Self-assembling human heart organoids generated by Wnt signaling modulation. Our method was designed to meet four initial milestones: 1) high organoid quality and reproducibility; 2) high-throughput/high-content format; 3) relative simplicity (no need for special equipment outside of traditional cell culture instrumentation); 4) defined chemical conditions for maximum control and versatility for downstream applications. We started by assembling hPSCs into embryoid bodies by centrifugation in ultra-low attachment 96-well plates followed by a 48-hour incubation at 37 °C and 5% CO2 prior to induction. This incubation allowed for spheroid stabilization and was important to increase efficiency, as other incubation times (12 hours, 24 hours) provided inferior results once differentiation started. After induction, two-thirds of spent medium was removed and replaced with fresh medium for each medium change, resulting in gradual transitions in exposure to the different signals employed. Induction of mesoderm and cardiogenic mesoderm was achieved by sequential exposure to CHIR99021, a canonical Wnt pathway activator (via specific GSK3 inhibition), and Wnt-C59, a Wnt pathway inhibitor (via PORCN inhibition) (Fig. 1a). Brightfield and immunofluorescence imaging of hHOs showed a significant increase in size throughout the differentiation protocol (Fig. 1b). Confocal microscopy of hHOs stained with cardiomyocyte specific TNNT2 antibody showed that organoids started to develop sarcomeres as early as day 7 (Fig. 1b), with clear sarcomere formation and fiber assembly readily apparent by day 15 (Fig. 1c). Beating hHOs appeared as early as day 6 of the differentiation protocol, with robust and regular beating appearing by day 10 in all samples (Suppl. Videos 1 & 2). To determine optimal conditions for initial Wnt activation, we exposed embryoid bodies to different concentrations of CHIR99021 (4 µM, 6.6 µM and 8 µM) on day 0 for 24 hours. On day 15, hHOs were evaluated for cardiac lineage formation by confocal microscopy (Suppl. Figure 1a). Optimal cardiogenic mesoderm induction for all hESC and hiPSC lines tested occurred at lower CHIR99021 concentrations than previously reported for cardiomyocyte monolayer differentiation protocols, which typically range from 10 to 12 µM CHIR7–13. A 4 µM CHIR99021 exposure resulted in the highest cardiomyocyte content with 64 ± 5% TNNT2+ cells at day 15, versus 9.6 ± 5% and 2.4 ± 2% for 6.6 µM and 8 µM CHIR99021, respectively (Fig. 1d, Suppl. Figure 1a). This difference is probably due to endogenous morphogen production and paracrine signaling within the developing hHOs, bestowed by the 3D environment and inherent self-assembling properties of the organoids. hHOs treated with 4 µM CHIR99021 also displayed the best functional properties of the three concentrations (Suppl. Figure 1b, c). Our initial hHO differentiation protocol was reproducible across multiple hPSC lines (iPSC-L1, AICS-37-TNNI1-mEGFP, iPSCORE_16_3, H9). hHOs derived from different hPSC lines exhibited similar differentiation efficiencies, beat metrics, and sizes (Fig. 1e, f).
Controlled induction of epicardial lineage in human heart organoids. To increase organoid complexity and produce more developmentally relevant structures, we adapted methods that have been used successfully in monolayer hPSC differentiation for specific induction of epicardial cells11. The method consists of a second activation of canonical Wnt signaling on differentiation days 7–9 to induce secondary cardiac lineages. To determine if this second activation would prime our hHOs to increase complexity and better recapitulate heart development, we tested the effects of a second CHIR99021 exposure on day 7 (Fig. 2a). CHIR99021 was added to developing hHOs at varying concentrations (2, 4, 6 and 8 µM), and exposure lengths (1, 2, 12, 24, and 48 hours). Efficiency of epicardial cell and cardiomyocyte formation was evaluated using confocal imaging and quantification for well-established epicardial (WT1, ALDH1A2, TJP1) and cardiomyocyte (TNNT2) markers at day 15 (Fig. 2b, c; Suppl. Figure 2a, b). We found that the treatment robustly promoted the formation of proepicardium and epicardial cells (Fig. 2b-d; Suppl. Figure 2a). We found that high concentrations or long exposure times led to marked inhibition in the formation of other cardiac cell types other than epicardial cells, particularly affecting cardiomyocytes formation. We found that a single 2 µM CHIR99021 treatment for 1 hour on differentiation day 7 produced the most physiologically relevant epicardial to myocardial ratio (60% cardiomyocytes, 10–20% epicardial cells) (Fig. 2b, c; Suppl. Figure 2b, c). Structurally, a significant part of the epicardial tissue was found on external layers of the organoid and adjacent to well-defined myocardial tissue (TNNT2+) (Fig. 2d), thus recapitulating the structural organization found in the heart. The robust expression of TJP1 on epicardial cell membranes also confirmed the epithelial phenotype of these cells (Fig. 2c, d).
Transcriptomic analysis reveals hHOs closely model human fetal cardiac development and produce all main cardiac cell lineages. We decided to perform transcriptomic analysis at different stages of organoid formation to better characterize developmental steps and the molecular identity of cells in the organoids. hHOs were collected at different timepoints (day 0 through day 19) of differentiation (Fig. 3). Unsupervised K-means clustering analysis revealed organoids progressed through three main developmental stages: day 0 – day 1, associated with pluripotency and early mesoderm commitment; day 3 – day 7, associated with early cardiac development; and day 9 – day 19, associated with fetal heart maturation (Fig. 3a, Suppl. Figure 3). Gene ontology biological process analysis identified important genetic circuitry driving cardiovascular development and heart formation (Fig. 3a & Supp. Table 1; raw data deposited in GEO under GSE153185). To compare cardiac development in hHOs to that of previously existing methods, we performed RNA-seq on monolayer iPSC-derived cardiac differentiating cells using well-established protocols3. We also compared our RNA-seq results to publicly available datasets from previously reported monolayer cardiac differentiation protocols and human fetal heart tissue (gestational age days 57–67)14 (GSE106690). In all instances, hHO cardiac development transcription factor expression regulating first and second heart field specification (FHF, SHF, respectively) was similar to that observed in monolayer PSC-derived cardiac differentiation and corresponded well to that observed in fetal heart tissue (Fig. 3b, Suppl. Figure 3a). Interestingly, gene expression profiles showed that hHOs had higher cardiac cell lineage complexity than cells that underwent monolayer differentiation, especially in the epicardial, endothelial, endocardial, and cardiac fibroblast populations (Fig. 3c, Suppl. Figure 3b-c). These data suggest a significant enrichment in the structural and cellular complexity of our hHOs, thus bringing them in line with fetal hearts. This was confirmed by extending our gene expression analysis to look at several widespread critical gene clusters involved in classic cardiac function, including conductance, contractile function, calcium handling, and cardiac metabolism, among others (Fig. 3d). Of special interest, hHOs produced significant amounts of heart-specific extracellular matrix, a feature present in the fetal hearts but completely absent in monolayer differentiation protocols (Fig. 3d, Suppl. Figure 3d). Principle component analysis showed a clear progression in development in the hHOs from day 0 to 19 (Suppl. Figure 3e). Overall, hHOs had individual expression profiles best matching those of fetal hearts, and the global hHO transcriptome was closer to that observed in fetal hearts than in any of the monolayer protocols, as determined by hierarchical clustering (Fig. 3e).
Human heart organoids recapitulate heart field specification and atrial and ventricular chamber formation. The first and second heart fields are two cell populations found in the developing heart. Cells from the FHF contribute to the linear heart tube formation, followed by migrating cells belonging to the SHF that contribute to further expansion and chamber formation15. We found evidence of cells representing both heart fields in our organoids. HAND1 (FHF) and HAND2 (SHF) are members of the Twist family of basic helix-loop-helix (bHLH) transcription factors that play key roles in the regulation of numerous cell types in the developing heart16. Immunofluorescence of Day 8 hHOs showed well-differentiated, segregated regions of HAND1 (Fig. 4a) and HAND2 (Fig. 4d) cells, suggesting that both FHF and SHF progenitors are present and separate into their respective heart fields. In human hearts, the left ventricle ultimately forms from FHF progenitors and the atria form from SHF progenitors17. We therefore sought to determine whether our hHOs contain cardiomyocytes committed to either the atrial or ventricular lineages. Immunofluorescence for MYL2 (which encodes myosin light chain-2, ventricular subtype) and MYL7 (encodes myosin light-chain 2, atrial subtype) in Day 15 hHOs showed cardiomyocytes positive for both subtypes. The two different populations localized to different regions of the organoid and were in close proximity, which mirrors the expression pattern seen in human hearts (Fig. 4c). The expression of HAND1, HAND2, and MYL7 transcripts in the hHOs increased throughout the differentiation protocol and were similar to that observed in human fetal hearts, while MYL2 increased to a lesser degree (Suppl. Figure 3a, c). Adding India ink to the media for contrast, we recorded the beating organoids under a light microscope and observed central chamber-like structures surrounded by beating tissue (Suppl. Video 3). Taken together, these data suggest that the differentiation of our hHOs involves heart field formation, chamber specification and cardiomyocyte specification into atrial and ventricular subpopulations, both of which further emphasize their recapitulation of human cardiac development.
Heart organoids produce multiple cell cardiac lineages and acquire cardiac-specific morphological functionality. Results from the transcriptomic analysis (Fig. 3) suggested that the second CHIR99021 exposure led to the formation of other mesenchymal lineages and higher complexity in hHOs. To evaluate this finding, we performed immunofluorescence analysis for secondary cardiac cell lineages. Confocal imaging confirmed the presence of cardiac fibroblasts positive for THY1 and VIM (Fig. 5a), which made up 12 ± 2% of the tissues in the hHOs (Fig. 5e), similar to the composition of the fetal heart described in the literature18. Further confocal imaging revealed a robust interconnected network of endothelial cells (PECAM1+), and vessel-like formation throughout the organoid (Fig. 5b). Higher magnification images uncovered a complex web of endothelial cells adjacent to or embedded into myocardial tissue (Fig. 5c, Suppl. Video 4). 3D reconstruction of confocal imaging stacks showed a well-connected endothelial network intertwined in the hHO tissue (Suppl. Videos 4 & 5). These results strongly indicate that during hHO development, self-organizing endothelial vascular networks emerge in response to the 3D cardiovascular environment, adding a coronary-like vascular network to the organoids (a phenomenon not observed before). We also observed chamber-like areas within the TNNT2+ and suspected they might possess chamber-like qualities and mimic early heart chamber formation. Immunofluorescence analysis for the endocardial marker NFATC1 revealed the formation of an endocardial layer of NFATC1+ cells lining these spaces, similar to the endocardial lining of the heart (Fig. 5d). Figure 5e shows a quantification of the contribution of these different cell populations to the organoids. Next, we employed optical coherence tomography (OCT) to characterize chamber properties using minimally invasive means, thus preserving chamber physical and morphological properties. OCT showed clear chamber spaces within the hHOs, typically with one or two large chambers near the center of the organoids (Fig. 6a, Suppl. Figure 4a-c). 3D reconstruction of the internal hHO topology revealed a high degree of interconnectivity between these chambers (Suppl. Videos 6–8). The presence of chambers was further confirmed using light-sheet imaging of whole organoids (Fig. 6b). Given the relatively large size of our heart organoids (up to 1 mm), we decided to verify whether the formation of these chambers could be associated to internal cell death. To do this, we created a transgenic hiPSC line expressing FlipGFP, a non-fluorescent engineered GFP variant which turns fluorescent upon effector caspase activation and is thus a reporter for apoptosis19. FlipGFP organoids in control conditions exhibited no fluorescence indicating that there is no significant programmed cell death (Suppl. Figure 4d). This observation is further supported by the lack of internal cellular debris observed during confocal imaging (data not shown). Doxorubicin-treated hHOs were used as a positive control for apoptosis (Suppl. Figure 4d), with evident signs of cell death.
Ultrastructural analysis of hHOs showed features similar to those typically found in age-matched human fetal hearts, with well-defined sarcomeres surrounded by mitochondria, and containing gap junctions and T-tubules (Fig. 6c). We also measured electrophysiological activity to determine functionality. We utilized an in-house multi-electrode array sensor technology develop (Suppl. Figure 5) to show that hHOs exhibit normal electrophysiological activity with well-defined QRS complexes and T and P waves, and regular action potentials (Fig. 6d).
BMP4 and activin A improve heart organoid chamber formation and vascularization. The growth factors bone morphogenetic protein 4 (BMP4) and activin A have frequently been used as alternatives to small molecule Wnt signaling manipulation since they are the endogenous morphogens that pattern the early embryonic cardiogenic mesoderm and determine heart field specification in vivo20,21. We suspected that BMP4 and activin A, in combination with our small molecule Wnt activation/inhibition protocol, could synergistically improve the ability of hHOs to recapitulate cardiac development in vitro. We tested the effect of BMP4 and activin A in the context of our optimized protocol by adding the two morphogens at 1.25 ng/ml and 1 ng/ml, respectively (recommended concentrations found in the literature20), at differentiation day 0 in conjunction with 4 µM CHIR99021. No significant differences were found in formation of myocardial (TNNT2+) or epicardial (WT1+/TJP1+) tissue between control and treated hHOs (Fig. 7a). However, significant differences in organoid size were observed as hHOs treated with growth factors were about 15% larger in diameter (Fig. 7b, c). This difference may correspond with the increase in microchamber number and connectivity, as BMP4/Activin A-treated hHOs had more microchambers that were ~ 50% more interconnected with other chambers compared to control hHOs (Fig. 7d, e, g). Immunofluorescence and confocal analysis of organoids treated with BMP4 and activin A showed a 400% increase in the area of PECAM1 + tissue, indicating a significant effect on organoid vascularization (Fig. 7f, h), which might also account for the increase in hHO size.
Modeling DDP-induced CHD using human heart organoids. As proof-of-concept on the utility of our system, we decided to use our hHO model to study the effects of diabetes during pregnancy on cardiac development. Diabetes affects a large number of the female population in reproductive age and there is significant epidemiological evidence linking diabetes during the first trimester of pregnancy to increased risk of CHD (up to 12% in some cases, a 12-fold increase)22, but little understanding of the underlying mechanisms. To do this, we first modified hHO culture conditions to reflect reported normal physiological levels of glucose and insulin (3.5 mM glucose, 170 pM insulin, NHOs)23 and reported diabetic conditions (11.1 mM glucose and 58 nM insulin, DDPHOs)23,24. Interestingly, normal conditions also differed from the original protocol’s glycemic and insulin conditions (due to most media being originally developed for cancer cell culture and containing abnormally high levels of glucose). NHOs developed at a slower pace than their higher glucose counterparts but presented better physical organization, with formation of heart tube-like structures and later segmentation into different well-defined areas reminiscent of heart looping and chamber formation (Suppl. Figure 6a) without evidence of cell death or abnormal physiology (Fig. 8 and data not shown). However, their structure was also significantly more delicate and could easily be damaged (Fig. 8a). NHOs and DDPHHOs showed significant morphological differences as early as day 4 of differentiation. NHOs were slower to grow and exhibited patterning and elongation between days 4 and 8, while DDPHOs remained spherical throughout the two-week period (Fig. 8a). DDPHOs were also significantly larger in size after 1 week of differentiation (Fig. 8b), suggesting hypertrophy, a common outcome of diabetes in newborns, which typically suffer from macrosomia in all organs. Electrophysiology analysis showed increased amplitude and frequency in action potentials in DDPHOs (Fig. 8c and Suppl. Figure 6b, c and d) suggesting higher activity in the diabetic organoids. Metabolic assays for glycolysis and oxygen consumption revealed decreased oxygen consumption rate in DDPHOs and increased glycolysis when compared to NHOs (Fig. 8d, e and Suppl. Figure 6e). TEM imaging revealed DDPHOs had a reduced number of mitochondria surrounding sarcomeres (Fig. 8f) and a significantly larger number of lipid droplets, suggesting dysfunctional lipid metabolism. None of these phenotypes were found in NHOs. Confocal microscopy of myocardial and epicardial markers revealed a drastic difference in morphological organization as DDPHOs contained epicardial tissue surrounded by myocardial tissue, whereas NHOs contained epicardial tissue on top of or beside myocardial tissue as expected (Fig. 8g). Furthermore, compared with normal glycemia conditions, diabetic hHOs showed decreased MYL2+ ventricular cardiomyocytes and enlarged chambers, again suggesting a dilated cardiomyopathy-like phenotype (Fig. 8h). These differences in impaired structural/developmental organization and lipid metabolism in DDPHOs are consistent with expected phenotypes found in diabetic patients and newborns exposed to high glucose/insulin. Taken together, our data suggest significant molecular and metabolic perturbations between NHOs and DDPHOs consistent with previous studies on DDP suggesting increased oxidative stress, cardiomyopathy and altered lipid profiles25–27, and constitute a significant step forward to model metabolic disorders in human organoids.