Therapeutic effect of a histone demethylase inhibitor in Parkinson’s disease

Iron accumulation in the substantia nigra is recognized as a hallmark of Parkinson’s disease (PD). Therefore, reducing accumulated iron and associated oxidative stress is considered a promising therapeutic strategy for PD. However, current iron chelators have poor membrane permeability and lack cell-type specificity. Here we identified GSK-J4, a histone demethylase inhibitor with the ability to cross blood brain barrier, as a potent iron suppressor. Only a trace amount of GSK-J4 significantly and selectively reduced intracellular labile iron in dopaminergic neurons, and suppressed H2O2 and 6-OHDA-induced cell death in vitro. The iron-suppressive effect was mainly mediated by inducing an increase in the expression of the iron exporter ferroportin-1. In parallel, GSK-J4 rescued dopaminergic neuron loss and motor defects in 6-OHDA-induced PD rats, which was accompanied by reduction of oxidative stress. Importantly, GSK-J4 rescued the abnormal changes of histone methylation, H3K4me3 and H3K27me3 during 6-OHDA treatment although the iron-suppressive and neuroprotective effects were sensitive to H3K4me3 inhibition only. Also, upregulating H3K4me3 increased ferroportin-1 expression and neuroprotection. Taken together, we demonstrate a previously unappreciated action of GSK-J4 on cell-specific iron suppression and neuroprotection via epigenetic mechanism. Compared with conventional iron chelators, this compound has a stronger therapeutic potential for PD.


Introduction
Parkinson's disease (PD) is the second most common neurodegenerative disease, which affects millions of people worldwide and has been posing a heavy burden on the healthcare system in many regions 1,2 . PD patients typically suffer from resting tremor, rigidity, bradykinesia, and posture instability. It is widely believed that most typical PD symptoms are the results of dopaminergic (DA) neuron loss in the midbrain substantia nigra (SN). At present, PD still remains incurable and only symptomatic treatments, such as L-DOPA medication and deep brain stimulation, are available. With disease progression, however, these symptomatic medications lose efficacy 2,3 .
There is thus a significant unmet need for new medications capable of slowing or preventing PD progression by blocking SN neuron death.
Intense research effort in the past has led to the discovery of several central hallmarks in PD, including, alpha-synuclein aggregation, mitochondrial dysfunction, and iron accumulation 2,4,5 . Among all, iron accumulation in SN is one of the most invariable pathological findings in PD 4,[6][7][8] . Iron is vital to the function of a wide variety of proteins. However, labile iron, through Fenton reaction, can generate free radicals from by-products of mitochondrial respiration, triggering oxidative stress 9 . Although it is known that iron accumulation can induce oxidative stress and therefore is toxic, the exact role of iron accumulation in SN of PD patients still remains largely unknown. According to transcranial sonography and other live brain iron detection methods, it is believed that increased level of iron in SN can be one of the earliest pathological changes in PD 6,7,10 . Moreover, increased non-transferrin-bound iron (NTBI) uptake via divalent metal transporter 1 (DMT1) 11 , transferrin-bound iron (TBI) uptake into mitochondria mediated by transferrin/ transferrin receptor 2 (Tf/TfR2) 12 and reduced iron export via ferroportin-1 (Fpn1), the sole iron export protein identified 13 , have been found associated with iron accumulation in PD 7,10 . The discovery of the genetic association between iron-related proteins and PD occurrence further supports the hypothesis that iron dysregulation could play a key role in causing PD.
A number of reports have demonstrated that oxidative damages, potentially caused by iron-induced oxidative stress, may appear to be one of the earliest pathological changes in PD 6,8 . Recent findings that iron accumulation and oxidative stress may trigger and accelerate pathological changes of PD, together with the discovery of the association of a number of iron-related and oxidative stress-related genes with PD, provide further supporting evidence that iron accumulation and elevated oxidative stress may be early PD pathogenic events causing DA neurons loss 2,8,14 . Consistently, treatments with iron chelators and antioxidants have been found effective in PD, suggesting that iron chelators and antioxidants might be promising targets for disease-modifying treatment of PD 2,15,16 . However, iron-suppressing treatment has yet been widely adopted due to major limitations concerning blood-brain-barrier (BBB) permeability and cell-type specificity of current iron-suppressing drugs 17,18 .
GSK-J4 is a newly developed cell-permeable prodrug that, once entered into cells, can be hydrolyzed by esterase to GSK-J1, a potent inhibitor of H3K27me3/me2 and H3K4me3/me2 histone demethylase (also known as KDM6A/B and KDM5B/C, respectively) 19,20 . A number of reports have suggested a possible application of GSK-J4 in cancer treatment and inflammation suppression based on its inhibitory effects on histone demethylases 21 . In some studies, silencing histone demethylases, such as KDM5C, the most abundant form of KDM5 in the brain, is associated with beneficial effect in neurodegenerative diseases such as Huntington's disease 22 . Due to the capability of removing Fe 2+ from targeted demethylases, GSK-J4/1 has also been suggested to be a possible iron chelator 19 . Therefore, in this study, based on the well-established models of PD, experiments were designed to investigate if GSK-J4 treatment is beneficial in models of PD and to elucidate the detailed underlying mechanisms.

Selective iron suppression effect of GSK-J4 on DA neurons
To examine the effects of GSK-J4 on the potentially toxic intracellular labile iron pool (LIP) in cells, a calcein-AM assay was performed. Since the fluorescence of calcein is partially quenched upon binding to iron, compounds that displace intracellular iron from its complex with calcein would result in an increased fluorescence signal. As shown in Fig. 1a, b, SH-SY5Y cells treated with the iron supplement FAC (50 μg/ml) for 12 h displayed a lower calcein fluorescence level than the control group (P < 0.05). However, calcein fluorescence increased with the addition of 0.5 μM GSK-J4 (P < 0.001, compared with control group), suggesting a decrease in free labile iron in the SH-SY5Y cells with the treatment of GSK-J4. Moreover, the addition of FAC (50 μg/ml) inhibited the GSK-J4-mediated increase of calcein fluorescence (P < 0.01, compared with GSK-J4 group), further indicating that the elevated calcein signal by GSK-J4 was caused by the iron suppressing effect of GSK-J4.
We further characterized the effect of GSK-J4 on cellular iron homeostasis by examining the expressions of the iron importer protein TfR1 and the iron storage protein ferritin light chain (FtL). As shown in Fig. 1c, d in SH-SY5Y cells treated with the iron chelator deferoxamine (DFO, 50 μM), TfR1 synthesis was increased compared to untreated control cells (P < 0.01), whereas synthesis of FtL is strongly inhibited (P < 0.01). Interestingly, similar to DFO, only a trace amount of GSK-J4 (0.5 μM) treatment caused a significant increase of TfR1 (P < 0.001) and a decrease of FtL (P < 0.01). These findings are consistent with that under the condition of free labile iron depletion, IRP1 stabilizes TfR1 mRNA to prolong its half-life, while inhibiting the translation of FtL mRNA 23,24 . These effects enhance iron uptake into cells and stops it from being sequestered away for storage. However, the effects of GSK-J4 treatment on FtL and TfR1 were not found in HEK293 (Fig. 1e, f) and HepG2 cell line (Fig. 1g, h). In contrast, the classical iron chelator DFO (50 μM) caused a significant increase of TfR1 (P < 0.01, compared with each control group) and a significant decrease of FtL (P < 0.01, compared with each control group), in all these cell lines ( Fig. 1c-h). In addition, the iron suppression effect was also found in the MES23.5 cell line ( Supplementary Fig. 1a, b). Since SH-SY5Y and MES23.5 both are widely used DA neuronal cell lines, these results indicate that GSK-J4 has a selective iron suppressing effect on DA neurons.

GSK-J4 displays a potent neuroprotective effect in 6-OHDA-induced PD model in vitro
Considering the cell-type specificity of the ironsuppressing effect of GSK-J4, we hypothesize that it is beneficial in PD. To investigate the effect of GSK-J4 on a PD model in vitro, SH-SY5Y cells were treated with the neurotoxin 6-hydroxydopamine (6-OHDA). As shown in Fig. 2a, 6-OHDA induced a dose-dependent shrinkage, detachment, and cell death in SH-SY5Y cells in 24 h. Compared with 6-OHDA-treated cells, pre-treatment with GSK-J4 (0.5 μM) clearly protected cells against 6-OHDA toxicity, even at a high concentration of 6-OHDA (40 μM). The protective effect of GSK-J4 was also In addition, the common iron chelator deferoxamine (DFO, 50 μM, for 24 h) caused a decrease of FtL in all tested cell lines. The relative protein levels of FtL was normalized to β-Actin. *P < 0.05; **P < 0.01; ***P < 0.001; ns means no significant difference. Data are presented as mean ± SEM. n = 3-4 for each group. quantitatively assessed by the MTT assay (Fig. 2b). For example, upon 40 μM of 6-OHDA treatment for 24 h, 0.5 μM of GSK-J4 significantly increased cell viability from 30.5 ± 10.9 to 86.6 ± 8.4% (P < 0.001). On the other hand, treatment with GSK-J4 alone (0.5 μM) did not have a significant effect on morphology and cell viability. These data suggest that GSK-J4 showed a powerful neuroprotection in the 6-OHDA-treated SH-SY5Y cells. We also examined the expressions of two important markers of cell apoptosis, namely cleaved Caspase-3 and Bcl-2 by Western blot. 6-OHDA treatment increased the level of cleaved caspase-3 to 160% (Fig. 2c, d) and decreased the level of Bcl-2 to 51% when compared with the control group, while pretreatment with GSK-J4 significantly prevented these changes (Fig. 2c, d). These data suggest that GSK-J4 decreased the 6-OHDA-induced apoptosis in SH-SY5Y cells.

Neuroprotective effect of GSK-J4 in PD models in vivo
To explore the neuroprotective and therapeutic effects of GSK-J4 in in vivo PD model, SD rats were micoinjected with 6-OHDA (8 μg/2 μl) at the medial forebrain bundle (MFB) in the right side of the brain. GSK-J4 (1 μg/ 1 μl) was injected into MFB one day before the 6-OHDA administration (Fig. 3a). Three weeks after injection, the relevant behavior tests were performed before sacrifice of the animals for histological and biochemical assessments. In the cylinder test, unilateral 6-OHDA injection decreased the use of the contralateral forelimb to about 20% while GSK-J4 treatment partially but significantly rescued this motor defect, in which the use of contralateral forelimb was almost 40% (Fig. 3b). In the apomorphine-induced rotation test, GSK-J4-treated rat exhibited far fewer number of contralateral rotation (15 rpm) in 5 min when compared with the unilateral 6-OHDA-treated rat, which was about 36 rpm in the same period of time (Fig. 3c). These results indicate that GSK-J4 can prevent the 6-OHDA-induced motor defects in rats. At the same time, based on tyrosine hydroxylase (TH) staining, it was found that 6-OHDA injection caused a significant decrease in the number of TH-positive neurons in SN, which was about 25% of the saline-treated side. With GSK-J4 treatment, 50% of TH-positive neurons survived after 6-OHDA injection indicating a neuroprotective effect of GSK-J4 (Fig. 3d, e).
To establish the link between the beneficial effect of GSK-J4 and iron level in the PD animal model, we first asked whether iron was altered in the unilateral 6-OHDAinduced rat model. Iron staining (Fig. 3f, g) showed that the number of iron-positive cells significantly increased in the 6-OHDA-lesioned SN, indicating increased iron content after 6-OHDA treatment. However, the treatment of GSK-J4 significantly inhibited this increase of iron, suggesting that GSK-J4 can reduce iron accumulation in response to 6-OHDA treatment.

Anti-oxidative stress is involved in the neuroprotection of GSK-J4
Iron-induced oxidative stress as a result of iron accumulation is regarded as an important cause of neurodegeneration in many experimental PD models. Given the iron suppression effect of GSK-J4 that we observed, we investigated whether anti-oxidative stress plays a role in the neuroprotective effect of GSK-J4 in PD model.
We determined the levels of different oxidative stress markers in the SN of 6-OHDA-treated rats in the absence and presence of GSK-J4 pre-administration. It was found that GSK-J4 could significantly suppress the 6-OHDAinduced increase of reactive oxygen species (ROS) production, as measured by DCF fluorescence, from 176 ± 7.9% to 131 ± 6.3% (Fig. 4a), the increase of MDA level from 0.79 ± 0.10 to 0.52 ± 0.12 nmol (Fig. 4b), the increase of protein carbonyl level from 5.84 ± 0.54 to 3.52 ± 0.25 nmol (Fig. 4c) and the decrease of total GSH level from 10.27 ± 1.05 to 15.27 ± 1.08 μg/mg (Fig. 4d). These results indicate that GSK-J4 acts as a strong antioxidant in the 6-OHDA-induced PD model.
To further confirm the antioxidant nature of GSK-J4, we tested the effect of this compound on SH-SY5Y cells treated with H 2 O 2 . We found that GSK-J4 at 0.25 μM prevented H 2 O 2 -induced changes in cell death reflected by changes in morphology (Fig. 4e). In both 150 and 300 μM concentrations of H 2 O 2 , pretreatment with GSK-J4 maintained a high level of cell viability, confirmed with MTT assay (Fig. 4f).

GSK-J4 prevents abnormal histone methylation in 6-OHDA model
As a selective inhibitor of histone demethylases including KDM5 and KDM6, GSK-J4 could increase the levels of H3K4me3 and H3K27me3, respectively 19,20 . Therefore, apart from the effect on iron level and oxidative stress, we asked whether the effects of GSK-J4 on H3K27me3 and H3K4me3 contribute to its therapeutic action in PD animals. As shown in Fig. 5a treatment when compared with the control group but pretreatment with GSK-J4 significantly attentuated this decrease (P < 0.05, when compared with the 6-OHDAlesioned group, Fig. 5a, b). These data reveal abnormal levels of H3K4me3 and H3K27me3 in PD model that could be rectified by GSK-J4. The cylinder test showed that GSK-J4 rescued the unilateral 6-OHDA-induced asymmetry of the forelimbs. c Consistent with the above, the apomorphine-induced rotation in the unilateral 6-OHDA-treated rat was suppressed by the treatment of GSK-J4. d, e Tyrosine hydroxylase staining showed that MFB injection of 6-OHDA caused a significant decrease of the TH-positive neuron in SN (−4.80 to −6.12 mm from Bregma), while GSK-J4 treatment rescue the level of TH neuron. Scale bar, 100 μm. f, g 6-OHDA treatment caused a significant increase in iron accumulation in the SN, which was prevented by the treatment of GSK-J4, as reflected by Perls'-DAB staining. *P < 0.05; **P < 0.01; ***P < 0.001. Data are presented as mean ± SEM. Scale bar, 200 μm. n = 8 for the behavioral test, and n = 6 for TH staining.  To further explore and distinguish the roles of H3K27me3 and H3K4me3 in the protective effect of GSK-J4 on the 6-OHDA-induced PD model, two sets of experiments were performed. First, SH-SY5Y cells were treated with GSK-J4 (0.5 μM) followed by treatment with a highly selective EZH2 methyltransferase inhibitor (GSK126, 1 μM) that could potently inhibit H3K27me3 25,26 . As shown in Fig. 5c, in 6-OHDA-treated cells, the increase of H3K27me3 induced by GSK-J4 was inhibited by GSK126 (P < 0.001, when comparing the 6-OHDA+GSK-J4 group with 6-OHDA+GSK-J4+GSK-126). However, MTT assay showed that there was no significant difference in the cell viabilities of these two groups (Fig. 5d). Next, we performed similar experiments to determine the role of H3K4me3. In SH-SY5Y cell line, GSK-J4 treatment (0.5 μM, for 24 h) increased not only the basal levels of H3K27me3 and H3K4me3 but also rectified the decrease in their levels under 6-OHDA-treatment (25 μM, for 24 h). c, d SH-SY5Y cells were pretreated with GSK-J4 (0.5 μM, for 1 h), followed by the treatment of 6-OHDA (25 μM, for 24 h) with or without GSK126 (1 μM, a specific inhibitor of H3K27me3). As reflected by Western blot c, H3K27me3 was inhibited under the treatment of GSK126. ***P < 0.001. Data are presented as mean ± SEM. n = 3 for each group. However, d MTT assay showed that no inhibition effect from GSK126 was observed on the neuroprotection of GSK-J4 in the 6-OHDA-induced model. e, f SH-SY5Y cells were pretreated with GSK-J4 (0.5 μM, 1 h), followed by the treatment of 6-OHDA (25 μM, 24 h) with or without DZNep (1 μM, decreasing H3K4me3 level). H3K4me3 was inhibited under the treatment of DZNep e, while MTT assay revealed a partial inhibitory effect by DZNep f. *P < 0.05; **P < 0.01; ***P < 0.001. Data are presented as mean ± SEM. ns means no significant difference. n = 3-4 for each group.

H3K4me3 plays a role in iron metabolism under GSK-J4 treatment
Since GSK-J4 was found to selectively regulate iron metabolism in SH-SY5Y cells, while histone modifications is often locus type and cell type specific [28][29][30] . To fully elucidate the therapeutic mechanism of action of GSK-J4, we explored the relationship between H3K4me3 and iron homeostasis. First, in SH-SY5Y cells, Western blot analysis revealed that, when compared with the control cells, GSK-J4 treatment (0.5 μM, 24 h) significantly increased the levels of iron uptake proteins TfR1 (P < 0.01), the iron export protein Fpn1 (P < 0.01), and decreased the iron storage protein FtL (P < 0.01) (Fig.  6a, b). Together, these data are consistent with that increased iron export by Fpn1 is the primary cause of iron-suppressing effect of GSK-J4. Indeed, iron storage protein FtL was largely increased (P < 0.01) by 6-OHDA treatment when compared with control cells, while a marked decrease of FtL was found in the 6-OHDA plus GSK-J4 treatment group (P < 0.01, when compared with 6-OHDA-treated cells). Moreover, GSK-J4 could significantly prevent the 6-OHDA-induced decrease of Fpn1 and TfR1. In another set of experiments, it was found that the increase of Fpn1 could be inhibited by the addition of DZNep (decreasing H3K4me3) (Fig. 6c, d) while not by the addition of GSK126 (decreasing H3K27me3) (Fig. 6e,  f). Consistent with the effect on iron metabolism, the effects of GSK-J4 on H3K4me3 and H3K27me3 were only found in SH-SY5Y cell line but not in HEK293 or HepG2 cell lines (Supplementary Fig. 2a, b). Furthermore, realtime PCR experiments showed that GSK-J4 could also upregulate the expression of Fpn1 at the mRNA level (Fig. 6g). These results together implicate that H3K4me3 plays a role in the iron metabolism under GSK-J4 treatment.

Upregulation of H3K4me3 increases iron exporter Fpn1 and exhibits neuroprotection
To further probe the role of H3K4me3 in cell-specific iron regulation of GSK-J4, SH-SY5Y cells were pretreated with CPI-455, a specific inhibitor of KDM5, which could increase the global level of H3K4me3 31 . We found that CPI-455 (1 and 5 μM) caused a significant increase (over 2-fold) in the global level of H3K4me3 (P < 0.01, when compared with the control group, Fig. 7a, b). Under this condition, CPI-455 treatment also resulted in an increase in the iron exporter Fpn1. In addition, CPI-455 treatment improved cell viability in response to 6-OHDA exposure (Fig. 7c).
KDM5C is the main histone demethylase that can specifically alter H3K4 methylation in several cell types 31,32 . To further probe the role of H3K4me3 in regulating iron metabolism, the effect of AAV-KDM5Csilence virus was tested. As shown in Fig. 7d, KDM5C silencing strongly decreased KDM5C levels to 17% of control based on the mRNA level and to 16% of control based on the protein level (Fig. 7e, f).
Remarkably, significant increases of Fpn1 and H3K4me3 were observed under this condition (Fig. 7e, f). Also, KDM5C silencing improved the viability of cells under 6-OHDA treatment (Fig. 7g). Thus, KDM5Cmediated upregulation of H3K4me3 is involved in the increase of iron exporter (Fpn1) underlying the neuroprotective effect of GSK-J4.

Discussion
In this study, we identified GSK-J4 as a potent and cell type-specific suppressor of intracellular labile iron, which confers neuroprotection from oxidative stress in 6-OHDA-induced PD model, alleviating Parkinsonian motor deficits. Importantly, we found that this effect is largely attributable to the upregulation of H3K4me3 by GSK-J4. A higher level of H3K4me3 expression leads to an increased iron export via the iron exporter Fpn1. Therefore, we discovered a novel and effective mechanism of neuron-specific suppression of iron level with clear clinical implication for neurodegenerative diseases.
The decrease in intracellular iron caused by GSK-J4 is likely due to alterations in iron transport and storage proteins. If the iron exporter (Fpn1) increases, it will cause a decrease in cellular iron and a corresponding decrease in iron storage (FtL) and an increase in iron uptake (TfR1). However, if the iron uptake TfR1 increases, it will cause an increase in cellular iron with a corresponding increase in FtL and an increase in Fpn1. Thus, it is reasonable to assert that an increase of Fpn1 is the initial effect of GSK-J4. In fact, our data showed that Fpn1 is epigenetically regulated in cells treated with GSK-J4. A major consequence of this change is the reduction in oxidative stress, which is believed to be a common mechanism involved in many neurodegenerative diseases, including PD 4,6,7,14 . In fact, clinical studies have found that elevated iron in SN can trigger and exacerbate oxidative stress, leading to DA neuronal death 7,14,16,17 . Consistently, in this In this study, the iron-suppressive effect of GKS-J4 is cell type-specific and only trace amounts of this compound can significantly reduce the iron storage protein FtL. These findings strongly suggest that the effect of GSK-J4 cannot be solely explained by its ability to directly c MTT assay showed CPI-455 showed neuroprotection against the toxicity of 6-OHDA in SH-SY5Y cells. d-g SH-SY5Y cells were infected by AAV-KDM5C-silencing virus, and KDM5C was significantly decreased at the mRNA level as reflected by RT-PCR d and at the protein level by Western blot e, f. In addition, Fpn1 and H3K4me3 were increased in AAV-KDM5C-silencing virus infected cells e, f. Moreover, g silencing KDM5C showed a slight neuroprotection against 6-OHDA-induced PD model in SH-SY5Y cells. *P < 0.05; **P < 0.01; ***P < 0.001. Data are presented as mean ± SEM. n = 3-4 for each group in Western blot and n = 5 for each group in MTT assay. chelate iron. In fact, the most important discovery of the present study is the epigenetic cause of the ironsuppressive and neuroprotective effect of GSK-J4, as reflected by the involvement of H3K4me3. Cell-type specificity is known to be an important feature of epigenetic modification in regulating gene expression 30 . Our study indicated that the effects of GSK-J4 on H3K4me3 and H3K27me3 were only found in SH-SY5Y cell line but not in HEK293 or HepG2 cells. Although the reason for this cell specificity is not known, different microenvironments in different cell lines may underlie different responses to the treatment of GSK-J4. Further exploration on this issue is needed. It is noteworthy that, similar to the phenomenon we observed, Fpn1 could be regulated at both the transcription and post-translation level 33 , and in a cell-type-dependent manner at the transcription level 34 . GSK-J4 is currently regarded as a useful small molecular tool of epigenetic modification, especially for histone methylation, including the H3K4 and H3K27 trimethylation levels 19 . Methylation of histones can induce different changes in gene expression, depending on the specific site of methylation 35 . In general, trimethylation of H3K4 is associated with transcriptional activation of gene expression, while trimethylation of H3K27 is associated with transcriptional inhibition of gene expression 28,35 . With respect to regulation of iron level, GO gene analysis in KDM5 silencing in flies revealed its close relationship with the metal iron transport pathway 36 , which further indicated the potential role of H3K4me3 in the regulation of iron metabolism.
Most PD cases are idiopathic in nature and do not have a clear genetic link. On the other hand, since epigenetic modification could account for the interaction between gene expressions and environmental insults, epigenetic changes have recently been suggested to play a role in PD pathogenesis 37 . Among all examined epigenetic histone modifications in PD, histone methylation remains relatively unknown. In this study, it was found that 6-OHDA itself caused a significant decrease in H3K4me3 and H3K27me3. Indeed, other reports found that perinucleolar and perinuclear H3K27me3 levels decreased in the SN upon 6-OHDA injection 38 . Although previous studies have shown that L-DOPA can induce phosphorylation of H3K27meS28 in the striatum in the unilateral 6-OHDAlesioned model, local changes in H3K27me3 and H3K4me3 have not been established 39 . This may be due to the different expression patterns between the different cells in these brain regions 40 . Our present findings strengthen the possibility that histone methylation may play a role in the pathogenesis of PD.
KDM5 is involved in regulating transcription and chromatin remodeling, mainly through the regulation of H3K4me3 32 . Among the KDM5 family, KDM5C is highly expressed in the central nervous system. Previous studies have shown that mutations in KDM5C are associated with X-linked mental retardation 41 . KDM5C-knockout mice showed abnormal dendritic arborization, spine anomalies, and altered transcriptomes. These mice also exhibited impaired social behavior, memory deficits, and aggression 42 . However, in some studies, silencing KDM5C, the most abundant form of KDM5 in brain, is associated with beneficial effects in neurodegenerative diseases such as Huntington's disease, which increases the level of neurotrophic factor (BDNF) and synaptic proteins (SYP) through the increased level of H3K4me3 in Huntington's disease 22 . In addition, it was found that the level of H3K4me3 in the striatum is significantly reduced in MPTP-induced mouse or nonhuman primate PD models 43 . These and our own findings are in notion with a significant role of KDM5-H3K4me3 in the pathogenesis and treatment of PD.
Chelating accumulated iron thus reducing oxidative stress has been considered a feasible treatment for PD 16,44 . Although several iron chelating agents such as DFO can effectively suppress neuronal loss in the SN and the striatum in several PD animal models 15,45 , due to their low efficiency in chelating iron, the existing iron chelating agents require a high dose when used, resulting in notable side effects. Besides, most of these current iron chelators cannot pass through the BBB, severely limiting their application in neurodegenerative diseases 17,44,46 . Compared with other iron chelators as a neuroprotective agent, GSK-J4 possesses several advantages. First, GSK-J4 exhibited a higher capacity to suppress free labile iron when compared with current iron chelators used in clinical practice, therefore minimizing side effects. Second, different from other iron chelating agents, GSK-J4 could suppress iron with cell-type specificity. Indeed, cellspecific iron chelators have not been reported as only organelle-specific chelating agents, such as mitochondrial iron chelators are available 47 . At present, GSK-J4 is the only agent that can specifically affect iron metabolism in neuronal cells. Finally, GSK-J4 has been reported to be able to pass through the BBB 48 .
In conclusion, we demonstrate a previously unappreciated mechanism of GSK-J4 in the selective iron suppression, conferring neuroprotection from oxidative stress in PD animal model. Our results not only facilitate the development of new drug targets that may eventually slow down, stop, or even reverse PD progression but also shed light on pathogenic mechanism of this disorder.

Animals
Sprague-Dawley (SD) rats weighing 320-350 g were used in this study. The animals were housed in standard cages with ad libitum access to both food and water. All animals were housed at a temperature of 25°C under a 12-h light-on-light-off schedule. After that, all animals were randomly divided into different groups. All animal procedures were performed in accordance with the protocol approved by the Chinese University of Hong Kong Animal Ethics and Experimentation Committee.

MTT assay
To study the effect of GSK-J4 on neuroprotection, MTT assays were performed to assess cell viability as previously described 49 . In brief, SH-SY5Y cells were cultured in the 96-well plates at a density of 1 × 10 5 cells per well at 37°C. Two days later, cells were placed in a fresh medium with different treatments. After a particular treatment, cell viability was assessed by the MTT assay (11465007001 ROCHE, Cell Proliferation Kit I (MTT)). First, the supernatant in each culture well was removed and the wells were washed with PBS twice. 20 μl MTT solution (5 mg/ml in PBS) was added to each well (containing 100 μl culture medium). After incubating for 4 h, the medium was removed carefully. Then, 150 μl DMSO was added to each well and shaken for 10 min to allow the crystals to be fully melted. Finally, the absorbance intensity was measured at 490 nm with a microplate reader (Bio-RAD 680, USA) and together with a reference wavelength of 620 nm. Control wells (cells with the same concentration of drug in medium) were set at the same time. The cell viability (%) was expressed as a percentage relative to the control group.
PD animal models 6-OHDA-induced PD rat model 6-OHDA PD rat model was established according to our method described previously 50,51 . Briefly, SD rats were deeply anesthetized with ketamine (75 mg/kg, i.p.) and xylazine (6 mg/kg, i.p). They then received stereotaxic injections of 6-OHDA (2 μl of 4 μg/μl) dissolved in 0.9% saline containing 0.2 mg/ml ascorbic acid to the unilateral MFB, at a site relative to the bregma: anteroposterior (AP), −4.4 mm; mediolateral (ML), +1.2 mm; and dorsoventral (DV), 8.0 mm from dura. The injection was administered within 5 min and the Hamilton syringe needle was kept in situ for 5 min. The animals were monitored for 2 h postsurgery before returning to the animal housing facility. Separate groups of control rats received saline sham injections following the same procedures.

Behavioral test Cylinder test
To evaluate motor dysfunction, a forelimb use asymmetry test (the cylinder test) was used 52 . In short, 14 days after injection of 6-OHDA, animals were placed in a quiet, welllit room at least half an hour in advance. Then, the rats were placed in a clean, open-top plexiglass cylinder (diameter: 20 cm, height: 30 cm). The behavior was recorded for 5 min for each session. Three observers who were blind to the treatment scored the numbers that the animal used the left and the right forelimbs to touch the cylindrical walls.

Apomorphine-induced circling test
Apomorphine-induced circling test was used to reflect DA denervation in rodents with unilateral nigrostriatal lesion 50,52 . In brief, 14 days after surgery, rats were injected intraperitoneally with apomorphine (M8750 Sigma, 5 mg/kg) and placed in a clean open arena (120 × 120 cm). The whole experiment was performed in a quiet, well-lit room. Thirty minutes after the injection, the rats were recorded for 10 min. The number of rotations towards the contralateral side of lesion was counted.

Immunostaining
Immunostaining was carried out as reported before with some modifications 53 . Briefly, the rats having received different treatments were deeply anesthetized with 4% chloral hydrate (0.4 mg/g, i.p.) followed by perfusion with normal saline through the left ventricle. The brains were removed and post-fixed in 4% paraformaldehyde (PFA) followed by cryoprotection with serial sucrose solutions. After that, brain tissues were immersed in cryo-embedding media (OCT) and brain sections (30 μm) that contained the SN (AP: −4.80 mm to AP: −6.12 mm from bregma) were prepared by a cryotome. Every fourth section was stained for TH and analyzed. The brain slices were then incubated in blocking solution followed by overnight incubation at 4°C with primary antibodies such as anti-TH (1:1000, Millipore). After washing with PBS, the slides were incubated with Alex 546 or FITC-conjugated secondary antibody for 1 h at room temperature. The nuclei were counterstained with DAPI. The slides with immunofluorescence staining were visualized with a Nikon C-1 confocal laser scanning microscope (Nikon). The number of TH-positive cells as well as the areas of SN (in mm 2 ) were analyzed by using ImageJ software (from National Institutes of Health and available at http://rsb.info.nih.gov/ij/).
cDNA amplification and detection were fulfilled using MyiQ Real-time PCR Detection System (Bio-Rad Laboratories, CA) and iQ SYBR Green Supermix (Bio-Rad Laboratories, CA). The initial denaturation was 95°C for 5 min, then 95°C for 10 s followed by 60°C for 45 s, which cycled for 40 times.
The β-actin cDNA (3′-primer, 5′-CTCTCAGCTG TGGTGGTGAA-3′; 5′-primer, 5′-GTCGTACCACTGGC ATTGTG-3′) was simultaneously amplified as the internal control. Relative quantification exploited the comparative Cx method. The mRNA level of KDM5C of each sample was normalized to that of the level of β-actin and control group samples were used as calibrator. Relative mRNA level was calculated by the 2 −ΔΔCT method.

DAB-enhanced Perls' iron staining
DAB-enhanced Perls' iron staining was performed as described previously 54 . Briefly, the slides were fixed with 4% PFA for 10 min. After three washes with PBS, the slides were incubated in Perls' solution containing 8% potassium ferrocyanide and an equal volume of 1.2 mmol/l hydrochloride acid solution for 16 h at 4°C and then washed with deionized water three times, followed by dehydration through 95% alcohol and mounting with xylene. Afterwards, the sections were washed and incubated in a solution of DAB to enhance the signals. Iron staining was observed under a Nikon Eclipse TE2000-U microscope (Nikon, UK).

Calcein-AM assay
The ferrous iron measurement on cell was performed using calcein-AM method as previously described, with some modifications 55 . Briefly, SH-SY5Y cells were grown in DMEM supplemented with 10% heat-inactivated fetal bovine serum, and cultured at 37°C under humidified 5% CO 2 atmosphere. Cells were treated with ferric ammonium citrate (FAC) or GSK-J4 or GSK-J4 plus FAC for 12 h in serum-free media. Then they were washed with PBS twice, followed by the addition of calcein-AM (50 nM), and incubated for 30 min at 37°C. After that, fluorescence was measured using a microplate fluorometer (Thermo Fisher Scientific Inc).

ROS, protein carbonylation determination
ROS were measured by the cell-permeant 2′,7′dichlorodihydrofluorescein diacetate (H 2 DCFDA) assay as previously described with minor modifications 56 . Briefly, brain, cell, or protein extraction were incubated with 5 μM H 2 DCFDA in dark at 37°C for 30 min and then washed three times with PBS. DCF fluorescence was observed under Nikon C-1 confocal laser scanning microscope and quantified using a fluorescence microplate reader (LS55 fluorescence Spectrometer, Perki-nElmer) with an excitation/emission wavelength of 485/ 515 nm in 96-well fluorescent plates. Protein carbonyls were measured by using an OxyBlotTM protein oxidation detection kit (Chemicon), in which carbonyl groups in the protein side chains are derivatized to 2,4-dinitrophenylhydrazone (DNP) by reaction with 2,4-dinitrophenylhydrazine (DNPH).

MDA level
Lipid peroxidation was determined in total brain lysates using the N-methyl-2-phenylindole based LPO-586™ lipid peroxidation kit (Oxis International, OR, USA) 57 . This assay measures the level of MDA. Standard curves of MDA were established using 1,1,3,3 tetramethocypropane. To minimize non-specific oxidation during sample preparation, 5 mm butylated hydroxytoluene (BHT dissolved in acetonitrile) was added to the extraction buffer. The assay was performed in triplicates according to manufacturer's recommendation using 4 mg of total brain lysates per reaction, and the results were calculated as picomoles of MDA per mg of protein. Briefly, 200 µl of brain lysates (4 mg) was added to 10 µl of 0.5 M BHT and 650 µl of 0.1 mM N-methyl-2-phenylindole, followed by gentle mixing. For MDA measurement, 150 µl of reagent grade HCl (∼36%) was added. The sample was mixed and incubated for 1 h at 45°C. After incubation, the sample was centrifuged, and the supernatant was extracted and measured at 586 nm.

GSH level
Total reduced GSH was measured as described previously 57 . Briefly, the acid-soluble fraction was obtained from cell lysates and tissue homogenates by adding perchloric acid to a final concentration of 3%, followed by centrifugation at 14,000 × g for 10 min. The acid-soluble fraction was neutralized to pH 7 with 0.5 M KOH/50 mM Tris. After the removal of precipitate (potassium perchlorate) by a second centrifugation, 50 μl aliquots of sample was combined with 100 μl of reaction mixture consisting of 2.5 ml of 1 mM 5,5′-dithio-bis-(2-nitrobenzoic acid) (DTNB), 2.5 ml of 5 mM NADPH, 2.5 ml of phosphate buffer solution (100 mM NaPO4, pH 7.5, 1 mM EDTA), and glutathione reductase (5 U/ml final volume). GSH-mediated reduction of DTNB was measured at 412 nm at 30 s intervals over 30 min. GSH content was normalized to protein.

Construction of AAV virus
The coding silence sequence of KDM5C will be first ligated into a shuttle plasmid pAAV with EGFP-tag (Addgene). Insertion of the coding sequence was then verified by sequencing. The positive recombinant plasmid was co-transfected with pAAV-RC, pAAV-helper (Addgene) into HEK293 cells. AAV was harvested 3 days later and designated as "AAV-KDM5C". The titers of the virus were measured according to standard procedure.

Statistical analysis
All data were presented as mean ± standard error of the mean (SEM). Graphpad Prism was used for statistical analysis. Unpaired Student's t test or one-way ANOVA with a post hoc Tukey's test was performed, as appropriate, to determine significant differences between groups. A probability value of P < 0.05 was taken to be statistically significant. All the experiments consisted of at least three replicates.