OUP user menu

Cardia bifida, defective heart development and abnormal neural crest migration in embryos lacking hypoxia-inducible factor-1α

Veerle Compernolle , Koen Brusselmans , Diego Franco , Antoon Moorman , Mieke Dewerchin , Désiré Collen , Peter Carmeliet
DOI: http://dx.doi.org/10.1016/j.cardiores.2003.07.003 569-579 First published online: 1 December 2003


Objectives: Previous studies have revealed the essential role of hypoxia-inducible factor-1α (HIF-1α), a basic helix-loop-helix transcription factor, in cardiovascular development. We attempted to further characterize the underlying mechanisms resulting in abnormal cardiogenesis and defective angiogenesis in mice deficient for HIF-1α (HIF-1α−/−). Methods: We analyzed cardiovascular development in HIF-1α−/− embryos at both the macroscopic and microscopic level. Gene expression was determined by RT-PCR, in situ hybridization and immunohistochemistry. Embryonic survival was studied using whole embryo culture. Results: HIF-1α deficiency caused cardia bifida in some embryos, while cardiac looping was disturbed in others. These defects did not result from abnormal cardiomyocyte commitment or differentiation, but may relate to defective ventricle formation caused by reduced expression of myocyte enhancer factor 2C (MEF2C) and eHAND. In addition, remodeling of the aortic outflow tract and cephalic blood vessels was abnormal in HIF-1α−/− embryos. These malformations, together with the hypoplastic pharyngeal arches, are presumably induced by defective neural crest cell (NCC) migration. Impaired migration might be related to insufficient levels of semaphorin-3A (Sema3A). Hyperoxia prolonged survival but only partially rescued the developmental program of cultured HIF-1α−/− embryos. Conclusion: HIF-1α is essential for proper cardiac development by modulating both neural crest migration and ventricle formation.

  • Developmental biology
  • Embryology
  • Morphogenesis
  • Myocytes
  • Endothelins

1. Introduction

Low oxygen concentrations induce an entire spectrum of cellular and systemic responses [1]. Hypoxia-inducible factor-1 (HIF-1), a heterodimeric transcription factor formed by HIF-1α and HIF-1β, is stabilized under hypoxic conditions and binds the hypoxia-response element (HRE) in hypoxia-sensitive genes, thereby enhancing their transcription [1]. While previous studies already documented an essential role of HIF-1α in cardiovascular development [2,3], developmental and molecular mechanisms resulting in abnormal cardiogenesis remain undetermined.

In vertebrates, the formation of the heart initiates soon after gastrulation. Cardiogenic commitment of progenitor cells within the anterior lateral plate mesoderm results in the formation of horseshoe-shaped cardiac crescent, which, upon folding of the embryo, forms the primary linear heart tube. Subsequently, the heart tube bends to the right, a process called cardiac looping. Five distinct segments form along the heart tube: the sinus venosus, atria, atrioventricular canal, ventricles and outflow tract [4]. Although HIF-1α is expressed in the early heart and in cultured embryonic ventricular myocytes [5], the role of HIF-1α in cardiogenic commitment, heart tube formation, looping and ventricle formation is unknown. Remodeling of the outflow tract requires proper development of neural crest cells (NCC), which populate mesenchyme in the conotruncus and around the aortic arch arteries. Abnormal neural crest development has been suggested in the HIF-1α−/− embryos, but has not been established to date [6].

Several angiogenic growth factors including vascular endothelial growth factor (VEGF) are upregulated by hypoxia and responsible for the formation of new blood vessels. HIF-1α−/− embryos display severe defects in angiogenesis [2,3], but surprisingly, previous studies documented an increase in VEGF expression in these embryos [6]. Therefore, it remains undetermined which angiogenic candidate genes are responsible for the abnormal vascular remodeling in HIF-1α−/− embryos.

2. Materials and methods

2.1. Genotyping and animals

DNA was isolated from mouse tail tips or embryonic yolk sacs and subjected to PCR analysis for genotype determination using the following primers: 5′-CAA GCA TTC TTA AAT GTG GAG CTA TCT-3′; 5′-TTG TGT TGG GGC AGT ACT GGA AAG ATG-3′; and 5′-CGA AGG GGC CAC CAA AGA ACG CAG CCG-3′ yielding a 380-bp band specific for the HIF-1α+/+ allele and 230-bp band specific for the HIF-1α−/− allele. Housing and procedures involving experimental animals were approved by the Institutional Animal Care and Research Advisory Committee of the K.U. Leuven, Belgium.

2.2. Quantitative real time PCR analysis

Gene expression was quantified, relative to the expression of β-actin, by real time PCR as previously described [7]. Primers and probes are listed in S1.3

2.3. Embryo culture, immunohistochemistry and in situ hybridization

Embryos were dissected at E8.5 and cultured in sterile 50-ml Pyrex bottles (Merck Eurolab) filled with 3.2-ml human serum and 0.8-ml rat serum. Culture bottles were flushed every 12 h with humidified gas mixtures containing 20% O2–5% CO2–75% N2 (normoxia) or 40% O2–5% CO2–55% N2 (hyperoxia). Culture medium contained 300 mg/dl glucose (normoglycemia) or 900 or 1200 mg/dl glucose (hyperglycemia). The presence of both heartbeats and circulating red blood cells were used to evaluate the viability of the embryos.

Antibodies against desmin (ICN/Cappel, Asse, Belgium), myosin (light and heavy chain) (Sigma, Bornem, Belgium), smooth muscle cell actin (DAKO, Carpinteria, CA, USA), platelet endothelial cell adhesion molecule (PECAM, Dejana) and P75 (D. Anderson) were used for immunostaining. Apoptotic cells were determined using the In Situ Cell Death detection Kit (Boehringer Mannheim, Mannheim, Germany). Dead cells were detected by incubating embryos in Nile blue sulfate (NBS, 0.01% in PBS) for 30 min. Whole mount in situ hybridization was done as described [8].

2.4. Statistical analysis

Numerical data are represented as mean±standard error of the mean (S.E.M.). Statistical analysis of data was performed using the Students' t-test or Fisher's Exact test (only if indicated). A P-value<0.05 was considered significant.

3. Results

3.1 Generation of HIF-1α−/− mice

Targeted inactivation of the HIF-1α gene was achieved by deletion of the basic helix-loop-helix domain, responsible for DNA binding and dimerization [9]. Targeted embryonic stem cell clones were used for the generation of heterozygous HIF-1α-deficient (HIF-1α+/−) mice, which appeared normal and healthy. Analysis of embryos, obtained by HIF-1α+/− intercrosses, at embryonic day (E) 8.5 and E9.5 revealed a Mendelian distribution (S2a)3. Our HIF-1α−/− embryos had a mixed 129/SvJ × Swiss genetic background in contrast to the previously generated HIF-1α−/− embryos, which had a mixed C57BL/6 and 129 genetic background [2,3] (S2b)3. All HIF-1α−/− embryos had a normal number of somites at E8.5 (n = 13) but were growth-retarded thereafter (S2c)3. Macroscopic examination of E9.5 HIF-1α−/− embryos revealed severe defects including an open neural tube, heart malformations and a single abnormal pharyngeal arch, as previously reported [2,3,6].

3.2 Cardia bifida and abnormal cardiac looping in HIF-1α−/− embryos

Gross anatomical analysis of 10 wild type (WT) and 14 HIF-1α−/− embryos at E8.5 and of 18 WT and 28 HIF-1α−/− embryos at E9.5 revealed that cardiac development was severely abnormal in HIF-1α−/− embryos (Table 1), even after correction for growth retardation by evaluating embryos with a comparable number of somites. Loss of HIF-1α primarily caused two distinct types of cardiac defects. The earliest and most severe defect was the occurrence of a cardia bifida in 30% of mutant embryos. Whereas in WT embryos, a horseshoe-shaped cardiac crescent is formed (Table 1) and endocardial cells continuously lined the lumen of the primitive heart at the 5–6 somite stage (Fig. 1a,c), the horseshoe-shaped cardiac crescent failed to form in one third of HIF-1α−/− embryos, even when they developed to the 5–10 somite stage (Table 1) (occurrence of cardia bifida in 17 WT vs. 28 HIF-1α−/− embryos with 5–10 somites; P<0.05; Fisher's Exact test). As a result, two separate myocardial tubes developed that were each lined by endocardial cells (Fig. 1b). In some HIF-1α−/− embryos (3 out of 7), a single myocardial tube developed around a labyrinth of endocardial channels (Fig. 1d).

Fig. 1

Cardia bifida and defective looping and ventricle formation in HIF-1α−/− embryos. (a–d) At E9.0–E9.5, the trabeculated myocardium (arrow) surrounded a single endocardial channel (arrowhead) in WT embryos (SMA staining in a and PECAM in c). In contrast, in one third of HIF-1α−/− embryos, the horseshoe-shaped primordium failed to form, leading to two endocardial tubes (indicated by “1” and “2”), which were each surrounded by cardiomyocytes (SMA staining in b). In a subgroup of HIF-1α−/− embryos, a single myocardial tube developed around a labyrinth of endocardial channels (arrowhead in d; PECAM staining at E9.0). The myocardium is not trabeculated in b and d. SV: sinus venosus; V: primitive ventricle. Scale bar: 50 μm. (e–k) In situ hybridization expression patterns in E9.25 embryos. (e and f) ANF is normally expressed in WT (e) and HIF-1α−/− (f) embryos. Cardiac looping occurred normally in WT (e) but not in HIF-1α−/− (f) littermates. (g and h) In situ hybridization for eHAND on embryos from which the extraembryonic membranes were only partially removed. Expression of eHAND was detectable in the extraembryonic membranes of WT and HIF-1α−/− embryos (arrowheads). eHAND was also expressed in the left ventricle of WT (arrow in g) but not of HIF-1α−/− (arrow in h) embryos. (i–k), MEF2C expression was detectable in the heart of WT (arrow in i) but not of HIF-1α−/− embryos (arrow in j). MEF2C was present in the brain of WT embryos (asterisk in i), but only slightly detectable in localized regions in the hindbrain in HIF-1α−/− embryos (arrowhead in k). (l) Transcript levels of MEF2C were decreased in HIF-1α−/− embryos as compared to WT littermates. Data represent the mean±S.E.M. of the number of mRNA copies per 100 mRNA copies of β-actin (N = 6; *:P<0.05).

View this table:
Table 1

Abnormal cardiac development in HIF-1α−/− embryos

  • The percentages indicate the incidence of WT and mutant embryos at E8.5 and E9.5 exhibiting various cardiac developmental phenotypes: cardia bifida (A), horseshoe-shaped heart (B), single unlooped heart tube positioned at the midline (C), looping heart (D), looped heart (E) and the embryonic chambered heart (F). The latter indicates the presence of macroscopically visible primitive ventricles.

In contrast to WT embryos, whose heart tubes were completely and normally looped at the 11–20 somite stage (N = 9), only two out of fifteen 11–20 somite-stage HIF-1α−/− embryos had a correctly looped heart (P<0.05, Fisher's Exact) (Table 1; ANF staining in Fig. 1e,f). Myocardial trabeculation, initiated in WT embryos at E9.0 (Fig. 1a,c), was significantly retarded in half or absent in the other half of the HIF-1α−/− embryos at E9.5 (Fig. 1b,d). Since similar cardiac defects were documented in angiopoietin-1 (Ang-1)-deficient embryos [10], we analyzed whether Ang-1 levels were abnormal in the HIF-1α−/− embryos. Retarded trabeculation in HIF-1α−/− embryos was associated with decreased expression of Ang-1 (copies Ang-1 per 100 mRNA copies of β-actin: 18±1.7 in WT vs. 9.3±0.7 in HIF-1α−/− embryos, N = 6; P<0.05) and its receptor TIE2 (copies TIE2 per 100 mRNA copies of β-actin: 12±0.7 in WT vs. 9.3±0.7 in HIF-1α−/− embryos, N = 6; P<0.05). Cardiac development was more severely affected than overall development, suggesting that cardiac defects were not merely a-specific. HIF-1α−/− embryos with looping defects were less retarded than the cardia bifida mutant embryos and died around E9.75–E10.

To analyze which gene could be involved in abnormal cardiac development, in situ analysis of a number of candidate genes was performed. Genes involved in the commitment of cardiac precursors (GATA-4, Nkx2.5) were normally expressed in HIF-1α−/− embryos at E9.25 (Fig. S2a,b)3. In addition, in situ hybridization for myosin (β-MHC) (not shown), atrial natriuretic factor (ANF) (Fig. 1e,f) and myosin light chain-2a (MLC2a) (Fig. S2c,d)3 and immunostaining for smooth muscle cell α-actin (Fig. 1a,b), desmin and myosin (not shown) revealed that cardiomyocytes differentiated normally in E9.25 HIF-1α−/− embryos, suggesting that cardiac lineage development and differentiation occurred in the absence of HIF-1α. At E9.25, expression of eHAND, which has been implicated in looping of the cardiac tube [11], was expressed in the future left ventricle of the heart in WT embryos, but was undetectable in HIF-1α−/− embryos at E9.25 (Fig. 1g,h). Expression of eHAND in the extraembryonic membranes was comparable in both genotypes (Fig. 1g,h). Heart tubes of embryos lacking myocyte enhancer factor 2C (MEF2C) fail to loop and the future right ventricle in these embryos fails to develop properly [12]. Although MEF2C was expressed in the heart of WT embryos, it was undetectable in HIF-1α−/− embryos at E9.25 by whole mount in situ hybridization and reduced by quantitative RT-PCR (Fig. 1i–l). Other genes, previously involved in cardiac looping, were also abnormally expressed. For instance, while expression of dHAND was confined to the future right ventricle in WT embryos, it remained widely expressed throughout the entire heart in HIF-1α−/− embryos at E9.25 (Fig. S2e–g)3 [13]. Thus, loss of HIF-1α deregulates expression of a number of genes essential for proper cardiac looping and ventricle formation.

3.3 Abnormal neural crest migration in HIF-1α−/− embryos

Consistent with previous reports [2,3,6], defective formation of pharyngeal arches was observed in all HIF-1α−/− embryos. Macroscopically, the head folds of E9.5 HIF-1α−/− embryos failed to close, often leading to enlarged and abnormally positioned neural folds. The second pharyngeal arches were small or absent, and the third pharyngeal arches were never present. Such pharyngeal arch abnormalities are often associated with neural crest defects. Transcript levels of HNF3 forkhead homologue 2 (hfh-2) (hfh-2 copies per 100 mRNA copies of β-actin: 1.5±0.2 in WT vs. 1.3±0.1 in HIF-1α−/− embryos, N = 6; P = NS), a marker of pre-migratory and early migratory NCC [14], were comparable in both genotypes, suggesting that NCC accumulated in both genotypes. In contrast, expression of Id2, a marker of migratory cranial NCC [15], was reduced in HIF-1α−/− embryos at E9.5 (Fig. 2f).

Fig. 2

Abnormal neural crest cell migration in HIF-1α−/− embryos. (a–e) Whole mount p75 immunostaining, displaying the migration pattern of NCC. NCC arising from the neuroepithelial junction (arrows in a, E8.5) migrated to the first and second pharyngeal arch (arrowhead in b, E8.75, somite-matched control for e). At E9.5 (c, age-matched for e), p75 expression was detectable in the third and fourth pharyngeal arch and in the rostral part of the somites. The pharyngeal arches are marked by arrowheads in c. At E8.5, expression of p75 was normal in HIF-1α−/− embryos (d). The neuroepithelial junction is marked by arrows in d. By E9.5, NCC failed to migrate in HIF-1α−/− embryos (e). Some expression of p75 is observed in the malformed first pharyngeal arch (arrowhead in e). (f) Expression levels of Id2 were severely decreased in HIF-1α−/− embryos compared to WT littermates. (g and h) In situ hybridization of endothelin-A receptor (ETA), revealing expression (*) in the pharyngeal arch region of E9.25 WT embryos but not in HIF-1α−/− embryos. The region of the pharyngeal arches are delineated by a dashed line, those of the head by a full line in g and h. H: head; O: otic vesicle. (i) Expression levels of endothelin-1 (ET-1) were significantly decreased in HIF-1α−/− embryos as compared to WT littermates. Data in (i) represent means ±S.E.M. of the number of mRNA copies per 100 mRNA copies of β-actin (N = 6, *:P<0.05). (j and k) Double immunostaining for TUNEL (green) and p75 (pink-red), revealing the absence of apoptopic cells in E8.75 embryos in both wild type (j) and HIF-1α−/− embryos (k). At this stage, migartion of NCC was retarded in HIF-1α−/− embryos (pink staining in k) compared to stage matched WT embryos (pink staining in j). (l and m) TUNEL staining in E9.25 wild type (l) and HIF-1α−/− embryos (m), demonstrating the presence of apoptotic cells at E9.25 in HIF-1α−/− embryos (arrowhead in m) but not in WT embryos (l). (n–p) Combined immunostaining for p75 (blue) and Semaphorin-3A (Sema3A; brown), demonstrating that Sema3A is slightly expressed in the head mesenchyme of WT embryos at E8.75 (n). Sema3A expression became prominent at E9.25 (o). In contrast, no expression of Sema3A was detectable in the head mesenchyme of E9.0 HIF-1α−/− embryos (p). The head mesenchyme is delineated by a dashed line in n and p. (q) Expression levels of Sema3A were lower in HIF-1α−/− than WT embryos. Data in (q) represent the mean ±S.E.M. of the number of mRNA copies per 100 mRNA copies of β-actin (N = 6, *:P<0.05). Scale bar is 100 μm.

Whole mount immunostaining for the low-affinity p75 NGF receptor, which marks undifferentiated NCC [16], revealed that NCCs were abnormally distributed in HIF-1α−/− embryos as early as E8.75. In WT embryos, NCCs were found at the neuro-epithelial junction in the forebrain at E8.5 (Fig. 2a), and subsequently migrated into the entire head (to become the future head mesenchyme) and into the first and second pharyngeal arch (to become the future pharyngeal arch mesenchyme) (Fig. 2b). At E9.5, p75 immunoreactivity became expressed more widespread in the brain, disappeared in the first pharyngeal arch (reflecting their differentiation) [16] and was detectable in the second and third pharyngeal arch and in the rostral half of each somite (Fig. 2c). In HIF-1α−/− embryos, a comparable p75-staining pattern was observed at E8.5 (Fig. 2d). Beyond E8.75 and at E9.5, p75-positive cells failed, however, to migrate and remained localized at the neuroepithelial junction (Fig. 2e). MEF2C, which has been previously documented in cephalic NCC at E9.25 [17], was present in the brain in WT embryos (Fig. 1i), but only detectable in localized regions in the hindbrain in E9.25 HIF-1α−/− embryos (Fig. 1k). The endothelin receptor-A (ETA), which marks ectomesenchyme after transformation of NCC [18], was detectable in the entire brain and in the pharyngeal arches at E9.25 in WT embryos, but not in E9.25 HIF-1α−/− embryos (Fig. 2g,h). Transcript levels of endothelin-1, which binds to ETA and is upregulated by HIF-1α, were also significantly lower in HIF-1α−/− than in WT embryos (Fig. 2i).

To analyze whether impaired NCC migration is caused by mesenchymal cell death, 9–10 somite-staged embryos were incubated in Nile blue sulfate (NBS). At this stage, defects in neural crest migration were present in HIF-1α−/− embryos (Fig. 2j,k). In none of the heads of the 9–10 somite-staged embryos (n = 3) was NBS staining detectable, indicating that neural crest migration defects occur prior to neural crest cell death. In addition, TUNEL staining was combined with P75 staining. Whereas P75 staining clearly documented the presence of defects in neural crest migration in 9–10 somite-staged HIF-1α−/− embryos, TUNEL staining on sections of these embryos confirmed the absence of apoptotic cells in the head of HIF-1α−/− embryos (E8.75; n = 3) (Fig. 2k). In concordance with previous findings, apoptotic cells were observed in E9.5 HIF-1α−/− embryos (Fig. 2m). Therefore, although the primary defect in neural crest migration is not due to cell death, we cannot exclude the possibility that, during further development, cell death impaired neural crest cell migration.

To further explore the mechanism causing neural crest cell defects, neuropilin-1 (NP-1) and semaphorin-3A (Sema3A) transcript levels were determined. NP-1, which is expressed by migrating NCC [19], is a receptor for Sema3A, known to repulse cultured NCC [19] and loss of NP-1 causes a neural crest ablation-like syndrome [20]. NP-1 transcript levels were comparable in both genotypes (NP-1 copies per 100 mRNA copies of β-actin: 21±2 in WT vs. 28±3 in HIF-1α−/− embryos, N = 6; P = NS), but Sema3A transcript levels in HIF-1α−/− embryos were only half of those found in WT littermates (Fig. 2q). Decreased expression of Sema3A was also found by immunohistochemistry: in WT embryos, Sema3A was slightly expressed in the head mesenchyme at E8.75 and became prominent at E9.25. In contrast, Sema3A remained undetectable in the head mesenchyme of E9.0 HIF-1α−/− embryos (Fig. 2n–p). Therefore, the decreased Sema3A expression potentially contributes to the abnormal localization of NCC in mutant embryos.

3.4 Impaired embryonic angiogenesis in HIF-1α−/− embryos

Whole-mount staining for platelet endothelial cell adhesion molecule type-1 (PECAM-1) revealed that loss of HIF-1α did not prevent the differentiation of angioblasts to endothelial cells and their subsequent assembly into a primary vascular labyrinth in the yolk sac and in the embryo proper at E8.5, indicating that vasculogenesis could proceed without HIF-1α. As others have shown [2,3], vascular defects became apparent in HIF-1α−/− embryos beyond E9.0, when the primitive vascular network in WT embryos normally expands via sprouting angiogenesis and remodels in a branching network of small and large vessels (Fig. S1a–b) 3. Consistent with previous findings, VEGF levels were not reduced in HIF-1α−/− embryos (VEGF copies per 100 mRNA copies of β-actin: 4.6±0.5 in WT vs. 6.5±0.9 in HIF-1α−/−, N = 6; P = NS), presumably because ischemia, resulting from the severe vascular defects, compensatorily upregulated VEGF via HIF-1α-independent mechanisms [6,21]. The expression of the VEGF receptors Flt-1 and Flk-1 (Flt-1 and Flk-1 copies per 100 mRNA copies of β-actin, respectively: 2.1±0.1 and 14±0.7 in WT vs. 1.4±0.2 and 9±1.7 in HIF-1α−/− embryos, N = 6; P<0.05), which are essential for the differentiation and assembly of endothelial channels [22,23], was significantly reduced in HIF-1α−/− embryos, suggesting that mutant embryos contained fewer or immature endothelial cells. Nevertheless, vasculogenesis occurs in HIF-1α−/− embryos. The remodeling of the immature capillary plexus into a mature vascular bed was, however, disturbed. Concordantly, transcript levels of the angiopoietins Ang-1 and Ang-2 and its receptor Tie2 were reduced in HIF-1α−/− embryos (Fig. S1c–e) 3. These molecules have been implicated in the remodeling of the expanding vasculature and their reduced expression might therefore have contributed to the abnormal patterning and remodeling of the vascular network in the mutant embryos. Furthermore, low levels of the endothelial-specific junctional VE-Cadherin (VE-cadherin copies per 100 mRNA copies of β-actin: 12±0.5 in WT vs. 8.2±1.1 in HIF-1α−/− embryos, N = 6; P<0.05), as well as reduced levels of MEF2C in HIF-1α−/− embryos (Fig. 1i–l), may contribute to the impaired remodeling of the yolk sac vasculature [24,25]. These changes in angiogenic gene expression were specific as expression of other angiogenic markers was either unchanged (TGF-β1 copies per 100 mRNA copies of β-actin: 1.7±0.2 in WT vs. 1.7±0.1 in HIF-1α−/− embryos, N = 6; P = NS) or slightly increased (PDGF-B copies per 100 mRNA copies of β-actin: 0.25±0.04 in WT vs. 0.38±0.02 in HIF-1α−/− embryos, N = 6; P<0.05). More details about the vascular development are given in S33 and Fig. S1f–i3.

3.5 Rescue of impaired survival of HIF-1α−/− embryos by hyperoxia

Early stage WT embryos develop in vitro from E8.5 to E9.5 when the culture medium contains 300 mg/dl glucose and is equilibrated with a gas mixture of 20% oxygen (referred to as “normoxia”), 5% CO2 and 75% N2. Since majority of E8.5 HIF-1α−/− embryos have no vascular or other developmental defects yet, we expected that they might survive in culture. However, none of the E8.5 HIF-1α−/− embryos (N = 4) was able to survive for more than 6–10 h in normoxic, normoglycemic culture conditions (Table 2). To examine whether the supply of oxygen or glucose was limiting, we cultured E8.5 HIF-1α−/− embryos in culture medium supplemented with increased glucose levels (“hyperglycemia”) or equilibrated with increased oxygen levels (“hyperoxia”). Hyperglycemia failed, however, to rescue the survival and development of HIF-1α−/− embryos (N = 6; P = NS vs. normoglycemia, Fisher's Exact test; Table 2), possibly because transcript levels of hexokinase-1, a rate-limiting enzyme in anaerobic glycolysis, and lactate dehydrogenase were reduced in HIF-1α−/− embryos (Fig. 3g). In contrast, “hyperoxia” prolonged the survival of HIF-1α−/− embryos (n = 11) to E9.25 (25% of embryos) or for the entire culture period until E9.5 (75% of embryos) (P<0.05 vs. normoxia Fisher's Exact test; Table 2). HIF-1α−/− embryos cultured in “hyperoxia” contained as many somites (17±1, n = 17) as WT embryos cultured in “hyperoxia” (18±1 somites; n = 23, P = NS). Culture in hyperoxic conditions at least partially rescued the development of the embryonic heart, evidenced by the absence of embryos with cardia bifida in HIF-1α−/− embryos (Fig. 3a). In general, cardiac development remained delayed in hyperoxic HIF-1α−/− embryos (Fig. 3a). Nevertheless, some embryonic HIF-1α−/− hearts progressed into the embryonic chambered heart stage in hyperoxic conditions (Fig. 3a), indicating that culture in hyperoxia is capable of rescuing chamber formation. In concordance with these findings, the expression of eHAND, dHAND and MEF2C switched on in 66% of the HIF-1α−/− embryos (Fig. 3b–f). Culture in hyperoxic conditions did not rescue the remodeling of the yolk sac vessels in HIF-1α−/− embryos (Fig. 3h). Also, processes involving NCC remained affected: wild-type embryos cultured in hyperoxia had two or three pharyngeal arches, whereas majority of the hyperoxic HIF-1α−/− embryos developed only a single pharyngeal arch (Fig. 3i). Histological analysis indicated that the outflow tract was incorrectly formed in the heart of hyperoxic HIF-1α−/− embryos (Fig. 3j,k). In addition, remodeling of cephalic vessels, which depends on structural support by surrounding pericytes and mesenchyme, both derived from neural crest [26], was also not rescued in hyperoxic HIF-1α−/− embryos (Fig. 3j,k). Expression of Sema3A remained absent in HIF-1α−/− embryos cultured in hyperoxic conditions (Fig. 3l,m). Overall, these findings suggest that neural crest cell development was not completely rescued by culture in hyperoxic conditions.

Fig. 3

Hyperoxia partially rescued cardiac development but not neural crest development and vascular remodeling in HIF-1α−/− embryos. (a–c) Bar diagram displaying the percent of WT and HIF-1α−/− embryos with the indicated cardiac phenotypes, when cultured in hyperoxic conditions. Cardia bifida or horseshoe hearts were not detected in HIF-1α−/− embryos cultured in hyperoxia indicating that hyperoxia rescued the development of the heart tube. Although chamber formation was rescued in a subgroup of HIF-1α−/− embryos, in general, cardiac development in HIF-1α−/− embryos cultured in hyperoxia remained retarded. (b and c) After culture in hyperoxic conditions, the expression of MEF2C (in situ hybridization) was detectable in the heart of WT and HIF-1α−/− embryos (arrows in b and c). (d–f) eHAND was expressed in the heart of all WT embryos cultured in hyperoxia (arrow in d). Culture in hyperoxic conditions rescued eHAND expression in some (arrow in e) but not all HIF-1α−/− embryos (arrow in f). (g) Expression levels of hexokinase-1 were lower in HIF-1α−/− than in WT embryos . Data represent means ±S.E.M. of the number of mRNA copies per 100 mRNA copies of β-actin (n = 6, *;P<0.05). (h) Bar diagram displaying the percent of WT and HIF-1α−/− embryos exhibiting normal vascular remodeling of the yolk sac after culture in hyperoxia. Vitelline vascular remodeling failed to occur in HIF-1α−/− embryos cultured in hyperoxic conditions. (i) Bar diagram displaying the percent of WT and HIF-1α−/− embryos with the indicated pharyngeal arches (PA) phenotypes after culture in hyperoxia. 1st PA: first PA; 2nd PA: second PA; 3rd PA: third PA. (j and k) H&E, E9.5. WT embryos developed normally after 24 h in culture (j) (AA: aortic arch, *:cardinal vein). After culture in hyperoxia, cardiac looping improved in some embryos and the caudal dorsal aorta developed normally (1 in k) in most HIF-1α−/− embryos. However, rescue of organogenesis was incomplete since the outflow tract is interrupted (arrow in k) and the cardinal vein remains dilated (* in k). (l and m) Immunostaining for Sema3a reveals that Sema3A is present in the mesenchyme of WT embryos (arrowheads in l), but is undetectable in HIF-1α−/− embryos (arrowheads in m). Scale bar is 100 μm.

View this table:
Table 2

Embryonic viability after 24 h of culture in normoxia, normoglycemia, hyperoxia or hyperglycemia

Culture conditionsNumber of surviving embryos after 24 h
Normoxia normoglycemia5/50/4
Hyperoxia normoglycemia12/128/11
Normoxia hyperglycemia2/20/6
  • Embryos were cultured in indicated conditions. After 24 h, the viability of the embryos was examined. Embryos with a beating heart and visible circulation were considered alive.

4. Discussion

4.1. Role of HIF-1α in cardiogenesis and NCC

Congenital heart disease is the leading non-infectious cause of death in children, but the underlying mechanisms remain often undetermined. HIF-1α might contribute to cardiac development by (at least) two mechanisms: firstly, by directly affecting the myocardium and, secondly, by affecting NCC (see below). HIF-1α is expressed in the early heart [27] and in cultured embryonic ventricular myocytes. Yet, its role in cardiomyocyte commitment and differentiation seems redundant since several genes, known to be involved in commitment and differentiation of cardiomyocytes, including Nkx2.5 [28], GATA-4 [29], ANF [30], β-MHC [31], MLC2a [8] and desmin [31], were normally expressed in mutant embryos. Nevertheless, formation of the horseshoe-shaped cardiac crescent failed to occur properly in 30% of HIF-1α−/− embryos, resulting in cardia bifida. The occurrence of this type of cardiac defect is the earliest discernable effect of this transcription factor in embryogenesis, e.g. before it affects vascular development. Remarkably, the developmental program of these precardial cells to assemble into a tube of myocardial cells, surrounding an endocardial layer, progressed, indicating that HIF-1α did not specifically block cardiogenesis. The molecular and cellular mechanisms of cardia bifida remain poorly understood. Cardia bifida has been documented in embryos lacking GATA-4 [32], MesP1 [33] and in anti-fibronectin antibody-treated embryos [34]. The expression pattern of GATA-4 was, however, normal in HIF-1α−/− embryos and therefore the precise mechanism of action of HIF-1α in the formation of a horseshoe-shaped cardiac crescent remains to be determined. In the remaining HIF-1α−/− embryos, in which a single heart tube formed, the primitive hearts failed to loop at all or, if it looped, looping occurred abnormally. Impaired looping did not appear to result from specific growth retardation, since cardiac looping of HIF-1α−/− embryos was retarded even when compared to somite-matched controls (Table 1). Several genes, involved in cardiac looping and proper ventricle development, were abnormally expressed in mutant embryos. For instance, eHAND normally becomes restricted to the future left ventricle, but was undetectable in HIF-1α−/− embryos. Conversely, expression of dHAND becomes confined to the future right ventricle in WT embryos, but remained widely expressed throughout the entire heart in HIF-1α−/− embryos. MEF2C levels were also reduced in mutant embryos. Loss of MEF2C prevents cardiac looping and formation of the future right ventricle [12]. Furthermore, retarded trabeculation in HIF-1α−/− embryos might relate to decreased expression of angiopoietin-1 [10] and its receptor TIE2. None of the affected genes has thus far been documented to be regulated by HIF-1α, but this may be due to insufficient characterization. In addition, regulation of the expression of (some of) these genes may be indirect, e.g. via intermediate signals, which themselves are direct targets of HIF-1α. Nevertheless, since genes involved in differentiation of cardiomyocytes were normally expressed in HIF-1α−/− embryos, HIF-1α specifically affects cardiac looping and formation of future ventricles. We can also, obviously, not exclude the possibility that other genes, whose expression is regulated directly or indirectly by HIF-1α, are involved.

NCCs contribute to the development of the outflow tract and cardiac septation and are known to be sensitive to hypoxia [35]. HIF-1α is expressed in the neuroepithelium (from which NCCs derive) and in the pharyngeal arches (containing neural crest-derived cells) [27]. Some degree of hypoxia might be beneficial for the development of NCC, since culturing these cells in hypoxia improves their survival and proliferation in vitro [36]. The similar transcript levels for hfh2 in WT and HIF-1α−/− embryos suggest that NCCs proliferate and accumulate in HIF-1α−/− embryos. In contrast, migration of NCC is impaired. Indeed, the p75 receptor of NGF was abnormally expressed and transcript levels of Id2, normally expressed in migrating cranial neural crest and involved in the transformation from neuroepithelial to NCC [15], were reduced in HIF-1α−/− embryos. Impaired migration of NCC might be responsible for the underdevelopment of the first pharyngeal arch, complete absence of the second to fourth pharyngeal arches and abnormal remodeling of the aortic arches and cephalic vessels. NCC migration defects are not secondary to mesenchymal cell death since migration defects appear prior to cell death. Previous studies have suggested an important role for neuropilin-1 (NP-1) and its ligand semaphorin-3A (Sema3A) in neural crest migration [19]. While the loss of NP-1, expressed by migrating NCC, causes a neural crest ablation-like syndrome [20], Sema3A is known to repulse NCC in vitro and delineates territories non-permissive for NCC in vivo [19]. Immunohistochemistry and RT-PCR revealed that Sema3A expression was decreased in mutant embryos, suggesting that it might therefore contribute to the abnormal migration of NCC in mutant embryos. The ET-1/ETA-mediated signaling also plays an essential role in aortic arch patterning by affecting the differentiation of postmigratory cardiac NCC [37]. Since the expression levels of endothelin-A (ET-A) receptor and its ligand endothelin-1, a known HIF-1 target, were also decreased in HIF-1α−/− embryos, postmigratory in situ differentiation of NCC might also have been affected by loss of HIF-1α. However, we cannot exclude the possibility that the decreased expression levels of these targets were secondary to the impaired NCC migration.

4.2. Role of HIF-1α in angiogenesis

Hypoxia is a significant stimulus for the formation of new blood vessels. HIF-1α is not essential for vasculogenesis, since angioblasts differentiated to endothelial cells and assembled in a primitive vascular bed in HIF-1α−/− embryos. However, as others have shown [2,3], HIF-1α is required for the subsequent remodeling of the vascular bed into a complex network of small and large vessels. Apparently, HIF-1α modulates, directly or indirectly, an entire program of vascular gene expression, including Ang-1, Ang-2, TIE2 and MEF2C, which have been implicated in maturation and stabilization of the primitive vascular network by affecting periendothelial cells [25] as well as VE-Cadherin. In contrast, the expression of other angiogenic markers was either unchanged (TGF-β1) or slightly increased (PDGF-B). Further analysis should point out whether these genes are HIF-1α targets or whether their reduced expression is caused indirectly. Whatever the mechanisms, the reduced expression levels of these angiogenic factors in HIF-1α−/− embryos likely contributed to the abnormal vascular remodeling.

4.3. HIF-1α is essential to induce proper differentiation pathways

HIF-1α regulates the expression of several key enzymes in anaerobic glycolysis. We demonstrate that hexokinase-1, a rate-limiting enzyme in anaerobic glycolysis, is severely reduced in HIF-1α−/− embryos. This could possibly explain why hyperglycemia failed to rescue the survival and development of HIF-1α−/− embryos. Hyperoxia, on the other hand, prolonged survival and abolished growth retardation of HIF-1α−/− embryos. Increased atmospheric oxygen levels are known to overcome the deteriorating effects of insufficient glycolysis, the principal energy pathway at this early stage of embryonic development, by promoting oxidative metabolism and are capable of compensating insufficient oxygen delivery due to placental or vascular defects [38]. Therefore, culture in hyperoxia allows to distinguish between phenotypes mediated by metabolic factors and phenotypes directly dependent on HIF-1α. Morphological analysis and in situ hybridization of MEF2C, eHAND and dHAND demonstrated that hyperoxia partially rescues cardiogenesis, indicating that hypoxia or energy depletion contributes to the phenotype. Since metabolic deficiencies are well known to affect cardiac development [39] and hexokinase-1 is highly expressed in the embryonic heart [40], HIF-1α may regulate cardiogenesis through regulation of hexokinase-1. Interestingly, in contrast to its beneficial effect on general growth, survival and cardiac development, hyperoxia failed to rescue yolk sac vessel remodeling or pharyngeal arch development, a process involving NCC. These data suggest that the defects in these systems are not primarily due to hypoxia or energy depletion, but rather due to the absence of differentiation pathways induced by HIF-1α.


The authors thank D. Anderson for the p75 antibody (California Institute of Technology), E. Dejana for the PECAM antibody (Mario Negri Institute), Corrie de Gier-de Vries (AMC, Amsterdam) and A. Bouché, B. Hermans, S. Jansen, L. Kieckens, A. Manderveld, K. Maris, S. Terclavers, A. Vandenhoeck and P. Van Wesemael (CTG, Belgium) for assistance. This work was supported in part by the European Union (Biomed BMH4-CT98-3380), the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen (FWO # G.0125.00), KULeuven Geconcerteerde Onderzoeksacties (GOA # 2001/09) and the Bristol-Meyers Squibb Research Foundation. V.C. and K.B. were supported by fellowships from FWO and from the Flemish Institute for Promotion of Scientific Research (IWT), respectively.



  1. [1]
  2. [2]
  3. [3]
  4. [4]
  5. [5]
  6. [6]
  7. [7]
  8. [8]
  9. [9]
  10. [10]
  11. [11]
  12. [12]
  13. [13]
  14. [14]
  15. [15]
  16. [16]
  17. [17]
  18. [18]
  19. [19]
  20. [20]
  21. [21]
  22. [22]
  23. [23]
  24. [24]
  25. [25]
  26. [26]
  27. [27]
  28. [28]
  29. [29]
  30. [30]
  31. [31]
  32. [32]
  33. [33]
  34. [34]
  35. [35]
  36. [36]
  37. [37]
  38. [38]
  39. [39]
  40. [40]
View Abstract