HIF prolyl hydroxylase inhibition prior to transient focal cerebral ischaemia is neuroprotective in mice



This study investigated the effects of 2-(1-chloro-4-hydroxyisoquinoline-3-carboxamido) acetic acid (IOX3), a selective small molecule inhibitor of hypoxia-inducible factor (HIF) prolyl hydroxylases, on mouse brains subject to transient focal cerebral ischaemia. Male, 8- to 12-week-old C57/B6 mice were subjected to 45 min of middle cerebral artery occlusion (MCAO) either immediately or 24 h after receiving IOX3. Mice receiving IOX3 at 20 mg/kg 24 h prior to the MCAO had better neuroscores and smaller blood–brain barrier (BBB) disruption and infarct volumes than mice receiving the vehicle, whereas those having IOX3 at 60 mg/kg showed no significant changes. IOX3 treatment immediately before MCAO was not neuroprotective. IOX3 up-regulated HIF-1α, and increased EPO expression in mouse brains. In an in vitro BBB model (RBE4 cell line), IOX3 up-regulated HIF-1α and delocalized ZO-1. Pre-treating IOX3 on RBE4 cells 24 h before oxygen–glucose deprivation had a protective effect on endothelial barrier preservation with ZO-1 being better localized, while immediate IOX3 treatment did not. Our study suggests that HIF stabilization with IOX3 before cerebral ischaemia is neuroprotective partially because of BBB protection, while immediate application could be detrimental. These results provide information for studies aimed at the therapeutic activation of HIF pathway for neurovascular protection from cerebral ischaemia.


We show that IOX3, a selective small molecule (280.66 Da) HIF prolyl hydroxylase inhibitor, could up-regulate HIF-1α and increase erythropoietin expression in mice. We further demonstrate that HIF stabilization with IOX3 before cerebral ischaemia is neuroprotective partially because of blood–brain barrier (BBB) protection, while immediate application is detrimental both in vivo and in vitro. These findings provide new insights into the role of HIF stabilization in ischaemic stroke.

Abbreviations used

blood–brain barrier


cerebral blood flow








external carotid artery


middle cerebral artery occlusion


oxygen–glucose deprivation

The hypoxia-inducible factors (HIFs) are transcription factors that regulate gene expression in response to cellular hypoxia (Semenza 1999; Pugh and Ratcliffe 2003; Kaelin and Ratcliffe 2008; Greer et al. 2012). A mechanism by which the human HIFs ‘sense’ oxygen levels is enabled by four oxygen-sensitive hydroxylases: three prolyl hydroxylases (PHDs) and one asparaginyl hydroxylase, factor inhibiting HIF (FIH) (Schofield and Ratcliffe 2004). In normoxia, hydroxylation of either of two prolyl residues in the oxygen-dependent degradation domains of HIF-1α promotes its interaction with the von Hippel–Lindau ubiquitin E3 ligase complex. Thereafter, HIF-1α is targeted for ubiquitination, and subsequent proteasomal degradation (Semenza 1999; Pugh and Ratcliffe 2003; Kaelin and Ratcliffe 2008; Greer et al. 2012). In addition, hydroxylation of an asparaginyl residue in the C-terminal transcriptional activation domain (CAD) of HIF-1α reduces the association of HIF-1α with transcriptional coactivator proteins, thus inhibiting HIF-mediated transcription (Lando et al. 2002). In hypoxia, activity of the HIF hydroxylases is reduced, and levels of transcriptionally active HIF rise, causing induction of a gene array that contributes to cell protection via multiple mechanisms, including angiogenesis, vascular remodelling, metabolic regulation and erythropoiesis (Manalo et al. 2005). Inhibition of HIF hydroxylases has been shown to activate HIF and protect the adult rat kidney, mouse bowel and heart, rat and mouse brain from ischaemic and oxidative stress-induced injury without toxic effects upon systemic administration (Siddiq et al. 2005; Bernhardt et al. 2006, 2010; Baranova et al. 2007; Fraisl et al. 2009).

We have previously applied dimethyl-oxalylglycine (DMOG), a prodrug diester form of the broad-spectrum 2-oxoglutarate (2-OG) oxygenase inhibitor (N-oxaylglycine), to rats subjected to focal cerebral ischaemia (Nagel et al. 2011). We found that DMOG protected the rat brains from ischaemia/reperfusion injury and up-regulated a number of HIF-regulated genes and proteins, but that the observed protection was probably not simply related to HIF-1α levels present at the time animals were killed (Nagel et al. 2011). The liberated form of DMOG, N-oxaylglycine, inhibits many 2-OG-dependent enzymes (including chromatin-modifying enzymes) (Rose et al. 2012) and may therefore affect other signalling pathways and important cellular processes, e.g. the nuclear factor erythroid 2-related pathway for activating antioxidant gene expression in microvascular endothelial cells (Natarajan et al. 2009).

The above study prompted us to undertake further research employing a much more selective inhibitor of the HIF PHDs in a mouse model of ischaemic stroke to evaluate its effects on cerebral ischaemia. 2-(1-chloro-4-hydroxyisoquinoline-3-carboxamido) acetic acid (IOX3) (Thalhammer et al. 2012) is a member of a novel class of potent small molecules that inhibit HIF PHDs (Stubbs et al. 2009; Tian et al. 2011). The molecular structure of IOX3 (mass 280.66 Da) is believed to be identical to that of FG-2216 (Rose et al. 2011). FG2216 induces erythropoietin (EPO) production in mice and rhesus macaques (Hsieh et al. 2007), as well as in healthy human subjects and haemodialysis patients (Bernhardt et al. 2010). This compound has been studied under different names [bicyclic isoquinolinyl inhibitor (BIQ or BIC), 2-(1-chloro-4-hydroxyisoquinoline-3-carboxamido) acetate] by different research groups (Stubbs et al. 2009; Leung et al. 2010; Wang et al. 2012). The aim of our study was to define the effects of IOX3 on mouse brain following transient focal cerebral ischaemia, and to investigate the molecular mechanisms for these effects.


Synthesis of IOX3

IOX3 was synthesized as reported in Stubbs et al. (2009). IOX3 was dissolved first in dimethylsulfoxide (DMSO) and then in Fisher Scientific Buffer (Fisher scientific UK Ltd, Loughborough, UK) pH 7.0 (5% DMSO/buffer).

In vivo transient focal cerebral ischaemia

Transient focal ischaemia

A single dose of IOX3 (20 or 60 mg/kg) or vehicle (5% DMSO/buffer) was given to male, 8- to 12-week-old C57/B6 mice through a tail vein injection. One day or immediately after IOX3 injection, the mice were subjected to 45 min of middle cerebral artery occlusion (MCAO) under anaesthesia with 1.5% isoflurane in O2/N2O (1 : 3) as described (Chen et al. 2012). Briefly, under the operating microscope, the right common carotid artery, the right external carotid artery (ECA) and the right internal carotid artery (ICA) were isolated. A silicone rubber-coated monofilament (Doccol Corp., Redlands, CA, USA) was introduced into the ECA and pushed up the ICA until resistance was felt, effectively blocking the middle cerebral artery. Transcranial measurements of cerebral blood flow (CBF) were made by laser-Doppler flowmetry (LDF) (Oxford Optronix, Oxfordshire, UK). The surgical procedure was considered adequate if ≥ 70% reduction in regional CBF (rCBF) occurred immediately after placement of the intraluminal occluding suture; otherwise, mice were excluded. The suture remained inserted for 45 min, after which it was removed to allow reperfusion over the ICA and the circle of Willis and the ECA was permanently tied. Sham-operated (SO) mice underwent the same anaesthesia and surgical regime; however, the monofilament was only temporarily inserted into the intracranial portion of the internal carotid and withdrawn immediately to control for direct damage of the arterial intima. All procedures were in accordance with the UK Home Office Animals (Scientific Procedures) Act 1986.

Temperature regulation

Mice were implanted with intraabdominal radiofrequency probes (TA10TA-F20; DSI, St. Paul, MN, USA) 7 days before MCAO. Core temperature was sampled every 20 s using receivers (RLA-1020; Data Sciences Int., St. Paul, MN, USA) interfaced to a computer running ART 2.2 (DSI). This telemetry system allows temperature monitoring/control in the freely moving animal (Chen et al. 2012).


Twenty four hours after the surgery, mouse behaviour was assessed as described (Chen et al. 2012). Behavioural assessment consisted of scoring: forelimb flexion, reduced resistance to lateral push, gait towards the paretic side and rotational behaviour (Bederson et al. 1986; Barber et al. 2004; Chen et al. 2012). Scoring was performed by a trained individual unaware of treatment allocation.

Tissue processing

Mice were given sodium pentobarbital (70 mg/kg intraperitoneally) 24 h after MCAO, and were perfused with chilled (4°C) phosphate-buffered saline (PBS), followed by 4% formalin (in PBS). The brains were then removed and stored in chilled (4°C) 10% formalin (in PBS) before embedding in paraffin wax. Representative coronal 6- and 10-μm sections were cut from each 1-mm slice of embedded brain along the rostral-caudal axis using a microtome.

Infarction volume measurement

Nissl stain of 10-μm sections from each 1 mm of brain slice was used for assessing infarction volume (Chen et al. 2012). Photomicrographs were obtained with a Nikon Eclipse E110M microscope (Nikon, Surrey, UK). Areas of infarction were delineated by a blinded investigator using a NIS elements imaging software (Nikon). The infarction volume was calculated as the sum of section volumes made by multiplying the area of infarction on each section by the thickness of the corresponding slice. The infarction volume was corrected for oedema using the following equation: corrected infarct volume = total infarct volume − [(right (ipsilateral) hemisphere volume − left (contralateral) hemisphere volume)]. The infarction volume was presented as a percentage of the contralateral hemisphere.

TUNEL assay for the detection of apoptotic cells

The terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed on neighbouring 6-μm sections using an ApopTag® Plus Fluorescein in situ Apoptosis Detection Kit (Millipore, Temecula, CA, USA) according to the manufacturer's protocol. The sections were mounted with medium containing 4,6-diamidino-2-phenylindole (DAPI; Vector Laboratories, Peterborough, UK) as a counter-stain. Fluorescence was detected using a Nikon Eclipse E110M fluorescent microscope (Nikon) and images were processed using the NIS elements imaging software (Nikon). Four standard non-overlapping high power fields from the ischaemia boundary and two in the ischaemic core were counted. The ratio of TUNEL-positive cells to DAPI-positive nuclei was determined.

Blood–brain barrier disruption measurement

Blood–brain barrier (BBB) disruption was analysed using mouse IgG immunohistochemistry on 6-μm sections (Chen et al. 2012). Briefly, sections were de-waxed, rehydrated and the antigen binding sites were exposed. Both endogenous peroxidase activity and non-specific binding sites were blocked before incubating in biotinylated anti-mouse IgG antibody (1 : 100; Vector Laboratories) for 2 h at room temperature (20°C). Immunohistochemical staining was performed following the protocol of the ABC staining kit (Vector Laboratories). Photomicrographs were obtained with a Nikon Eclipse E110M microscope (Nikon). Areas of BBB disruption were delineated by a blinded investigator using a NIS elements imaging software (Nikon). The BBB disruption volume was calculated as the sum of section volumes made by multiplying the area of mouse IgG extravasation on each section by the thickness of the corresponding slice. BBB disruption volume was corrected for oedema and presented as a percentage of the contralateral hemisphere.

HIF-1α immunohistochemistry and immunofluorescence

Neighbouring 6-μm sections were treated as above for HIF-1α immunohistochemistry and without blockage of endogenous peroxidase activity for HIF-1α immunofluorescence. Sections were incubated with a rabbit polyclonal HIF-1α antibody (1 : 100; Novus Biologicals, Littleton, CO, USA) at 4°C overnight followed by appropriate secondary antibodies, at room temperature (20°C) for 2 h. Immunostaining was detected and images were processed as above.


In separate mouse groups, at 6 and 24 h after receiving IOX3 through tail veins, mice were killed with sodium pentobarbital (70 mg/kg intraperitoneally). The brain and kidney were rapidly removed, rinsed in cold PBS and immediately snap-frozen in liquid N2 and stored at −80°C. RNA was extracted using the RNeasy RNA isolation kit (Qiagen, Valencia, CA, USA) as per the manufacturer's instructions. cDNA was synthesized using Agilent Technologies' (Santa Clara, CA, USA) AffinityScript cDNA synthesis kit as per the manufacturer's instructions. cDNA was used to carry out real-time PCR for genes of interest using Agilent Technologies' Brilliant II SYBR Green QPCR Master Mix as per the manufacturer's instructions. Primers for nine HIF-regulated genes were designed in exons only and were designed to be of different sizes for genomic DNA and cDNA (Table S1). The relative gene expression was calculated using the efficiency-corrected calculation model (Pfaffl 2001).

In vitro brain endothelial cell culture

Cell culture

The rat brain endothelial cell line RBE4 was cultivated in 50 : 50 α-minimal essential medium/Ham's F-10 medium mixture (Gibco, Zug, Switzerland) supplemented with 10% fetal bovine serum, 300 μg/mL Geneticin (Gibco) and 1 ng/mL basic fibroblast growth factor (PeproTech, Rocky Hill, NJ, USA) on rat tail collagen-coated Petri dishes and coverslips as described (Al Ahmad et al. 2009). For hypoxia exposure, O2 concentration was constantly monitored and maintained at 1% in a purpose-built hypoxic glove-box chamber (InVivO2 400; Ruskinn Technologies, Pencoed, UK). Immediately prior to oxygen–glucose deprivation (OGD), maintenance media were replaced with glucose-free Dulbecco's modified Eagle's medium (Gibco) containing all other supplements. OGD experiments were carried out in the hypoxic chamber at 1% O2, 5% CO2 and 37°C for 24 h.

Drug administration

IOX3 was prepared as described above. Ten or 50 μM IOX3 or vehicle (5% DMSO/buffer) was added to the culture media under either normoxic or hypoxic conditions. For preconditioning experiments, 10 or 50 μM IOX3 or vehicle (5% DMSO/buffer) was added to the maintenance culture media 24 h prior to the OGD. For immediate exposure, drug was added directly to OGD media at the start of the experiment.


Micrograph pictures of fluorescence immunocytochemistry for zonula occludens (ZO)-1 (rabbit anti-ZO-1, 1 : 100; Invitrogen, Basel, Switzerland) were taken using an inverted fluorescence microscope (Axiovert 200M; Zeiss, Feldbach, Switzerland) coupled to an 8-bit CCD camera (Axiocam HR; Zeiss). DAPI was used as a counter-stain.

Permeability assay

Barrier function was assessed by paracellular permeability of RBE4 monolayers grown on rat tail collagen-coated Transwells (Corning, Schiphol, The Netherlands) to sodium fluorescein (376 Da) as described previously (Al Ahmad et al. 2009). All obtained measurements were normalized to the 0-h time point of the corresponding treatment. A clearance slope was established from the measurements obtained at the different time points and used to calculate permeability values (Pe) (Rist et al. 1997).

MTT assay

Cell viability was determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay. Briefly, 15 000 RBE4 cells were seeded and cultured on 96-well culture plates in different conditions for 48 h (i.e. to confluency). Cells were then exposed to MTT solution (0.5 mg/mL) for 1 h. Subsequently, the medium was removed and 1/10 volume of DMSO was added to the wells for 5 min to dissolve the dye incorporated in the cells. The absorbance was then measured at 570 nm with background subtraction performed at 630 nm.


Confluent monolayers were washed with ice-cold PBS and homogenized in radioimmunoprecipitation assay buffer (Cell Signaling Technology, Danvers, MA, USA) supplemented with a protease inhibitor cocktail (Calbiochem, Merck, Darmstadt, Germany). Accurate protein concentration was determined using Pierce (Rockford, IL, USA) bicinchoninic acid Protein Assay (Thermo Fisher Scientific Inc., Rockford, IL, USA). Thirty micrograms of total protein were separated by denaturing sodium dodecyl sulphate–polyacrylamide gel electrophoresis and subsequently transferred onto nitrocellulose membranes. Membranes were blocked in 5% non-fat dry milk dissolved in Tris-buffered saline followed by incubation at 4°C overnight in the presence of antibodies directed against β-actin (1 : 5000; Sigma-Aldrich, St. Louis, MO, USA) or HIF-1α (1 : 1000; Novus Biologicals). Membranes were washed with 0.1% Tween-20 in Tris-buffered saline and incubated in the presence of horseradish peroxidase-conjugated secondary antibody (Jackson ImmunoResearch, Suffolk, UK). Band detection was performed by enhanced chemiluminescent substrate and visualized using luminescent image analyser LAS-3000 (Fujifilm, Dielsdorf, Switzerland). Blot quantification was performed using ImageJ software (ImageJ; NIH, Bethesda, MD, USA). Immunoreactivity was correct for loading using the beta-actin quantification.

Statistical analysis

Data are represented as mean ± SEM or median with interquartile range where appropriate. Parametric data were analysed using Student's t-test for single comparisons or one-way anova followed by the Bonferroni's test for multiple comparisons. Non-parametric data were assessed using a Mann–Whitney test or the Kruskal–Wallis test with Dunn's multiple comparison test. Continuous data of body temperature, activity and rCBF were analysed by anova for repeated measurements (rm-anova) with Bonferroni post hoc comparisons. Mortality rates were compared with the Fisher's exact test. Pearson correlations were used to assess the relationships among histology data. Statistical analyses were performed using the SPSS for Windows program (version 17). A p-value < 0.05 was considered statistically significant.


Intravenous injection of IOX3 and vehicle caused no apparent physiological changes (e.g. temperature, behaviour, weight), while MCAO resulted in a greater than 70% reduction in rCBF from baseline in each mouse of all experimental groups as required by the protocol. The reduction of rCBF in IOX3 and vehicle groups was statistically indistinguishable, nor was the return of rCBF during reperfusion (Figure S1). The core temperature was reduced in all mice after the surgery but rapidly returned to the normal range with the aid of the telemetric temperature control system (Figure S2). No significant differences were observed in mortality following MCAO, which were about 10% in all animals. There was similar loss of body weight in the IOX3 and vehicle-treated groups 24 h after MCAO, which ranged from 8% to 15% of baseline bodyweight (Table S2).

Behavioural tests assessed at 24 h after reperfusion showed that the mice treated with IOX3 at 20 mg/kg 1 day before (preIOX3-20) had significantly better neuroscores than those with vehicle (preVEH-20) (Fig. 1b). Although there was some apparent improvement with preIOX3-60 compared to preVEH-60, it did not reach statistical significance (Fig. 1b). There were no significant differences in neuroscores between mice when the drug was given immediately before MCAO (immIOX3) and those having the vehicle (immVEH) (Fig. 1a). Interestingly, immIOX3-60 had significantly worse neuroscores than immIOX3-20 (Fig. 1a). Further TUNEL assay analyses indicated that there was no significant difference in the apoptotic ratio between these two groups (Figure S3).

Figure 1.

Mouse behaviour was assessed with neuroscores at 24 h after middle cerebral artery occlusion (MCAO). (a) There was no significant difference in neuroscore between mice receiving IOX3 and vehicle immediately before the MCAO [n values: in the IOX3 (20 mg/kg) comparison, five each; in the IOX3 (60 mg/kg) comparison, four received IOX3 and five had vehicle]. While mice receiving the higher dose of IOX3 (i.e. 60 mg/kg) (n = 4) had worse functional recovery than those having the lower dose of IOX3 (i.e. 20 mg/kg) (n = 5); (b). There was significant improvement in neuroscores in mice receiving the lower dose (i.e. 20 mg/kg) of IOX3 given 24 h prior to the MCAO (n = 8) compared to those receiving the vehicle (n = 5), but no difference was observed in the higher dose (i.e. 60 mg/kg) comparison (four each). *< 0.05.

Nissl staining showed clear infarct areas in brain sections of mice having MCAO and no infarction in SO mice (Fig. 2a). The infarct volume of preIOX3-20 was significantly less than preVEH-20, while there were no differences between preIOX3-60 and preVEH-60 (Fig. 2c), and between immIOX3 and immVEH (Fig. 2b).

Figure 2.

Infarction volume was measured in mouse brain sections with Nissl stains. (a) Representative mouse brain sections stained with Nissl: (i) a SO mouse brain section showing no infarct volume; (ii) a middle cerebral artery occlusion (MCAO) mouse brain section having IOX3 at 20 mg/kg 24 h before the surgery showing a small infarct area; (iii) a MCAO mouse brain section having the vehicle 24 h before the surgery showing a large infarct area. (b) There was no significant difference in infarct volume between mice receiving IOX3 and vehicle immediately before MCAO [n values: in the IOX3 (20 mg/kg) comparison, five each; in the IOX3 (60 mg/kg) comparison, four received IOX3 and five had vehicle]. (c) Infarct volumes in mice received IOX3 at 20 mg/kg 24 h before MCAO (n = 8) were significantly less than those receiving the vehicle (n = 5), but there was no significant change in mice receiving IOX3 at 60 mg/kg compared to those receiving the vehicle (five each). *< 0.05.

Mouse IgG extravasation, an index of BBB disruption, was seen in brain sections of mice having MCAO but was not seen in brain sections of SO mice (Fig. 3a). The volume of BBB disruption in preIOX3-20 was significantly smaller than in preVEH-20 (Fig. 3c), while no significant difference was found in other comparisons (Fig. 3b and c).

Figure 3.

Blood–brain barrier (BBB) disruption was analysed by mouse IgG extravasation on mouse brain sections. (a) Representative mouse brain sections immunostained with mouse IgG: (i) a SO mouse brain section showing no mouse IgG extravasation; (ii) a middle cerebral artery occlusion (MCAO) mouse brain section having IOX3 at 20 mg/kg 24 h before the surgery showing a small area of mouse IgG extravasation; (iii) a MCAO mouse brain section having the vehicle 24 h before the surgery showing a large area of mouse IgG extravasation. (b) The volume of mouse IgG extravasation in mice receiving IOX3 immediately before MCAO was not different from those receiving the vehicle [n values: in the IOX3 (20 mg/kg) comparison, five each; in the IOX3 (60 mg/kg) comparison, four received IOX3 and five had vehicle]. (c) The volume of mouse IgG extravasation in mice receiving IOX3 at 20 mg/kg 24 h before MCAO (n = 8) were significantly less than those receiving the vehicle (n = 5), but there was no significant change in mice receiving IOX3 at 60 mg/kg compared to those receiving the vehicle (five each). *< 0.05.

HIF-1α was not observed in brain sections of SO mice receiving the vehicle (Fig. 4a), but was detected in brain sections of SO mice receiving IOX3 24 h prior to the SO operation (Fig. 4b). In the MCAO mice, HIF-1α was up-regulated in the ipsilateral hemisphere of mice either receiving the vehicle (Fig. 4c) or IOX3 (Fig. 4d) 24 h prior to the MCAO operation.

Figure 4.

Hypoxia-inducible factor (HIF)1α immunostains on 6-μm mouse brain sections 24 h after the surgery from representative SO mice receiving vehicle (a), receiving IOX3 20 mg/kg 24 h before the surgery (b), middle cerebral artery occlusion (MCAO) mice receiving the vehicle (c) and receiving IOX3 (20 mg/kg) (d) 24 h before the surgery. There was no HIF-1α signal on brain sections of SO mice receiving the vehicle (a1, under 4× magnification), while HIF-1α was detected in brain sections of SO mice receiving IOX3 20 mg/kg (b1), and in the ipsilateral hemispheres of MCAO mice (c1, d1). Under higher magnification (10× and 40×), HIF-1α was observed to accumulate in the cell nucleus (b2, b3; c2, c3; d2, d3). 4,6-Diamidino-2-phenylindole (DAPI) (blue) is nuclear counter-stain. n = 4.

RT-PCR was performed with nine representative HIF target genes on mouse brains and kidney in separated mouse groups. There was substantial up-regulation of the EPO gene in both brain and kidney of mice receiving IOX3 compared to those that received vehicle only (Fig. 5). There was no significant up-regulation in eight other HIF target genes (data not shown).

Figure 5.

EPO gene expression in mouse brains and kidneys was measured by RT-PCR. There were significantly higher amounts of EPO gene expression related to the house keeper gene (SADH) in mouse brains and kidneys at 6 and 24 h after receiving IOX3 than those receiving the vehicle. n = 4. *< 0.05, **p < 0.01.

To further investigate the direct effects of IOX3 on BBB morphology and function, we then applied IOX3 in an in vitro BBB model. IOX3 at 10 and 50 μM significantly up-regulated HIF-1α in the RBE4 cells under both normoxia and hypoxia (Fig. 6).

Figure 6.

Hypoxia-inducible factor (HIF)-1α induction in RBE4 cells exposed to IOX3 and hypoxia (1% O2) for 1.5–24 h. (a) A summary of normalized HIF-1α expression measured at 1.5, 3, 6 and 24 h after exposure to IOX3 (10 and 50 μM) in both normoxia and hypoxia (n = 4). One-way anova indicated that IOX3 had significant effects on HIF-1α induction at both normoxia (6 h: F = 14.8, p < 0.01; 24 h: F = 16.9, p < 0.01) and hypoxia (1.5 h: F = 6.8, p < 0.05; 3 h: F = 7.9, p < 0.05; 6 h: F = 5.16, p < 0.05; 24 h: F = 16.3, p < 0.01). Post hoc analysis suggested that IOX3 50 μM significantly up-regulated HIF-1α, *p < 0.05, **p < 0.01. (b and c) Representative HIF-1 blots are shown together with the beta-actin blot that was used for quantification.

Under normoxic conditions without IOX3 treatment, ZO-1 staining was continuous at cell–cell borders throughout the monolayer. When exposing to 10 μM IOX3, the monolayer remained largely intact, whereas exposure to 50 μM IOX3 induced disruption of ZO-1 localization with loss of staining between some cells (Fig. 7). Under hypoxic conditions, gap formation at cell–cell borders was visible and exposure to 50 μM IOX3 further exaggerated the hypoxic effect with significant delocalization of ZO-1 from endothelial cell junctions (Fig. 7).

Figure 7.

Effects of IOX3 on ZO-1 localization in BRE4 cells. Immunocytochemistry showed that the localization of ZO-1 (green) at cell–cell borders was continuous under normoxic conditions without IOX3 treatment. On exposure to 10 μM IOX3 under normoxic conditions, the monolayer remained largely intact, whereas in contrast, exposure to 50 μM IOX3 resulted in increased ZO1 disruption with stretched appearance of staining between cells (denoted by arrows). Under hypoxic conditions ZO-1 disruption at cell–cell borders was evident but not further influenced by 10 μM IOX3 (denoted by arrows). On the contrary, exposure to 50 μM IOX3 further exaggerated the hypoxic effect with significant delocalization of ZO-1 from endothelial cell junctions, i.e. frequent loss of cell–cell border staining and gap formation (denoted by arrows). 4,6-Diamidino-2-phenylindole (DAPI) (blue) is nuclear counter stain. n = 4.

Glucose deprivation disrupted ZO-1 localization resulting in the appearance of stretched staining at the contacts and gaps (Fig. 8). Hypoxia further exaggerated this response with a more obvious loss of ZO-1 at cell junctions particularly when glucose was simultaneously withdrawn (Fig. 8). Exposure to IOX3 prior to OGD clearly improved ZO-1 localization and prevented the effects of OGD (Fig. 8). In contrast, exposure to IOX3 immediately prior to OGD did not improve ZO-1 localization and, in some cases, seemed to aggravate the condition with the majority of cells still exhibiting widespread loss of expression at cell–cell contacts with frayed staining and clumping frequently observed (Fig. 8).

Figure 8.

Effects of IOX3 on ZO-1 localization in BRE4 cells during oxygen and glucose deprivation (OGD). When glucose was withdrawn from the media, partial loss of ZO-1 was observed under normoxia. Additional oxygen deprivation (OGD) resulted in dramatic disruption of ZO-1 with consistent and widespread loss of the protein from cell–cell borders. Pre-treatment with IOX3 significantly improved ZO-1 stability and localization. On the contrary, simultaneous exposure to IOX3 could not prevent the OGD effect: complete loss of cell–cell border staining and gap formation was clearly observed. 4,6-Diamidino-2-phenylindole (DAPI) (blue) is nuclear counter-stain. n = 4.

Combined oxygen and glucose withdrawal (OGD) significantly increased barrier permeability by two to threefold (Fig. 9). Pre-treatment of the cells with IOX3 for 24 h before OGD prevented the cells from barrier dysfunction, while immediate exposure to the drug and OGD had no such protection (Fig. 9).

Figure 9.

Effects of IOX3 on RBE4 monolayer permeability. Exposure of RBE4 cells to hypoxia (Hx+G) caused only a moderate change in RBE4 monolayer integrity, whereas oxygen and glucose deprivation (OGD, Hx-G) significantly increased paracellular flux by two to threefold compared to normoxia (Nx+G). Twenty-four-hour pre-treatment with IOX3 abrogated barrier disruption during OGD significantly at 50 μM compared to Hx-G, while immediate drug exposure had no effect. n = 4. **p < 0.01 compared to Hx+G; *p < 0.05 compared to Hx-G.

To rule out the possibility that the changes observed were because of alterations in cell survival, an MTT assay was performed. Endothelial survival was reduced when cells were exposed to hypoxia and further impaired when glucose was withdrawn from the media. However, drug treatment did not affect cell survival at either concentration during preconditioning or immediate exposure (Figure S4).


In this study, we found that (at 20 mg/kg) IOX3, a reported selective inhibitor of the HIF PHDs, only exerts neuroprotective properties in a mouse model of focal cerebral ischaemia, when given 24 h prior to the ischaemic insult. Immediate treatment with IOX3 before the onset of MCAO was ineffective. Moreover, higher doses of IOX3 (60 mg/kg) with pre-treatment failed to show this neuroprotective effect. One potential downstream mediator of the observed protective effect was EPO, which showed an elevated expression profile in the brain and the kidney. Further in vitro studies on endothelial brain cells revealed that one potential mechanism of the abolished protection at 60 mg/kg might be increased BBB permeability. In the in vitro studies, in correction with significant delocalization of ZO-1 from endothelial cell junctions, HIF-1α was up-regulated by IOX3. Pre-treating IOX3 on RBE4 cells 24 h before OGD had a protective effect on BBB integrity with ZO-1 being better localized, while simultaneous IOX3 treatment did not.

Hypoxia/ischaemic preconditioning is considered the next most powerful experimental neuroprotective strategy after hypothermia (Dirnagl et al. 2009). HIF is of central importance in the response to hypoxia/ischaemia, and is essential for cerebral ischaemia tolerance induced by hypoxic preconditioning (Ratan et al. 2004). The HIF-α isoforms act as transducers of the response to hypoxia, whereas the HIF hydroxylases (PHDs) are responsible for regulating HIF activity in an oxygen-dependent manner (Fraisl et al. 2009). Pharmacological inhibition of HIF PHDs can act as mimics of hypoxic/ischaemic conditioning (Li et al. 2008; Nagel et al. 2011; Wacker et al. 2012; Engelhardt et al. 2014). Analogues of 2-OG such as DMOG inhibit the PHDs but they are not specific for HIF PHDs and may affect other signalling pathways (Loenarz and Schofield 2008; Nigel et al. 2011; Rose et al. 2012). IOX3 has been reported to specifically inhibit the HIF PHDs over FIH, and likely, at least some other human 2-OG oxygenases (Tian et al. 2011), including the fat mass and obesity associated protein (FTO) (Aik et al. 2013).

Hypoxic/ischaemic preconditioning has two time windows of protection, an immediate one mediated mostly by local factors, including adenosine and the ATP-sensitive potassium channel, and a later one lasting for 1–3 days which is mediated by genetic reprogramming (Dirnagl et al. 2009). For pharmacological preconditioning through HIF PHD inhibition, only the latter one can be assumed to be effective, unless the applied drug has some HIF-unrelated protective effects. To the best of our knowledge, no direct neuroprotective effects of IOX-3 have been reported to date. IOX3 treatment immediately before MCAO had no neuroprotective effects at 24 h after cerebral ischaemia in vivo. Similarly, in the in vitro experiments, IOX3 did not improve the barrier function when added to RBE4 cells immediately before the OGD; however, IOX3 improved ZO-1 localization and prevented the effects of OGD when given 24 h prior to the OGD. This is consistent with a recent study on neurons, in which preconditioning with DMOG for 24 h prior to OGD induces protein stabilization of HIF-1α, and significantly reduces OGD-induced neuronal death (Ogle et al. 2012). Notably, a single post-treatment of FG 4497 was sufficient to decrease tissue damage at a later time point (i.e. 7 days) after permanent MCAO (Reischl et al. 2014). It will be interesting to further study the effects of IOX3 treatment started in the pivotal therapeutic time window after onset of ischaemic stroke at some later time points (e.g. 3 or 7 days).

To detect putative protective downstream targets acting as effectors of the HIF response, we analysed a set of nine HIF-regulated genes. Interestingly, within the limits of detection, only EPO showed a significant up-regulation in the brain and the kidney. Although this was a little surprising, the factors that up-regulate the sets of HIF target genes in given cell types/contexts are only at early stage of being identified, and it appears that factors other than limitation of PHD activity may be important in the context of our investigations. HIF is a highly pleiotropic transcription factor, likely regulating 500 genes directly and many others indirectly (Geiger et al. 2011). It is important to note that the sets of HIF target genes that are up-regulated are highly context dependent, depending on factors such as HIF-1α/HIF-2α levels, the PHD isoform present and whether the C-terminal transcriptional activating domain of HIF-α cofactors is hydroxylated (Mole et al. 2009). Furthermore, there is strong evidence that the activity of the HIF hydroxylase can be regulated by factors other than oxygen availabilities, e.g. tricarboxylic acid cycle intermediate levels (Tannahill et al. 2013). IOX3 specifically inhibits PHDs over the FIH, as shown in cellular studies (Tian et al. 2011). Since FIH is reported to have a comparatively smaller effect on HIF-2α than HIF-1α (at least in some contexts) (Koivunen et al. 2004; Bracken et al. 2006), it is possible that IOX3 could induce production of more active HIF-2α than HIF1-α. Notably, it is HIF2 (rather than HIF1) that is reported to regulate EPO expression in the brain (Yeo et al. 2008; Kunze et al. 2012). Although further studies must be performed to confirm this, the strong up-regulation of EPO upon hypoxia/ischaemia in mouse brains indicates that EPO plays an important role in neurovascular protection from ischaemia (Kunze et al. 2012). EPO increases neuronal survival, fosters ischaemic tolerance (Ruscher et al. 2002; Kumral et al. 2004; Ogunshola and Bogdanova 2013) and increases neurogenesis and oligodendrogliosis of subventricular zone precursor cells after ischaemic stroke (Gonzalez et al. 2013). EPO preserves endothelial cell integrity (Chong et al. 2002), promotes angiogenesis (Wang et al. 2004) and protects BBB from disruption during injury and maintains the establishment of cell–cell junctions (Martínez-Estrada et al. 2003). However, in a double-blind, placebo-controlled, randomized German Multicentre EPO Stroke Trial (Phase II/III; ClinicalTrials.gov Identifier: NCT00604630), ischaemic stroke patients receiving EPO did not show favourable effects (Ehrenreich et al. 2009). Therefore, unless other HIF-mediated mediators of neuroprotection for IOX3 are detected in further studies (and given the number of HIF target genes this is a possibility), the observation of EPO up-regulation in isolation cannot be used to advocate testing IOX3 as a neuroprotectant in a clinical setting at this point in time.

Our in vivo and in vitro studies reveal similar but not identical outcomes using IOX3 at different concentrations. This could reflect the limited extent of our studies, or the highly pleiotropic HIF system, or inhibition of other enzymes in addition to the PHDs, including those inhibiting nucleic acid demethylation (Aik et al. 2013). It is difficult to match IOX3 concentration in vivo with that in vitro, as the former is affected by numerous factors, including context-dependent pharmacokinetics. Notably, 60 mg/kg is the maximal amount of IOX3 that can be given in a single injection to a mouse when considering the maximal volume being injected as well as maximal concentration of the drug being dissolved. Both the in vivo and in vitro data agree that drug application immediately after cerebral ischaemia should be avoided. IOX3 pre-treatment is neuroprotective partially because of better maintaining BBB integrity during ischaemia. BBB function and integrity contribute substantially to neuroprotection (Gidday 2006). The BBB is rapidly affected by ischaemia/reperfusion, and has been shown to precede neuronal damage (Latour et al. 2004). BBB disruption in ischaemic stroke leads to haemorrhagic transformation with enhanced brain injury in both animals and humans (Jickling et al. 2014). Our study suggests that HIF stabilization with IOX3 before cerebral ischaemia is neuroprotective partially because of BBB protection, while immediate application could be detrimental. Indeed, we have recently shown that immediate HIF stabilization leads to delocalization and decreased phosphorylation of tight junctions, and compromises barrier function (Engelhardt et al. 2014).

In vivo mouse models of cerebral ischaemia are very sensitive to confounding changes in physiological variables like body temperature and cerebral blood flow. Both variables were very well controlled in our experiments, which is the strength of the study. We did not observe any differences in rCBF during ischaemia and reperfusion between IOX3 and vehicle groups. While our previous study on PHD isoform genetic modified mice found that PHD2+/− mice (PHD2 is the most important human isoform of PHD) have more rapid rCBF return during reperfusion than wild-type mice (Chen et al. 2012). This indicates that mechanisms of neuroprotection in PHD2+/− mice and by IOX3 treatment are different. This observation is important since we and others have previously found that PHD2+/− mice have increased vessel density in organs including the brain by a long-term response through genetic reprogramming, which enables the body to form collateral arteries (Takeda et al. 2011; Chen et al. 2012). After MCAO, mice had hypothermia. Hypothermia decreases the cerebral metabolic rate, reduces infarct volume, improves neurological recovery and hence is neuroprotective (Yenari and Han 2012). Previously, we have reported that temperature-regulated mice had larger infarct volumes, and worse histological and behavioural scores compared to non-temperature-regulated mice (Barber et al. 2004). To reduce variables that may affect the outcomes, we have controlled the temperature through a telemetric feedback regulated system in this study.

In conclusion, we have found that a selective HIF PHD inhibitor, IOX3, at low concentrations protects the brain from ischaemia/reperfusion damage when given preconditionally. Immediate treatment before the ischaemic insult is ineffective possibly because there is not enough time for ‘genetic reprogramming’ which occurs on a timescale of hours to days (Dirnagl et al. 2009). Out of nine measured HIF-targets, only EPO was found to be a potential protective effector gene, although in addition to EPO, other HIF target genes are likely up-regulated. IOX3 up-regulates HIF-1α in RBE4 cells and causes significant delocalization of ZO-1 from endothelial cell junctions with rising doses. Pre-treatment of IOX3 on RBE4 cells prevents barrier disruption from a subsequent OGD, while immediate application does not have the protection. Pharmacological preconditioning with HIF PHD inhibitors remains a promising strategy; however, further experiments on the detailed mechanisms of neuroprotection are needed before IOX3 is tested in a clinical setting for prevention of ischaemic strokes. The fine regulation of the HIF system during permanent cerebral ischaemia or ischaemia followed by reperfusion in the setting of simultaneous application of HIF manipulating drugs is still only poorly understood. In the absence of a detailed molecular understanding, one way forward will be to vary the nature and dose of the HIF hydroxylase inhibitors used.

Acknowledgments and conflict of interest disclosure

We are grateful for the funding received from the Dunhill Medical Trust, MRC, the Foundation Leducq, the Wellcome Trust, British Heart Foundation, Biotechnological and Biological Research Council, the European Union, and the Henry Smith Medical Research. CJS was a cofounder of ReOx, a company that was founded to exploit basic science work on hypoxia for therapeutic benefit. The underlying [original] research reported in the article was funded by The National Institute for Health Research (NIHR) – Professor Alastair M Buchan is supported by an NIHR Senior Investigator award and the NIHR Oxford Biomedical Research Centre.

All experiments were conducted in compliance with the ARRIVE guidelines.