Hypoxia Signaling in Vascular Homeostasis

Abstract

Hypoxia signaling in the vasculature controls vascular permeability, inflammation, vascular growth, and repair of vascular injury. In this review, we summarize recent insights in this burgeoning field and highlight the importance of studying the heterogeneity of hypoxia responses among individual patients, distinct vascular beds, and even individual vascular cells.

Introduction

Vascular homeostasis during adult life is the result of balancing vascular injury and damage with repair and regeneration while also integrating environmental cues to optimize vascular function and blood vessel growth, thus ensuring adequate supply of nutrients and oxygen to tissues. Vascular regeneration may occur either through the de novo formation of blood vessels via endothelial progenitor cells (vasculogenesis), by the expansion of the existing vascular network (angiogenesis), or by the combination of both (19). Vascular homeostasis refers to the maintenance of vascular function over time and thus also includes the adaptation to persistent environmental signals such as chronic hypoxia. It is noteworthy that the vasculature also has the ability to rapidly respond to acute signals to restore homeostatic function, such as what is observed during acute hypoxia. Recent studies have identified several adaptive vascular responses to chronic hypoxia, including changes in vascular permeability, arterial vs. venous specification of the developing or growing vasculature, and the regulation of inflammatory processes. These recent findings not only provide important insights into basic mechanisms by which environmental oxygen levels regulate vascular function but also suggest potential novel therapeutic approaches that target hypoxia-regulated pathways in the vasculature to minimize vascular injury or enhance vascular repair and regeneration. With the emergence of precision medicine and precision science—popular catchphrases that signify the growing specificity as well as greater appreciation of biological complexity and heterogeneity—it is becoming apparent that the heterogeneity of cellular responses to hypoxia may be a critical factor in determining how tissues respond to low oxygen levels. This heterogeneity has been observed at an inter-individual level in which distinct patients may exhibit genetic differences in the key signaling pathways mediating hypoxia responses. In addition to inter-individual heterogeneity, hypoxic responses of the vasculature also differ widely between distinct vascular cell types and tissue beds, and even between individual cells in one vascular bed (FIGURE 1). Finally, a word of caution is also required since systemic manipulation of hypoxic signaling pathways might have untoward side effects due to the central role of hypoxic signaling in homeostasis. Understanding the complexity and heterogeneity of tissue responses to hypoxia is a prerequisite for designing targeted therapies that minimize unintended collateral effects.

FIGURE 1.

Heterogeneity of hypoxia signaling

Under normal oxygen levels, HIF-1α and HIF-2α are hydroxylated on conserved prolyl residues by a set of oxygen-dependent prolyl hydroxylases (PHD1–3). This results in the recognition of HIF-1α and HIF-2α by the von Hippel-Lindau protein, leading to ubiquitination and subsequent proteasomal degradation. In response to lower oxygen levels, PHD enzymes will be inhibited, allowing for the accumulation of HIF-1α and HIF-2α. They subsequently form a complex with constitutively expressed HIF-1β and induce the expression of target genes. These genes are involved both in systemic responses such as increased angiogenesis and in changes of cellular energy metabolism. There is considerable heterogeneity in the hypoxic response between individuals, between different vascular beds, and even between individual endothelial cells. Understanding this complexity and heterogeneity is a prerequisite for designing targeted therapies that minimize unintended collateral effects.

Hypoxia Signaling Pathways

Cells respond to low oxygen levels in several ways, and probably the best-studied system to sense hypoxia consists of the hypoxia-inducible factors (HIFs), which are transcription factors consisting of the HIF-1α or HIF-2α subunits that dimerize with the HIF-1β subunit to initiate the transcription of hypoxia-responsive genes. Oxygen-dependent prolyl hydroxylases (PHDs) hydroxylate HIF-1α and HIF-2α when oxygen is present at high levels, leading to ubiquitination by the von Hippel-Lindau (VHL) protein and subsequent proteasomal degradation. Low oxygen levels inhibit PHDs, allowing for the accumulation of HIF-1α and HIF-2α (FIGURE 1), and the expression of specific genes involved in blood vessel growth, anaerobic metabolism, and enhanced delivery of oxygen to tissues (13, 52, 66). In addition, factor inhibiting HIF (FIH) is an asparagine hydroxylase that negatively regulates the transcriptional activity of HIF and can also serve as a regulatory mechanism for modulating hypoxia responses (84).

It is noteworthy that some animals such as sponges and comb jellies maintain their gene expression levels despite exposure to extremely low oxygen levels, probably because they lack a VHL ortholog and their HIF orthologs lack hydroxylation domains (49, 64). Therefore, HIF-mediated oxygen-sensing pathways were not present when single-celled ancestors started to evolve into the first animals ~800 million years ago and appear to have evolved since then.

In addition to HIF-mediated shifts in gene expression, cells can also respond more rapidly to changes in oxygen levels by alternate mechanisms. For example, hypoxic pulmonary vasoconstriction is mediated by the mitochondria (17) and complex III of the electron transport chain has been identified as an important oxygen sensor (76), indicating that hypoxia can also be sensed independently of the PHD/HIF pathway and can lead to rapid hypoxic responses that do not require de novo gene expression. Similarly, hypoxia can modulate reactive oxygen species (ROS) in vascular cells (24, 77) and thereby modulate vascular barrier integrity (51, 80) and the angiogenic response (46). Therefore, vascular homeostasis depends on both HIF-dependent and -independent mechanisms, which can act in concert to regulate vascular function and adaptation to environmental oxygen levels.

Role of Hypoxia Signaling on Blood Vessel Permeability and Stability

During acute lung injury (ALI)/acute respiratory distress syndrome (ARDS), the endothelial barrier in the lung is disrupted, which in turn leads to edema formation and the infiltration of inflammatory cells. Lung edema impairs the gas-exchange function of the lung, thus resulting in local and systemic hypoxia. The high levels of oxygen consumption by inflammatory neutrophils can also induce local tissue hypoxia even in the absence of environmental hypoxia, as has been reported for mucosal inflammation (8). It is thus likely that the combination of edema-impaired gas exchange and inflammation-induced oxygen consumption results in profound hypoxia of the lung vasculature during ALI/ARDS. This prompted researchers to study whether hypoxic signaling in the vascular endothelium may regulate endothelial barrier permeability as a possible adaptive mechanism to contain tissue injury during ALI/ARDS. Mice with HIF-2α-deficient ECs develop normally but exhibit increased vessel permeability and focal lymphocyte infiltration in the lungs and adipose tissue, whereas vascular permeability in the liver and kidney were not altered (67). Endothelial cells lacking HIF-2α showed reduced adhesion in hypoxic conditions and expressed lower levels of the NOTCH ligand Dll4 (67). Reduced Dll4 levels can increase vessel branching, but the resulting vessels have a smaller diameter, are often not perfused, and fail to recruit pericytes (65). Similar findings were made in tumors where Dll4 inhibition also leads to increased vascularization but poor perfusion of the blood vessels (53). Therefore, Dll4 downregulation appears to destabilize blood vessels and can ultimately lead to increased tissue hypoxia.

At a molecular level, vascular endothelial cadherin (VE-cadherin) homodimer formation leads to the formation of endothelial adherens junctions (AJs), which are critical regulators of endothelial barrier integrity. Vascular endothelial protein tyrosine phosphatase (VE-PTP) is an important stabilizer of AJs, which prevents VE-cadherin phosphorylation and internalization (21). It was recently shown that endothelial-specific deletion of HIF-2α increased lung vascular permeability in ALI/ARDS and that, conversely, activation of HIF-2α by inhibiting PHD2 suppresses lung vascular leakiness in ALI/ARDS, thus indicating an important adaptive barrier-stabilizing role of hypoxic signaling in the lung vasculature (21). In contrast, HIF activation has been found to lead to a disruption of the blood-brain barrier in response to inhaled anesthetics such as isoflurane by downregulating the tight-junction protein occludin as well as collagen type IV in brain blood vessels (9). In line with these findings, deletion of HIF-1α in endothelial cells can prevent blood-brain barrier leakiness after transient middle cerebral artery occlusion of diabetic mice (86). These findings demonstrate that each vascular bed may respond to hypoxia differently and that HIF-1α and HIF-2α can act in a distinct manner in each vascular bed.

Arterial vs. Venous Specialization of Endothelial Cells

Early embryonic development occurs in a relatively hypoxic environment, which stimulates the formation of endothelial cells (37, 59). For example, a reduced oxygen environment (5% O₂) during the first 6 days of human-induced pluripotent stem cell differentiation into endothelial cells leads to increased numbers of endothelial cells that are more functional in forming a vascular network (37).

Recent work has shown that oxygen-derived free radicals stimulate the differentiation of pluripotent stem cells by inducing the degradation of the master pluripotency regulator OCT4 (45). Pluripotent stem cells maintain a reduced intracellular environment, probably as a protection against reactive oxygen species. Metabolic shifts that enhance the cellular oxidation state significantly enhance the loss of pluripotency and facilitate vascular differentiation of pluripotent embryonic stem cells (45).

Hypoxia also guides endothelial cells toward an arterial cell fate, as evidenced by the upregulation of arterial markers, likely indicating an adaptive response to enhance arterial blood flow and oxygenation (72). Subsequent work in mouse embryonic stem cells has shown that increased formation of endothelial cells due to hypoxia is the result of a HIF-1α-dependent downregulation of the pluripotency markers Oct4 and Nanog, and upregulation of VEGF (41). The initial phase of endothelial differentiation is of critical importance, and genetic deletion of HIF-1α prevented the upregulation of the Etv2 transcription factor, which regulates endothelial differentiation (72).

During embryonic development, the arterial or venous identity of endothelial cells is established very early, even before blood is circulated through the vasculature, as evidenced by the expression of Ephrin B2 in arteries and its receptor EphB4 in veins (74). One of the earliest signals that differentiates between both fates is the expression of Sonic Hedgehog (Shh), which is required for arterial differentiation (39). Downstream of Shh, VEGF-mediated induction of Notch signaling was shown to be required for the formation of arterial endothelial cells (39), and loss of Notch signaling leads to ectopic expression of venous markers in the dorsal aorta of zebrafish (38). Continuous hypoxia induced the differentiation of mouse embryonic stem cells into arterial endothelial cells in hypoxia, an effect that depended on HIF-1α-mediated upregulation of Notch1 (72).

In a mouse brain endothelial cell line, hypoxia induced the arterial-specific Notch ligand Dll4, resulting in the upregulation of several Notch target genes. These target genes in turn suppress the venous regulator Coup-TFII (16). Therefore, hypoxia-induced Notch activation can lead to a switch between a venous and arterial phenotype, but it remains to be seen whether hypoxic enhancement of vascular barrier function (21) is also in part mediated by a phenotype switch of endothelial cells.

Role of Hypoxia Signaling in Inflammation

The inflammatory environment is often characterized by low oxygen levels due to a high metabolic demand of the inflammatory cells (8), and, in more severe cases, thrombosis or vascular damage can further impair oxygen delivery to the inflammatory environment. Interestingly, several observations point to an induction of HIF in inflammatory cells even in a normoxic environment. Stimulation of macrophages with both Gram-positive and Gram-negative bacteria leads to a robust induction of HIF-1α (58), which depends on the activation of NF-κB (62). Furthermore, macrophages lacking the NF-κB activator IκB kinase beta (IKKβ) demonstrate blunted HIF-1α accumulation in response to hypoxia (62). Several cytokines, including TNF-α (88) and IL-1β (32), similarly upregulate HIF-1α, and it is noteworthy that this response is to a large extent dependent on increased HIF-1α mRNA transcription and not due to a suppression of PHD-mediated HIF-1α degradation (32, 88). Conversely, macrophages lacking HIF-1α have impaired bacterial killing, and mice lacking HIF-1α in myeloid cells were not able to contain a skin infection of group A Streptococcus (58). Pharmacological activation of HIF-1α with AKB-4924 in contrast protected mice from Staphylococcus aureus infection (55). These findings indicate that signaling via HIF-1α is critical for intact immune function.

Much of the work on the role of hypoxic signaling in inflammation has focused on leukocytes, and because leukocytes are required to migrate through the endothelial layer to reach inflamed tissues, targeted modulation of hypoxic signaling in endothelial cells could emerge as a potential novel approach to regulating acute or chronic inflammation. Interestingly, inflammatory activation and tissue hypoxia may also create an environmental niche that not only recruits additional leukocytes but also recruits endothelial progenitor cells, which may participate in blood vessel repair and growth. Hypoxia-mediated upregulation of intercellular adhesion molecule-1 (ICAM-1) on endothelial cells, for example, allows endothelial progenitor cells to home to ischemic areas (83), which is dependent on their expression of β2 integrins (10).

Radiation Sensitivity of Endothelial Cells

Radiation therapy dosage in cancer patients is limited by the risk for normal tissue injury. For example, intestinal injury is often seen after irradiation of patients with abdominal and pelvic tumors. Several studies have shown that irradiation-induced endothelial apoptosis and capillary loss adversely affects the repair capacity of the epithelium (75). Moreover, endothelial deletion of HIF-1α protects against injury induced by localized intestinal irradiation by improving vascular integrity, which reduced both intestinal hypoxia and macrophage infiltration (71). Reduction of HIF-1α signaling led to a less inflammatory profile of endothelial cells, with lower levels of P-selectin, IL-1β, and VCAM1 expression after irradiation (71). Deletion of HIF-1α in gut epithelial cells, on the other hand, had no impact on disease severity.

When all three PHD enzymes were removed in gut epithelial cells, mice showed a remarkably improved survival after abdominal irradiation, an effect that was almost entirely dependent on the upregulation of HIF-2α (70). HIF signaling was further shown to improve epithelial integrity of the gastrointestinal tract and prevent hypernatremia and hyperglycemia. This study also showed that treating mice with DMOG to activate HIF signaling did not enhance xenograft growth of colorectal cancer cells or inhibit the tumoricidal effect of irradiation (70). These studies underscore the complexity of modulating HIF signaling in vivo: on the one hand, enhanced HIF-2α signaling might improve epithelial integrity, but, on the other hand, reducing HIF-1α signaling would improve endothelial function. Therefore, it might be necessary to develop therapeutic approaches that can be targeted to specific cell types and specifically modulate HIF-1α or HIF-2α.

Microvascular Repair After Organ Transplantation

Upon organ transplantation, the blood vessels of the transplanted organ are exposed to the host immune system, and alloimmune rejection negatively impacts graft survival. Microvascular dropout is an important factor in chronic rejection for transplantation of several solid organs, and HIF activation can mitigate this response. For example, after lung transplant, local ischemia and loss of microvasculature is linked to small airway fibrosis or bronchiolitis obliterans, a major complication that determines long-term survival of transplanted lungs (23). By using an orthotopic tracheal transplantation model in mice, researchers found that donor endothelial cells increased HIF-1α levels and grafts with adenoviral-mediated enhanced HIF-1α levels had better microvascular perfusion and less fibrotic remodeling, whereas the opposite was true if HIF-1α was deleted (30).

Mechanistically, HIF-1α increased the expression of proangiogenic factors and enhanced the recruitment of recipient Tie2⁺ angiogenic cells to the allograft to replace donor endothelial cells (30). In addition, VHL-haplodeficient pulmonary endothelial cells are more angiogenic and resistant to serum withdrawal-induced cell death (29), whereas VHL-haplodeficient mice were better able to contain a pulmonary infection of Aspergillus fumigatus, possibly because of the improved microvascular integrity (29). Topical application of the HIF activator desferrioxamine (DFO) on airway transplants at the time of transplantation promoted microvascular repair, suggesting that HIF activation might be a clinically relevant therapy (31). Similarly, treating a donor rat with a PHD inhibitor (FG-4497) improves the long-term survival of allogeneic kidney transplants (7).

Genetic Instability in Response to Hypoxia

Hypoxia has been studied extensively as a driver of genetic instability of tumors (43), and several genes involved in DNA damage repair were found to be regulated by hypoxia. However, similar observations about hypoxia-mediated downregulated DNA repair enzymes have been made in non-cancer cells, suggesting that endothelial cells exposed to hypoxia might be more prone to accumulate mutations.

Normal human fibroblasts maintained in hypoxia have an impaired ability to repair DNA double-strand breaks in response to irradiation, and this results in increased chromosomal instability (36). Hypoxia was subsequently shown to downregulate several proteins involved in homology directed repair, including Rad51, BRCA1, and BRCA2 (47).

In a mouse dermal fibroblast cell line, exposure to 48 h of hypoxia or desferrioxamine led to a substantial downregulation in MLH1 and PMS2 proteins involved in DNA mismatch repair (48). This resulted in a twofold increase in mutation frequency and increased replication slippage at DNA repeat regions. Hypoxic downregulation of the MLH1 protein depended on histone deacetylation and could largely be prevented by the histone deacetylase inhibitor TSA (48). A similar HIF-1α-dependent inhibition of mismatch repair enzymes was also found in colon cancer cells (35).

Endothelial cells from renal cell carcinomas were shown to exhibit a high proportion of aneuploidy (2), and further work from the same group showed that, when tumor-associated endothelial cells from high-metastatic and low-metastatic melanoma xenografts were compared, endothelial cells isolated from high-metastatic tumors proliferated faster and had a more invasive phenotype (54). Interestingly, freshly isolated endothelial cells from high-metastatic tumors also had a strikingly higher aneuploidy rate and even double-minute chromosomes (54). Intravital imaging of oxygen levels using pimonidazole further showed that blood vessels in high-metastatic tumors were exposed to hypoxia, suggesting that the tumor oxygen environment can impact the genetic stability of tumor endothelial cells (54), and these genetic changes in endothelial cells might in turn stimulate metastasis (44).

Certain types of tumor cells appear to have the ability to differentiate into endothelial-like cells as well. For example, endothelial cells present in B-cell lymphomas contained similar chromosomal translocations as the lymphoma cells, suggesting that tumor endothelial cells may also be derived from malignant cells (69). The precise origin and phenotype of the tumor endothelium can vary among tumors, and some studies suggest that tumors can form their own microvascular channels that do not require the recruitment of preexisting tissue endothelial cells, a process that is referred to as vasculogenic mimicry. Vasculogenic mimicry has been observed in several types of tumors (18) and correlates with a poor survival (15). During vasculogenic mimicry, tumor cells are even able to acquire typical endothelial markers such as VE-cadherin, but it is not clear that they have necessarily fully transdifferentiated into mature endothelial cells and likely represent a hybrid cell type (15).

It should also be noted that a clonal expansion of endothelial cells has been observed in plexiform lesions of patients with pulmonary artertial hypertension (PAH) (40), and, since hypoxia is a widely used model to study PAH, it would be interesting to see whether hypoxia-induced genomic instability might be involved in the formation of hyperproliferative endothelial cells in patients with PAH.

Genetics of Hypoxia Responses and Implications for Personalized Medicine

Modulating HIF signaling might be an attractive target, yet it is important to appreciate that there is considerable heterogeneity in the individual response to hypoxia, and this needs to be taken into account when customizing care to patients (FIGURE 1). Several mutations in the HIF signaling pathway have a clear phenotype, including the mutation in VHL that leads to Chuvash disease (3, 4). In addition, recent genome-wide association studies (GWAS) have identified several genetic variants in the vicinity of genes involved in the cellular and physiological response to hypoxia. Further work is necessary to clarify the importance of these SNPs and how the presence of these SNPs affects hypoxia signaling in individuals.

HIF-1α and HIF-2α

Mutations in HIF-2α are associated with erythrocytosis and cancer development. For example, a heterozygous G1609->T mutation in exon 12 of HIF-2α was identified in a family with erythrocytosis, and the resulting amino acid change led to a weaker binding of PHD2 to Pro531 and reduced degradation of HIF-2α (57). Similar gain-of-function mutations in HIF-2α have also been found in two patients with multiple paragangliomas (catecholamine-producing tumors from the chromaffin cells of the extra-adrenal paraganglia) with or without somatostatinoma (somatostatin-producing tumors derived from neural crest cells) (89). Both patients were found to have mutations in exon 12, leading to a disruption of prolyl hydroxylation of HIF-2α and increased half-life of HIF-2α (89).

More subtle variants within or in the proximity of HIF-2α have been identified by using GWAS. For example, Tibetans have adapted to living at high altitude by having lower hemoglobin levels preventing blood hyperviscosity. Several studies have implicated HIF-2α in this adaptation (56, 78, 82). In the biggest of these studies, a GWAS on >1,000 native Tibetans identified several variants in HIF-2α that were much less frequent in people with a Han Chinese or Japanese ancestry (56). Variants in the promoter region of HIF-2α have similarly been associated with high-altitude adaptation (79).

A multi-ancestry GWAS study of >150,000 individuals also identified that a SNP upstream of HIF-2α is associated with birth weight (26), whereas an intronic SNP in HIF-2α is associated with atrial P-wave duration in a GWAS study comprising 44,000 individuals (14). Craniofacial microsomia is a congenital condition for which the pathogenesis is unknown; however, a case/control GWAS identified a SNP upstream of HIF-2α (85).

The association of this large number of seemingly unrelated conditions with variants in the vicinity or intronic regions of HIF-2α might be surprising but probably reflects variants that influence tissue-specific enhancers. It also points at how modulating HIF-2α can affect multiple tissues and the potential variability in how individuals will respond.

Overall, few GWAS studies have identified SNPs in the vicinity of HIF-1α, perhaps pointing to the importance of tightly controlling HIF-1α expression levels. In one exception, a GWAS in a subpopulation of the Framingham heart study found that a SNP in an intronic region of HIF-1α is correlated with CD40 ligand serum levels (6). Serum CD40 ligand levels are biomarkers for increased cardiovascular risk (81), although changes in HIF-1α levels themselves have not been associated with increased cardiovascular risk.

PHDs and VHL

An intronic SNP in the PHD1 gene is associated with smoking behavior (number of cigarettes/day) (20), and a SNP upstream of PHD1 is associated with the development of chronic obstructive pulmonary disease (COPD) (12). However, it should be pointed out that the major nicotine-inactivating enzyme Cyp2A6 lies just 30 kb downstream of PHD1 and that these SNPs might simply be in linkage disequilibrium. Further research needs to be done to determine the exact role of PHD1 in COPD.

A GWAS looking at hematological traits in >170,000 individuals of European ancestry found a strong association between a 5′ UTR variant of the PHD2 gene and hematocrit, hemoglobin levels, and red blood cell count. A separate intronic SNP was similarly associated with hematocrit levels and red blood cell counts (5).

No SNPs in the vicinity of VHL have been identified using GWAS studies; however, both loss-of-function and hypomorphic mutations in VHL lead to disease. For example, VHL mutations lead to the development of a number of tumors, including highly vascular hemangioblastomas and clear-cell renal carcinomas (33). In addition, the hypomorphic VHL R200W allele leads to the development of Chuvash polycythemia and pulmonary hypertension (3, 4), probably because of the activation of HIF-2α (25). Also, neutrophils from patients with heterozygous mutations in VHL are apoptosis resistant and have an increased phagocytic activity (73), and a full loss of VHL results in an almost complete resistance to viral infection of renal clear-cell carcinoma and glioblastoma cell lines (27).

Novel Approaches to Study Hypoxia in the Vasculature

Local oxygen levels can be measured indirectly by detecting the incorporation of pimonidazole in cells exposed to a Po₂ of <10 Torr (22). This approach presents an all or nothing measurement at a single time point, and novel tools have been developed recently to circumvent these limitations (50). For example, the phosphorescent PtP-C343 probe has a phosphorescence decay time that depends on local oxygen levels, and this can be imaged in vivo, for example, inside the bone marrow, using two-photon microscopy (68). This technique allows oxygen concentration measurements along an individual blood vessel in a live animal.

The development of large-scale transcriptomic analyses using single-cell RNA-Seq is revolutionizing our understanding of vascular heterogeneity at a single-cell level (34, 87). For example, endothelial cells from different organs could readily be distinguished based on their G-protein-coupled receptor (GPCR) expression levels, but even within one organ there were significant differences between the expression levels in individual endothelial cells (34). Similarly, endothelial cells from different organs have clearly distinguishable gene regulatory networks (28). This heterogeneity is likely also present for the response of endothelial cells to hypoxia, but it is an open question whether some endothelial cells are primed to respond to hypoxia, whereas others might be relatively insensitive.

Importantly, endothelial cells in vivo are exposed to spatial hypoxia gradients and not to uniform oxygen levels. This spatial complexity of the oxygen environment can be mimicked in vitro by culturing cells in a microfluidic device consisting of an oxygen-permeable polydimethylsiloxane membrane perfused with gas of different oxygen concentrations (60, 61). By combining different channels and gas mixtures, linear, exponential, and more complex gradients can be generated, and these gradients can be modified over time (1). The use of hydrogels even allows the generation of 3D hypoxia gradients, where the thickness of the hydrogel can be manipulated to establish different gradients, and this system can be studied for up to 2 wk (42). In addition to regulating local oxygen concentrations using gas mixtures, it is also possible to use chemical reactions to locally produce or consume oxygen (11). Even more complex 3D structures can be generated by seeding cells together with collagen into a porous cellulose scaffold. This scaffold can then be rolled up to generate a 3D structure. Unrolling the sheet allows for the isolation of cells from specific locations within the 3D structure (63). These approaches enable the generation of ex vivo oxygen gradients, which permits the dissection of how spatial oxygen gradients may differentially impact hypoxia responses.

Translational Perspective

Enhancing HIF-1α or HIF-2α signaling has a clear beneficial effect in the settings of organ transplant, irradiation, or pathological increases in vascular permeability. Desferrioxamine, which activates both HIF-1α and HIF-2α, is already in clinical use as an iron chelator, and therefore safety has already been established. Nevertheless, it should be clear from the aforementioned studies that such a broad activation of both HIF-1α and HIF-2α throughout the body might have unintended consequences, such as activating inflammation, which limit the therapeutic efficacy. The endothelium can easily be reached through the circulation, and we expect that novel drug delivery methods will be developed that allow specific activation of HIF-1α and HIF-2α signaling in the endothelium. Moreover, because of the divergence between HIF-1α and HIF-2α signaling pathways in endothelial cells, it should be a high priority to develop drugs that specifically inhibit or activate either isoform.

Understanding the cellular and tissue-specific heterogeneity of hypoxia responses on a molecular level in the vasculature should be critical for future therapeutic approaches that take advantage of the adaptive and reparative pathways activated by hypoxia while containing hypoxia-induced tissue injury.