The miR-143-adducin3 pathway is essential for cardiac chamber morphogenesis

Developing organs, including the heart, adopt a characteristic size and shape that is essential for their function. Even subtle perturbations in cardiac architecture can lead to congenital heart defects and, in the most extreme forms, to embryonic lethality (Manner, 2009). Despite the obvious clinical implications, little progress has been made in defining the genetic pathways that drive cardiac morphogenesis.

During development, the heart assembles as a linear tube of myocardial cells that are uniform in shape and size (Auman et al., 2007; Manasek, 1976). This tube then undergoes the concurrent morphogenetic events of looping and chamber ballooning to give rise to the mature organ (Manner, 2009). During looping, the heart tube bends away from the body in a process associated with larger, more elongated cells on the convex side versus smaller, more rounded cells on the concave side (Manasek et al., 1972; Soufan et al., 2006; Taber, 2006). Simultaneously, the cardiac chambers balloon into bean-shaped structures that consist of a convex, greater surface termed the outer curvature (OC) and a concave, lesser surface called the inner curvature (IC) (Auman et al., 2007; de la Cruz and Sanchez-Gomez, 1998; Taber, 2006). As in the bent heart tube, OC chamber myocardial cells are larger and more elongated than their smaller, rounded IC cell counterparts (Auman et al., 2007; Manasek et al., 1972; Meilhac et al., 2004). Thus, the initial differences in cell size and shape of the bending heart tube correspond to the future ICs and OCs of the mature chambers (Christoffels et al., 2000; Taber, 2006). How these localized changes in cell size and shape are achieved to produce accurately contoured chambers remains elusive.

Both extrinsic and intrinsic mechanisms have been shown to affect changes in cell morphology (Auman et al., 2007; Christoffels et al., 2000). The uniformly distributed forces of blood flow and contractility are known to affect chamber formation in zebrafish by either promoting or restricting cell size and elongation, respectively, in the OC (Auman et al., 2007). Intrinsic, regionally expressed markers, such as Irx5, nppa (anf) and Chisel (Smpx), are found exclusively on the convex side of the bent heart tube and in corresponding OC cells of the mature chamber (Auman et al., 2007; Christoffels et al., 2000). Intriguingly, explanted chick hearts have been shown to bend in culture by a process involving regionalized changes in cell morphology (J. K. Butler, MA thesis, University of Texas, 1952) (Manning and McLachlan, 1990). These data indicate that cell growth and elongation are active events driven by developmental programs intrinsic to the heart. Together, these results support a model in which both genetic programs and physical forces influence the bulging of OC myocardium from the linear heart tube via a process that involves active changes in myocardial cell size and shape (Auman et al., 2007; Manasek et al., 1972; Meilhac et al., 2004; Soufan et al., 2006).

Several lines of experimentation suggest that F-actin dynamics within OC myocytes supply sufficient force to bend the heart tube by promoting cell growth and elongation (Itasaki et al., 1989; Manasek et al., 1972; Taber, 2006; Voronov et al., 2004). Heart tube bending from explanted chick hearts is blocked by the addition of actin polymerization inhibitors (Itasaki et al., 1991; Latacha et al., 2005; Manasek et al., 1972). Moreover, each chamber curvature has been shown to display a unique F-actin distribution in chick: actin fibers retain a circumferential localization in the IC, whereas OC myocytes redistribute their actin filaments in a net-like pattern throughout the cytoplasmic space (Itasaki et al., 1989; Shiraishi et al., 1992). This rearrangement of actin fibers is thought to actively drive cell growth and elongation in OC cells, while their static localization is thought to maintain cell size and circularity in IC cells (Taber, 2006). However, no genetic programs have been reported to regulate F-actin dynamics during chamber emergence.

microRNAs (miRNAs) are known to play important roles in cardiac development and disease (Bartel, 2004; Divakaran and Mann, 2008). These non-coding RNAs bind to target transcripts, inhibit their translation and often cause nonsense-mediated mRNA decay (Bartel, 2009; Giraldez et al., 2006). Binding sequences for individual miRNAs have been predicted based on algorithms that account for nucleotides 2-8 at the 5′ end, referred to as the `seed sequence' (Bartel, 2009). More recently, studies in zebrafish (Choi et al., 2007) and cultured cells (Baek et al., 2008) have suggested that a major function of miRNAs is to serve as rheostats to fine-tune target mRNA translation.

Here, we show that miR-143 is expressed in the zebrafish heart during cardiac chamber ballooning. Using morpholino knockdown analyses, we find that miR-143 is required for ventricular chamber formation and function. Our studies reveal a primary role for miR-143 in promoting OC myocardial cell growth and elongation via repression of its conserved target, adducin3 (add3), which encodes a cytoskeletal F-actin capping protein. Protection of endogenous add3 transcripts from miR-143-mediated repression, or mosaic overexpression of add3 in cardiomyocytes (CMs), leads to a block in cell elongation and F-actin rearrangements. Our findings suggest that Add3 acts in a cell-autonomous manner to regulate cytoskeletal growth and remodeling, a process tightly regulated by miR-143 function.

Zebrafish were maintained under standard conditions. Ethical approval was obtained from the Institutional Animal Care and Use Committees of Massachusetts General Hospital. The following lines were used: Tg(cmlc2::GFP), Tg(cmlc2::DsRed^nuc), Tg(flk1::GFP) and wild-type AB, TuAB and Wik strains.

Total RNA was isolated from wild-type (WT) and experimental animals at 48 hpf, 20 μg run on a 15% polyacrylamide TBE-urea gel and then probed with ³²P-labeled oligomeric DNA probes complimentary to mature miR-143 or miR-145 as described (Kloosterman et al., 2007).

In situ hybridization was performed as described (Thisse et al., 1993) using digoxygenin-labeled antisense RNA probes to cmlc2, nppa and add3. Probes were synthesized using a DIG RNA Labeling Kit (SP6/T7; Roche). miR-143 in situ hybridizations were performed as described (Wienholds et al., 2005) using 2 pmol of LNA probe with additional EDC fixation steps (Pena et al., 2009). Following in situ hybridization, embryos were embedded in JB-4 plastic resin (Polysciences) and 10 μm sections were cut using a Leica microtome. Double fluorescence in situ hybridization was performed as previously described (Auman et al., 2007; Schoenebeck et al., 2007).

Embryos were injected with either 0.5 mM MO¹⁴³ or 0.25 mM MO^add3TP, collected at 48 hpf, lysed in TRIzol reagent (Invitrogen) and total RNA purified. cDNA was synthesized with random hexamers using Superscript III reverse transcriptase (Invitrogen). cDNA (2 ng) was used to evaluate add3 expression by real-time quantitative (q) PCR (7900HT Fast Real-Time PCR System, ABI) using forward primer 5′-GCGGCAAGACTTCAACATG-3′ and reverse primer 5′-TTCCCCTTGTTCATCTGCTC-3′. bactin1 served as an internal control (Burns et al., 2005a).

Immunohistochemistry was performed as described (Alexander et al., 1998), using primary antibodies against DM-GRASP (zn-8, 1:15 dilution; Developmental Studies Hybridoma Bank, University of Iowa) and α-actinin (1:100; Santa Cruz Biotechnology). Secondary antibodies were obtained from Invitrogen (1:1000). F-actin was visualized with Alexa Fluor 546-conjugated phalloidin (Invitrogen) at 1:75 using an established protocol (Horne-Badovinac et al., 2003). Embryos were mounted in 1% low melting point agarose and imaged using an LS5 confocal microscope (Zeiss).

Antisense morpholinos were synthesized as follows (shown 5′ to 3′): multi-blocker miR-143 (MO¹⁴³, CGAGCTACAGTGCTTCATCTCAGACTCCCAG); guide Dicer miR-143 (gDi¹⁴³, TACAGTGCTTCATCTCAGACTCCCA); star Dicer miR-143 (sDi¹⁴³, GTTGACCAGAGATGCAGCACTGCAC); and target protector morpholino (MO^add3TP, ACATGAGAGATGAGTGGTAGATAGA) (Gene Tools). MOs were injected at 0.5-1.0 mM in 1× Danieau medium/1× Phenol Red into the yolk of 1- to 2-cell embryos. A miR-143 Dharmacon mimic was injected at 8-20 μM (Thermo Scientific #C-300611-05). MO¹⁴³ and MO^add3TP were co-injected at 0.16 μM and 0.2 mM, respectively. For plasmid injections, WT embryos were injected with 150 pg/nl of tol2-flanked Tg(cmlc2::add3;cmlc2::GFP), Tg(cmlc2::GFP-2A-add3) or Tg(cmlc2::GFP) circular plasmid DNAs together with 100 pg/nl of transposase RNA.

Short movies were acquired and original software used for their conversion to still-motion mode (M-Mode) images as described (Seguchi et al., 2007). End-systolic ventricular width was measured over a series of cardiac cycles and divided by the end-diastolic ventricular width to yield percentage fractional shortening (FS%).

Microangiography was performed as described (Cheng et al., 2001; Isogai et al., 2001). Approximately 3 nl of a 50% bead solution (0.36 μm fluoresceinated, carboxylated, latex micro-beads, Invitrogen) in double-distilled water/1% BSA was injected through the yolk into the sinus venosus of 36 and 48 hpf embryos mounted in tricane/2% methyl cellulose.

One-cell embryos were injected with 1 nl of 1 mM dextran-coupled Calcium Green (MW3000, Molecular Probes). Embryos were incubated in the dark until 48 hpf, at which time the heart was removed, mounted and imaged in Tyrode's medium containing blebistatin to stop physical contraction in the live heart. Fluorescence was imaged on an inverted microscope (Nikon TE2000) and recorded at 800 frames per second using a high-speed CCD camera (CardioCCD-SMQ, RedShirt Imaging, Decatur, GA, USA). Images were analyzed and isochronal maps generated with Metamorph software (Universal Imaging, Downington, PA, USA) as previously described (Milan et al., 2006). The fluorescence intensity in the ventricle and atrium was measured and plotted versus time.

Embryos were fixed, dehydrated, embedded, sectioned and imaged as previously described (Majumdar and Drummond, 2000).

Morphometric analyses were performed on zebrafish hearts after immunostaining with an anti-DM-GRASP antibody. ImageJ software (NIH, http://rsbweb.nih.gov/ij) was used to trace cell outlines and measure surface area and perimeter as described (Auman et al., 2007). Circularity was calculated as a normalized ratio of area (A) to perimeter (P), with a ratio of 1:1 representing a perfect circle (circularity=4πA/P²). Using this system, the value distinguishes cell surfaces that are more circular from those that are more elliptical. Cells were selected to be measured based on the clarity of their borders in the xy plane of view.

Sample sizes: WT, 147 IC cells and 209 OC cells from six hearts; MO¹⁴³, 185 IC cells and 191 OC cells from six hearts; MO^add3TP, 124 IC cells and 189 OC cells from five hearts; miR-143 mimic, 171 IC cells and 227 OC cells from eight hearts. Fourteen Tg(cmlc2::add3,cmlc2::GFP) mosaic animals were imaged to count 58 GFP⁺ Add3⁺ OC cells. Seven Tg(cmlc2::GFP-2A-add3) mosaic animals were imaged to count 14 GFP⁺ Add3⁺ cells in the OC. Six Tg(cmlc2::GFP) embryos were imaged to count 18 GFP⁺ cells in the OC of control animals. For blastula transplantation: WT, six IC cells and 14 OC cells from two hearts; MO¹⁴³, five IC cells and 11 OC cells from two hearts.

Standard techniques were used to create chimeric embryos at mid-blastula stages (Ho and Kane, 1990). Ten to fifteen blastomeres from uninjected or MO¹⁴³-injected Tg(cmlc2::GFP) donors were placed in the margin of WT hosts and developed until 54 hpf, fixed in 4% paraformaldehyde, immunostained with DM-GRASP antibodies, and imaged as described above. No differences in viability were observed between recipients receiving WT or morphant donor cells.

RNA was synthesized using mMessage Machine (Ambion) and injected at 150 pg/nl (GFP WT or MUT sensor RNA) or 200 pg/nl (DsRed control RNA). miR-143 mimic (#C-300611-05, ThermoFisher Scientific, Lafayette, CO, USA) was injected at 8 μM, whereas control miR-21 (Ambion) was injected at 2 μM. Average pixel intensity for the embryo was quantified in ImageJ, background corrected, and normalized to red control levels as described (Giraldez et al., 2006) (n=6 for each group). All statistics were performed using unpaired Student's t-tests.

A large-scale effort to characterize miRNA localization in the zebrafish embryo revealed that miR-143 localizes to the heart at 5 days post-fertilization (dpf) (Wienholds et al., 2005). We more precisely defined the miR-143 expression profile by whole-mount in situ hybridization and first detected transcripts throughout the heart tube at 36-38 hours post-fertilization (hpf), a developmental stage coinciding with the onset of looping and chamber ballooning (Fig. 1A,B). At 48 and 72 hpf, miR-143 became restricted to the ventricle and outflow tract (Fig. 1C-E). At 5 dpf, miR-143 was visualized in plastic sections in the ventricle, outflow region and in an anterior population of extra-cardiac pharyngeal cells (Fig. 1F). High-magnification images revealed miR-143 transcripts throughout the thickness of the ventricular wall (Fig. 1G).

To determine whether miR-143 is required for cardiogenesis, we used an antisense morpholino (MO¹⁴³) to knockdown processing of mature miR-143 (see Fig. S1 in the supplementary material). To confirm that MO¹⁴³ efficiently blocks miR-143 biosynthesis, we isolated total RNA from control and MO¹⁴³-injected embryos and performed northern blot analysis (Fig. 2A). Mature miR-143 was visible in control embryos, but absent from morphants. Another miRNA, miR-145, which shares the same primary cistron with miR-143 (Cordes et al., 2009; Thatcher et al., 2008), was present in both samples. We next analyzed heart development in MO¹⁴³-injected Tg(cmlc2::GFP) embryos, which express GFP specifically in the myocardium (Burns et al., 2005b). Although the body axes of miR-143 morphants appeared normal at 48 hpf, pericardial edema was apparent (Fig. 2D, black arrowhead). Closer examination revealed contracting, dilated atria, and collapsed ventricles exhibiting severely decreased function (n=860/1095) (Fig. 2B-G; see Movies 1 and 2 in the supplementary material). We confirmed the specificity of the MO¹⁴³-induced cardiac defect by injecting two non-overlapping morpholinos, both of which showed loss of miR-143 transcripts by northern analyses and resulted in a gross phenocopy of the cardiac defect (gDi¹⁴³, n=80/102; sDi¹⁴³, n=160/185) (Fig. 2F and see Fig. S1 in the supplementary material).

To quantify the output deficiency, end-diastolic and end-systolic dimensions of wild-type (WT) and morphant ventricles were measured with M-Mode software (Seguchi et al., 2007) and a percentage fractional shortening (FS%) was calculated. At 36 hpf, no significant difference in FS% was observed (MO¹⁴³, n=8; WT, n=5) (Fig. 2G). Microangiography confirmed blood flow in the morphants (MO¹⁴³, n=40) (Fig. 2H,J), suggesting proper cardiovascular function. However, at 48 hpf a substantial decrease in FS% was observed (MO¹⁴³, n=14; WT, n=14) (Fig. 2G), and microangiography revealed restricted blood flow as evidenced by the trapping of fluorescent beads in the yolks and atria (MO¹⁴³, n=19) (Fig. 2I,K). These data demonstrate that miR-143 promotes normal ventricular contractility and defines a critical interval for miR-143 function between 36 and 48 hpf.

A series of assays was performed to examine the possible causes of diminished ventricular output in MO¹⁴³ animals. We examined CM differentiation and proliferation by counting CM nuclei from Tg(cmlc2::DsRed2nuc) reporter embryos (Mably et al., 2003) at 72 hpf. No difference in CM number was found in MO¹⁴³ hearts compared with controls (WT, n=20; MO¹⁴³, n=22) (Fig. 3A-C). The endocardium, which was assessed using Tg(flk1::GFP) reporter embryos (Jin et al., 2005) (flk1 is also known as kdrl – Zebrafish Information Network), was also found to be present at 48 hpf (n=26/31 for morphants; n>30 for controls) (Fig. 3D,E). Analysis of calcium flux using a fluorescent indicator showed regular waves of fluorescence in both WT and MO¹⁴³ hearts at 48 hpf (Fig. 3F). Additionally, the normal delay of conduction at the atrioventricular node was preserved in MO¹⁴³ animals, as readily observed on a shortened time-scale (not shown). Sarcomere assembly was also found to be normal in morphant ventricles at 48 hpf, as evidenced by the presence of Z-bands visible in α-actinin-stained hearts (Fig. 3G,J) and transmission electron micrographs (TEMs) (Fig. 3H,K). TEMs also showed properly organized myofiber bundles containing myosin and actin filaments (Fig. 3I,L). These data demonstrate that the output defect observed in miR-143 morphants does not result from gross abnormalities in endocardial or myocardial specification, conduction system development or sarcomere assembly and maintenance.

Regionalized changes in cell size and shape have been shown to correlate with chamber morphogenesis (Auman et al., 2007; Manasek et al., 1972; Soufan et al., 2006). To determine whether cell size and shape are altered in the absence of miR-143, we visualized the outlines of myocardial cells in WT and morphant Tg(cmlc2::GFP) animals by immunostaining with the zn-8 antibody that recognizes the cell-surface antigen DM-GRASP (Alcama – Zebrafish Information Network) (Beis et al., 2005; Fashena and Westerfield, 1999) (Fig. 4A-I). Confocal images were acquired and CM cell area and circularity were quantified using ImageJ as previously described (Auman et al., 2007). Because the ventricle does not balloon in morphant animals, the OC and IC were defined as regions distal and proximal to the atrioventricular junction, respectively.

At 48 hpf, both WT and morphant IC cells remained similarly spherical (Fig. 4B,E), showing no observable difference in circularity (Fig. 4J). IC morphant cells were slightly larger than WT cells (P=9.04×10^–7), indicating that some cell growth may be achieved in the absence of miR-143. Conversely, WT OC cells appeared larger and substantially more elongated than morphant OC cells, which remained small and circular (P=1.40×10^–50 and P=1.56×10^–193, respectively) (Fig. 4C,F,J,K). These data indicate that miR-143 is crucial for OC cell growth and elongation.

To test the sufficiency of miR-143 in promoting myocardial cell growth and elongation, we analyzed CM size and circularity in embryos overexpressing miR-143 mimic RNA at 48 hpf. miR-143 did not alter cell size or shape in the OC (Fig. 4G-K); however, IC cells were much larger (P=5.17×10^–20) and more elongated (P=1.46×10^–21) than their WT counterparts (Fig. 4G-K). These findings suggest that although miR-143 is sufficient to aberrantly enlarge and elongate IC cells, it is not capable of hyper-enlarging/extending OC cells.

Since the OC appeared defective in morphant ventricles, we determined whether a conserved marker of OC fate, natriuretic peptide precursor A [nppa; also known as atrial natriuretic factor (anf)], was also affected. Interestingly, MO¹⁴³ ventricles expressed normal levels of nppa (Fig. 4L,M), indicating that morphant OC myocytes display proper regional fate but cannot undergo the cellular processes of growth and elongation.

We next analyzed cell growth and shape of miR-143 morphant cells placed in a WT hemodynamic environment. WT control or MO¹⁴³ cells from Tg(cmlc2::GFP) donor embryos were transplanted into WT hosts at mid-blastula stages and donor (green) cell morphology was analyzed in the expanded ventricle at 54 hpf (Fig. 4N-P). IC cells were of similar shape and size in WT and morphant donors (Fig. 4Q,R). Conversely, MO¹⁴³ OC cells remained smaller and more spherical (P=0.05 and P=3.44×10^–6, respectively) (Fig. 4O-R) than WT OC cells, which grew and elongated properly (Fig. 4N,Q,R). Similar to the cell exclusion phenotype displayed by transplanted, non-contractile ventricular myosin heavy chain (vmhc) mutant cells (Auman et al., 2007), MO¹⁴³ cells were sometimes aberrantly extruded from the ventricular wall (Fig. 4P). These results indicate that miR-143 acts downstream of OC cell fate specification to primarily regulate myocardial cell growth and elongation independently of blood flow.

To understand the mechanism by which miR-143 regulates cardiac form and function, a list of candidate mRNA targets was generated by mining the TargetScan and PicTar databases (Bartel, 2009; Krek et al., 2005; Lall et al., 2006) for the presence of miR-143 binding sites conserved in 3′ UTRs between chicken and mammals. The hits from each database were cross-referenced and the 34 potential transcripts were analyzed for conservation of the target sequence in zebrafish. From this analysis, a highly conserved miR-143 binding site in the adducin3 (gamma) (add3) 3′ UTR was identified (Fig. 5A).

Adducin proteins have been shown to form a complex with the fast-growing ends of filamentous actin (F-actin), such that its binding prevents the loss or addition of actin subunits (Matsuoka et al., 2000). As most miRNAs repress their mRNA targets, we hypothesized that high Add3 levels might inhibit cellular growth, cellular shape changes and normal contractile force by hyper-stabilizing/polymerizing F-actin.

We utilized four approaches to test whether add3 is a physiologically relevant target of miR-143 in vivo. We first evaluated add3 expression during development. At the onset of cardiac looping (∼36 hpf), add3 was predominantly expressed in the head and pronephric ducts, and uniformly within the developing heart tube (Fig. 5B). By 48 hpf, when the cardiac chambers have ballooned and the heart has looped, add3 transcripts were found throughout both chambers (Fig. 5C). Double-fluorescence in situ hybridization analyses of add3 and a myocardial marker, cmlc2 (myl7 – Zebrafish Information Network), showed colocalization at 36 and 54 hpf (Fig. 5D,E). These data reveal that add3 is expressed in the correct spatiotemporal domain to be a physiologically relevant target of miR-143.

Second, we analyzed whether the MO¹⁴³ myocardial phenotype is dependent upon miR-143 regulation of add3. To specifically disrupt the miR-143-add3 interaction, we analyzed embryos injected with a target protector morpholino (Choi et al., 2007) complementary to the miR-143 binding site within the add3 3′ UTR (MO^add3TP) (Fig. 5F). This manipulation should only affect add3 regulation and leave other miR-143 targets unchanged. Protection of add3 transcripts from miR-143 phenocopied the MO¹⁴³-induced ventricular defect (n=75/112) (Fig. 5G,H). The atria became dilated and the ventricles became dysmorphic and showed decreased function (see Movie 3 in the supplementary material). IC cells in MO^add3TP animals were of normal size and shape, whereas OC cells failed to grow and elongate (P=3.73×10^–43 and P=3.83×10^–37, respectively) (Fig. 5I-L). Moreover, MO¹⁴³ and MO^add3TP co-injected embryos showed no synergistic affect on CM dimensions compared with either morpholino alone (Fig. 5L), suggesting that miR-143 and add3 act in the same genetic pathway.

Next, we analyzed the relative abundance of endogenous add3 transcripts in control and experimentally manipulated embryos by quantitative PCR (Fig. 6A). Overexpression of miR-143 resulted in a ∼3-fold decrease in add3 transcript levels (P=0.0001). Conversely, injection of MO¹⁴³ or MO^add3TP showed a ∼2-fold increase in add3 expression levels (P=0.007 and 0.0001, respectively). These data demonstrate that endogenous add3 transcript levels change in response to increases or decreases in miR-143 activity and reveal that miR-143 exerts a repressive effect on add3 in vivo.

Finally, we conducted a sensor assay (Flynt et al., 2007; Giraldez et al., 2005; Kloosterman et al., 2007; Kloosterman et al., 2004; Leucht et al., 2008; Yin et al., 2008) to determine whether miR-143 can directly interact with its predicted binding sequence in the add3 3′ UTR (Fig. 6B-D). A WT sensor mRNA consisting of GFP fused to the add3 full-length 3′ UTR (Fig. 6B) was injected into embryos in the presence or absence of the miR-143 mimic RNA or a control miR-21 RNA. DsRed2 mRNA was co-injected to allow for GFP normalization. At 28 hpf, GFP fluorescence was quantified from each group using ImageJ software and normalized to that of DsRed (Fig. 6C,D) as previously described (Giraldez et al., 2006; Yin et al., 2008). Injection of the WT sensor mRNA alone resulted in bright GFP fluorescence, which was significantly dampened by miR-143 (P=0.002), but not by miR-21 (Fig. 6C). Furthermore, the miR-143-induced dampening was abrogated by co-injection with MO¹⁴³ or MO^add3TP (Fig. 6C). These findings were confirmed by quantitative western blot analyses (not shown). Additionally, a single point mutation in the predicted miR-143 binding site blocked the observed GFP silencing by miR-143 (Fig. 6D). These data reveal a specific interaction between miR-143 and the add3 3′ UTR and support a model in which ventricular myocardial cell growth and elongation require miR-143-mediated repression of add3 in vivo.

Adducin proteins cap the fast-growing ends of F-actin and stabilize/polymerize the cytoskeleton to preserve cell morphology (Matsuoka et al., 2000). Accordingly, the Adducin-F-actin complex must be disrupted to permit changes in cell size and shape (Fukata et al., 1999). We hypothesized that knockdown of miR-143 results in a net increase in Add3 levels, causing hyper-stabilization of the cytoskeleton and a block in cell growth and elongation.

To test this model, we examined whether F-actin distribution correlates with CM cell shape by visualizing F-actin (phalloidin) localization over the course of chamber formation (see Fig. S2 in the supplementary material). At 24 hpf, when CMs of the linear heart tube are small and circular (Auman et al., 2007), F-actin was localized to the periphery (see Fig. S2A-C in the supplementary material). By 36 hpf, when the heart is beginning to bend, ventricular cells of the future OC showed remodeled F-actin throughout the cytoplasm (see Fig. S2D-F in the supplementary material), whereas future IC cells retained cortical F-actin (see Fig. S2G-I in the supplementary material). By 48 hpf, when the ventricle has fully ballooned, isotropic F-actin fibers were observed throughout the cytoplasm of OC myocytes (see Fig. S2J-L in the supplementary material), whereas F-actin remained peripheral in IC myocytes (see Fig. S2M-O in the supplementary material). These analyses are consistent with previously published findings in chick (Itasaki et al., 1989; Taber, 2006) and suggest that the basic cellular mechanisms driving chamber formation are conserved among vertebrates.

We next examined whether F-actin localization was altered in the OC of MO¹⁴³ and MO^add3TP animals. Wild-type OC myocytes showed randomly organized F-actin fibers running throughout the cytoplasm at 54 hpf (n>300 cells from 11 hearts) (Fig. 7A). Conversely, in MO¹⁴³ and MO^add3TP OC CMs, F-actin was largely localized to the cortical membrane (MO¹⁴³, n>300 in eight hearts; MO^add3TP, n>100 cells from three hearts) (Fig. 7B,C). These data suggest that unlike WT CMs, which show dynamic F-actin rearrangements, morphant CMs retain F-actin at the cortex.

We next tested whether overexpressing add3 in CM clones could promote a cell-autonomous block in cell growth, elongation and F-actin redistribution in an environment in which hemostatic pressures are normal. A transgene carrying two cmlc2 myocardial promoters driving either GFP or add3 expression in opposing orientations (Fig. 7D) was injected and the resulting mosaic animals were stained with phalloidin and anti-DM-GRASP. Control animals were injected with a similar construct expressing GFP alone. Control CMs (GFP alone) were larger and more elongated than those overexpressing Add3 (P=4.9×10^–5 and P=4×10^–3, respectively) (Fig. 7E,F). Whereas abundant cytoplasmic F-actin fibers were observed in elongated control cells (Fig. 7G,H), F-actin showed a strong circumferential orientation in Add3-overexpressing CMs (n=205 cells) (Fig. 7I-N). Similar to MO¹⁴³ transplanted cells (Fig. 4O), 50% of the Add3-overexpressing cells were extruded from the ventricular wall (n=102/205) (Fig. 7K-N). These results suggest that add3 acts autonomously within the CM lineage to affect F-actin distribution, cell growth and elongation via a process that is independent of alterations in blood pressure. Together, our data uncover the miR-143-add3 genetic pathway and show that it is essential for chamber formation through direct regulation of F-actin cytoskeletal dynamics.

Heart looping and chamber ballooning are vital morphogenetic events in cardiac development, with minor perturbations leading to substantial structural and functional abnormalities. Alterations in cytoskeletal dynamics have been suggested to actively drive cardiac morphogenesis by promoting regional changes in cell size and shape (Taber, 2006). Here, we identify and connect the miR-143-add3 genetic pathway to alterations in F-actin arrangements. To our knowledge, these data provide the first link between a developmental program and the cytoskeletal rearrangements needed for chamber emergence.

Although the mechanism by which IC and OC regional information is established is poorly understood, studies in chick and zebrafish suggest that both intrinsic and extrinsic factors are at play (J. K. Butler, MA thesis, University of Texas, 1952) (Auman et al., 2007; Manning and McLachlan, 1990). Using zebrafish mutants, myocyte contraction has been shown to promote circularity, whereas blood flow has been shown to act as an opposing force, promoting cell elongation (Auman et al., 2007). Specifically, weak atrium (wea; atrial myosin heavy chain; myh6) mutants lack ventricular blood flow and show small CMs with a highly circular morphology. Although these cellular defects are reminiscent of the miR-143 knockdown, the most striking and important difference between the phenotypes is expression of nppa. wea mutants fail to maintain normal levels of nppa in the ventricular chamber, suggesting that blood flow is required for maintaining OC chamber fate. MO¹⁴³ animals, by contrast, express normal levels of nppa in the ventricle, but fail to redistribute their actin cytoskeleton. These findings indicate that the blood flow observed at 36 hpf in miR-143 morphants is strong enough to maintain OC regional information and suggest that the defect in chamber morphogenesis lies downstream of cell fate decisions.

Transcriptional initiation of miR-143 commences as the linear heart tube undergoes chamber emergence. Whereas transcription factors such as Hand1, Nppa and Chisel are regionally enriched in emerging OC cells (Christoffels et al., 2000), nppa is upregulated in future OC cells prior to cell elongation in zebrafish (Auman et al., 2007). This suggests that regionalized genetic signals, such as nppa, create inherent differences among the initially similar cells of the linear heart tube before overt cell size and shape changes occur. These early transcriptional differences might subsequently result in alternative susceptibilities to evenly distributed cues, such as genetic signals or physical forces. Although the mechanism by which the uniformly distributed miR-143 transcript exerts a regional effect on cell size and elongation during chamber emergence is unclear, our data support a role for miR-143 in globally repressing new Add3 protein production.

Two recent reports provide evidence that miR-143 regulates smooth muscle (SM) cell phenotypic switching based on data drawn from expression studies, cell culture experiments and a postnatal arterial injury model (Cordes et al., 2009; Xin et al., 2009). Both studies demonstrate that miR-143 is expressed in mouse aortic SM cells on embryonic day 9.5, consistent with a role for miR-143 in SM biology. However, miR-143 is also expressed much more robustly in progenitor cells of the cardiac crescent on embryonic day 7.5, 2 days prior to SM differentiation, and very highly in all myocardial and endocardial cells of the linear heart tube 1 day prior to SM differentiation. As SM precursors and differentiated cells are not observed in the zebrafish heart until 60 hpf and 21 dpf, respectively (Grimes et al., 2006), a potential SM deficiency cannot explain the profound defect that is evident in MO¹⁴³ embryos by 48 hpf. Moreover, these expression data in mouse support an SM-independent role for miR-143 during cardiogenesis. Although mice deficient for miR-143 are viable and display no cardiac abnormalities (Xin et al., 2009), our zebrafish miR-143 loss-of-function data suggest that other programs might compensate for the absence of miR-143 during murine development. Thus, in addition to a role in SM biology, our findings show that myocardial miR-143 acts to drive the morphogenetic events of heart ontogeny by autonomously regulating cytoskeletal dynamics.

The most compelling argument that a primary function for miR-143 is to regulate myocyte growth and elongation is the identification of a cytoskeletal component, Add3, as one of its required targets. In agreement with our findings, add3 transcripts have been shown to be 2-fold more abundant in adult SM samples from miR-143;miR-145 double-knockout mice (Xin et al., 2009). In these studies, miR-143 and miR-145 were shown to actively repress numerous transcripts involved in actin dynamics. Interestingly, two additional miRNAs, miR-1 and miR-133, were recently shown to predominantly target actin-related and actin-binding proteins of the sarcomere in skeletal muscle (Mishima et al., 2009). Since Adducin proteins are known to alter actin cytoskeletal dynamics, Add3 represents an attractive target for converting both uniform and localized genetic signals into precise cell growth and shape changes.

One model for chamber emergence can be envisioned based on the collective results from a variety of model organisms. The linear heart tube consists of smaller, more circular cells that show polymerized F-actin at the periphery. In order for the chamber myocardium to balloon from the tube, OC CMs must break down their actin cytoskeleton to allow cell growth and elongation, a process that is dependent upon fine-tuning of Add3 function. We hypothesize that miR-143 acts throughout the heart tube to limit the production of new, intracellular Add3 protein, thus providing one level of Add3 regulation. Concomitantly, we speculate that other regionalized signals, such as a kinase (Fukata et al., 1999), might act to post-translationally promote the dissociation of Add3 from F-actin, thereby allowing the cytoskeleton to be redistributed throughout the cell. In the absence of miR-143 function, continued Add3 production might overwhelm other modes of post-translational regulation, leading to sustained Add3-actin complex formation and cytoskeletal polymerization. Discerning whether miR-143 acts in concert with other regionalized factors to influence cell morphology will be vital to appreciating the full role of the miR-143-add3 pathway in cardiac morphogenesis and, possibly, disease.

We thank Daniela Panakova, Andreas Werdich, Jeffrey Winterfield, David Milan, Heather Kolpa, Mary McKee and members of the Iain Drummond laboratory for technical assistance. M-Mode analysis software was kindly provided by Seiji Takashima. Electron microscopy was performed at the Microscopy Core of the Center for Systems Biology, which is supported by an Inflammatory Bowel Disease Grant DK43351 and a Boston Area Diabetes and Endocrinology Research Center Award DK57521. The zn-8 antibody developed by Bill Trevarrow was obtained from the Developmental Studies Hybridoma Bank under the auspices of the NICHD and maintained by the University of Iowa, Department of Biological Sciences, Iowa City, IA 52242, USA. This work was supported by a Basil O'Connor Award from the March of Dimes and by the Cardiovascular Research Center at Massachusetts General Hospital.