Activation of the Hepcidin-Ferroportin1 pathway in the brain and astrocytic–neuronal crosstalk to counteract iron dyshomeostasis during aging

During physiological aging, iron accumulates in the brain with a preferential distribution in regions that are more vulnerable to age-dependent neurodegeneration such as the cerebral cortex and hippocampus. In the brain of aged wild-type mice, alteration of the Brain Blood Barrier integrity, together with a marked inflammatory and oxidative state lead to increased permeability and deregulation of brain-iron homeostasis. In this context, we found that iron accumulation drives Hepcidin upregulation in the brain and the inhibition of the iron exporter Ferroportin1. We also observed the transcription and the increase of NCOA4 levels in the aged brain together with the increase of light-chain enriched ferritin heteropolymers, more efficient as iron chelators. Interestingly, in cerebral cortex and hippocampus, Ferroportin1 is mainly expressed by astrocytes, while the iron storage protein ferritin light-chain by neurons. This differential distribution suggests that astrocytes mediate iron shuttling in the nervous tissue and that neurons are unable to metabolize it. Our findings highlight for the first time that Hepcidin/Ferroportin1 axis and NCOA4 are directly involved in iron metabolism in mice brain during physiological aging as a response to a higher brain iron influx.

liver and spleen 9 . Recently, an extra-hepatic function of NCOA4 was demonstrated 10 . However, up to now, no data are available on brain NCOA4 and Hepc/Fpn1 expression and function during aging or neurodegeneration.
In the brain, iron regulates important functions such as neurotransmission, myelination and division of neuronal cells 11 . Iron reaches the brain crossing the Blood Brain Barrier (BBB) 12 . Iron up-take is then mediated by TfR1 expressed on the luminal side of brain capillaries 13 . Once inside the cell, iron is released into the cytoplasmic space and exported through the abluminal membrane by unknown mechanisms in which Fpn1 and other transporters may be involved 14 .
It has been shown that Hepc is present in the brain, in mature astrocytes and oligodendrocytes 15 , where it plays a role in the control of iron amount together with its own iron regulatory proteins 14 . However, it is not yet clear whether the Hepc acting on Fpn1 in the brain is the one produced in the liver or not 15 . Although the peptide size and its amphipathic cationic structure 16 would allow hepatic Hepc to pass the BBB, it has been shown that there is an endogenous cerebral Hepc expression 17 and that it responds to brain iron state 18 .
Several conditions which are typical of aging such as inflammation, BBB damage due to the release of inflammatory mediators, free radicals and vascular endothelial growth factor 19 cause iron redistribution and unbalance in the brain 20 .
With age, iron accumulates in the cerebral cortex (Ctx) and in the hippocampus (Hip), regions which are involved in neurodegenerative disorders 12 , but the underlying mechanism it is not yet known.
Here we demonstrated that NCOA4, Hepc and Fpn1 are activated in WT mice brain during physiological aging as a consequence of iron accumulation and that they participate to brain response to increased iron entry. Furthermore, we assessed the astrocytic-neuronal crosstalk and we found that the iron exporter Fpn1 co-localizes with astrocytes, while neurons are enriched in the iron deposit Ft-L heteropolymers, both in the Ctx and Hip. These data suggest that, while glial cells enhance iron export in the nervous tissue, neurons accumulate it, triggering neurodegenerative processes.

Results
Iron amount and distribution in the brain during aging correlates to the level of BBB permeability. Brain Iron Content (BIC) increases during aging at each experimental time point (Fig. 1A) in different brain areas as shown by histochemical analysis with DAB-enhanced Prussian blue Perls' staining ( Fig. 1B). WT O mice show an increased number of brown precipitates compared to WT A mice in specific parenchymal region such as Ctx, Hip CA regions, third ventricle (3 V) and striatum. Similarly, we also observed an agedependent increase in the levels of iron in liver and in serum of old mice indicating a general perturbation of iron metabolism with physiological aging (Supplementary Figure 1S). Since progressive BBB damage is occurring not only in neurodegeneration 21 but also during aging, we analysed Zonula occludens-1 (ZO-1) protein, whose role is to maintain the compactness of BBB acting as a bridge connecting Claudin and Occludin proteins to the actin cytoskeleton in order to stabilize the tight junction (TJ) structure 22 . ZO-1 levels significantly decrease during aging (Fig. 1C), therefore, we can hypothesize that age-dependent BBB altered permeability, could contribute, together with age-dependent metal dyshomeostasis, to iron accumulation in specific areas of the brain during physiological aging.
Increased inflammatory and oxidative stress state during brain aging. The two main markers of neuroinflammation and oxidative stress, Serum amyloid A1 (SAA1) 23 and Nuclear factor erythroid 2-related factor 2 (Nrf2) 24 , are overexpressed in aged brains. SAA1 expression levels is more than 20 times higher in WT O animals compared to WT A ( Fig. 2A) and Nrf2 expression levels are constantly increasing during aging (Fig. 2B).
Moreover, we performed immunohistochemistry to selectively label reactive intermediate filament protein (GFAP)-positive astrocytes. In fact, GFAP is an indicator of neuroinflammation in the CNS 25 and it is also involved in the progression of neurodegeneration in ischemia, AD, MS, Amyotrophic Lateral Sclerosis (ALS) and PD [26][27][28] . In addition, we also checked the expression of the Ionized calcium-binding adaptor molecule 1 (IBA-1), a microglia/macrophage-specific calcium-binding protein which is also a key molecule in proinflammatory processes 29 . We identified high astrocytes activation and an increased expression of microglia in both parenchymal regions of WT O mice where iron accumulated, Ctx and Hip, compared to those of WT A ( Fig. 2C and D).
These data show that iron accumulation in the brain is accompanied by the neuroinflammatory and antioxidative stress response.
Hepc/Fpn1 activation and ferritins response to iron accumulation during brain aging. In order to evaluate if the Hepc/Fpn1 axis has a role during brain physiologic aging, we measured both Hepc and Fpn1 in the whole brain of aged mice. Interestingly, we observed that Hepc gene expression significantly increases in WT M-A and WT O mice brain (Fig. 3A), while Fpn1 protein decreases (Fig. 3B). To investigate how neuronal cells responded to the increase of iron amount, we analysed in the total brain also the iron deposit protein ferritins (Ft) and separately evaluating the two polymers: ferritin light-chains (Ft-L) and ferritin heavy-chains (Ft-H). As expected, we observed a significant increase in Ft-L amount (Fig. 3C), but, surprisingly a 40% reduction of Ft-H in WT O animals' brains compared to the WT A (Fig. 3D). We also checked for NCOA4 levels of transcription and translation in the brain. NCOA4 gene results to be highly transcribed in the brain and its expression is comparable to that of the liver (Ct values 25 ± 1 and 24 ± 1.5 respectively) ( Fig. 3E and 9 ).
Furthermore, NCOA4 protein amount is also significantly increased in WT O mice brain compared to WT A (Fig. 3F).
Altogether, these results demonstrate that in old mice brain iron accumulation together with the inflammatory condition ( Fig. 2A-C and D) induces Hepc expression and, consequently, Fpn1 degradation; therefore, activation

Discussion
During aging and in neurodegenerative diseases with old age onset such as PD and AD, an increase in iron content was observed in multiple brain regions 30,31 . In pathological conditions, it was demonstrated to be the cause of motor deterioration 12 and of proteins aggregation 32 leading to cellular stress 33 . Parallel to deposition of iron in the brain, in the periphery, systemic iron levels decrease and old subjects are subjected to anemia 34 . www.nature.com/scientificreports/ Systemic iron regulation is based on a complex protein regulatory system in which the hepatic Hepcidin (Hepc) plays a major role. Indeed, the iron dependent modulation of Hepc expression determines de facto iron availability in the body 35 . In the brain, iron homeostasis is regulated by the same proteins network that acts at the systemic level 30 and the Hepc regulatory system is active also in the CNS 31 . Indeed, Hepc is expressed by glial cells and neurons from different brain regions and, under brain iron accumulation, it is activated and it induces Fpn1 decrease 36,37 . However, it is not clear yet whether this rely on brain or hepatic Hepc 15 . Moreover, it is not known how this regulatory system respond to intracerebral iron increase during aging. Intrigued by this question, we studied the brain expression of proteins involved in systemic iron homeostasis in wild type (WT) mice during aging.
We characterized the state of the brain at different ages by studying BBB integrity, brain inflammation and oxidative state, all key features related to the process of aging and that influence iron homeostasis (Figs. 1 and 2). It is known that BBB mitigates iron entry from the blood to the brain through highly regulated and selective systems: iron crosses the BBB bound to Tf through TfR-mediated endocytosis 38 and brain vascular endothelial cells (BVECs) export intracellular iron using Fpn1, whose activity is conditioned by the iron ferroxidases ceruloplasmin and hephaestin 31 . Finally, iron is acquired by nervous cells through iron transporter proteins, as DMT1, and released from these cells through Fpn1 31 .

Figure 2.
Iron-dependent inflammatory response and oxidative stress during aging. (A) Real-time PCR of SAA1 in total brain from all genotypes. (B) Nrf2 mRNA expression levels in total brain. The expression levels of the two genes were normalized to levels of β-glucuronidase (Gus-β) housekeeping gene (material and methods section). (C) Immunofluorescence anti-GFAP (green) and anti-IBA1 (pink) antibodies in cerebral cortex (Ctx) and hippocampus (Hip); 4,6-diamidino-2-phenylindole (DAPI) (blue) was used to counterstain cell nuclei. Scale bars:40X. *Statistically significant vs WT A control group *P < 0.05; **P < 0.01 ***P < 0.001 using OneWay www.nature.com/scientificreports/ It is also known that iron accumulation in the brain, triggers the release of pro-inflammatory cytokines, determining an environment prone to neurodegeneration 39 . www.nature.com/scientificreports/ Indeed, we demonstrated the progressive accumulation of iron during physiological aging in the Ctx, Hip, third ventricle and striatum and the parallel decrease of the BBB integrity. As a consequence of iron accumulation, the transcription of SAA-1, a protein related to acute inflammation and marker of neuroinflammation 40,41 , described also in AD as able to stimulate the release of cytokines and chemokines 23,41 , increases up to 1000 times in old mice brain. Moreover, the transcription of Nrf2, a redox-sensitive transcription factor 24 , is also increased, supporting the evidence of a stressful condition in WT O mice brain.
During aging, a consistent activation of astrocytes and a generalized neuroinflammation are evident 42 . In line with these findings, in both Ctx and Hip of WT O mice we observed high astrocytic and microglial activation.
Interestingly, in this context of increased iron deposition and inflammation in the brain, we found the activation of the Hepc/Fpn1 pathway: brain Hepc transcription increases and brain Fpn1 amount gradually decreases during aging. These observations are in line with what Sato and colleagues observed in the cerebral cortex and in mitochondria isolated from the brain of aged mice 43 . To better decipher the mechanism of the regulation of iron content in neuronal tissue during physiological aging, we also analyzed the iron deposit protein Ft and a newly characterized protein, NCOA4, since it is involved in Ft degradation and its inactivation in mice causes iron accumulation in the liver 9 . Specifically, NCOA4 promotes autophagic ferritin degradation through its binding to Ft-H subunit 7,44 . Ferritin levels are enhanced in a cellular model (HeLa cells) in which NCOA4 is silenced, suggesting that ferritin is constantly degraded by an NCOA4-dependent pathway 45 . Moreover, when NCOA4 knockdown is selectively targeting hepatocytes, the protein silencing promotes an increase in both iron amount and ferritin levels 46 .
Surprisingly, in old mice's brains we found an increased amount of NCOA4, contrary to what happens in the liver 9 . Furthermore, specifically evaluating the ferritin polymers, we observed an increase of Ft-L and a decrease of Ft-H chains in the aged mice brains. These data demonstrated that a differential Ft chains degradation occurs www.nature.com/scientificreports/ in both cortical and hippocampal neurons of old animals. We can suppose that Ft-L enriched heteropolymers are more efficient in iron chelation 3 and are also more abundant in cortical and hippocampal neurons. Interestingly, when we analysed Fpn1 localization in the brain by immunofluorescence, we found that Fpn1 colocalized with astrocytes, both in the Ctx and Hip. On the contrary, Ferritin accumulated in cortical and hippocampal neurons close to the soma, but not in astrocytes. We suggest that this could be due to a different detoxifying mechanism carried out by neurons and astrocytes, aimed either to store or remove iron excess, respectively. On the whole, this data revealed that aging dependent brain iron accumulation compromises the cells specific response: astrocytes, which are less susceptible than neurons to iron deposits-related toxicity 47 and which play a protective role towards neurons 42 , have an increased iron export, while, neurons increase the metal storage in Ft-L rich heteropolymers. These deposits could trigger the neuronal death in Ctx and Hip evidenced during aging and even more during neurodegeneration 30 .
Furthermore, we showed for the first time that NCOA4 is transcribed in brain cells and that its expression is increased in WT O animals brain.
In conclusion, we demonstrated that even during physiologic aging, iron accumulates in the brain and that its accumulation, selectively localized in the Ctx and Hip, triggers neuroinflammation and the modification of the Hepc/Fpn1 pathway, all this enhancing iron availability imbalance and oxidative stress that could lead to neurodegeneration (Fig. 5).
In perspective, since NDs are characterized by inappropriate Hepc production 48 , a therapeutic approach aimed at modifying the Hepc response could be taken in consideration. Different strategies could be used, such as mini-Hepc and Hepc agonists 49,50 by the stimulation/inhibition of Hepc production by targeting its regulators 35,48,[51][52][53] . Additional research studies in animal models of NDs are required to clarify the CNS response to the increased iron aimed to exploit the results for the prevention and clinical management of patients with these diseases.

Methods
Animals. C57BL/6 J mice (WT) used for the study were purchased from the Jackson Laboratory and subdivided for age according to its classification (https:// www. jax. org): until 2 months of age mice are considered Young (WT Y n = 5); from 2 to 6 months of age Adult (WT A n = 7); from 6 to 12 months of age Middle-aged (WT M-A n = 5) and from 12 months of age (between 18-24 months of age) Old (WT O n = 5) (Table 1S). Since www.nature.com/scientificreports/ from a pilot analysis on potential gender-related issues, no gender bias was observed (Supplementary Figure 2S), both male and female mice were analysed and grouped according to their age. Immunoblotting. The Fpn1, Ft-H, Ft-L, NCOA4 and Zonula occludens-1 (ZO-1) proteins' amount in the whole brain homogenates was evaluated by Western Blotting using specific antibodies. 50 μg of total brain lysates were separated on 6-12% SDS polyacrylamide gel and immunoblotted 55 58 , sections were exposed to Cy2-, Cy3-(Jackson Immu-noResearch Laboratories, West Grove, PA) and 647 Alexa Fluor-conjugated secondary antibodies (Molecular Probes Inc, Eugene Oregon) for 1 h at room temperature. DAPI (4,6-diamidino-2-phenylindole, Fluka, Italy) was used to counterstain cell nuclei. After processing, sections were mounted with Tris-glycerol supplemented with 10% Mowiol (Calbiochem, LaJolla, CA). The samples were examined by a Leica TCS SP5 confocal laser scanning microscope (Leica, Mannheim); z-stacks images were taken at 40X and 63X magnification.
Iron parameters. Brain nonheme iron content (BIC) was evaluated using 20 mg of dissected and dried murine whole brains 59 . Perfused brains were stained for nonheme ferrous iron by Prussian blue Perl's using a commercial kit (Bio-Optica, Milan, Italy). To improve the sensitivity, an intensification step with DAB (3-3′-diaminobenzidine tetrahydrochloride) 60 was performed. Images were taken at 10X magnification using a Leica DM4000B automated microscope with IM50 program for acquisition (Leica Microsystems, Wetzlar, Germany).
Statistical analysis. One-way ANOVA followed by Bonferroni's post hoc analysis or two-tailed Student's t-test were applied according to the experimental group's number. P values of < 0.05 were considered as statistically significant. Analyses were performed with Image Lab 4.0.1 and GraphPad Prism 7.00. Data were expressed as average ± SD of the mean. Significance was defined as *P < 0.05, **P < 0.01 and ***P < 0.001. WT adult (A) mice were used as normalizer. The number of samples in each experimental condition is indicated in the figure legends. In each Western Blotting experiment, we reported 3 samples per group.

Data availability
The data regarding reference genes for Nrf2, NCOA4 and Gus-B primers are openly available in the repository "Nucleotide" at https:// www. www.nature.com/scientificreports/