Muscleblind-like 1 is a negative regulator of TGF-β-dependent epithelial–mesenchymal transition of atrioventricular canal endocardial cells



The development of the valves and septa of the heart depends on the formation and remodeling of endocardial cushions. Here, we report that the alternative splicing regulator muscleblind-like 1 (MBNL1) exhibits a regionally restricted pattern of expression in canal region endocardium and ventricular myocardium during endocardial cushion development in chicken. Knockdown of MBNL1 in atrioventricular explants leads to a transforming growth factor β-dependent increase in epithelial–mesenchymal transition (EMT) of endocardial cells. This reveals a novel role for MBNL1 during embryonic development, and represents the first evidence that an alternative splicing regulator is a key player in endocardial cushion development. Developmental Dynamics 238:3266–3272, 2009. © 2009 Wiley-Liss, Inc.


The heart is the first functional organ to develop during embryonic life, forming initially as a simple tube consisting of an outer muscle layer (the myocardium) separated from an inner endothelial layer (the endocardium) by extracellular matrix. The transformation of this simple tube into a four-chambered heart is accomplished through a series of morphogenetic changes. These include cardiac looping and formation and remodeling of endocardial cushions in the atrioventricular canal (AVC) and outflow tract (OFT) regions to give rise to the valves and septa of the heart. Initial outgrowth of the endocardial cushions occurs when localized deposition of a specialized extracellular matrix called cardiac jelly in the AVC and OFT pushes the endocardium away from the myocardium (Person et al., 2005). These cushions are subsequently invaded by a subpopulation of endocardial cells within the cushions that respond to signals from the local myocardium by undergoing an epithelial–mesenchymal transition (EMT). Improper regulation of EMT or subsequent remodeling of the mesenchyme in the endocardial cushions can lead to valve and septal defects (Person et al., 2005). In the United States, approximately 5% of infants are born with heart defects each year (Pierpont et al., 2007), and heart defects are the leading cause of birth defect-related deaths.

The development of an ex vivo invasion assay in which the AVC region is explanted onto a collagen gel before the initiation of EMT (Bernanke and Markwald, 1982) has provided insight into the spatial regulation of EMT in the early heart, and has facilitated in the identification of key players in this process. Both the inductive capacity of the myocardium and the ability of endocardial cells to respond to inductive signals are restricted to the canal regions, as ventricular myocardium cannot induce EMT in AVC endocardial cells and ventricular endocardial cells do not undergo EMT in response to AVC myocardium (Mjaatvedt et al., 1987). Members of the transforming growth factor β (TGF-β) family have been shown to be critical for both initial endocardial cell activation, in which cell–cell contacts are lost, and invasion into the matrix (Mercado-Pimentel and Runyan, 2007). In this study, we report the identification of a negative regulator of TGF-β-dependent EMT, the alternative splicing factor muscleblind-like 1 (MBNL1).

Alternative splicing is a common means of generating protein diversity and modulating gene expression. It is currently estimated that more than 90% of multiexon genes undergo alternative splicing, with the majority of alternative splicing events being differentially regulated in a tissue-specific manner (Pan et al., 2008; Wang et al., 2008). Aberrant regulation of alternative splicing has been implicated in diseases such as muscular dystrophy and cancer (Cooper et al., 2009). MBNL1 is one of three members of a highly conserved family of zinc finger-containing RNA binding proteins that regulate tissue- and developmental stage-specific alternative splicing (Pascual et al., 2006). MBNL1 has gained much attention in recent years due to its contribution to the pathogenesis of myotonic dystrophy, a genetic disorder in which loss of MBNL activity contributes to a reiteration of fetal splicing patterns in adult tissues that has been linked directly to disease symptoms (Pascual et al., 2006; Cooper et al., 2009). MBNL1 is broadly expressed in adult tissues in mice and humans, with the strongest expression evident in heart, skeletal muscle, and brain (Fardaei et al., 2002; Kanadia et al., 2003). MBNL1 is detected in both embryonic and adult hearts, and has been implicated in regulating fetal-to-adult transitions in alternative splicing of cardiac transcripts during late fetal and postnatal heart development in birds and mammals (Ladd et al., 2005a; Kalsotra et al., 2008). Here, we present evidence that MBNL1 also plays a role in cardiac morphogenesis in the embryo.

We investigated the expression patterns of MBNL1 in the chick embryo, a widely used model system for the study of early heart development. In situ hybridization of early chick embryos revealed regionally restricted patterns of MBNL1 expression in both the myocardium and endocardium during cardiac morphogenesis that suggested a role for MBNL1 in endocardial cushion development. Knockdown of MBNL1 in AVC explants led to a TGF-β–dependent increase in endocardial cell EMT, indicating that MBNL1 is a negative regulator of TGF-β–mediated EMT in the endocardial cushions. This is a novel role for MBNL1 during vertebrate embryogenesis.


MBNL1 Expression Is Regionally Restricted in the Heart During Cardiac Morphogenesis

We used reverse transcriptase-polymerase chain reaction (RT-PCR) to amplify a full-length open reading frame (ORF) for MBNL1 from Hamburger and Hamilton (H&H) stage 35 chicken heart cDNA using primers against the chicken MBNL1 sequence in the NCBI database (accession no. NM_001031322). A full-length ORF was obtained for MBNL1 that is approximately 96% identical to human and mouse MBNL1 proteins at the amino acid level.

MBNL1 protein has been found in both embryonic mouse and chicken hearts by Western blotting (Ladd et al., 2005a). By in situ hybridization, high MBNL1 expression has been detected in mouse embryos in the developing nervous system, limbs, and muscles (Kanadia et al., 2003), but the spatial and temporal expression patterns in the early heart remain unclear. To determine where and when MBNL1 is first expressed in the embryonic heart, whole-mount in situ hybridization was performed on chick embryos at H&H stages 10, 14, and 18. In situ hybridization was also performed on sections from embryos at H&H stages 18 and 23. Controls processed in parallel and hybridized with sense probes were negative at all stages (data not shown). MBNL1 expression was not detected at H&H stage 10, when the primitive heart tube fuses and differentiates and the embryo undergoes neurulation (data not shown). This suggests that MBNL1 expression is not activated in the heart as part of the initial cardiac differentiation program. MBNL1 was faintly detected in the developing heart at H&H stage 14, when cardiac looping is under way (Fig. 1A). Expression was also strongly detected in the head. At H&H stage 18, MBNL1 expression continues in the developing heart and head, and is also seen in the forming limb buds (Fig. 1B), consistent with reports from the mouse (Kanadia et al., 2003).

Figure 1.

Muscleblind-like 1 (MBNL1) is expressed in ventricular myocardium and nonmesenchymal cells in the endocardial cushions. A: Ventral view of Hamburger and Hamilton (H&H) stage 14 chick embryo hybridized with MBNL1 antisense probe. The heart is shown in magnified inset. B: Left side of H&H stage 18 chick embryo hybridized with MBNL1 antisense probe. C: The heart region of a sagittal section of H&H stage 18 chick embryo hybridized with MBNL1 antisense probe. D,E: Higher magnification views of the atrioventricular canal (D) and the ventricular wall (E) from the section in C are shown. F: The heart region in a sagittal section of H&H stage 23 chick embryo hybridized with MBNL1 antisense probe. G,H: Higher magnification views of the atrioventricular canal (G) and the ventricular wall (H) from the section in F are shown. Arrowheads in E and H highlight differences in endocardial cell expression between stages 18 and 23. AV, atrioventricular canal; OFT, outflow tract; Vent, ventricle.

In situ hybridization of sections from stage 18 and 23 embryos revealed that MBNL1 expression is not uniform throughout the embryonic heart (Fig. 1C–H). In the myocardium, at stage 18 MBNL1 is absent from the muscle in the AVC (Fig. 1D), but is faintly detected in the ventricular wall (Fig. 1E). At stage 23, MBNL1 expression continues to be low or absent from the myocardium underlying the endocardial cushions in the AVC and OFT regions (Fig. 1F,G), but is now high in the ventricular wall (Fig. 1H). An intriguing possibility is that MBNL1 is involved in restricting myocardial-derived inductive signals that act on the endocardial cushions, such as TGF-β proteins or specialized extracellular matrix, to the canal regions to prevent ectopic EMT. Alternatively, MBNL1 may play a role in regulating myocardial cell proliferation or cell state changes during trabeculation and compaction of the ventricular muscle.

In contrast to its myocardial expression, in the endocardium strong MBNL1 expression is seen in the OFT and AVC regions at stage 18 (Fig. 1C,D), and is maintained at stage 23 (Fig. 1F,G). In noncanal regions, a low level of expression is detected in the endocardial lining of the atrium and ventricle at stage 18 (Fig. 1E), but is lost by stage 23 (Fig. 1H). MBNL1 is also low or absent in the mesenchymal cells that have undergone EMT and invaded the endocardial cushions. The differential and progressively more restricted pattern of expression of MBNL1 in the canal regions suggests that MBNL1 plays a role in endocardial cushion development.

Loss of MBNL1 Expression in AVC Explants Stimulates Endocardial Cell EMT

To determine whether MBNL1 expression in the AVC is important for endocardial cell EMT, MBNL1 was knocked down in AVC explants from H&H stage 14 hearts by RNA interference. Two independent MBNL1 siRNAs designed against different target sequences and with different GC content were used, and the results were compared with mock-treated explants (i.e., explants exposed to the transfection reagent in the absence of siRNA) and explants transfected with an equal concentration of a transfection control siRNA. After 38 hr of incubation, the explants were fixed and the extent of EMT was determined using Hoffman Modulation Contrast optics. Because only AVC and OFT endocardial cells undergo EMT (Mjaatvedt et al., 1987), and MBNL1 expression is highest in AVC and OFT endocardial cells (Fig. 1), we initially predicted that MBNL1 would be required for EMT. However, both siRNAs directed against MBNL1 stimulated a mild increase in endocardial cell activation and a larger increase in invasion of mesenchymal cells into the collagen matrix, while the control siRNA had no effect (Fig. 2). siRNAs against CUG binding protein 1 (CUG-BP1), an unrelated alternative splicing regulator that is also expressed in the embryonic heart (Brimacombe and Ladd, 2007), had no effect on endocardial cell activation or invasion (data not shown).

Figure 2.

Loss of muscleblind-like 1 (MBNL1) in atrioventricular canal (AVC) explants stimulates endocardial cell activation and invasion. A–D: Images of mock- (A,B) or MBNL1 siRNA (C,D) -treated AVC explant cultures were taken with the plane of focus on the surface of the gel (A,C) to show activated endocardial cells, and beneath the surface (B,D) to show mesenchymal cells that have invaded the gel. E: Activated and invaded cells were counted, and the mean was determined for each treatment group. The number of explants in each group is indicated. An asterisk indicates a significant difference from mock-transfected controls (P ≤ 0.05).

The increase in the number of mesenchymal cells in MBNL1 siRNA-treated explants could be due either to an increase in the fraction of endocardial cells undergoing EMT or an increase in the total number of endocardial cells due to increased proliferation. To determine whether loss of MBNL1 stimulates endocardial cell proliferation, 5-ethynyl-2′-deoxyuridine (EdU) assays were performed. EdU is an alternative to the commonly used bromodeoxyuridine (BrdU) assay. We found that neither the percentage of EdU-positive (i.e., proliferating) endocardial cells (14.4 ± 2.1% in mock vs. 12.4 ± 1.4% in MBNL1 siRNA-treated explants) nor the total number of endocardial cells counted for all explants in each group (12,003 in mock explants vs. 10,046 in MBNL1 siRNA-treated explants, n = 15) differed significantly between mock- and MBNL1-siRNA–treated explants (P = 0.21 and 0.18, respectively). These data indicate that MBNL1 is in fact a negative regulator of EMT, and not endocardial cell proliferation, in the AVC.

We propose that MBNL1 plays an important role during cardiac morphogenesis in limiting the production of cushion mesenchyme by preventing excessive or precocious EMT. Only a small subset of the total endocardial cell population in the cushions normally undergoes EMT (Wunsch et al., 1994). Because MBNL1 is not expressed in endocardially derived mesenchymal cells, it is tempting to speculate that this subset of endocardial cells is able to respond to EMT-inducing signals in part because they have lower basal levels of MBNL1 activity. Alternatively, MBNL1 expression may be down-regulated in this subset of cells in response to inductive cues, allowing EMT to proceed. The low level of expression and subsequent down-regulation of MBNL1 in the endocardium in noncanal regions during cardiac morphogenesis indicates that loss of MBNL1 alone is insufficient to trigger EMT in endocardial cells that are not responsive to the EMT-inducing signals produced by the cushion myocardium.

Induction of EMT Following Loss of MBNL1 Expression in AVC Explants Is TGF-β–dependent

Numerous studies have demonstrated that TGF-β proteins are required for endocardial cell activation and matrix invasion during EMT (Mercado-Pimentel and Runyan, 2007). To determine whether the loss of MBNL1 is sufficient to stimulate EMT in the absence of TGF-β, we used a pan anti–TGF-β neutralizing antibody to inhibit TGF-β signaling in mock- or MBNL1 siRNA-treated AVC explants. Activation and gel invasion were nearly eliminated by the anti–TGF-β treatment in both control and MBNL1-depleted explants (Fig. 3), demonstrating that the stimulatory effect of MBNL1 knockdown on EMT is TGF-β–dependent.

Figure 3.

Stimulation of epithelial–mesenchymal transition (EMT) in atrioventricular canal (AVC) explants by reduction of muscleblind-like 1 (MBNL1) expression is transforming growth factor β (TGF-β) -dependent. A,B: Endocardial cell activation and invasion is inhibited in mock- (A) and MBNL1 siRNA- (B) -treated AV canal explant cultures treated with an inhibitory pan anti–TGF-β antibody. C: Activated and invaded cells were counted, and the mean was determined for each treatment group. The number of explants in each group is indicated. An asterisk indicates a significant difference in the number of invaded cells relative to mock-transfected controls without additional treatment, whereas a number sign indicates a significant difference in invasion relative to MBNL1 siRNA-transfected explants without additional treatment (P ≤ 0.05). P values given in the figure above the brackets indicate comparisons in the number of invaded cells between mock-transfected explants with and without noggin, between mock- and MBNL1 siRNA-transfected explants without additional treatment, and between MBNL1 siRNA-transfected explants with and without noggin.

In the chick, early TGF-β2 signals mediate endocardial cell activation, whereas TGF-β3 induces later mesenchymal cell invasion (Mercado-Pimentel and Runyan, 2007). To determine whether the loss of MBNL1 is sufficient to stimulate invasion in the absence of TGF-β3 but the presence of earlier TGF-β signals, we used an anti–TGF-β3 neutralizing antibody to specifically inhibit TGF-β3 activity in mock- or MBNL1 siRNA-treated AVC explants. In these explants, activation was only marginally affected, but gel invasion was significantly inhibited in both control and MBNL1-depleted explants (Fig. 3C).

Bone morphogenetic protein 2 (BMP2) is thought to synergize with TGF-β3 to induce endocardial cell EMT (Yamagishi et al., 1999; Nakajima et al., 2000). Inhibition of BMP signaling with the BMP antagonist noggin has been shown to reduce the number of invaded mesenchymal cells in AVC explants from both chick and mouse (Sugi et al., 2004; Sakabe et al., 2008). To determine whether BMP signaling is also required for MBNL1 siRNA-induced EMT, we added noggin to mock- or MBNL1 siRNA-treated AVC explants. Noggin treatment did not significantly reduce cell invasion in MBNL1 siRNA-treated explants (Fig. 3C). It is worth noting that, in our hands, noggin treatment resulted in only a small but statistically nonsignificant decrease in cell invasion in mock-transfected explants, consistent with a report that noggin inhibits EMT in explants from stage 16, but not stage 14, embryos (Romano and Runyan, 2000). These data suggest that, unlike TGF-β3, BMP2 is not essential for the enhancement of EMT induced by the loss of MBNL1.

MBNL1 is primarily expressed in the AVC in the endocardium (Fig. 1D,G), which is the tissue that responds to inductive TGF-β signals. These results, therefore, suggest that MBNL1 may act on a downstream component of the TGF-β signaling pathway. A survey of the databases reveal that several downstream mediators of TGF-β signaling are alternatively spliced, including the TGF-β receptors TGFBR1 (also known as ALK5) and TGFBR2, the co-receptor endoglin, and the downstream effectors Smad2 and Smad3. A goal of future studies is to identify the specific splicing targets of MBNL1 that mediate its effects on EMT. Other interesting possibilities to explore include whether TGF-β signaling induces the down-regulation of MBNL1 expression in the subset of endocardially derived cells that undergo the mesenchymal transition, or whether localized TGF-β signaling in the canal regions is required to maintain MBNL1 expression in the cushion endocardium while it is down-regulated elsewhere.

Taken together, our data reveal an important new role for MBNL1 during embryonic heart development as a negative regulator of endocardial cell EMT. We propose that MBNL1 directs alternative splicing events that are critical for preventing the overproduction of cushion mesenchyme that would lead to valve or septal malformations. MBNL1 is the first alternative splicing regulator identified as a key player in endocardial cushion development.


In Situ Hybridization

The chicken MBNL1 open reading frame was amplified from embryonic day 8 chicken heart RNA by RT-PCR as previously described (Ladd et al., 2001) using the primers ATGGCCGT CAGCGTCACGCCCATC and CTACA TCTGTGTAACATACTTGTG, cloned into the pCR-Blunt II-TOPO vector using the Zeroblunt TOPO kit (Invitrogen), and confirmed by sequencing. MBNL1 antisense and sense probes were transcribed from the TOPO clone and in situ hybridizations on whole embryos and sections were performed as previously described (Brimacombe and Ladd, 2007). Fertilized Babcock B-300 chicken eggs (Case Western Reserve University Squire Valleevue and Valley Ridge Farms) were incubated in 50% humidity at 100°F until desired stages were reached. Chicken embryos were staged according to Hamburger and Hamilton (Hamburger and Hamilton, 1992).

Collagen Gel Preparation

Collagen was extracted from adult rat tail tendons by incubation at 4°C in 100 ml/g 0.5 M acetic acid with stirring for 3–4 days, followed by vacuum filtration through cotton gauze and centrifugation at 10,000 × g for 15 min. Supernatant was dialyzed for 3 days at 4°C in 10× volume 0.1× Medium 199 (Mediatech), pH 4.0, with 1 ml of chloroform added. Collagen stock solution was aliquotted and stored at −20°C. Working dilutions for suitable gel polymerization were determined empirically. Collagen gels were prepared by adding 1 ml of collagen stock solution diluted 1:3 in water to 200 μl of 2.2% sodium bicarbonate diluted 1:1 in 10× Medium 199. A total of 290 μl of the gel mixture was added per well of Nunclon *D Multidishes (Fisher) and allowed to solidify at 37°C for at least 15 min. A total of 500 μl of 1× M199+ (1× Medium 199 supplemented with ITS (Invitrogen), pen/strep antibiotic solution, and 1% chick serum) was added to each well 1 hr before use, and removed before explants were placed on the gel.

Knockdown of MBNL1 in AVC Explants

Embryos were collected at stage 14. The AVC regions were excised, cut open, preincubated at room temperature for 45 min with or without 10 nM siRNA using siPORT NeoFX (Ambion), and placed endothelial-side down on the collagen gels, avoiding fluid carryover. After incubation at 37°C for 18 hr, explants were boosted with a second siRNA transfection. After 38 hr of total incubation, explants were fixed in 4% paraformaldehyde for 45 min at room temperature. Explants were washed twice with 1× PBS and stored at 4°C until counted. Endothelial cell activation (defined as isolated cells on the surface of the gel) and mesenchymal cell invasion (defined as cells wholly within the gel matrix) were analyzed using Hoffman Modulation Contrast optics.

siRNA duplexes against MBNL1 (5′-GGGAAUUCCUCAAGCUGUAdT dT-3′ and 5′-CAUAAUAUCUGCCG AACAUdTdT-3′) and CUG-BP1 (5′-GGGUGCUGUUUUGUUACAUdTdT- 3′ and 5′-GAGCCGAGGUUGUGCA UUUdTdT-3′) were obtained from Dharmacon RNAi Technologies. Efficient transfection was confirmed using siGLO Green Transfection Indicator (Dharmacon). siGLO Green was also used as a control siRNA that does not effect MBNL1 expression. To confirm the knockdown of MBNL1 in siRNA-treated explants, three to four mock- or siRNA-treated explants were pooled and RNA was extracted by the method of Chomczynski and Sacchi (Chomczynski and Sacchi, 1987). Semi-quantitative RT-PCR was carried out as previously described (Ladd et al., 2005b), except that instead of end-labeling the forward primer, 5 μCi of [α-32P]dCTP was added to each reaction. Total MBNL1 transcript levels were normalized against the level of 18S RNA. Conditions were optimized for each primer set for amplification in the linear range: MBNL1 = AGGTGGAGAACG GACGTGT and GCCATGTTCTTCTG CTGGAT, 61.8°C, 20 cycles; 18S = G GTAACCCGTTGAACCCCATTC and GGACCTCACTAAACCATCCAATCG, 56.7°C, 8 cycles. All PCR products were resolved on 5% nondenaturing polyacrylamide gels, scanned by a Phosphorimager (Molecular Dynamics), and quantified using ImageQuant software. Reduction of MBNL1 transcript levels was approximately 60% in siRNA-treated explants relative to mock-treated explants.

The efficacy of protein knockdown was confirmed in primary chicken embryonic cardiomyocytes. Cardiomyocytes were cultured as previously described (Ladd et al., 2005a) and transfected using Lipofectamine 2000 (Invitrogen) according to manufacturer's protocol. Total protein lystates were collected after 24 or 48 hr and Western blots were performed as previously described (Ladd et al., 2001) using the anti-MBNL1 antibody A2764 (kind gift of Charles Thornton, University of Rochester) or the anti-CUG-BP1 antibody 3B1 (Invitrogen). Each duplex exhibited ≥75% knockdown of its target protein relative to mock- or siGLO siRNA-transfected plates at 10 nM concentrations (data not shown). Semi-quantitative RT-PCR for MBNL1 performed on total RNA samples collected from parallel plates demonstrated a corresponding reduction of MBNL1 of ∼60%, similar to that seen in AVC explants.

Assessment of Endocardial Cell Proliferation in AVC Explants

AVC regions were transfected with or without MBNL1 siRNA and explanted onto collagen gels as described above. At 38 hr after explantation, 100 μM EdU was added and explants were incubated for an additional 4 hr before fixation in 4% paraformaldehyde. EdU incorporation was detected using the Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen) following manufacturer's directions. The total number and number of EdU-positive endocardial cells were determined for 15 explants per treatment group, and the average percentage of EdU-positive cells per explant was calculated.

Anti–TGF-β and Noggin Treatments

AVC regions were transfected with or without MBNL1 siRNA and explanted onto collagen gels as described above. After 6 hr, either 10 μg/ml of a neutralizing antibody that recognizes TGF-β1, -β2, and -β3 (R&D Systems, catalog no. MAB1835), 10 μg/ml of a neutralizing antibody that recognizes only TGF-β3 (R&D Systems, catalog number AF-243-NA), or 500 ng/ml human recombinant noggin (Invitrogen) in M199+ or M199+ alone was added and explants were returned to 37°C. At 18 hr after explantation, the treatment was removed and explants were boosted with siRNA for 45 min as above. M199+ with or without fresh anti-TGFβ antibody or noggin was then added and explants were incubated at 37°C until fixation at 38 hr.


Data are reported as mean ± standard error of the mean. Comparisons between means were performed by means of two-tailed t-tests assuming unequal variances using Microsoft Excel software. Differences were considered statistically significant when P ≤ 0.05.


We thank Donna Driscoll for helpful comments on the manuscript. A.N.L. was funded by the March of Dimes, the American Heart Association, and the National Institutes of Health.