Hypoxia induces profound changes in the cellular gene expression profile. The discovery of a major transcription factor family activated by hypoxia, HIF (hypoxia-inducible factor), and the factors that contribute to HIF regulation have greatly enhanced our knowledge of the molecular aspects of the hypoxic response. However, in addition to HIF, other transcription factors and cellular pathways are activated by exposure to reduced oxygen. In the present review, we summarize the current knowledge of how additional hypoxia-responsive transcription factors integrate with HIF and how other cellular pathways such as chromatin remodelling, translation regulation and microRNA induction, contribute to the co-ordinated cellular response observed following hypoxic stress.
- hypoxia-inducible factor (HIF)
- microRNA (miRNA)
Hypoxia, or decreases in the oxygen concentration, activates a variety of complex pathways at both the cellular and organism level, with the ultimate aim of reinstating oxygen homoeostasis. Although physiological responses to hypoxia have been appreciated for a long time, the molecular processes activated within the cells are still under investigation. However, this area has been greatly advanced by the discovery of a major transcription factor that responds to hypoxia, HIF (hypoxia-inducible factor). Importantly, the finding of hypoxia-responsive HIF regulators has uncovered the complex mechanism of control over this transcription factor family. A novel class of dioxygenases called PHDs (prolyl hydroxylases) were identified as molecular oxygen sensors within the cell. How these enzymes are regulated within the cell and what other possible targets they possess is now under intense investigation. Apart from the HIF system, other transcription factors are hypoxia-responsive. In addition, other molecular processes are activated following hypoxia that will ultimately determine the transcription profile of a given cell. In the present review, we describe several of the pathways that are activated following hypoxic stress that influence the pattern of gene expression achieved.
THE HIF SYSTEM
Molecular research into hypoxia-induced cellular responses was greatly enhanced by the seminal discovery of HIF-1 by the Semenza laboratory in the early 1990s . Since this discovery, a number of laboratories have contributed to the delineation of the HIF system as we presently know it. The HIF family of transcription factors are composed of a heterodimer of α and β subunits [2,3]. There are three known isoforms of HIF-α: HIF-1α, HIF-2α and HIF-3α. HIF-3α has several splice variants. HIF-1β, also known as ARNT (aryl hydrocarbon receptor nuclear translocator), also possesses several splice variants and is constitutively expressed [2,4]. HIF-α subunits share highly similar domain regions, characterized by the presence of bHLH (basic helix–loop–helix)–PAS (Per/ARNT/Sim) domains (Figure 1). In addition, they possess an ODD (oxygen-dependent degradation domain), rendering these proteins labile in the presence of oxygen (see below). Whereas HIF-1α and HIF-2α contain two transactivation domains, HIF-3α lacks the C-terminal transactivation domain, and, as such, is thought to act as an inhibitor of HIF-1α and HIF-2α . HIF-1 and HIF-2 have non-redundant functions in the cell, and, although HIF-1α is the best understood isoform, recent studies have unveiled important functions for HIF-2α [5,6]. A number of recent reviews have focused on HIF-mediated targets and function [2,3,7] and therefore we will not focus on this matter. Instead, we discuss how HIF integrates with other transcription factors and how HIF is, or could be, modulated by other hypoxia-induced cellular processes which contribute to the gene expression pattern.
PHD AND FIH-1 (factor inhibiting HIF-1) REGULATION OF HIF
The discovery of HIF led to an intensification of the research on the molecular mechanisms behind hypoxia cellular sensing. The identification of the players involved in HIF regulation was the subject of extreme interest. The identification of a novel class of dioxygenases called PHDs made these proteins prime candidates for the task of cellular oxygen sensors. PHDs catalyse the hydroxylation reactions in the ODD of HIF [8,9]. They require molecular oxygen, iron and 2-oxoglutarate as co-factors in this reaction [9,10]. The hydroxylation of HIF acts as a signal for recognition by the tumour suppressor VHL (von Hippel–Lindau protein). VHL acts as an E3 ligase, promoting Lys48-linked ubiquitination and proteasomal-mediated degradation (Figure 2). There are four PHD isoforms identified so far: PHD1–PHD4. Only isoforms 1, 2 and 3 have been shown to hydroxylate HIF . Genetic and biochemical studies have demonstrated differences between these enzymes [11–14]. Biochemical studies have shown that PHD2 has higher affinity for HIF-1α, whereas PHD1 and PHD3 have higher affinity for HIF-2α . In addition, the importance of PHDs in development was also elucidated recently [12–14]. Although homozygous deletion of PHD2 results in embryonic lethality between E (embryonic day) 12.5 and E14.5, PHD1- or PHD3-knockout mice are viable . However, despite not causing embryonic lethality, a double deletion of PHD1 and PHD3 induced moderate erythrocytosis . In a more recent study, PHD3 deletion was shown to result in abnormal sympathoadrenal development . This study also confirmed the previous suggested increased affinity for HIF-2α . These genetic models will be extremely helpful in determining the importance of the individual PHDs in terms of acute compared with prolonged hypoxia.
In addition, to the prolyl hydroxylation catalysed by the PHDs, another oxygen-dependent modification occurs in the most C-terminal transactivation domain of HIF-α [15–17]. This is an asparaginyl hydroxylation, catalysed by another dioxygenase called FIH-1. Asparaginyl hydroxylation prevents binding of HIF to its co-activators p300 or CBP [CREB (cAMP-response-element-binding protein)-binding protein], and hence prevents full target gene activation (Figure 2). However, several studies have shown that, in the absence of FIH-1  or absence of p300 binding , a number of target genes are still HIF-inducible. These studies have raised the possibility that either the N-terminal transactivation domain is sufficient for induction of certain target genes or, alternatively, binding of unknown co-activators can occur even in the presence of FIH-1-mediated asparaginyl-modified HIF. Further exploration of these possibilities over the next few years should provide valuable insights into this aspect of HIF regulation. As with the PHDs, a genetic knockout model would reveal the importance of this HIF regulator in development and hypoxia responses.
Since the discovery of this class of dioxygenases, a large number of research laboratories have been investigating alternative and novel targets to HIF. Interestingly, FIH-1 has been shown to be able to modify a number of other targets, all of which contain ankyrin repeats, such as IκB [inhibitor of NF-κB (nuclear factor κB)]-like molecules  and Notch . Again, future studies will determine the relative contribution of these novel targets to the gene expression profile following hypoxic stress.
HYPOXIA-RESPONSIVE TRANSCRIPTION FACTORS
Apart from the HIF family, hypoxia activates a number of other important transcription factors, such as NF-κB, AP-1 (activator protein 1), p53 and c-Myc, among others (Figure 3). Given the importance of HIF for the cellular response to hypoxia, cross-talk between these different transcription factors should exist, and, in fact, it has been documented for several of them. In this section, we will review the recent findings in this area.
NF-κB is the collective name for a family of seven proteins, encoded by five genes: RelA, RelB, c-Rel, NF-κB1 (p105/p50) and NF-κB2 (p100/p52) (reviewed in ). These proteins all share a conserved region of homology at their N-terminus called the Rel homology domain. The C-terminus of these proteins is poorly conserved between family members, with three of the proteins possessing transactivation domains (RelA, RelB and c-Rel). The remaining two (NF-κB1 and NF-κB2) require proteolytic cleavage to achieve their active forms (p50 and p52). However, since p50 and p52 do not possess transactivation domains, they generally need to heterodimerize to activate transcription. NF-κB is best known for its role in the immune system and in inflammatory responses. However, recent studies have demonstrated its prominent role in other disorders such as cancer .
Hypoxia-mediated induction of NF-κB was reported over a decade ago . Despite this, the mechanism of hypoxia-induced NF-κB has not been clearly defined, although a number of different laboratories have demonstrated that NF-κB can contribute to the cellular response to hypoxia [25–28]. Although in some tissues and cell systems, hypoxia-induced NF-κB acts as a survival signal , in neuronal systems, NF-κB acts as a pro-death factor [26,27,29]. This dual nature of NF-κB is now well established ; however, what triggers the different and often opposing responses in NF-κB function is a matter of intense research in the field. Most studies have concentrated on the RelA–p50 heterodimer of NF-κB [25,28,30]. However, to date, no study has investigated the function of NF-κB2 (p100/p52) in hypoxia. This research could have important implications, as NF-κB2 has significant links to cancer and cell-cycle progression [31–33]. The function of RelB in hypoxia has also not been investigated so far. Importantly, RelB activation was recently shown in invasive breast tumours , suggesting that this NF-κB subunit also plays a role in cancer. Only a few studies have documented the involvement of c-Rel in the hypoxia response. These have shown that the c-Rel–p50 heterodimer regulates the expression of a known target and important anti-apoptotic gene, Bcl-X . A more in-depth analysis of NF-κB complexes would reveal which of the subunits are important for HIF co-operation.
Interestingly, we, and others, have shown that NF-κB can directly modulate HIF-1 transcriptionally [36–38]. This occurs not only in response to stimuli such as TNFα (tumour necrosis factor α) [36,38] and H2O2 , but also in response to hypoxia . Whether HIF-1 or HIF-2 can regulate directly any of the NF-κB subunits has not been investigated so far. It is possible that HIF can induce NF-κB via the activation of HIF targets such as TNFα and IL (interleukin)-1β [39,40], two potent NF-κB inducers. Studies over the next few years should determine whether a reciprocal and more direct link exists.
AP-1 refers to the combination of dimers formed between the Jun, Fos and ATF (activating transcription factor) families of transcription factors. As such, its regulation and biological functions are highly complex. AP-1 activation has been associated with a great number of biological processes, such as proliferation, apoptosis and even tumorigenesis. Genetic knockout and transgenic experiments have determined the importance of some AP-1 members for development . Although c-Jun, JunB and Fra-1 are indispensable for embryonic development, c-Fos, FosB and JunD are dispensable, despite harbouring other defects (for a review, see ).
It is well accepted that hypoxia induces AP-1 activity (reviewed in [42,43]). The mechanisms of AP-1 activation, however, are generally cell-type-specific and even dependent on the experimental conditions used. Transcriptional activation for c-Jun  and JunB  has been demonstrated in certain cell types, whereas increased post-translational modifications owing to activation of upstream signalling cascades has been shown to occur for c-Jun .
In general terms, AP-1 exerts its effects by acting in co-operation with additional transcription factors to modulate their activity. In fact, co-operation between AP-1 and HIF-1 has been reported [46,47] and is thought to occur in hypoxia. Although physical association between these transcription factors has not been formally demonstrated, functional co-operation has been widely documented [46,47]. In addition, AP-1 and NF-κB co-operation is observed in the activation of common target genes, such as IL-8 . The impact of AP-1 on the gene expression profile following hypoxia has not been established. Given the number of AP-1 family members, this question is almost impossible to answer. However, depletion of certain AP-1 components does not seem to affect the cellular response to hypoxia .
One additional and important transcription factor activated by hypoxia is the tumour suppressor p53. Although the activation of p53 following hypoxia is to be expected, given its sensitivity to most types of stress within the cell (reviewed in ), p53 activation following hypoxia is very atypical. Hypoxia-induced p53 has been suggested to be dependent on the DNA-checkpoint kinases ATM (ataxia telangiectasia mutated)/ATR (ATM- and Rad3-related). However, hypoxia does not induce detectable DNA damage, the stimulus for these sensor kinases. Hypoxia mostly induces a G1 arrest in the cell cycle, through HIF-1-dependent and -independent mechanisms . In addition, hypoxia-induced p53 has been shown to be both dependent and independent of HIF-1, in several different studies (reviewed in ).
Surprisingly, experiments investigating the transcriptional activity of p53 found that, under hypoxic conditions, p53 did not induce the same set of genes as a typical p53-activating stimulus such as UV light [50,51]. Most studies have found that p53 activated by low oxygen levels acts to repress certain genes, such as α-tubulin, while failing to induce others, such as p21 . However, it has been demonstrated that p53 does activate a subset of genes different from those normally seen with DNA-damaging agents [52–54]. For example, BNIP3L (Bcl2/adenovirus E1B 19 kDa interacting protein 3-like) was identified as a key p53 target gene activated under hypoxia that is responsible for p53-mediated apoptosis following hypoxic stress . Furthermore, the Harris laboratory has shown that hypoxia-induced p53 activates a very similar subset of genes to those seen following nitric oxide exposure . Interestingly, ChIP (chromatin immunoprecipitation) experiments have shown that p53 is recruited to target gene promoters, despite no increase in mRNA production being detected [55,56]. Therefore alternative mechanisms for p53-mediated responses in hypoxia, which do not necessitate an increase in mRNA production, warrant consideration. These mechanisms would include regulation of translation rates, mRNA maturation processes or even miRNA (microRNA) induction (see below).
Although the transcriptional targets are currently not well defined, it is clear that p53 is important for hypoxia-induced apoptosis. However, further and more detailed analysis will be required to identify the mechanisms by which p53 mediates hypoxia-induced apoptosis.
The Myc family
Genes of the Myc family of transcription factors are involved in a wide variety of cellular processes such as cell growth, cell proliferation, inhibition of cell differentiation, promotion of angiogenesis and genomic instability (reviewed in [57,58]), and as such are subject to very tight control .
In mammalian cells, there are four members of the Myc family: c-Myc, N-Myc, L-Myc and S-Myc . No monomers or homodimers of Myc proteins are found in vivo; instead, Myc proteins heterodimerize with a partner protein, Max, through contacts between their bHLHZ (bHLH zipper) domains [59,60]. Myc–Max heterodimers bind to specific DNA sequences, termed E-boxes, within the promoters of target genes to activate or repress transcription . Myc can also bind to non-consensus sequences through protein–protein interactions to alter expression of target genes . Max itself is also under tight regulation through protein–protein interactions with Mad1, Mxi1 (Mad2), Mad3 or Mad4, which compete for binding to Max, and function to antagonize Myc activity ([63–65], but see [65a]).
Mammalian cells respond to hypoxic stress by reducing cell growth and by inducing gene products, which causes cell-cycle arrest. Both of these activities are usually promoted by high Myc activity [66–68]. A body of work has suggested that Myc function is compromised under low oxygen conditions and therefore contributes to the hypoxic response. This antagonism of Myc can occur through a variety of mechanisms. Under hypoxic conditions, cell proliferation ceases through the induction of cyclin-dependent kinase inhibitors normally repressed in proliferating cells under normoxic conditions [66,68]. Myc and HIF complexes can compete for binding sites at the promoters of target genes to alter their expression profile. This has been demonstrated in the regulation of the cyclin-dependent kinase inhibitor p21. Under normoxic conditions, p21 is repressed by c-Myc binding to its promoter, but when cells are exposed to low oxygen, c-Myc is replaced by HIF-1, resulting in induction of p21 expression and hence cell-cycle arrest . Indeed, HIF-1 overexpression, even in normoxia, is sufficient to induce p21 expression . Interestingly, HIF-1α transcriptional activity or its DNA binding are not essential for the cell-cycle arrest, indicating a novel function for HIF-1α .
HIF-1α can block Myc binding to its DNA-binding partners directly. HIF can bind to Myc itself, but also to its binding partners Max and Sp1, in hypoxic cells [68–71]. This acts to disrupt Myc–Max complex formation and DNA binding and hence reduce expression of Myc target genes. HIF can also disrupt Myc function indirectly by mediating the induction of the Myc antagonist, Mxi . This results in a decrease in the levels of several Myc target genes and inhibition of transformation .
In contrast with HIF-1, elevated HIF-2 can promote c-Myc activity in a variety of cell lines . It can achieve this by binding to and stabilizing the Myc–Max heterodimer, allowing it to transactivate its target genes. HIF-2α was shown to act in concert with Myc to promote transformation of mouse embryonic fibroblasts . As HIF-1 is expressed ubiquitously and HIF-2 is only expressed in a subset of cell types, the different effects of these factors on Myc activity in hypoxia may reflect the different requirements of different tissues.
Although there are many genes and processes in which Myc and HIF have opposite effects, both are elevated in tumour cell types and promote processes such as angiogenesis. Furthermore, it has been shown that HIF-1α is essential for c-Myc-mediated tumorigenesis . Detailed analysis of cells expressing elevated Myc showed that HIF and c-Myc can co-operate to induce shared target genes VEGF (vascular endothelial growth factor), HK2 (hexokinase 2) and PDK1 (pyruvate dehydrogenase kinase 1) . These observations reveal a complex interplay between the Myc and HIF transcription factors, which will ultimately be cell-type- and cell-context-dependent.
The co-operative or antagonistic nature of the connection between HIF and the additional transcription factors is bound to determine the gene expression profile of each cell (Figure 3). Whether a cell progresses or stops its cell cycle, survives or dies following hypoxia exposure is tightly linked to the cross-talk between HIF and transcription factors such as p53, NF-κB and c-Myc, among others.
We have discussed how hypoxia activates HIF, as well as other transcription factors, which act in concert to alter the gene expression profile of the cell. Although the identities of some of these DNA-binding factors are known, it is still unclear by which precise mechanisms they activate or repress transcription of their target genes. For transcription factors to activate gene expression, they must be recruited to the promoters or enhancers of their target genes through direct binding to specific DNA sequences. As a mammalian cell packages approx. 2 m of DNA into a compact chromatin structure that is contained in a nucleus of less than 10 μm in diameter, these sequences may not always be accessible. Therefore different mechanisms exist to allow the transcription machinery to access these areas.
The fundamental unit of the chromatin is the nucleosome, which consists of 146 bp of DNA wrapped twice around an octomer formed from the globular domains of two molecules of each of histones H2A, H2B, H3 and H4 . The globular domains of the histones form the core of the nucleosome, and the basic flexible tail regions of the histones protrude from the octomer. These tail regions are thought to be associated with the negatively charged DNA wrapped around the outside . Each nucleosome is connected with 10–80 bp of linker DNA associated with a linker histone, such as histone H1 or H5. These complexes are progressively folded into higher-order and more condensed structures . The higher-order packed chromatin is generally repressive for transcription, because of the inaccessibility of the DNA [78,79]. Appropriate gene expression therefore requires interplay with complexes that when recruited by transcriptional activators, or repressors, can adjust the chromatin structure to alter its accessibility.
Chromatin structure can be altered by three main mechanisms: through mobilization of the nucleosomes utilizing energy from ATP ; post-translational modification of histone proteins (predominantly within the flexible tail regions) ; or replacement of histone proteins in the octomer with histone variants .
Nucleosome remodelling is performed by various complexes, which utilize the energy from ATP to alter the contacts between the histones and the DNA, thus facilitating disruption of the nucleosome structure . These can be subdivided into three groups on the basis of biochemical properties and the conservation of their catalytic subunits: SWI/SNF, ISWI and MI-2/CHD groups (for a review, see ). In general terms, the SWI/SNF complexes are involved in gene activation, whereas ISWI and MI-2/CHD are involved in transcriptional repression.
Another way to change the accessibility of DNA within the chromatin is through post-translational modification. Residues on histones have been found to be phosphorylated, acetylated, methylated, ubiquitinated, SUMOylated and poly(ADP-ribosyl)-ated. These covalent modifications of histones correlate with profound effects on transcription, either changing the charge and structure of the chromatin or providing binding sites for co-activator or co-repressor proteins to be targeted to the DNA. Each one of these modifications can be correlated with an effect on transcription and it has been proposed that the combination of modifications on the histones surrounding a gene can fine-tune its expression .
It is therefore surprising, considering the profound effects that chromatin has on transcription, that almost no work has been done to elucidate the effects of chromatin on gene expression following hypoxia.
The HAT (histone acetyltransferase) p300 has been shown to interact with HIF and is required for full activation of target genes . Several studies have demonstrated an increase in localized histone acetylation surrounding activated genes following hypoxia (Figure 4) [85–87], but the observation that disrupting the HIF–p300 interaction only alters expression of 35–50% of HIF-responsive genes suggests that p300-induced acetylation of promoter histones is not required for all HIF target genes . Several studies have also investigated the effects of HDACs (histone deacetylases) on HIF-1 function [88,89]. One of these studies identified HDAC7 as an interacting partner of HIF-1, and revealed its importance for HIF-1 activation. This apparent paradox between HDACs and HATs both being co-activators for HIF activity may reflect differing co-factor requirements for different target genes. The situation is complicated further by the demonstration, by co-immunoprecipitation analysis, that HIF, HDAC7 and p300 can exist in the same complex at the same time (Figure 4) [88,90]. This may reflect another level of control of the HIF response with different combinations of co-factors being available to activate different target genes. In vivo ChIP assays identifying co-factor recruitment to HIF target genes would help to elucidate how co-repressors and co-activators are targeted.
A recent study has also demonstrated changes in histone methylation on HIF target genes when cells are challenged with low oxygen . In this study, the promoters of the hypoxia-induced genes VEGF and EGR1 (early growth response 1) were investigated . When cells were exposed to hypoxia, there was an observed increase in the proportion of promoter histones displaying H3K4 (histone H3 Lys4) trimethylation, a mark of actively transcribing genes , and a decrease in H3K27 (histone H3 Lys27) trimethylation, a modification generally associated with genes in the inactive state (Figure 4) . The identity of the methylases and demethylases involved in this switch is unknown, but may indicate how important epigenetic changes can be in the regulation of hypoxia-induced genes.
The contribution of ATP-dependent chromatin-remodelling enzymes to the gene expression changes induced by hypoxia has also been severely underinvestigated. The one report to date that has investigated how this class of enzymes can alter transcription in hypoxia has focused on the EPO (erythropoietin) gene . This study demonstrated that the SWI/SNF complex is required for full induction of the EPO mRNA following hypoxia. It is also interesting to note that there was an increase in localized histone acetylation at the promoter of the EPO gene following hypoxia , perhaps indicating a series of chromatin changes induced by hypoxic stress. Although not looking directly at ATP-dependent remodellers, a study examining genes both repressed and activated by exposure to low oxygen has demonstrated an inverse relationship between nucleosome occupancy and transcriptional activation of these genes (Figure 4) . These findings may indicate a hypoxia-responsive remodelling of chromatin, with a link to the activation of downstream target genes.
Another important aspect of gene expression is translation. Following hypoxia, global protein synthesis is slowed down or inhibited (Figure 5). This is part of an energy-saving mechanism, by which cells adapt and attempt to survive stressful conditions such as lack of oxygen or nutrients.
Mechanistic studies on translation and overall protein synthesis have been mostly performed following severe lack of oxygen, ranging from 0.5 to 0% O2 [91,92]. However, moderate hypoxia (1–5% O2) can produce the same effects, although with delayed kinetics [93,94].
Translation is a complex process with multiple levels of regulation. Molecularly, the process of mRNA translation can be divided into three steps: initiation, elongation and termination. Initiation is the most complex step and is very tightly controlled by a number of molecules termed eIFs (eukaryotic initiation factors). In eukaryotes, most translation is cap-dependent, where the assembly of an active eIF4 at the m7GpppN mRNA structure (5′-end of the RNA) is required. This complex consists of a cap-binding protein eIF4E, a scaffold protein eIF4G and an ATP-dependent helicase eIF4A. This complex facilitates the recruitment of the 43S pre-initiation complex, composed of the small 40S ribosomal subunit, the initiation factor eIF3, and the ternary complex [eIF2–GTP–Met-tRNAi (initiator-methionyl tRNA)] . The function of the 43S complex is to scan through the 5′-UTR (untranslated region) of the mRNA until it encounters the AUG initiation codon. eIF5 then promotes the hydrolysis of GTP-bound eIF2 into GDP, releasing the initiation factors and promoting the recruitment of the large 60S ribosomal subunit. This defines the beginning of the elongation step.
The eIF4 and eIF2 complexes constitute the two major control points in initiation of translation. eIF4E protein levels are limiting, and the formation of the eIF4 complex can be prevented by a family of 4E-BPs (eIF4E-binding proteins), which compete with eIF4G for binding. 4E-BPs are regulated by phosphorylation, mediated mainly by mTOR (mammalian target of rapamycin), a global integrator of several signalling pathways .
mTOR is a serine/threonine protein kinase that responds to changes in energy status, nutrient availability, insulin and growth factors, and its activity is associated with cell-cycle progression and cell growth. mTOR-mediated phosphorylation of 4E-BPs induces a reduction in their affinity for eIF4E, thereby allowing it to bind eIF4G. In addition, mTOR also phosphorylates S6 kinase, which stimulates translation via the ribosomal protein S6 and the ribosomal recruitment protein eIF4B .
In hypoxia, mTOR activity is inhibited, leading to hypophosphorylation of S6K, S6 ribosomal protein and 4E-BP1 . The mechanism of mTOR inhibition is complex. However, the data published so far suggest that hypoxia can induce a decrease in cellular energy levels, leading to activation of the energy-responsive kinase AMPK (AMP-activated protein kinase) . AMPK phosphorylates the TSC1/2 (tuberous sclerosis complex 1/2), which down-regulates mTOR activity. AMPK activation is mainly regulated by the tumour suppressor protein kinase LKB1 . Additional kinases are also involved, although their identities are currently unknown .
Hypoxia also induces hyperphosphorylation of Ser51 of eIF2α [91,93]. This results in translation inhibition, since phosphorylation of eIF2α at this site prevents the exchange of GDP for GTP and sequesters eIF2B in an inactive complex. Phosphorylation of eIF2α can be achieved by several different kinases: PKR (interferon-inducible double-stranded RNA-activated kinase), HRI (haem-regulated inhibitor of translation), GCN2 (kinase activated by nutrient starvation) and PERK (PKR-like endoplasmic reticulum kinase), which is actively involved in the UPR (unfolded protein response). All of these kinases can respond to different types of stress. PERK is the main kinase responsible for eIF2α phosphorylation following hypoxia [91,98,99]. Several studies have identified the UPR as a component of the cellular response to hypoxia, leading to more mechanistic insights as to how cells survive such harsh conditions. Genetic studies with loss-of-function models for PERK and ATF4 (a transcription factor that is up-regulated following hypoxia in a HIF-independent manner and is also involved in the UPR) demonstrated that these proteins contribute to overall survival. In addition, cells deficient for these factors are more sensitive to hypoxia-induced cell death, and tumour xenograft models have demonstrated that PERK contributes to tumour growth in vivo . These studies have identified the UPR as a good candidate for therapeutic targeting in solid tumours.
Interestingly, a recent study demonstrated that 4E-BP1 and eIF4G are highly overexpressed in human breast cancers . This study suggested that this facilitates cap-independent translation in hypoxic tumours, and promotes translation of IRES (internal ribosome entry site)-containing mRNAs allowing survival and angiogenesis to occur . However, IRES-containing mRNAs cannot account for all the preferential translation that takes place in hypoxia. In fact, a more recent study investigated the activity of previously documented IRES elements in certain mRNAs that are preferentially translated in hypoxia . This study demonstrated that hypoxia-selective mRNA translation was independent of IRESs, even for IRES-containing transcripts such as VEGF and HIF-1α . One additional mechanism for controlling gene expression, which is dependent on eIF2α phosphorylation, was described recently . In this study, nonsense-mediated RNA decay was found to be inhibited by hypoxia, giving rise to increased stability of a number of mRNAs of stress-induced genes such as ATF4, ATF3 and CHOP [C/EBP (CCAAT/enhancer-binding protein)-homologous protein 10] . This suggests yet another pathway modulated by hypoxia (Figure 5).
The fact that translation still occurs in hypoxia by mechanisms that are cap-independent and IRES-independent gives rise to the possibility that other pathways contribute to the gene expression in these conditions. One such pathway is miRNA modulation. miRNA is a non-coding RNA composed of 21–24 nt, which is derived from an endogenous hairpin RNA precursor (usually approx. 70 nt long). The process of miRNA maturation involves several classes of proteins with enzymatic capability such as the RNase-III-type enzyme called Dicer and multifunction proteins such as Argonautes (for reviews on the mechanisms of miRNAs, see [103,104]). The miRNA mode of action seems to be more of a fine-tuning of expression rather than degradation of mRNA, a mechanism by which RNA interference works . miRNAs can prevent translation by several mechanisms, either by preventing initiation, or by preventing elongation by inducing ribosome drop from the polypeptide (Figure 6).
Despite the relatively recent discovery of this form of gene regulation, hypoxia induction of miRNAs has been documented [105–109]. Some studies have even identified a miRNA signature associated with hypoxia in certain cell types [108,110]. Furthermore, some miRNA targets have been identified, with VEGF being one of the most prominent . A study investigating mechanistic manipulations of miRNAs in hypoxia has been conducted recently . However, this study only focused on one particular miRNA, mir-210. The next few years should determine how much of the hypoxia-induced gene expression is dependent on miRNA function. In addition, the next big question shall be the determination of which miRNAs are important for cell transformation and tumour survival in hypoxic situations. One additional point to consider is the fact that several of the transcription factors activated by hypoxia can induce miRNAs; these include p53 and NF-κB, as well as c-Myc [111–114]. Induction of transcription-factor-mediated miRNA has not been investigated in the context of hypoxia stimulation. Despite this fact, miRNA induction is a very important aspect of the biology of these transcription factors that needs to be taken into account while testing for hypoxia responses.
Traditionally, gene expression analysis following hypoxia has focused on HIF-mediated transcriptional responses; however, many pathways are activated by hypoxic stress within the cell (Figure 7). In order to achieve a co-ordinated response under such stressful conditions, all of these pathways must be integrated. Chromatin reorganization, translational processes and miRNA induction are all new aspects that have to be considered in the analysis of the hypoxia response (Figure 7). In addition, cross-talk with other transcription factors also modulates HIF activity. Future research in these areas will offer much needed insight into how the gene expression profile of a cell under hypoxia is determined, with the hope of applying this knowledge to pathological situations.
We thank Professor Neil Perkins, Dr Cari Culver, Mrs Sharon Mudie and Mr Patrick van Uden for valuable comments on this manuscript. We thank all of the members of the Wellcome Trust Centre for Gene Regulation and Expression at the College of Life Sciences, University of Dundee, for their support and encouragement. Work in our laboratory is supported by the International Association for Cancer Research (AICR), the Medical Research Council (MRC), Research Councils UK (RCUK) and the University of Dundee.
Abbreviations: AMPK, AMP-activated protein kinase; AP-1, activator protein 1; ARNT, aryl hydrocarbon receptor nuclear translocator; ATF, activating transcription factor; ATM, ataxia telangiectasia mutated; bHLH, basic helix–loop–helix; ChIP, chromatin immunoprecipitation; eIF, eukaryotic initiation factor; 4E-BP, eIF4E-binding protein; EPO, erythropoietin; FIH-1, factor inhibiting hypoxia-inducible factor-1; HAT, histone acetyltransferase; HDAC, histone deacetylase; HIF, hypoxia-inducible factor; IL, interleukin; IRES, internal ribosome entry site; miRNA, microRNA; mTOR, mammalian target of rapamycin; NF-κB, nuclear factor κB; ODD, oxygen-dependent degradation domain; PHD, prolyl hydroxylase; PKR, interferon-inducible double-stranded RNA-activated kinase; PERK, PKR-like endoplasmic reticulum kinase; TNFα, tumour necrosis factor α; UPR, unfolded protein response; VEGF, vascular endothelial growth factor; VHL, von Hippel–Lindau protein
- © The Authors Journal compilation © 2008 Biochemical Society