Inhibition of HIF2 Prevents the Development of Neurodegenerative Disorder Induced by De ciency of IRP2

Jiaqi Shen Nanjing University Medical School Li Xu Nanjing University Medical School Yuxuan Li Nanjing University Medical School Weichen Dong Nanjing University Medical School Clinical College: East Region Military Command General Hospital Jing Cai Nanjing University Medical School A liated Nanjing Drum Tower Hospital Yutong Liu Nanjing University Medical School Hongting Zhao Nanjing University Medical School Tianze Xu Nanjing University Medical School A liated Nanjing Drum Tower Hospital Esther Meyron Holtz Technion Israel Institute of Technology Yanzhong Chang Hebei Normal University College of Life Sciences Tong Qiao Nanjing University Medical School A liated Nanjing Drum Tower Hospital Kuanyu Li (  likuanyu@nju.edu.cn ) Nanjing University https://orcid.org/0000-0001-9738-049X


Introduction
Iron is an indispensable element in mammals. Maintaining proper iron concentration in our body is of great signi cance because iron, in forms of heme, iron-sulfur cluster (Fe-S), or iron itself as important cofactors, are involved in multiple biochemical pathways, including hemoglobin synthesis and mitochondrial respiratory chain (Hentze et al., 2004;Darshan et al., 2010;Ganz and Nemeth, 2012;Rouault, 2013). For this reason, mammals have developed sophisticated mechanisms to maintain proper iron concentration in the body: (1) systemic iron homeostasis is maintained by hepcidin-ferroportin (hepcidin-FPN1, encoded by HAMP and SCL40A1) axis (Nemeth et al., 2004;Ganz and Nemeth, 2011); (2) cellular iron homeostasis is mediated by iron regulatory proteins (IRPs, IRP1 and IRP2, also called ACO1 and IREB2) through IRP-IRE (iron responsive element) system (Rouault, 2006;Wallander et al., 2006;Muckenthaler et al., 2008;Rouault, 2013). More tissue speci c strategies have also been developed, e. g. ferritinophagy to regulate erythropoiesis (Mancias et al., 2015). These ways function and interplay to netune iron levels in the body . IRP1 and IRP2 are both iron-regulatory RNA binding proteins that regulate the expression of a series of iron-related genes at the post-transcriptional levels (Hentze et al., 2010;Anderson et al., 2012). Under conditions of iron de ciency, the IRE in the target mRNA can be recognized and bound by IRPs, but the consequence of IRP binding depends on the position of the IRE on the mRNA of the target genes. If the IRE is in the 5'-untranslated region (UTR) of the target mRNA, the binding of IRPs may inhibit the translation of the genes, including L-and H-ferritin and FPN1; if the IRE is in the 3'-UTR, the binding of IRPs may stabilize the mRNA, such as that of transferrin receptor 1 (TfR1) (Casey et al., 1988;Müllner et al., 1989) and divalent metal transporter 1 (DMT1) (Tybl et al., 2020). When cellular iron is su cient, IRP1 binds to a [4Fe-4S] cluster, therefore, gains aconitase activity and loses the ability to bind IRE, whereas IRP2 is removed by iron and oxygen-mediated proteasome degradation (Salahudeen et al., 2009;Vashisht et al., 2009) to avoid the excessive iron uptake and to promptly stores excess intracellular iron and/or export excess iron. Recently, we found that mitochondrial dysfunction was closely associated with the reduced expression of a number of genes that are involved in Fe-S biogenesis and mitochondrial respiratory chain (Li et al., 2018). The further investigation revealed that Irp2 may function as a key to switch the metabolism between aerobic glycolysis and oxidative phosphorylation (OXPHOS), mediated by upregulation of hypoxia-inducible factor subunits Hif1α and Hif2α in mouse embryonic broblasts (MEFs) (Li et al., 2019). HIF1 and HIF2 are two important transcription factors that can regulate the expression of a series of genes. Active HIF is a heterodimer, composed of α subunit (HIF1α or Hif2α) and β subunit (HIF1β, also called ARNT) and can bind hypoxia responsive element (HRE), which is a very important mechanism in intestinal iron absorption (Mastrogiannaki et al., 2009;Shah et al., 2009) and under other conditions, such as cancer (Keith et al., 2011) and ischemia (Kapitsinou et al., 2014;Barteczek et al., 2017). IRP1 can bind to the IRE in the 5'-UTR of HIF2α mRNA to regulate HIF2α at the post-transcriptional level (Sanchez et al., 2007;Zimmer et al., 2008). In Irp1 −/− mice, elevated Hif2α up-regulates erythropoietin (EPO), causing the mice to develop polycythemia and pulmonary hypertension (Anderson et al., 2013;Ghosh et al., 2013;Wilkinson and Pantopoulos, 2013). Interestingly, in Irp2 −/− MEFs we found that Hif1α and Hif2α were both up-regulated, which switches the metabolism type from oxidative phosphorylation to glycolysis (Li et al., 2019). Inhibition of both Hif1α and Hif2α reversed the energy metabolism.
In this study, we con rmed the elevated Hif2α, not Hif1α, in Irp2 −/− mice. The upregulated glycolytic pathway-related proteins were also observed and associated with the enhanced glycolysis, while down-regulated frataxin (Fxn) and iron-sulfur cluster scaffold protein (IscU), respectively, were observed and associated with de ciency of iron-sulfur clusters. Consequently, the expression of electron transmit chain (ETC) subunits was reduced and OXPHOS was weakened. After the inhibition of Hif2α by PT-2385, the energy metabolism was shifted from glycolysis to OXPHOS in Irp2 −/− mice, the histological and behavioral indicators were restored, and neurodegenerative symptoms were alleviated. Our results indicate that the neurodegenerative disorder induced by the loss of Irp2 is, at least partially, mediated by the upregulated Hif2α. In the hang test, mice were allowed to grip a wire mesh square that was then inverted. The latency time that mice could hang on to an inverted wire mesh square before falling was measured. And each mouse was tested for three times with an interval of 5 min.

Rotarod Tests
The motor functions of balance and coordination were assessed using an accelerating rotarod (Jiangsu SANS Technology Co., Ltd.). Recorded the staying time of mice on the rotating rod (the rotating rod accelerated from 4 rpm to 40 rpm within 5 min), and each mouse was tested for three consecutive times.

Drug Treatment
Both PT-2385 and PX-478 were dissolved in DMSO, diluted with normal saline and injected intraperitoneally into six-month-old male Irp2 −/− mice. The injection doses of PT-2385 and PX-478 were 0.4 mg/kg body weight and 5 mg/kg body weight, respectively, and the injections lasted for one month every other day.

H&E Staining
In H&E staining, tissue sections were dealt with the following steps: dewaxed for 10 min in xylene twice; hydrated for 5 min in each 100% − 50% ethanol gradient buffers; rinsed for 5 min in running water at room temperature; stained with Hematoxylin for 10 min, then in Eosin Y for 10 min. Slides were dehydrated through gradual ethyl alcohol solutions for imagining.

Electron Microscopy
The cerebellum and spinal cord were separated in the size of rice grains, placed in a mixed solution of 2% paraformaldehyde and 0.1 M cacodylate for 30 min at room temperature, then stored at a constant temperature of 4˚C (refrigerator). The samples were rinsed once or twice, then dehydrated through a series of ethanol from 50-100%, and then propylene oxide was used instead of ethanol. The samples were stored in 50% propylene oxide and 50% EPON resin (1:1 mix) for 1 hour, and then placed in pure EPON. The samples were transferred to fresh EPON in molds or beem embedding capsules, which were lled carefully to avoid air bubbles, and kept at 60˚C for at least 24 h. Samples were observed and photographed by using HT7800 electron microscope at 80 keV, and electron micrographs were commented by Hitachi TEM system.

Western Blot Analysis
The total protein of each entire tissue was extracted and analyzed (25-35 µg total protein/lane) by 7.5%-12.5% SDS-PAGE at 100 V, transferred onto nitrocellulose membrane at 250 mA for 1. (cat# AJ1290b) from ABGENT (San Diego, CA), anti-Fxn, IscU, Irp1 and Irp2 (polyclonal, self-made, raised from rabbits). All the self-made antibodies were validated in previous studies (Li et al., 2018;Li et al., 2019). When it is necessary to detect multiple proteins in one blot and the molecular weight of the protein is different, we cut the blotted nitrocellulose membrane according to the molecular weight, and then incubate with different antibodies. When the molecular weights are very close, run multiple gels with the same prepared total protein samples, transfer them to nitrocellulose membranes, cut according to molecular weights, and then incubate with different antibodies. We used Tanon Science and Technology Co., Ltd. (Shanghai, China) ECL-plus reagent to visualize the detected proteins. The intensity of the western blot band was quanti ed by ImageJ software. Each experiment was repeated at least three times independently, and biological replicates were performed in parallel each time. The average intensity of the bands from replicate samples was rst normalized to an internal control (actin), and then normalized to a wild-type control, with the value set to 1. The nal value was the average value from at least three independent experiments.

Ferrozine Iron Assay
The serum, intestinal, cerebellum, and spinal cord iron content were detected by the Ferrozine Iron Assay. Took 50 µL lysate or serum (took double volume for lysis buffer as control). Added 11 µL concentrated HCL (11.6 M). Placed all tubes on the 95℃ heating block for 20 min. Centrifuged at the highest speed for 10 min, removed very gently from centrifuge. Removed 45 µL supernatant very carefully. Added 18 µL ascorbate (75 mM) to each tube, ascorbate acts as a reductant, moving Fe from 3 + state to the 2 + state. Vortex-quick spin, incubated for 2 min. Added 18 µL ferrozine (10 mM) to each tube. Ascorbate acts as an oxidant, taking Fe from the 2 + state to the 3 + state, incubated for 2 min. Added 36 µl saturate ammonium acetate (NH 4 Ac) to each tube, incubated for 2 min. Read samples at 562 nm using multifunctional uorescent microplate reader.

Enzymatic Activities
The activities of complex I and II were measured according to the manufacturer's protocols, respectively. Both kits were purchased from Comin Biotechnology Co. (Suzhou, Jiangsu, China).

Determination of ATP Content
The levels of ATP in tissues were detected by using ATP determination kit (Beyotime Biotech.). The reading is measured by GloMaxTM96-well plate luminescence detector (E6521)

Lactic Acid Production
The tissue lysates were collected and assayed according to the lactic acid production detection kit (Nanjing Jiancheng Bioengineering Institute). The assay was detected by multifunctional uorescent microplate reader at 530 nm.

Statistical Analysis
Student's t-test or one-way analysis of variance (ANOVA) was carried out using Graphpad prism 8. The measurement was expressed as the mean ± SEM; all the experiments were repeated more than 3 times independently. Signi cance was considered at p < 0.05. . We wonder if it is the case in vivo. First, we detected the expression levels of iron-related proteins in the CNS tissues (cerebrum, cerebellum, brainstem and spinal cord) of Irp2 −/− mice. Compared with that in wild type (WT) mice, ferritin expression was increased, while Tfr1 was decreased in Irp2 −/− mice ( Fig. 1A and B), which was in line with previous study (Jeong et al., 2011). Next, we detected the expression levels of Hif1α and glycolysis-related proteins and did not nd the same elevation of Hif1α as we observed in Irp2 −/− MEF (Li et al., 2019). However, glycolysis-related proteins, including lactate dehydrogenase A (LdhA), glucose transporter 1 (Glut1), hexokinase 2 (Hk2), were upregulated, compared with that in WT mice ( Fig. 1C and D), although these genes are the members of Hif1 regulon. Then, we detected the expression levels of Hif2α, the Fe-S biogenesis-related proteins (IscU and Fxn), and mitochondrial respiratory complex subunits (Ndufs1, SdhB, and Uqcrfs1). We found that the levels of IscU and Fxn were reduced and subunits of complex I (Ndufs1), II (SdhB) and III (Uqcrfs1) were also reduced in the CNS tissues of Irp2 −/− mice, compared with WT mice (Fig. 1E and F), suggesting a reduction of OXPHOS. Taken together, our results con rmed the biochemical changes in vivo related to energy metabolism, OXPHOS and glycolysis, in the tissues of the CNS of Irp2 −/− mice.
3.2 Administration of PT-2385 signi cantly improves the behavioral performance and anemia of Irp2 −/− mice As presented above, we only observed the upregulation of Hif2α in central nervous tissues of Irp2 −/− mice, but both inhibitors, PX-478 (5 mg/kg body weight) for Hif1α and PT-2385 (0.4 mg/kg) for Hif2α, were still injected into Irp2 −/− mice intraperitoneally every other day for one month, individually. During the one month, the mice weight was all monitored and found increased normally without difference compared with the vehicle treatment, indicating the safety of the drug and its dosage ( Fig. 2A). In terms of behavioral performance, the latency time of Irp2 −/− mice on the rotating rod and the hanging time on the wire mesh square were signi cantly shorter than that of the WT mice. However, it signi cantly recovered in Irp2 −/− mice after administration of PT-2385, while no e cacy was observed after PX-478 treatment in agreement with no change of Hif1α levels in the CNS tissues of mutant mice ( Fig. 2B and C). To con rm the effect of PT-2385 in Irp2 −/− mice is through rescuing Irp2 de ciency, we also treated the WT mice with PT-2385. The rotarod and hang tests did not show the signi cant difference between the vehicle and PT-2385 treatment (not shown). These data proved the critical role of Hif2α in CNS of Irp2 −/− mice. Since then, our work focused on Hif2α inhibition by PT-2385.
The anemia of Irp2 −/− mice, likely, resulted from decreased expression of TfR1 in erythroblasts and decreased bone marrow iron stores (Cooperman et al., 2005;Galy et al., 2005). Very interestingly, PT-2385 treatment corrected the anemia of Irp2 −/− mice as well, showing reversed number of red blood cells, hemoglobin, and hematocrit, but not the mean corpuscular volume (Fig. 2D, E, F and G). Then, we measured more parameters to evaluate the iron status, including the EPO mRNA in kidney, serum EPO and iron. Surprisingly, the iron status globally improved (Fig. 2H, I and J). To understand how the PT-2385 treatment corrected the iron insu ciency anemia of Irp2 −/− mice, we assessed the iron content in intestine and liver, HAMP mRNA level in liver, and serum interleukin 6 to determine whether the serum iron resulted from intestinal uptake or iron release from liver. The results showed that HAMP mRNA levels in liver, and serum interleukin 6 (IL-6), and iron content in intestine and liver all reduced after PT-2385 treatment (Fig. 2K, L, M and N), suggesting that both ways, intestinal uptake and iron release from liver, contributed to the elevation of the serum iron. This assumption was further supported by the increased NcoA4 ( Fig. 2O and P), which is involved in ferritinophagy for ferritin degradation to release iron (Mancias  (Fig. 3A). The results from the electron microscopy showed that the density of mitochondria in mutant cerebellum is lightened, which phenomenon was much more severe in the spinal cord of Irp2 −/− mice than that in WT mice. More affectedly, the morphology of the mitochondria in spinal cord of Irp2 −/− mice became swollen, vacuolated, and internal-cristae damaged. Interestingly, PT-2385 treatment signi cantly alleviated the poor presentation, including the deformed mitochondria and Wallerian and segmental demyelination (Fig. 3B, C and D), suggesting the bene cial effect of PT-2385 against motor neurodegeneration. Fxn, IscU, and complex subunits, was decreased in Irp2 −/− mice compared with WT. However, PT-2385 treatment inhibited all these biochemical changes compared with Irp2 −/− mice (Fig. 4A, B, C and D). In line with these results, the activities of mitochondrial complex I and II were also signi cantly restored in both tissues ( Fig. 4E and F). The exception is the coupled ETC product ATP. In cerebellum, ATP content was lower in Irp2 −/− mice than that in WT mice and PT-2385 administration increased it, which is correlated with the ETC activities (Fig. 4G). However, in spinal cord, ATP content increased signi cantly in Irp2 −/− mice (Fig. 4H), which is consistent with the previous studies (Li et al., 2018;Li et al., 2019) in MEFs though the ETC-related proteins and enzymatic activities were lower in Irp2 −/− mice than those in WT mice (Fig. 4C, D and F). The reason will be discussed further in Discussion. Very surprisingly, compared with Irp2 −/− mice, more ATP was produced after PT-2385 administration (Fig. 4H). Overall, the inhibition of Hif2α by PT-2385 rescues the weakened OXPHOS in Irp2 −/− mice to provide more ATP to ful ll the energy need.
3.5 Inhibition of Hif2α attenuates enhanced glycolysis in the cerebellum and spinal cord of Irp2 −/− mice Though Hif1α was not found to be upregulated in vivo, we observed the enhanced glycolysis-related gene expression (Fig. 1B), which is similar to the previous results in Irp2 −/− MEFs (Li et al., 2019). We still used the tissues cerebellum and spinal cord to check the effect of PT-2385 on the expression of LdhA, Glut1, Hk2, and endothelin 1 (Edn1), which genes are considered to be the members of Hif regulon. The results showed that the expression of these genes increased in Irp2 −/− mice and reduced to the WT levels after PT-2385 treatment (Fig. 5A and B), con rming the action of PT-2385 on Hif2α and the regulon relationship of Hif2α to the tested genes. The protein levels of glycolysis-related genes including LdhA, Glut1, and Hk2 were also signi cantly reduced after PT-2385 treatment (Fig. 5C, D, E and F). Accordingly, the lactic acid levels were signi cantly higher in both cerebellum and spinal cord of Irp2 −/− mice than those in WT, and PT-2385-treatment signi cantly lowered the levels in both tissues of Irp2 −/− mice ( Fig. 5G  and H). Comparing the two tissues, cerebellum and spinal cord, the upregulated Hif targeted genes seemingly responded stronger in spinal cord than in cerebellum since their mRNA levels elevated more in spinal cord (Fig. 5B) than in cerebellum (Fig. 5A), particularly for Hk2 and Edn1, after Irp2 depletion. The protein levels of LdhA and Glut1 in the spinal cord increased about twofold and vefold, respectively ( Fig. 5E and F), versus 1.3-and 2-fold in the cerebellum (Fig. 5C and D) after Irp2 depletion. The results suggest that the spinal cord might suffer more from the upregulated Hif2α and active glycolysis.

Discussion
In this study, we found that Irp2 ablation-induced Hif2α upregulation alone mediated the metabolism switch from OXPHOS to glycolysis in vivo. The protective effect of PT-2385 through Hif2α inhibition suggested that Hif2α is a potential target therapeutically in the treatment of IRP2 mutant-caused neurodegenerative syndrome. Hif2α also weakened glycolysis to avoid the toxicity of high levels of lactic acid through suppression of LdhA, Glut1, and Hk2. Therefore, PT-2385 administration dramatically protected from the progressive neurodegeneration as modeled in Fig. 6.
Interestingly, Hif2 inhibition also improved blood parameters overall from anemia. However, Hif2 is very important for erythropoiesis by regulating EPO production and for iron uptake in the small intestine by regulating DMT1, FPN1 and Dcytb. Surprisingly, we found that the lifted iron content in liver of Irp2 mutant reduced and the increased ferritin dropped back to WT levels. More profoundly, we found that NcoA4 expression increased in liver after Hif2 inhibition, suggesting the important role of NcoA4 for iron release from ferritin, very likely, through ferritinophagy.
Both Hif1α and Hif2α are regulated by oxygen and iron (Majmundar et al., 2010;Prabhakar and Semenza, 2012). In this study, we found the elevated Hif2α alone, not together with Hif1α, contribute to the switch of energy metabolism from OXPHOS to glycolysis. The rationale for Hif2α to regulate IscU has been demonstrated that IscU is a member of the miR-210 regulon (Chan et al., 2009) and the promoter of miR-210 contains a HRE for Hif1/2 binding (Kulshreshtha et al., 2007). Therefore, down-regulation of IscU may be explained to be through the miR-210-Hif2 axis in Irp2 −/− mice. The accompanied co-regulation of Fxn with IscU was often observed (Ferecatu et al., 2018;Li et al., 2018;Li et al., 2019), the detail regulation mechanism of Hif2α on Fxn remains to be explored. If mouse Fxn was regulated by Hif2α as reported (Oktay et al., 2007), it would expect that Fxn expression should be increased. Indeed, Irp2 depletion induced the downregulation of Fxn, which expression was reversed after inhibition of Hif2α by PT-2385.
The co-regulation of Fxn and IscU could be the key to response to PT-2385 treatment since the interaction of Fxn with IscU is important to facilitate Fe-S biogenesis (Fox et al., 2019;Gervason et al., 2019) to cure mitochondrial dysfunction. The similar work has been reported that neuronal Hif1α and Hif2α de ciency improves neuronal survival and sensorimotor function in the early acute phase after ischemic stroke (Barteczek et al., 2017). Wilkinson and Pantopoulos, 2013). However, the upregulated EPO expression in Irp2 −/− mice is probably invalid due to the iron limit in bone marrow (Cooperman et al., 2005). Moreover, the increased Hif2α endowed glycolysis-related genes, such as LdhA, Glut1, and Hk2, in the cerebellum and spinal cord of Irp2 −/− mice. Therefore, inhibition of Hif2 by PT-2385 did not only increase the expression of Fxn and IscU to strengthen mitochondrial function, but also decrease the expression of LdhA, Glut1, and Hk2 to weaken glycolysis to avoid the toxicity of high level of lactic acid. Although according to the astrocyte-neuron lactate shuttle hypothesis, lactic acid can be used as an energy metabolism substrate for neurons, Irp2 de ciency-induced mitochondrial dysfunction is insu cient to meet the energy needs of neurons.
Intriguingly, the elevated lactic acid was more in the spinal cord than in the cerebellum, which is consistent with the more severity in motor than in other behaviors (Jeong et al., 2011). The enhanced glycolysis could be the reason why the ATP production was slightly, but signi cantly, more in the spinal cord than in cerebellum after Irp2 ablation.
The patient with absence of IRP2 shows functional iron de ciency and mitochondrial dysfunction that emulate Irp2 −/− mice (Costain et al., 2019). The complete loss of IRP2 in patient-derived lymphoblasts also induces the decreased expression of complex subunits and activities of mitochondrial complex I and II (Costain et al., 2019), although the expression levels of Hifs, FXN, ISCU and glycolytic pathway-related proteins are not detected. We expect that Hif2α is upregulated in the tissues of CNS in patients as we observed in Irp2 −/− mice and the inhibition of Hif2 may be a therapeutic option for the IRP2-loss patients.

Conclusion
In summary, we have demonstrated that Irp2 ablation induces the expression of Hif2α, not Hif1α, in the tissues of CNS of Irp2 −/− mice and inhibition of Hif2α by PT-2385 dramatically protects from the neurodegenerative disorder through shifting the energy metabolism from glycolysis to oxidative phosphorylation, indicating that Hif2α is a potential target for neurodegenerative syndrome caused by loss of IRP2.