Oxygen supply and consumption are tightly regulated processes in multicellular organisms and when oxygen demand exceeds supply, a state of reduced oxygenation called hypoxia ensues. Hypoxia-inducible factors (HIFs) are master regulators of cellular and systemic adaptation to poor oxygen availability in physiological and pathological conditions. They promote expression of a growing number of tissue-specific genes that mediate adaptation to acute decrease in oxygen supply as well as molecular functions occurring in cell compartments that are physiologically exposed to low oxygen levels (1).

Breakthrough discoveries into the molecular mechanisms of how cells sense and adapt to oxygen availability led to the 2019 Nobel Prize in Physiology or Medicine being awarded to Drs. Gregg Semenza, William Kaelin, and Peter Ratcliffe (2).

Genes that tune cellular metabolism to atmospheric oxygen availability belong to an ancient pathway that co-evolved with oxidative phosphorylation in the first metazoans (3). Orthologues of various components of the HIF pathway (PHD, FIH, VHL, HIF, described later in detail) are conserved from simple organisms like nematodes to mammals (3) and along evolution many of these genes have been subjected to events of gene duplication. Such is the case of HIF factors, the central molecules in the hypoxia adaptation pathway, with HIF1A being shared by all metazoans, and EPAS1 (the gene encoding HIF2α) and HIF3A having appeared at later branches of metazoan evolution. Additionally, oxygen sensing domains embedded within HIF molecules display increased molecular complexity throughout evolution, with primitive forms of HIFα expressing a single oxygen-sensitive regulatory domain and genes that have evolved later containing repetitions of these sites (3). Ongoing studies are directed to define the functional specification of the three mammalian HIFα genes in different physio/pathological cell states and cell types. However, these studies are at the very beginning, and further analyses in normal and pathological tissues will lead to a better understanding of the diversification and full regulatory potential of this pathway.

Importantly, although HIF factors have been initially defined as central regulators of cellular adaptation to acute oxygen scarcity, it must be highlighted that low oxygen levels are physiological in certain tissues or tissue microenvironments (4) and that, besides acute hypoxia, HIF factors are also expressed in cells ordinarily exposed to low oxygen, where they take part to the regulation of physiological mechanisms. Along these lines, a common semantic error is to consider the atmospheric oxygen partial pressure (pO₂) of 160 mm Hg as normoxic, with hypoxia indicating pO₂ values below 38 mm Hg. Although in vitro this is the condition where HIF factors begin to accumulate in many cell types, we must remember that most cell lines are adapted to grow at oxygen tensions as in the atmosphere, even though the tissue they originated from was probably exposed to much lower oxygen levels. Thus, the terms “hypoxia” and “normoxia” should be used attentively (5).

One typical tissue that is considered physiologically hypoxic is the bone marrow (BM), where low oxygen levels are believed to be intrinsic, especially in microenvironments devolved to the control of hematopoietic stem cells (HSCs) homeostasis (4). As such, HIF factors have been implicated in the biology of HSCs and hematopoietic cells at large and, as we will review in this manuscript, this also applies to the pathological derivatives of hematopoietic cells: leukemic cells.

Before describing the known functions and regulation of hypoxia-inducible genes in leukemia, we will use this chapter to briefly summarize the way these genes work. HIF proteins are a family of evolutionarily conserved DNA-binding transcription factors belonging to the bHLH/PAS (basic-helix-loop-helix/Per-ARNT-Sim) family. They function as heterodimers composed of an oxygen-sensitive HIFα subunit and the constitutively expressed HIF1β subunit (or aryl hydrocarbon receptor nuclear translocator, ARNT) (6). Three different HIFα genes exist in the human genome: HIF1A, EPAS1 (endothelial PAS domain protein 1 gene, encoding HIF2α), and HIF3A. HIF1α and HIF2α bind HIF1β to assemble HIF1 and HIF2 active transcription factors, which recognize hypoxia responsive elements (5’-(A/G)CGTG-3’, HREs) in the regulatory regions of HIF-target genes (7). HIF3α is less studied than its two siblings and has been suggested to act as a dominant-negative regulator of hypoxia signaling via competition for HIF1β binding by some of its splicing variants and transcriptional inhibition (8). More recently, HIF3α has been identified as a lipid sensor, with endogenous lipids stabilizing its heterodimerization with HIF1β (9), which may lead to regulation of lipolysis genes in adipocytes (10). However, because information on the function of HIF3α is still scarce, in this review we will focus on HIF1α and HIF2α.

HIF proteins share homologous amino-terminal (N-terminal) domains, whilst diverging at their carboxyl-terminal (C-terminal) region. At the N-terminus, all HIF proteins carry a bHLH domain that is required for DNA binding and PAS-A and PAS-B domains that mediate heterodimerization and formation of the transcriptional complex. In the central region, an oxygen-dependent degradation domain (ODDD) is crucial for oxygen-dependent degradation of HIFα proteins as described below, and is lacking in HIF1β, thus explaining its oxygen-independent expression. At the C-terminus, HIF1α and HIF2α have two transactivation domains (N-TAD and C-TAD) that are essential to promote transcription of HIF-target genes, while HIF3α and HIF1β have only one TAD domain (11). Importantly, the C-terminal half of HIFα and HIF1β proteins contains intrinsically disordered regions (IDRs) that exert critical functions in transcriptional regulation by mediating diffusion properties and interaction with chromatin factors (12).

As their name suggests, HIF proteins are importantly regulated by oxygen levels. However, HIFs are not oxygen sensors per se, rather they are post-translationally regulated by enzymes that use molecular oxygen to promote protein hydroxylation, thus acting as oxygen sensors (13). When oxygen is available, prolyl hydroxylases (PHD1-3) modify two conserved prolyl-residues located in the ODDD of HIFα proteins (P402/P564 in human HIF1α and P405/P531 in HIF2α) and, in so doing, promote their recognition by the von Hippel-Lindau (pVHL) E3 ubiquitin ligase that triggers their degradation by the proteasome via poly-ubiquitination (14, 15). In addition, an asparagine residue located in the C-TAD of HIF1α and HIF2α (N803 in human HIF1α and N851 in HIF2α) confers further oxygen-dependent regulation upon hydroxylation by the asparaginyl hydroxylase FIH (factor inhibiting HIF), which blocks interaction with the transcriptional coactivator complex p300/CBP and suppresses transcriptional activity (11, 16, 17) ( Figure 1A ). Both PHDs and FIH belong to the 2-oxoglutarate (2-OG)-dependent dioxygenase superfamily that uses oxygen and 2-OG as co-substrates together with ferrous iron (Fe²⁺) and ascorbate as obligate cofactors to catalyze protein hydroxylation (15). The activity of PHDs and FIH is inhibited under low oxygen tensions due to lack of the essential co-substrate. As a consequence, HIFα subunits are stabilized, dimerize with HIF1β and bind HREs on specific target genes together with several coactivators, including CBP/p300 ( Figure 1B ). Of note, PHD2 and PHD3 are themselves hypoxia-inducible proteins, thus providing a self-sufficient mechanism to control HIFα protein levels and ensure rapid removal of HIFα upon reoxygenation (18). Also, PHDs exhibit tissue-specific expression (for example, PHD3 is strongly expressed in the heart and PHD1 is the only isoform expressed in the testis) and differential HIFα hydroxylation, with PHD1 and PHD3 being more active on HIF2α and PHD2 hydroxylating more efficiently HIF1α (18, 19). Consequently, oxygen-dependent HIFα induction may also reflect differences in PHDs tissue distribution and activity.

Active HIF heterodimers promote transcription of an ever-growing number of genes involved in a variety of cellular functions, with important implications for both physiology and disease. Hypoxia signaling has been linked to control of cellular metabolism (with a switch from mitochondrial oxidation to glycolysis), neo-angiogenesis, regulation of cell migration and tissue invasion, regulation of cell proliferation and apoptosis, and regulation of stem cell maintenance, amongst other functions (20). Mechanistically, HIF1α and HIF2α show high sequence homology in their DNA-binding domains (21) and bind identical HRE motifs in vitro (7, 22), thus regulating common target genes (23, 24). However, emerging evidence indicates that HIF1α and HIF2α are endowed with important target selectivity, with HIF1α mainly promoting the expression of genes involved in glycolytic metabolism, pH regulation and apoptosis, whilst HIF2α regulates genes involved in stem cell maintenance, cell cycle and invasion (24), albeit this division of functions does not occur in all tissues and a strong element of tissue specificity is emerging in the HIF transcriptional program. The functional specification of HIFα factors has been connected to specific interactions with transcriptional coactivators occurring in their C-terminal transactivation domains, which have a lower degree of sequence homology (25). In addition, more recently it was proposed that HIFα proteins display different diffusion properties and DNA binding profiles thanks to the different amino acidic compositions of their IDRs (12). Specifically, it was revealed that HIF2α has slower nuclear diffusion than HIF1α and a tendency to bind negatively charged chromatin-associated RNA and proteins due to the preeminent presence of positively charged amino acids in its IDR (12). Adding to these intrinsic properties, tissue- and developmental-specific expression and different sensitivity to oxygen levels lead to an emerging pattern of differential gene regulation by HIF1α and HIF2α (23, 24, 26).

Interestingly, it has been suggested that the diversification of HIFα functions may originate from their evolutionary history (3). Having evolved to adapt cellular physiology to acute oxygen scarcity, HIF1A, which is the most ancient member of the family, is presumed to be the primary regulator of metabolic rewiring, while the later appearance of EPAS1 may have led to a wider spectrum of functions, including processes that occur within cell types residing physiologically at low oxygen levels (i.e. stem cells). However, as we will describe later, such a clear diversification of functions may not apply to all tissues or cells.

In conclusion, it is worth mentioning that, beside their role as transcriptional activators, HIFα have also been described as transcriptional repressors for certain genes (e.g. peroxisome proliferator activated receptor (PPAR)α, α-fetoprotein, leukemia inhibitory factor receptor) via binding to HREs located on the minus DNA strand (also called reverse HREs, rHREs) (27–29). Recently, binding of HIFα subunits to rHREs has been shown to repress transcription via epigenetic silencing promoted by recruitment of histone modifiers like HDAC1 and EZH2 (the catalytic subunit of Polycomb repressive complex 2, PRC2) (30–32) ( Figure 1C ). Because HIF factors also promote the expression of a number of epigenetic factors, epigenetic mechanisms are emerging as a relevant way of controlling gene expression downstream hypoxia signaling (33).

In humans, oxygen gradients differ greatly among tissues and tissue microenvironments. In the BM, oxygen levels are overall lower than in many other organs, and hematopoietic cells experience physiological pO₂ ranging from 32 mm Hg in arteriole-rich endosteal zones to 9.9 mm Hg in deeper regions (34, 35). The BM is home to a variety of non-hematopoietic cell types that constitute specialized microenvironments, known as niches, where stem cells, progenitors and more differentiated hematopoietic cells reside and are physiologically regulated (36). Despite its extensive vascularization, the BM is considered a tissue low in O₂, likely due to the extent of oxygen consumption caused by its high cellularity. Initial studies using in silico modeling of oxygen BM gradients and quantification of BM perfusion led to the assumption that HSCs reside near the poorly perfused endosteal niche and show signs of hypoxia due to their distance from oxygen-rich blood vessels (37–39). Despite this general assumption, measuring the exact oxygen tension and physical location of niches low in O₂ in the BM is technically challenging, therefore indirect measurements have been often utilized, such as evaluation of HIF1α expression or incorporation of hypoxia markers like pimonidazole (Pimo) and Hoechst 33342 (35, 38, 39). However, use of chemical agents, such as Pimo and Hoechst 33342 may be misleading as they do not provide a direct assessment of O₂ levels and may rather offer a readout of specific cellular conditions. As an example, Pimo, which competes with oxygen as an electron acceptor, may display increased reductive activation not only as a consequence of decreased intracellular oxygen, but also upon distinctive metabolic shifts from mitochondrial to glycolytic catabolism (40). Similarly, uptake of the fluorescent DNA intercalant Hoechst 33342, which has been used to measure blood perfusion as an indirect marker of tissue oxygenation, may vary depending on cell proliferation rates. Conversely, use of two-photon phosphorescence lifetime microscopy, which directly measures oxygen levels, revealed that the BM is overall a tissue low in O₂ where intravascular oxygen tensions are quite similar in endosteal and sinusoidal vessels (21.9 and 17.7 mm Hg respectively), and a moderate extravascular oxygen gradient exists, with higher oxygen levels in the endosteal region and decreased oxygen towards the sinusoidal region (34). Of note, these experiments have been performed in the calvaria, which may not represent all BM compartments, as Pimo staining revealed that hypoxic cells are less frequent in this site compared to long bones (41). Therefore, it remains unclear whether the current knowledge on BM oxygen levels may be relevant for most bones.

Nonetheless, the notion that hematopoietic cells thrive in an environment that is low in O₂ has been validated over the years by a number of in vitro or ex vivo experiments demonstrating that low oxygen levels: i) promote HSCs quiescence and maintenance (42–48), while at the same time supporting a balanced differentiation program towards many hematopoietic lineages (49–52); ii) favor a metabolic switch to (anaerobic) glycolysis thus exerting protection from oxidative stress that may cause DNA damage (35). Also, hypoxia efficiently enhances in vitro expansion and in vivo engraftment of HSCs (53–55).

More recently, it was revealed that, besides external oxygen availability, HSCs display an inherent hypoxic state that is independent of their exact location within the BM and may be rather dictated by internal metabolic activity (40). In this respect, it must be emphasized that expression and activity of HIF factors is not only regulated by oxygen levels but also by a variety of oxygen-independent mechanisms that promote their expression and transcriptional activity. Providing a list of these mechanisms is beyond the scope of this review, as it has been discussed elsewhere (56, 57), but suffice it to say that they include extracellular stimuli like growth factors, cytokines and chemokines. Thus, the molecular milieu of different BM niches may participate considerably to the regulation of hypoxia signaling within hematopoietic cells. Whatever the triggering event, one certain finding is that HIF factors are highly expressed in hematopoietic stem and progenitor cells (HSPCs), where they play important functions. This was clearly demonstrated in knock-out mice, where it was shown that HIF1α exerts a critical cell-autonomous functions in promoting HSCs quiescence (58), and HIF2α promotes HSCs maintenance predominantly via regulation of the BM microenvironment (59).

In the case of hematological malignancies, leukemic cells are great oxygen consumers, but at the same time promote BM neoangiogenesis, which increases nutrients and oxygen availability (60). Thus, it is not clear if abundance of proliferating leukemic cells and high oxygen and nutrient consumption may further decrease oxygen availability in the leukemic BM. By measuring BM blood gas levels in healthy volunteers and AML patients, two independent reports observed similar pO₂ of around 47 mm Hg (61, 62). Conversely, in vivo intracellular hypoxia labeling with the indirect marker 2-nitroimidazole in a rat model of AML showed increased reactivity during disease progression and in comparison with healthy animals (63). More recently, direct measurement of oxygen tension in leukemic BM was obtained in a BCR-ABL B-ALL mouse model by intravital fast scanning two-photon phosphorescence lifetime imaging microscopy, which showed that oxygen levels vary from the initial to the final stages of leukemia, with a transient increase in oxygen levels at intermediate leukemia burden, correlating with expansion of the vasculature network, and a later significant decrease of BM oxygenation as leukemia cellularity increased at disease end-stage (64). By suggesting that different oxygen levels characterize different stages of leukemia development, these findings may provide an explanation to the apparently contradictory findings described above. However, as previously discussed for normal BM, also and particularly in a leukemic BM where the cytokine milieu is perturbed towards pro-inflammatory and leukemia-sustaining cytokines, intracellular hypoxic pathways may be activated via additional routes. In this respect, work from our and other laboratories has shown that HIF1α is transcriptionally upregulated when leukemic cells are cocultured with BM mesenchymal cells (65–67). Also, a number of studies that we will describe in the next sections have reported that upregulation of HIF factors in leukemic blasts does not necessarily occur at the post-translational level, thus implying oxygen-independent mechanisms. In the following chapter, we will provide a summary of the regulation and function of HIF1α and HIF2α across blood malignancies.

Since their cloning and molecular characterization, the function of HIF factors has been intensely investigated in solid tumors. More recently, increasing research efforts have also focused on the involvement of hypoxia signaling in leukemia, albeit with somewhat controversial results that fail to provide a unanimous consensus. Depending on the type/subtype of leukemia, or the experimental approach that has been utilized (knock-out animals versus knock-down experiments in cell lines/primary cells), HIFs have been described as tumor suppressors, oncogenes, or neutral genes even within the same leukemia. In the next paragraphs, we will summarize current knowledge on the expression and activity of HIF factors in different leukemic contexts (acute and chronic, myeloid and lymphoid) and discuss possible explanations to controversial evidence recently published. Of note, the majority of the studies that we will describe have focused on defining the functions of HIF1α, whereas HIF2α has been much less investigated and detailed analyses on the functions of this factor in leukemia establishment and progression are still lacking.

Hypoxia-responsive transcription factors are being increasingly implicated in the regulation of many normal and pathological cellular processes that expand well beyond mediating metabolic adaptation to oxygen deprivation, which led to their evolutionary emergence. Their increasing pervasiveness in cell biology is probably due to a number of reasons. First and foremost, their constitutive expression in conditions of physiological hypoxia. Most cell lines in use in laboratories around the world have been adapted to in vitro culture at pO₂ of ambient air (160 mm Hg). However, direct measurement of oxygen levels in vivo reveals that much lower oxygen tensions are homeostatic in many tissues. The BM is a typical example of an organ where low oxygen levels appear directly implicated in maintenance of both normal hematopoietic and leukemic cells. This is particularly true in tissue microenvironments such as stem cell niches, where HIF factors promote a condition of metabolic “dormancy” based on a shift from mitochondrial metabolism to glycolysis that lowers production of DNA-damaging reactive oxygen species (ROS), and at the same time activate molecular programs implicated in stem cell maintenance. In addition to local hypoxia, growing evidence indicates that HIF1α and HIF2α expression is also promoted by extracellular stimuli or cell adhesion molecules provided by stromal cells, which probably reinforce HIF activity in specific BM microenvironments. As a consequence, hypoxic responses may be intrinsically wired in some hematopoietic and leukemic cells and occur at least in part independently of oxygen levels.

HIF1α and HIF2α share substantial homology and a similar structural organization, yet emerging literature is revealing diverging transcriptional outputs. The growing complexity of HIFα specific functions appears to be governed at different levels: i) tissue-specific expression of HIFα genes; ii) different sensitivity to oxygen concentrations of HIFα proteins; iii) different oxygen-independent regulation; iv) different nuclear diffusion properties and DNA binding profiles; v) binding to specific partners that promote cooperative transcriptional activation; vi) tissue-specific epigenetic modulation of target genes. The sum of all of these events leads to a diversification of HIF1α and HIF2α functions with non-overlapping consequences that are constantly emerging in physiological and pathological conditions.

With this in mind, it is to be expected that the importance of this pathway in biology will continue to grow. As per the role of hypoxia signaling in oncology, at current time a large amount of work has described the implication of HIFα factors in solid tumors, where in most cases they promote features of tumor aggressiveness, such as metastatic spread and relapse after treatment. Conversely, although hypoxia is a physiological hallmark of hematopoietic organs, the contribution of HIF1α and HIF2α to hematological malignancies has been underestimated for a long time. In recent years, many research groups have started to look into the function of hypoxia-responsive genes in blood malignancies. As we have summarized in this manuscript, important leukemia-promoting functions of HIF factors have emerged across distinct types of leukemia, including protection of LSCs, leukemia dissemination, leukemia expansion and impaired sensitivity to apoptosis and chemotherapy. Nonetheless, parallel work suggests that HIFα factors may not be essential or rather play tumor-suppressive functions in leukemia. Explanations to these contradictory results may rely on divergence of functions in distinct leukemia subtypes or at the preleukemic stage versus overt leukemia. In this respect, most of the studies that describe tumor-suppressive functions of HIFα factors in leukemia are based on genetic inactivation of these genes at a preleukemic stage, or concomitantly to leukemia initiation by mutated oncoproteins, a circumstance where it may be difficult to disentangle the function of HIFα genes in hematopoiesis or leukemia. In conclusion, further investigation is required to fully elucidate the extent and relevance of hypoxia-responsive gene activation in leukemia. In so doing, we should also aim to address the translational implications of inhibiting these pathways for the treatment of blood malignancies.

In this respect, an important note is that the vast majority of studies performed in leukemic contexts have focused on the HIF1α gene, with much less effort into elucidating the functions of HIF2α. Because HIF2α is specifically expressed in some cell types and its transcriptional output may differ from that of HIF1α, additional work may uncover specific context-dependent functions of HIF2α that are different from those exerted by HIF1α. Of relevance, because a specific small molecule inhibitor of HIF2α (belzutifan) has been recently approved by the U.S. FDA (Food and Drugs Administration) for the treatment of adult patients with VHL disease associated with renal cell carcinoma or pancreatic tumors, future studies into this pathway may pave the way for the use of this compound in other malignancies including blood cancers.

Both authors collected and reviewed relevant literature and wrote this manuscript.

This work was supported by the Italian Ministry of Health (Ricerca Finalizzata, RF-2019-12369841).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.