To address the role of GSK3 during zebrafish cardiogenesis, we designed gsk3α- and gsk3β-MO for specifically inhibiting the translation of gsk3α and gsk3β, respectively. When the protein lysate was extracted from gsk3α – and gsk3β-MO-injected embryos at 24 hours postfertilization (hpf), Western blot analysis was performed by using isoform-specific antibodies. Results showed that the protein levels of GSK3α and GSK3β were largely reduced in the protein extracts from gsk3α – and gsk3β -morphants, respectively (Fig. 1), suggesting that the MOs we designed in this study were isoform-specific.

Similar morphological defects of the heart were observed in gsk3α- and gsk3β-MO-injected zebrafish embryos at 72 hpf, such as a thin and string-like shape, pericardial edema, and blood pooling (Fig. 2F, G, I, J). These defects occurred initially in some 2 days postfertilization (dpf) morphants, and then were predominantly observed in most 3- and 4-dpf morphants. Although the heart defects were similar between gsk3α- and gsk3β-MO-injected zebrafish embryos, the defects of the gsk3α morphants were more severe than those of the gsk3β morphants (Fig. 2B, C, F vs. 2D, G). Around 40% of the gsk3α morphant defects were lethal due to an absent body axis during 24 hpf (Fig. 2B), and the remainder of the surviving gsk3α morphants suffered from an incomplete formation of axis (Fig. 2C), suggesting that gsk3α and gsk3β may function differently during cardiogenesis, although they cause similar heart defects. We also noticed that the percentage of heart abnormalities was dependent on the concentration of the injected gsk3α- and gsk3β-MO (Table 1). When 0.5 ng of gsk3α-MO was injected into 1-celled stage embryos, we found that 41.8% (n = 146) of embryos displayed a string-like-shape heart; whereas when 2 ng gsk3α-MO were injected, 88.2% (n = 212) of embryos appeared similar heart defect. Similarly, 2 ng gsk3β-MO caused 30.2% (n = 126) of embryos to suffer a string-like-shape heart at 72 hpf; whereas 6 ng gsk3β-MO caused 87.5% (n = 288) of embryos to have similar heart defect. These results indicated that the effect of gsk3-MO on embryogenesis was dosage-dependent and specific.

We investigated whether the MO-induced defects could be rescued by co-injecting synthetic gsk3α or gsk3β-mRNA with its corresponding MOs, and vice versa. Results showed that co-injection of gsk3α-MO with synthetic gsk3α-mRNA could effectively rescue the defects caused by the injection of gsk3α-MO alone (Table 1). Similarly, the gsk3β-MO-induced defects were rescued by injection of gsk3β-mRNA. However, the synthetic gsk3α-mRNA did not rescue the gsk3β-MO-induced phenotype, and vice versa (Table 1). This evidence clearly demonstrates that two isoforms of GSK3 are necessary for heart development, but the function of GSK3α and GSK3β is not redundant, suggesting that GSK3α and GSK3β play specific roles in cardiogenesis during zebrafish development.

We injected either gsk3α- or gsk3β-MO into embryos derived from the transgenic line Tg(cmlc2: gfp), in which the GFP is expressed specifically in heart, resulting in a good material to monitor cardiac development of zebrafish 29. In the wild-type embryos, the heart precursor cells completed their in situ formation, elongated, and jogged to the left at 24 hpf; started looping at 30 hpf; and completed looping at 48 hpf 34. However, we observed that heart development was retarded, failing to elongate at 24 hpf (Fig. 3B) and even ceasing at heart-cone stage without further morphogenesis to a heart tube at 30313233343536 hpf (Figs. 3E, H) in the gsk3α-MO-injected embryos. We observed defective hearts as stretched to a thin and string-like shape at 72 hpf (Fig. 3L). Nevertheless, unlike in gsk3α morphants, elongation of the heart tube in gsk3β morphants at 24 hpf was as normal as in wild-type zebrafish (Fig. 3C), but heart looping was incomplete from 30 to 36 hpf (Figs. 3F, I), resulting in a stretched heart at 72 hpf (Fig. 3M).

In addition, we have designed to an experiment for using a standard negative control morpolino (MO) injection: 5'-CCTCTTACCTCAGTTACAATTTATA-3' (Gene Tools, USA). This oligo has no target, no significant biological activity. After 2 and 6 ng of this control MO were injected, no any defects were observed at 24 hpf. The morphology and development of heart appeared normally (see Additional file 1 and Figure 3N). These results reveal that the defects induced by the gsk3α- and gsk3β-MO are specific in this study.

Compared to that of wild-type and gsk3β morphants, the GFP signals in cardiomyocytes of gsk3α morphants were greatly reduced (Fig. 3B). To investigate whether the reduced GSK3α level affects the cardiomyocyte number, we used a cardiomyocyte marker, cardiac myosin light chain 2 (cmlc2), to detect cells at heart-field and heart-cone stages. We found that the number of cmlc2-positive cells was greatly reduced in gsk3α morphants at both heart-field and heart-cone stages (Fig. 4B, E), indicating that the cardiomyocyte number was greatly reduced in the gsk3α morphants. These results suggest that the retarded heart development in gsk3α morphants is due to the decreased number of cardiomyocytes during early cardiogenesis. In contrast, gsk3β morphants displayed normal cmlc2 staining (Fig. 4C, F), indicating that cardiomyocyte number remains unchanged in gsk3β morphants. These results also clearly demonstrate that GSK3α and GSK3β play distinct roles during cardiogenesis.

The pronounced degeneration in the head of gsk3α morphants at 18192021222324252627282930 hpf were also observed (Fig. 2C). To confirm whether the reduced cardiomyocyte number in gsk3α morphants was due to apoptosis, the embryos were analyzed by Terminal deoxynucleotidyl Transferase Biotin-dUTP Nick End Labeling (TUNEL) assay after MO injection. In wild-type embryos at 20 hpf, apoptosis was low (Fig. 4G). However, in gsk3α morphants at the same stage, apoptosis was pronounced throughout the axis, especially in the head (Fig. 4H) but was limited in the head of controls (G) and gsk3β morphants (I). Moreover, In gsk3a-morphants, the GFP signal was very faint at 24 hpf (Fig. 4N). The apoptotic signals were co-localized with the heart-specific GFP signal, indicating that the reduced cardiomyocyte numbers was due to apoptosis in heart (Fig. 4O). Taken together, the heart defects in gsk3α morphants was due to the reduced number of cardiomyocytes, which results from apoptosis in the head.

Although the heart of gsk3β-MO-injected embryos eventually becomes a string-like shape, we found that the cardiomyocyte development was not affected in the gsk3β morphants during early cardiac development, suggesting that GSK3β may play a unique role in cardiac morphogenesis. Whole-mount in situ hybridization of the cmlc2 probe at 36 hpf, outlining cardiac looping, was marked by a rightward bending in the ventricle in wild-type embryos (Fig. 5A). However, no looping was observed in gsk3β morphants (Fig. 5B–D). Upon detailed analysis of the early (jogging) and late (looping) stages of cardiac positioning in the gsk3β morphant heart (Table 2), we found that heart positioning was severely disrupted in gsk3β morphants and that the extent of the defect was proportional to the amount of gsk3β-MO we injected. The majority of gsk3β morphant hearts failed to jog (69.9%; 65/93). Moreover, this defect was frequently accompanied by no looping (45.3%) or L-looping (14.4%) of the heart tube, compared to wild-type, which has correct left-jogging (93.1%; 67/72) and D-looping (92.4%; 122/132). These results indicate that knockdown of GSK3β resulted in a severe disruption of jogging and looping of cardiac positioning. However, we found that the ventricle-specific marker vmhc and the atrium-specific marker amhc were normally transcribed in gsk3β morphants (Fig. 5E, G vs. 5F, H), suggesting that GSK3β might not affect the chamber-specific pattern of gene expression, although normal heart looping was not completed. We also noted that the heart positioning in gsk3α morphants was delayed but that correct jogging (left-jog) and looping (D-loop) were observed at 36373839404142434445464748 hpf, indicating that GSK3α was not involved in heart positioning.

Cardiac bmp4 is an integral component involved in the asymmetric signaling pathway and interprets left-right information for the zebrafish embryo heart 35. The bmp4 transcripts became markedly asymmetric, with far more on the left side than on the right side of the heart ring at 20 hpf (Fig. 6A, B), just before jogging. This left-predominant asymmetry persists through the stages of jogging (25 hpf, Fig. 6G). However, the pattern of bmp4 expression in gsk3β morphants was symmetrical before jogging (Fig. 6D, E) and ectopic around the heart-tube stage at 25 hpf, thereby disrupting the pattern of left-predominant asymmetry (Fig. 6H, I). Moreover, another asymmetric marker, lefty-1 36, lost its expression domain in the left side of the midline in gsk3β morphants (Fig. 6C, F). We propose that GSK3β mediates bmp4 and lefty-1 in cardiac positioning.

Many morphological defects of heart were found in the gsk3β-MO-injected zebrafish embryos. Moreover, when we used the valve markers bmp4 and versican to detect the gsk3β-MO-injected embryos at 60–72 hpf, we found that these valve markers were markedly up-regulated in the heart (Fig. 7A–D), suggesting that GSK3β might also be involved in the formation of cardiac valves. Thus, we used a two-photon fluorescence image to directly observe the valve formation of embryos derived from the transgenic zebrafish line Tg(cmlc2:Hc-RFP; 28). The yellow color shown in our nonlinear microscopy image (valves and red blood cells) is corresponding to the image modality taken by the Third-Harmonic-Generation Microscopy. Valves were normally formed in the wild-type embryos (Fig. 7E), but valves of embryos injected with gsk3β-MO were totally absent (Fig. 7F).

Hurlstone et al 37 reported that cardiac valve formation is severely affected in zebrafish APC mutants (apc^mcr). Furthermore, when axin1, another key component in the Wnt pathway, is knocked down, either a reduction or absence of heart positioning of the heart tube was frequently observed (see Additional file 2A–D). GSK3 is known to be important in the canonical Wnt pathway, and the defective valves and hearts in gsk3β-MO-injected embryos were identical to those observed in the apc^mcr mutants and axin1 morphants, suggesting that GSK3β modulates cardiac development through Wnt/β-catenin signaling.

Although, knockdown of gsk3α and gsk3β causes similar defective phenotypes, such as an unlooped and stretched heart, pericardial edema, blood pooling. We used gsk3α-MO and gsk3β-MO in the transgenic zebrafish line Tg(cmlc2:GFP), in which GFP is expressed in the myocardium specifically, to modulate and observe, in real-time, the different defective phenotypes. The hearts of gsk3α morphants failed to elongate at 24 hpf. We prove that the heart defects induced by the gsk3α-MO are due to a decreased number of cardiomyocytes. On the other hand, the gsk3β-MO-injected embryos develop normally before the onset of cardiac jogging. Defective heart positioning is observed after 26 hpf. Rescue experiments revealed that GSK 3α and GSK3β do not function redundantly. Taken together, we conclude that each isoform of GSK3 plays its own distinct role during cardiogenesis of zebrafish.

GSK3 plays an important role in the regulation of apoptosis/cell survival through the activation of caspase3 394142. These findings support a role of GSK3β in controlling apoptosis. Many studies reporting the affect of GSK3β on apoptosis have been confirmed by using GSK3 inhibitors, including lithium, the first known inhibitor, and many synthetic ones 434445. However, these inhibitors have many effects on cells and are not isoform-specific. Thus, whether GSK3α and GSK3β function redundantly or distinctly on cell survival is still ambiguous. In our study, extensive apoptosis is observed throughout the head region in the gsk3α morphants. On the other hand, only slight apoptosis is noticed in the gsk3β morphants, suggesting that GSK3α, but not GSK3β, is greatly involved in apoptosis during early embryogenesis. Moreover, embryos that are co-injected with gsk3α-MO and gsk3β mRNA do not show reduced apoptosis, suggesting that GSK3α and GSK3β do not function redundantly in cell survival.

The Wnt signaling are involved in cell proliferation and in apoptosis 46474849. On the other hand, PKB/Akt, a major regulator of GSK3, also triggers a network that regulates cell cycle progression through inactivation of GSK3β 50. It has been shown that PKB/Akt promotes cell survival in cardiac myocytes 5152. In zebrafish, apc^mcr mutant's hearts are morphologically normal during early cardiogenesis, but they fail to undergo looping morphogenesis 37. Both apc^mcr and axin1 mutants (mbl) display cardiac defects that are similar to those of gsk3β morphants. However, no information is provided about apoptosis in apc^mcr and mbl mutants. In this report, we find that apoptosis occurs in the head of gsk3α morphants. In addition, the axin1-MO-injected embryos and the mbl mutant of zebrafish have defects of looping morphogenesis in the heart, which are similar to defects occurring in the gsk3β morphants but are unlike defects occurring in the gsk3α morphants (see Additional file 2). Therefore, we know that GSK3α may not mediate apoptosis implicated in Wnt signaling because apoptotic signals do not increase in axin1 morphants (data not shown). It is worth studying which pathway is implicated in GSK3α-mediated apoptosis.

The phenotypes of apc^mcr and mbl mutants are similar to our results in that inhibition of GSK3β also causes unlooping heart tube, pericardial edema, and blood pooling 37. In addition, valve development is totally lost in gsk3β morphants (Fig. 7), which is similar to that of apc^mcr mutants. Ectopic expression of bmp4 in the heart at 24–72 hpf and ectopic expression of versican in the valve at 60–72 hpf are also observed in the apc^mcr mutant and in the gsk3β morphant (Figs. 7A–D). Moreover, the retention of bmp4 symmetry is associated with disordered jogging, and we observe that bmp4 retention occurred in the gsk3β morphant. In addition, bmp4 is downstream of Wnt/β-catenin signaling in several systems 5354. Therefore, it is reasonable to conclude that GSK3β might regulate zebrafish cardiac development by means of the canonical Wnt/β-catenin signaling pathway.

Our study reveals that knockdown of gsk3β causes a string-like heart. This phenotype is similar to the heartstrings mutant, caused by mutation of the tbx5 55. Patients with Holt-Oram syndrome, one of the autosomal dominant human "heart-hand" disorders, are caused by mutations of tbx5 56. Both loss and gain of tbx5 functions result in an absence of heart looping and an alteration in cardiac-specific genes 5758. In our study, we demonstrate that gsk3β morphants appear to have multiple heart defects, such as a non-looping or reversed looping heart, slower heart rate, and no blood circulation (Figs. 3, 4). In addition, after we probe with fin markers, we prove that the pectoral fin of the GSK3β morphant fails to differentiate (see Additional file 3). In chick, Tbx5 and Tbx4 trigger limb initiation through activation of the Wnt/Fgf signaling cascade 59. Therefore, we propose that GSK3β and Tbx5 might be involved in the same regulatory mechanism during cardiogenesis.

GSK3 is a target of prominent drugs for treating many diseases, including Alzheimer's disease and diabetes mellitus. Substrate-competitive inhibitors, which compete for the substrate binding site of the kinase, are more likely to be highly specific inhibitors. Several ATP-competitive inhibitors of GSK3 have also been defined 1718. However, the development of new drug not only requires the identification of the target, but also requires validation in an in vivo system. Recently, Atilla-Gokcumen et al., 60 performed phenotypic experiments in zebrafish embryo which is served as an in vivo experiment to analyse the functions of novel GSK3 inhibitor, organometallic reagent (R)-7. In this study, we clearly distinguish the morphological defects in zebrafish GSK3α- and GSK3β-knockdown embryos. Therefore, these findings will surely provide new criteria for the in vivo validation of potential isoform-specific inhibitors of GSK3.

In this report, we have defined that GSK3α and GSK3β play distinct roles during zebrafish cardiogenesis. Moreover, the defective valves and hearts in gsk3β-MO-injected embryos were identical to those observed in the apc^mcr mutants and axin1 morphants, suggesting that GSK3β modulates cardiac development through Wnt/β-catenin signaling. In addition, GSK3 is a critical regulator of Wnt signaling mechanism, several recent studies have shown that the components of the Wnt signaling play an important role in heart development 3. However, heart defects are not reported in the GSK3β-knockout mice. One of reasons is that mice GSK3α might function redundantly to GSK3β during the heart development of mice. We also notice that the expression profiles of GSK3β in zebrafish and in Xenopus are different: zebrafish gsk3β is weakly detected until 50–60% epiboly, but Xenopus gsk3β is expressed strongly and constantly throughout embryogenesis 6162. Taken together, although GSK3 isoforms share highly conserved in their functional domain, the biological functions of GSK3 isoforms in different species are not identical.

The zebrafish AB strain, transgenic lines Tg(cmlc2:Hc-RFP) and Tg(cmlc2:GFP) were raised and staged as previously described (28–30). The heart formation were observed under a fluorescent stereomicroscope MZ FLIII (Leica) and two-photon fluorescence microscope and Third-Harmonic-Generation Microscopy 28.

The following morpholino antisense oligonucleotides (MOs) were obtained from Gene Tools: gsk3α-MO, CCGCTGCCGCTCATTTCGGGTTGCA; gsk3β-MO, GTTCTGGGCCGACCGGACATTTTTC; axin1-MO, GCTAATGCGGTCATATCTCCTCTGC; standard negative control-MO, CCTCTTACCTCAGTTACAATTTATA. All MOs were prepared at a stock concentration of 1 mM and diluted to the desired concentration for microinjection into each embryo.

The embryos were dechorionated and deyolked with two extra washing steps as described in Link et al. 31. Deyolked samples were dissolved in 2 μl of 2 × sodium dodecyl sulfate (SDS) sample buffer per embryo and incubated for 5 min at 95°C. After full-speed centrifugation for 1 min in a microcentrifuge to remove insoluble particles, samples were loaded on a 12% SDS gel (seven embryos per lane). Antibodies used were anti-GSK3 (Santa Cruz, SC-7291, 1:750) and anti-α-tubulin (Sigma-Aldrich, T9026, 1:750).

Whole-mount in situ hybridization techniques have been described previously 32. The probes were digoxigenin-labeled, after which we cloned their partial DNA fragments.

Capped mRNAs of gsk3α, gsk3β, and RFP were synthesized according to the protocol of the manufacturer (Epicentre). The resultant mRNAs were diluted to 44 ng/μl with distilled water. Approximately 2.3 nl was injected into one-cell stage embryos.

The apoptosis assay was performed using The DeadEnd™ Colorimetric TUNEL System (Promega) and has been described previously 33.