Introduction

Hypoxia refers to a lower-than-normal level of oxygen in an organ and is a common feature of solid tumors. Normal oxygen levels in healthy tissues range widely between organs from approximately 14.5% O2 in the lung alveolus to 3% O2 in lymphoid tissues but average around 5% O2 [1,2,3]. As the rapid proliferation of cancer cells outgrows the development of new blood vessels in a tumor, the O2 levels in cancerous tissue lesions become significantly lower than in healthy tissue, falling between 0.3% and 4.2% O2 [3]. As early as the 1930s, researchers discovered that hypoxic conditions could protect cancer cells from radiation-induced cell death [4]. Later studies suggested that this radiation-resistant phenotype results from reduced DNA damage and cell cycle arrest, and increases in stemness-related gene products [5, 6]. Hypoxic cancer cells are also resistant to chemotherapeutic agents due to increased drug efflux pumps, DNA over-replication, and the upregulated expression of glucose- and oxygen-regulated proteins [7,8,9,10]. More recently, additional studies have confirmed that hypoxia can impair anti-cancer immune responses by regulating the tumor microenvironment (TME) [11,12,13].

Hypoxia-inducible factors (HIFs) are key transcription factors mediating the activation of many oxygen-dependent genes. HIF activation occurs in many pathological disease states, including ischemia, anemia, pulmonary hypertension, kidney disease, and cancer [14,15,16,17,18]. In 1992, Wang GL and Semenza GL first identified HIF-1 as a DNA-binding protein that regulates erythropoietin (EPO) gene transcription in response to hypoxia [19]. Later studies showed that HIF-1 is a heterodimer composed of a hypoxia-inducible alpha subunit and a constitutively expressed beta subunit HIF-1β (also known as the aryl hydrocarbon receptor nuclear translocator, ARNT) [20]. To date, three isoforms have been described, HIF-1α, -2α, and -3α subunits that all bind to HIF-1β [21,22,23]. Among these isoforms, HIF-1α and HIF-2α are the most well-characterized. During canonical HIF signaling, hypoxia prevents the degradation of HIF-1α or HIF-2 α, which causes their accumulation. Subsequently, HIF-1α or HIF-2α form heterodimers with HIF-1β and bind to DNA motif(s) within target gene promoters known as hypoxia-response elements (HREs) [24]. Data from knockout studies in mice suggested non-redundant roles of HIF-2α during embryonic development. HIF-1α−/− mice showed cardiac and vascular malformations resulting in embryonic lethality at E10.5 [25]. Interestingly, HIF-2α−/− mice also exhibit embryonic lethality, with severe vascular defects and abnormal lung maturation [26,27,28].

In this review, we describe an overview of the structure, expression, and DNA binding of the HIF-1 and HIF-2 complexes. We use the terminology ‘canonical’ to describe hypoxia-induced HIF regulation and non-canonical for hypoxia-independent HIF regulation. We further discuss the most recent therapeutic approaches targeting HIF-1 and HIF-2, to provide new insights into cancer treatment.

figure a

Structure, expression patterns, and DNA binding of HIF subunits

Both HIF-1 and HIF-2 are heterodimeric proteins composed of an O2-sensitive alpha subunit and a constitutively expressed beta subunit. Full-length HIF-1α or HIF-2α consists of 826 or 870 amino acids, respectively. Both HIF-1α and HIF-2α contain a basic helix-loop-helix (bHLH) domain that is essential for binding to the HRE in the DNA sequence of target gene promoters. They both also contain Per-ARNT-Sim (PAS) domains in the N-terminus, allowing them to heterodimerize with HIF-1β [24, 29]. The C-terminus of HIF-1α and HIF-2α include two transactivation domains (TADs), an N-terminal transactivation domain (N-TAD) and a C-terminal transactivation domain (C-TAD). TADs regulate the interaction of HIFs with co-activators [30]. Moreover, all HIF-1/2α subunits contain an O2-dependent degradation domain (ODDD) that overlaps with the N-TAD, structurally distinguishing HIF-1/2α subunits from HIF-1β. The ODDD serves as the recognition site of the von Hippel-Lidau tumor suppressor protein (pVHL) and mediates oxygen-regulated HIF-1/2α subunit stability. Proteasomal degradation of HIF-1/2α will rapidly proceed when two proline residues in the ODDD are hydroxylated by oxygen-sensitive prolyl hydroxylase domain (PHD)-containing proteins [31]. Unlike the short half-life of HIF-1/2α, the protein level of the HIF-1β subunit remains constant due to the lack of ODDD and N-TAD (Fig. 1).

Fig. 1
figure 1

Structure of HIF isoforms and functional domains. HIF-1α shares a high level of similarity with HIF-2α, contain bHLH motif for DNA binding, two PAS domains (PAS-A and PAS-B) for heterodimerization, TADs (N-TAD and C-TAD) for co-activator binding and transactivation, and ODDD for proteasomal degradation. HIF-3α only has bHLH, two PAS domains, ODDD, and N-TAD. HIF-1β does not contain ODDD or N-TAD

Full size image

Although HIF-1α and HIF-2α both contain bHLH, PAS, ODDD, and TAD domains, the domains differ in their sequence homology [32]. For example, the DNA binding, dimerization, and C-TADs have 83%, 70%, and 67% similarity respectively [33]. The main structural difference between HIF-1α and HIF-2α lies in the ODDD. The distinct regions within this domain lead to differences in their regulation and stability. HIF-1α degradation is dependent on the hydroxylation of proline residue Pro402 and Pro564, while HIF-2α is hydroxylated on Pro405 and Pro531 in the ODDD. p300/CBP (E1A binding protein p300/CREB-binding protein) is one of the essential co-activators that bind to the C-TAD of both HIF-1α and HIF-2α enhancing HIF-1 and HIF-2 transcriptional activities [34]. In addition to p300/CBP, some specific co-activators that only bind to HIF-1α or HIF-2α have also been found. Signal transducer and activator of transcription 3 (STAT3) interacts with the N-TAD of HIF-1α and increases the recruitment of p300/CBP and RNA polymerase II (Pol II), thereby increasing HIF-1, but not HIF-2-mediated hypoxic transcriptional responses [35]. Likewise, upstream stimulatory factor 2 (USF2) activates HIF-2 target genes by binding to HIF-2α N-TAD recruiting p300/CBP to form enhanceosome complexes at HIF-2 target gene promoters [36]. In addition, HIF-2α contains two bipartite-type nuclear localization signals (NLS), the N-terminus NLS (amino acids 14–50) and the C-terminus NLS (amino acids 705–742) [37]. Under hypoxia, the NLS mediates nuclear accumulation of HIF-2α, which improves HIF-2α binding to the promoters of target genes. HIF-1 is often associated with the acute response to hypoxia, promoting cellular adaptation to low oxygen conditions by enhancing glycolysis, reducing oxygen consumption, and inducing angiogenesis [38,39,40,41]. HIF-2 has been implicated in more chronic responses to hypoxia and regulates erythropoiesis (production of red blood cells) and vascularization [42, 43].

There are also differences in the tissue distribution of each subunit. HIF-1α is widely expressed in human organs including brain, heart, lung, liver, kidney, and pancreas [44]. HIF-2α has tissue-specific expression and is restricted primarily to embryonic cells, endothelium, lungs, kidney, and liver [45]. The overexpression of HIF-1α has been observed in a variety of cancer types and the prognostic relevance of HIF-1α in cancer progression has been demonstrated [46,47,48]. For example, a retrospective study of 745 patients suggested that HIF-1α is expressed in breast carcinomas with higher expression in patients with poor survival rates [49]. The expression of HIF-1α is significantly higher in lung cancer compared to normal lung tissue according to a systematic review and meta-analysis [50].

Results from a ChIP-seq-based high-resolution genome-wide mapping suggest that compared to HIF-2, HIF-1 preferentially binds to the transcriptional start site (TSS) of genes [51]. Approximately 40% of HIF-1-binding sites are located within 2.5 kb of the TSS, while only 20% of HIF-2-binding sites lie within the same range. HIF-1 predominantly binds to genes regulating carbohydrate metabolism, while HIF-2 binds to genes that play an essential role in stem cell pluripotency by mediating the Octamer transcription factor 4 (Oct4) pathway. Additional studies show that different cancer cell lines exhibit diverse binding patterns for HIF-1 and HIF-2. For example, in A549 (lung cancer) cells a higher proportion of DNA was bound by HIF-2-bound rather than HIF-1 whereas the opposite is true in HCT116 (colorectal cancer) cells [52]. Additionally, HIF-1 and HIF-2 have distinct chromatin binding preferences based on specific histone modifications. For example, HIF-1 associates more strongly with histone H3K4me3 and H3K9ac modifications (marks primarily associated with promoters and TSSs); HIF-2 is more strongly related with H3K27me1 and H3K27ac (marks primarily associated with enhancers and other distal regulatory elements) [53]. On the other hand, the methylation status of HREs can alter the ability of HIF-1/HIF-2 to promote the transcriptional regulation of hypoxia-inducible genes under hypoxic conditions [54, 55].

In summary, HIF-1α shares a high level of similarity with HIF-2α. These common features make both HIF-1α and HIF-2α biomarkers of poor prognosis in cancer patients [56]. However, their differences in structure and expression patterns may explain their unique response to hypoxia and thus differences in their functional roles [38,39,40,41,42,43].

Regulation of HIF-1 and HIF-2

The regulation of HIF-1/2α predominantly depends on protein stability and accumulation through post-translational modifications. The canonical oxygen-dependent regulation involving pVHL pathway has been well established; however, non-canonical pathways that regulate HIFs are complex and continue to require further investigation.

Canonical regulation

Under physiological O2 concentrations, HIF-1α and HIF-2α are maintained at low levels due to constant ubiquitination-dependent degradation via interactions with pVHL [57,58,59]. pVHL functions as the recognition component of an E3 ubiquitin-ligase complex and recognizes and binds hydroxylated HIF-1/2α at proline residues (Pro402 and Pro564 in HIF-1α and Pro405 and Pro531 in HIF-2α) [60]. HIF-1/2α proline hydroxylation is catalyzed by PHDs (PHD1, 2, and 3) whose total enzymatic activity requires the presence of O2, ferrous iron (Fe (II)), and ascorbate as cofactors [61]. PHD2 preferentially hydroxylates HIF-1α. The regulation of HIF-2α by PHDs might involve a more complex interplay than HIF-1α. Some studies suggest that PHD2 is the primary regulator of HIF-2α stability, while others also propose a role for PHD1 [62,63,64]. Under hypoxic conditions, PHD activity is suppressed, which decreases HIF-1/2α proline hydroxylation and degradation, resulting in HIF-1/2α protein accumulation. The stabilized HIF-1/2α subunit translocates into the nucleus, dimerizes with HIF-1β, and binds to HREs, leading to the transactivation of hypoxia-responsive genes (Fig. 2) [65].

Fig. 2
figure 2

Adapted from The transcriptional factors HIF-1 and HIF-2 and their novel inhibitors in cancer therapy, by Albadari N, et al., 2019, Expert opinion on drug discovery, 14(7):667–682

Canonical regulation of HIF-1/2α activity under normal and hypoxic conditions. Under normal O2 concentrations, HIF-1/2α is hydroxylated by PHDs and FIH, which allows the formation of HIF-pVHL complexes, resulting in proteasomal degradation. Under hypoxia, the lack of oxygen limits PHD and FIH activities, leading to the accumulation of HIF-1/2α. After translocating to the nucleus, HIF-1/2α forms a heterodimer with HIF-1β. The heterodimer then binds to HRE and recruits co-activators such as p300/CBP to the promoter region of target genes to regulate transcription.

Full size image

In addition to prolyl hydroxylases, HIF-1/2α oxygen-dependent stability is mediated by factor inhibiting HIF (FIH) which belongs to the Fe (II) and 2-OG-dependent dioxygenase family of hydroxylases [66]. This asparagine hydroxylase modifies Asn803 or Asn847 residues within the C-TADs of HIF-1α and HIF-2α, respectively. This prevents HIFs from associating with transcriptional co-activators such as p300/CBP (Fig. 2) [67, 68]. FIH is activated only in the presence of molecular oxygen. The limited oxygen availability during hypoxia leads to the suppression of FIH, decreases in HIF-1/2α asparagine hydroxylation, and thus stabilization of the transcriptional complex. FIH preferentially hydroxylates HIF-1α potentially due to the valine surrounding the asparagine hydroxylation site in contrast to the alanine present in HIF-2α [69]. In addition, FIH binds to pVHL and inhibits HIF-1α transactivation function by recruiting histone deacetylases [70]. This dual regulation by PHDs and FIH-1 adds another layer of complexity to fine-tuning HIF-mediated responses to changes in O2 levels.

Non-canonical regulation

Apart from the regulation by PHDs and FIH-1, multiple non-canonical mechanisms contribute to the regulation of HIF-1/2α protein stability, synthesis, and transcriptional activity (Fig. 3).

Fig. 3
figure 3

Non-canonical regulation of HIF-1 and HIF-2. HIFs are regulated by multiple interacting proteins and mechanisms independent of O2 regulation. For example, growth factors promote receptor tyrosine kinase activity and drives downstream pathways such as PI3K/AKT/mTOR and RAS/RAF/MEK/ERK which affect the levels of HIF-1α and/or HIF-2α protein

Full size image

Post-translational modifications of HIF-1/2α alter protein stability and transactivation

Increasing evidence indicates that the alpha subunits of HIF-1 and HIF-2 undergo posttranslational modifications including acetylation, SUMOylation, and phosphorylation. Arrest defective-1 (ARD1) N-acetyltransferase acetylates Lys532 of the HIF-1α ODDD which stabilizes its interaction with pVHL and inhibits HIF-1α transcriptional activity under both normal and hypoxic conditions [71, 72]. By contrast, the metastasis-associated protein 1 (MAT1) induces the deacetylation of HIF-1α at Lys532 residue, which counteracts the acetylation function of ARD1 [73]. The MAT1-induced deacetylation of HIF-1α is mediated by increasing the recruitment of histone deacetylase 1 (HDAC1), thereby promoting HIF-1α stabilization and transcription [74]. Additionally, Lys647 was shown to be deacetylated by an NAD-dependent deacetylase sirtuin 1 (SIRT1), blocking p300 recruitment and resulting in repressed HIF-1α transcriptional activity in HT1080 and HEK 293 cells [75]. However, another study demonstrated that SIRT1 selectively deacetylates HIF-2α and promotes its signaling during hypoxia in Hep3B cells [76]. More recently, using an overexpression system incorporating a Gal4-luciferase reporter, SIRT1 was found to inactivate HIF-1α CAD but to activate HIF-2α CAD in HEK293T cells. In follow-on studies using 10 different cell lines, SIRT1 repressed HIF-1α transcriptional activity but SIRT1 only affected HIF-2α activity in 3 of the 10 cell lines tested. [77]. Geng H, et al. reported that p300 acetylates HIF-1α at Lys709, stabilizing HIF-1α protein levels by decreasing polyubiquitination under both normal and hypoxic conditions [78]. The peptidyl-prolyl cis–trans isomerase NIMA-interacting 1 (Pin1) was found to prolong the stability of HIF-2α by modulating the levels of ubiquitinated HIF-2α, with no significant effect on HIF-1α [79]. However, another study in human colon cancer (HCT116) cells showed that Pin1 interacts with HIF-1α, stabilizing HIF-1α protein and subsequently increasing its transcriptional activity [80].

There is conflicting evidence on the role of SUMOylation in HIF-1α regulation; some reports suggest that SUMOylation can positively regulate HIF-1α stability, while others find that SUMOylation leads to HIF-1α degradation [81,82,83,84]. Studies mutating lysine residues in the ODDD led to the identification of SUMOylated lysine residues (391 and 477) in HIF-1α, enhancing its stability and transcriptional activity [81]. On the other hand, Cheng J, et al. report that hypoxia-induced SUMOylation promotes HIF-1α binding to pVHL in the absence of proline hydroxylation, resulting in ubiquitination and degradation [82]. HIF-2α is SUMOylated at Lys394 located in the SUMOylation consensus site LKEE, leading to its rapid degradation via SUMO-targeted ubiquitin ligases [85, 86].

Reactive oxygen species alter HIF-regulation

The role of reactive oxygen species (ROS) in HIF-1/2α regulation has been demonstrated in many studies. Induced by prolonged hypoxia, accumulation of ROS increases the expression of redox factor-1 (Ref-1), which promotes HIF-1α transcription [87]. However, the activation of HIF-1α induces PHD2 and FIH-1, causing a negative feedback loop that promotes HIF-1α degradation. As discussed in the Sect. “Canonical regulation”, Fe (II) is required for the hydroxylation of HIF-1α by PHD. Ascorbate, a redox active molecule, interacts with ferric iron Fe (III) and reduces it to Fe (II), causing the increased activity of PHDs and suppressed hypoxia-induced HIF-1α activity [88]. In an in vivo study, supplementing ascorbate reduces tumor growth, metastasis, and inflammatory cytokine secretion in ascorbate-deficient Gulo−/− mice [89]. Additionally, nitric oxide (NO) is also able to regulate HIF-1α stabilization and transcriptional activity by S-nitrosylation (SNO) of Cys533 located in ODDD of HIF-1α [90,91,92].

HIF-regulation by tumor suppressor genes: VHL, p53, and PTEN

Perhaps the most well-studied oxygen-independent regulation of HIFs occurs due to a mutation in the tumor suppressor gene, VHL, which can be either somatic or germline [93]. Mutations in the VHL gene result in a non-functional or absent VHL protein. Without functional VHL protein, hydroxylated HIF-1/2α cannot be targeted for degradation resulting in HIF accumulation in the cytoplasm [94]. Individuals with a hereditary VHL mutation have an increased risk of developing tumors, particularly in the central nervous system (CNS) and kidneys [95,96,97].

The tumor suppressor, p53, has also been reported to regulate HIF signaling [98]. p53 impairs HIF-1α protein stability through both mouse double minute 2 homolog (MDM2)-dependent and MDM2-independent pathways [99, 100]. In HCT116 colon carcinoma cells, p53 acts as a scaffold protein complexing the E3 ligase, MDM2 with HIF-1α, thereby leading to the HIF-1α ubiquitination and degradation [101]. Mutating the transactivation domain of p53 abolishes its ability to interact with MDM2, and reduces HIF-1α degradation, thus increasing the levels of HIF-1α [102]. However, Choy M, et al. demonstrated that ectopic expression of p53 promotes HIF-1α ubiquitylation and proteasomal degradation in p53/MDM2 double-null mouse embryonic fibroblasts, indicating other potential p53-mediated mechanisms that promote HIF-1α degradation [103]. A p53 target gene, Parkin (also known as PARK2), interacts with HIF-1α at Lys477 and ubiquitinates HIF-1α, consequently promoting HIF-1α proteasomal degradation [100]. Moreover, mutant-p53 inhibits the expression of SHARP1, a bHLH transcription factor that serves as a HIF-presenting factor to the proteasome and mediates proteasomal degradation of HIF-1α and HIF-2α [104,105,106]. In addition, p53 regulates the transcriptional activity of HIF-1α by competing for binding to the co-activator p300 [107, 108]. Transglutaminase-mediated p53 depletion increases HIF-1α-p300 binding and stabilizes HIF-1α transcriptional activity in renal cell carcinoma (RCC) [109]. Whole genome sequencing was performed on ten patients with clear cell RCC (ccRCC) on long-term treatment with the HIF-2α inhibitor, PT2358 [110]. One patient developed a HIF-2α mutation that prevents PT2358 from blocking the dimerization of HIF-2α with HIF-1β. A second patient acquired a p53 mutation, suggesting that p53 mutations could promote resistance to HIF-2α inhibitors.

Another tumor suppressor protein, phosphatase and tensin homolog (PTEN) is also involved in HIF-1α regulation. Loss of PTEN in glioblastoma-derived cell lines facilitates hypoxia-mediated HIF-1α stabilization, thereby contributing to tumor progression [111]. Treating ovarian cancer cells with gallic acid (GA) which induces PTEN expression by suppressing protein kinase B (AKT) phosphorylation, results in HIF-1α stabilization [112]. PTEN has also been shown to regulate HIF-1α transcription by inhibiting the PI3K/AKT/mammalian target of rapamycin (mTOR) pathway [113]. Myeloid cell-specific PTEN knockout in mice induces HIF-1α and HIF-2α stabilization in bone marrow-derived macrophages [114].

Epigenetic regulation of HIF-1/2α

DNA methylation is an epigenetic change that can alter HIF-1/2α protein stability and/or transcriptional activity. For example, Li C, et al. determined that methylation of both CpG and non-CpG sites in the HIF-1α promoter contributes to HIF-1α expression [115]. HIF-1α promoter methylation is higher in the luminal subtype as compared to triple-negative breast cancer (TNBC) samples. This may explain why TNBC cells express higher levels of HIF-1 compared to luminal cell lines and tissue samples. On the other hand, methylated CpG binding protein 3 (MBD3) binds to the HIF-2α promoter and promotes its transcription by demethylating CpG islands located around the TSS in breast cancer cells (MDA-MB-468) [116]. Hypermethylation of the VHL promoter leads to VHL silencing, resulting in HIF-1α constitutive expression and activation, which enhances the transcription of downstream target genes including carbonic anhydrase 9 (CA9) and glucose transporter type 1 (GLUT1) [117, 118].

Another epigenetic mechanism that regulates HIF-1α activity is histone methylation/demethylation. The histone lysine methyltransferases, G9a and G9a-like protein (GLP), directly bind to HIF-1α and methylate HIF-1α at Lys674 but do not bind to HIF-2α. This modification reduces HIF-1α transcriptional activity without altering HIF-1α protein degradation or binding to hypoxia response elements [119]. On the other hand, lysine-specific demethylase 1 (LSD1), also known as KDM1A, demethylates HIF-1α at Lys391, inhibiting its proteasomal degradation. This accumulation of HIF-1α causes an increase in miR-146a expression, which promotes the metastasis of papillary thyroid carcinoma (PTC) in mouse models [120].

PI3K and MAPK regulation of HIFs

Activation of the PI3K and the mitogen-activated protein kinase (MAPK) signaling pathways promotes an increase in HIF-1α protein translation [121, 122]. PI3K activation induces downstream AKT and mTOR leading to the phosphorylation of the eukaryotic translation initiation factor 4E (eIF-4E) binding protein (4E-BP1), a protein translation repressor. Phosphorylated 4E-BP1 lacks the ability to form a complex with eIF-4E [123] and increases HIF-1α protein translation [124] potentially by increasing cap-dependent mRNA initiation. Another study shows that calcitriol reduces HIF-1/2α protein levels by inhibiting AKT and eIF-4E, which indicates a potential mechanism of action that eIF-4E regulates both HIF-1α and HIF-2α protein translation [125]. Moreover, activation of mTOR promotes phosphorylation of p70 ribosomal protein S6 kinase (p70S6K), which subsequently phosphorylates the ribosomal protein S6 (rpS6) and enhances HIF-1α translation [126]. mTOR forms two distinct protein complexes, mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2), which regulate different cellular processes. In VHL-deficient RCC cell lines, HIF-1α regulation is dependent on both mTORC1 and mTORC2, while HIF-2α is dependent only on mTORC2 [127]. Likewise, insulin-like growth factor (IGF) has been reported to drive HIF-2α expression in hypoxic neuroblastoma cells via insulin-like growth factor receptor (IGFR)-PI3K-mTORC2 signaling [128]. Similar to the PI3K pathway, the induction of the RAS-RAF-MEK-ERK kinase cascade leads to 4E-BP1 and p70S6K phosphorylation thus enhancing HIF-1α translation [129, 130]. In addition to IGF, a variety of growth factors including vascular endothelial growth factor (VEGF), heregulin, and androgens, regulate HIF-1/2α protein translation through PI3K and MAPK signaling pathways [131,132,133].

The MAPK/ERK pathway can also lead to increasing transcriptional activity of HIF-1α by either phosphorylating the co-activator p300/CBP or directly phosphorylating the C-terminal domain of HIF-1α at serine 641 and serine 643 [130, 134]. Similarly, HIF-2α is phosphorylated by ERK1/2 at serine 672 which enables its chromosomal region maintenance 1 (CRM1)-mediated nuclear shuttling, increasing HIF-2α transcriptional activity [135]. Reptin52 binds to the HIF-2α protein in the absence of ERK phosphorylation. Therefore, the inhibition of ERK1/2 phosphorylation promotes HIF-2-Reptin52 binding reducing HIF-2α protein stability thus reducing HIF-2α activity [136]. Interestingly, Reptin is methylated by the methyltransferase G9a under hypoxia, which results in Reptin binding to the promoters of a subset of hypoxia-responsive genes, and consequently suppressing HIF-1α transcriptional activity [137].

Apart from PI3K and MAPK, cAMP-dependent protein kinase A (PKA) phosphorylates Thr63 and Ser692 on HIF-1α, thus inhibiting proteasomal degradation of HIF-1α which enhances its transcriptional activity [138]. Casein kinase 1 (CK1) belongs to the casein kinase family of serine/threonine-selective enzymes [139]. CK1δ has been reported to phosphorylate serine 247 in the N-terminal PAS domain of HIF-1α which inhibits HIF-1α activity without affecting its stability or nuclear accumulation [140, 141]. On the contrary, CK1δ promotes HIF-2α nuclear accumulation by phosphorylating HIF-2α at Ser383 and Thr528 which blocks the CRM1-dependent export of HIF-2α from the nucleus [142]. HIF-2α is phosphorylated at Thr844 by casein kinase 2 (CK2), and as a result, its transcriptional activity is increased in an O2-independent manner [143].

Hsp90 and RACK-dependent regulation of HIFs

HIF-1α degradation is regulated through a pVHL-independent pathway involving heat shock protein 90 (Hsp90) and the receptor for activated protein C kinase 1 (RACK1). Initial findings demonstrated that HIF-1α associates in vitro with the molecular chaperone, Hsp90 [144]. Further studies demonstrate that Hsp90-HIF-1α binding leads to enhanced coupling with HIF-1β to promote HIF-1 transactivation [145]. RACK1 competes with Hsp90 for binding to the HIF-1α PAS domain, which promotes ubiquitination and degradation of HIF-1α [146, 147]. Therefore, the Hsp90 inhibitor 17-(allylamino)-17-demethoxygeldanamycin (17AAG) reduces the levels of HIF-1α in an oxygen-independent manner [148]. Co-immunoprecipitation studies showed that HIF-2α also interacts with Hsp90 [149]. Hsp90-deficient embryonic stem cells (ESCs) showed a delayed accumulation of HIF-1α in response to hypoxia exposure as compared to wild-type ES cells. However, heat shock protein 70 (Hsp70) was identified as a HIF-1α-interacting protein that selectively mediates HIF-1α degradation by recruiting the ubiquitin ligase carboxyl terminus of Hsp70-interacting protein (CHIP), with no effect on HIF-2α stability [150, 151].

Inflammatory cytokine regulation of HIFs

Experimental data support the concept that inflammatory cytokines directly regulate the HIF pathway. Hellwig-Burgel T, et al. first described the involvement of proinflammatory cytokines interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF-α) in oxygen-independent HIF-1 regulation [152]. They demonstrated that both IL-1β and TNF-α elevate HIF-1 activity by inducing HIF-1 DNA binding. Moreover, IL-1β increases HIF-1α protein levels in human hepatoma cells [152]. Both the PI3K and MAPK pathways are required for IL-1β’s regulation of HIF-1. For example, LY294002, a PI3K inhibitor, inhibits IL-1β-induced HIF-1α activation in a dose-dependent manner [153]. Likewise, Qian D, et al. reported that IL-1β prompts ERK phosphorylation, leading to increased HIF-1α protein expression, which can be inhibited by treating with the ERK inhibitor, PD98059 [154]. IL-1β induces HIF-1α accumulation in a highly invasive human breast cancer cell line, MDA-MB-231, as well as in mouse models [155]. A study in renal cancer suggested that IL-1β could facilitate RCC metastasis through AKT/p65/HIF-2α activation [156]. Haddad J, et al. determined that TNF-α increases ROS production in alveolar epithelial cells thereby promoting HIF-1α translocation into the nucleus, and finally activating downstream target genes [157]. More recent studies show that TNF-α upregulates HIF-1α mRNA and protein levels via the NF-κB pathway [158, 159].

It’s important to note that these non-canonical regulatory mechanisms are often interconnected with the canonical regulation pathways and can contribute to the overall regulation of HIFs in a context-dependent manner. Additionally, the specific details of these pathways may vary depending on the cell type and environmental conditions.

Targeting HIF-1 and HIF-2

Considering the pleiotropic effects of HIF-1/2α in tumor progression, metastasis, and therapeutic resistance, there has been tremendous interest in developing inhibitors targeting this pathway as potential anti-cancer agents. According to their mechanism of action, HIF-1/2α inhibitors can be divided into six major groups that modulate: i) HIF-1/2α transcriptional activity, ii) HIF-1/2α mRNA expression, iii) HIF-1/2α protein translation, iv) HIF-1/2α DNA binding, v) HIF-1/2α stabilization, vi) HIF-2α/HIF-1β binding. Here, we describe small molecule HIF inhibitors that have been recently reported or that have been under clinical investigation over the last 5 years (Table 1). For HIF inhibitors that have been reported on prior to 2018 please see DiGiacomo JW, et al. [160].

Table 1 Newly reported small molecule HIF inhibitors
Full size table

HIF-1/2 inhibitors

Inhibitors of HIF transcriptional activity

Blocking HIF-1α binding to co-activator proteins, including p300 and CBP, represents a potential mechanism in which small molecules can impair HIF-1α activity. Chetomin, a natural product isolated from Chaetomium species, was found to inhibit the interaction of HIF-1α and HIF-2α with p300 by binding to the cysteine–histidine-rich domain 1 (CH1) domain of p300 and disrupting its structure [161]. A recent study demonstrated that Chetomin blocks Hsp90 binding to HIF-1α, thus markedly suppressing tumor formation in a spontaneous KrasLA1 lung cancer model, H1299 cell-derived flank xenograft model, and an in vivo tumor-initiating assay [162]. Cardenolides, isolated from a medicinal plant C. gigantea, show strong HIF-1α inhibitory activity and a potent cytotoxic effect in a human breast cancer cell line (MCF-7 cells) [163]. Six new non-classical cardenolides have been developed and most effectively suppress HIF-1α transcriptional activity [164]. Recently, Xue Z, et al. developed a series of benzo[d]isoxazole derivatives that exhibit strong inhibitory activities against HIF-1α transcription, among which, six compounds showed IC50 values below 100 nM in HEK 293 T cells [165]. The effects of cardenolides and benzo[d]isoxazole derivatives on HIF-2α have not been tested to date.

Inhibitors of HIF mRNA expression

A synthetic anti-sense oligonucleotide EZN-2968 that specifically binds and inhibits the expression of HIF-1α mRNA inhibits cancer cell growth under normal and hypoxic conditions in vitro. EZN-2968 also partially suppresses tumor growth in nude mice bearing xenografts of human DU145 prostate cancer cells when administered intraperitoneally (i.p.) twice weekly for 5 wk [166]. This study also suggested that EZN-2968 has no significant effect on HIF-2α mRNA levels in vitro and only a weak effect on HIF-2α higher doses of EZN-2968 in vivo. EZN-2968 has been evaluated in two phase I clinical trials in lymphoma and advanced solid tumors in patients with liver metastasis. The results show that EZN-2968 is well tolerated in patients with advanced malignancies and a reduction in HIF-1α mRNA was observed in 4 of 6 patients with paired pre- and post-treatment tumor biopsies [167, 168]. A third phase I trial for EZN-2968 was conducted in patients with hepatocellular carcinoma but it terminated early because the primary endpoint (a reduction in HIF-1α mRNA expression level change after 1 cycle of treatment was not met) [169]. Like EZN-2968, a phytochemical celastrol (tripterine) was shown to significantly reduce the level of HIF-1α mRNA in cancer cells. It also inhibits the hypoxia-induced accumulation of nuclear HIF-1α protein under both normal and hypoxic conditions [170]. However, another study using human hepatoma cells indicates that celastrol inhibits HIF-1α protein synthesis without affecting its mRNA levels. In vivo studies confirmed the inhibitory effect of celastrol on HIF-1α with an accompanying decrease in tumor growth in a xenograft model established with HEP3B cells [171]. Compound C6, derived from modifications to the C-29 carboxyl group of celastrol, exerted a higher inhibition rate (74.03%) compared to 5-fluorouracil treatment (59.58% inhibition rate) in a mouse tumor xenograft model, with little toxicity [172]. Toxicity data in preclinical studies are promising, nevertheless, clinical studies have not been pursued to date [173,174,175]. The effects of celastrol on HIF-2α have not been tested to date.

Inhibitors of HIF-1 protein translation

Inhibitors of topoisomerase I and II, including Camptothecin (CPT) analogs, Topotecan (Hycamtins–TPT), and EZN-2208 (an active metabolite of irinotecan), have been found to prevent HIF-1α accumulation and decrease HIF-2α protein levels in cell lines and xenograft mouse models [176,177,178,179]. In mouse models of chronic lymphocytic leukemia (CLL), EZN-2208 was shown to improve the response to fludarabine [180]. Six phase I/phase II clinical trials have been conducted to study EZN-2208 in different solid tumors including metastatic breast cancer (NCT01036113) and metastatic colorectal cancer (NCT00931840). The studies have demonstrated that ENZ-2208 is well tolerated in most patients. The objective response rate (RR) and clinical benefit rate (CBR) were 22.5% and 36.7% among patients with metastatic TNBC [181]. However, in patients with advanced colorectal cancer, EZN-2208 in combination with cetuximab did not lead to a survival advantage compared to irinotecan plus cetuximab [182]. The clinical trials did not test the effect of EZN-2208 treatment on HIF signaling or expression in patient samples. CRLX101 is a nanoparticle comprised of CPT, pendant carboxylic acid groups, and linear, cyclodextrin-polyethylene glycol (CD-PEG) blocks. In a model of subcutaneously implanted A2780 human ovarian cells in nude mice, it reduces HIF-1α and HIF-2α protein expression. [183]. In a phase II clinical trial, CRLX101 demonstrated acceptable tolerability when used as a single agent or in combination with bevacizumab in patients with ovarian cancer. CRLX101, as a single agent led to an overall response rate (ORR) of 11%. The addition of bevacizumab increases the ORR to 18% [184]. Clinical trials that have focused on CRLX101 did not test the HIF-1 or HIF-2 protein levels or activity in patient samples.

HIF-1 DNA binding inhibitors

Originally isolated from Streptomyces echinatus, echinomycin binds to the core of the HIF-1/2α recognition sequence (5′–CGTG-3′), thus inhibiting the binding of HIF-1α/HIF-2α to cognate HREs [185]. In 2020, Bailey C, et al. reported that liposomal-echinomycin significantly inhibits primary tumor growth and metastasis in MDA-MB-231 and SUM-159 xenograft models [186]. Anthracyclines are widely used chemotherapeutic agents that can prevent HIF function. Several anthracyclines including Doxorubicin, Daunorubicin, Epirubicin, and Idarubicin have been demonstrated to disrupt HIF-1α and HIF-2α binding to DNA [187,188,189]. The daily injection of anthracyclines at low doses in tumor-bearing mice results in the inhibition of HIF-1α target gene expression and angiogenesis.[190].

Inhibitors of HIF-1 stability

An active component isolated from Angelica gigas roots, decursin, can regulate HIF-1α protein stability and promote its degradation, by increasing oxygen-dependent hydroxylation and ubiquitination [191]. However, the effects of decursin on HIF-2α have not been tested. Another HIF inhibitor 32-134D has been developed and effectively inhibits both HIF-1 and HIF-2 activity by promoting the degradation of HIF-1α and HIF-2α protein in HCC cells [192]. Previous studies in our lab demonstrated that mebendazole (MBZ), an FDA-approved anti-parasite drug, can inhibit the transcriptional activity of HIFs in breast cancer cell lines by decreasing HIF-1α, HIF-2α, and HIF-1β protein expression without affecting HIF mRNA levels [193].

HIF-2 specific inhibitors

To date, HIF-2α-specific inhibitors have progressed the farthest in clinical trials. Most exert their function by preventing HIF-2α heterodimerization with HIF-1β.

HIF-2 to HIF-1β DNA binding inhibitors

PT2385 (2,3-dihydro-1H-inden-4-yl-oxy derivative) is a HIF-2α-selective inhibitor currently under phase II clinical trials. This selective and orally active small-molecule inhibitor blocks HIF-2α dimerization with HIF-1β by binding the lipophilic cavity of the HIF-2α PAS-B domain [194]. PT2385 treatment reduces the expression of HIF-2α downstream gene products, including VEGF-A, PAI-1, and cyclin D1, and subsequently causes dramatic tumor regression in ccRCC [195]. Notably, preclinical studies showed favorable pharmacokinetic properties and a good safety profile of PT2385 with no adverse effect on cardiovascular performance. In phase I clinical trial (NCT02293980), PT2385 inhibited HIF-2α in ccRCC but also in normal tissue as measured by a reduction of EPO levels [196]. This resulted in anemia as one of the most frequent adverse events. Fortunately, the anemia was predominantly low-grade (grade 1 or 2, 35%; grade 3, 10%; no grade 4), and no patients discontinued treatment because of adverse events. To improve the pharmacokinetic profile of PT2385, a second-generation HIF-2α inhibitor PT2977 has been developed, and later renamed MK-6482 or belzutifan [195]. The first-in-human phase I clinical trial (NCT02974738) of PT2977 showed encouraging outcomes in patients with advanced ccRCC at a dose of 120 mg once daily [197]. In the phase II trial, MK-6482–004 (NCT03401788), patients with germline VHL mutations that had non-metastatic RCC, CNS hemangioblastomas, or pancreatic neuroendocrine tumors (pNETS) were treated with 120 mg of PT2977, once daily [198]. The median follow-up was 37.8 months. The ORR was 64% (39 out of 61) for RCC, 44% (22 out of 50) for CNS hemangioblastomas, and 91% (20 out of 22) for pNETS [199]. A randomized open-label phase III of PT2977 versus everolimus (NCT04195750) has been conducted in advanced ccRCC, which shows the progression-free survival (PFS) and ORR were superior with PT2977 vs everolimus [200]. In August 2021, the FDA approved PT2799 (belzutifan; WELIREG™) as the first HIF-2α inhibitor for the treatment of VHL disease as a therapy for RCC, CNS hemangioblastomas, or pNETS, that do not require an immediate surgery [201]. Meanwhile, work continues to develop even more specific inhibitors for HIF-2α. For example, YQ-0629 interacts with the PAS-B pocket of the HIF-2α protein and prevents its dimerization with HIF-1β thus abolishing a hypoxia-induced stem-like phenotype in breast cancer with a Kd of 57.5 uM [202]. Another small molecule, AB521, has recently been developed as a highly potent and selective HIF-2α inhibitor that avidly binds to the HIF-2α PAS-B domain, thereby inhibiting its heterodimerization with HIF-1β [203]. In preclinical experiments, AB521 causes a significant regression of the ccRCC tumors xenografts in mice. This orally available inhibitor is being tested in a phase I clinical trial for safety and pharmacokinetics as a monotherapy in patients with solid tumors and will be assessed for clinical efficacy in patients with ccRCC that have been previously treated (NCT05536141) [204].

Conclusions and perspectives

Hypoxia alters the biology of cancer cells within a solid tumor to promote angiogenesis, metastasis, and resistance to chemotherapy and radiotherapy. Numerous studies have provided convincing evidence that hypoxia-induced cellular adaptations are mediated via both HIF-1α and HIF-2α downstream signaling events. HIF-1α and HIF-2α have structural similarities and both undergo oxygen-dependent regulation. They have similar but also unique downstream target genes. This can be explained in part by different chromatin binding preferences for each HIF subunit. Both HIFs contribute to the formation of new blood vessels and subsequent invasion, with HIF-1α primarily responsible for directly activating pro-survival genes [38,39,40,41]. On the other hand, HIF-2α has additional roles in modulating vascular endothelial cells and macrophages in the TME [42, 43]. It is worth noting that both HIF-1 and HIF-2 have been shown to participate in various processes of tumor metabolism, angiogenesis, and metastasis. Due to space limitations, we will not elaborate on these aspects in detail here.

There has been an extensive array of inhibitors that have been documented for their ability to decrease HIF-1/2 activation in preclinical studies. Many of the inhibitors have also been shown to be efficacious in preclinical models for the treatment of cancer. Whether or not the reduction in HIF expression or activity is causative for the inhibitors’ efficacy is not always clear. In fact, most clinical trials testing the inhibitors described in this review do not include a pre-treatment or post-treatment measurement of HIF-1/2 levels or activity (see Table 1). If HIF-1/2 levels or activity have been considered as part of a clinical trial protocol, they have not been evaluated together or simultaneously stained in the same tissue sample. More recently, the ability to perform multiplexing in order to image multiple targets proteins of interest in the same tissue section has become more commonplace. In 2024, we developed a multiplex staining protocol for HIF-1 and HIF-2 that could be employed in the future to evaluate the expression of HIF-1 and -2 levels in the same FFPE tissue section [205]. In the future, this approach could also be useful for stratifying which patients would benefit from a HIF-1 versus a HIF-2 inhibitor. In our studies, multiplexed HIF-1 and HIF-2 staining in the biopsies of 15 patients diagnosed with TNBC revealed that HIF-1 and HIF-2 expression is colocalized in the same cell less than 50% of the time (unpublished observations). This is in line with recent studies discussed in this review that demonstrate that although several stimuli including hypoxia promote the expression of both HIF-1 and HIF-2, there are pathways that regulate HIF-1 and HIF-2 independently. A recent review summarized experimental data indicating that HIF-1 and HIF-2 play complementary roles in many cancers, which suggests that inhibiting both HIF-1 and HIF-2 could offer greater therapeutic benefits compared to targeting either factor alone [206]. HIF-2α specific inhibitors have been quickly progressing through clinical trials, and act by inhibiting the dimerization between HIF-2α and HIF-1β. Given the promising results with HIF-2α, HIF-1α inhibitors are likely to be next in line.

It is also important to note that when HIF-1α or HIF-2α expression has been evaluated in tumor biopsies, some samples show a pattern of HIF expression that is likely due to hypoxia. For example, there is a gradient in HIF-1α expression with cells located furthest from a blood vessel showing high HIF-1α expression and those farther from a blood vessel with less HIF expression. Other biopsies show patterns of HIF-1α or -2α expression that are independent of O2 gradients, suggesting that O2-independent mechanisms may be responsible for increased expression [207]. Knowing the stimulus that leads to HIF-1/2 α overexpression will be beneficial in designing an effective therapeutic and to have a better arsenal against therapeutic resistance. For example, if aberrant expression of tyrosine kinases such as PI3K/AKT and/or MAP kinases cause HIF-overexpression, inhibitors targeting these specific pathways may be a promising treatment option.

Given the recent success of immunotherapies, there has been a growing interest in determining the role of HIF-1/2α expression in the immune microenvironment. The cross-talk between HIF and PD-1/PD-L1 pathways has become well recognized [208,209,210]. A preclinical study showed that the combination of HIF-1α inhibition and immune checkpoint inhibitors can significantly improve tumor regression in the setting of non-small cell lung cancer [211]. Furthermore, a phase 3 clinical trial (NCT05239728) began in 2022 and aims to access the efficacy and safety of the combination of the HIF-2α-specific inhibitor, belzutifan, and PD-1 inhibitor, pembrolizumab [212]. The results of these trials will indicate whether immunotherapies combined with HIF inhibitors will be beneficial as a novel cancer therapy.

In general, gaining a more profound insight into the biology of both HIF-1α and HIF-2α will be crucial for providing additional treatment strategies for cancer. Future work should focus on the development of more specific and selective HIF-1/2α inhibitors.