The heart is a resident tissue for hematopoietic stem and progenitor cells in zebrafish

Abstract

The contribution of endocardial cells (EdCs) to the hematopoietic lineages has been strongly debated. Here, we provide evidence that in zebrafish, the endocardium gives rise to and maintains a stable population of hematopoietic cells. Using single-cell sequencing, we identify an endocardial subpopulation expressing enriched levels of hematopoietic-promoting genes. High-resolution microscopy and photoconversion tracing experiments uncover hematopoietic cells, mainly hematopoietic stem and progenitor cells (HSPCs)/megakaryocyte-erythroid precursors (MEPs), derived from EdCs as well as the dorsal aorta stably attached to the endocardium. Emergence of HSPCs/MEPs in hearts cultured ex vivo without external hematopoietic sources, as well as longitudinal imaging of the beating heart using light sheet microscopy, support endocardial contribution to hematopoiesis. Maintenance of these hematopoietic cells depends on the adhesion factors Integrin α4 and Vcam1 but is at least partly independent of cardiac trabeculation or shear stress. Finally, blocking primitive erythropoiesis increases cardiac-residing hematopoietic cells, suggesting that the endocardium is a hematopoietic reservoir. Altogether, these studies uncover the endocardium as a resident tissue for HSPCs/MEPs and a de novo source of hematopoietic cells.

Fig. 1. Hematopoietic cells are stably maintained and grow in numbers on the endocardium.

A Single-cell RNA sequencing analysis of EdCs sorted by FACS for Tg(kdrl:nls-mCherry) expression at 48 and 72 hpf. The UMAPs reveal a subcluster of 390 cells that expresses genes affiliated with hematopoietic differentiation mostly at 72 hpf. B UMAP of the hematopoietic endocardial cell cluster indicates 5 subclusters. C Violin plots showing that genes enriched in hematopoietic cells at 72 hpf, such as myb (HSPCs/MEPs), mpx (neutrophils), gata1 (erythrocytes), and mpeg1.1 (macrophages), are enriched in this endocardial hematopoietic cluster, with HSPCs/MEPs and erythrocytes being the predominant cell types. D–F Representative confocal images of Tg(cd41:GFP) expression labeling HSPCs (low GFP; red asterisks) and platelets (high GFP; white asterisks) attached to the endocardium marked by Tg(kdrl:nls-mCherry) expression. D At 50 hpf, cd41:GFP⁺ cells are almost absent in the heart. E, F From 74 hpf onwards, cd41:GFP⁺ cells are attached to the endocardium. G HSPC (cd41:GFP^low+) numbers progressively increase from 56 hpf and remain until at least 10 dpf. n = 10 (50 hpf), 13 (56 hpf), 19 (74 hpf), 27 (98 hpf), 17 (120 hpf), 33 (7 dpf), and 19 (10 dpf) hearts. H Schematic heat map representing the positional mapping of cd41:GFP⁺ cells, the majority of which (66.5%) attaches to the outer curvature of the ventricle. n = 20 hearts. (I, J) Quantification of co-surface localization of cd41:GFP⁺ cells and endocardium labeled by Tg(kdrl:BFP-CAAX) expression in the ventricle. HSPCs share 11%, 27.4%, and 29.3% of surface areas with the ventricular endocardium at 74, 96, and 120 hpf, respectively. n = 20 (74 hpf), 28 (96 hpf), and 23 (120 hpf) hearts, respectively. K, K’ Representative confocal images of EdU pulse labeling assays (50–74 hpf) in Tg(cd41:GFP), Tg(kdrl:nls-mCherry) larvae. Most cd41:GFP⁺ cells attached to the heart are EdU-positive at 74 hpf (yellow asterisks). L Quantification of the proportions of cd41:GFP⁺ cells positive for EdU after a pulse assay between 50–74 and 74–98 hpf. n = 14 (50–74 hpf) and 17 (74–96 hpf) hearts. One-way ANOVA with Tukey’s multiple comparison test (G, I, J) or Unpaired, two-tailed t test (L) was used. Scale bars, 30 µm (D–F, K). Source data are provided as a Source Data file.

Fig. 2. Endocardial cells contribute to cardiac-residing hematopoietic cells.

A–B’ Confocal Z-stack projection (A) and single confocal plane (A’) of photoconverted endocardium at 22 hpf, and follow-up imaging of the same animals (B, confocal Z-stack projection; B’, single confocal plane) 52 h post-conversion. Extruding Kaede-red⁺ cells are observed on the endocardium at 74 hpf (white asterisk). C, D Confocal image of 74 hpf fli:Kaede⁺ caudal hematopoietic tissue (CHT; C) and thymus (D) from larvae with their endocardium photoconverted at 22 hpf. Kaede-red⁺ cells (white asterisk) surrounded by endothelial cells are visible in the CHT (C); few Kaede-red⁺ cells visible in the thymus (white asterisks) (D). E, F Quantification of Kaede-red⁺ cells from the photoconverted endocardium (22 hpf) reveals 16.7 and 0.92 endocardial-derived cells in the CHT (74 hpf, E) and thymus (74 hpf, F), respectively, 52 h post-conversion. n = 14 CHT and 14 thymi. G–H’ Single confocal plane of a heart with photoconverted endocardium in the ventricular outer curvature (G, G’) or the atrium (H, H’) at 48 hpf and at 48 h post-conversion (G’–H”). G”, H” Closeup single plane of the 96 hpf fli:Kaede⁺ heart, where EdCs in the ventricular outer curvature (G”) or the atrium (H”) of the same animals had been photoconverted at 48 hpf. G” Kaede-red⁺ cells observed on the ventricular endocardium (white asterisks) near Kaede-green⁺ cells (red asterisks). H” Kaede-red⁺ cells derived from the atrium observed close to the atrial and ventricular endocardium (white asterisks). Kaede-green⁺ cells (red asterisks) are also present in the ventricle. I, J Quantification of Kaede-red⁺ cells derived from the ventricular outer curvature (I) or the atrium (J) found in other regions of the heart and the CHT, 48 h post-conversion. Several Kaede-red⁺ cells from the ventricular outer curvature found in the CHT, but not in other heart regions. Atrial-derived Kaede-red⁺ cells are present in the CHT and ventricular outer curvature. n = 14 hearts with photoconverted ventricular outer curvature and 14 hearts with photoconverted atrium, respectively. One-way ANOVA with Tukey’s multiple comparison test (I, J) was used. Scale bars, 30 µm (A–D, G–H”). Source data are provided as a Source Data file.

Fig. 3. Ex vivo transplants and longitudinal light sheet imaging of the beating heart show endocardial cells activating HSPC markers.

A Schematics of ex vivo growing of hearts extracted at 48 hpf and cultured for 24 h. B Representative confocal images of a 48 hpf extracted Tg(Runx1:mCherry), Tg(cd41:GFP), Tg(kdrl:BFP-CAAX) heart, showing no Runx1:mCherry⁺ or cd41:GFP⁺ cells. After culturing for 24 h, Runx1:mCherry⁺ and cd41:GFP⁺ cells are present in the heart. C Significantly higher numbers of Runx1:mCherry⁺ and cd41:GFP⁺ cells are observed in hearts at the end of the 24 h culture compared with 48 hpf extracted hearts. n = 13, 17, 12, and 19 hearts, from left to right. D Schematics of a 72 hpf zebrafish larva at the illumination axes of the light sheet microscope. Movie stack acquisitions (one movie per plane) were used to reconstruct a synchronization of the beating heart every 30 min for 7 h. E–J Single plane of an endocardial cell in the ventricular outer curvature (E–G, magenta, white arrow) or in the ventricular inner curvature (H–J, magenta, white arrow). At 0 min (E, H), no Tg(cd41:GFP) expression is observed. At 30 min (F, I), Tg(cd41:GFP) expression appears noticeable (F); at 60 min (G, J), strong Tg(cd41:GFP) expression is visible in endocardial cells that remain integrated in the heart. The hearts are outlined in dashed lines. C Unpaired, two-tailed t test was used. Scale bars, 30 µm (B, E–J). Source data are provided as a Source Data file. Figure A, D created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.

Fig. 4. Dorsal aorta endothelial cells are major contributors to cardiac-residing hematopoietic cells.

A, D, G, J Representative confocal images of Kaede green-to-red photoconversion of Tg(fli1:Kaede)⁺ endothelial cells in the anterior-most region (A), middle region (D), posterior-most region (G), or combined middle and posterior regions (J) of the dorsal aorta (DA) at 22 hpf. B–C’ Z-stack projections and single confocal planes of the heart at 74 (B, B’) and 98 (C, C’) hpf show minimal to no contribution of anterior DA-derived Kaede-red⁺ cells to cardiac-residing cells. E–F’ Z-Stack projections and single confocal planes of the heart at 74 (E, E’) and 98 (F, F’) hpf show moderate contribution of the middle DA region to the heart (white asterisks). H–I’ Z-Stack projections and single confocal planes of the heart at 74 and 98 hpf demonstrate a significant contribution of posterior DA cells to the heart. K–L’ Photoconversion of endothelial cells in the middle and the posterior DA (combined) regions at 22 hpf lead to the highest numbers of DA-derived cells attached to the endocardium (white asterisks). M Quantification of DA-derived Kaede-red⁺ cells attached to the endocardium at 74 and 98 hpf. The posterior DA subregion contributes the highest numbers of cardiac-residing cells. Photoconverting endothelial cells in both the middle and posterior DA (combined) regions yields the most cardiac-residing cells. n = 14 (anterior), 24 (middle), 23 (posterior), and 12 (combined) hearts at 74 hpf, and 14 (anterior), 13 (middle), 16 (posterior), and 15 (combined) hearts at 96 hpf. A one-way ANOVA with Tukey’s multiple comparison test was used. Scale bars, 30 µm (A–L). Source data are provided as a Source Data file.

Fig. 5. HSPC attachment to the endocardium is dependent on Itga4-Vcam1b ligand-receptor interaction.

A Violin plots from the endocardial single-cell RNAseq dataset show that itga4 expression is enriched in the hematopoietic endocardial population (Cluster 5), whereas vcam1b expression is elevated in other endocardial cell populations. B RNAscope in situ hybridization for vcam1b mRNA in the 74 hpf zebrafish heart. vcam1b appears to be strongly expressed in the EdCs of the ventricular outer curvature (B’, labeled with Tg(kdrl:GFP) expression, white arrows), but appears undetectable in the valves (B”). C Representative confocal images of control, itga4^−/−, and vcam1b^−/− Tg(Runx1:mCherry), Tg(kdrl:GFP) hearts at 74 hpf. White asterisks indicate Runx1:mCherry⁺ cells attached to the endocardium. D Significantly fewer Runx1:mCherry⁺ cells were found in itga4^−/− and vcam1b^−/− hearts compared with control. n = 12 control, 20 itga4^−/−, and 9 vcam1b^−/− hearts. E Representative confocal images of Tg(cd41:GFP), Tg(kdrl:nls-mCherry) hearts in control, itga4, and vcam1b morphants at 74 hpf. F Quantification of cd41:GFP⁺ cells on the endocardium reveals a significant reduction in itga4 or vcam1b morphants compared with control. n = 38 control, 14 itga4, and 18 vcam1b morphant hearts. G, I, K Representative confocal images of photoconverted endothelial cells in the middle and the posterior regions of the DA at 38 hpf in control (G), itga4 morphants (I), and vcam1b morphants (K). H Photoconverted Kaede-red⁺ cells (white asterisks) are attached to the endocardium in 74 hpf control hearts. J, L Noticeably lower numbers of Kaede-red⁺ cells from the DA are present in 74 hpf hearts of itga4 (J) or vcam1b (L) morphants. M DA-derived Kaede-red⁺ cell numbers in the heart are significantly reduced in itga4 and vcam1b morphant larvae compared with control. n = 13 control, 14 itga4, and 12 vcam1b morphant hearts. D, F, M One-way ANOVA with Tukey’s multiple comparison test was used. Scale bars, 30 µm (B, C, E, G–L). Source data are provided as a Source Data file.

Fig. 6. HSPC attachment to the endocardium is at least partly independent of cardiac trabeculation and contraction-induced biomechanical forces.

A–C Representative confocal image of 74 hpf Tg(cd41:GFP), Tg(kdrl:nls-mCherry) hearts. In nrg2a^−/− mutants or larvae treated with the ErBB2 inhibitor PD168393 (PD), both leading to hypotrabeculation, cd41:GFP⁺ cells remain attached to the endocardium. D, E Quantitative analysis reveals no significant difference in endocardial-attached cd41:GFP⁺ cells in both models of hypotrabeculated hearts, PD drug treatment (D) or nrg2a mutants (E), compared with control. D n = 12 control, 18 PD-treated hearts; (E) 7 control, 10 nrg2a^−/− hearts. F–I Suboptimal concentrations of MS-222 (G), tnnt2a morpholino (H), or Blebbistatin (I) were used to reduce cardiac contractions compared with control (F). No noticeable changes in cd41:GFP⁺ cell numbers were observed. J Quantifications show no significant changes in cd41:GFP^low+/HSPC/MEP numbers upon reduction of cardiac contraction. n = 23 control, 13 tnnt2a morphant, 6 Blebbistatin-treated, and 14 MS222-treated hearts. Two-way ANOVA with Sidak’s multiple comparison test (D, E) or One-way ANOVA with Tukey’s multiple comparison test (J) was used. Scale bars, 30 µm (A–C, F–I). Source data are provided as a Source Data file.

Fig. 7. Perturbing primitive erythropoiesis enhances the presence of cardiac-residing hematopoietic cells independently of blood flow.

A–C Representative confocal images of 74 hpf control (A) and gata1 morphant (B) Tg(cd41:GFP), Tg(kdrl:nls-mCherry) hearts, as well as gata1 morphant Tg(cd41:GFP), Tg(kdrl:nls-mCherry) hearts treated with MS-222 (C). Noticeable increase of cd41:GFP⁺ cells present in gata1 morphant hearts treated with DMSO (B) or MS-222 (C). D Quantification of cd41:GFP⁺ cells in untreated or MS222-treated control and gata1 morphants. Untreated gata1 morphant hearts exhibited significantly increased cd41:GFP^low+ cell numbers compared with control, whereas MS-222 treated gata1 morphant hearts exhibited a trend towards increased numbers of cd41:GFP^low+ cell numbers. n = 17 control, 15 gata1 morphant, and 13 MS222-treated gata1 morphant hearts. E, F Representative confocal images of control and gata1 morphant Tg(cd41:GFP), Tg(kdrl:nls-mCherry) CHTs at 74 hpf. G In contrast with the heart, total numbers of cd41:GFP⁺ cells did not change in the CHT upon loss of primitive erythrocytes in gata1 morphants. n = 11 control and 17 gata1 morphant CHTs. H, I Quantification of cd41:GFP^low+ and gata1:DsRed⁺ cell numbers show a significant increase in the total numbers and proportions of double-positive cells upon gata1 knockdown. (H) n = 12 control, 9 gata1 morphant hearts; (I) n = 11 control, 9 gata1 morphant hearts. J, K Representative confocal images of control and gata1 morphant Tg(cd41:GFP), Tg(gata1:DsRed), Tg(kdrl:BFP-CAAX) hearts at 74 hpf. Asterisks indicate cells positive for both Tg(cd41:GFP), Tg(gata1:DsRed) expression. Noticeably higher numbers of double-positive cells were observed in gata1 morphant hearts. One-way ANOVA with Tukey’s multiple comparison test (D), two-way ANOVA with Sidak’s multiple comparison test (G, H) or unpaired, two-tailed t test (I) was used. Scale bars = 30 µm (A–C, J, K), 100 µm (E, F). Source data are provided as a Source Data file.

Fig. 8. Schematic model showing the heart as a resident tissue for and a contributor of hematopoietic stem and progenitor cells.

The endocardium in red and the myocardium in gray. Endocardial cells undergo EHT (flat magenta cells), leading to the formation of HSPCs (round magenta cells in the endocardium). Additional HSPCs originating from the DA, particularly in the posterior DA region, attach to and stably integrate into the endocardium (green round cells). HSPCs originating from the endocardium and the DA contribute to the HSPC population in the CHT. Figure 8 created with BioRender.com released under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International license.