Hypoxia-dependent sequestration of an oxygen sensor by a widespread structural motif can shape the hypoxic response - a predictive kinetic model
© Schmierer et al; licensee BioMed Central Ltd. 2010
Received: 17 May 2010
Accepted: 18 October 2010
Published: 18 October 2010
The activity of the heterodimeric transcription factor hypoxia inducible factor (HIF) is regulated by the post-translational, oxygen-dependent hydroxylation of its α-subunit by members of the prolyl hydroxylase domain (PHD or EGLN)-family and by factor inhibiting HIF (FIH). PHD-dependent hydroxylation targets HIFα for rapid proteasomal degradation; FIH-catalysed asparaginyl-hydroxylation of the C-terminal transactivation domain (CAD) of HIFα suppresses the CAD-dependent subset of the extensive transcriptional responses induced by HIF. FIH can also hydroxylate ankyrin-repeat domain (ARD) proteins, a large group of proteins which are functionally unrelated but share common structural features. Competition by ARD proteins for FIH is hypothesised to affect FIH activity towards HIFα; however the extent of this competition and its effect on the HIF-dependent hypoxic response are unknown.
To analyse if and in which way the FIH/ARD protein interaction affects HIF-activity, we created a rate equation model. Our model predicts that an oxygen-regulated sequestration of FIH by ARD proteins significantly shapes the input/output characteristics of the HIF system. The FIH/ARD protein interaction is predicted to create an oxygen threshold for HIFα CAD-hydroxylation and to significantly sharpen the signal/response curves, which not only focuses HIFα CAD-hydroxylation into a defined range of oxygen tensions, but also makes the response ultrasensitive to varying oxygen tensions. Our model further suggests that the hydroxylation status of the ARD protein pool can encode the strength and the duration of a hypoxic episode, which may allow cells to memorise these features for a certain time period after reoxygenation.
The FIH/ARD protein interaction has the potential to contribute to oxygen-range finding, can sensitise the response to changes in oxygen levels, and can provide a memory of the strength and the duration of a hypoxic episode. These emergent properties are predicted to significantly shape the characteristics of HIF activity in animal cells. We argue that the FIH/ARD interaction should be taken into account in studies of the effect of pharmacological inhibition of the HIF-hydroxylases and propose that the interaction of a signalling sensor with a large group of proteins might be a general mechanism for the regulation of signalling pathways.
In animals, the response to hypoxia is mediated by an α,β-heterodimeric transcription factor, the hypoxia inducible factor or HIF. In humans, there are three different HIFα isoforms, with HIF1α and HIF2α being better characterised than HIF3α. The HIFβ subunit is identical with the aryl hydrocarbon receptor nuclear translocator (ARNT). Both the level and transcriptional activity of HIF are regulated by post-translational hydroxylation of the HIFα, but not HIFβ, subunit. In the presence of sufficient oxygen, HIF1α and HIF2α undergo hydroxylation of two proline-residues in their oxygen-dependent degradation domain (ODD), reactions catalysed by three Fe(II)- and 2-oxoglutarate-dependent prolyl hydroxylase domain (PHD1-3 or EGLN1-3) enzymes . In healthy mammalian cells, PHD2 is the most important regulator of the hypoxic response as shown by cellular  and animal studies . HIF1α and HIF2α also undergo asparaginyl hydroxylation  of the C- terminal (CAD) of the two transactivation domains found in HIFα (CAD-hydroxylation). This reaction is catalysed by factor inhibiting HIF (FIH), which is also an Fe(II)- and 2-oxoglutarate-dependent oxygenase [5, 6]. HIFα prolyl hydroxylation by PHDs very substantially increases its binding to the von Hippel Lindau protein (pVHL), which acts as a targeting component for an E3 ubiquitin ligase complex and thus mediates rapid degradation of HIFα by the proteasome. When PHD catalysis is limited by oxygen availability, i.e. in hypoxia, HIFα degradation is slowed, its level rises, it dimerises with HIFβ and upregulates HIF-target gene transcription. In contrast to the PHD-dependent ODD-hydroxylation, FIH-dependent CAD-hydroxylation does not affect the stability of HIFα, but more directly decreases the transcriptional activity of HIF by blocking the recruitment of the transcriptional co-regulator p300/CBP to the CAD [4, 6], thus disrupting CAD-dependent target gene expression. In contrast, target genes that depend on the N- terminal transactivation domain (NAD) of HIFα are not affected by FIH activity . For reviews see [8–10].
More recently, it has become clear that HIFα is not the only FIH substrate, but that FIH also catalyses the hydroxylation of a wide range of other proteins [11–16]. With the notable exception of HIFα itself, all FIH substrates identified to date contain an ankyrin-repeat domain (ARD), an evolutionarily ancient structural domain found in all kingdoms of life . ARDs seem to predominantly mediate protein-protein interactions , and occur in proteins as diverse as signal transducers, ion channels, cell cycle regulators, transcriptional regulators and chromatin-associated proteins. ARD proteins contain varying numbers of ankyrin repeats (ARs). ARs are one of the most commonly occurring protein repeats in animals . The stereotypical AR consists of 33 amino acid residues and has an L-shaped fold, which is formed by two short α-helices, arranged in an anti-parallel fashion, and, perpendicular to the helices, a protruding loop region followed by a β-hairpin. The asparagine residue targeted by FIH in some ARs is located in the loop region. ARs stack together to form an ARD, which, in humans, can contain up to 28 ARs. Several studies suggest that AR-hydroxylation by FIH is widespread [11–16], however its biological significance is unclear. Studies with consensus ARDs suggest that hydroxylation may cause an increase in the thermodynamic stability of the ARD fold [19, 20], and some evidence points to a potential role for ARD hydroxylation in signalling crosstalk in the cases of Notch [13, 16] and NFκB/IκBα . Because the inhibition of HIFα CAD-dependent transcription remains the only well-defined functional outcome of the catalytic activity of FIH, ARD proteins have been speculated to fine-tune FIH activity towards HIFα by binding and sequestering FIH [12, 21]. The discovery that FIH interacts with multiple ARD proteins raises major questions as to the role of FIH as an oxygen sensor. To our knowledge, the proposal that the interaction of multiple proteins with a sensor has a regulatory role, is unprecedented. It is unclear what effect the competitive inhibition of FIH by ARD proteins would have on signal processing and on the input/output relation of the network. Because the proposed regulatory effect of ARD proteins on FIH involves multiple interactions it is difficult to study via classical approaches. We therefore devised and analysed a rate equation model of HIFα CAD-hydroxylation.
Our model predicts that the presence of ARD proteins and their hydroxylation by FIH can indeed fine-tune HIFα CAD-hydroxylation, provided that the affinity of FIH for ARD proteins is significantly weakened by their hydroxylation. The simulations highlight unexpected functional consequences of the FIH/ARD protein interaction for the hypoxic response: By creating an oxygen threshold, HIFα CAD-hydroxylation is predicted to be focused into a defined range of oxygen tensions (range finding mechanism), the signal/response curves of HIFα CAD-hydroxylation is predicted to become significantly sharpened (ultrasensitivity), and, upon reoxygenation, FIH-release is predicted to occur with a time-delay, the length of which depends on the duration and the strength of the preceding hypoxic period (memory effect).
A database of human ankyrin repeats (ARs)
The SMART , PFAM  and Uniprot  databases were searched for human AR sequences. SMART contains 1766 human ARs corresponding to 341 distinct protein entries, PFAM Version 24 contains 2337 ankyrin repeats corresponding to 646 protein entries in Uniprot. Sequences that are represented incompletely in the PFAM and SMART databases were extended to the canonical length of 33 residues. To eliminate redundancy, ARs corresponding to different entries for identical proteins were removed. All individual ARs were assembled into a database (Additional File 1), which includes the amino acid sequences of all repeats, the protein names and identifiers, their position of the AR within the ARD protein and a classification according to sequence motifs. ARs longer than 34 or shorter than 32 residues (less than 10% of repeats) were excluded from the set of sequences used to obtain the consensus sequence and the sequence logos.
Nomenclature of kinetic parameters and reaction species
List of model species.
unhydroxylated FIH target repeats
hydroxylated FIH target repeats
all FIH target ankyrin repeats
Dimensionless parameter values used for the Full Model.
Dissociation constant of PHD/HIFα binding
(k cat for ODD-hydroxylation)/(basal HIFα degradation rate constant)
Dissociation constant of FIH/HIFα binding
set to same as 2
(k cat for HIFα CAD-hydroxylation)/(k cat for HIFα ODD-hydroxylation)
(KM of FIH for oxygen)/(KM of PHD for oxygen)
total concentration of FIH target ARs
0 - 500
Dissociation constant of FIH/ARD protein binding (unhydroxylated ARs)
set to same as 2
(FIH affinity for hydroxylated)/(FIH affinity for unhydroxylated ARs)
0 - 0.1
(basal HIFα degradation rate constant)/(AR degradation rate constant)
1 - 10
Dissociation constant of HIF/HRE binding
0 - 1
Modelling and modelling assumptions
For the HIF-hydroxylases to act as oxygen sensors in the proposed manner, their activity in cells must be limited by oxygen availability. Moreover, to ensure that HIFα-levels and the amount of HIF bound to DNA reflect the intracellular oxygen tension at all times, ODD-hydroxylation must be rate-limiting rather than degradation of ODD-hydroxylated HIFα, complex formation with HIFβ or DNA-binding of HIF. These requirements allow us to make the following simplifying assumptions: Firstly, HIFβ and the hydroxylase co-substrates, 2-oxoglutarate and Fe(II), are not limiting, and secondly, the degradation of ODD-hydroxylated HIFα as well as binding of HIF to hypoxia response elements are fast compared to the ODD-hydroxylation reaction. Although we appreciate that under some conditions these assumptions may not be valid, for instance in some tumour cells [25, 26], a body of evidence suggests that these assumptions are reasonable for normal cells. All simulations were done using the open source software XPP-AUT . Steady state values were calculated by running time course simulations at different oxygen-tensions until a steady state was reached.
The Full Model
For the hydroxylation rate functions in the Full Model, we take into account that free HIFα concentration decreases by binding to the enzymes ("full model kinetics", see also Additional File 2). Unlike classical Michaelis-Menten kinetics, this approach is also valid if there is no substrate excess, a situation that is frequently encountered in protein-protein interaction networks . Because of the expected excess of ARD proteins over FIH, we use the Michaelis-Menten approximation for FIH-catalysed AR-hydroxylation. The Full Model is defined by three ordinary differential equations, which are given in dimensionless form. The CAD-hydroxylated forms of HIFα and the ARD proteins are defined by mass conservation. "Hat" (^) indicates non-dimensional quantities expressed relative to the maximal amount of HIFα present in the absence of oxygen, and "prime" (') indicates non-dimensional quantities expressed relative to the basal degradation rate constant for HIFα. Oxygen is given relative to the KM of PHD for oxygen, which is indicated by "tilde" (~). For details, refer to Additional File 2.
- 1.PHD-dependent HIFα ODD-hydroxylation()(4)(5)
- 2.FIH-dependent HIFα CAD-hydroxylation in the presence of competing ARs ()(6)(7)(8)
Skeleton Model 1 (SKM1)
the sum of which is the fraction of HREs occupied by either form. Here, is the dissociation constant of the HIF/HRE interaction, assuming all forms of HIF bind to DNA with the same affinity. These expressions can also be interpreted as the probability of a specific HRE to be occupied by either species.
Skeleton Model 2 (SKM2)
Experimental information restricts the biologically relevant range of parameter values and allows estimates. The KM of PHDs for oxygen has been reported to be in the range of approximately 220-250 μM  and references therein. This is slightly higher than the maximal solubility of oxygen in water, which sets a theoretical upper limit for intracellular oxygen concentrations. Thus, the biologically relevant range of oxygen-tensions must be below this value. The KM of FIH for oxygen has been reported to be lower than the KM of the PHDs , which we take into account. We normalise to the maximal amount of HIFα, , which is reached at steady state in the absence of oxygen-dependent degradation. Three parameters define HIFα ODD-hydroxylation, the dissociation constant of the PHD/HIFα interaction, ; the maximal reaction rate, ; and the PHD concentration, P tot . We used the reported value of 1 μM for the PHD/HIFα binding affinity , i.e. . We then chose such that a good agreement with measured signal/response curves was obtained . PHD expression levels relative to HIFα are not known, and the concentration of PHD was set to 0.2, an assumption which does not affect any of our conclusions. In our simulations, the overall rate of FIH hydroxylation must be higher than the overall rate of PHD hydroxylation for significant HIFα CAD-hydroxylation to occur. The means by which high FIH activity is achieved are irrelevant for our conclusions (data not shown), whether by higher expression levels, higher affinity for HIFα or a faster turnover rate. We thus introduce an arbitrary five-fold excess of FIH over PHD, which allows us to clearly illustrate the inhibitory effect of the ARD proteins on FIH activity. The binding affinity of FIH for HIFα and the turnover rate are set to match the values for PHD. In vivo, FIH activity might be lower than assumed in the model, however all our conclusions are qualitative and thus entirely independent of these assumptions. Table 2 summarises the dimensionless parameter values that were used for calculating the graphs shown, unless indicated otherwise in the figure legends. The model can be found in the BioModels Database, http://www.ebi.ac.uk/biomodels/ accession number MODEL1008170000.
Potential ankyrin-type FIH targets in the human proteome
To investigate the potential extent of ARD protein interaction with FIH, we initially carried out bioinformatic analyses. Searching the SMART , PFAM  and Uniprot  databases for human AR-sequences, we found 1505 annotated ankyrin repeats (ARs) mapping to 252 distinct human ARD proteins. All 1505 ARs were assembled into a database (Additional File 1) and analysed for the presence of a potential FIH hydroxylation site. Diagnostic features were then extracted from experimentally verified target sequences to aid the prediction of AR-type FIH substrates and give an estimate of their overall abundance.
Classification of ankyrin-repeats according to sequence motifs.
% of all AR
of which L(X)4(AC)(DEN)(ILV)N
Asn-repeats, non L-8N
Skeleton Model 1 - HIFα CAD-hydroxylation in the absence of the FIH/AR-interaction
Experimentally, CAD-dependent genes, but not NAD-dependent genes, are induced by FIH knockdown and repressed by FIH overexpression . Interestingly, a third group of HIF target genes have been described, which show the opposite behaviour to CAD-dependent genes. This group, one member of which is BNIP3 (BCL2/adenovirus E1B 19 kD interacting protein 3), is repressed by FIH knockdown and de-repressed by FIH overexpression [7, 35]. To explain this unexpected behaviour, a CAD-dependently expressed repressor was postulated, which would be present only at very low oxygen concentrations . Our model suggests a more parsimonious explanation, which can explain the unusual behaviour of some HIF target genes. We propose that BNIP3 belongs to a third, novel class of target genes, which are specifically activated by the CAD-hydroxylated HIFα (CADOH-dependent, Figure 3B, red shading), and thus expressed in a p300/CBP-independent manner. This proposal assigns a potential direct function to CAD-hydroxylated HIFα, and we studied the behaviour of this species in more detail. Because FIH activity is a monotonically increasing function of oxygen, the naive expectation is that the amount of CAD-hydroxylated HIFα should also increase monotonically (at least until a saturation point) with increasing oxygen-levels. The model predicts however that this is not the case and that the level of CAD-hydroxylated HIFα will peak at intermediate oxygen tensions. The location and magnitude of the peak of CAD-hydroxylated HIFα are parameter dependent, but its existence is a generic system property. The postulated group of CADOH-dependent genes is thus predicted to be expressed at intermediate hypoxia only (Figure 3B), and to show the bell-shaped signal/response curve observed experimentally .
Skeleton Model 2 - FIH sequestration by ARD proteins and oxygen-dependent FIH-release
In addition to this steady state analysis, we also simulated the temporal response of the system to step changes in oxygen concentrations, i.e. to sudden hypoxia and to sudden reoxygenation after a hypoxic episode. SKM2 predicts that the time-resolved response to hypoxia is rather insensitive to variations of κ and β, with the family of curves showing a hyperbolic decrease in free FIH with time, either reaching distinct steady state values at low oxygen when κ is varied (Additional File 3, Figure S1A), or starting out from distinct steady state values at high oxygen if β is varied (Additional File 3, Figure S1B). More interestingly however, the temporal response to an increase of oxygen after hypoxia (reoxygenation) is affected significantly by both, κ and β. Although variations in κ affect half-response times for FIH-release only moderately (Figure 4C), varying κ does significantly mould the shape of the time-response curves, which change from a gradual FIH-release if κ is small, to a delayed, switch-like FIH-release if κ is large (Figure 4C). In contrast to varying κ, varying β substantially affects half response times - the larger β, the faster the FIH release in response to reoxygenation (Figure 4D). Thus, SKM2 predicts that oxygen-dependent FIH-release from ARD proteins upon sudden reoxygenation of hypoxic cells can occur in a switch-like manner, with a time delay relative to the reoxygenation event. The reason for this time delay is that, shortly after reoxygenation, the concentration of unhydroxylated, FIH-accessible ARs is still high enough to allow rebinding of any released FIH. SKM2 thus predicts that a permanent release of FIH is only achieved once the FIH-accessible ARs are hydroxylated to a significant extent, and the concentration of unhydroxylated ARs starts to become limiting for FIH sequestration, which, due to the excess of ARs over FIH, can only happen if hydroxylation of FIH-accessible ARs approaches completion. The time required for FIH to hydroxylate ARs to a sufficient degree to overcome sequestration explains the time delay between reoxygenation and FIH-release. This behaviour also allows modulation of the timing of FIH-release as a function of the strength and duration of the preceding hypoxic episode, a feature we explore using the Full Model at the end of the following section.
The Full Model - the effects of the FIH/ARD protein interaction on HIFα CAD-hydroxylation
Up to now we have assumed that FIH does not bind at all to hydroxylated ARs. To make the Full Model more realistic, we now allow such product binding to happen and introduce a parameter 0 < γ < 1, which is the binding affinity of FIH for hydroxylated ARs relative to its affinity for unhydroxylated ARs. It becomes clear that even weak binding of FIH to hydroxylated ARs (γ < 1) can strongly attenuate the levels of free FIH at high oxygen tensions (Figure 5B, left hand panel). The peak values of CAD-hydroxylated HIFα (Figure 5B, right hand panel) are also decreased. Importantly however, while binding of FIH to hydroxylated ARs attenuates the steepness of the response curve, the existence of an oxygen threshold is entirely independent of such binding. Experiments suggest that there is indeed a substantial differential between the FIH binding affinity for non-CAD-hydroxylated versus CAD-hydroxylated ARs , and we predict that this difference is essential for achieving significant HIFα CAD-hydroxylation and the sharpest possible response curves.
FIH-dependent Asn-hydroxylation is, to the best of our knowledge, irreversible and can only be reversed indirectly by degradation of the Asn-hydroxylated proteins and de novo synthesis of unhydroxylated proteins. Thus, the mean life times of the competing substrates, ARD-proteins and HIFα, are expected to be important parameters. To test this, we varied the parameter ε, which is the ratio of the mean life time of ARD-proteins relative to the mean life time of HIFα under basal turnover conditions, i.e. in the absence of oxygen. The Full Model predicts that long-lived ARD proteins (large ε ) decrease the oxygen-tension at which FIH-release occurs (Figure 5C, left hand panel). Consequently, if ε is large, FIH is already released at oxygen-tensions at which PHD-activity is still only moderate, and a substantial amount of HIFα is available for CAD-hydroxylation, leading to increased peak values of CAD-hydroxylated HIFα and a sharper response (Figure 5C, right hand panel). The exact value of ε then determines both the degree of hypoxia at which HIFα CAD-hydroxylation can occur, and the extent to which CAD-hydroxylated HIFα can accumulate.
In summary, steady-state simulations with the Full Model suggest that, in the absence of other variables, a low binding affinity of FIH for already hydroxylated ARs, as well as a fast basal turnover of HIFα compared to the average turnover of ARD proteins are essential requirements for efficient HIFα CAD-hydroxylation. The model predicts two functionally significant effects of oxygen-dependent FIH-release from ARD proteins, which are emergent system properties that shape the hypoxic response: The process of HIFα CAD-hydroxylation is focused into a defined range of oxygen-tensions (range-finding mechanism), and the signal/response curves are significantly sharpened (ultrasensitivity). Time course simulations further suggest that a delay in FIH-release in response to reoxygenation provides a readout of the hydroxylation status of the ankyrin-pool, which in turn encodes or "memorises" duration and strength of the preceding hypoxic episode.
Several kinetic models of the HIF-pathway have been reported, each focusing on different aspects of this signalling system [35–38]. Only one of these models considers FIH and HIFα CAD-hydroxylation, but this does not include ARD proteins as competing FIH substrates . Predictions derived from SKM1, which ignores the FIH/AR-interaction, support previous experimental  and theoretical work . Despite its simplicity, SKM1 successfully reproduces the findings that, first, expression of CAD-dependent genes requires more severe hypoxia when compared to NAD-dependent genes, and second, that higher FIH activity increases the magnitude of this shift towards severe hypoxia (Figure 3). These features are generic properties that are valid whether or not the FIH/AR interaction is taken into account. Our proposal that there is an additional group of target genes that relies on the hydroxylated CAD for transcriptional activation in a CBP/P300-independent manner provides a possible explanation for the unusual activation profile of HIF target genes such as BNIP3. The existence of such a target gene group however requires experimental verification. Because FIH needs to compete with PHD-induced degradation of HIFα, SKM1 predicts that HIFα CAD-hydroxylation can only occur to a significant extent if FIH-dependent hydroxylation is at least as effective as PHD-dependent hydroxylation (see α = 1 in Figure 3A for identical activities of FIH and PHDs). This is true even in the hypothetical absence of an inhibitory FIH/AR-interaction, i.e. if FIH-activity towards HIFα is constitutive and maximal. Indeed, FIH has been suggested to be a more efficient enzyme than the PHDs because of its lower KM for oxygen, for which there is in vitro evidence . Our SKM1 predictions modify this notion, indicating that efficient HIFα CAD-hydroxylation does not specifically depend on the KM of FIH for oxygen, but more generally on intracellular FIH activity. Expression levels of FIH relative to PHDs thus are predicted to play a major role in vivo in determining the extent of HIFα CAD-hydroxylation at a given oxygen tension, and accurate measurements of relative expression levels of FIH and PHDs are desirable.
The focus of our work was on the regulation of FIH activity towards HIFα through competitive inhibition by ARD proteins. The large number of ARD proteins, a substantial fraction of which we predict to be FIH targets, and the fact that at least some isolated ankyrin-domains bind more tightly to FIH than to HIFα [13, 39] make it seem possible that ARs will out-compete HIFα and prevent significant HIFα CAD-hydroxylation irrespective of the oxygen concentration. However, HIFα clearly is CAD-hydroxylated by FIH in vivo, and a conclusive theory needs to consolidate this fact with the presence and the action of FIH-binding ARD proteins. Several factors can potentially contribute to weakening the competition by ARD proteins for FIH. First, not all ARD proteins are expected to be expressed in a particular cell type. Second, the ARD is a protein-protein interaction motif, and only a fraction of a given ARD protein is expected to be accessible to FIH in vivo. Third, affinity constants measured in vitro with isolated protein domains might be misleading. These considerations make it difficult to give an estimate of a realistic concentration of FIH binding target repeats, and FIH sequestration could be much less pronounced than expected from idealised theoretical considerations. Irrespective of its actual extent however, FIH sequestration by ARD proteins must be oxygen-regulated in order to tune HIFα CAD-hydroxylation, other than just repressing it at all oxygen tensions, and experimental evidence points to hydroxylation-dependent release of FIH from ARD-proteins as the responsible mechanism. SKM2 predicts that such FIH-release only occurs if hydroxylation of FIH-accessible ARs approaches completion. Importantly, this prediction is not in contrast with experimental findings indicating that the hydroxylation of an individual AR is often far from complete, even under conditions of high oxygen [11, 13, 15]. Because the ARD is a protein-protein interaction domain, the access of FIH to the ARDs will be restricted by the presence of ARD-interactors other than FIH, which will partly protect ankyrin repeats from FIH-dependent hydroxylation. Thus, it is possible that near-complete hydroxylation of a specific AR may not be observed experimentally, but the hydroxylation of its FIH-accessible fraction might still approach completion.
We identify several potential functional effects of the FIH/AR-interaction on HIFα CAD-hydroxylation. The Full Model predicts that the FIH/AR-interaction introduces an oxygen-threshold, below which HIFα CAD-hydroxylation is marginal and CAD-dependent genes are fully active. The exact threshold concentration of oxygen, above which HIFα CAD-hydroxylation can occur and CAD-dependent genes are turned off, is predicted to be determined by the total amount of FIH-accessible, hydroxylatable ARs, their binding affinity for FIH, as well as by the relative turnover rates of HIFα and ARD proteins and should thus be tunable by modulating these system features. A range finding property is essential for the HIF hydroxylases to act as oxygen sensors in environments with different physiologically relevant ranges of oxygen concentrations . Although it is unlikely that the FIH/AR-interaction is the only range-finding mechanism in the HIF system, our results demonstrate that it is one possible such mechanism. The change in CAD-hydroxylation at the oxygen threshold is significantly sharpened compared to the absence of an FIH/AR-interaction. Such ultrasensitivity, as characterised by a sigmoid signal/response curve, is an important feature of many signalling pathways and cellular decision making processes. Mechanisms giving rise to ultrasensitive behaviour include cooperativity, multi-site modification, zero-order ultrasensitivity, and positive feedback. In our case, the mechanism causing ultrasensitivity is similar to a proposed "ultrasensitivity by substrate competition", where stoichiometric inhibition of an enzyme by a competing substrate can make enzyme activity for other substrates nonlinearly dependent on the enzyme level . Finally, time course simulations using SKM2 (Figure 4) and the Full Model (Figure 6) predict that FIH-release can occur in a switch-like fashion with a time delay after reoxygenation of hypoxic cells. The length of this delay depends on the intensity and the duration of the preceding hypoxic period. Functionally, this time delay in FIH-release can prevent a premature FIH-release during a brief and perhaps only transient reoxygenation event, and only sustained reoxygenation will trigger bulk FIH-release.
Overall, the combined modelling results reveal that the interaction of multiple ARD proteins with FIH has the potential to significantly input on the dynamics of the hypoxic response in human cells. The FIH/AR-interaction can provide a mechanism by which the oxygen-threshold for the hypoxic response can be varied (range finding mechanism), it can confer substantial sharpening of the signal/response curves (ultrasensitivity), and it can create a time-delay for CAD-hydroxylation after reoxygenation, the length of which can encode the strength and duration of the preceding hypoxic episode (a memory effect). The modulation of HIF hydroxylase activity is of considerable interest with respect to therapeutic intervention. Inhibition of the HIF hydroxylases may be useful for treatment of anaemia and ischemic disease, via the upregulation of erythropoiesis and angiogenesis, respectively. Presently, it is unclear whether HIF hydroxylase inhibitors should target individual HIF hydroxylases or combinations of enzymes. Our results reveal that the FIH/AR-interactions should be taken into account in analyses of the cellular and physiological effects of HIF hydroxylase inhibitors, whether or not these inhibitors are selective for PHDs and/or FIH. Interfering with the binding of FIH to ARD-proteins is predicted to increase HIFα CAD-hydroxylation, whereas interfering with AR-hydroxylation by FIH is predicted to decrease HIFα CAD-hydroxylation. Thus, the FIH/AR-interaction itself is a promising potential target for pharmacological modulation of the HIF pathway. Finally, we note that if the concept of multiple proteins regulating a signalling sensor occurs in the oxygen sensing HIF-system, it is likely to occur in other signalling pathways.
List of Abbreviations Used
ankyrin repeat domain
C-terminal transactivation domain
Asn-hydroxylated C-terminal transactivation domain
egl nine homolog
factor inhibiting HIF
hypoxia inducible factor
hypoxia response element
N-terminal transactivation domain
oxygen-dependent degradation domain
prolyl hydroxylase domain.
We thank Ming Yang for sharing unpublished results and for discussions, Peter Ratcliffe for discussions, and Orsolya Kapuy and Maria Rosa Domingo Sananes for critically reading the manuscript. This work was funded by the Biotechnology and Biological Sciences Research Council (BBSRC) and the European Union.
- Schofield CJ, Ratcliffe PJ: Signalling hypoxia by HIF hydroxylases. Biochem Biophys Res Commun. 2005, 338: 617-626. 10.1016/j.bbrc.2005.08.111View ArticlePubMedGoogle Scholar
- Berra E, Benizri E, Ginouves A, Volmat V, Roux D, Pouyssegur J: HIF prolyl-hydroxylase 2 is the key oxygen sensor setting low steady-state levels of HIF-1alpha in normoxia. EMBO J. 2003, 22: 4082-4090. 10.1093/emboj/cdg392PubMed CentralView ArticlePubMedGoogle Scholar
- Takeda K, Ho VC, Takeda H, Duan LJ, Nagy A, Fong GH: Placental but not heart defects are associated with elevated hypoxia-inducible factor alpha levels in mice lacking prolyl hydroxylase domain protein 2. Mol Cell Biol. 2006, 26: 8336-8346. 10.1128/MCB.00425-06PubMed CentralView ArticlePubMedGoogle Scholar
- Lando D, Peet DJ, Whelan DA, Gorman JJ, Whitelaw ML: Asparagine hydroxylation of the HIF transactivation domain a hypoxic switch. Science. 2002, 295: 858-861. 10.1126/science.1068592View ArticlePubMedGoogle Scholar
- Hewitson KS, McNeill LA, Riordan MV, Tian YM, Bullock AN, Welford RW, Elkins JM, Oldham NJ, Bhattacharya S, Gleadle JM, et al.: Hypoxia-inducible factor (HIF) asparagine hydroxylase is identical to factor inhibiting HIF (FIH) and is related to the cupin structural family. J Biol Chem. 2002, 277: 26351-26355. 10.1074/jbc.C200273200View ArticlePubMedGoogle Scholar
- Lando D, Peet DJ, Gorman JJ, Whelan DA, Whitelaw ML, Bruick RK: FIH-1 is an asparaginyl hydroxylase enzyme that regulates the transcriptional activity of hypoxia-inducible factor. Genes Dev. 2002, 16: 1466-1471. 10.1101/gad.991402PubMed CentralView ArticlePubMedGoogle Scholar
- Dayan F, Roux D, Brahimi-Horn MC, Pouyssegur J, Mazure NM: The oxygen sensor factor-inhibiting hypoxia-inducible factor-1 controls expression of distinct genes through the bifunctional transcriptional character of hypoxia-inducible factor-1alpha. Cancer Res. 2006, 66: 3688-3698. 10.1158/0008-5472.CAN-05-4564View ArticlePubMedGoogle Scholar
- Kaelin WG, Ratcliffe PJ: Oxygen sensing by metazoans: the central role of the HIF hydroxylase pathway. Mol Cell. 2008, 30: 393-402. 10.1016/j.molcel.2008.04.009View ArticlePubMedGoogle Scholar
- Lendahl U, Lee KL, Yang H, Poellinger L: Generating specificity and diversity in the transcriptional response to hypoxia. Nat Rev Genet. 2009, 10: 821-832. 10.1038/nrg2665View ArticlePubMedGoogle Scholar
- Semenza GL: Life with oxygen. Science. 2007, 318: 62-64. 10.1126/science.1147949View ArticlePubMedGoogle Scholar
- Cockman ME, Lancaster DE, Stolze IP, Hewitson KS, McDonough MA, Coleman ML, Coles CH, Yu X, Hay RT, Ley SC, et al.: Posttranslational hydroxylation of ankyrin repeats in IkappaB proteins by the hypoxia-inducible factor (HIF) asparaginyl hydroxylase, factor inhibiting HIF (FIH). Proc Natl Acad Sci USA. 2006, 103: 14767-14772. 10.1073/pnas.0606877103PubMed CentralView ArticlePubMedGoogle Scholar
- Cockman ME, Webb JD, Kramer HB, Kessler BM, Ratcliffe PJ: Proteomics-based identification of novel factor inhibiting hypoxia-inducible factor (FIH) substrates indicates widespread asparaginyl hydroxylation of ankyrin repeat domain-containing proteins. Mol Cell Proteomics. 2009, 8: 535-546. 10.1074/mcp.M800340-MCP200PubMed CentralView ArticlePubMedGoogle Scholar
- Coleman ML, McDonough MA, Hewitson KS, Coles C, Mecinovic J, Edelmann M, Cook KM, Cockman ME, Lancaster DE, Kessler BM, et al.: Asparaginyl hydroxylation of the Notch ankyrin repeat domain by factor inhibiting hypoxia-inducible factor. J Biol Chem. 2007, 282: 24027-24038. 10.1074/jbc.M704102200View ArticlePubMedGoogle Scholar
- Ferguson JE, Wu Y, Smith K, Charles P, Powers K, Wang H, Patterson C: ASB4 is a hydroxylation substrate of FIH and promotes vascular differentiation via an oxygen-dependent mechanism. Mol Cell Biol. 2007, 27: 6407-6419. 10.1128/MCB.00511-07PubMed CentralView ArticlePubMedGoogle Scholar
- Webb JD, Muranyi A, Pugh CW, Ratcliffe PJ, Coleman ML: MYPT1, the targeting subunit of smooth-muscle myosin phosphatase, is a substrate for the asparaginyl hydroxylase factor inhibiting hypoxia-inducible factor (FIH). Biochem J. 2009, 420: 327-333. 10.1042/BJ20081905View ArticlePubMedGoogle Scholar
- Zheng X, Linke S, Dias JM, Gradin K, Wallis TP, Hamilton BR, Gustafsson M, Ruas JL, Wilkins S, Bilton RL, et al.: Interaction with factor inhibiting HIF-1 defines an additional mode of cross-coupling between the Notch and hypoxia signaling pathways. Proc Natl Acad Sci USA. 2008, 105: 3368-3373. 10.1073/pnas.0711591105PubMed CentralView ArticlePubMedGoogle Scholar
- Sedgwick SG, Smerdon SJ: The ankyrin repeat: a diversity of interactions on a common structural framework. Trends Biochem Sci. 1999, 24: 311-316. 10.1016/S0968-0004(99)01426-7View ArticlePubMedGoogle Scholar
- Li J, Mahajan A, Tsai MD: Ankyrin repeat: a unique motif mediating protein-protein interactions. Biochemistry. 2006, 45: 15168-15178. 10.1021/bi062188qView ArticlePubMedGoogle Scholar
- Hardy AP, Prokes I, Kelly L, Campbell ID, Schofield CJ: Asparaginyl beta-hydroxylation of proteins containing ankyrin repeat domains influences their stability and function. J Mol Biol. 2009, 392: 994-1006. 10.1016/j.jmb.2009.07.070View ArticlePubMedGoogle Scholar
- Kelly L, McDonough MA, Coleman ML, Ratcliffe PJ, Schofield CJ: Asparagine beta-hydroxylation stabilizes the ankyrin repeat domain fold. Mol Biosyst. 2009, 5: 52-58. 10.1039/b815271cView ArticlePubMedGoogle Scholar
- Cockman ME, Webb JD, Ratcliffe PJ: FIH-dependent asparaginyl hydroxylation of ankyrin repeat domain-containing proteins. Ann N Y Acad Sci. 2009, 1177: 9-18. 10.1111/j.1749-6632.2009.05042.xView ArticlePubMedGoogle Scholar
- SMART Database. http://smart.embl.de/
- PFAM Database. http://pfam.org
- Uniprot Database. http://uniprot.org
- Pollard PJ, Briere JJ, Alam NA, Barwell J, Barclay E, Wortham NC, Hunt T, Mitchell M, Olpin S, Moat SJ, et al.: Accumulation of Krebs cycle intermediates and over-expression of HIF1alpha in tumours which result from germline FH and SDH mutations. Hum Mol Genet. 2005, 14: 2231-2239. 10.1093/hmg/ddi227View ArticlePubMedGoogle Scholar
- Zhao S, Lin Y, Xu W, Jiang W, Zha Z, Wang P, Yu W, Li Z, Gong L, Peng Y, et al.: Glioma-derived mutations in IDH1 dominantly inhibit IDH1 catalytic activity and induce HIF-1alpha. Science. 2009, 324: 261-265. 10.1126/science.1170944PubMed CentralView ArticlePubMedGoogle Scholar
- XPP-AUT Homepage. http://www.math.pitt.edu/~bard/xpp/xpp.html
- Borghans JA, de Boer RJ, Segel LA: Extending the quasi-steady state approximation by changing variables. Bull Math Biol. 1996, 58: 43-63. 10.1007/BF02458281View ArticlePubMedGoogle Scholar
- Ehrismann D, Flashman E, Genn DN, Mathioudakis N, Hewitson KS, Ratcliffe PJ, Schofield CJ: Studies on the activity of the hypoxia-inducible-factor hydroxylases using an oxygen consumption assay. Biochem J. 2007, 401: 227-234. 10.1042/BJ20061151PubMed CentralView ArticlePubMedGoogle Scholar
- Koivunen P, Hirsila M, Gunzler V, Kivirikko KI, Myllyharju J: Catalytic properties of the asparaginyl hydroxylase (FIH) in the oxygen sensing pathway are distinct from those of its prolyl 4-hydroxylases. J Biol Chem. 2004, 279: 9899-9904. 10.1074/jbc.M312254200View ArticlePubMedGoogle Scholar
- Jiang BH, Semenza GL, Bauer C, Marti HH: Hypoxia-inducible factor 1 levels vary exponentially over a physiologically relevant range of O2 tension. Am J Physiol. 1996, 271: C1172-1180.PubMedGoogle Scholar
- Crooks GE, Hon G, Chandonia JM, Brenner SE: WebLogo: a sequence logo generator. Genome Res. 2004, 14: 1188-1190. 10.1101/gr.849004PubMed CentralView ArticlePubMedGoogle Scholar
- Coleman ML, Ratcliffe PJ: Signalling Cross Talk of the HIF System: Involvement of the FIH Protein. Curr Pharm Des. 2009, 15: 3904-3907. 10.2174/138161209789649448View ArticlePubMedGoogle Scholar
- Webb JD, Coleman ML, Pugh CW: Hypoxia, hypoxia-inducible factors (HIF), HIF hydroxylases and oxygen sensing. Cell Mol Life Sci. 2009, 66: 3539-3554. 10.1007/s00018-009-0147-7View ArticlePubMedGoogle Scholar
- Dayan F, Monticelli M, Pouyssegur J, Pecou E: Gene regulation in response to graded hypoxia: the non-redundant roles of the oxygen sensors PHD and FIH in the HIF pathway. J Theor Biol. 2009, 259: 304-316. 10.1016/j.jtbi.2009.03.009View ArticlePubMedGoogle Scholar
- Kohn KW, Riss J, Aprelikova O, Weinstein JN, Pommier Y, Barrett JC: Properties of switch-like bioregulatory networks studied by simulation of the hypoxia response control system. Mol Biol Cell. 2004, 15: 3042-3052. 10.1091/mbc.E03-12-0897PubMed CentralView ArticlePubMedGoogle Scholar
- Qutub AA, Popel AS: A computational model of intracellular oxygen sensing by hypoxia-inducible factor HIF1 alpha. J Cell Sci. 2006, 119: 3467-3480. 10.1242/jcs.03087PubMed CentralView ArticlePubMedGoogle Scholar
- Qutub AA, Popel AS: Three autocrine feedback loops determine HIF1 alpha expression in chronic hypoxia. Biochim Biophys Acta. 2007, 1773: 1511-1525. 10.1016/j.bbamcr.2007.07.004PubMed CentralView ArticlePubMedGoogle Scholar
- Wilkins SE, Hyvarinen J, Chicher J, Gorman JJ, Peet DJ, Bilton RL, Koivunen P: Differences in hydroxylation and binding of Notch and HIF-1alpha demonstrate substrate selectivity for factor inhibiting HIF-1 (FIH-1). Int J Biochem Cell Biol. 2009, 41: 1563-1571. 10.1016/j.biocel.2009.01.005View ArticlePubMedGoogle Scholar
- Kim SY, Ferrell JE: Substrate competition as a source of ultrasensitivity in the inactivation of Wee1. Cell. 2007, 128: 1133-1145. 10.1016/j.cell.2007.01.039View ArticlePubMedGoogle Scholar
- Mosavi LK, Minor DL, Peng ZY: Consensus-derived structural determinants of the ankyrin repeat motif. Proc Natl Acad Sci USA. 2002, 99: 16029-16034. 10.1073/pnas.252537899PubMed CentralView ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.