach and help avoid its expression in the intestine. These results prompted the question of which miRNAs might specifically limit the duration of Barx1 expression in mouse stomach development and our studies reveal that miR-7a and miR-203 are likely to perform this function. The accumulation of these miRNAs in fetal stomach mesenchyme coincides with declining Barx1 mRNA levels and they interact with discrete sequences in the Barx1 3′-UTR. miR-7a and miR-203 also repress expression of Barx1 and of linked reporter genes and repress Barx1 function in stomach mesenchymal cells. Targeted disruption of the Mir7 and Mir203 gene loci might formally prove the role that our study supports.

Because single miRNAs can target many transcripts with partial complementarity (Stefani and Slack, 2008), their own expression must be closely balanced with that of crucial and incidental targets. This example of Barx1 downregulation in the late embryonic mouse stomach might represent a widely used mechanism to modulate the levels and timing of genes with potent developmental activities.

We thank Warren Zimmer for providing Bapx1^+/Cre mice and Andrew McMahon for providing the pCIG vector. Supported by National Institutes of Health grant R01DK081113 (to R.A.S.) and a predoctoral training grant from the Harvard Stem Cell Institute (to J.W.). Deposited in PMC for release after 12 months.