Live-imaging reveals Coordinated Cell Migration and Cardiac Fate Determination during Mammalian Gastrulation

Heart development involves the specification of distinct sets of cardiac progenitors at various times and locations during ontogeny. Here, we used live imaging in mice from the initiation of gastrulation to heart tube formation stages to investigate the origin and migratory paths of cardiac progenitors. We tracked individual mesodermal cells, reconstructing the lineage tree of the cells and fates for up to four generations. Our findings revealed independent unipotent progenitors originating at specific times that exclusively contribute to the left ventricle/atrioventricular canal (LV/AVC) or atrial myocytes. LV/AVC progeny differentiated into myocytes early, forming the cardiac crescent, while atrial progenitors differentiated later and contributed to the venous poles of the heart tube during morphogenesis. We also identified short-lived bipotent and tripotent mesodermal progenitors that contribute to a diverse array of mesodermal cell types, illustrating early plasticity during gastrulation. Sister cells generated by multipotent progenitors dispersed more and adopted more diverse migratory trajectories within the anterior mesoderm space than those from unipotent progenitors. Together the data reveal the regulation of individual cell directionality and cardiac fate allocation within the seemingly unorganised migratory pattern of mesoderm cells.


Introduction
How cell fate specification and morphogenesis are coordinated in time and space to generate tissues and organs of unique forms and functions is central to developmental biology.This is evident during gastrulation when mesodermal cells acquire diverse cardiac fates and engage in complex cell movements to generate spatial patterns, such as a cohesive cardiac crescent, which transforms into a primitive heart tube.
Clonal analysis establishes lineage relationships between progenitors and their derivatives.In the gastrulating mouse embryo, tracking the derivatives of single progenitors led to the finding that progenitors are assigned to specific anatomical locations in the heart prior to the formation of the heart fields [1][2][3][4].Unipotent Mesp1+ progenitors are solely destined for the left ventricle (LV) and atria myocardium can be distinguished from unipotent endocardium progenitors [2].
Additional clonal analysis of Hand1+ progenitors located at the embryonic/extraembryonic boundary in the early gastrulating embryo identified bipotent and tripotent progenitors.These progenitors generated LV/AVC myocytes in addition to pericardium, epicardium, and extraembryonic tissues [1].
One limitation of clonal analysis, however, is that the history of the cells is deduced by analysing descendants at the endpoint.It does not allow the identification of the progenitors' initial locations or subsequent migratory paths in the embryo.Single-cell tracking in liveimaging is needed for this and is the most rigorous approach to reconstituting cell lineages, identifying when cardiac progenitors become lineage-restricted during gastrulation and enabling migration analysis [5].
A recent live-imaging analysis uncovered the dynamics of mesodermal cell migration during mouse gastrulation [6].This analysis revealed that cells dispersed extensively in the embryo, with clearly separate movements of daughter cells, suggesting cell identity may not be fixed but instead influenced by the position of the cells at the end of the migration period.
However, as Dominguez et al. discussed, the motility of mesodermal cells is unlikely to be completely random.There may be some regulation of directionality of individual cell migration to ensure progeny migrate to their correct locations and establish spatial patterns, including the cardiac crescent and distinct LV/AVC and atrial progenitor domains [7][8][9].Indeed, a previous migration analysis showed mesodermal cell migrate with directionality during mouse gastrulation [10].Thus, early mammalian mesoderm migration may exhibit some degree of determinism.This is reminiscent of an evolutionarily distinct species, the ascidian, in which a small number of genealogically related and determined heart progenitors migrate with predetermined directionally [11].However, it is not known whether progenitors adopt more stereotypical migratory trajectories, once committed to specific cardiac fates in the context of mammalian gastrulation.
Here, through long-term live imaging and single-cell tracking in mice, spanning from the initiation of gastrulation to the stages of heart tube formation, our goal was to reconstitute the lineage tree of cells and assess how the migratory paths of cells relate to their eventual cardiac fate within the seemingly unorganised migration pattern.

Development and characterisation of cTnnT-2a-eGFP mice
To track cardiomyocytes in vivo, we developed a knock-in mouse reporter line cTnnT-2a-eGFP where the eGFP sequence is inserted downstream of the endogenous cardiac troponin T (cTnnT) loci.A virus-derived 2a self-cleaving peptide inserted between the cTnnT and eGFP coding sequence ensures co-expression of both cTnnT and eGFP proteins (Fig. 1 A) [12].The cTnnT-2a-eGFP line was maintained as homozygotes.Animals are viable and indistinguishable from heterozygotes.Whole-mount immunostaining for cTnnT confirmed specific eGFP expression in cTnnT+ cardiomyocytes at E8 -heart tube stage-and E12.5 (Fig. 1 B-C).
We first analysed cardiac differentiation dynamics in real time using multiphoton liveimaging and the cardiomyocyte cTnnT-2a-eGPF reporter line (Fig. 1 D).We combined the cTnnT-2a-eGPF reporter with the Bre:H2BCerulean BMP reporter [7], which expressed cerulean in the lateral plate mesoderm.We found initial sparse GFP positive cells appearing within the Bre-Cerulean positive lateral plate mesoderm at E7.5 -consistent with initial sparse cTnnT protein distribution found in the lateral plate mesoderm [13].This was followed by the establishment of the cardiac crescent epithelium-like structure and primitive heart tube (Fig. 1 D).We conclude that the cTnnT-2a-eGFP reporter faithfully identifies cardiomyocytes among a population of lateral plate mesodermal cell derivatives.

Establishing long term light-sheet live microscopy for cardiac lineages analysis
Using live-imaging and single-cell tracking in conjunction with the cardiomyocyte cTnnT-2a-eGPF live-reporter, we set out to reconstitute the lineage trees of mesodermal cells and identify the initial mesodermal progenitors that contribute solely to the heart tube in the gastrulating mouse embryo (Fig. 2 A).To achieve this required culturing and imaging early mouse embryos from the onset of gastrulation to the heart tube stage (i.e.~25-35 hours).
We found the Viventis LS1 open-top light-sheet microscope allowed the culture of early mouse embryos over long periods of embryonic development (>24 hours).Incubation media was stable and could be exchanged during acquisition.A large media volume (~1ml) improved embryonic viability for long-term imaging.Embryos cultured from E6.5 and for 35 hours developed normally; a cardiac crescent formed and generated a heart tube corresponding to E8 embryos.
To permanently label mesodermal cells and their progeny at a density suitable for live cell tracking, we used an inducible T 2a-cre/ERT2 mouse combined with R26 tdomato reporter and administered intermediate doses of tamoxifen (0.002mg/bw) at stages encompassing E5 to E6.5 [7].We first administered tamoxifen earlier -(at E5) -and cultured embryos in tamoxifen-free culture media, from before the start of the gastrulation period and onset of T/Bra expression in the primitive streak and mesoderm -(at E6) -.In these conditions, no tdTomato expressing cells could be identified in the intra-embryonic mesoderm over ~11 hours of live-imaging acquisition (Figure 2-supplementary figure 1 A).This confirms that creERT2 activity in T 2a-cre/ERT2 embryos requires T/Bra expression [7].In the absence of tamoxifen, rare tdTomato-positive cells were identified in only one embryo (not shown), confirming that tdTomato widespread mesodermal expression in T 2a-cre/ERT2 R26 tdTomato embryos requires tamoxifen.
We generated three light-sheet live-imaging datasets spanning 23 to 35 hours of mouse embryonic development from gastrulation -(E6.5-E7)-to heart tube stage (Fig. 2 B and Videos 1-3).Embryos were imaged for 1 minute at 2 minutes intervals.Raw data amounts to 5-7 terabytes per experiment, representing up to half a million images.To correct for drift during acquisition, BigSticher [14] was used to register the datasets in 4D as previously described [6].Movies were synchronised according to the timing of appearance of the cardiac crescent and heart tube inflows (Figure 2-supplementary figure 2 A-K).
Each movie contains up to ~1000 time points, and a small percentage (<1%) of linkage inaccuracy between cells could lead to lineage misinterpretation that propagates over the course of the movie.Automated cell tracking methods have seen advancements [15,16].
However, achieving the level of precision necessary to reconstitute cell lineages remains challenging.To obtain accurate cell lineages, we manually tracked single cells by visualising them at successive time points from the beginning to the end of the movie using Massive Muti-view Tracker (MaMut) [5] (Videos 1-3).We interrupted a track when it was impossible to identify the same cell across two successive time points unequivocally.A total of 61 mother cells were tracked for up to 4 generations.
We determined the identity of the final daughters based on their location in the heart tube.
The heart tube is formed by an inner endocardial layer ensuring the presence of a circulatory system, a myocardial layer formed by cTnnT positive cardiomyocytes and derived from the splanchnic mesoderm, and an outer layer derived from the somatic mesoderm called the pericardium.We could discriminate these cell types in our live-imaging datasets (Videos 4-6).Moreover, endocardial cells had distinct spindle-like shapes, displaying protrusions and transmigrating across the myocardium (Fig. 4D and Video 6, n=4).Cardiomyocytes were further distinguished by their higher levels of cTnnT-2a-eGFP reporter expression (Material and Methods and Figure 2-supplementary figure 3 A-D).Finally, the locations of the myocyte cTnnT-2a-eGFP+ cells within the heart tube indicated their fates.The heart tube has an inverted Y shape, and the two arms of the Y -or inflows-, positioned inferiorly, are fated to become the atria, with the stem of the Y becoming the left ventricle (LV) and atrioventricular canal (AVC) [17].In what follows, we describe the lineage trees and timing for mesodermal progenitors' specification into distinct cardiac lineages.

Distinct mesoderm contributes to the heart tube and inflows myocardium.
We first addressed the timing of LV/AVC and atria myocyte lineage segregation.From our 3 datasets, we identified 29 progenitors contributing to at least one cTnnT-2a-eGFP+ myocyte (Fig. 3A).We analysed their progenies' locations in the heart tube and found they established a clonal boundary located at the junction between the LV/AVC and atria myocyte compartments suggesting that atrial and LV/AVC progenitors have distinct mesodermal origins (Fig. 3 Bi-iii and Figure 3-supplementary Figure 1 Ai-vi).Among these 29 progenitors, 14 contributed to the LV/AVC and 13 to the atria.None contributed to clones spanning the LV/AVC and atria compartments.Two additional progenitors generated cTnnT-2a-eGFP+ cells in deeper z-locations within the heart tube (~ -250 μm).
The LV/AVC and atrial myocyte lineages have distinct temporal origins in the mesoderm.
Early mesoderm contributed to the LV/AVC, while late mesodermal cells generated atrial myocytes.(Fig. 3Bi).This observation is in line with the previous hypothesis that atrial and LV/AVC compartments have distinct spatial and temporal origins during gastrulation in the mouse [7][8][9].
The LV/AVC progenitors are born first and differentiate into cTnnT-2a-eGFP+ myocytes before other cardiomyocytes.This establishes the initial cardiac crescent within a ~10-hour period with the first LV/AVC progeny differentiating at ~E6.5 + 15 hours and the last at ~E6.5 + 25-27 hours.Atrial progenitors are born the latest, differentiate the latest -from ~E6.5 + 25-27 hours -, and are recruited to posterior regions during the folding of the cardiac crescent into the heart tube.This establishes the inflows (Fig. 3 D-E and H-I).A subset of mesoderm progenitors located in the inflow regions did not become cTnnT-2a-eGFP+ (n=6); however, these cells may become positive at later stages.
Myocytes developed concurrently within each lineage (Fig. 3 F).The time intervals between the first and the last daughter to transition into cTnnT-2a-eGFP+ myocytes were 2 hours and 30 minutes on average for LV/AVC lineages, and 2 hours and 35 minutes on average, for atrial lineages.Notably, the differentiation timing varied among lineages, with some displaying greater synchrony than others.For instance, in 3 out of 29 lineages, the mother cell generated LV/AVC cTnnT-2a-eGFP+ myocyte daughters in more than 5 hours.In contrast, in 8 out of 26 lineages, all daughters transitioned into cTnnT-2a-eGFP+ myocytes in less than one hour.
Together, the live-imaging analysis shows that the heart tube is established by at least two sets of independent LV/AVC and atrial myocyte progenitors generated from early and late mesoderm and differentiating into myocytes at different embryonic stages.
Bipotent progenitors contributing to cardiac and extra-embryonic mesodermal cells were preferentially located at the extra embryonic/embryonic border (Fig. 3 C).However, other bipotent/tripotent progenitors were intermingled with unipotent LV/AVC progenitors within the proximal mesoderm, with no clear spatial pattern identifiable.The initial cells' locations in the early proximal mesoderm seems, therefore, not to correlate with specific mesodermal fates.However, tracking a greater number of cells would be required to fully address this question.In line with previous live-imaging analyses [6,7], the distal mesoderm migrated to more medial locations (Fig. 3 C -dark red cells), known to provide progeny for the right ventricle, outflow tract and branchiomeric muscles [19][20][21][22].
In the late mesoderm, we identified 6 unipotent atrial myocyte progenitors out of the 13 progenitors contributing to the cTnnT-2a-eGFP+ atria myocytes.Longer tracks encompassing later stages are required to determine if the remaining progenitors contribute to atria cTnnT-2a-eGFP+ myocytes entirely or also to additional lineages.Moreover, we found additional progenitors contributing exclusively to cTnnT-2a-eGFP-daughters located in the inflow's regions of the heart tube and posterior lateral plate mesoderm (n=16 lineages, identified as meso GFP-in Fig. 3 A).Additional markers or longer movies will be required to determine the identity of these cells.A subset of these cells had spindle-like shapes and were identified as endothelial-like cells (n=6, Fig. 3 A).
Together, the live imaging analysis of lineages shows that the early mesodermal cells harbour a previously underappreciated plasticity and diversity of fates during gastrulation [1].Yet, their ability to alternate fates seems to be rapidly reduced.20 out of 35 initial mother cells contributing to cardiac fates were unipotent, and in all cardiac lineage trees with two or three fates (n=8), progeny become lineage-restricted early, during migration, before the onset of cTnnT-2a-eGFP+ expression in the embryo (Fig. 3 G).All 6 bipotent progenitors analysed generated unipotent progenitors after the first cell division; and the two tripotent progenitors generated unipotent daughters after the first and second cell division (Fig. 3 A).These results are consistent with previous clonal analyses suggesting that early mesodermal progenitors are rapidly specified into discrete fates after the initiation of gastrulation [2][3][4].

Migration analysis in lineages reveal hidden patterns of cell migration.
Previous live analysis of cell trajectories during gastrulation revealed apparently chaotic individual cell movements during migration [6].Consistent with this analysis, we found that mesodermal cells dispersed extensively during migration with strong separating movements between daughter cells (Fig. 5 A).We analysed distances between the first two daughters (coordinates were taken 10 minutes after the first cell division -Timepoint 0 -, and last time point before the daughters' subsequent cell division -Timepoint 1 -) and granddaughters (final time point at which all granddaughter cells exist, we only considered branches lasting at least 4 hours into the cell cycle to allow sufficient cell migration -Timepoint 2 -) in each lineage.Distances between daughters and granddaughters gradually increased, reaching considerable distances within a single lineage (up to 365 μm) (Fig. 5 A).We noted, however, that distances were highly heterogenous; a proportion of the progeny generated less dispersive daughters with separating distances of less than 50 μm between them (11 out of 27 cardiac lineages).
One possibility for the observed heterogeneity in these distances is that the daughters generated by unipotent progenitors exhibit less dispersive migratory paths than those generated by bipotent progenitors (Fig. 5 B).To test if this correlation was true, we analysed cell movements in lineages, taking advantage of our lineage analysis from the live-imaging data.
We first analysed if distances between daughters and granddaughters generated by bipotent cardiac progenitors were greater than those generated by unipotent cardiac progenitors.
Immediately after the first cell division, we found no distance differences between sisters generated from unipotent or bipotent progenitors.Distances between sisters with shared cardiac fates became smaller on average compared to distances between sisters of distinct cardiac fates (Fig. 5 C).A similar difference was found when cell coordinates were sampled before the onset of cTnnT-2a-eGFP expression (Fig. 5 D-H).Thus, unipotent cardiac progenitors generated less dispersed daughter and granddaughter cells than bipotent progenitors.Distances between sisters could be high in non-cardiac lineages generating only extra embryonic mesoderm, occasionally reaching distances of over 300 μm (Fig. 5 G).
The observation that sister cells sharing the same fate end up in the same position in the embryo may be due to these cells following the same migratory paths.Alternatively, pairs of sister cells could independently follow distinct migratory trajectories and, by chance, converge into similar embryonic territories.To address this question, we analysed if sisters harbouring a shared cardiac fate migrated in closer proximity than sisters with divergent fates.We used a dynamic time warping (DTW) algorithm (Fig. 5 I and Material and Methods) that accounts for instances where cells may exhibit similar behaviours but with temporal shifts [23,24].We selected time periods such that sisters' trajectories started and ended at the same time and calculated the cumulative distances between the two trajectories that yielded the optimal alignment in each pair of sister cells (i.e. a DTW distance).
On average, unipotent progenitors generated sisters with lower DTW distances than bipotent progenitors (Fig. 5 J-L).We observed four case (out of 20) where a unipotent progeny generated sisters with a distinct migratory trajectory (log DTW value > 8.5), and two cases (out of 10) where bipotent progenitors produced sisters with similar migratory paths but distinct fates (log DTW value < 8.5) (Fig. 5 J).This suggests that sisters can occasionally diverge in their migratory paths yet adopt a similar cardiac fate; or adopt similar paths and contribute to different fates.(Fig. 5 J-L and Figure 5 -supplementary 1 A-B).However, the majority of sister pairs with the same cardiac fate exhibited notably similar migratory trajectories (Fig .5 N-S).To rigorously assess this observation, we employed a permutation test.We created 100.000 permutations by pooling all log DTW distances and randomly assigning them to either unipotent or bipotent conditions.In each permutation, we computed the difference in average log DTW values between sister cells with shared fates and those with distinct fates.This iterative process generated a null distribution, allowing us to test whether unipotent progenitors are producing sisters with more similar trajectories compared to bipotent progeny (Fig. 5 Supplementary Figure 1C).Our analysis revealed that this was the case (p-value= 0.00027).Plotting the DTW values over time shows that the migratory trajectories of sister cells sharing the same fates are similar throughout their entire migratory periods and any observed similarity is not attributed to systematic smaller DTWcontributions towards the end of the trajectory (Fig. 5 M).Moreover, the DTW values diverged early in sister cells with distinct fates, indicating that their migratory trajectories are rapidly distinct at the beginning of their migration when the different fates head off towards different destinations (Fig. 5 M).The results suggest that pairs of sisters with shared fate maintain closer proximity throughout their entire migration periods, exhibiting more analogous migratory paths compared to sister cells that adopt distinct fates (Fig. 5 M and N-S).

Discussion
Our findings illustrate the progression of cardiac mesodermal lineages during gastrulation (Fig. 6).Using live-imaging and single cell tracking, we reconstructed cardiac mesodermal lineages and migratory paths of the cells over extended periods encompassing gastrulation and heart tube morphogenesis (~35 hours).Culturing embryos in large volumes of media culture (~1ml) using an open top light sheet microscope was critical to achieve these experiments.
These findings suggest that an early segregation of the ventricular and atrial cells has been conserved during evolution; an early segregation of these progenitor populations was previously shown at single cell resolution in the zebrafish [29][30][31][32].They also align with in vitro differentiation experiments demonstrating that modulating pathways known to induce mesoderm can generate molecularly distinct mesoderm favouring the generation of ventricular or atrial-like cardiomyocytes respectively [33][34][35][36][37].
We found that the early proximal mesoderm harbours multipotent progenitors generating extraembryonic mesoderm, pericardial, endocardial and endothelial-like cells in addition to LV/AVC myocytes.These progenitors became rapidly restricted into unique cardiac fates during migration -prior to the establishment of the cardiac crescent and onset of myocyte differentiation.The observation of short-lived multipotent cardiac progenitors is consistent with clonal analysis results of Hand1+ and Mesp1+ mesodermal progenitors [1][2][3].
Previous migration analysis noted opposing cell density and motility gradients in the mesoderm [6].According to this model, cells continually exchange neighbours and disperse widely until their movements gradually diminish, eventually settling in positions and fates as gastrulation concludes.Our live-imaging analysis builds upon these findings, offering a more detailed and prolonged evaluation of the migratory paths of cells in relation to their future fates.The analysis revealed that progenitors contributing to LV/AVC and atrial myocytes remain as separated cell populations throughout migration, establishing two distinct progenitor domains in the heart tube without mixing.During ontogeny, the pericardial and myocardial layers, along with the subjacent plexus of elongated endocardial cells, emerge in close proximity within the cardiac crescent [38].The migratory trajectories of progeny contributing to these three distinct cardiac fates were more deterministic than previously recognised.Sister cells contributing to the same fate tended to exhibit similar migratory paths.In the future, it will be crucial to discern whether early mesodermal progenitors exhibit similar migratory paths because of similarities in their initial internal state and their ability to interpret environmental cues in a cell-specific manner.
Moreover, BMP, Nodal, and FGF morphogen gradients regulate cell migration independently of cell fate [42][43][44][45][46]. Thus, it seems that mesodermal cells respond to morphogen cues with precision, providing determinism to the morphogenetic cell behaviours.Simultaneously, they demonstrate plasticity regarding their final fate [47].While progenitors are seen giving rise to only one cardiac cell type, they could potentially generate additional cardiac fates when no longer constrain by positional cues.We propose that achieving a delicate balance between determinism and plasticity is essential to ensure robust morphogenesis.This balance enables cells to follow specific developmental pathways while also maintaining the flexibility needed to adapt to changing external cues.
Together, our live-imaging analysis of migration and cardiac lineages provide evidence that some regulation of directionality of cell movements and fate allocation may exist early within the mesoderm.The findings have broader implications for our understanding of organogenesis since they address how initial differences between progenitors and signalling cues may ultimately affect the fate and movements of cells.

Experimental model and subject details
All animal procedures were performed in accordance with the Animal (Scientific Procedures) Act 1986 under the UK Home Office project licenses PP8527846 (Crick) and PP3483414 (UCL) and PIL IA66C8062.

Generation of the cTnnT-2a-eGFP line
The C -terminal tagging of cardiac troponin cTnnt2 with eGFP was generated in the Genetic Modification Service using CRISPR-Cas9 strategy.This editing was performed by cotransfection of a Cas9-gRNA vector and a donor vector comprising the T2A self-cleaving peptide and eGFP into B6N 6.0 embryonic stem cells using Lipofectamine 2000.The donor vector contained a 786bp insert of T2A-eGFP with 1kb homology arms either side.The guide sequence used was 5'-TTTCATCTATTTCCAACGCC -3'.Two correctly targeted ESC clones were microinjected into blastocysts which were then transfered into the uterus of pseudo pregnant BRAL (C57BL/6 albino) females.Generated chimeras were crossed to BRAL and F0 chimera offspring were initially screened for the proper integration of T2A -eGFP by Sanger sequencing.The F0 were again crossed to BRAL to confirm germline transmission of the mutation and F1 mice were validated by Sanger sequencing.
Embryos were systematically imaged throughout from top to bottom.Images were processed using Fiji software [48].

Embryo culture
Embryos were dissected at E6.5 in a solution of DMEM (Dulbecco's Modified Eagle Medium-D5921 Sigma-Aldrich) with 10% FBS (Fetal Bovine Serum-A5256701 Thermo Fisher), 25 mM HEPES-NaOH (pH 7.2), penicillin, and streptomycin.The dissection was done on a stereoscope microscope with a Tokai Hit thermoplate at 37 °C.Dissection was completing within 5 minutes and immediately transfer to media culture to preserve the embryo's developmental potential.Media culture was a mixture of 75% freshly prepared rat serum (filtered through a 0.2-mm filter) and 25% DMEM (containing 1 mg/ml D-glucose and pyruvate, without phenol red and L-glutamine-D5921 Sigma-Aldrich), supplemented with 1× glutamax, 100 units/ml penicillin, 100 μg/ml streptomycin, and 11 mM HEPES.The rat serum was prepared according to established protocols [49,50], stored at −80 °C, heatinactivated at 56 °C for 30 minutes, and filtered through a 0.22-μm filter before use.All media were equilibrated with a mixture of 5% O2, 5% CO2, and 90% N2, and warmed to 37 °C before adding embryos.

Multiphoton microscopy
To hold embryos in position during time-lapse acquisition, we made bespoke plastic holders with holes of different diameters (0.3 to 05 mm) to ensure a good fit for the ectoplancental cone similarly to the traps developed by Nonaka and colleagues [51].Embryos were mounted with their anterior side facing up.To avoid evaporation, the medium was covered with mineral oil (Sigma-Aldrich; M8410).Before starting the time-lapse acquisition, embryos were precultured for at least 2 hours in the microscopy culture set up.For the acquisition, we used the multiphoton Olympus FVMPE-RS equipped with a 5% CO2 incubator and a heating chamber maintaining 37°C.The objective lens used was a HCX APO L 20x/1.00 W dipping objective, which allowed a 2-mm working distance for imaging mouse embryos.A SpectraPhysics MaiTai DeepSee pulsed laser was set at 880 nm and used for one-channel two-photon imaging.Image settings was: output power: 250 mW, pixel dwell time: 7 μs, line averaging: two and image dimension: 610 × 610 μm (1,024 × 1,024 pixels).The z step was 6 μm.
Oral gavage was performed (0.02 mg/body weight) at indicated embryonic stages.Embryos were dissected at least 12 hours after.Before the time-lapse acquisition, embryos were precultured for at least 2 hours in the microscopy culture set-up.To hold embryos in position during the acquisition, we embedded part of the ectoplacental cone in Matrigel growth factor reduced phenol red-free (Corning Cat.No 356231) diluted two times with culture medium in a dedicated open-top FEP sample chamber containing an array of four chambers.Typically, 1 embryo was mounted per well, totalling 4 per experiment.Pixel size was 2 μm x 0.347 x 0.347 (z, x, y).The Viventis LS1 used a single view and dual illumination light sheet.
Detection was done using a Nikon 25X NA 1.1 water immersion objective with final 18.7X magnification generating a field of view of 800 x 800 μm.Ilumination with the 488 and 561 lasers was sequential.Image acquisition was performed every 2 mins and in stacks totalling 500 μm.Exposure times for GFP and tdTomato detection was 50ms and 100ms.Light sheet thickness was 3.3 µm.Embryos were incubated at 37°C in 8% CO2 with humidification throughout.
The midpoints between D1's and D2's progeny were first determined by averaging their coordinates.Progenitors' coordinates were sampled before cTnnT-2a-eGFP signal was observed.Subsequently, a midpoint distance between D1 and D2's progeny's midpoints was calculated.

Dynamic time warping (DTW)
We analyzed migratory trajectories within cardiac lineages by comparing the paths of sister cells from the time immediately following the division of their mother cell to the final time point before the first sister cell division.Trajectory similarity was assessed using the 'dtwpython' (version 1.3.0)Python package, measuring the dynamic time warping distance (DTW) between the cells.We opted for DTW distance over Euclidean distance, as DTW accommodates cells with spatially similar yet temporally nonsynchronous trajectories [52,53].We applied the 'symmetricP1' step pattern, a slope-constrained step pattern [22,23].To determine whether DTW distances between sister cells, whether sharing the same fate or having distinct fates, exhibited significant differences, we performed a permutation test using the 'scipy' Python package (Version 1.11.1)[54].This involved 100,000 resamplings and the calculation of differences between log mean DTW distances for unipotent and bipotent cells in each permutation.The resulting p-value was computed as the proportion of permuted values greater than or equal to the observed values + 1, divided by the total number of permutations + 1.Our null hypothesis assumed no difference in log mean DTW distances between sister cells generated by unipotent and bipotent progenitors, while the alternative hypothesis suggested a significant difference.

Figure 4 .
Figure 4. Lineage analysis in time-lapse movies demonstrates unipotent and bipotent

Figure 5 .
Figure 5. Cell migration analysis in lineages reveals hidden patterns.(A) Dispersion

Figure 6 .
Figure 6.Working Model of early cardiac development.(A) Early proximal mesodermal