Non-canonical approaches to targeting hypoxic tumors

Abstract

Hypoxia is a common characteristic in solid cancers. Hypoxia-inducible factors (HIFs) are involved in various aspects of cancer, such as angiogenesis, metastasis and therapy resistance. Targeting the HIF pathway has been regarded as a challenging but promising strategy in cancer treatment with recent FDA approval of a HIF2α-inhibitor. During the past several decades, numerous efforts have been made to understand how HIFs participate in cancer development and progression along with how HIF signaling can be modulated to achieve anti-cancer effect. In this chapter, we will provide an overview of the role of hypoxia and HIFs in cancer, summarize the oxygen-dependent and independent mechanisms of HIF-1α regulation, and discuss emerging approaches targeting hypoxia and HIF signaling which possess therapeutic potential in cancer. We will emphasize on two signaling pathways, involving cyclin-dependent kinases (CDKs) and heat shock protein 90 (HSP90), which contribute to HIF-1α (and HIF-2α) stabilization in an oxygen-independent manner. Through reviewing their participation in malignant progression and the potential targeting strategies, we discuss the non-canonical approaches to target HIF signaling in cancer therapy.

Hypoxia in cancer

Angiogenesis is among the key hallmarks of cancer. Despite the intensive vascularization, blood supply is often inadequate in tumors because of the abnormal blood vessel formation with the unbalanced level between pro-angiogenic and anti-angiogenic molecules. The blood vessels can be leaky, twisted, and lacking of the regular structures composed of arterioles, capillaries and venules. Meanwhile in tumors, the diffusion radius of oxygen is limited to 100 μm away from blood vessels [1]. The cancer cells located at a further distance become deprived of oxygen, due to a combination of excessive oxygen consumption during the rapid cell division and the disorganized tumor-associated vasculature, resulting in intratumoral hypoxia (Figure 1). More than half of the locally advanced solid tumors contain diversely distributed hypoxic regions [2], which contributes to the heterogeneity within the tumor mass.

Figure 1.

Intra-tumoral hypoxia and HIF signaling is implicated in most of the cancer hallmarks. Oxygen concentration is low at regions within a tumor far away from the blood supply, which is termed as intra-tumoral hypoxia. The tumor cells survive the hypoxic tensions by inducing adaptive responses mostly mediated by hypoxia-inducible factors. Hypoxia/HIF signaling plays a role in various hallmarks of cancer. (Picture created at Biorender.com).

In adaptation to hypoxia, a coordinated set of cellular responses is triggered. The main mediator in such response is the hypoxia-inducible factor (HIF). Hypoxia/HIF signaling affects a majority of the cancer hallmarks (Figure 1). It selects for the cell population with defective apoptotic capabilities (with p53-deficiency) [3]. It downregulates proapoptotic molecules (e.g. Bid & Bax) to promote survival [4]. It switches the energetics (e.g. increasing the expression of GLUT1) [5]. It stimulates angiogenesis [6] (e.g. induction of VEGF). It contributes to invasion [7] and metastasis [8]. It decreases DNA repair and increases genomic instability [9]. Notably, it also suppresses immune reactivity (e.g. driving the expression of PD-L1) [10,11]. In clinic, it is linked with chemo- and radio-therapy resistance as well as aggressive disease and poor patient prognosis [12-14].

Given the involvement of hypoxia/HIF in many aspects of the cancer hallmarks and its association with more aggressive and resistant disease, hypoxic environment serves as an appealing target in cancer treatment. One potential approach to target hypoxia in cancer is through the bioreductive prodrugs based on certain chemical moieties (e.g. nitro groups) that can be metabolized in low-oxygen conditions [15]. These kinds of prodrugs undergo one-electron reduction by cellular reductases. The generated prodrug radical can be re-oxidized by oxygen in normoxic condition but is converted to the active drug form in hypoxia, by which way the treatment selects particularly for hypoxic cells. However, challenges exist with this concept, such as the need for determining biomarkers in patient selection and improving the pharmacokinetic properties (e.g. delivery into cells at the targeted site) of the prodrug. A main limitation of this approach is that conventionally the active drugs being released are DNA-damaging cytotoxins, which contributes to off-target toxicity as well as eliminates the possibility of combining with traditional chemotherapy. Replacing the active drug with protein components may provide new options [16].

Hypoxia-inducible factors

Hypoxia-inducible factors (HIFs) are the major mediators in hypoxia-induced cellular responses. The discovery of HIF dates back to the early 90s during the study of an enhancer upstream to the erythropoietin (EPO) gene, which identified a nuclear factor that activates the transcription of EPO in response to hypoxia, termed as hypoxia-inducible factor 1 (HIF-1) [17,18]. Subsequent purification and characterization of the protein have shown that HIF-1 is composed of two separate subunits: HIF-1α and a smaller HIF-1β (Figure 2A) [19], both of which harbor a basic-helix-loop-helix-PAS (bHLH-PAS) structure [20]. HIF-1β is also named aryl hydrocarbon nuclear translocator (ARNT), since it was discovered as a heterodimer with aryl hydrocarbon receptor (AhR) before the identification of HIF-1α. As a transcription factor, HIF-1 acts by binding to the hypoxia-responsive element (HRE) in promoters and stimulates the transcription of its target genes. The base domain in the HIF subunits has DNA binding capabilities while the HLH domain is responsible for the protein dimerization. The core motif of HRE is a consensus sequence of 5’-(A/G)CGTG-3’. HIF-1α subunit contains two transactivation domains (TAD), referred to as N-TAD (N-terminal TAD) and C-TAD (C-terminal TAD) (Figure 2B). Hypoxia allows the interaction between C-TAD and co-activators p300/CBP, which recruits the transcriptional machinery and thus increases gene expression [21,22]. N-TAD is also involved with transcriptional activity and is overlapping with the oxygen-dependent degradation domain (ODDD) which is essential for the protein stability regulation of HIF by oxygen availability. Interestingly, N-TAD is found to modulate the specificity of target gene selection between HIF-1α and HIF-2α [23].

Figure 2.

Hypoxia-inducible factor (HIF) subunits. (A) The heterodimer of HIF-1α and HIF-1β binds to the HRE region in target DNA to elicit HIF-1 signaling. (B) Comparison of domains in HIF subunits. (A is drawn in Bioblender. B is drawn with reference to Linda Ravenna, et al. 2015 with permission from John Wiley and Sons).

The second main HIF, HIF-2α, was identified later in 1997 [24-27]. It is also known as endothelial PAS domain protein-1 (EPAS1). HIF-2α shares a similar structure composition as HIF-1α. HIF-2α also contains an ODDD domain. The sequence similarity in their DNA binding domains is 83% [23], while the similarity in the dimerization domains is 70% [23]. Their C-TAD is also comparable (67%) [23]. Regions surrounding their oxygen-dependent regulatory sites are alike as well (70%) [23]. This sequence homology is in line with the observation that HIF-2α heterodimerizes with HIF-1β and recognizes the same HRE. However, while HIF-1α serves as the “master regulator” [28] in maintenance of the oxygen homeostasis, HIF-2α, although can be induced under hypoxia [29], seems to play distinct roles [30] and show specific tissue distributions (e.g. expressed in endothelial cells). While HIF-1α and HIF-2α have shared transcription activities, they each target unique genes as well [31-34].

Another HIF-α subunit, HIF-3α, unlike HIF-1α or 2α, carries a LZIP (leucine zipper) domain instead of the C-TAD domain at the C-terminal. It exists as several different variants according to the involvement of different promoters, transcription start sites, and splicing patterns. The full-length HIF-3α contains the ODDD domain and can exert oxygen-regulated transcription activities [35]. Some truncated variants lacking ODDD can act as a dominant negative inhibitor on HIF-1α/2α. Other variants are able to dampen the HIF-1α/2α signaling through competitive binding with HIF-1β. Further investigation is needed to unravel the complexity of HIF-3α.

HIF-1α regulation

Oxygen-dependent regulation of HIF-1α

Regulation of stability

All HIF-α subunits containing the ODDD domain go through post-translational regulation in an oxygen-dependent manner, while HIF-1β lacks the ODDD domain and is constitutively expressed. A group of prolyl hydroxylases, named as prolyl hydroxylase-domain protein (PHD) or HIF-1 prolyl hydroxylases (HPH), rely on molecular oxygen (as well as the activating metal iron, the co-substrate α-ketoglutarate, and the co-factor ascorbate) to modify the conserved proline residues at the LXXLAP motif in the ODDD domain [36-41]. In HIF-1α, for example, two proline residues (P402 & P564) are subject to such hydroxylation under normoxia, which allows the interaction with the von Hippel Lindau (VHL) protein. VHL is the substrate recognition component within an E3 ubiquitin ligase complex and is known as a tumor suppressor. The VHL complex (VHL together with the adaptor protein Elongin B/Elongin C, the scaffold protein Cullin-2, and the RING protein Rbx1) catalyzes polyubiquitination on HIF-1α, which results in protein degradation by the proteasome [42-44]. Under hypoxia, insufficient oxygen amount limits the PHD hydroxylation activity and hence prevents the VHL-dependent ubiquitination and degradation (Figure 3).

Figure 3.

Regulation of the HIF-1α expression in an oxygen-dependent and independent manner. In normoxia, HIF-1α is hydroxylated and subsequently targeted by VHL for ubiquitination, resulting in proteasomal degradation (middle). In hypoxia, hydroxylation of HIF-1α is inhibited, which allows it to translocate into the nucleus and activate target gene transcription in a heterodimer with HIF-1β (left). Alternatively, HIF-1α is regulated through VHL-independent mechanisms at transcriptional, translational and post-translational levels. In this regard, HIF-1α stabilization can be achieved even in the presence of oxygen. (The three icons of transcription, translation and stabilization are adapted from https://innovativegenomics.org/glossary/).

Additionally, a lysine residue, K532, located in the ODDD region of HIF-1α, is an acetylation target site of the arrest-defective-1 (ARD-1) acetyltransferase. The acetylation of K532 by ARD-1 enhances the binding of VHL to HIF-1α, and thus contributes to HIF-1α ubiquitination and degradation [45]. The finding that K532R mutation stabilizes HIF-1α in normoxia further supports the role of K532 acetylation in the regulation of HIF-1α [45]. ARD-1 does not require oxygen as a substrate for its function, however, hypoxic conditions downregulate the expression of ARD-1 at the transcriptional level [45].

Regulation of activity

Other than stability regulation, HIF-1α transactivation activity can be regulated by oxygen availability through its C-TAD domain. Asparagine residue N803 in the C-TAD domain of HIF-1α can be hydroxylated by an enzyme called factor inhibiting HIF (FIH) [46-48]. Modification of this residue is critical because it sits within the molecular interface in the binding between HIF-1α and CBP [49]. FIH also senses oxygen concentration in a similar way as PHD. However, it is noteworthy that the Km value of FIH for O₂ is three times lower than those of PHD hydroxylases, which indicates that such regulation of HIF-1α signaling is still able to occur even when HIF-1α is already stabilized by the PHD activity inhibition.

Oxygen-independent regulation of HIF-1α

Aside from the oxygen-sensing mechanisms, HIF-1α is as well-regulated at multiple levels in an oxygen-independent manner.

Regulation of HIF-1α transcription

At the transcriptional level, HIF1A (the gene encoding HIF-1α) is upregulated through the JAK-STAT3 pathway [50,51]. Pro-inflammatory cytokine interleukin-6 (IL-6) induces the transcription of HIF-1α in lymphocytes [52] and MCF7 breast cancer cells [50], which requires the participation of STAT3. Moreover, HIF1A transcription is stimulated by the NF-κB pathway in myeloid cells [53]. The endotoxin lipopolysaccharide (LPS) increases the expression of HIF-1α mRNA [54] through toll-like receptor (TLR)-induced NF-κB activity [55]. Furthermore, the reactive oxygen species (ROS) increases the HIF1A transcription under both normoxic and hypoxic conditions through the extracellular signal-regulated kinase (ERK) and phosphatidyl inositol 3-kinase (PI3K) pathways [56,57].

Regulation of HIF-1α translation

The PI3K pathway is also involved in the translation of HIF-1α. PI3K acts in a signaling cascade including its downstream target, protein kinase B (Akt), and a subsequent factor, mammalian target of rapamycin (mTOR) [58]. Activated PI3K-Akt-mTOR signaling is then transduced to mRNA translation activity. For example, mTOR releases the eukaryotic translation initiation factor 4E (eIF-4E) from its binding protein (4E-BP1) through 4E-BP1 phosphorylation, which allows the cap-dependent translation of mRNA. Alternatively, mTOR can activate the p70 S6 Kinase (S6K) which phosphorylates ribosomal protein S6, resulting in mRNA translation. The PI3K-Akt pathway is exploited by certain hormones (e.g. Insulin [59]) and growth factor signaling (e.g. human epidermal growth factor receptor 2 (HER2) [60]) to induce the translation of HIF1A and thus increases the expression of HIF-1α protein.

Regulation of HIF-1α stabilization

As in the canonical regulatory mechanism, targeting HIF-1α protein stability plays an essential part in the non-canonical HIF-1α regulation.

The HIF-1α-targeting PHD hydroxylases require oxygen as well as other co-factors such as ascorbate for their enzymatic activity. Exposure to Cobalt (II) has been found to deplete ascorbate and increase HIF-1α expression without the requirement for a hypoxic environment [61]. As a matter of fact, Cobalt chloride (CoCl₂) has been used in multiple circumstances as a chemical inducer to elevate the level of HIF-1α.

The VHL-dependent ubiquitination of HIF-1α is a downstream event to the oxygen-caused hydroxylation. Therefore, defective VHL function leads to the accumulation of HIF-1α regardless of oxygen concentration. For instance, the VHL protein is a substrate of WD repeat and SOCS box-containing protein 1 (WSB1) for ubiquitination and degradation, which consequently facilitates HIF-1α stabilization and promotes cancer cell migration and invasion [62]. Genetic deletion or mutations of the VHL gene leads to constantly stabilized HIF-1α (as well as HIF-2α). A clinical condition with VHL gene alterations is known as the VHL syndrome, which is related to human cancers such as hemangioblastomas (affecting brain, spinal cord, and retina) and clear cell renal cell carcinoma (ccRCC) [63].

In addition to VHL, several other E3 ubiquitin ligases have been revealed in HIF-1α regulation. The hypoxia-associated factor (HAF) ubiquitinates HIF-1α but not HIF-2α independently of oxygen [64]. The carboxy terminus of Hsp70-interacting protein (CHIP) ubiquitinates HIF-1α in high-glucose conditions with aid of Hsp40/Hsp70 [65]. The mouse double minute 2 homolog (MDM2) is recruited to ubiquitinate HIF-1α through interaction with the tumor suppressor p53 [66]. The receptor of activated protein kinase C (RACK1) acts both similarly as VHL to ubiquitinate HIF-1α by recruiting elongin-C and promote proteasomal degradation and distinctly from VHL in that it targets unhydroxylated HIF-1α [67]. The F-box and WD protein Fbw7 (FBW7) ubiquitinates the glycogen synthase kinase 3 (GSK-3)-phosphorylated HIF-1α [68-70]. The X-linked inhibitor of apoptosis (XIAP) also serves as an E3 ubiquitin ligase that targets HIF-1α for Lys63-linked polyubiquitination [71].

Other than proteasomal degradation, HIF-1α can be digested in the lysosome. The co-localization of HIF-1α and the lysosome-associated membrane protein type 2A (LAMP2A) was observed in RCC4 VHL-deficient clear cell renal cancer cells [72]. The heat shock cognate 70-kDa protein (HSC70) and LAMP2A bind to HIF-1α and escort it to lysosome via chaperone-mediated autophagy (CMA).

Another chaperone protein, the heat shock protein 90 (HSP90) promotes the stabilization of HIF-1α instead. HSP90 has been acknowledged to interact with HIF-1α since 1996 [73]. The binding is suggested to induce a conformational change that assists in the dimerization of HIF-1α and HIF-1β [74]. Inhibition of HSP90 by geldanamycin (GA) accelerates the destabilization rate of HIF-1α in UMRC2 ccRCC cells that lack a functional VHL [75], which indicates the stabilization of HIF-1α by HSP90 is VHL-independent. It is proposed that HSP90 stabilizes HIF-1α through competitively binding to it against RACK1, preventing the recruitment of elongin-C and subsequent ubiquitination [76].

Our lab unraveled a previously unrecognized mechanism by which CDK1 stabilizes HIF-1α through phosphorylation at its Serine 668 residue [77]. Knockdown or inhibition of CDK1 decreases the expression of HIF-1α in hypoxia. Overexpression of CDK1 or cyclin B1 raises the level of HIF-1α in normoxia, which is reversed by CDK1 inhibitor, Ro3306. The interaction of CDK1 and HIF-1α phosphorylates Ser668 which is located within a sequence (RTASPNR) that contains the CDK1 consensus motif pS/T-P-x-R. Point mutation S668E enhances the stabilization of HIF-1α while S668A inhibits HIF-1α expression, indicating the Ser668 phosphorylation plays an important role in HIF-1α stability. In consistence with the presence of CDK1 activity throughout cell cycle, HIF-1α expression is increased after 8 hours upon release from the synchronization at S phase, at which point most of the cells are in G2/M phase. In this study, CDK4 knockdown is also shown to inhibit HIF-1α expression.

Recently, we have shown a novel VHL-independent mechanism of HIF-1α regulation, which involves the SMAD specific E3 ubiquitin protein ligase 2 (Smurf2) [78]. Smurf2 is a E3 ubiquitin ligase that is known to have a classic substrate of Smad 2/3. We identified Smurf2 in the overlap of a HIF-1α-interacting protein mass spectrometry upon CDK4/6 inhibition and a bioinformatic prediction of HIF-1α-targeting E3 ubiquitin ligases. Smurf2 targets HIF-1α for ubiquitination and destabilization in vitro. The study provides additional components to HIF-1α-regulating profile and suggests potential strategies to therapeutically target HIF signaling.

HIF-1α in cancer

HIF-1α overexpression in cancer

HIF-1α is induced in a broad spectrum of solid tumors as a result of the hypoxic nature inside the malignant mass at both primary and metastatic sites. In breast cancer, for example, the median oxygen partial pressure is as low as 10 mm Hg compared to that being 65 mm Hg in normal breast tissue [79]. However, the intra-tumoral expression of HIF-1α is not limited to the deoxygenated core, but also appears under normoxia in the process of immune selection and can be located at tumor margins as well [80]. Thus HIF-1α signaling is upregulated both by hypoxia and through other mechanisms in normoxia in cancer.

In a manner similar to how HIF-1α is regulated independently of oxygen, HIF-1α overexpression in cancer can be attributed to non-hypoxic signaling pathways as well. Alterations in different oncogenes and tumor suppressors give rise to the upregulation of HIF-1α through multiple pathways. (1) As mentioned, VHL loss-of-function mutations result in constitutive HIF-1α stabilization [63]. (2) TP53 (encoding p53) is the most commonly mutated gene in human cancer [81]. And p53 deletion disrupts MDM2 binding with HIF-1α, which increases HIF-1α expression in colon cancer cells [66]. (3) Mutations in the isocitrate dehydrogenase-1 (IDH1) gene impair the catalytic activity of the enzyme and reduce the production of α-ketoglutarate (α-KG) [82]. α-KG is the indispensable co-substrate in PHD-mediated HIF-1α hydroxylation. IDH1 mutation enhances HIF-1α transcriptional activity and is associated with high HIF-1α level in human gliomas [82]. (4) Tuberous sclerosis complex 2 (TSC2) is a tumor suppressor that inhibits the activation of mTOR. Mutation in the TSC gene causes the tuberous sclerosis disease, with which patients develop benign tumors and have an increased risk for giant cell astrocytoma [83] and ccRCC kidney cancer [84]. It is known that mTOR is involved in the stimulation of HIF-1α transcription and translation. Genetic inactivation of TSC2 increases the expression of HIF-1α mRNA and protein as well as boosts its target gene transcription [85]. (5) Another tumor suppressor, Phosphatase and tensin homolog (PTEN), is a negative regulator of the PI3K/Akt pathway and is frequently mutated in glioblastoma, endometrial cancer, and prostate cancer. Nearly 70% of prostate cancer patients carry a single-allele PTEN loss [86]. Defects in PTEN activity contribute to the expression of HIF-1α which leads to the transcription of responsive genes in prostate cancer cells [87]. (6) Amplification of the c-myc oncogene occurs in more than 15% of breast cancer cases [88]. It encodes the protein c-Myc which post-transcriptionally increases HIF-1α expression regardless of hypoxia in breast cancer cells [89]. (7) The RAS genes are the most commonly mutated oncogenes in human cancer [90]. In colorectal cancer, KRAS mutation occurs at a frequency of more than 40% [91]. The KRAS G12V mutation activates both Akt and ERK pathways and induces HIF-1α transcription in colorectal cancer cells [92].

An elevated level of HIF-1α enables cell survival and promotes cancer progression by orchestrating an extensive set of target genes involved in energetic switch, angiogenic induction, apoptotic counteraction, immune evasion, proliferative capabilities, invasive and metastatic properties, therapy resistance, and stemness maintenance (Table 1). In patients, HIF-1α overexpression has been correlated with adverse prognosis among multiple cancer types, such as brain [93], head-and-neck [12], lung [94], breast [13,95-97], gastric [98], pancreatic [99], colorectal [100-102], cervix [103-105], ovarian [106], bladder [107], and prostate cancers [108].

Table 1.

HIF-1α target genes in cancer progression

Involvement in cancer hallmarks	HIF-1α target gene product
Angiogenesis	Angiopoietin 2, Inhibitor of differentiation 2, Placental growth factor, Stromal-derived factor 1, Vascular endothelial growth factor
Cell survival & proliferation	Insulin-like growth factor 2, Survivin, WSB1, Platelet-derived growth factor B, Transforming growth factor-α
Immortalization	Telomerase
Invasion & Metastasis	Angiopoietin-like 4, C-ME, CXCR4, Endothelin 1, Fibronectin 1, Lysyl oxidase, Matrix metalloproteinase 2 and 14, TWIST1, Urokinase plasminogen activator receptor, ZEB1 (ZFHX1A), ZEB2 (ZFHX1B)
Glucose uptake & metabolism	Glucose phosphate isomerase, Glucose transporter 1, Hexokinase 1 and 2, Lactate dehydrogenase A, Pyruvate dehydrogenase kinase 1, Pyruvate kinase M2
Genomic instability	DEC1
Immune evasion	NT5E (ecto-5’-nucleotidase/CD73)
Stemness	ABCG2, JARID1B, Kit ligand (stem cell factor), OCT4
Epigenetic reprogramming	JMJD1A, JMJD2B, JMJD2C, PLU-1 [190]
Senescent cells	IL8, CXCR2, GROα, IL6, PAI1 [191]

Adapted from G L Semenza, 2010 [189].

Targeting HIF in cancer treatment

Apart from the attempts to develop pro-drugs that are activated in the hypoxic microenvironment, efforts have been made to target HIF signaling (Table 2). Anti-angiogenic drugs that block VEGF and its receptor represent an example of indirect targeting of HIF signaling through the downstream components that mediate the HIF-induced cancer-promoting effects. Some other approaches target the upstream of HIF signaling to inhibit HIF-α mRNA and protein levels. For instance, it is not unexpected to see the mTOR inhibitor, rapamycin, decreases the expression of HIF-1α in normoxic and hypoxic conditions [109], given the role of PI3K/Akt/mTOR pathway in HIF-1α transcription and translation. Temsirolimus, a soluble ester of rapamycin, is the first mTOR inhibitor approved by the FDA to treat cancer (advanced renal cell carcinoma) [110]. Other compounds that target the same pathway may have similar effects. FDA-approved topotecan is a camptothecin analogue that suppresses topoisomerase I and inhibits HIF-1α translation [111]. ENZ-2968 is a RNA oligonucleotide that selectively binds to the HIF-1α mRNA, reducing the expression of HIF-1α mRNA and protein in various cancer cells [112]. 2-Methoxyestradiol is an estrogen metabolite but does not bind with estrogen receptors. It decreases the nuclear accumulation of HIF-1α and inhibits its transcriptional activity [113]. Histone deacetylase (HDAC) inhibitors have been clinically used in cancer treatment. They are shown to reduce the HIF-1α expression [114]. Moreover, HIF-1α transcriptional activity can be regulated through the interruption of its binding with HIF-1β (e.g. acriflavine [115]), DNA sequences (e.g. echinomycin [116]), or p300/CBP (e.g. Chetomin [117]). Until the present time, several targeted therapies have become standard of care, some of which induce changes that are involved in hypoxia/HIF signaling. Meanwhile developing direct HIF destabilizers remains a challenge. Although some attempts targeting HIF pathway did not eventually make it to clinical application because of the toxicity, lack of efficacy, undefined patient selection, or discontinuation of the development, targeting the HIF signaling represents a promising approach in cancer therapy as it provides a way to generally and simultaneously target multiple malignant properties. Notably, targeting HIF-2α is thought to be important in certain tumors (e.g. ccRCCs). And certain inhibitors are selected for a HIF-2α specificity (e.g. PT2385). An orally available selective HIF-2α inhibitor, Belzutifan (MK-6482, developed by Merck), has been approved by FDA for cancers associated with von Hippel-Lindau (VHL) disease in 2021. The inhibitor showed favorable tolerance and anti-tumor activity in patients with previously treated clear cell renal cell carcinoma (ccRCC) in phase I/II study [118] and had a 49% objective response rate in patients with renal cell carcinoma in a phase II trial [119]. On the other hand, a recent transgenic mouse model that monitors tumor evolution has revealed an essential role of HIF-1α at the tumor-initiating phase in clear cell renal cell carcinoma [120]. It has also reinforced the involvement of HIF-1α in glycolysis and revealed HIF-2α involvement in other mechanisms such as lipoprotein metabolism. Overall, HIF-1α and HIF-2α are both potential targets with context-dependent importance in cancer therapy.

Table 2.

Molecular targeting of HIF signaling

Inhibitory mechanism	Molecular targets	Targeting agents
mRNA/protein expression	PI3K	Wortmannin, LY94002, GDC-0941, PI-103
	mTOR	Rapamycin, PP242, Glyceollins
	Topoisomerase 1	Topotecan, PEG-SN38
	HIF-1α mRNA	EZN-2968
	Microtubules	2ME2, ENMD-1198
	Hsp90	Geldanamycin and analogs
	HDAC	Vorinostat
	Thioredoxin-1	PX-12, pleurotin
	IRP1/IRE interaction	HIF-2α translational inhibitors
HIF-α/HIF-1β dimerization	HIF-1α/2α PAS-B domain	Acriflavine
	HIF-2α PAS-B domain	PT2385
DNA binding	HRE	Echinomycin, Polyamides
Transcriptional activity	p300 recruitment	Chetomin, Bortezomib
	FIH-1 interaction and p300 recruitment	Amphotericin B
	Hsp70	Triptolide
	Thioredoxin-1	AJM290, AW464

Adapted from Caroline Wigerup, et al., 2016 [192].

Cyclin-dependent kinases

CDKs in cell cycle regulation

Eukaryotic cells proliferate by going through cell cycles where they duplicate the genetic materials as well as important organelles and divide equally into two. Each cell cycle is composed of G1, S, G2, and M phases. DNA replication takes place in the S phase, in between of the G1 and G2 gap phases during which cells grow in size, produce proteins and assemble organelles to prepare for DNA synthesis and mitosis, respectively. Mitosis in the M phase results in segregation of the chromosome replicas, followed by cytokinesis where cytoplasm is partitioned in cell division. Cells in the G1 phase can exit the cell cycle and enter a resting phase referred to as G0, where they become quiescent or senescent and not dividing.

Progression of the cell cycle requires participation of the cyclin-dependent kinases (CDKs) (Figure 4). The first member among the CDK family was discovered by screening for mutants in a yeast species, Schizosaccharomyces pombe, in 1986 [121], and was named as cdc2 with regard to its role in the cell division control. Next year, another protein serine kinase was identified as PSK-J3 (putative serine/threonine kinase, filter J colony 3) by homologous probing [122]. Soon after, it was demonstrated that the activation of cdc2 requires its association with a protein that belongs to a previously recognized group, termed as cyclins [123,124]. In 1991, to unify the nomenclature, cdc2 was designated as CDK1, and PSK-J3 was designated as CDK4, along with other members designated as CDK2 [125-128], CDK5 and CDK6 [129-131]. To date, CDK1 through CDK20 have been identified, together with a series of cyclins. Cyclins bind to CDKs and allow them to catalyze the phosphorylation on their substrates. Cyclins are synthesized and degraded by ubiquitin-mediated proteolysis throughout the cell cycle, and thereby regulate the activity of different CDKs at corresponding stages. There are two categories of CDKs, with one being involved in cell cycle regulation (CDK1-CDK6, CDK14-CDK18) and the other one controlling gene transcription (CDK7-CDK13, CDK19 and CDK20) [132].

Figure 4.

The role and regulation of CDKs through cell cycle progression. CDK4/6-cyclin D promotes G1-S transition, which is further mediated by CDK2 activity. CDK1 is essential in G2-M progression, which results in the accomplished mitosis. CDK1, CDK2 and CDK4/6 activities are regulated through stimulation upon synthesis of and interaction with cyclins, as well as inhibition by checkpoint kinase signaling. The suppression of CDK1 is through inhibitory phosphorylation, while CDK4/6 and CDK2 are inhibited by protein binding. INK4A protein targets CDK4/6, while p21 and p27 suppress CDK2. Phosphorylation of Rb relieves its inhibitory interaction with E2F and allows E2F transcriptional activity. (Picture created at Biorender.com).

The activity of CDK1 plays an essential part in the G2-M transition. At the entry to G2 phase, CDK1 interacts with A-type cyclins (e.g. cyclin A2). As cell cycle progresses from G2 to M phase, A-type cyclins are degraded and B-type cyclins are synthesized. Consequently, CDK1 binds to B-type cyclins (e.g. cyclin B1) in M phase. CDK1-cyclin B activity has been implicated in multiple events that are fundamental to mitosis such as centrosome separation, nuclear envelope disassembly, chromosome condensation and spindle formation [133]. To prevent premature mitosis, CDK1 activity is tightly regulated. In addition to the cyclin-mediated control, cell cycle checkpoints are key determinants to assess the cell status, ensure the proper pace and restrict the cell cycle progression in response to DNA damage. For instance, checkpoint kinase 1 (CHK1) phosphorylates and activates WEE1, which in turn confers the phosphorylation on CDK1 and inhibits its activity [134,135]. CDK1 can be derepressed by CDC25 through its phosphatase activity [134]. Later in the M phase, elimination of B-type cyclins leads to the exit from mitosis.

CDK4 and CDK6 share high amino acid homology and the ability to bind D-type cyclins. The CDK4/6 activity is a key determinant in the G1-S phase transition. In the G1 phase, mitogenic signals upregulate the synthesis of D-type cyclins (D1-D3), which interact with CDK4/6. The complex phosphorylates the downstream substrates in the Rb (retinoblastoma) “pocket protein” family, including Rb, p107 and p130 proteins [136]. Rb functions by binding with the E2F transcription factors and thus inhibits the transcription of E2F-responsive targets [136]. The phosphorylation of Rb releases the sequestered E2F which activates the expression of genes that are required in G1-S progression. Among these stimulated genes are E-type cyclins, which subsequently upregulate CDK2 activity. CDK2-cyclin E hyper-phosphorylates Rb [137], resulting in irreversible cell cycle commitment. Activated CDK4/6 also derepresses CDK2 by interacting with p21^WAF1 and p27^KIP1 which inhibit CDK2 activity through interactions. As the cell cycle progresses, E-type cyclins are degraded while A-type cyclins are synthesized and bound by CDK2. The CDK2-cyclin A activity then assists the progression in S phase. The activity of CDK4/6 is regulated through binding by the INK4 family (p16^INK4A, p15^INK4B, p18^INK4C and p19^INK4D) [138-140]. Interaction with p18^INK4C disrupts the CDK4/6-cyclin D association and distorts the kinase catalytic region to suppress CDK4/6 activity [141-143].

Altered CDK4/6 signaling in cancer

Rapid and uncontrolled cell proliferation is a key trait of cancer. Examination of malignant tissues has revealed overexpression of promoting factors and abnormally activated progression signals in the cell cycle. After the activation of CDK4/6 in G1-S phase, the positive feedback loop between Rb phosphorylation and CDK2 activation forms a “restriction point”, from where the extracellular signal becomes dispensable for cell division. This feature has made the CDK4/6-Rb axis a major hotspot of dysregulation exploited by cancer cells. The CCND1 gene (encoding cyclin D1) is amplified in 10-20% in breast cancer (BC), and the overexpression of cyclin D1 has been reported in up to 81% of the BC cases [144-147]. CDK4 amplification occurs in 15-20% glioblastomas [148]. Homozygous deletion of CDKN2A (encoding p16^INK4A) is found in up to 43% IDH-mutant gliomas [149]. RB1 (encoding Rb protein) is the earliest identified tumor suppressor gene. Its mutation was detected in 75% in a cohort of small cell lung cancer (SCLC) [150].

Targeting CDKs in cancer therapy

The involvement of deregulated cell cycle and CDKs in tumorigenesis and cancer progression has prompted the development of CDK inhibitors in cancer treatment. Trials of Pan-CDK inhibitors (e.g. flavopiridol) suffer from the lack of specific biomarkers for patient selection and the undetermined therapeutic window. Flavopiridol also has significant toxicity. Whereas four of the selective CDK4/6 inhibitors have acquired FDA approval over the recent years, including palbociclib (Ibrance; Pfizer) (Figure 5A) [151], abemaciclib (Verzenio; Eli Lilly) (Figure 5B) [152], ribociclib (Kisqali; Novartis) (Figure 5C) [153] and trilaciclib (Cosela; G1 Therapeutics) (Figure 5D) [154]. Palbociclib and ribociclib have a higher selectivity to CDK4/6 than abemaciclib. However, abemaciclib has the advantage in crossing the blood-brain barrier which permits potential application on brain tumors and metastases [155]. These three CDK4/6 inhibitors are approved for use in hormone receptor-positive, HER2-negative advanced breast cancer and are taken orally. Trilaciclib is approved in February 2021 for the treatment of patients with small cell lung cancer (SCLC) in combination with chemotherapy. It is administered intravenously prior to chemotherapeutic agents to protect from chemo-induced myelosuppression [156,157].

Figure 5.

Approved CDK4/6 inhibitors. (A) Palbociclib, (B) abemaciclib, (C) ribociclib and (D) trilaciclib chemical structures are shown. Structures are drawn with Chem Space.

A remaining issue with targeted therapy inhibiting CDK4/6 is the development of resistance in patients. Some of the potential resistance mechanisms include amplification of CDK4/6, overexpression of cyclin D1/D2, upregulation of CDK2 and cyclin E1/E2, silencing of Rb, and activation of the growth factor signaling (Figure 6) [158]. To overcome the resistance, exploration of the combination treatment strategies is being investigated such as adding PI3K pathway inhibitors to the therapeutic plan of CDK4/6 inhibition plus endocrine therapy. Interestingly, inhibition of CDK4/6 has been shown to induce immune activity [159-161], which suggests a prospective strategy to combine CDK4/6 inhibitors with immunotherapy (e.g. Anti-PD1/PDL1). To date, there are still more CDK4/6 inhibitors and therapeutic plans with existing approved drugs under investigation in ongoing clinical trials [162] (Figure 7).

Figure 6.

Potential mechanisms linked with the resistance to CDK4/6 inhibition. The red “+” and “-” indicate where alterations could occur. “+” implicates activation and overexpression, and “-” implicates loss of function. (Picture created at Biorender.com).

Figure 7.

CDK4/6 inhibitors investigated in clinical trials. The status of CDK4/6 inhibitors in clinical trials for cancer treatment. Graph is generated based on information in the ClinicalTrials.gov database.

Heat-shock proteins

HSPs as chaperone proteins

The heat shock proteins (HSPs) were discovered unexpectedly in Drosophila melanogaster upon a temperature increase. In response to heat (and other cellular) stress(es), HSPs play a protective role on newly synthesized and misfolded proteins. HSPs are traditionally named as per their molecular weights. Accordingly, HSP90 proteins refer to a group of 90-kDa HSPs. Members within this family are highly conserved and assist the folding of their clients. Mammalian HSP90 isoforms include cytosolic HSP90α (HSP90α₁ & HSP90α₂) and HSP90β, mitochondrial TRAP, and GRP94 that is located in endoplasmic reticulum [163]. The cytoplasmic HSP90s represent the primary HSP90 expression and interact with more proteins than the isoforms residing in the ER and mitochondria. HSP90 homologues contain an N-terminal domain (NTD), a middle domain (MD), a C-terminal domain (CTD), and a charged linker (CR) region (Figure 8). The CTDs mediate the homodimerization of HSP90, which in turn serves as an ATPase. The NTDs provide an ATP binding site and serve as motifs for co-chaperone associations. Client proteins and co-chaperones are recruited to the MD region. For example, the activator of HSP90 ATPase homologue 1 (AHA1) interacts with the NTD and the MD of HSP90 and promotes the ATPase activity of the HSP90 dimer through the induction of dynamic conformational changes [164].

Figure 8.

The model of HSP90 cycle with dynamic ATP and co-chaperone associations. AHA1 accelerates the conformational switch, which results in the ATPase activity of HSP90 dimer. (Picture created at Biorender.com).

HSP90 chaperoning in cancer

During the initiation and progression of cancer, increased malignancy is coupled with the overexpression and various mutations of numerous proteins. HSP90 chaperones a variety of oncoproteins, such as cancer-related kinases and transcription factors, maintains their activated forms and prevents them from misfolding and degradation [165,166]. The functions of HSP90 clients span multiple hallmarks of cancer and alleviate external and internal stresses to facilitate cell survival (Figure 9). For example, the epidermal growth factor receptor (EGFR) is a tumor driver that typically induces PI3K/Akt signaling. Amplification and mutation of the EGFR gene often results in increased activation of the pathway and promotes cancer especially in lungs and breasts [167]. As an HSP90 client protein [168], EGFR is degraded upon HSP90 inhibition [169]. Another client of HSP90, DNA (cytosine-5)-methyltransferase 1 (DNMT1) [170], induces the hypermethylation at CpG islands to silence tumor-suppressive genes [171] and promote cancer. Overexpression of DNMT1 has been reported in multiple types of cancer (e.g. colon [172], breast [173] and liver [174]). As mentioned, HIF-1α is stabilized by HSP90, through which the role of HSP90 is linked to angiogenesis and metabolism alterations in cancer. The overexpression of HSP90 itself has also been associated with low survival rate in breast cancer [175].

Figure 9.

HSP90 regulates proteins involved in various cancer hallmarks. HSP90 clients are implicated in multiple pathways to help cancer cells survive the environmental stresses as well as to promote tumor growth and progression.

Targeting HSP90 in cancer therapy

HSP90 has been long perceived as a potential target in cancer treatment. A classic HSP90 inhibitor, geldanamycin (Figure 10A), was extracted as a natural product from Streptomyces hygroscopicus half a century ago [176]. It is structurally similar to ATP and able to dock at the ATP-binding pocket in the NTD of HSP90 [177-179]. Although the instability and hepatotoxicity have limited its applications in clinic, geldanamycin has been used in vitro for decades to elucidate the effects of HSP90 inhibition. The derivative of geldanamycin, 17-N-Allylamino-17-demethoxygeldanamycin (17-AAG) was the first HSP90 inhibitor to be tested in clinical trials two decades ago. A concern about 17-AAG is its poor aqueous solubility [180]. The development of 17-AAG was eventually discontinued. Over the past years, multiple HSP90 inhibitors have emerged and have been tested in more than 170 clinical trials (Figure 11) [181]. Some of the inhibitors developed during the recent decade include ganetespib (also known as STA-9090) [182] (Figure 10B), onalespib (also known as AT13387) [183] (Figure 10C), luminespib [184] (Figure 10D), XL888 [185] (Figure 10E), and TAS-116 [186] (Figure 10F). While most HSP90 inhibitors are similar in the way that they bind to the NTD region competitively against ATP, TAS-116 is unique in that it is the first-in-class to enter clinical trials with a selectivity for cytosolic HSP90 (HSP90α & HSP90β) [186]. An anti-cancer effect has been observed with TAS-116 in non-small cell lung cancer and gastrointestinal stromal tumor (GIST) in a Phase I study [187].

Figure 10.

Structures of HSP90 inhibitors. (A) Geldanamycin, (B) ganetespib, (C) onalespib, (D) luminespib, (E) XL888 and (F) TAS-116 chemical structures are drawn with Chem Space.

Figure 11.

HSP90 inhibitors investigated in clinical trials. The status of HSP90 inhibitors in clinical trials for cancer treatment. Graph is generated based on information in the ClinicalTrials.gov database.

Summary

Accumulation of HIF-1α is a signature of hypoxia and potentiates tumor propagation and malignant progression. HIF-1α overexpression is associated with poor prognosis in various cancer types. Malignant induction of HIF-1α signaling is not restricted to intratumoral hypoxic regions. Understanding and targeting the mechanisms of cancer-related HIF-1α stabilization potentially help to improve current cancer treatments. Targeting HIF-1α in malignant cells under both normoxia and hypoxia represents a promising approach to generally target tumordriving molecules and pathways involved with cancer hallmarks.

We have previously shown that cyclin-dependent kinase 1 (CDK1) stabilizes HIF-1α through phosphorylation of its Ser668 residue in a VHL-independent manner. The stabilization occurs both under hypoxia and at G2/M under normoxia. Meanwhile we showed that CDK4/6 proteins also a role in HIF-1α stabilization. We have recently shown a convergence of CDK1 or CDK4/6 and HSP90 signaling on HIF-1α, provides the rationale and preclinical reference for targeting HIF-1α by strategies of combination therapy [188].

Understanding the different mechanisms of HIF-1α regulation will help to pave the way for therapeutic targeting of the related components in cancer as well as other physiological disorders which involve HIF-1α activity. The therapeutic targeting of HIF1-alpha and the non-canonical mechanism(s) of its destabilization during cancer therapeutic intervention is an area ripe for much research.

Disclosure of conflict of interest

None.