Correspondence to: Loydie A. Jerome-Majewska, Department of Pediatrics and Human Genetics, McGill University Research Institute, Place Toulon, 4060 Ste. Catherine West PT 420, Montreal, QC H3Z 2Z3, Canada. E-mail: firstname.lastname@example.org
During development, alternative splicing allows cells to rapidly change expression of proteins without concurring changes in transcription. In fact, over 95% of human genes are alternatively spliced (Pan et al., 2008; Wang et al., 2008). ESRP1 (RBM35a) and ESRP2 (RBM35b) are members of the RNA recognition motif family of proteins. ESRP1 was initially described in a luciferase-based high-throughput cDNA expression screen for factors that promote the splicing of FGFR2, to contribute to generation of an epithelial specific isoform of FGFR2 (Warzecha et al., 2009a, 2009b). Since then, ESRP1 and ESRP2 have been shown to positively or negatively regulate alternative splicing of 148 alternative splicing events by microarrays (Warzecha et al., 2009b). More recently, RNA-seq allowed identification of an additional 281 cassette exons alternatively spliced by ESRP1 (Dittmar et al., 2012). UGG- or GGU-rich motifs on the pre-mRNA were shown to be required for recognition and splicing of these exons by ESRP1. Furthermore, knockdown of ESRP1 or both ESRPs using shRNAs induced a phenotypic change resembling an epithelial to mesenchymal transition (EMT) and increased expression of mesenchymal-specific proteins (Warzecha et al., 2010), supporting a role for ESRPs in EMT. In fact, a large amount of work suggest that ESRP1 and its homolog ESRP2 are part of a subset of proteins that regulates alternative splicing in several different human cancers (Brown et al., 2011; Gemmill et al., 2011; Shapiro et al., 2011; Lekva et al., 2012) and may promote metastasis (Reinke et al., 2002200120052012).
Orthologs of Esrp1 and Esrp2 have been identified in Caenorhabditis elegans and Drosophila, and are known as sym-2 and fusilli, respectively. Sym-2 was identified in a synthetic lethal screen with another splicing factor, mec-8 (Davies et al., 1999). Sym-2 is more highly expressed in embryos than in the first larval stages, the L1 stages (Jiang et al., 2001). Although mutation in sym-2 changes splicing of a specific subset of genes between the embryo and L1 stages (Barberan-Soler and Zahler, 2008), it does not result in an obvious homozygous mutant phenotype (Yochem et al., 2004; Barberan-Soler et al., 2011). In contrast, fusilli is detected in ovary and nurse cells of Drosophila and is required for embryonic survival (Wakabayashi-Ito et al., 2001). However, targets of fusilli have not been identified and the specific function of this gene during Drosophila development remains to be determined.
Esrp1 has been found in a number of developmental screens. In a screen for markers of definitive and visceral endoderm, Sherwood et al. reported that Esrp1 was highly enriched and expressed in endoderm at E8.5 (Sherwood et al., 2007). Ohazama et al. also reported expression of Esrp1 in pharyngeal ectoderm during dentition (Ohazama et al., 2010). In a genome-wide study on the transcriptome of developing human embryos in the first third of organogenesis, the expression of ESRP1 was found to decrease more than six-fold between Carnegie stages (S) 9 and S14 (Embryonic day (E)20–E32 for humans, approximately E9 to E11.5 in mice) (Fang et al., 2010). More recently, ESRP1 expression was shown to correlate with splicing events in the developing amnion of humans (Kim et al., 2012). However, a detailed expression of this gene during development has not been described.
We previously reported that a large number of genes are alternatively spliced in embryos and placentas during organogenesis (Revil et al., 2010). In this report, we describe our finding that expression of Esrp1 and Esrp2 is higher in embryos compared to placentas during organogenesis and we report ubiquitous expression of Esrp1 between E5.5 and E6.5 and cell-type and tissue-specific expression of Esrp1 from E7.5–E13.5. Furthermore, we show that a number of validated Esrp1 targets are alternatively spliced in a fashion that is consistent with expression of this gene in embryo and placenta.
To identify RNA-binding proteins that may regulate alternative splicing in our previous study (Revil et al., 2010), we analyzed the expression levels of over 380 genes for known RNA-binding proteins previously reported (Galante et al., 2009). In this list, several known or predicted alternative splicing factors are regulated either in a tissue-dependent way (between placenta and embryo), or in a stage-specific manner (Fig. 1). Furthermore, overall expression of these genes diverges widely, with over twelve-fold difference of expression levels between the lowest expressed gene at E11.5, Esrp2, and the highest, Rbmx, a gene shown to be involved in neural and muscle development (Dichmann et al., 2008). We found that several mRNAs show tissue-dependent expression changes: Rbmx was more highly expressed in embryos when compared to placentas (Fig. 1); in contrast, Esrp1 and Esrp2 are more highly expressed in placentas compared to embryos (Fig. 1). We chose to focus our attention on Esrp1 and Esrp2, two closely related genes implicated in EMT, an important developmental process.
Real-time quantitative RT-PCR was used to confirm the observed changes in expression of these mRNAs at E9.5 and E10.5, and to analyze these genes in earlier stages of development, E3.5 and E6.5 (Fig. 2). We observe low levels of Esrp1 at E3.5, which increase sharply in E6.5 embryos. At E9.5, Esrp1 expression is lower than at E6.5 and further decreases by E10.5, similar to what we observed in our microarray data (Figs. 1 and 2). Esrp2 mRNA levels also follow this general trend, while remaining lower than those observed for Esrp1 (Fig. 2). In addition, Esrp1 and Esrp2 are both more highly expressed in placentas compared to embryos at E9.5 and E10.5, further corroborating our microarray data (Figs. 1 and 2).
We found that a number of validated target exons of ESRP1 and ESRP2 (Warzecha et al., 2009b) are alternatively spliced in our microarray dataset (Table 1). For these exons, we compared the probe detection level in placentas and embryos of stage E11.5, a developmental stage for which the Esrp1 and Esrp2 genes' expression are significantly different (P < 0.0001) (Fig. 1). The inclusion or exclusion of the alternative exons was then compared with the observed effect of the knockdown of ESRP1 and ESRP2 in the human cancer cell lines (Table 1). We found changes of expression in 12 putative target exons, which correlate with previously observed results, 4 exons for which there are no major changes in splicing, and 5 target exons that do not correlate.
Table 1. Correlation of the Alternative Splicing of Known Validated ESRP-Sensitive Alternative Exons With Esrp Expression Levels During Mouse Development
We chose three genes with alternatively spliced exons, Dock7, Cask, and Osbpl3, for further validation by RT-PCR (Fig. 3) in E3.5, E6.5, E9.5, and E10.5 embryos, as well as placentas from E9.5 and E10.5 embryos. We found that alternative splicing of these previously identified ESRP1 target exons was consistent with the decrease of Esrp1 and Esrp2 expression in embryos during development, and the higher levels of Esrp1 and Esrp2 in the placenta versus the embryos (Fig. 2). Indeed, inclusion of the alternative target exons of Dock7 and Osbpl3 is decreased during embryogenesis and is favoured in the placenta when compared to the embryo. We found that the amplified region of the Cask pre-mRNA possesses two alternative exons; relative inclusion of both of these exons is decreased in the placentas expressing increased Esrp levels.
Esrp1 Expression Becomes Restricted at Gastrulation
Expression of Esrp1 was previously reported in the endoderm and forming tooth buds of E8.5 and E10.5 mouse embryos (Sherwood et al., 2007; Ohazama et al., 2010). In order to determine the expression pattern of Esrp1 and Esrp2 throughout mouse development, we performed in situ hybridizations on wholemount and sectioned embryos between E6.5–E14.5 and on developing placentas from E9.5–E10.5.
We did not detect Esrp2 expression in developing embryos using several probes (data not shown), whereas Esrp1 is initially broadly expressed and becomes regionally restricted. Wholemount in situ hybridization reveals broad expression of Esrp1 in embryonic ectoderm and extraembryonic ectoderm of E6.5 embryos (Fig. 4A,B). In situ hybridizations on sectioned E6.5 deciduas confirm that Esrp1 expression is excluded from parietal and visceral endoderm (Fig. 4C). In early bud E7.5 embryos, Esrp1 expression is reduced in the primitive streak and is not found in the visceral and definitive endoderm (Fig. 4D). In head-fold-staged embryos, expression of Esrp1 is observed in the chorionic ectoderm, and regions underlying the headfolds, but not in the visceral endoderm and ectoplacental cone (Fig. 4E,F). Section in situ hybridizations reveal restricted expression of Esrp1 in the epiblast but not in the primitive streak (Fig. 4G). Expression is also restricted to the junction between the neural ectoderm and surface ectoderm at E7.5 (Fig. 4G). It is also at this stage that expression of Esrp1 is found for the first time in the definitive endoderm (Fig. 4G). In older embryos, expression of Esrp1 is found in the open neural folds, in the hinge points where neural fold closures will initiate, and in the foregut (Fig. 4H,I). In the extraembryonic region, Esrp1 is highly expressed in the chorionic ectoderm in all E7.5 embryos (Fig. 4E,F,I), and in a subset of cells between the epiblast and the growing amnion (Fig. 4H,I). Esrp1 is not found in the allantois and yolk sac (Fig. 4H,I).
Esrp1 Is Expressed in a Subset of Epithelial Cells
At E8.5, the first of the five pharyngeal arches that will form in the mouse is apparent. At this stage, Esrp1 expression is found in the definitive endoderm, including the foregut, and first pharyngeal pouch (Fig. 5A,B). Expression of Esrp1 is also found in the first pharyngeal arch (Fig. 5B) and the chorion (Fig. 5D). At E9.5, when the body plan of the embryo has been set up, expression of Esrp1 is found throughout the endoderm, from the foregut to the hindgut (Fig. 6A–D). In the developing pharyngeal arches, Esrp1 expression is ventrally restricted to the pharyngeal pouches (Fig. 6A,B,F). In addition, Esrp1 is also expressed in the pharyngeal cleft of embryos at this stage (Fig. 6F). Additional sites of expression in the developing head region include the otic vesicles (Fig. 6E) and the surface ectoderm of the first pharyngeal arch (Fig. 6A,F). In the developing posterior region of the embryo, Esrp1 expression is also found in the nephric (Wolfian) duct that forms by mesenchymal to epithelial transition from the intermediate mesoderm (Fig. 6D). In the developing placenta, Esrp1 is expressed in differentiated trophoblast cells surrounding the maternal sinuses in the distal end of the developing labyrinth layer of the placenta and in a subset of trophoblast cells in the proximal region of the labyrinth layer (Fig. 6G).
Expression of Esrp1 During Organogenesis
At E10.5, organ specific differentiation has begun in developing embryos. At this stage, Esrp1 continues to be highly expressed in the developing head region, in the ectoderm of the pharyngeal clefts (Fig. 7A,G), the endoderm of the pharyngeal pouches (Fig. 7A,E,G), and the surface ectoderm of the first pharyngeal arch (Fig. 7A,E,G). In the otic vesicle, which will form the inner ear, expression of Esrp1 is ventral-laterally restricted (Fig. 7I). Novel sites of expression in the developing head region are found in Rathke's pouch, the precursor of the pituitary gland (Fig. 7B), the olfactory pit, the precursor of the nose (Fig. 7A,E), and in the nasopharyngeal groove, or surface ectoderm distal to the developing lens (Fig. 7B). Expression is also found along the anterior-posterior length of the endoderm. In the developing urogenital region, Esrp1 is expressed in the ureteric bud that emerges from the nephric duct (Fig. 7C) and high levels of Esrp1 expression are also found in the urogenital sinus that forms from the hindgut (Fig. 7C,D). In the forelimb region, Esrp1 expression is found in the surface ectoderm of the ventral surface between the outgrowing limb and the body wall (Fig. 7F), a pattern reminiscent of Lef1 expression in the first forming mammary bud (Mailleux et al., 2002).
In later stages of organogenesis, Esrp1 is maintained in the derivatives of the structures where it was expressed at E9.5. Esrp1 is expressed in anterior and posterior derivatives of the endoderm, including: the salivary glands, the developing alveoli of the lung, the pancreas, and the intestines (Fig. 8C–F). In addition, Esrp1 is expressed in the pinnae (Fig. 8A), which form from the first pharyngeal cleft, the developing ureteric branches of the kidney (Fig. 8F), and the cords of the testis, which forms from differentiation of the Wolfian duct (Fig. 8F). Novel sites of expression include the forming lens and vibrissae, the sensory placodes of the face, the mammary buds, and the hair placodes (Fig. 8A).
The ESRP family of splicing factors was discovered during a screen in human cancer cells for proteins regulating the alternative splicing of Fgfr2, with ESRP1 and ESRP2 increasing the epithelial-specific exon IIIb, while decreasing the mesenchymal-specific exon IIIc (Warzecha et al., 2009a). Since then, it has been postulated that ESRP1 and ESRP2 regulate inclusion levels of a subset of alternative exons that could be important in the process of MET, or the converse, EMT (Warzecha et al., 2010). Consistent with the importance of these processes to embryogenesis, we found that these two genes are differentially expressed during development and that alternative splicing of a subset of their known targets is consistent with a role for these two proteins and their targets during development. Interestingly, we also found by qRT-PCR that the expression of these genes was higher in placentas compared to embryos, as also observed in human (Kim et al., 2012), and that their expression in embryos decreased during organogenesis. This is contrary to our expectation that the levels of Esrp1 and Esrp2 would be higher in older embryos that contain more epithelial cells. Indeed, the high relative levels of Esrp1 mRNA in E6.5 embryos, as observed by real-time PCR, are consistent with the almost ubiquitous expression of this gene, as detected by RNA in situ hybridization. This suggests that the observed reduction of the levels of Esrp1 mRNA in later stages of embryos is due to the restriction of its expression to developing or developed organs, similar with what has been previously reported in adult mice (Warzecha et al., 2009b).
The high levels of expression of Esrp1 detected by microarray and qRT-PCR in the placenta when compared to embryos are interesting. In the developing placenta, Esrp1 expression was restricted to a group of ectodermal cells in a pattern that was reminiscent of Gcm1, syncytin-A and -B (Simmons et al., 2008). Intriguingly, Gcm1 is not alternatively spliced but both syncytin-A and syncytin-B are alternatively spliced and required for placental labyrinth development (Dupressoir et al., 2005, 2009, 2011). The potential co-localization of Esrp1 with these two genes suggests that it may be implicated in their alternative splicing. Esrp1 is also expressed in human placenta and was demonstrated to be expressed at over five-fold higher levels in the amnion than the average of other analyzed human tissues (Kim et al., 2012). Intriguingly, Esrp1 was not highly expressed in the amnion of mouse embryos and was restricted to a region where the amnion contacts the embryonic ectoderm. We confirmed that at least three genes, Dock7, Cask, and Osbpl3, contained exons that were alternatively spliced between the placenta and embryo. The mutant phenotype for Osbpl3 is not known but embryos with targeted deletions in Cask and Dock7 are born, suggesting that these genes are not required for placental development. However, Dock7 and Cask are required for normal morphogenesis and embryos with mutations in these genes show abnormalities in the craniofacial and the reproductive systems (Laverty and Wilson, 1998; Blasius et al., 2009). We are performing in situ hybridization to analyze the expression patterns of specific alternative exons in order to confirm this hypothesis. The role of ESRP1 and its alternatively spliced target exons in the developing placenta is presently unclear, but the very specific localization of Esrp1 expression observed suggests that this role is cell-type restricted.
Esrp1 was previously shown to be expressed during embryogenesis (Ohazama et al., 2010) and it was also proposed to be a marker of definitive endoderm (Sherwood et al., 2007). Our work extends these findings and shows that Esrp1 expression can be found before the onset of endoderm differentiation. The initial expression of Esrp1 was quite broad and appeared to become more refined as development progress. Intriguingly, although Esrp1 expression was epithelial-specific, as predicted by its expression in adult tissues, its expression was tissue-specific. Thus, while pre-gastrulation staged E6.5 embryos are composed of two epithelial cell types—the visceral endoderm, a secretory epithelium, and the epiblast—Esrp1 expression was restricted to the epiblast. Similarly, at E8.5, Esrp1 was expressed in the pharyngeal endoderm, an epithelial cell type but not in the neuro-epithelium. Our findings confirmed the previous observations that Esrp1 was expressed in definitive endoderm in E8.25 and E9.5 embryos, and extended these findings to show that its expression is maintained in most derivatives of the endoderm: the salivary gland, the lung epithelium, the pancreas, and the intestines. Intriguingly, we did not find expression of Esrp1 in the liver, which, similar to the pancreas, forms from the ventral foregut. However, unlike the pancreas, the liver forms by endoderm cells undergoing EMT in the hepatic mesenchyme (Zaret, 2002). Esrp1 expression was also found in the nephrogenic cord at E9.5 and was maintained in the metanephric kidney and urogenital sinus that forms from this structure. The expression pattern of Esrp1 suggests that it is expressed in epithelial precursors that will not be undergoing EMT or is rapidly downregulated when cells in these precursors initiate EMT.
Several transcriptional repressors have been found to down-regulate Esrp1 expression during TGF-β-induced EMT: δEF1, SIP1 (Horiguchi et al., 2012), and ZEB1. Recently, the transcription repressor Snail was also reported to bind to E-boxes in the Esrp1 promoter, decreasing its transcription in cells and influencing the splicing of CD44 pre-mRNA (Reinke et al., 2012). However, little is known about the genes that regulate expression of Esrp1 during development. We are currently investigating the regulation of Esrp1 expression during mouse development.
ESRP1 maps to human chromosome 8q22.1 a region commonly deleted in patients with Nablus mask-like facial syndrome (NMFLS, OMIM: 608156). NMFLS Patients display mild to severe craniofacial abnormalities in the derivatives of the pharyngeal arches including cleft palate, abnormal teeth, nose, and pinnae, as well as aortic arch malformations. Similar malformations are found when genes expressed in the pharyngeal endoderm, the signalling center that patterns the pharyngeal arches, are mutated (Jerome, 2001; Jerome-Majewska et al., 2005). Furthermore, these patients exhibit anomalies in the lacrimal duct, the nipple, and the genital, which are all derived from Esrp1-expressing tissues. However, these patients have a 2.4-MB deletion, which includes at least 10 additional genes, suggesting that this could be a contiguous deletion syndrome, with several genes in this region contributing to abnormalities in the patients. In addition, since a number of patients have been found to have less severe abnormalities or to be unaffected, it is possible that additional mutations can contribute to variable expressivity (Jain et al., 2010). Similar observations have been found in TAR and 22q11.2 syndrome (Albers et al., 2012; McDonald-McGinn et al. 2012).
Although ESRP1 and ESRP2 are predicted to function redundantly in cell culture models, we expect that this is not the case during embryogenesis. Esrp2 is detected by qRT-PCR. However, we do not detect Esrp2 expression in developing embryos with three different probes and several different in situ protocols, suggesting that Esrp2 is not highly expressed during embryogenesis or is uniformly expressed at a low level. Further studies will be necessary to assess the role, if any, of Esrp2 during mammalian development. Our findings suggest that Esrp1 will be required in a subset of epithelial cells in developing embryos.
383 RBP Microarray Expression
A list of 383 human RNA-binding proteins was taken from Galante et al. (2009). Orthologs in mice were found using BioMart on Ensmbl (http://ensembl.org/biomart/martview/). Metaprobeset IDs for the mouse homologs on the Affymetrix Mouse Exon 1.0 ST array were then identified with NetAffx (https://www.affymetrix.com/analysis/netaffx/). Average expression values were then computed from our previous microarray analyses (Revil et al., 2010) and hand curated for interesting candidates.
RNA Extraction and Reverse Transcriptase
RNA extraction was performed on embryos using TriZOL (Invitrogen, Carlsbad, CA) according to the manufacturer's instructions. The extracted total RNA was treated with DNase I, Amplification grade (Invitrogen) prior to the reverse transcriptase (RT). RNA was quantified on a NanoDrop ND1000 (ThermoScientific, Waltham, MA) and 1 &gr;g was used in a RT reaction using iScript RT Supermix (Bio-Rad, Hercules, CA).
Expression of Esrp1, Esrp2, and 18S rRNA, as an internal control, was quantified using 1/20th of the cDNA reaction, with the QuantiFast SYBR Green PCR Kit (Qiagen, Valencia, CA) on a LightCycler 480 (Roche, Indianapolis, IN). The primers used are in Supp. Table S1, which is available online. ΔCt values represent target gene (Esrp1 or Esrp2) Ct values minus rRNA Ct values. Standard deviations for averages were calculated as and error bars were represented as .
Splicing profiles of putative target genes were analyzed from cDNAs using recombinant Taq DNA polymerase (Invitrogen) according to the manufacturer's instructions, with a Tm of 58°C and extension time of 1 min. The primers used are available in Supp. Table S1. Analyses of the products were done on a 3% agarose gel and imaged using a GelDoc 2000 (BioRad).
Probes for Esrp1 and Esrp2 were made by amplifying E10.5 cDNAs with Platinum Taq HiFi (Invitrogen) using the primers available in Supp. Table S1, and cloned in the TOPO® TA Cloning Dual Promoter Kit (Invitrogen). The resulting constructs were analyzed by restriction analysis and sequencing. Transcription of the probes was done using the DIG RNA Labeling Mix (Roche) on plasmids cut with either BamHI or EcoRV (NEB) to produce the sense or antisense probes. All protocols used were those supplied by the manufacturers.
Preparation of Embryos for In Situ Hybridization and Embedding
Dissected embryos were fixed in 4% paraformaldehyde overnight and dehydrated using a graded methanol series for wholemounts or ethanol series for sections, the latter being then embedded in a 1:1 mix of paraffin and Paraplast. Seven-micromolar sections were performed on a Leica RM 2155 microtome and mounted on coated slides. In situs were performed as previously described (for wholemount in situs, see Wilkinson et al., 1990; for section in situs, see Simmons et al., 2007).
The authors thank Hélène Defalque and Benjamin Chin-Yee for their help in setting up the conditions for some experiments, the Jerome-Majewska lab and Aimee Ryan and Indra Gupta as well as their labs for comments on the figures, and Yojiro Yamanaka and Jacek Majewski for helpful comments on the manuscript and figures. L.J.M. is a member of the Research Institute of the McGill University Health Centre, which is supported in part by the Fonds de la recherche en santé du Québec (FRSQ). T.R. is supported in part by a fellowship from the Foundation of Stars at the Montreal Children's Hospital and an award from the CSR.