The myocardium utilizes Pdgfra-PI3K signaling to steer towards the midline during heart tube formation

Abstract

Coordinated cell movement is a fundamental process in organ formation. During heart development, bilateral myocardial precursors collectively move towards the midline (cardiac fusion) to form the primitive heart tube. Along with extrinsic influences such as the adjacent anterior endoderm which are known to be required for cardiac fusion, we previously showed that the platelet-derived growth factor receptor alpha (Pdgfra) is also required. However, an intrinsic mechanism that regulates myocardial movement remains to be elucidated. Here, we uncover an essential intrinsic role in the myocardium for the phosphoinositide 3-kinase (PI3K) intracellular signaling pathway in directing myocardial movement towards the midline. In vivo imaging reveals that in PI3K-inhibited zebrafish embryos myocardial movements are misdirected and slower, while midline-oriented dynamic myocardial membrane protrusions become unpolarized. Moreover, PI3K activity is dependent on and genetically interacts with Pdgfra to regulate myocardial movement. Together our findings reveal an intrinsic myocardial steering mechanism that responds to extrinsic cues during the initiation of cardiac development.

Introduction

During organogenesis, cell progenitor populations often need to move from their origin of specification to a new location in order to form a functional organ. Deficient or inappropriate movement can underlie congenital defects and disease. Directing these movements can involve extrinsic factors such as chemical and mechanical cues from neighboring tissues and the local environment as well as intrinsic mechanisms such as intracellular signaling and polarized protrusions (1). Progenitor cell movement occurs during cardiac development, where myocardial cells are specified bilaterally on either side of the embryo (2). To form a single heart that is centrally located, these bilateral populations must move to the midline and merge (3, 4). As they move, myocardial cells undergo a mesenchymal-to-epithelial (MET) transition forming intercellular junctions and subsequently moving together as an epithelial collective (5–9). This process is known as cardiac fusion and occurs in all vertebrates (10, 11).

External influence from the adjacent endoderm is essential for the collective movement of myocardial cells towards the midline. Mutations in zebrafish and mice which inhibit endoderm specification or disrupt endoderm morphogenesis result in cardia bifida – a phenotype in which the bilateral myocardial populations fail to merge (9, 12–21). Similar phenotypes also occur in chicks and rats when the endoderm is mechanically disrupted (22–24). Studies simultaneously observing endoderm and myocardial movement have found a correlation between the movements of these two tissues, suggesting a model in which the endoderm provides the mechanical force that pulls myocardial cells towards the midline (24–27). Yet, these correlations do not occur at all stages of cardiac fusion, indicating that myocardial cells may also use intrinsic mechanisms to actively move towards the midline. Indeed, recent studies revealing a role for the receptor tyrosine kinase, platelet-derived growth factor receptor alpha (Pdgfra) in the movement of myocardial cells have suggested a paracrine chemotaxis model, in which the myocardium senses chemokine signals from the endoderm and responds to them (28). However, the existence and identity of these intrinsic myocardial mechanisms remain to be fully elucidated.

We have sought to identify the intracellular pathways downstream of Pdgfra that regulate the collective movement of the myocardium. The phosphoinositide 3-kinase (PI3K) pathway is known as an intracellular signaling mediator of receptor tyrosine kinases (e.g. Pdgfra). PI3K phosphorylates phosphatidylinositol (4,5)-bisphosphate (PIP2) to create phosphatidylinositol (3,4,5)-trisphosphate (PIP3), a regulator of cellular processes such as proliferation and cell migration (29). The PI3K pathway has been shown to be important for both individualistic cell migration such as in Dictyostelium and neutrophils (30, 31) as well as collective cell migration such as in the movement of border cells in Drosophila and the movement of the anterior visceral endoderm during mouse gastrulation (32, 33).

Using the advantages of external development and ease of live-imaging in the zebrafish model system (34), our studies reveal that myocardial PI3K signaling is required for proper directional movement towards the midline during cardiac fusion. In particular we find that inhibition of the PI3K pathway, throughout the embryo or only in the myocardium, results in bilateral cardiomyocyte populations that fail to reach the midline (cardia bifida) or have only partially merged by the time wild-type myocardial cells are fully merged. High-resolution live imaging in combination with mosaic labeling further reveals that the orientation of myocardial membrane protrusions during cardiac fusion is dependent on PI3K signaling. Furthermore, we find that PI3K signaling and Pdgfra genetically interact to facilitate cardiac fusion. Altogether our work supports a model by which intrinsic Pdgfra-PI3K signaling regulates the formation of membrane protrusions that facilitate the collective movement of the myocardium towards the midline. Insight into the balance of extrinsic and intrinsic influences for directing collective movement of myocardial cells has implications for understanding a wide set of congenital and environmental cardiac defects as well as the pathogenic mechanisms of diseases broadly associated with collective movement.

Results

The PI3K pathway is required for proper cardiac fusion

In a search for intracellular signaling pathways that are important for cardiac fusion we examined the role of the phosphoinositide 3-kinase (PI3K) signaling pathway, by pharmacological inhibition of PI3K with LY294002 (LY) (35). Treatments were started at the bud-stage (10 hours post-fertilization - hpf), in order to exclude effects on mesodermal cells during gastrulation (36). In wild-type or DMSO-treated embryos, bilateral myocardial populations move towards the midline and merge to form a ring structure between 20–21 hpf, which corresponds to the 20–22 somite stage (s) (Fig. 1A, A’, F). However, in LY-treated embryos myocardial movement is disrupted and the bilateral myocardial populations fail to properly merge by 22s (Fig. 1B, B’, F, Suppl. Fig. 1A–C, M). To rule out possible off-target phenotypic artifacts of LY (37), we exposed bud stage embryos to two other PI3K inhibitors, Dactolisib (Dac) or Pictilisib (Pic) (38, 39). Exposure to either of these inhibitors also causes cardiac fusion defects (Fig. 1C, C’, F, G, Suppl. Fig. 1D–F, N; Fig. 1D, D’, F, G, Suppl. Fig. 1G–I, O, respectively), as does the mRNA injection of a truncated form of p85 (Fig. 1E, E’, F, G, Suppl. Fig. 1J–L, P), which acts as a dominant negative inhibitor of PI3K (dnPI3K) activity (40). Furthermore, to ensure our analysis was not complicated by a developmental delay, we used developmentally stage-matched embryos (somite stage) rather than time-matched embryos (hours post-fertilization; hpf) to assess cardiac fusion phenotypes (see Suppl. Fig. 2 for embryos analyzed at 20 hpf).

Fig. 1. The PI3K pathway is required for cardiac fusion.

A-E Dorsal views, anterior to the top, of the myocardium labeled with myl7 (A-E) at 22 somite stage (s) or Tg(myl7:egfp) (A’-E’) at 20s. In contrast to a ring of myocardial cells in DMSO-treated embryos (A, A’), in embryos treated with PI3K inhibitors LY294002 (LY, B, B’), Dactolisib (Dac, C, C’), or Pictilisib (Pic, D, D’) at bud stage or injected with dnPI3K mRNA (750pg) at the one-cell stage (E, E’) cardiac fusion fails to occur properly with embryos displaying either cardia bifida (B, C) or fusion only at the posterior end (D, E). F, G Graphs depict the percentage (F) and range (G) of cardiac fusion defects in control and PI3K-inhibited embryos. Dots represent the percent of embryos with cardiac defects per biological replicate. Total embryos analyzed n = 37 (DMSO), 31 (20μM LY), 39 (40μM Dac), 38 (50μM Pic), 86 (dnPI3K). Blue – cardiac ring/normal; Orange – fusion only at posterior end/mild phenotype, Red – cardia bifida/severe phenotype. H Representative immunoblot and ratiometric analysis of phosphorylated Akt (pAkt) to Akt protein levels in DMSO and LY treated embryos reveals a dose-dependent decrease in PI3K activation. Bar graphs indicate mean ± SEM, dots indicate pAKT/AKT ratio per biological replicate, normalized to DMSO. Three biological replicates per treatment. One-Way ANOVA tests – letter changes indicate differences of p < 0.05 (F, H). Scale bars, 40 μm (A-E), 42 μm (A’-E’). Raw data and full p-values included in the source file.

We also examined the morphology of the cardiac ring in PI3K-inhibited embryos and cellular processes known to be regulated by PI3K signaling. During the later stages of cardiac fusion as part of the subduction process, medial myocardial cells form a contiguous second dorsal layer (26) and develop epithelial polarity in which intercellular junction proteins such as ZO1 are localized to the outer-edge of the myocardium (5, 6) (Suppl. Fig. 3A–C). In PI3K-inhibited embryos, we found that myocardial cells form this second dorsal layer however, the localization of polarity markers and the tissue organization can appear mildly disorganized (Suppl. Fig. 3D–F). Furthermore, the PI3K signaling pathway is known to promote cell proliferation and cell survival (29) however, we did not find a difference in the number of cardiomyocytes in DMSO- or LY- treated embryos at 20s (Suppl. Fig. 3G–I). Similarly, no apoptotic cardiomyocytes were observed in DMSO- nor in 20 μM LY- treated embryos (n = 17, 19 embryos, respectively from 3 biological replicates). Apoptotic cardiomyocytes were however observed in DNAse-treated controls. These experiments reveal that PI3K signaling is required for proper cardiac fusion.

The extent and duration of PI3K inhibition determines the penetrance and severity of cardiac fusion defects.

PI3K-inhibited embryos display cardiac phenotypes at 22s that range from severe, in which the myocardial populations remain entirely separate (cardia bifida) (Fig. 1G – red; examples - Suppl. Fig 1C, F, I, L), to more mildly affected hearts in which the myocardial populations form a U-shaped structure, having merged at the posterior but not anterior end (Fig. 1G – orange; examples Suppl. Fig. 1B, E, H, K). A subset of the PI3K-inhibited embryos also appear phenotypically normal (~25% for 20 μM LY, Fig. 1F, G) indicating incomplete penetrance. Increasing concentrations of PI3K inhibitor or dnPI3K mRNA increases the severity and penetrance of these phenotypes in a dose-dependent manner (Suppl. Fig. 1.). Similarly, we confirmed that LY inhibits PI3K activity in a dose-dependent manner, as measured by the ratio of phosphorylated AKT (pAKT) to AKT (Fig. 1H). AKT is phosphorylated as a direct consequence of PI3K activity (41). Thus, the severity and penetrance of cardiac fusion defects depends on the efficacy of PI3K inhibition.

Since differing modes of movement (9) as well as cellular processes such as MET (5, 6) and subduction (26) occur at different times during cardiac fusion, we also evaluated the developmental stages over which PI3K signaling is required. Short exposures (<3 hours) just prior to 22s or starting at bud stage had no effect on cardiac fusion. However, progressively longer times of exposure ending at 22s or starting at bud stage result in correspondingly more severe phenotypes and higher penetrance (Fig. 2A, B). These addition and wash-out experiments indicate that both the severity and penetrance of cardiac fusion phenotypes correlate with the duration of LY-incubation and not a specific developmental stage inside the 3–20s window. Thus, myocardial movement is responsive to both the levels and duration of PI3K signaling throughout cardiac fusion.

Fig 2: PI3K is required in the myocardium throughout cardiac fusion.

A, B Graphical representation of the PI3K inhibitor addition (A) and wash-out (B) experiments used to determine the developmental stage over which PI3K is required. In (A) LY is added to embryos at different developmental stages and incubated until 22s, when cardiac fusion is assessed. In (B) LY is added at bud stage and washed-out at different developmental stages, after which embryos are incubated in normal media till 22s, when cardiac fusion is assessed. Bar graphs indicate the average proportion of embryos displaying different phenotypes. Blue – cardiac ring/normal; Orange – fusion only at posterior end/mild phenotype, Red – cardia bifida/severe phenotype. n = 45 embryos per treatment condition from three biological replicates. C Schematic outlines experimental design to test requirement for PI3K in the myocardium. Pink – cells with the Tg(myl7:dnPI3K) transgene. F0 animals are mosaic for the transgene, while all cells in F1 embryos either have the transgene (pink) or do not (clear). The myl7 promoter restricts dnPI3K expression to the myocardium in Tg(myl7:dnPI3K) embryos. D-G Dorsal view of the myocardium labeled with myl7 in embryos at 22s from 4 different founder pairs (D-D’, E-E’, F-F’, G-G’). F1 embryos without the Tg(myl7:dnPI3K) transgene (as determined by genotyping) display normal cardiac fusion (D-G, n = 23/24, 16/16, 16/16, 16/16, per founder pair), while F1 siblings with the Tg(myl7:dnPI3K) transgene display cardiac fusion defects (D’-G’, n = 6/6, 13/13, 11/11, 13/13), indicating that PI3K signaling is required in myocardial cells. H Graph indicating the average % of wild-type and Tg(myl7:dnPI3K)+ embryos with cardiac fusion defects. Letter difference indicates a significant Fisher’s exact test p = 5.56 × 10⁻³¹. Scale bar, 40μm.

PI3K signaling is required in the myocardium for proper cardiac fusion.

Mutations affecting the specification or morphology of the anterior endoderm result in myocardial movement defects (13, 19, 42, 43), revealing a non-autonomous role for the anterior endoderm in cardiac fusion. However, when PI3K signaling is inhibited with 15 or 25 μM LY starting at bud stage we did not observe differences in the expression of endoderm markers such as axial/foxa2 or Tg(sox17:egfp) compared to DMSO-treated embryos (Suppl. Fig 3J–L, N–P). Additionally, the overall morphology of the anterior endoderm appeared intact and the average anterior endoderm width was similar between PI3K-inhibited and DMSO-treated embryos (Suppl. Fig 3M, Q).

To determine if PI3K signaling is specifically required within the myocardium, as opposed to the endoderm, we created a myocardial-specific dominant negative transgenic construct, Tg(myl7:dnPI3K). Our experimental design is outlined in Fig. 2C. In F1 embryos at 22s we observed embryos with normal cardiac rings and embryos with cardiac fusion defects (Fig. 2D–G’). Genotyping revealed that F1 embryos with normal cardiac rings (Fig. 2D–G) did not have the transgene (n = 71/71), while almost all sibling embryos with cardiac fusion defects (Fig. 2D’–G’) were positive for the Tg(myl7:dnPI3K) transgene (n=40/41). And all embryos with the Tg(myl7:dnPI3K) transgene have a cardiac fusion defect (Fig. 2H). (F1 embryos from 4 independent founder pairs were analyzed since stable transgenics could not be propagated due to loss of viability, likely due to a requirement for PI3K signaling in cardiac contraction at later stages (44)). Statistical analysis reveals that the Tg(myl7:dnPI3K) transgene is significantly associated with a cardiac fusion defect (Fisher’s test p = 5.56 × 10⁻³¹), indicating that PI3K signaling acts in the myocardium to regulate its movement during cardiac fusion.

PI3K signaling is responsible for the steering and velocity of myocardial movements during cardiac fusion.

Our analysis points to a role for PI3K signaling in the movement of myocardial cells. To identify the properties of myocardial movement regulated by PI3K signaling, we analyzed myocardial movement by performing in vivo time-lapse imaging with the Tg(myl7:egfp) transgene, which labels myocardial cells. A time-series using hand2 expression to compare myocardial movement in DMSO-treated and PI3K-inhibited embryos reveals dramatic differences in myocardial movement beginning after 12s (Suppl. Fig. 4). We thus focused our time-lapse imaging on the 14–20s developmental window. In time-lapse movies of DMSO-treated embryos, myocardial cells display coherent medially directed movement (Fig. 3A–B, E, Suppl. Fig. 5A–A”’, Video-1) with an average velocity of 0.2334 ± 0.007 microns/min, which is consistent with previous studies (9, 28). In PI3K-inhibited embryos myocardial cells also display coherent, coordinated movement and do move in the general direction of the midline, however they make dramatically less progress (Fig. 3C–D, Suppl. Fig. 5B–B”’, Video-2). Quantitative analysis of these myocardial cell tracks reveals that myocardial cells are slower (0.1879 ± 0.008 microns/min) and less efficient. Therefore, myocardial cells ultimately move less in LY-inhibited embryos compared to DMSO-treated embryos (Fig. 3E, F, Video-3). Differences in velocity occur throughout cardiac fusion (Suppl. Fig. 5C). However, the most dramatic difference between PI3K-inhibited and DMSO-treated myocardial cells is in the direction of their movement. Tracks of myocardial cells in DMSO-treated embryos are predominately oriented in a medial direction (average of 31.1 ± 1.65 degrees), while tracks in LY-treated embryos are mostly oriented in an angular anterior direction (60.6 degrees ± 1.73, p-value = 2.77 × 10⁻¹², Fig. 3G, H). Differences in directional movement occur mainly in the early stages of cardiac fusion when wild-type myocardial movement is mostly medial (Suppl. Fig. 5D). Together this analysis of myocardial cell tracks suggests that PI3K signaling is responsible for both steering and propelling myocardial cells towards the midline.

Fig 3: PI3K signaling regulates the medial movement and velocity of the myocardium during cardiac fusion.

A-D Time points from a representative time-lapse video of myocardial cells visualized with the Tg(myl7:egfp) transgene in embryos treated with DMSO (A, B, Video-1) or 20μM LY (C, D, Video-2) from bud - 22s. 3D reconstructions of confocal slices (A, C) reveal the changes in conformation and location of the myocardium at 3 major stages of cardiac fusion: early medial movement towards the embryonic midline (A-A’, C-C’), posterior merging of bilateral populations (A”, C”) and anterior merging to form a ring (A”’, C”’). Representative tracks (B, D) show the paths of a subset of myocardial cells over ~2.5 hr timelapse. Yellow dots indicate the starting point of each track. E-H Graphs depict box-whisker plots of the velocity (E), efficiency index (F) and angle of movement (H) of myocardial cells. The direction of movement is visualized by rose plots (G). Myocardial cells in PI3K-inhibited (LY-treated) embryos show an overall direction of movement that is angular (60–90 degrees) and is slower than in DMSO-treated embryos. Scale bar, 60μm. 96 and 125 cells were analyzed from five DMSO- and six 20μM LY- treated embryos, respectively. Two sample t-test, letter change indicates p < 0.05. Raw data and full p-values included in the source file.

Myocardial membrane protrusions are medially polarized by PI3K signaling

The role of PI3K signaling in regulating the polarity of migratory protrusions in the dorsal epithelium in Drosophila and prechordal plate in zebrafish (36, 45) along with previous reports of the existence of myocardial membrane protrusions (7, 26), led us to next look for these protrusions in myocardial cells during cardiac fusion and to examine if they are disrupted in PI3K-inhibited embryos. To visualize membrane protrusions in the myocardium, we performed in vivo time-lapse imaging during cardiac fusion of embryos injected with myl7:lck-egfp plasmid DNA in order to mosaically label the plasma membrane of myocardial cells. Despite myocardial cells being connected via intercellular junctions (5, 28), we observed that the lateral edges of myocardial cells in wild-type/DMSO-treated embryos are highly dynamic; transitioning from appearing smooth and coherent to undulating and extending finger-like membrane protrusions away from the cell (Fig. 4A–A””, Video-4). These protrusions are dynamic, actively extending and retracting, and are prevalent occurring on average 20.3 ± 6.7 times per hour per cell and lasting for an average of 2.3 ± 0.6 mins (Fig. 4A). In LY-treated embryos we observed similar membrane protrusions extending from myocardial cells (Fig. 4B,Video-4), which occur at a similar rate (17 ± 7.4 per hour per cell, p-value = 0.36), but with slightly longer persistence (3.23 ± 0.84 mins, p-value = 0.008).

Fig 4: Myocardial membrane protrusions are misdirected in PI3K-inhibited embryos.

A-B”” timepoints from representative timelapse videos (see Video-4) of myocardial cells whose membrane has been labeled with myl7:lck-eGFP (black), medial to the right, in a DMSO- (A-A””) or a 20μM LY- (B-B””) treated embryo. Red arrowheads indicate representative protrusions, which are oriented medially, coincident with the direction of movement in DMSO-treated embryos (A-A””) but are oriented in all directions in LY-treated embryos (B-B””). C, D Rose (C) and Bar (D) graphs displaying the orientation of membrane protrusions in DMSO- (left) or LY- (right) treated embryos. The length of each radial bar in (C) represents the percentage of protrusions in each bin. Bar graph displays the total percentage of forward or backward protrusions. Forward protrusions: 270–90 degrees, pink. Backward protrusions: 90–270 degrees, black. n = 425 protrusions from 11 cells in 5 embryos (DMSO), and 480 protrusions from 11 cells in 4 embryos (20μM LY). Fisher’s exact test, P value < 0.0001. Error bars, mean ± SEM. Scale bar, 30 μm. Raw data and full p-values included in the source file.

We further observed that in DMSO-treated embryos membrane protrusions occur predominantly in the medial direction (77.25 ± 21.76% of protrusions were in the forward direction, Fig. 4A–A””, C, D), suggesting an association with the medial movement of the myocardial tissue. In contrast, in LY-treated embryos myocardial membrane protrusions do not display the same medial polarity, instead extending from all sides of a myocardial cell equally (only 46 ± 11.6% of protrusion were in the forward direction, Fig 4B–B””, C, D). The finding that myocardial membrane protrusions are medially polarized in wild-type embryos but not in PI3K-inhibited embryos where myocardial cells are misdirected and slower to reach the midline suggests that PI3K signaling helps to steer and propel myocardial cells towards the midline through the polarization of these active protrusions.

PI3K signaling is regulated by Pdgfra during cardiac fusion

The improperly directed myocardial cells in PI3K-inhibited embryos (Fig. 3) are reminiscent of the steering defects observed in pdgfra mutant embryos (28). This similarity led us to investigate whether Pdgfra activates PI3K signaling to regulate myocardial movement. We found that PI3K activity as measured by the ratio of phospho-AKT to AKT levels (41), is severely diminished in pdgfra mutant embryos during cardiac fusion (Fig. 5A). Conversely, when Pdgfra activity is increased during cardiac fusion through the over-expression of pdgf-aa, PI3K activity is up-regulated (Fig. 5B).

Fig 5: Pdgfra activates and genetically interacts with PI3K signaling to regulate cardiac fusion.

A, B Immunoblot and ratiometric analysis of phosphorylated Akt (pAkt) compared to total Akt levels reveals reduced pAkt levels in loss-of-function pdgfra^sk16 heterozygous (−/+) or homozygous (−/−) mutant embryos at 22s (A), and elevated pAkt levels at 22s when PDGF signaling is activated with the hs:pdgfaa transgene (B). Bar graphs display averages from three separate experiments. C-E Dorsal views, anterior to the top, of the myocardium labeled with myl7 at 22s. In contrast to a normal ring of myocardial cells in wild-type embryos treated with 10μM LY starting at bud stage (C) or pdgfra heterozygous embryos (D), when pdgfra heterozygous mutants are exposed to 10μM LY, cardiac fusion is defective with embryos displaying severe phenotypes such as cardia bifida (E). F Bar graph depicts the average distribution of cardiac fusion defects in DMSO-treated wild-type and pdgfra heterozygous mutants as well as 10μM LY-treated wild-type and pdgfra heterozygous mutant embryos. The total number of embryos examined over three separate biological replicates are 47 (DMSO, +/+), 25 (DMSO, −/+), 36 (10μM LY, +/+), and 31 (10μM LY, −/+). Blue - cardiac ring/normal; Orange - fusion only at posterior end/mild, Red - cardia bifida/severe. Bar graphs, mean ± SEM. One-way ANOVA (A, C) or Student’s T-test (B), letter change indicates p < 0.05. Scale bar, 60μm. Raw data and full p-values included in the source file.

To determine if Pdgfra’s influence on PI3K activity is important for myocardial movement towards the midline, we examined whether these two genes genetically interact while regulating cardiac fusion. When pdgfra heterozygous mutant embryos are exposed to DMSO cardiac fusion occurs normally (Fig. 5D, F), even though there is a small reduction in PI3K activity (Fig. 5A). When wild-type embryos are exposed to 10μM LY, PI3K activity is modestly reduced (Fig. 1H) and a small percent of embryos display mild cardiac fusion defects (Average of 10.9 ± 7.39% of 10μM LY-treated embryos display mild U-shaped cardiac fusion defects, n = 36, 3 replicates, Fig 5C, F). However, when pdgfra heterozygous mutant embryos are exposed to 10μM LY, there is a synergistic increase in both the severity and penetrance of cardiac fusion defects. 100% of pdgfra heterozygous embryos exposed to 10μM LY display cardiac fusion defects, with the majority of embryos displaying severe cardia bifida phenotypes (Fig. 5E, F). Together these results suggest that PDGF signaling activates PI3K activity to promote myocardial movement towards the midline.

Discussion

Our studies reveal an intrinsic PI3K-dependent mechanism by which the myocardium moves towards the midline during the formation of the primitive heart tube. Together with our previous studies revealing a role for the PDGF pathway in facilitating communication between the endoderm and myocardium (28), our current work suggests a model in which Pdgfra in the myocardium senses signals (PDGF ligands) from the endoderm and via the PI3K pathway directs myocardial movement towards the midline through the production of medially oriented membrane protrusions. While genetic and imaging studies in zebrafish and mice (5, 13, 18, 19, 25–27, 46, 47) along with embryological studies in chicks and rats (22–24, 48) have identified the importance of extrinsic influences – such as the endoderm and extracellular matrix, on myocardial movement to the midline, our studies using tissue-specific techniques identifies an active role for myocardial cells, providing insight into the balance of intrinsic and extrinsic influences that regulate the collective movement of the myocardial tissue during heart tube formation.

Specifically, we found a requirement for PI3K signaling in cardiac fusion which is complemented by previous studies in mice examining Pten, an antagonist of PI3K (29). Pten mutant mice also display cardia bifida (33), indicating that appropriate PIP3 levels and localization are required for proper cardiac fusion. Our spatial and temporal experiments further build on these studies by revealing a requirement for PI3K specifically in the myocardium and throughout the duration of cardiac fusion (Fig. 2). We also observed a mild disorganization of the sub-cellular localization of intercellular junctions in the myocardium of PI3K-inhibited embryos (Suppl. Fig. 3). This finding is consistent with previous studies linking epithelial polarity to PI3K signaling (49). However, myocardial cells defective in apical-basal polarity still form a cardiac ring (50, 51), suggesting that an apical-basal defect is unlikely to be the primary reason for myocardial movement defects. Instead, our studies showing that PI3K-inhibited myocardial cells move slower and most prominently are misdirected during the early stages of cardiac fusion indicate a role for PI3K signaling in the steering of myocardial movements medially towards the midline. Our finding that steering in PI3K-inhibited embryos is perturbed in the early stages of cardiac fusion is furthermore consistent with the different phases of myocardial movement identified by Holtzman et al. (9) and suggests that PI3K signaling could be part of a distinct molecular mechanism that drives these early medial phases of myocardial movement. We also found that similar to loss-of-function pdgfra mutants, inhibition of PI3K signaling causes defects in directional movement. However, inhibition of PI3K signaling affects myocardial velocity and efficiency (~ 20% μ/min decrease in velocity, and a 25% decrease in efficiency compared to DMSO-treated embryos) more noticeably than pdgfra mutants, in which no significant difference in velocity or efficiency were detected (28). These differences could simply be a result of differences in the extent of PI3K inhibition by 20μM LY compared to extent of loss-of-pdgfra function by the ref mutation. Alternatively, similar to the role of PI3K signaling in the velocity of gastrulating mesoderm cells as well as in migrating dictyostelium and neutrophil cells (30, 31, 36, 52) these differences could also indicate a Pdgfra-independent PI3K function in regulating the velocity of myocardial movement.

Myocardial membrane protrusions were postulated by De Haan et al. ~50 years ago as a mechanism by which myocardial cells move towards the midline (53). Here using mosaic membrane labeling of myocardial cells to visualize membrane protrusions, we have observed myocardial membrane protrusions that are oriented in the medial direction in a PI3K-dependent manner, confirming his hypothesis. These studies are complemented by previous studies in zebrafish which have observed myocardial protrusions prior to and after cardiac fusion (26, 54) as well as recent studies in the mice (7) indicating that these cellular processes are likely conserved. Indeed, similar observations of PI3K signaling orienting and stimulating protrusion formation in migrating Dictyostelium and neutrophil cells as well as in the collective movement of endothelial tip cells, the prechordal plate and the dorsal epithelium (30, 31, 36, 45, 55, 56) support a conserved role for PI3K signaling in regulating protrusion formation.

However, the question of how active membrane protrusions facilitate the collective medial movement of the myocardium to the midline remains to be addressed. Our studies indicate that directionality and to a lesser extent velocity and efficiency are compromised, when membrane protrusions are improperly oriented in PI3K-inhibited embryos (Fig. 3). These observations could suggest that the observed membrane protrusions are force generating, similar to protrusions from leader cells in the lateral line or in endothelial and tracheal tip cells (57–59). Alternatively, these protrusions could act more like filopodia sensing extrinsic signals and the extracellular environment (60). Future studies examining myocardial protrusions and their role in the biomechanical dynamics of the myocardium will help to elucidate the role of membrane protrusions in the collective movement of the myocardium during cardiac fusion.

Overall, our studies delineate a role for the PDGF-PI3K pathway in the mechanisms by which myocardial precursors sense and respond to extracellular signals to move into a position to form the heart. These mechanisms are likely relevant to other organ progenitors including neural crest cells, endothelial precursors, endodermal progenitors, and neuromasts, all of which must move from their location of specification to a different location for organ formation. Although varying in their morphogenesis, many of these movements are collective in nature. Indeed, a similar Pdgfra-PI3K signaling cassette is important in the collective directional migration of several organ progenitors including the migration of mesoderm and neural crest cells (36, 61–67). Receptor tyrosine kinase (RTK)-PI3K pathways are also important across several cardiac developmental processes, including epicardial development, cardiac neural crest addition, cardiomyocyte growth, cardiac fibroblast movement and cardiomyocyte contraction (44, 68–72). Similarly, PDGF-PI3K and more generally RTK-PI3K signaling cassettes are activated in several diseases including glioblastomas, gastrointestinal stromal tumors and cardiac fibrosis (73–76). Thus, the role of this RTK-PI3K cassette in sensing and responding to extracellular signals is likely to be broadly relevant to the etiology of a wide array of developmental processes as well as congenital diseases.

Materials and methods

Zebrafish husbandry, microinjections and plasmid construction:

All zebrafish work followed protocols approved by the University of Mississippi IACUC (protocol #21–007). Wildtype embryos were obtained from a mixed zebrafish (Danio rerio) AB/TL background. The following transgenic lines of zebrafish were used: Tg(myl7: eGFP)^twu34 (RRID:ZFIN_ZDB-GENO-050809–10)(77), Tg(sox17:eGFP)^ha01 (ZFIN_ZDB-GENO-080714–2)(78), Tg(hsp70l:pdgfaa-2A-mCherry;cryaa:CFP)^sd44 abbreviated hs:pdgfaa (ZDB-GENO-170510–4), and ref (pdgfra^sk16) (ZDB-GENO-170510–2) (28). All embryos were incubated at 28.5 °C unless otherwise noted. Transgenic Tg(myl7:dnPI3K; Cryaa:CFP) F0 founders were established using standard Tol2-mediated transgenesis (79). F0 founders pairs were screened by intercrosses looking for a high percentage of F1 embryos with CFP+ eyes and cardiac edema. Stable transgenic lines could not be propagated due to loss of viability. Based on the germline mosaicism of the F0 parents, only a proportion of the F1 embryos are expected to have the transgene. Embryos from 4 different F0 pairs were analyzed for cardiac fusion phenotypes. Due to germ-line mosaicism F1 embryos were genotyped after in situ hybridization for the presence of the transgene using standard PCR genotyping. Primer sequences are provided in Suppl. Table 1.

Truncated p85 (dnPI3K) capped mRNA was synthesized from the pBSRN3-Δp85 construct (40) and injected at the 1-cell stage. To mosaically label cells in the myocardium for protrusion imaging, myl7:lck-eGFP (30 ng/μl) DNA was injected along with Tol2 transposase (40 ng/μl) into Tg(myl7:eGFP) heterozgyous embryos at the 1 cell stage and embryos were subsequently allowed to develop at 28.5 °C.

Plasmids were constructed by using gibson assembly (NEB, E2621) to transfer lck-eGFP (80) or a truncated version of p85 (40) into the middle-entry vector of the tol2 gateway system (81), which were verified by sequencing. Primer sequences are provided in Suppl. Table 1. Then gateway recombination between p5E-myl7 promoter, the constructed middle-entry clones, p3E-polyA and either pDESTTol2pA2 (81) or pDESTTol2pA4-Cryaa:CFP (28) was used to produced plasmids containing myl7:lck-eGFP or myl7:dnPI3K; Cryaa:CFP, respectively.

Inhibitor treatments:

The following inhibitors were used: LY294002 (LY, Millipore-Sigma 154447-36-6), Dactolisib (Dac, Millipore-Sigma 915019-65-7), and Pictilisib (Pic, Millipore-Sigma 957054-30-7). For each treatment, inhibitors were freshly diluted serially from stocks such that the same percentage (0.1%) of DMSO (Goldbio 67-68-5, Millipore-Sigma 67-68-5) was added to 1X E3 in glass vials (Fisherbrand 03-339-22B). 0.1% DMSO was used as a control. 15 dechorionated embryos per vial were incubated in the dark at 28.5 °C. In the course of these studies, we noticed that incubation with pharmacological PI3K-inhibitors caused a delay in trunk elongation and somite formation along with defects in cardiac fusion (Suppl. Fig. 2). To ensure our analysis was not obfuscated by a developmental delay, we used somite number to stage match embryos. PI3K-inhibited embryos thus develop approximately 2–3 hours longer than DMSO-treated embryos, prior to analysis.

Immunoblot, Immunofluorescence, in situ hybridization:

Embryos at 22s were prepared for immunoblots by deyolking (82). Primary and secondary antibodies include phospho-AKT (1:2000, Cell Signaling 4060, RRID: AB_2315049) and pan-AKT (1:2000, Cell Signaling 4691, RRID: AB_915783), Anti-rabbit HRP-conjugated (1:5000, Cell Signaling 7074, RRID: AB_2099233). pAKT and AKT immunoblots were visualized (Azure 600 Imaging system, Azure Biosystems) and quantified using ImageJ (83) by calculating the ratio of pAkt to Akt. Ratios were normalized to DMSO. To identify pdgfra/ref heterozygous and homozygous embryos, embryo trunks were clipped and genotyped as described (28). The body of the embryo including the heart was snapped frozen and stored at −80 °C. After genotyping, they were pooled via their genotype and analyzed via immunoblot. To activate Pdgfra, embryos expressing the Tg(hsp701: pdgfaa-2A-mCherry) transgene were heat-shocked at bud stage as described (28) and collected at 22s.

Immunofluorescence performed on transverse sections used standard cryoprotection, embedding and sectioning (46). Primary, secondary antibodies and dyes include: anti-GFP (1:1000, Abcam ab13970, RRID: AB_300798), anti-ZO-1 (1:200, Thermo Fisher Scientific 33–9100, RRID: AB_87181), donkey anti-chicken-488 (1:300, Thermo Fisher Scientific A32931TR, RRID: AB_2866499), donkey anti-mouse-647 (1:300, Thermo Fisher Scientific A32728, RRID:AB_2633277). TUNEL was performed using the Cell Death detection kit, TMR red (Millipore Sigma 12156792910). Addition of DNaseI was used to confirm we could detect apoptotic cells.

In situ hybridization was performed using standard protocols (Alexander et al., 1998), with the following probes: myl7 (ZDB-GENE-991019–3), axial (ZDB-GENE-980526–404) and hand2 (ZDB-GENE-000511–1). Images were captured with Zeiss Axio Zoom V16 microscope (Zeiss) and processed with ImageJ.

Fluorescence Imaging:

To analyze cardiac fusion (Fig. 1A’–E’) Tg(myl7:eGFP) embryos were fixed, manually deyolked and imaged with a Leica SP8X microscope. To analyze the anterior endoderm (Suppl. Fig. 3N–P) Tg(sox17: eGFP) embryos were fixed and imaged with an Axio Zoom V16 microscope (Zeiss).

For live imaging, Tg(myl7:eGFP) embryos were exposed to DMSO or 20μM LY at bud stage and mounted at 12 somite stage as described (84). Mounted embryos were covered with 0.1% DMSO/20μM LY in Tricaine-E3 solution and imaged using a Leica SP8 X microscope with a HC PL APO 20X/0.75 CS2 objective in a chamber heated to 28.5 °C. GFP and brightfield stacks were collected approximately every 4 min for 3 hours. After imaging, embryos were removed from the mold and incubated for 24 hrs in E3 media at 28.5 °C. Only embryos that appeared healthy 24 hours post imaging were used for analysis. The tip of the notochord was used as a reference point to correct embryo drift in the Correct 3D direct ImageJ plugin (85). Embryos were handled similarly for imaging protrusions, except 15 confocal slices of 1μm thickness were collected every 1.5 min with a HC PL APO 40X/1.10 CS2 objective.

Image analysis:

Embryonic length (Suppl. Fig. 2) was measured from the anterior tip of the head to the posterior tip of the tail of each embryo using the free-hand tool of ImageJ. The endoderm width was measured 300 microns anterior from the posterior point of intersection of the two sides of the endoderm. The distance between the hand2 expressing domains was measured at three equidistant positions (~200 microns apart) along the anterior posterior axis. Tg(myl7:eGFP)+ cardiomyocytes were counted from blinded and non-blinded 3D confocal images of 20s embryos from 4 biological replicates using the cell counter addon in ImageJ. No difference between the blinded (1) and non-blinded (3) replicates was detected.

For live imaging of cell movements – the mTrackJ addon in ImageJ (86) was used. 20–25 cells per embryo whose position could be determined at each timepoint were chosen from the two most medial columns of myocardial cells on each side of the embryos. From these tracks, cell movement properties including overall displacement, velocity (displacement/time), efficiency (displacement/distance) and direction (atan(Δy/Δx)×57.295) were calculated. Rose plots in Fig. 3 display the direction of movement of the overall trajectory of individual cells. In these plots individual cells are grouped into 6 bins based on their net direction of movement; the length of each radial bar represents the percentage of cells in each bin.

For live imaging of myocardial membrane protrusions – stacks were processed in Leica LASX and/or Imaris Viewer (Bitplane) to position the medial edge to the right of the image. Videos of the myocardium were inspected frame by frame in ImageJ for a protrusion. Only cells that were not neighbored by other labeled cells on their medial and lateral edges were analyzed. The direction of protrusion was measured using the “straight line” function a line to draw a line from the bottom of the protrusion to the tip. All protrusions of each cell over the entire recording were measured. Graphs, cartoons and figures were created with Prism (Graphpad), Excel (Microsoft), and Indesign (Adobe).

Statistics and replication:

All statistical analysis was performed in R or Prism (Graphpad). Sample sizes were determined based on prior experience with relevant phenotypes and standards within the zebrafish community. Deviation from the mean is represented as standard error mean or box-whisker plots. In box-whisker plots, the lower and upper ends of the box denote the 25^th and 75^th percentile, respectively, with a horizontal line denoting the median value and the whiskers indicating the data range. All results were obtained from at least three separate biological replicates, blinded and non-blinded. All replicates are biological. Samples were analyzed before biological sex is determined (87). Raw data and full p-values included in the source file.

Supplementary Material

Supplement 1

Video 1. Myocardial cells in DMSO-treated embryos collectively move towards the midline and form a ring during cardiac fusion. A-C Representative time-lapse movie of myocardial cells visualized with Tg(myl7:egfp) during cardiac fusion in a DMSO-treated embryo (A), tracks show movement of selected cells from the timelapse video (B) and overlay of eGFP and tracks (C). Time-lapse images are of three-dimensional reconstruction of confocal slices taken at 4:32 min intervals for 2.5 hours, beginning at 14s.

Video 2. Myocardial cells in PI3K-inhibited embryos fail to move properly towards the midline. A-C Representative time-lapse movie of myocardial cells visualized with Tg(myl7:egfp) (A), tracks of selected cells (B), and overlay of tracks and eGFP (C) from an embryo treated with 20μM LY from bud-20s. Time-lapse was acquired as described in Video 1.

Video 3. PI3K signaling promotes the medial directional movement of myocardial cells towards the midline. A, B Side-by-side comparison of myocardial movement in DMSO- (A, video 1) and LY- (B, video 2) treated embryos reveals that inhibition of PI3K signaling by LY prevents myocardial cells from being adequately directed towards the midline. Selected analyzed tracks (white lines) overlaying 3D reconstructions of the Tg(myl7:egfp) transgene (green) in DMSO (A) and 20μM LY (B) treated embryos.

Video 4. Dynamic medially oriented myocardial membrane protrusions are lacking in PI3K-inhibited embryos. Representative time-lapse movies of myocardial membrane protrusions during cardiac fusion, visualized by injecting myl7:lck-eGFP plasmids at the 1-cell stage, in DMSO- (left panel) or 20μM LY – (right panel) treated embryos. Left panel highlights membrane protrusions (red arrowheads) in a set of posterior myocardial cells in a DMSO-treated embryo. Myocardial membrane protrusions in DMSO-treated embryos are mostly directed in the medial orientation (towards the right in both panels). Right panel highlights myocardial membrane protrusions (red arrowheads) in PI3K-inhibited embryos during cardiac fusion. Medial membrane protrusions (towards the right) are lacking in PI3K-inhibited embryos. DMSO or LY treatment from bud-20s. Time-lapse movies are 3D reconstruction of confocal images of membrane protrusions taken at ~90s intervals for 2 hours. Scale bar, 10μm.

1

Supplemental Fig. 1: The penetrance and severity of cardiac fusion defects in PI3K-inhibited embryos is dose-dependent. A-L Dorsal views, anterior to the top, of the myocardium labeled with myl7 at 22s. Incubation of embryos with LY (A-C), Dac (D-F), Pic (G-I) from bud stage to 22s or injection of embryos with dnPI3K mRNA (J-L) at the one-cell stage results in dose-dependent cardiac fusion defects at 22s. M-P Graphs depict the distribution of cardiac fusion defects in embryos treated with increasing concentrations of LY (M), Dac (N), Pic (O) or injected with increasing amounts of dnPI3K mRNA (P). Graphs reveal that both the percentage of embryos displaying cardiac fusion defects and the severity of those defects are dose-dependent. Total number of embryos analyzed (n) from > 3 treatments or injections at the indicated concentrations in (M-P): LY- 40, 40, 30, 31, 31; Dac: 38, 34, 39; Pic: 37, 39, 38, dnPI3K mRNA: 73, 52, 61, 57, 52, respectively. Dots indicate the percent of embryos displaying a specific phenotype per incubation. Blue - Cardiac ring/normal; Orange - fusion only at posterior end/mild, Red - cardia bifida/severe. Bar graphs, mean ± SEM. One-Way ANOVA comparing percent of cardiac fusion defects- letter change indicates p < 0.05. Scale = 60 μm. Full p-values included in the source file.

Supplemental Fig. 2: LY-incubation results in trunk extension and somite formation delays. A-B, D-E Lateral brightfield views of 20 hours post fertilization (hpf) embryos treated with DMSO (A, D) or 20μM LY (B, E) at bud stage. C, F Box-whisker plot depicting the average embryonic length (yellow curved line in A, B) or somite number (yellow dots in D, E) at 20 hpf. Total number of embryos (n) from > 3 separate incubations = 40 (DMSO), 40 (20μM LY) for (C), and 39 (DMSO), 42 (20μM LY) for (F). Dots = measurements from individual embryos. Two sample t-test; p-value = 4.527×10⁻⁴ and 7.624×10⁻⁵, respectively. G-H Dorsal views, anterior to the top, of the myocardium labeled with myl7 at 20 hpf. Embryos treated with DMSO at bud stage show cardiac rings (G) whereas those treated with 20μM LY show cardia bifida at 20 hpf (H). I Graph depicts the average percentage of cardiac fusion defects in embryos treated with DMSO or 20μM LY. The total number of embryos examined from three separate incubations (n) = 45 (DMSO), 45 (20μM LY). Two sample t-test; p-value = 4.56×10⁻⁵. Dots indicate the percent of embryos with cardiac fusion defects per incubation. Letter changes (C, F, I) indicate p-values < 0.05. Raw data and full p-values included in the source file.

Supplemental Fig 3: The morphologies of the myocardium and anterior endoderm are not compromised in PI3K-inhibited embryos. A-F Representative transverse cryosections, dorsal to the top, compare the morphology of the myocardium, visualized with Tg(myl7:eGFP) (green), ZO1 (purple) and DAPI (blue) between DMSO- (A-C) and 20μM LY- (D-F) treated (bud-20s) embryos. Box (A, D) indicate region magnified in B, C, E, F. G-I Representative 3D confocal images of the myocardium at 20s, which were used to count myocardial cells in DMSO- (G) or 20μM LY - (H) treated embryos. Yellow dots indicate individual myocardial cells counted using ImageJ. Graph depicts box-whisker plots for the average number of myocardial cells. 21 (DMSO) and 25 (LY) embryos from 4 separate bud-20s incubations were analyzed (I). J-Q Dorsal views, anterior to the top, of the anterior endoderm labeled with axial (J-L) or the Tg(sox17:eGFP) transgene (N-P) at 30s. Embryos incubated with either DMSO (J, N) or 15μM LY (K, O) or 25μM LY (L, P) from the bud stage to 30s show no observable difference in the appearance or width of the anterior endoderm. Box-whisker plots of the average width of the anterior endoderm labeled with either axial (M) or Tg(sox17:eGFP) (Q). 47 (axial) and 42 (Tg(sox17:egfp)) embryos per inhibitor concentration from three separate incubations were analyzed. Yellow lines: width of the endodermal sheet. Purple dots (I, M, Q) indicate individual embryos. Letter differences indicate a p-value < 0.05 as tested by 1-way ANOVA. Scale bars, 10 (A-F), 42 (G-H), 60 (J-L), and 50 (N-P) μm. Raw data and full p-values included in the source file.

Supplemental Fig. 4: Myocardial movement towards the midline is disrupted in PI3K-inhibited embryos throughout cardiac fusion. A-H Dorsal views, anterior to the top, of embryos displaying the expression of hand2 in the anterior lateral plate mesoderm (ALPM) at (A, E) 12s, (B, F) 14s, (C, G) 18s and (D, H) 20s, treated with either DMSO (A-D) or 20μM LY (E-H) at bud stage. I Box-whisker plots depict the average distance between the sides of the ALPM. Although, hand2 is properly expressed in LY-exposed embryos, ALPM convergence is affected as early as the 10s stage, with a dramatic difference in convergence starting at 12s. The total number of embryos analyzed (n), from 3 separate incubations at the noted stages (I) are: n = 34, 33, 31, 32, 34, 34 (DMSO); 32, 29, 30, 34, 31, 28 (20μM LY), respectively. Dots indicate the distance between ALPM sides per embryo. Student’s t-test: asterisk indicates p values < 0.05. Scale bar, 100μm. Raw data and full p-values included in the source file.

Supplemental Fig. 5. PI3K signaling directs myocardial movement during the early stages of cardiac fusion and regulates velocity throughout cardiac fusion. A-B Time-lapse confocal reconstructions from Figure 3 overlaid with cell movement tracks, starting at t = 0 (yellow dots). Scale bar, 60μm. C, D Box-whisker plots display the average velocity (C) and direction of movement (D) sub-divided at three distinct stages of cardiac fusion: early movement (0 – 48min), posterior fusion (49 – 99min) and anterior fusion (100 – 153min). The average velocity of myocardial cells in LY-treated embryos is consistently slower than the velocity of DMSO-treated embryos which is consistent throughout cardiac fusion (C). However, in LY-treated embryos myocardial cells display a more angular average direction of movement compared to DMSO-treated embryos during the early stages of cardiac fusion (Early movement, Posterior fusion), after which wild-type myocardial cell movement becomes more angular matching myocardial cell movements in LY-treated embryos. Two sample t-test, letter change indicates p < 0.05.