Sox7‐positive endothelial progenitors establish coronary arteries and govern ventricular compaction

Abstract The cardiac endothelium influences ventricular chamber development by coordinating trabeculation and compaction. However, the endothelial‐specific molecular mechanisms mediating this coordination are not fully understood. Here, we identify the Sox7 transcription factor as a critical cue instructing cardiac endothelium identity during ventricular chamber development. Endothelial‐specific loss of Sox7 function in mice results in cardiac ventricular defects similar to non‐compaction cardiomyopathy, with a change in the proportions of trabecular and compact cardiomyocytes in the mutant hearts. This phenotype is paralleled by abnormal coronary artery formation. Loss of Sox7 function disrupts the transcriptional regulation of the Notch pathway and connexins 37 and 40, which govern coronary arterial specification. Upon Sox7 endothelial‐specific deletion, single‐nuclei transcriptomics analysis identifies the depletion of a subset of Sox9/Gpc3‐positive endocardial progenitor cells and an increase in erythro‐myeloid cell lineages. Fate mapping analysis reveals that a subset of Sox7‐null endothelial cells transdifferentiate into hematopoietic but not cardiomyocyte lineages. Our findings determine that Sox7 maintains cardiac endothelial cell identity, which is crucial to the cellular cross‐talk that drives ventricular compaction and coronary artery development.


Introduction
The mammalian heart is composed of a range of different cell types including cardiomyocytes, fibroblasts, endothelial cells, and perivascular cells. Although cardiomyocytes occupy 70-85% of the heart volume, recent studies showed that cardiac endothelial cells are the most abundant cells in both murine and human hearts (~60%) (Pinto et al, 2016). Cardiac endothelial cells contribute to the endocardium (the inner cell layer of the heart), and the endothelium which forms the inlet and outlet vessels of the coronary plexus that comprises lymphatics, veins, arteries, and capillaries (Colliva et al, 2020). Given the close proximity of the endothelial lining and cardiomyocytes, factors secreted from cardiac endothelial cells have the potential to modulate cardiomyocyte contractile behavior, shape, and function. In support of this cross-signaling activity, recent studies have identified the endocardium as the critical tissue controlling the morphogenetic processes taking place during cardiac trabeculation (Del Monte-Nieto et al, 2018;Qu et al, 2019). In particular, Del Monte-Nieto et al (2018) identified endocardial Notch and Nrg1 pathways as key regulators, respectively, of endocardial extracellular matrix (ECM) degradation and myocardial ECM synthesis processes that are critical for trabecular myocardium growth and organization. These data suggest that the highly dynamic process of trabecular myocardium morphogenesis relies on a critical and tightly controlled interplay between endothelial cells and cardiomyocytes (Del Monte-Nieto et al, 2018).
Given the essential role of cardiac endothelial cells in controlling the interplay of differential cell type activity during heart formation, a central question remains as to whether endothelial cell identity and differentiation state are key prerequisites for cardiac morphogenesis. While endocardial cells are differentiated by E8.5, it is now known that the cardiac endothelial phenotype is highly plastic. During heart formation, both sinus venosus endothelium and ventricular endocardium are the source of cells for the endothelium forming the coronary vessels (Red-Horse et al, 2010;Wu et al, 2012). In the endocardial cushions, differentiated endocardial cells transdifferentiate into mesenchymal cells, ultimately giving rise to cardiac valves and septa (de Lange et al, 2004;Lincoln et al, 2004). In addition, a subset of endocardial cells in the outflow tract and atria will contribute to myeloid and erythroid cells during embryogenesis (Nakano et al, 2013). This suggests that maintenance of endocardial cell identity is finely regulated upon cell specification, ensuring only a subpopulation will acquire the potential to form other cardiac cell types without compromising the integrity of heart tissue homeostasis.
Although a handful of transcription factors and signaling pathways have been shown to induce early endocardial cell fate specification (Ferdous et al, 2009;Kataoka et al, 2011), little is known about those regulators necessary to maintain endocardial commitment and direct subsequent, specialized cell differentiation. The SOXF group of transcription factors (Sox7, Sox17, and Sox18) are known key players of endothelial cell differentiation and specification (François et al, 2008;Corada et al, 2013;Chiang et al, 2017). In the heart, SOXF group members are widely expressed in cardiac tissues in progenitor populations of distinct cell lineages (e.g., cardiac and endothelial/endocardial) (Sakamoto et al, 2007;Gonz alez-Hernandez et al, 2020). Studies in Xenopus have implicated sox7 and sox18 in cardiogenesis (Zhang et al, 2005;Afouda et al, 2018). In mice, Sox17 has been shown to play a critical role in endocardial differentiation during early heart development and coronary artery formation (Saba et al, 2019;Gonz alez-Hern andez et al, 2020). This body of work suggests that the SOXF transcription factors share an important role in controlling the differentiation of blood vascular endothelial cells and cells of the cardiac endothelium.
Constitutive inactivation of Sox7 in the mouse results in pericardial edema and embryonic lethality, suggesting a potential role for SOX7 protein in early heart development (Wat et al, 2012). Some studies in embryoid bodies have implicated SOX7 as a regulatory switch between cardiac and endothelial lineages during cardiac mesoderm differentiation, through the positive regulation of Wnt and Bmp signaling (Nelson et al, 2009;Doyle et al, 2019). Further recent studies have established a role for Sox7 in endocardial cushion formation and the modulation of atrioventricular septum formation via the modulation of the Wnt4/Bmp2 signaling pathways (Hong et al, 2021). In humans, SOX7 variants have been also reported in patients displaying conotruncal defects, with some in vitro evidence suggesting that direct transcriptional regulation of VE Cadherin by this TF is essential for the endothelial-to-mesenchymal transition that takes place in the outflow tract formation during development (Jiang et al, 2021). Despite the clear association of Sox7 gene function and cardiac development, it remains elusive how Sox7-dependent pathways maintain endothelial cell identity in cardiac vascular beds, and whether Sox7-positive endothelial cells instruct morphogenic events that underpin heart tissue organization.
Here, we assessed the role of SOX7 molecular role in committed cardiac endothelial cells during heart morphogenesis. Our data suggest that Sox7 acts as a central regulator of cardiac endothelial cell differentiation and lineage maintenance and that this function is critical for governing endothelial-to-cardiomyocyte interactions, which in turn governs ventricular myocardium trabeculation and compaction. These findings identify Sox7 as a novel endothelialspecific master regulator of mammalian cardiogenesis.

Sox7 is expressed by the endocardium and is indispensable for cardiac development
To investigate the role of Sox7 in heart morphogenesis, we first examined the expression of the SOX7 protein during development. It has been reported that all members of the SOXF family are expressed in coronary vessels but are differentially expressed in the endocardium (Gonz alez-Hern andez et al, 2020). To detect endogenous SOX7 expression, we used the Sox7-V5-tagged transgenic reporter mice (Chiang et al, 2023). At E11.5, SOX7-positive signals (white) were found exclusively in endomucin-positive endocardial cells (green), but could not be detected in the tropomyosin-positive myocardium (red) (Fig 1A). SOX7 was also observed at the outer lining of the endocardial cushion (ECC) (Fig 1A). We also investigated expression of SOX17 by immunofluorescence analysis on tissue sections, using a SOX17 antibody ( Fig 1B). Interestingly, it has been previously reported that Sox17 is critical for endocardial differentiation in endocardium precursor cells (Saba et al, 2019). SOX17 protein was virtually undetectable in the endocardial lining by E11.5 ( Fig 1B). Our data are consistent with work from Sharma et al (2017), showing that Sox17 is a coronary vessel marker not expressed in the bulk of the endocardium under physiological conditions, but is induced in a subset of activated endocardium to form coronary vessels when the sinus venosus-derived coronary growth is stunted. In contrast, SOX17 protein is highly expressed in the dorsal aorta (Fig 1B) (Corada et al, 2013).
Constitutive inactivation of Sox7 in mouse embryos resulted in developmental delay, pericardial edema, and lethality as early as E10.5, consistent with a previous report ( Fig 1C) (Wat et al, 2012). The onset of pericardial edema occurs around E9.5, prior to the general gross phenotype ( Fig 1D). However, the early lethality found in constitutive Sox7 mutant embryos prevented any further investigation on the role of Sox7 at later stages of heart development.
generate Cdh5-CreERT2:Sox7 fl/fl mice (Sox7 iECKO ). To minimize confounding defects from impaired vascular development shown in Fig 1D (Kim et al, 2016;Lilly et al, 2017), tamoxifen-induced excision of Sox7 exon 2 was performed at E9.5 and E10.5, corresponding to early to mid-organogenesis. In particular, this time point corresponds to the extension phase of trabecular myocardium and therefore, Sox7 iECKO embryos will have normal establishment of trabecular units, allowing investigation of the role of endocardial Sox7 in the later stages of trabecular myocardium development (Del Monte-Nieto et al, 2018). In addition, this time point is prior to the formation of coronary vessels; therefore, allowing the assessment of the role of endothelial Sox7 during coronary vessel formation (Vir agh & Challice, 1981).
To characterize the cardiac phenotype of Sox7 iECKO mutants, we performed a time-course series of morphological analyses and morphometric quantifications in order to address not only general heart morphology but also differences in the relative area occupied by trabecular, compact, and total myocardium between Sox7 iECKO mutant embryos and control littermates (Figs 2 and EV1). No morphological defects or area differences in trabecular, compact, and total myocardium in either ventricle were found at E11.5 and E12.5 in the Sox7 iECKO mutants compared to control embryos (Fig EV1A-L). However, a statistically non-significant trend toward an increase in trabecular myocardium area was found in the left ventricle of Sox7 iECKO mutant hearts compared to the control littermates at E12.5 ( Fig EV1L). This increase in trabecular myocardium area was statistically significant by E13.5 in both left and right ventricles, suggesting that loss of Sox7 in the endocardium promotes a hyper-trabeculation phenotype (Fig EV1M-R). Similar analysis at E14.5 and E16.5 confirmed the hyper-trabeculation phenotype in Boxed areas SOX17 Figure 1. Sox7 is expressed in cardiac endothelial cells and is necessary for heart development.
A SOX7 expression in the E11.5 heart based on the Sox7-V5-tagged transgenic reporter mice. SOX7 (white) is expressed in the endothelial lining (delineated by endomucin (EMCN), green) of the atrium (A), ventricle (V), and endocardial cushions (ECC) but not in the myocardium (marked by tropomyosin (TPM), red). B Unlike SOX7, SOX17 (white) is not expressed in the endocardium at E11.5 (lower panels), although is expressed in the dorsal aorta (DA, upper panels). C, D Sox7-null embryos show cardiac and vascular defects at E9.5 and developmental defects at E10.5 (white arrowheads). The number of embryos showing the illustrated phenotype among the total examined is indicated. Whole-mount fluorescence and confocal analysis of Sox7 À/À mutant embryos at E9.5, with the panendothelial marker, endomucin (green), and the pericyte-marker, smooth muscle actin (red) reveals pericardial edema (white arrowheads) and blood vasculature defects (white arrows). Optical projection tomography (OPT) analysis using the pan-endothelial marker PECAM (blue) shows a lack of large arteries in the head vasculature (asterisks) in Sox7 À/À mutants.
Ó 2023 The Authors EMBO reports 24: e55043 | 2023 Sox7 iECKO mutants and identified a significant reduction in the compact myocardium area in both ventricles (Figs 2A-F and EV1S-X). These results suggest that loss of endothelial Sox7 promotes trabecular myocardium defects from E12.5, resulting in a non-compaction phenotype later in development.
Hyper-trabeculation defects in Sox7 iECKO mutants are associated with abnormal allocation of chamber cardiomyocytes to the trabecular layer The hyper-trabeculation phenotype identified in Sox7 iECKO mutants could be caused by multiple factors including increased size of trabecular cardiomyocytes (hypertrophy) or increased number of cells forming the trabecular myocardium. The latter could be due either to an increase in the number of cells entering the trabecular layer from the highly proliferative compact layer (de Boer et al, 2012), or an increase in cardiomyocyte proliferation in the trabecular layer.
To determine the cause for the increased area of the trabecular layer found in Sox7 iECKO mutants, we performed quantification of cell number and proliferation rate in the ventricular chamber myocardium, in both the trabecular and compact myocardium ( Tissue-specific staining for endocardium (CD31) and myocardium (MF-20 or SMA) were used to generate myocardial and endocardial masks to classify the identity of cell nuclei forming the chambers. Further semiautomatic segmentation allowed the classification of the nuclei into trabecular and compact myocardial nuclei. This allowed the quantification of the number of cells forming each myocardial subregion. Once all nuclei were classified and counted, we performed analysis of Ki67 staining inside each nucleus in order to determine the proliferation rate in each myocardial subregion. Quantification of cell number in both ventricular chambers revealed no significant difference in the total number of cells (endocardium and myocardium) forming either ventricular chamber in mutant versus control hearts. Likewise, separate quantification of the ventricular myocardium revealed no significant difference at any of the stages analyzed, although there is a trend toward an increase in the number of cardiomyocytes forming both chambers at E13.5 and E14.5 (Appendix Fig S1D, E, I and J). However, similar quantification following separation of cells forming either the compact or the trabecular myocardium identified a significant reduction in the number of cardiomyocytes forming the compact myocardium of both ventricles in Sox7 iECKO mutants at E14.5 (Fig 2G and H, left panels). Concomitantly, there was a significant increase in the number of cardiomyocytes forming the trabecular myocardium of Sox7 iECKO mutants in the left ventricle at E13.5 with an increase (non-significant) in the right ventricle that becomes significant in both ventricles at E14.5 (Fig 2G and H, right panels). These results suggest that the reduction in compact myocardium area and increase in the trabecular myocardium area observed morphologically are due to changes in the number of cells forming the compact and trabecular myocardium of Sox7 iECKO mutants, and not to cell hypertrophy.
To identify the mechanism underlying the change in cell number observed, we next analyzed cell proliferation in the Sox7 iECKO mutants. General proliferation rates in the ventricular chamber (endocardium and myocardium) and in the chamber myocardium specifically were not significantly different at E12.5 and E13.5 (Appendix Fig S1N, O, S and T). However, there was a significant reduction in cell proliferation rates in the cardiac chamber in the left ventricle and the chamber myocardium of both ventricles at E14.5 (Appendix Fig S1N, O, S and T). Similar analysis on separated compact and trabecular myocardial cells identified normal proliferation rates in both tissues at E12.5 and E13.5, but a significantly reduced rate of cardiomyocyte proliferation both in the compact and trabecular myocardium of both ventricles of Sox7 iECKO mutants at E14.5 (Fig 2I and J). These results suggest that the increase in cardiomyocyte number observed in the trabecular myocardium is not a result of a local hyper-proliferative response. Since we observed decreased proliferation in both compartments (trabecular and compact), this cannot explain the differential increase in cell number in the trabecular layer and decrease in the compact zone. This elevated number of cells in the trabecular layer is most likely due to an increase in the allocation of cardiomyocytes from the compact layer into the trabecular layer, hence resulting in the reduction in compact myocardium cells (Fig 2G-J).

Endothelial-specific loss of SOX7 function perturbs coronary artery formation
To further investigate the generalized reduction in cell proliferation we observed in both the trabecular and compact layers at E14.5, we turned toward a possible vascular phenotype we noted while characterizing the cardiac tissue architecture. We observed a dramatic reduction in the number of coronary vessels in both the subepicardial space (coronary veins) and within the compact myocardium (coronary arteries), suggesting that the loss of endothelial Sox7 function may also cause defects in coronary vessel formation (Figs 2C and EV1U; empty arrows). The significant reduction in cell proliferation in the trabecular and compact layers at E14.5 may be related to the lack of proper nourishment and oxygenation of the fast-growing myocardium due to defects in the formation of the coronary plexus found in Sox7 iECKO mutants. The coronary vasculature starts to form very early in development, but it only connects to the blood flow at around E13.5 when the coronary plexus connects to the base of the aorta (Vir agh & Challice, 1981). Therefore, defects in the proper formation of the coronary vascular plexus may explain the cardiac defects found in Sox7 iECKO mutants after E13.5. This observation prompted us to investigate the mechanism underlying the vascular defects in the Sox7 iECKO mutants, and we next set out to analyze the coronary plexus.
To further investigate the coronary vessel defects identified during the morphological characterization of Sox7 iECKO mutant hearts (Figs 2C and EV1U, empty arrows), we studied the patterning of the coronary artery plexus. We performed whole heart immunofluorescence staining using antibodies against the arterial marker, CX40, and a general endothelial marker, VEGFR2, to visualize the coronary arteries and coronary vessel plexus, respectively ( Fig 3A). In Sox7 iECKO mutant hearts, we consistently failed to detect distal coronary arteries (marked by arterial marker CX40 in red), while formation of the coronary stems appeared intact ( Fig 3A, Appendix Fig S2,  arrows). In contrast, the surrounding coronary vessels, labeled with VEGFR2 and Histone H3, appeared grossly normal (Fig 3A and B, Appendix Fig S2). The seemingly normal vessel network suggests that sprouting angiogenesis is not compromised in the cardiac tissue of Sox7 iECKO embryos. We confirmed this by quantifying the coronary vessel coverage, ERG + endothelial cell number, and their proliferation rate in hearts from mutant and littermate control E13.5 embryos (Fig 3C-E). These results suggest an important role for Sox7 in the coronary endothelium in the process of distal coronary artery assembly, but not coronary plexus formation. These results were further supported by morphometric analysis of the epicardial and intramyocardial coronary vessels (Appendix Fig S3). This analysis was done at E14.5 comparing Sox7 iECKO to control embryonic hearts and identified a significant reduction in intramyocardial vessels, subepicardial vessels, and subepicardial area. The results suggest that coronary and subepicardial defects are associated with the loss of Sox7 function in the endothelium (Appendix Fig S3A and B). However, all these changes were not maintained at E16.5 (Appendix Fig S3C and D), suggesting that the reduction in these parameters at E14.5 may be due to a delay in epicardium, subepicardium, and coronary vessel formation most likely due to the absence of proper growth of the compact layer. Indeed, at E16.5, the compact myocardium of Sox7 iECKO hearts is perfused by coronary vessels and both the subepicardium and its vessels look normal (Appendix Fig S3C). Interestingly, no major coronary vessels were present inside the compact myocardium of Sox7 iECKO hearts compared to control (Appendix Fig S3C, arrow), suggesting that Sox7 may play a role in the maturation and hierarchization of coronary vessels.
To determine whether the loss of distal coronary arteries observed in the Sox7 iECKO is a cardiac-specific defect or a more general vascular defect, we analyzed the process of blood vessel formation at earlier stages in constitutive Sox7 À/À embryos. Detailed analysis of vibratome sections of Sox7 À/À embryos as early as 17-18 somite stage (ss) identified arterio-venous shunts between the dorsal aorta (DA) and the cardinal vein (CV) (Figs 3F and EV2A). This type of vascular by-pass is a classic phenotype of compromised arterio-venous identity, suggesting that the loss of distal coronary arteries is likely due to the loss of arterial identity, akin to the phenotypic outcome in the Sox17 loss of function scenario (Corada et al, 2013; Gonz alez-Hern andez et al, 2020).

SOX7 directly regulates early markers of future coronary arteries
To investigate putative downstream targets of SOX7 TF involved in arterial specification, we performed bulk RNA-Seq analysis on whole Sox7 knockout embryos at E8.5 before any traces of gross severe vascular defects (Appendix Fig S4A and B). This approach revealed 15 down-and 50 up-regulated genes. Gene ontology analysis revealed that the differentially expressed genes have been primarily implicated in blood vessel development and angiogenesis. Of these genes, Dll4, Cx37, and Cx40 have a documented role in arterial specification (Benedito et al, 2008;Gkatzis et al, 2016;Fang et al, 2017), with Dll4 being a known target directly regulated by SOX7 (Sacilotto et al, 2013). Furthermore, Cx40 has been shown to mark the "pre-artery" cells that build coronary arteries (Su et al, 2018). To further confirm SOX7-dependent cellautonomous regulation of these genes in the endothelium, we performed RNA-Seq analysis of CD31 + cells sorted from Sox7 iECKO and sibling control embryos at E10.5 from dams pulsed with tamoxifen at E9.5. Of the handful of differentially expressed genes, we repeatedly observed the downregulation of Cx40 (Appendix Fig S4C). We confirmed the reduction of Cx40 transcript levels in sorted endothelial cells from Sox7 iECKO and sibling controls by qPCR ( Fig 4A). In addition, we also validated the expression levels of a closely related gene Cx37, identified as dysregulated in the whole-embryo RNA-Seq analysis. To further assess the regulation of Cx37 and Cx40 by SOX7 in an arterial specified cell line, we performed gene knockdown of SOX7 using a siRNA-based approach in the human arterial line HUAECs. This led to a reduction in CX37, CX40, and DLL4 expression levels, while expression of endothelial markers such as ERG and VEGFR2 remained unchanged ( Fig 4B). This series of in vitro and in vivo experiments position both Cx37 and Cx40 as downstream effectors of the Sox7 pathway in endothelial cells.
To further validate the positive regulation of Cx37 and Cx40 by SOX7 in vivo, we next assessed Cx37 transcript levels in Sox7 À/À embryos by whole mount in situ hybridization at E8.5 ( Fig 4C). In the constitutive knockout, loss of Sox7 function was validated at the phenotypic level by the presence of arterio-venous shunts (Figs 3F and EV2A). In these experiments, Cx37 expression appeared reduced in the mutant dorsal aorta at E8.5. SOX7-dependent downregulation of CX40 at the protein level was revealed by immunofluorescence analysis on tissue sections or whole-skin staining from E11.5 and E14.5 (Figs 4D and EV2B). Furthermore, Cx40 transcript levels were downregulated specifically in the ventricular coronary vessels, as shown by in situ hybridization on sections of Sox7 iECKO mutant hearts at E14.5 (Fig 4E).
◀ Figure 2. Endothelial-specific loss of Sox7 function leads to non-compaction cardiac defects.
A-C Morphological characterization of the cardiac defects in Sox7 iECKO hearts compared to littermate controls at E14.5 after Cre induction by tamoxifen injection at E9.5 and E10.5, by immunofluorescence analysis with tissue-specific markers CD31 (endocardium/endothelium; cyan), MF20 (myocardium; magenta), and nuclei (Hoechst; blue). (A) Low-magnification images showing the general heart morphology. (B, C) High-magnification images showing the morphology of the compact (arrows) and trabecular (arrowheads) myocardium, and coronary arteries (empty arrow) in the right (RV; B) and left (LV; C) ventricles. Note the clear increase in trabecular myocardium (arrowheads), reduction of compact myocardium (arrows, white contour), and the absence of intramyocardial coronary arteries (empty arrow) in Sox7 iECKO hearts compared to control hearts. D-F Quantification of the area occupied by total, compact, and trabecular myocardium in each ventricular chamber, showing a statistically significant increase in total (D) and trabecular (E, F, right graphs) myocardium area, and a statistically significant decrease in compact myocardium area (E, F, left graphs) in Sox7 iECKO hearts compared to control hearts. G, H Quantification of the number of cardiomyocytes forming the compact (left graphs) and trabecular (right graph) myocardium of the right (G) and left (H) ventricles in Sox7 iECKO hearts compared to control hearts at E12.5, E13.5, and E14.5. Cell counts confirmed that the increase in trabecular myocardium and decrease in compact myocardium areas in Sox7 iECKO hearts is due to an increase and decrease in cardiomyocytes in the trabecular and compact myocardium, respectively. I, J Quantification of the number of Ki67 + -proliferating cardiomyocytes compared to the total number of cardiomyocytes in the compact (left graphs) and trabecular (right graph) myocardium of the right (I) and left (J) ventricles in Sox7 iECKO hearts compared to control hearts at E12.5, E13.5, and E14.5. A significant reduction in cardiomyocyte proliferation in both the compact and trabecular myocardium was observed in Sox7 iECKO hearts compared to control hearts at E14.5.
6 of 21 To examine if Cx40 is a potential direct target gene of SOX7, we analyzed the SOX7-mCherrry ChIP-Seq dataset available from the human venous endothelial cell line (HUVECs) (https://www.ebi.ac. uk/arrayexpress/experiments/E-MTAB-4480/). This approach was designed to uncover putative SOX7-dependent regulatory elements associated with the CX40 gene. The most common binding motif identified in the SOX7 ChIP-Seq dataset corresponds to the reported SOX motif 5 0 -A/TTTGTT-3 0 ( Fig EV2C). From this dataset, we identified a 500 bp putative CX40 regulatory element situated 21 kb upstream from the transcription start site (CX40-21). The CX40-21 site appears to coincide with an open chromatin region (revealed by DNAseI footprint, black peaks) associated with the binding of active histone marker, H3K27Ac (light blue), and strong enhancers (ChroMM, orange bar) (EVD). To test the activity of this DNA element in vivo, we cloned a fragment of this region into the ZED enhancer reporter system (Bessa et al, 2009) and generated stable transgenic reporter zebrafish lines. The ZED vector has a minimal gata2 promoter driving GFP expression in the presence of the A Whole-mount immunostaining of control and Sox7 iECKO hearts at E14.5 after Cre induction by tamoxifen injection at E9.5, E10.5. Coronary vessels are stained with VEGFR2 (green), coronary arteries (arrows) and trabecular myocardium (TM) are both stained with CX40 (red). Asterisks show where the major distal coronary arteries are lost in the Sox7 iECKO heart. The number of embryos showing the illustrated phenotype among the total examined is indicated. B Whole-mount immunostaining of control and Sox7 iECKO hearts at E13.5 after Cre induction at E9.5 and E10.5. Coronary vessels are stained with VEGFR2 (white) and endothelial nuclei with ERG (red). The coronary plexus front is outlined by the dotted lines. C-E Graphs showing quantification of coronary vessel density (C), ERG-positive endothelial cells (D), and % of H3 + proliferative endothelial cells (E) in the coronary plexus. Scored sibling control, n = 4; Sox7 iECKO mutant embryos, n = 4 embryos; Mean AE SEM; Mann-Whitney U-test. ns = not significant. F Serial sections of a wild-type and a Sox7 À/À embryo at 17-18 ss (E9.0), stained with endomucin (blue) to detect the endothelial lining, ERG (white) for endothelial nuclei, and SOX17 (red) for arterial nuclei. The dorsal aorta (DA) is labeled by the red arrows and white arrowheads, while cardinal vein (CV), blue arrows. Fusion between DA and CV is indicated by orange arrowheads.
Ó 2023 The Authors EMBO reports 24: e55043 | 2023 CX40-21 region and enabled us to assess enhancer activity of this regulatory element during development. As shown in Figs 4F and EV2E, the 500 bp CX40-21 DNA element has the ability to drive GFP expression into cardiac endothelium and blood vessels in zebrafish larvae (Figs 4F and EV2E). Altogether, these data indicate that SOX7 plays an indispensable role in arterial specification of the distal A Quantitative PCR analysis on FACs-sorted PECAM + CD45 À endothelial cells from Sox7 iECKO mutants and sibling controls at E10.5, injected with tamoxifen at E9.5. Expression is normalized to the endothelial marker Pecam and Cdh5. Scored sibling controls, n = 7 embryos; Sox7 iECKO mutants, n = 7 embryos. The dashed line refers to the expression level of gene expression in control samples and is arbitrarily set to 1. B Quantitative PCR analysis on the human arterial endothelial cell line (HUAECS) transfected with SiSOX7 or SiCTRL for 48 h. Expression is relative to HPRT and GAPDH.
Biological replicates, n = 6. The dashed line refers to the expression level of gene expression in control samples and is arbitrarily set to 1. C Representative images of Cx37 in situ hybridization on whole E8.5 (8-10 ss) embryos. The staining in Sox7 À/À dorsal aortae is markedly reduced (asterisks) when compared with the wild-type controls (black arrows). D Section immunostaining of control and Sox7 iECKO early neuronal vascular plexus at E11.5. Blood vessels are stained with endomucin (green); SOX7 (white); and CX40 (red). Scale bars = 40 lm. E Representative images of Cx40 section in situ hybridization on E14.5 hearts. The staining in Sox7 iECKO coronary vessel in the ventricle free wall is markedly reduced (asterisks) when compared with the controls (black arrows). The number of embryos showing the illustrated phenotype among the total examined is indicated. Scale bar = 75 lm. F The CX40-21:GFP transgene directs GFP fluorescence expression to the heart endothelium and vascular endothelium, in transgenic zebrafish larvae at 4 dpf. OFT, outflow tract; A, atrium; V, ventricle.

Endothelial-specific deletion of Sox7 depletes a subpopulation of endocardial cells during heart development
While it is evident that SOX7 plays a pivotal role in arterial specification of the coronary artery, it is unlikely that the hypertrabeculation and non-compaction phenotype described above is a result of coronary artery malformation. The myocardium phenotype is initiated as early as E12.5 before coronary arteries are formed. To further dissect the molecular function of SOX7 in cardiac endothelial cells and its role in shaping the transcriptomic profiles of cardiac cell populations, we performed single-nuclei (sn) RNA-Seq on whole hearts from Sox7 iECKO and sibling control embryos at E12.5 3 days after Cre-mediated excision by tamoxifen treatment (Fig 5A). The gene-targeted disruption was confirmed by the presence of Sox7 transcripts missing exon 2 (the exon flanked by the loxP sequences) in the Sox7 locus of the Sox7 iECKO hearts (red track) ( Fig EV3A). snRNA-Seq analysis was performed using the Chromium system (10× Genomics), with individual nuclei collected from a pool of four hearts per genotype. A total of 6,414 control and 5,986 Sox7 iECKO nuclei were used for subsequent downstream analysis. Using the tdistributed stochastic neighbor embedding (t-SNE) approach, we identified a total of 13 distinct cell clusters, annotated based on the expression of top differentially expressed and known marker genes. These include three cardiomyocyte clusters (Ttn, Myh6, Nppa), three cardiac endothelial clusters (Flt1, Pecam1, Fli1), three hematopoietic cell clusters (Hba-a2, Itga2b, Ly86), two fibroblast clusters (Ptn, Pdgfra), one epicardial cluster (Wt1, Tbx18), and one cluster corresponding to cardiac valvular cells (Vcan, Tbx20) (Figs 5B-D and EV3B-E). Interestingly, all cell clusters were represented in the hearts of embryos of both genotypes, although their relative proportions varied between mutant and control samples.
It has been reported that the cardiac endothelial cell population is the most enriched population in the adult mouse heart (Pinto et al, 2016). Strikingly, general analysis of the relative contribution of the different cell populations in Sox7 iECKO mutants compared to control embryos identified a redistribution in the relative proportions of different cell types in Sox7 iECKO mutant hearts, with cardiomyocytes being the most abundant cell type outnumbering the diminishing endothelial cell population in this genotype (Figs 5E and F,and EV3B). We next interrogated the dataset to further investigate the molecular signatures contributing to this shift in the relative proportion of each cell population contributing to the Sox7 iECKO mutant heart.
The vast majority of endothelial cells have an endocardial cell identity as shown by C2, C5, and C7 clusters that express the Npr3 marker ( Fig 5D) (Zhang et al, 2016). This is consistent with the fact that the coronary vessels originating from the sinus venosus and endocardium have just started to infiltrate the cardiac tissue at E11.5 (Red-Horse et al, 2010). Of the three cardiac endothelial clusters, C5 is not proliferative as shown by the low levels of Daiph3, Ki67, and Ttk expression (Fig EV3C). A small subpopulation of endothelial cells expressed coronary vessel markers such as Fabp4 and Apelin (Npr3 low) within the proliferative population C7 (Fig 5D, red  Interestingly, we found a dramatic reduction in the C2 cardiac endothelial cluster in the Sox7 iECKO sample, while the other two clusters (C5, C7) remained unchanged (Fig 5B,E and F). Further analysis of the C2 endocardial population reveals that these cells display higher expression levels of genes such as Sox9 and Pdgfra that define endovascular progenitor cells (Fig EV3C) (Patel et al, 2017). In addition, cells in the C2 cluster also express Gpc3, a prominent liver stem cell marker (Grozdanov et al, 2006;Su et al, 2017), suggesting that the C2 population might represent a multipotent cell state. Unlike C5 and C7, heatmap analysis showing the expression profile of the C2 cell population reveals a blend of signals common to other populations (Fig 5C), suggesting that C2 represents a more immature endocardial population compared to C5 and C7.
To validate the presence of the C2 population in vivo, we analyzed the co-expression of Sox7, Sox9, and Gpc3 in the heart of E12.5 Sox7-V5 transgenic reporter embryos. Gene co-expression analysis of the C2 cluster shows that around 6% of the endocardial progenitor cells are triple positive for these markers. smFISH for Gpc3 and Sox9 combined with V5 immunofluorescence confirms the presence of a subset of the trabecular endocardium that is triple positive for SOX7, Sox9, and Gpc3 ( Fig EV4A, orange arrows). Consistent with what we observed in the snRNA-Seq analysis when comparing control and Sox7 iECKO , we observed a general decrease in Gpc3-and Sox9-positive cells in the mutant condition. This is especially evident in the right ventricle endocardium, but not as pronounced in the left ventricle ( Fig EV4B). This observation confirms that a subset of the C2 population is lacking expression of genes with stemness potential during Sox7 loss of function.
To further address possible changes in the endocardial composition in Sox7 iECKO mutant hearts stemming from the diminished C2 endocardial cell population, we quantified cell number and proliferation rate in the endocardium as described above (Appendix Fig S1F-H, K-M, P-R and U-W). In these experiments, quantification was performed separately in both the basal and apical endocardium, which are known to display differential expression of genes and function (Del Monte-Nieto et al, 2018). Quantification of cell number identified no significant differences in the number of cells forming the endocardium or the basal and apical endocardium of Sox7 iECKO embryos at E12.5 and E13.5. However, there was a significant reduction in the number of endocardial cells in the right ventricle of Sox7 iECKO mutants at E14.5 with a clear decrease (non-significant) in the left ventricle (Appendix Fig S1F and K). Interestingly, this reduction in endocardial cells was significant exclusively in the basal endocardium of the right ventricle at E14.5, with no significant difference in the apical endocardium (Appendix Fig S1G,  Endothelial-specific deletion of Sox7 promotes an increase in the specification of trabecular cardiomyocytes in the developing heart The changes observed across the C2, C5, and C7 endothelial populations in Sox7 iECKO hearts are paralleled by an enrichment of Ó 2023 The Authors EMBO reports 24: e55043 | 2023 subpopulations representing cardiomyocyte (C4) and hematopoietic cell lineages (C9 and C13) (Figs 5B, E and F, and EV3B). This is consistent with the increase in cardiomyocyte populations characterized in the trabecular myocardium described above (Figs 2E-H and EV1Q, R, W and X). Unlike the other two cardiomyocyte clusters (C1 and C8), C4 shows higher expression of cardiac progenitor markers such as Nkx2.5 and Desmin (Lien et al, 1999;Fuchs et al, 2016). In contrast, higher expression of genes such as Slc8a1, Actn2, Cacna1c, and Ryr2 involved in electrophysiology and calcium handling suggests that C1 and C8 populations are in a more mature state ( Fig EV3D)  suggesting that these cells actively engage with protein translation and are metabolically active (Fig EV3D). It has been shown that early differentiating cells display higher levels of ribosomal protein expression (Sampath et al, 2008;Ingolia et al, 2011;Buszczak et al, 2014) suggesting that C4 is undergoing active differentiation. snRNA-Seq analysis revealed in the C4 cluster a more pronounced expression level of Nppa and Hopx, two known trabecular markers.
In order to further validate in vivo changes in the C4 cardiomyocytes subpopulation, we analyzed Nppa and Hopx gene expression pattern comparing control and Sox7 iECKO E12.5 embryos. As expected, Hopx and Nppa transcripts were found restricted to the trabecular myocardium in the ventricular chambers (Appendix Fig S5). Interestingly, whereas in control embryos, the expression of both markers is restricted toward the trabecular apex (arrows in Appendix  Fig S5A-F, left panels), in Sox7 iECKO mutant hearts, we observed an expansion of these markers to the base of the trabeculae (arrows in Appendix Fig S5A-F, right panels). Lastly, in parallel to this ectopic expression of trabecular markers, we observed a significant reduction in the expression levels of three markers specific to the compact myocardium (Kcnd2, Hey2, and Mycn) in the C1 population (Appendix Fig S5G) when comparing control and Sox7 iECKO , consistent with the thinning of the compact myocardium (Figs 2 and EV1).
Altogether these results suggest that the loss of Sox7 function in the endothelium compartment causes a dramatic expansion of the C4 trabecular cardiomyocytes at the expense of the C1 compact myocardium population. This result suggests that the C2 endothelial subpopulation is essential to preserve a proper ratio between the trabecular and compact myocardium cell populations.
Here, we established a correlation between the depletion of the C2 endocardial population and a parallel increase in a cardiomyocyte subpopulation undergoing differentiation in the developing ventricle of the Sox7 iECKO mutant. This suggests that the cardiomyocyte maturation process is under the influence of both the number of cardiac endocardial cells and their degree of differentiation. This observation is strengthened by analysis of the snRNA-Seq dataset using the monocle algorithm, which infers cell fate trajectory. To account for a bias caused by C2 depletion, we excluded the C2 cluster from the pseudo-time analysis. The loss of Sox7 function in the endothelial compartment strongly influences cell state transition within the cardiomyocyte subpopulations (Appendix Fig S6A  and B).
To discount the possibility that the loss of endothelial Sox7 may lead to the aberrant transdifferentiation of the C2 endocardial cells into C4 cardiomyocytes, we performed lineage-tracing analysis of Sox7 iECKO cells labeled with the mT/mG reporter system (Appendix Fig S7). In this assay, the Cre activity arrests the membrane-bound Tomato expression and instead, triggers the expression of membrane-bound EGFP. Analysis of EGFP-positive cells in the Sox7 iECKO samples did not reveal the presence of any discrete pools of EGFP + cardiomyocytes, suggesting that it is unlikely that the increase in the C4 cardiomyocyte population is caused by a fate shift in the endocardial progenitor population.
Endothelial-specific deletion of Sox7 alters the balance of erythro-myeloid lineages in the developing heart Since it is well established that Sox7 function maintains endothelial cell identity at the expense of the hemogenic endothelium compartment in the dorsal aorta, it is possible that the lack of SOX7 function prompts C2 cells to differentiate into an erythro-myeloid lineage. Consistent with these findings, the C9 cell cluster shows expression of hemoglobin genes (hba-a2, hbb-bt) and erythroblast markers (Ermap, Slc4a1, Nudt4, Tfrc), and is proportionately overrepresented in the Sox7 iECKO hearts (Figs 5B, C, E and F, and EV3E). Likewise, the C13 cell population that displays high levels of expression of genes associated with heart macrophages (Ptprc, Csf1r, and Lair1) is over-represented in Sox7 iECKO relative to control hearts (Figs 5C and F, and EV3E).
To validate the observed variation in the cellular composition of Sox7 iECKO hearts, we next performed fluorescence-activated cell sorting (FACS) analysis in E13.5 hearts using the pan-hematopoietic marker, CD45 (Ptprc), the early erythroblast marker, CD71 (Tfrc), and the pan-endothelial marker, CD31 (Pecam1) (Fig 6). To avoid contamination from circulating blood cells, all embryonic hearts were flushed with saline buffer prior to analysis (Fig 6A). Similar to the snRNA-Seq analysis where the C2 cluster is diminished in Sox7 IECKO hearts relative to controls, there was a consistent depletion of a fraction of the endothelial cell population (CD71 À CD45 À CD31 + ) in the Sox7 iECKO hearts. This was paralleled by an increase in a hematopoietic population that expresses CD31 (CD71 À CD45 low CD31 + ) ( Fig  6B-E), whereas the CD71 À CD45 low CD31 À remained unchanged (Appendix Fig S8A and  B). CD71 is a well-established marker of early erythroblast cells (Dong et al, 2011;Chao et al, 2015). Quantification of the CD71 high population also revealed a higher number of cells in this subpopulation in Sox7 iECKO hearts compared to control hearts (Fig 6F-I), thereby confirming at the experimental level, our initial observations. This suggests that loss of Sox7 function promotes the depletion of the C2 cardiac endothelial cell population, which may in turn give rise to an increase in the erythro-myeloid lineages in the heart at around E12.5-13.5. These results are supported by the known role of Sox7 as a gatekeeper of endothelial cell identity at the expense of the hemogenic endothelium (Gandillet et al, 2009;Costa et al, 2012;Lilly et al, 2016). Furthermore, the identification of the endocardium with hemogenic potential (Nakano et al, 2013;Zamir et al, 2017) suggests that SOX7 most likely plays a similar function to its role in the dorsal aorta in order to preserve a balanced distribution of endothelial and hemogenic compartments.
To investigate the possibility that the loss of endothelial Sox7 may lead to the aberrant transdifferentiation of the C2 endocardial cells into hematopoietic cells, we performed lineage-tracing analysis of Sox7 iECKO cells labeled with the mT/mG reporter system (Fig EV5A and B). The principle of this assay is similar to the lineage-tracing approach performed above with the cardiomyocyte populations. Analysis of the Sox7 iECKO ; mT/mG samples showed the presence of EGFP + /CD34 + -or EGFP + /CD45 + -positive cell pools in the coronary vasculature, suggesting that the increase in the C9/C13 hematopoietic populations is caused by a fate shift in the SOX7positive endocardial progenitor population.
Paracrine signaling from endothelial cells modulates cardiomyocyte behavior in vitro in a Sox7-dependent manner Given the marked changes in cardiomyocyte response when Sox7 function is compromised in the endothelial cell compartment, we next set out to assess the functional consequences in  Figure 6. FACs analysis confirms a higher proportion of hematopoietic lineage cells in Sox7 iECKO hearts.
A Schematic representation of the tamoxifen injection course, where the pregnant dams were injected with tamoxifen at E9.5 and E10.5 and hearts were harvested at E13.5 (left). Hearts were flushed with saline buffer to remove circulating blood cells before dissociation into single cells for FACs' analysis (right). B Flow analysis of hematopoietic and endothelial cells in control and Sox7 iECKO hearts. Results displayed a dramatic increase in the number of events associated with CD71 À CD45 low CD31 + hematopoietic cells (circled) but fewer CD71 À CD45 À CD31 + endothelial cells (rectangular box) in Sox7 iECKO hearts relative to the controls. C Graphs showing the proportion of CD31 + cells that express CD45 low or CD45 À in E13.5 control and Sox7 iECKO hearts. D, E Graphs showing the proportion of CD71 À CD45 À CD31 + endothelial cells (D) and CD71 À CD45 low CD31 + hematopoietic cells (E), both normalized to the CD71 À CD45 À/low parent events. F-I Flow analysis of early erythroblast (CD71 high , F; CD71 high CD45 À CD31 À , H) population, showing an increase in such events in the E13.5 control and Sox7 iECKO hearts.
(G) Graphs showing the proportion of CD71 high cells in CD71-positive population of the E13.5 control and Sox7 iECKO hearts. (I) Graphs showing the proportion of CD71 high CD45 À CD31 À cells in CD71-positive population of the E13.5 control and Sox7 iECKO hearts.
12 of 21 EMBO reports 24: e55043 | 2023 Ó 2023 The Authors cardiomyocytes resulting from depleting Sox7 in endothelial cells, by taking advantage of a human cardiac organoid (hCO) system (Mills et al, 2017). Briefly, SOX7 was knocked down in HUAECs using siRNA-mediated gene silencing ( Fig 7A). SOX7 levels in HUAECs were consistently depleted to 60-70% of control levels upon siRNA treatment, with these levels maintained across 5 days of cell culture (Fig 7B). Knockdown was performed using a transient transfection at D0 with a single SOX7 siRNA. These SOX7-depleted endothelial cells together with their respective controls were next seeded with human pluripotent stem cell-derived cardiomyocytes at Day 2 after transfection. The endothelial cell (EC)/human cardiac organoid (HCO) co-cultures were allowed to form for another 4 days before batch analysis of cardiac contractile parameters was performed for each condition (Fig 7A). Interestingly, the co-culture of control-treated endothelial cells and cardiomyocytes significantly increased the contraction rate as well as the activation and relaxation time of cardiomyocytes (Fig 7C-F), supporting the idea that endothelial cells enhance the overall physiological performance of cardiomyocytes. In the presence of endothelial cells with siRNA-depleted SOX7 levels, cardiomyocyte contractility showed a significant increase in both activation time and relaxation time when compared to control endothelial/cardiomyocyte co-culture conditions. This supra-physiological increase in relaxation time is comparable to effects caused by known pro-arrhythmic drugs such as propranolol or E4031 (Hoang et al, 2018). The dramatic increase in relaxation and activation time seen in SOX7-depleted conditions may relate to alterations in the extra cellular matrix composition and/or stiffness, or to altered levels of angiocrine signaling from endothelial cells. Our data suggest that in vitro depletion of SOX7 function in endothelial cells is sufficient to influence the contractile properties of cardiomyocytes co-existing in culture.
Overall, several lines of experimental evidence support the idea that modulating endothelial cell identity by interfering with Sox7 function has a dual outcome: (i) It leads to a lack of maintenance in endocardial cell identity which in turn disrupts the fine balance in the distribution of cardiac cell types, and (ii) these perturbations of cell identity promote morphological and functional cardiac defects caused by the alteration of heterotypic cell interaction between cardiac tissues.

Discussion
In this study, we have identified Sox7 as a key regulator of cardiac endothelial cell identity and showed that SOX7-dependent cross-talk relaxation time (F). Overall contraction parameters AE SD from n = 31 endo/cardio-hCO cultured for 5 days in CTRL medium (10-11 hCO per condition). ****P < 0.0001 using one-way ANOVA with Tukey's post-test.
Ó 2023 The Authors EMBO reports 24: e55043 | 2023 between endothelial cells and cardiomyocytes controls the process of cardiac trabeculation and chamber myocardium compaction. At the cellular level, we showed that endothelial-specific loss of Sox7 perturbs the finely tuned balance of different cardiac cell types during heart embryogenesis. Loss of endothelial Sox7 promotes the reduction of an endothelial progenitor population and this results in the dysregulation of cardiomyocyte populations and erythromyeloid lineages. We demonstrated that the dysregulation of Sox7 in the endocardium promotes a hyper-trabeculation phenotype from E12.5-E13.5 that is not associated with increased proliferation of cells in the trabecular myocardium, but due to an increase in the allocation of cardiomyocytes to the trabecular layer from the compact myocardium.
Our results identified a reduction in the number of endocardial cells at the trabecular base, suggesting that the endocardial population depleted in the Sox7 iECKO mutant embryos may derive from this location, and therefore, be involved in the control of cellular process occurring in the trabecular myocardium base. Cells of the trabecular myocardium are in an advanced differentiation state compared to the compact myocardium. Growth of the trabecular myocardium occurs primarily from the trabecular base, with oriented cell division allocating compact layer cardiomyocytes into the trabecular layer, where they differentiate into trabecular cardiomyocytes (de Boer et al, 2012;Li et al, 2016).
Our results suggest that endocardial Sox7 may control the process by which cardiomyocytes are allocated to and differentiate in the trabecular layer. The increased undifferentiated cardiomyocyte population identified in the Sox7 iECKO mutant samples in snRNA-Seq experiments may therefore represent a population of cardiomyocytes that have entered the trabecular layer but failed to totally differentiate into trabecular myocardium due to the lack of inductive signals from the missing basal endocardial population controlling the process. Therefore, this study positions Sox7 as the key endothelial regulator of chamber myocardial morphogenesis and shows that signaling downstream of Sox7 controls myocardial cell lineage determination and cellular behavior during heart development.
Our data further show Sox7 is essential for coronary artery specification, and suggest that this is likely due to direct transcriptional regulation of Notch signaling and Connexin molecules. However, unlike the phenotype caused by the chromatin remodeler Ino80 loss of function, the hyper-trabeculation and non-compaction defects in the Sox7 iECKO mutant hearts are not related to coronary angiogenesis impairment. Instead, the coronary molecular and morphological phenotype identified in Sox7 iECKO mutants suggest that the established molecular relationship among SOXF, Notch, and Connexin signaling controls the process of arterialization and hierarchization of coronary vessels during the late stages of heart formation.
The role of sox7 in arterial specification has been reported in zebrafish vascular endothelium (Hermkens et al, 2015). Studies in mice confirmed its role and showed that Sox7 controls arterial development through transcriptional regulation of the Notch ligand, Dll4, and its receptor, Notch1 (Sacilotto et al, 2013;Chiang et al, 2017). In addition to its known role in the vascular endothelium, here we show that Sox7 is also required for coronary artery formation. Our data also confirm Dll4 as a downstream transcriptional target of SOX7. We also report additional direct target genes, including Cx37 and Cx40, which encode gap junction proteins that play a key role in cell-cell communication.
Both Cx37 and Cx40 are expressed in the blood vascular endothelium, including in arterial endothelial cells and smooth muscle (Simon & McWhorter, 2002;Haefliger et al, 2004;Buschmann et al, 2010). Cx37 has been shown to enable arterial specification through cell cycle arrest in a mouse retinal model (Fang et al, 2017). Cx40 has been linked to flow-driven arteriogenesis and collateral arterial network development (Buschmann et al, 2010). Loss of Cx40 was further shown to potentiate the appearance of arteriovenous shunts in Alk1haploinsufficent mice (Gkatzis et al, 2016), implicating Cx40 as an important component in the maintenance of arterial cell identity. Recently, through a single-cell transcriptomic analysis during heart development, Su et al (2018) identified Cx40 as a marker of arterial progenitor cells in the sinus venosus during coronary artery formation. This study showed that expression of Cx40 is a critical intermediary step for arterial specification. In line with and expanding on these findings, our data suggest that Sox7 modulates the emergence of prearterial cells via the direct regulation of a gene regulatory network that involves Cx40, Cx37, and Notch effectors (Sacilotto et al, 2013;Chiang et al, 2017).
Similar to the phenotype observed in the present study, failure to form the coronary arteries was reported in a mouse model of endothelial-specific loss of Sox17 function (Gonz alez-Hern andez et al, 2020). Although the downstream molecular cascade regulated by SOX17 to instruct coronary artery formation remains to be elucidated, our study establishes that both Sox7 and Sox17 functions are indispensable for this process. While it is possible that SOX17 might regulate overlapping SOX7 target genes, such as those involved in Notch signaling (Corada et al, 2013), the lack of compensation by either molecule when the other is inactivated suggests they play non-redundant roles in artery formation. This is reinforced by our finding that SOX17 protein expression levels ( Fig EV2B) remained unaffected in the absence of Sox7. Hence, the loss of coronary arteries due to altered Sox7 function is not mediated through perturbed Sox17 activity. Although both Sox7 and Sox17 are involved in coronary artery development, only Sox7 is expressed in endocardium by E11.5 (Fig 1A and B; Gonz alez-Hern andez et al, 2020), suggesting that Sox7 and Sox17 have distinct molecular roles in the cardiac endothelium, especially at later developmental stages. It still remains to be determined if depleting a Sox7-positive endocardial cell pool directly contributes to the lack of coronary artery progenitors, or whether there is a failure of arterial differentiation independent from the endocardial contribution.
A role for SOXF transcription factors during endocardium differentiation has been reported for both SOX17 and SOX7 (Doyle et al, 2019;Saba et al, 2019;Hong et al, 2021). While Sox17 is thought to regulate cardiomyocyte maturation in a non-cell autonomous manner, Sox7 was proposed to directly promote the endothelial lineage cell fate at the expense of the cardiac lineage through positive regulation of Wnt and BMP signaling in embryoid bodies (Doyle et al, 2019). Echoing this in vitro finding, we report that deletion of Sox7 in cells committed to an endothelial lineage causes an increase in a subpopulation of cardiomyocytes. Nonetheless, lineage-tracing analysis using the Cdh5-CreERT2:Sox7 fl/fl ;mT/mG mouse did not reveal any GFP-positive cardiomyocytes, suggesting that the expansion of the cardiomyocyte population observed in Sox7 iECKO mutants is not due to transdifferentiation from a committed endothelial lineage into the cardiomyocyte lineage (Appendix Fig S7). However, we cannot discount the possibility that SOX7 plays a critical role at an early time point prior to fate 14 of 21 EMBO reports 24: e55043 | 2023 Ó 2023 The Authors specification when progenitor cells retain the bipotentiality to give rise to either endothelial or cardiac lineages. Our data are consistent with the possibility that the increase in cardiomyocyte number in the absence of Sox7 is due to non-cell-autonomous effects from the cardiac endothelium to the myocardium, akin to what has been described in mice that are deficient for Sox17, Ino80, and Notch mutants, Fkbp1a and Jarid2 (Mysliwiec et al, 2011;Chen et al, 2013;Saba et al, 2019;Rhee et al, 2021). In addition to a failure to specify coronary arteries, our work shows that loss of Sox7 function depletes a subpopulation of endothelial cells in the developing heart. The enriched expression of endocardial marker Nrp3 and the absence of coronary markers Fabp4 and Apelin suggest that this population is mainly endocardial in nature. This observation coincides with an increase in the overall proportion of hematopoietic cells, specifically the CD45 low CD31 + hematopoietic cells and CD71 high CD45 À CD31 À erythroblasts (Fig 6), and a subset of cardiomyocytes. Cells expressing both CD45 and CD31 have been previously associated with the hematopoietic stem cell niche (Shaw et al, 2004). Interestingly, Sox7 has previously been shown to play a critical role in directing endothelial specification and lineage maintenance in the hemogenic endothelium, with its downregulation required for the subsequent emergence of hematopoietic progenitors (Gandillet et al, 2009;Costa et al, 2012;Lim et al, 2016). Akin to both the dorsal aorta and the yolk sac vasculature, the heart endocardium has been shown to have the capacity to undergo endothelial-tohematopoietic (EHT) transition from E9.5 in the mouse (Nakano et al, 2013). This transition is observed in a specific group of cells integrated into the outflow cushion and atria called the hemogenic endocardial cells. Later in development between E11 and E14, aggregates of endothelial and hematopoietic cells called the blood islands are formed, emerging from the endocardium (Red-Horse et al, 2010;Jankowska-Steifer et al, 2015). Consistent with the role of Sox7 in the hemogenic niche in other organs, analysis of E13.5 embryonic hearts from a Cdh5-CreERT2:Sox7 fl/fl ; mT/mG reporter line shows that GFPpositive endothelial cells (e.g., Sox7 null) harbor the expression of hematopoietic stem cell markers CD34 and pan-hematopoietic marker CD45 (Fig EV5). These results suggest an important role for SOX7 in the maintenance of the endocardial identity by preventing its transdifferentiation toward a hematopoietic fate.
In summary, we have identified Sox7 as a key regulator of cardiac endothelium identity. The present study further supports the pivotal requirement for a specified, differentiated endothelium to instruct heart morphogenesis (Mysliwiec et al, 2011;Chen et al, 2013;Qu et al, 2019;Sandireddy et al, 2019;Rhee et al, 2021). The finely tuned regulation of the number and state of cardiac endothelial cells by Sox7 and its downstream effectors is intricately linked to the distribution of other cardiac cell types, which in turn determines tissue architecture, assembly, and function during cardiac development. This finding sheds light on a potential novel etiological component of non-compaction cardiomyopathies.

Transgenic mice and fish
Mice used in this study were Sox7:tm1 (Chiang et al, 2017), Cdh5-CreERT2 (Wang et al, 2010), Sox7 fl/fl (Lilly et al, 2017), Sox7-V5 (Chiang et al, 2023), Sox7 Dex2/Dex2 (Wat et al, 2012), and mT/mG (Muzumdar et al, 2007). Sox7 endothelial-specific conditional knockout (C57BL/6) was a cross between Cdh5-CreERT2 and Sox7 fl/ fl . To delete Sox7 in endothelial cells, two consecutive intraperitoneal injections of tamoxifen (T5648, Sigma Aldrich), at 2 mg per pulse, were administered to dams at E9.5 and E10.5, and embryos were collected at the indicated time points. To detect the primary gene targets of SOX7, pregnant dams were pulsed with tamoxifen at E9.5, and embryos were harvested for FACs sorting 24 h after Cre induction. Cx40-21:GFP was generated as described in Chiang et al (2017) by cloning a 566 bp PCR fragment from human genomic DNA into the zebrafish enhancer detector (ZED) vector (Bessa et al, 2009). All animal work was approved by the relevant ethics committees of the University of Queensland and University of Sydney ethics ID: 084/19 Experimental Analysis of Lymphangiogenesis in Development and Regeneration.

Immunofluorescence staining and tissue sectioning
Mouse embryos were fixed in 4% paraformaldehyde for either 1 h or overnight at 4°C, depending on the developmental stage, and then washed three times in PBS. Tissue embedding and sectioning were performed according to standard protocols. Immunofluorescence staining of mouse embryos was performed as described in Chiang et al (2023). Immunofluorescence of tissue sections for the characterization is shown in Figs 2  Secondary antibodies were as follows: donkey anti-rat IgG Alexa 488 (A21208), goat anti-mouse IgG Alexa 594 (A11005), donkey anti-rabbit IgG Alexa 647 (A31573), donkey anti-goat IgG Alexa 647 (A21447), and donkey anti-rabbit Alexa 594 (A21207). Secondary antibodies were sourced from Invitrogen and used at 1:300 unless specified otherwise. For morphological characterizations, CD31 was amplified using anti-rat-HRP (1:100, A18745, Thermo Fisher Scientific) followed by tyramide-Cy5 amplification (NEL745001KT; Perkin Elmer); Ki67 was amplified using anti-rabbit-Cy5 (1:100, A31572, Thermo Fisher Scientific).

Cell isolation for snRNA-Seq
Cdh5-CreERT2, Sox7 fl/fl males were crossed with Sox7 fl/fl females, who were dosed with tamoxifen at E9.5 and E10.5. Hearts were harvested from E12.5 embryos, and immediately snap frozen in liquid nitrogen before storage at À80°C. Subsequently, four hearts from similar genotypes were pooled, and single nuclei extraction Ó 2023 The Authors EMBO reports 24: e55043 | 2023 and isolation were performed as described, with modifications (Sim et al, 2021). Briefly, isolation buffer was 0.3 M sucrose, 10 mM Tris-HCl, 5 mM magnesium acetate, 5 mM CaCl 2 , 2 mM EDTA, 0.5 mM EGTA, and 1 mM DTT. Pooled hearts were suspended in isolation buffer and homogenized with a 15 ml tissue grinder (357544, Edwards Wheaton). Samples were then filtered through a 40 lM cell strainer, washed, and finally, resuspended in PBS saline buffer containing Hoechst stain (1:1,000, Invitrogen). Nuclei were then sorted using a BD Influx TM Cell Sorter with a 70 lM nozzle, before loading onto the 10× chromium single-cell chip (v3, 10× Genomics).

Imaging and data analysis
Images were captured with a Zeiss LSM710 META BIG, Zeiss LSM 710 FCS or Leica TCS SP8 HyD confocal microscopes with the 10×, 20× or 40× oil objectives. Images were analyzed using the Bitplane IMARIS suite and Image J. Imaging was performed in the Australian Cancer Research Foundation (ACRF)'s Dynamic Imaging Facility at the Institute for Molecular Bioscience (University of Queensland) and the Sydney Cytometry facilities at Centenary Institute (University of Sydney). All graphs and statistical tests were performed with Graphpad Prism 9 and illustrated with Adobe Illustrator CS6. Morphological, cell count, and proliferation experiments were imaged using a Leica SP8 inverted microscope at Monash Microimaging platform.

Image quantification and data analysis
Image analysis to perform tissue area, quantification of cell number, and actively proliferating cells was performed in Fiji. Image channels were isolated and overlaid with a mask for area quantification and nuclear classification performed via intensity thresholding. We generated endocardial, myocardial, and epicardial masks. Myocardial masks were further segmented into trabecular and compact myocardial masks by manual selection of the compact layer boundary following endocardial touchdowns as described in Del Monte-Nieto et al (2018). The luminal boundary of the trabecular myocardium and endocardium was found by iteratively expanding this compact layer selection in a pixel-wise manner until it encompassed the entirety of the tissue; the distance at which all tissue was enclosed was labeled as the luminal apical boundary, with the basal/apical boundary defined as halfway between the compact myocardium and the apex of the trabeculae. This allowed further sub-segmentation of the endocardial and trabecular myocardial masks into basal and apical layers. Nuclear segmentation was performed with the StarDist plugin in Fiji (Schmidt et al, 2018). Individualized nuclei were subsequently classified using the previously generated tissue masks. This identified nuclei from cells forming the compact myocardium, basal and apical trabecular myocardium, epicardium, and basal and apical endocardium. The average intensity of Ki67 staining within each nucleus was then measured, and nuclei were further classified as either Ki67 + or Ki67 À based on a calibrated threshold; positive nuclei were labeled red and negative blue. The data collected were scaled to the characteristic length of the specific tissue area analyzed and normalized to the control value. Once scaled and normalized, data were plotted using GraphPad Prism. Data were analyzed for normal distribution using Shapiro-Wilk test. Once confirmed, statistical analysis was performed using parametric Student's t-test with significance achieved when P-value was < 0.05. Data were represented in the graph as mean AE SEM. For a summary of the quantification process, see Appendix Fig S9.
To validate the number of hematopoietic and cardiac endothelial cells observed in snRNA-Seq (Figs 5 and EV3), control and Sox7 iECKO mutants hearts at E12.5 were dissected in ice-cold PBS, and circulating blood cells were flushed using a microinjection needle filled with saline PBS. Hearts were dissociated in 0.071 mg/ml liberase in PBS buffer, supplemented with 0.25 mg/ml collagenase II, 0.25 mg/ml collagenase IV, 1 mg/ml deoxyribonuclease I, 0.9 mM CaCl 2 , and 10 mM Hepes, followed by incubation at 37°C for 5 min and pipetting with P200 pipette tips every 2 min to assist tissue dissociation. Tissue digestion was stopped by addition of 2 mM CaCl 2 and 10% fetal calf serum. Digestion buffer was next removed before passing the dissociated samples through a 70 lm cell strainer, washed once then resuspended in Fc Block (1:200, BD Bioscience) followed by 10 min blocking on ice. Dissociated cells were stained with appropriate antibodies for 0.5 h on ice. Antibodies used were Alexa 488-conjugated rat anti-CD31 (1:200, 563607, BD Pharmigen), Brilliant Violet 510-conjugated rat anti-CD71 (1:200, R17217, BioLegend), and PE/Cyanine 7-conjugated rat anti-CD45 (1:400, 30-F11, BioLegend). Live/dead cells were detected by DAPI (1:1,000, D9542, Sigma Aldrich). For compensation, UltraComp eBeads (Invitrogen) were used as per manufacturer's instruction. Samples were washed twice, and then two hearts from similar genotypes were pooled (following genotyping by PCR). Pooled samples were finally resuspended in 200 ll 1% FCS/PBS/2 mM EDTA solution containing DAPI. Cell sorting was performed on a BD FACSAria TM Cell Sorter with a 100 lM nozzle. FACs sorting was performed in Sydney Cytometry Facilities at Centenary Institute (University of Sydney).
Monocle3 (Trapnell et al, 2014) package (v3.0.0) in R (version 4.1.1) was used to analyze pseudo-time of development on integrated data in Appendix Fig S6. For the primary analysis, all cell clusters from either control or Sox7 iECKO mutant hearts were subsetted into separate Seurat objects and then analyzed with Monocle3 using the default parameters. For C2 negative analysis, the subsetted control or Sox7 iECKO Seurat objects were then further subsetted to remove all but five C2-defined cells before being subjected to Monocle3 analysis. When displaying Monocle3 data, we replaced the Monocle3-defined clusters with our predefined cluster identifications.

RNA-Seq and analysis
For RNA-Seq on FACS-sorted mouse endothelial cells at E10.5, triplicate samples were processed for whole transcriptome sequencing using the Smart-Seq 2 method as described in Picelli et al (2014). Samples were sequenced using the Illumina TM HiSeq 2500 platform. Reads were then mapped to GRCm38/mm10 mouse reference genome using STAR aligner (Dobin et al, 2013). Only unique aligned reads were considered. Next, transcripts were assigned to mouse genes using htseq_count in the HTSeq python package (Anders et al, 2015) and differential expression between sibling control and Sox7 iECKO mutants was calculated using DEseq2 (Love et al, 2014) and EdgeR (Robinson et al, 2010). Only genes with adjusted Pvalue < 0.05 in both DEseq2 and EdgeR were considered significant.
Heatmaps for the RNA-Seq results in Appendix Fig S4A and C were generated using pheatmap R package (Kolde et al, 2012), while the bubble chart depicting gene ontological terms in Appendix Fig S4B was plotted with ggplot2 R package (Wickham, 2009).

Heart dyno assay and hCO fabrication
Cardiac cells were differentiated, and hCO and Heart-Dyno culture inserts were fabricated as described in Mills et al (2017).
All cell lines used in this study have tested negative for mycoplasma contamination.

In situ hybridization assay
Whole-mount and section in situ hybridization was performed as described in Fowles et al (2003) and Metzis et al (2013). Cx37 (clone ID: 4971608) probe was from Dharmacon. In situ hybridization on tissue sections was performed as described in Kanzler et al (1998). Cx40 probe was generated by Dr. Gonzalo del Monte-Nieto while working at Prof. Richard Harvey Laboratory. Please contact Dr. Gonzalo del Monte-Nieto for more details.

Graphics
The synopsis graphics were created with BioRender.com

Data availability
Bulk RNA-seq and snRNA-seq datasets have been deposited to the GEO repository under unique identifiers: (i) snRNAseq and bulk RNAseq in conditional Sox7 knock-out: Gene Expression Omnibus GSE231636 (   Figure EV2. SOX7 transcriptionally regulates the arterial specification marker, Cx40. A Serial sections of a wild type and Sox7 À/À hearts at E10.5, stained with endomucin (EMCN, white) to detect the endothelial lining. The dorsal aorta (DA) is labeled by red arrows, and the cardinal vein (CV) by blue arrows. Fusion between the DA and CV is indicated by orange arrowheads. B Whole-mount immunostaining of Sox7 iECKO mutant and sibling control skin at E14.5, after Cre induction by tamoxifen injection at E9.5 and E10.5. The blood plexus is marked by endomucin (green), arteries by Cx40 (red), and SOX17 staining is shown in white. C Schematic showing that SOX transcription motifs are the top binding motifs in SOX7 HUVECs ChIP-Seq data. D Schematic representation of the human CX40 locus showing a 500 bp putative regulatory element situated 21 kb upstream from the transcription start site (TSS) (denoted as CX40-21 region) from UCSC Genome browser. The H3K27Ac is denoted in light blue, DNAseI hypersensitive hotspots are indicated by black/gray boxes, where the darkness is proportional to the maximum signal strength observed in any cell line. The chromatin state in HUVECs is shown in orange (indicates strong enhancer), green (weak transcribed), yellow (weak/poised enhancer), purple (inactive/poised promoter), and gray (polycomb repressed regions). E The CX40-21:GFP transgene directs GFP fluorescence expression to vascular endothelium in transgenic zebrafish larvae at 4 dpf.

A
There is a lack of transcripts harboring the exon 2 (floxed exon) of the Sox7 locus in the Sox7 iECKO hearts (red track). B Graphs showing the different cell types comprising E12.5 control and Sox7 iECKO hearts. C-E Dot plots showing expression of genes that define each subcluster within major cell type population: cardiac endothelial cells (C), cardiomyocytes (D), and blood/ immune cells (E).   Figure EV4. Triple-positive Gpc3, Sox9, and Sox7 endocardial cells in 12.5 developing hearts are depleted in the Sox7 iECKO .
A smFISH for Gpc3 and Sox9 combined with V5 immunofluorescence in Sox7-V5 transgenic reporter embryos shows the presence of a subset of trabecular endocardial cells that are triple positive for SOX7, Sox9, and Gpc3 (orange arrows). Pink asterisks, SOX7 single-positive cells. High magnification from inset with 1, 2, and 3 showing individual channels for each marker; 4 shows the plot profile of fluorophore distribution for triple-or single-positive cells. Scale bar in 3 and 10 lm. B Expression pattern of C2 population markers. Analysis by smFISH of the C2 endocardial markers Gpc3 (left panels), Sox9 (right panels) in combination with immunofluorescence of endocardial (endomucin, blue), and myocardial (SMA, red) markers on tissue sections of control and Sox7 iECKO mutant hearts. The first row panels are low-magnification images showing the entire heart tissue. The second row panels show images of the right ventricles. The third row panels show the left ventricle. The two bottom row panels show a high magnification from inset in RV panels and reveal the loss of Gpc3 and Sox9 expression in the endocardium of the Sox7 iECKO . Markers (green), myocardium (red), and endocardium ( A Transverse section of an E13.5 heart from a triple transgenic SOX7 iECKO ; mT/mG embryo stained for GFP, CD45/CD34 (same fluorophore), and DAPI. This staining shows the presence of GFP-positive endothelial cells in the coronary vasculature that express either CD34 or CD45 hematopoietic markers (red and black or white arrows). Scale bar = 20 lm. B Plot profile analysis of a cell from the inner lining of a vessel showing the colocalization of the fluorescent signal. These regions are blown-up areas indicated by the red arrows. Scale bar = 10 lm.