Meta-Analysis of Parkinson’s Disease and Alzheimer’s Disease Revealed Commonly Impaired Pathways and Dysregulation of NRF2-Dependent Genes

Abstract

Many lines of evidence suggest that Parkinson’s disease (PD) and Alzheimer’s disease (AD) have common characteristics, such as mitochondrial dysfunction and oxidative stress. As the underlying molecular mechanisms are unclear, we perform a meta-analysis with 9 microarray datasets of PD studies and 7 of AD studies to explore it. Functional enrichment analysis revealed that PD and AD both showed dysfunction in the synaptic vesicle cycle, GABAergic synapses, phagosomes, oxidative phosphorylation, and TCA cycle pathways, and AD had more enriched genes. Comparing the differentially expressed genes between AD and PD, we identified 54 common genes shared by more than six tissues. Among them, 31 downregulated genes contained the antioxidant response element (ARE) consensus sequence bound by NRF2. NRF2 is a transcription factor, which protects cells against oxidative stress through coordinated upregulation of ARE-driven genes. To our surprise, although NRF2 was upregulated, its target genes were all downregulated. Further exploration found that MAFF was upregulated in all tissues and significantly negatively correlated with the 31 NRF2-dependent genes in diseased conditions. Previous studies have demonstrated over-expressed small MAFs can form homodimers and act as transcriptional repressors. Therefore, MAFF might play an important role in dysfunction of NRF2 regulatory network in PD and AD.

Keywords

Parkinson’s disease Alzheimer’s disease meta-analysis MAFF NRF2

INTRODUCTION

Neurodegenerative diseases are one of the most common diseases of the elderly and seriously affect quality of life. Two typical diseases are Parkinson’s disease (PD) and Alzheimer’s disease (AD) [1]. It is widely accepted that PD and AD are two different types of diseases. The clinical manifestations of PD are bradykinesia, resting tremor, and postural instability [2], while AD patients predominantly show progressive cognitive dysfunction and memory impairment. According to current research, the major pathological changes of PD are dopamine neuron degeneration and loss [3] in the substantia nigrapars compacta in the midbrain, leading to decreased nigrostriatal dopamine content [4]. Residual dopaminergic neurons form Lewy bodies [5], the major component of which is alpha-synuclein. In contrast, the typical pathological hallmarks of AD are (1) extensive neuronal loss leading to brain tissue atrophy; (2) neurofibrillary tangles in the neocortex and hippocampus [6]; and (3) extracellular amyloid-β (Aβ) deposits, which form dispersed senile plaques and amyloid plaques [7]. However, an increasing number of studies have found that PD and AD have several similarities [8]. For example, (1) PD and AD are both age-related diseases; (2) PD and AD both display 5% –10% autosomal dominant inheritance [8]; (3) PD and AD both exhibit protein aggregation and inclusion bodies, which are formed by abnormal phosphorylation of the skeleton [9]; (4) PD and AD both display mitochondrial dysfunction and oxidative stress [10 –12]; and (5) some PD patients also have dementia [13]. However, the pathogenic mechanisms of PD and AD have not been completely elucidated. What are the molecular mechanisms underlying these common features?

Therefore, it is necessary to carry out a thorough and accurate exploration of the molecular mechanisms of the common features of PD and AD. We searched the relevant gene expression databases and collected 9 microarray datasets from PD studies and 7 from AD studies to perform meta-analyses. As a result, we confirmed several common characteristics between PD and AD, including synaptic vesicle cycle impairment, oxidative phosphorylation damage, TCA cycle reduction, and GABAergic synapse dysfunction. Meanwhile, we also found that they have their own pathogenic characteristics. PD was obviously impaired in dopamine synthesis and transport, and AD showed dysfunction in cognition and learning or memory. Importantly, we obtained 31 overlapping genes in most tissues of PD and AD, which all contained an antioxidant response element (ARE) consensus sequence bound by NRF2. NRF2 was upregulated, while the 31 genes were all downregulated, which contrasted with the results of previous studies. NRF2 plays an important role in cell resistance to oxidative and electrophilic stress by binding to the ARE and upregulating ARE-driven genes [14, 15]. Further exploration found that MAFF participated and influenced the gene network regulated by NRF2.

MATERIALS AND METHODS

Data collecting

We used “Parkinson” and “Alzheimer” as keywords and searched all genome-wide expression datasets such as NCBI-GEO (http://www.ncbi.nlm.nih.gov/geo/) and EMBL-EBI ArrayExpress database (https://www.ebi.ac.uk/arrayexpress/) andliterature databases such as PubMed (http://www.pubmed.gov/). The inclusion criteria were as follows: (1) all datasets were genome-wide; (2) the series type was expression profiling by array; (3) raw data cell-intensity file (CEL) files were available, with control and disease samples of human brain; and (4) for each study, total number of available samples were≥10. With the inclusion criteria, we screened out 11 datasets from PD studies and 8 datasets from AD studies. Among them, we performed meta-analysis of 10 datasets from PD studies (GSE7621, GSE8397, GSE20291, GSE20292, GSE20163, GSE20164, GSE20146, GSE20141, GSE20168, and GSE24378) and 7 datasets from AD (GSE1297, GSE28146, E-MEXP-2280, GSE39420, GSE4757, GSE5281, and GSE48350) and the rest datasets (GSE54282 of PD and GSE36980 of AD) as validation. For the details of the datasets, see Supplementary Table 1.

Data preprocessing

All datasets were processed using the robust multichip average algorithm [16] to generate gene expression values. Gene annotation and integration of AD and PD datasets were carried out using custom written Python code [17]. First, we removed probes that with no gene annotation or that matched multiple gene symbols. Next, we calculated the average expression value of multiple probe IDs that matched to an official gene symbol, and last took this value to represent the expression intensity of the corresponded gene symbol.

Processed gene expression data matrixes of all studies were classified according to the different brain regions of different diseases, and genes not shared by the same region were removed. Then, MetaQC [18] was used to assess the quality and consistency of the studies shared by the same region and excluded studies with poor quality. The results of MetaQC are shown in Supplementary Table 3. Last, 16 studies [substantia nigra (SN), frontal cortex (FC), putament (PT), and globus pallidus interna (GPI) of PD and posterior cingulate (PC), superior frontal gyrus (SFG), entorhinal cortex (EC), FC, hippocampus (HIP), medial temporal lobe (MTL), and primary visual cortex (PVC) of AD] were left and differentially expressed genes (DEGs) between disease samples and controls in every brain region were compared in different diseases using the RankProd (RP) method [19].

RP is a non-parametric statistical method that utilizes the rank of genes in each array instead of the actual expression value, which detects different genes that are consistently highly ranked in a number of studies, under two experimental conditions. Because the method utilizes the rank of genes, it can be flexibly applied to integrate multiple microarray studies from different laboratories or platforms [19]. RP has higher sensitivity and selectivity than the t-based method (random effects model) in both individual and meta-analysis, especially in the setting of small sample size and/or large between-study variation. The RP package was developed from the rank product method which was initially proposed to detect differentially expressed genes in a single experiment. Therefore, RP can perform meta-analysis of different studies as well as the analysis of a single study [20].

During the RP test, 1,000 random iterations were performed to determine the false discovery rate (FDR) for each gene. We utilized the “topgene” function to identify different genes based on a FDR correct p value (pfp)≤0.01 and the absolute value of fold change FC≥1.5.

KEGG pathways enriched by differentially expressed genes

ClueGO [21] is a plug-in of Cytoscape, which can interpret large lists of genes with a functionally organized Gene Ontology (GO) or pathway term network by integrating GO terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways [22]. In this study, we performed GO Biological Process and KEGG pathway enrichment analyses for the screened DEGs in four brain tissues of PD (FC, SN, PT, GPI) and in six brain tissues of AD (PC, MTL, PVC, SFG, EC, HIP), respectively. A process or a pathway was considered significantly enriched if it met the FDR corrected p value < 0.05.

Screen out common DEGs containing the ARE consensus sequence bound by NRF2

In PD and AD, DEGs of different tissues were intersected to obtain common DEGs shared by most tissues. To eliminate age-related DEGs, control samples containing age information of each study were collected with the inclusion criteria that the age of samples in each study should contain < 60 years and≥60 years respectively. Finally, the control samples from 5 studies including 6 tissues (SN, PT, FC of PD and SFG, HIP, EC of AD) were screened out. Then samples of different brain regions divided into two categories: the individuals’ age≥60 years as the elderly population, and the individuals’ age < 60 years as the non-elderly population; next, age-related DEGs of each brain region were analyzed and obtained with the RP algorithm; and last, age-related DEGs of different brain regions were took intersection with the obtained common DEGs. The mean age of case samples and control samples in AD and PD are shown in Supplementary Table 2.

Three (VSNL1, SNAP25, and SYT1) of the common DEGs have been proven to be positively regulated by NRF2 [23]. To identify whether the rest of the common DEGs shared by most tissues of PD and AD were also NRF2-dependent genes, first, we used the ARE consensus sequence ‘RTKAYnnnGCR’ [24] (where R = A or G; K = T or G; W = A or T; Y = A or C or G) to search the common DEGs upstream promoter elements and obtained the genes which included the motif region, then these genes were used to match with the ‘TGACWnRGCR’ ARE consensus sequence bound by NRF2 and finally genes containing the ARE consensus sequence bound by NRF2 were screened out.

To rule out underlying factor of the repressor protein KEAP1 and explore the correlation between MAFF and the common genes containing the ARE consensus sequence bound by NRF2, a univariate linear regression model was used to analysis the associations between NRF2, MAFF and KEAP1 and these genes. Significantly positive or negative correlation between the gene (NRF2, MAFF, or KEAP1) and the NRF2-dependent genes were considered as a correlation coefficient≥0.5 or≤–0.5.

Validation of the correlation between MAFF and NRF2-dependent genes

Validation of the correlation between MAFF and 31 NRF2-dependent genes was conducted in two studies (GSE54282 of PD and GSE36980 of AD). The study of GSE54282 contained three brain regions (cortex, striatum, and substantia nigra) and 28 samples and GSE36980 contained three brain regions (frontal cortex, hippocampus, and temporal cortex) and 79 samples. The correlation between MAFF and 31 NRF2-dependent genes in each brain region of PD and AD was conducted with the univariate linear regression model and results were shown in Supplementary Table 5.

RESULTS

Differentially expressed genes of PD and AD

To identify DEGs associated with PD, we used 9 datasets consisting of four different brain regions (FC, SN, PT, GPI) to perform meta-analyses. We obtained the DEGs from each tissue as shown in Table 1. Overall, the number of downregulated genes was greater than the number of upregulated genes (Table 1). Among them, tissues of the frontal cortex had the largest number of DEGs, including70 upregulated and 301 downregulated genes. Comparing the DEGs in every tissue (FC, SN, PT, GPI), we found there were three upregulated genes (HSPB1, MAFF, HSPA1A) shared by all tissues, and 33 downregulated genes were found in three tissues (FC, SN, PT) in PD.

AD is a neurodegenerative disease that involves damage to multiple tissues. Previous studies have conducted a comprehensive survey of brain and blood transcriptome in AD and got many valuable discoveries [25, 26]. However, the pathogenic mechanisms of AD have not been completely elucidated. To further elucidate the pathogenesis of AD, we performed a meta-analysis of 7 studies investigating 6 brain regions (MTL, PC, PVC, SFG, EC, HIP) using the RP method [19]. We screened the DEGs in six tissues as shown in Table 1. The number of downregulated genes was greater than that of the upregulated genes. Two tissues (MTL, PC) had the maximum number of DEGs (Table 1). Comparing the DEGs of the six tissues, we found that the number of common genes was 37. Among these genes, only one of them was upregulated, MAFF, and the remaining 36 were downregulated.

KEGG pathway enrichment analyses of differentially expressed genes in PD and AD

Functional enriched KEGG pathway annotation demonstrated that DEGs in PD were predominantly enriched in three pathways, which are shown in Fig. 1A: (1) the pathways involved in synaptic vesicle processes, for example, three tissues (FC, PT, SN) all showed dysfunction in the synaptic vesicle cycle [27], GABAergic synapse [28], and dopaminergic synapse [29], and the related genes were all downregulated; (2) the pathways related to cell energy metabolism, for example, two tissues (FC, SN) were both impaired in the citrate cycle (TCA cycle) and oxidative phosphorylation pathways [30], which was consistent with previous reports. (3) Some of the other new pathways related to PD, for example, PC, GPI, and PT, were all enriched in mineral absorption, and the related DEGs were mostly upregulated.The downregulated DEGs of FC and PT were both enriched in endocrine and other factor-regulated calcium reabsorption (EAOFRCR), and the downregulated DEGs of FC and SN were related to the phagosome. The SN DEGs were particularly enriched in Fc gamma R-mediated phagocytosis, the details were shown in Supplementary Table 4.

The results of AD DEG enrichment of KEGG pathways were divided into two categories. In the first group, KEGG pathways were consistent with previous studies. For example, the oxidative phosphorylation pathway [31, 32], synaptic vesicle cycle [33], and GABAergic synapse [34], were the most significantly enriched pathways of all tissues in AD; proteasome [35], long term potentiation [36], citrate cycle (TCA cycle), and pyruvate metabolism [37] were enriched by more than two organizations, which are shown Fig. 1B; TNF signaling pathway [38], cholinergic synapse [39], and glutamatergic synapse [36], were the significant pathways that enriched by up and downregulated DEGs of PC, which were shown in Supplementary Table 4. In the second group, new KEGG pathways associated with AD, such as phagosome, mineral absorption, glycolysis/gluconeogenesis, and endocrine and other factor-regulated calcium reabsorption, were enriched by more than two organizations, which are shown Fig. 1B; antigen processing and presentation pathway, axon guidance, calcium signaling pathway, and cAMP signaling pathway, were significantly enriched by up- and downregulated DEGs of PC; the up- and downregulated DEGs of MTL were significantly enriched in TGF-beta signaling pathway, hippo signaling pathway, signaling pathways regulating pluripotency of stem cells, and thyroid hormone signaling pathway; the significantly enriched KEGG pathways of downregulated DEGs of EC were closely associated with pentose phosphate pathway and neuroactive ligand-receptor interaction, which is shown in Supplementary Table 4.

Comparing the DEG enriched KEGG pathways of PD and AD, we identified several pathways shared by PD and AD, such as synaptic vesicle cycle, GABAergic synapse, oxidative phosphorylation, citrate cycle (TCA cycle), endocrine and other factor-regulated calcium reabsorption (EAOFRCR), mineral absorption, and phagosome pathways. Among them endocrine and other factor-regulated calcium reabsorption (EAOFRCR), mineral absorption, and phagosome pathways were some new pathways which have not been reported related to PD and AD. In terms of the gene assignment to each functional pathway, AD-enriched genes were significantly greater than those of PD. DEGs of PD and AD were also enriched in different KEGG pathways. For example, dopaminergic synapse pathway, Fc gamma R-mediated phagocytosis and protein processing in endoplasmic reticulum pathways were related to PD, while DEGs of AD were enriched in long time potentiation, glycolysis/gluconeogenesis, glutamatergic synapse, proteasome, pyruvate metabolism, axon guidance, TGF-beta signaling pathway, hippo signaling pathway, cholinergic synapse, and TNF signaling pathways and so on.

Common DEGs shared by most tissues containing the ARE consensus sequence bound by NRF2

Many lines of evidence have shown that mitochondrial dysfunction and oxidative stress are the greatest risk factors for neurodegenerative diseases [10 , 40], including AD, PD, Huntington’s disease, and amyotrophic lateral sclerosis. ARE has been identified as an enhancer element in the promoter region of genes encoding antioxidant and/or detoxication enzymes, such as NQO1, HMOX1, and GST [12], to respond to oxidative stress [41]. NRF2 is a cap’n’collar (CNC) basic-region leucine zipper (bZIP) transcription factor and a major regulator of the cellular response to oxidative stress through coordinated upregulation of ARE-driven genes, and it controls adaptive responses to various environmental stressors [42]. Activation of NRF2 inhibits inflammation and protects against oxidative stress in degenerative disease.

Comparing the DEGs in all tissues (FC, SN, PT, and GPI of PD; PVC, PC, MTL, HIP, SFG, and EC of AD), we screened out 81 identical genes shared by more than six tissues, including 28 upregulated DEGs and 53 downregulated DEGs. Among them 40 out of 53 downregulated genes contained ARE consensus sequence bound by NRF2. To eliminate the age-related DEGs, we screened out 6 tissues (SN, PT, and FC of PD and SFG, HIP, and EC of AD) containing the age information and reanalyzed. 116 upregulated and 36 downregulated age-related DEGs were shared by more than three tissues of PD and AD. 13 out of 28 upregulated and 41 out of 53 downregulated DEGs were obtained without age factors by the intersection outcome, of which 31 genes containing the ARE consensus sequence bound by NRF2. Among the 31 genes, VSNL1, SNAP25, and SYT1 were previously shown to be positively regulated by NRF2 using Nrf2^+/+ and Nrf2^-/- primary neuronal culture experiments [23], the results are shown in Table 2. Except for three genes, the other genes positively regulated by NRF2 identified with the Nrf2^+/+ and Nrf2^-/- primary neuronal culture experiments were also found to be significantly downregulated DEGs including in at least three tissues of PD and AD (Supplementary Figure 1), such as detoxification genes (HMOX1, NQO1, GSTA4, GSS, SOD1), calcium homeostasis genes (VSNL1, CALB1, SYT1, HPCA, SYT5, NUCB2, CALM3), signaling genes (GABBR1, GABRA2, GABRA1, SLC1A1, GNG3, ADCYAP1, KCNA1, PRKCB, ARHGEF3, MAPK10, PTPRO, MATK, SLC3A1), growth factor genes (FGF13, BDNF), and neuron-specific genes (CHGB, SCG2, SYP, SYN2, SNAP25, DCTN3, RPER). Therefore, the remaining 28 genes were new NRF2-dependent ARE-driven genes that need to be further verified experimentally. To our surprise, NRF2 was upregulated in all tissues of AD and PD but the 31 NRF2-dependent ARE-driven DEGs were downregulated (Fig. 2), which contrasted to previous studies showing that NRF2 protected cells through coordinated upregulation of ARE-driven genes. Therefore, NRF2 may be part of a complex regulatory network that was affected by other transcription factors. It was necessary to further explore the regulatory mechanism of NRF2.

MAFF was negatively correlated with the 31 NRF2-dependent genes in PD and AD

NRF2 is a member of the CNC bZIP transcription factors [42]. Under normal conditions, NRF2 protein is maintained at low levels due to ubiquitination and degradation by the repressor protein KEAP1 in the cytoplasm. When cells are under oxidative stress conditions, NRF2 is released from KEAP1 and enters the nucleus. NRF2 lacks the capacity to bind DNA on its own, requiring small MAFs (MAFF, MAFG, MAFK) [15, 43] as its heterodimer partners to bind to ARE elements, which initiates the transcription of its target genes. Previous studies have shown that because small MAFs possess a bZIP motif that lacks a transactivation domain [44], they can form homodimers among themselves and act as transcriptional repressors if overexpressed [45]. For example, positive or negative MARE-dependent transcriptional regulation is dependent on the abundance of small MAF proteins [46, 47], and overexpression of MAFK represses ARE-driven gene transcription [41].

MAFF was upregulated in all tissues in both PD and AD, and the expression of MAFF was almost twice that of NRF2. MAFK was only upregulated in PC and PVC of AD, and there was no difference of MAFG in every tissue of PD and AD. To rule out underlying factor of the repressor protein KEAP1 and explore the correlation between MAFF and the 31 genes containing the ARE consensus sequence bound by NRF2, a univariate linear regression model was used to analyze it. Results showed that the mean of correlation coefficient between MAFF and 31 NRF2-dependent genes was –0.5 in most tissues of PD and AD, and 0 in the control group. While the correlation between NRF2 and KEAP1 and the 31 NRF2-dependent genes varied in different tissues both of PD and AD and the control group. For example, in six tissues (FC, PT, and SN of PD; EC, MTL, PVC and SFG of AD), the correlation coefficient between NRF2 and the 31 NRF2-dependent genes was equal to or near to 0, and in the HIP and PC of AD, the ratio was≤–0.5; as for KEAP1, in five tissues (PT of PD; EC, MTL, PVC and SFG of AD), the correlation coefficient was equal to or near to 0, in the SN of PD and HIP and PC of AD, the ration was almost –0.5; and in the FC of PD, it was≥0.5. Details are shown in Fig. 3. The results showed that MAFF was significantly negatively correlated with the 31 NRF2-dependent genes, which was also verified in other data sets (Supplementary Table 5). There was low correlation between NRF2 and KEAP1 and the 31 NRF2-dependent genes in most tissues of PD and AD. In addition, there was a strong correlation between protein abundance and median RPKM of NRF2 and MAFF in normal tissues, which is shown in Supplementary Figure 2. Based on the above conditions, we put forward a hypothesis of the NRF2 regulatory network in PD and AD, which is shown in Fig. 4. Overexpression of MAFF resulted in the formation of MAFF homodimers, which acted as repressors of transcription by competing with MAFF-NRF2 heterodimer binding at the ARE site, thus inhibiting the expression of NRF2 upregulated genes. As a result, NRF2-dependent genes encoding antioxidant and detoxication enzymes were all downregulated, and subsequently, the nerve cells were much more sensitive to oxidative stress and displayed reduced detoxification.

DISCUSSION

In this study, we performed meta-analyses of gene expression data of two neurodegenerative diseases, PD and AD [1]. Result showed that DEGs varied in different brain regions both of PD and AD and the FC brain region of PD and MTL and PC regions of AD had the largest number of DEGs. On the whole, downregulated DEGs were far greater than the upregulated DEGs. Then, the DEGs of PD and AD were used to detect pathway enrichment for gene functional annotation. Using KEGG pathway analysis [22], we discovered that PD and AD were both dysfunctional in synaptic vesicle and mitochondrial oxidative metabolism pathways. But the enriched genes among AD were greater than those of PD, suggesting that although PD and AD have some common characteristics [8, 9], AD patient brain damage was more severe than that of PD. As is known to us all, cognition, learning and memory impairment are the major pathological features of AD [48, 49]. In our study, eight brain regions (PC, MTL, PVC, SFG, EC, and HIP of AD; SN and FC of PD) had downregulated DEGs that were significantly enriched in cognition (GO:0050890), and seven brain regions (PC, MTL, PVC, SFG, EC, and HIP of AD; SN and FC of PD) had downregulated DEGs that were significantly enriched in learning or memory (GO:0007611). Overall, almost all AD tissues were impaired in cognition and learning or memory, and the enriched genes of AD were significantly greater than those of PD. It indicated that cognition and learning or memory damage in AD was more severe than that in PD. DEGs of PD and AD were also enriched a number of other different pathways. For example, in PD, the significantly enriched GO terms of downregulated DEGs of SN were closely associated with dopamine synthesis and metabolism. Among these enriched genes, the rate limiting enzyme synthesis of catalytic dopamine (TH) [50] and solute carrier family genes (SLC6A3, SLC18A2) [51, 52] associated with dopamine transport and reuptake were all downregulated, consistent with reports from the literature. As for AD, DEGs of most tissues were significantly enriched in the proteasome pathway but only a few genes in PD were related to the proteasome pathway. Last, 31 overlapping downregulated genes in most tissues of PD and AD were identified containing an ARE consensus sequence bound by NRF2. And there were four miRNAs (hsa-miR-335-5p, hsa-miR-215-5p, hsa-miR-26b-5p, and hsa-miR-1301-3p) enriched by the 31 DEGs, which is shown in Supplementary Table 6.

By far the greatest risk factor for neurodegenerative diseases such as AD and PD is aging. Recent studies revealed that after the age of 40 there was a decrease in the expression of genes involved in synaptic plasticity, vesicular transport and mitochondrial function, followed by increased expression of stress-response, antioxidant, and DNA-repair genes [53]. In our study, the age-related DEGs of PD and AD enriched in the synaptic vesicle cycle and oxidative phosphorylation pathways, which was consistent with these studies. Age is also the best predictor of PD progression rate [4] and a cause of cognitive decline in the elderly, which remains the most prominent risk factor for AD [53].

Both AD and PD are characterized by abnormal protein deposition in various areas of the brain. Alpha-synuclein is the precursor protein of the non-Aβ component of AD amyloid (NAC) and is also the major protein component in the hallmark lesions of PD and Lewy bodies [54]. In our study, SNCA encoding alpha-synuclein was significantly downregulated in five tissues (FC and SN of PD; MTL, PVC, and PC of AD). Alpha-synuclein encoded by SNCA might play a role in the regulation of synaptic plasticity such as songbird learning, neuronal differentiation, upregulation of dopamine release, and mitochondrial deficits [55]. Defects in SNCA have been both implicated in the pathogenesis of PD and AD [55]. SNCB encoding β-synuclein was significantly downregulated in three tissues (FC of AD; MTL and PC of AD). Studies have proved that β-synuclein may be a natural negative regulator of α-synuclein aggregation and that a similar class of endogenous factors might regulate the aggregation state of other molecules involved in neurodegenerative diseases [56]. α- and β-synuclein are members of the synuclein family, which are abundant in the brain and can inhibit phospholipase D2 selectively. Therefore, it is possible that synucleins are important regulatory components of the vesicular transport processes [54]. In addition, amyloid-β precursor binding-proteins related genes (APBB3 and APBA2) were downregulated in MTL and PC of AD respectively. APBB3 encodes FE65L2, which is a member of the FE65s family. The FE65s (FE65, FE65L1, and FE65L2) are a family of multidomain adaptor proteins containing a WW domain and two PTB domains. Through the PTB2 domain, they can interact with the AD amyloid-β protein precursor (AβPP) intracellular domain (AICD) and can alter AβPP processing [57]. APBA2 encodes X11β which is a member of the X11 protein family. It is a neuronal adapter protein that interacts with AβPP. Overexpression of neuronal adaptor protein X11β has been shown to decrease the production of Aβ, a toxic peptide deposited in ADbrains [58].

Many genome-wide association studies have identified many genetic risk associated with PD and AD. For example, large-scale meta-analysis of genome-wide association data identifies six new risk loci (SIPA1L2, INPP5F, MIR4697, GCH1, VPS13C, and DDRGK1) for PD [59] and meta-analysis of 74,046 individuals identifies 11 new susceptibility loci (INPP5D, HLA-DRB5, MEF2C, NME8, ZCWPW1, PTK2B, CELF1, SORL1, FERMT2, SLC24A4, CASS4) for AD [60]. In order to identify whether the mutations of MAFF, NRF2, and 31 screened DEGs are correlated to AD and PD, we collected totally 220,953 SNPs across these genes in SNP-NCBI (https://www.ncbi.nlm.nih.gov/snp) and then mapped to the risk SNPs associated with PD and AD that were confirmed by previous studies in openSNP web server (https://opensnp.org/). Finally, we found there were 9 SNPs (rs10183914, rs1806649, rs2001350, rs2364722, rs2886161, rs35652124, rs6706649, rs6721961, rs6726395) of NRF2 and 1 SNP (rs4130047) of RIT2 that correlated with PD.

It is generally believed that oxidative stress [61] and mitochondrial dysfunction are likely to be the common mechanisms in PD and AD. In our study, some of the DEGs such as the downregulated detoxification genes (SOD1 [62], GSS), the upregulated heat shock protein genes [40] (HSPB1, HSPA1A) and metallothionein genes (MT1M, MT1X, MT1E, MT1H, MT2A) [11, 63] and downregulated DEG-enriched KEGG pathways (oxidative phosphorylation and citric acid cycle) [64], have demonstrated that the nerve cells of PD and AD patients are under oxidative stress. NRF2 is a CNC bZIP transcription factor [42] and plays an important role in cell resistance to oxidative and electrophilic stress by binding to the ARE and upregulating ARE-driven genes, especially in neurodegenerative diseases [14, 15]. In our study, NRF2 was upregulated, while most of its target genes were downregulated, such as detoxification genes (HMOX1, NQO1, GSTA4, GSS, SOD1), calcium homeostasis genes (VSNL1, CALB1, SYT1, HPCA, SYT5, NUCB2, CALM3), signaling genes (GABBR1, GABRA2, GABRA1, SLC1A1, GNG3, ADCYAP1, KCNA1, PRKCB, ARHGEF3, MAPK10, PTPRO, MATK, SLC3A1), growth factor genes (FGF13, BDNF), and neuron-specific genes (CHGB, SCG2, SYP, SYN2, SNAP25, DCTN3, RPER) [23], causing increased levels of oxidative stress and disturbances in brain function. Further analysis found that the expression of MAFF was almost twice that of NRF2. MAFK was only upregulated in PC and PVC of AD, and there was no difference in MAFG in every tissue of PD and AD. In addition, MAFF was significantly negatively correlated with the 31 NRF2-dependent genes in most tissues of PD and AD. As small MAFs possess a bZIP motif that lacks a transactivation domain [44], they can form homodimers among themselves and act as transcriptional repressors if overexpressed [45]. Therefore, we concluded that overexpression of MAFF represses NRF2-dependent gene transcription in PD and AD, leading to NRF2 target gene downregulation. As NRF2 target genes participate in diverse processes, such as immune and inflammatory responses, metabolism, tissue remodeling and fibrosis, metastasis, cognitive dysfunction, addictive behavior and calcium homeostasis and signaling [14], dysfunction of NRF2-regulated genes could cause various cellular function disorders. These provide a plausible explanation for many common characteristics between PD and AD.

Many studies have identified NRF2 as a potential therapeutic target for neurodegenerative disease [15, 65]due to two reasons: one, NRF2 acts as a transcription factor regulator of ARE-dependent expression of a battery of cytoprotective and antioxidant enzymes and proteins and has a protective effect on nerve cells [12 , 65]; two, L-sulforaphane derived from broccoli and licorice can induce the activation of NRF2, inhibiting the KEAP1/ NRF2 interaction and proteasomal degradation of NRF2 and increasing NRF2 target gene expression. In this study, NRF2 was activated while its target genes were significantly downregulated as a result of overexpressed MAFF repressing its transcriptional function. Thus, the quantitative balance of NRF2/MAFF interacting with the ARE determines NRF2 transcriptional activity and further determines the ability of the cells to respond to oxidative stress. Therefore, excess MAFF-mediated repression of NRF2-dependent gene transcription may be a new common mechanism of PD and AD, suggesting that MAFF could be a promising therapeutic target for PD and AD.

Footnotes

ACKNOWLEDGMENTS

This work was supported by the National Basic Research Program of China (Grant No. 2013CB835100), and the National Natural Science Foundation of China (Grant No. 31401142 and NO. 31401137).

Authors’ disclosures available online ().

Appendix

References

Nussbaum

, Ellis

(2003) Alzheimer’s disease and Parkinson’s disease. N Engl J Med 348, 1356–1364.

Jankovic

(2008) Parkinson’s disease: Clinical features and diagnosis. J Neurol Neurosurg Psychiatry 79, 368–376.

German

, Manaye

, Smith

, Woodward

, Saper

(1989) Midbrain dopaminergic cell loss in Parkinson’s disease: Computer visualization. Ann Neurol 26, 507–514.

Obeso

, Rodriguez-Oroz

, Goetz

, Marin

, Kordower

, Rodriguez

, Hirsch

, Farrer

, Schapira

, Halliday

(2010) Missing pieces in the Parkinson’s disease puzzle. Nat Med 16, 653–661.

Wakabayashi

, Tanji

, Odagiri

, Miki

, Mori

, Takahashi

(2013) The Lewy body in Parkinson’s disease and related neurodegenerative disorders. Mol Neurobiol 47, 495–508.

Dickson

(1997) Neuropathological diagnosis of Alzheimer’s disease: A perspective from longitudinal clinicopathological studies. Neurobiol Aging 18(4 Suppl), S21–26.

Hardy

, Selkoe

(2002) The amyloid hypothesis of Alzheimer’s disease: Progress and problems on the road to therapeutics. Science 297, 353–356.

Boller

, Mizutani

, Roessmann

, Gambetti

(1980) Parkinson disease, dementia, and Alzheimer disease: Clinicopathological correlations. Ann Neurol 7, 329–335.

Ross

, Poirier

(2004) Protein aggregation and neurodegenerative disease. Nat Med 10(Suppl), S10–17.

10.

Lin

, Beal

(2006) Mitochondrial dysfunction and oxidative stress in neurodegenerative diseases. Nature 443, 787–795.

11.

Khan

, Hassan

, Ahmad

, Perveen

, Aman

, Quddusi

, Alhazza

, Ashraf

, Aliev

(2016) Recent updates on the dynamic association between oxidative stress and neurodegenerative disorders. CNS Neurol Disord Drug Targets 15, 310–320.

12.

Denzer

, Munch

, Friedland

(2016) Modulation of mitochondrial dysfunction in neurodegenerative diseases via activation of nuclear factor erythroid-2-related factor 2 by food-derived compounds. Pharmacol Res 103, 80–94.

13.

Greenfield

, Vaux

(2002) Parkinson’s disease, Alzheimer’s disease and motor neurone disease: Identifying a common mechanism. Neuroscience 113, 485–492.

14.

Wang

, Campbell Michelle

, Lacher Sarah

, Cho

H-Y

, Wan

, Crowl Christopher

, Chorley Brian

, Bond Gareth

, Kleeberger Steven

, Slattery

, Bell Douglas

(2016) A polymorphic antioxidant response element links NRF2/sMAF binding to enhanced MAPT expression and reduced risk of Parkinsonian disorders. Cell Reports 15, 830–842.

15.

Buendia

, Michalska

, Navarro

, Gameiro

, Egea

, Leon

(2016) Nrf2-ARE pathway: An emerging target against oxidative stress and neuroinflammation in neurodegenerative diseases. Pharmacol Ther 157, 84–104.

16.

Irizarry

, Hobbs

, Collin

, Beazer-Barclay

, Antonellis

, Scherf

, Speed

(2003) Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 4, 249–264.

17.

, Dai

, Wang

, Guo

, Hong

, Zheng

, Liu

, Li

, Huang

(2016) Integrated analysis of ischemic stroke datasets revealed sex and age difference in anti-stroke targets. PeerJ 4, e2470.

18.

Kang

, Sibille

, Kaminski

, Tseng

(2012) MetaQC: Objective quality control and inclusion/exclusion criteria for genomic meta-analysis. Nucleic Acids Res 40, e15.

19.

Hong

, Breitling

, McEntee

, Wittner

, Nemhauser

, Chory

(2006) RankProd: A bioconductor package for detecting differentially expressed genes in meta-analysis. Bioinformatics 22, 2825–2827.

20.

Hong

, Breitling

(2008) A comparison of meta-analysis methods for detecting differentially expressed genes in microarray experiments. Bioinformatics 24, 374–382.

21.

Bindea

, Mlecnik

, Hackl

, Charoentong

, Tosolini

, Kirilovsky

, Fridman

, Pages

, Trajanoski

, Galon

(2009) ClueGO: A Cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics 25, 1091–1093.

22.

Kanehisa

, Goto

(2000) KEGG: Kyoto encyclopedia of genes and genomes. Nucleic Acids Res 28, 27–30.

23.

Lee

, Shih

, Murphy

, Johnson

(2003) NF-E2-related factor-2 mediates neuroprotection against mitochondrial complex I inhibitors and increased concentrations of intracellular calcium in primary cortical neurons. J Biol Chem 278, 37948–37956.

24.

Erickson

, Nevarea

, Gipp

, Mulcahy

(2002) Identification of a variant antioxidant response element in the promoter of the human glutamate-cysteine ligase modifier subunit gene. Revision of the ARE consensus sequence. J Biol Chem 277, 30730–30737.

25.

Feng

, Bai

, Wang

, Xie

, Sun

, Han

, Song

, Crack

, Duan

, Lei

(2014) Robust gene dysregulation in Alzheimer’s disease brains. J Alzheimers Dis 41, 587–597.

26.

Han

, Wang

, Zeng

, Feng

, Yu

, Cao

, Yi

, Zhou

, Jin

, Duan

, Wang

, Lei

(2013) Characteristic transformation of blood transcriptome in Alzheimer’s disease. J Alzheimers Dis 35, 373–386.

27.

Esposito

, Ana Clara

, Verstreken

(2012) Synaptic vesicle trafficking and Parkinson’s disease. Dev Neurobiol 72, 134–144.

28.

Emmanouilidou

, Minakaki

, Keramioti

, Xylaki

, Balafas

, Chrysanthou-Piterou

, Kloukina

, Vekrellis

(2016) GABA transmission via ATP-dependent K+channels regulates alpha-synuclein secretion in mouse striatum. Brain 139, 871–890.

29.

Gerlach

, Gsell

, Kornhuber

, Jellinger

, Krieger

, Pantucek

, Vock

, Riederer

(1996) A post mortem study on neurochemical markers of dopaminergic, GABA-ergic and glutamatergic neurons in basal ganglia-thalamocortical circuits in Parkinson syndrome. Brain Res 741, 142–152.

30.

Zheng

, Liao

, Locascio

, Lesniak

, Roderick

, Watt

, Eklund

, Zhang-James

, Kim

, Hauser

, Grunblatt

, Moran

, Mandel

, Riederer

, Miller

, Federoff

, Wullner

, Papapetropoulos

, Youdim

, Cantuti-Castelvetri

, Young

, Vance

, Davis

, Hedreen

, Adler

, Beach

, Graeber

, Middleton

, Rochet

, Scherzer

, Global PD Gene Expression (GPEX) Consortium (2010) PGC-1alpha, a potential therapeutic target for early intervention in Parkinson’s disease. Sci Transl Med 2, 52ra73.

31.

Chandrasekaran

, Hatanpaa

, Brady

, Rapoport

(1996) Evidence for physiological down-regulation of brain oxidative phosphorylation in Alzheimer’s disease. Exp Neurol 142, 80–88.

32.

Fisar

, Hroudova

, Hansikova

, Spacilova

, Lelkova

, Wenchich

, Jirak

, Zverova

, Zeman

, Martasek

, Raboch

(2016) Mitochondrial respiration in the platelets of patients with Alzheimer’s disease. Curr Alzheimer Res 13, 930–941.

33.

Selkoe

(2002) Alzheimer’s disease is a synaptic failure. Science 298, 789–791.

34.

, Sun

, Chen

, Xu

, Bu

, Zheng

(2016) Implications of GABAergic Neurotransmission in Alzheimer’s disease. Front Aging Neurosci 8, 31.

35.

Keller

, Hanni

, Markesbery

(2000) Impaired proteasome function in Alzheimer’s disease. J Neurochem 75, 436–439.

36.

Shankar

, Li

, Mehta

, Garcia-Munoz

, Shepardson

, Smith

, Brett

, Farrell

, Rowan

, Lemere

, Regan

, Walsh

, Sabatini

, Selkoe

(2008) Amyloid-beta protein dimers isolated directly from Alzheimer’s brains impair synaptic plasticity and memory. Nat Med 14, 837–842.

37.

Brooks

, Lynch

, Ingle

, Hatton

, Emson

, Faull

, Starkey

(2007) Gene expression profiles of metabolic enzyme transcripts in Alzheimer’s disease. Brain Res 1127, 127–135.

38.

Zhao

, Cribbs

, Anderson

, Cummings

, Su

, Wasserman

, Cotman

(2003) The induction of the TNFalpha death domain signaling pathway in Alzheimer’s disease brain. Neurochem Res 28, 307–318.

39.

Coyle

, Price

, DeLong

(1983) Alzheimer’s disease: A disorder of cortical cholinergic innervation. Science 219, 1184–1190.

40.

Calabrese

, Scapagnini

, Colombrita

, Ravagna

, Pennisi

, Giuffrida Stella

, Galli

, Butterfield

(2003) Redox regulation of heat shock protein expression in aging and neurodegenerative disorders associated with oxidative stress: A nutritional approach. Amino Acids 25, 437–444.

41.

Nioi

, McMahon

, Itoh

, Yamamoto

, Hayes

(2003) Identification of a novel Nrf2-regulated antioxidant response element (ARE) in the mouse NAD(P)H: Quinone oxidoreductase 1 gene: Reassessment of the ARE consensus sequence. Biochem J 374, 337–348.

42.

Hayes

, Dinkova-Kostova

(2014) The Nrf2 regulatory network provides an interface between redox and intermediary metabolism. Trends Biochem Sci 39, 199–218.

43.

Motohashi

, Katsuoka

, Engel

, Yamamoto

(2004) Small Maf proteins serve as transcriptional cofactors for keratinocyte differentiation in the Keap1-Nrf2 regulatory pathway. Proc Natl Acad Sci U S A 101, 6379–6384.

44.

Fujiwara

, Kataoka

, Nishizawa

(1993) Two new members of the maf oncogene family, mafK and mafF, encode nuclear b-Zip proteins lacking putative trans-activator domain. Oncogene 8, 2371–2380.

45.

Igarashi

, Kataoka

, Itoh

, Hayashi

, Nishizawa

, Yamamoto

(1994) Regulation of transcription by dimerization of erythroid factor NF-E2 p45 with small Maf proteins. Nature 367, 568–572.

46.

Kannan

, Solovieva

, Blank

(2012) The small MAF transcription factors MAFF, MAFG and MAFK: Current knowledge and perspectives. Biochim Biophys Acta 1823, 1841–1846.

47.

Motohashi

, Katsuoka

, Shavit

, Engel

, Yamamoto

(2000) Positive or negative MARE-dependent transcriptional regulation is determined by the abundance of small Maf proteins. Cell 103, 865–875.

48.

Mattson

(2004) Pathways towards and away from Alzheimer’s disease. Nature 430, 631–639.

49.

Loring

, Goldstein

, Chen

, Drane

, Lah

, Zhao

, Larrabee

, Alzheimer’s Disease Neuroimaging, Initiative (2016) False-positive error rates for reliable digit span and auditory verbal learning test performance validity measures in amnestic mild cognitive impairment and early Alzheimer disease. Arch Clin Neuropsychol 31, 313–331.

50.

Kastner

, Hirsch

, Agid

, Javoy-Agid

(1993) Tyrosine hydroxylase protein and messenger RNA in the dopaminergic nigral neurons of patients with Parkinson’s disease. Brain Res 606, 341–345.

51.

Habak

, Noreau

, Nagano-Saito

, Mejia-Constain

, Degroot

, Strafella

, Chouinard

, Lafontaine

, Rouleau

, Monchi

(2014) Dopamine transporter SLC6A3 genotype affects cortico-striatal activity of set-shifts in Parkinson’s disease. Brain 137, 3025–3035.

52.

Brighina

, Riva

, Bertola

, Saracchi

, Fermi

, Goldwurm

, Ferrarese

(2013) Analysis of vesicular monoamine transporter 2 polymorphisms in Parkinson’s disease. Neurobiol Aging 34, 1712. e1719–1713.

53.

, Pan

, Kao

, Li

, Kohane

, Chan

, Yankner

(2004) Gene regulation and DNA damage in the ageing human brain. Nature 429, 883–891.

54.

Lavedan

(1998) The synuclein family. Genome Res 8, 871–880.

55.

Kim

, Seo

, Suh

(2004) Alpha-synuclein, Parkinson’s disease, and Alzheimer’s disease. Parkinsonism Relat Disord 10(Suppl 1), S9–13.

56.

Hashimoto

, Rockenstein

, Mante

, Mallory

, Masliah

(2001) beta-synuclein inhibits alpha-synuclein aggregation: A possible role as an anti-parkinsonian factor. Neuron 32, 213–223.

57.

McLoughlin

, Miller

CCJ

(2008) The FE65 proteins and Alzheimer’s disease. J Neurosci Res 86, 744–754.

58.

Hao

, Chai

, McLoughlin

, Chan

HYE

, Lau

(2012) Promoter characterization and genomic organization of the human X11 beta gene APBA2. Neuroreport 23, 146–151.

59.

Nalls

, Pankratz

, Lill

, Do

, Hernandez

, Saad

, DeStefano

, Kara

, Bras

, Sharma

, Schulte

, Keller

, Arepalli

, Letson

, Edsall

, Stefansson

, Liu

, Pliner

, Lee

, Cheng

, International Parkinson’s Disease Genomics Consortium (IPDGC), Parkinson’s Study Group (PSG) Parkinson’s Research: The Organized GENetics Initiative (PROGENI), 23andMe, GenePD, NeuroGenetics Research Consortium (NGRC), Hussman Institute of Human Genomics (HIHG), Ashkenazi Jewish Dataset Investigator, Cohorts for Health, Aging Research in Genetic Epidemiology (CHARGE), North American Brain Expression Consortium (NABEC), United Kingdom Brain Expression Consortium (UKBEC), Greek Parkinson’s Disease Consortium, Alzheimer Genetic Analysis Group, Ikram

, Ioannidis

, Hadjigeorgiou

, Bis

, Martinez

, Perlmutter

, Goate

, Marder

, Fiske

, Sutherland

, Xiromerisiou

, Myers

, Clark

, Stefansson

, Hardy

, Heutink

, Chen

, Wood

, Houlden

, Payami

, Brice

, Scott

, Gasser

, Bertram

, Eriksson

, Foroud

, Singleton

(2014) Large-scale meta-analysis of genome-wide association data identifies six new risk loci for Parkinson’s disease. Nat Genet 46, 989–993.

60.

Lambert

, Ibrahim-Verbaas

, Harold

, Naj

, Sims

, Bellenguez

, DeStafano

, Bis

, Beecham

, Grenier-Boley

, Russo

, Thorton-Wells

, Jones

, Smith

, Chouraki

, Thomas

, Ikram

, Zelenika

, Vardarajan

, Kamatani

, Lin

, Gerrish

, Schmidt

, Kunkle

, Dunstan

, Ruiz

, Bihoreau

, Choi

, Reitz

, Pasquier

, Cruchaga

, Craig

, Amin

, Berr

, Lopez

, De Jager

, Deramecourt

, Johnston

, Evans

, Lovestone

, Letenneur

, Morón

, Rubinsztein

, Eiriksdottir

, Sleegers

, Goate

, Fiévet

, Huentelman

, Gill

, Brown

, Kamboh

, Keller

, Barberger-Gateau

, McGuiness

, Larson

, Green

, Myers

, Dufouil

, Todd

, Wallon

, Love

, Rogaeva

, Gallacher

, St George-Hyslop

, Clarimon

, Lleo

, Bayer

, Tsuang

, Yu

, Tsolaki

, Bossú

, Spalletta

, Proitsi

, Collinge

, Sorbi

, Sanchez-Garcia

, Fox

, Hardy

, Deniz Naranjo

, Bosco

, Clarke

, Brayne

, Galimberti

, Mancuso

, Matthews

, European Alzheimer’s Disease Initiative (EADI), Genetic and Environmental Risk in Alzheimer’s Disease, Alzheimer’s Disease Genetic Consortium, Cohorts for Heart and Aging Research in Genomic Epidemiology, Moebus

, Mecocci

, Del Zompo

, Maier

, Hampel

, Pilotto

, Bullido

, Panza

, Caffarra

, Nacmias

, Gilbert

, Mayhaus

, Lannefelt

, Hakonarson

, Pichler

, Carrasquillo

, Ingelsson

, Beekly

, Alvarez

, Zou

, Valladares

, Younkin

, Coto

, Hamilton-Nelson

, Gu

, Razquin

, Pastor

, Mateo

, Owen

, Faber

, Jonsson

, Combarros

, O’Donovan

, Cantwell

, Soininen

, Blacker

, Mead

, Mosley

Jr , Bennett

, Harris

, Fratiglioni

, Holmes

, de Bruijn

, Passmore

, Montine

, Bettens

, Rotter

, Brice

, Morgan

, Foroud

, Kukull

, Hannequin

, Powell

, Nalls

, Ritchie

, Lunetta

, Kauwe

, Boerwinkle

, Riemenschneider

, Boada

, Hiltuenen

, Martin

, Schmidt

, Rujescu

, Wang

, Dartigues

, Mayeux

, Tzourio

, Hofman

, Nöthen

, Graff

, Psaty

, Jones

, Haines

, Holmans

, Lathrop

, Pericak-Vance

, Launer

, Farrer

, van Duijn

, Van Broeckhoven

, Moskvina

, Seshadri

, Williams

, Schellenberg

, Amouyel

(2013) Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat Genet 45, 1452–U1206.

61.

Zawia

, Lahiri

, Cardozo-Pelaez

(2009) Epigenetics, oxidative stress, and Alzheimer disease. Free Radic Biol Med 46, 1241–1249.

62.

Milani

, Ambrosi

, Gammoh

, Blandini

, Cereda

(2013) SOD1 and DJ-1 converge at Nrf2 pathway: A clue for antioxidant therapeutic potential in neurodegeneration. Oxid Med Cell Longev 2013, 836760.

63.

Lee

, Koh

(2010) Roles of zinc and metallothionein-3 in oxidative stress-induced lysosomal dysfunction, cell death, and autophagy in neurons and astrocytes. Mol Brain 3, 30.

64.

Barrera

, Gentile

, Pizzimenti

, Canuto

, Daga

, Arcaro

, Cetrangolo

, Lepore

, Ferretti

, Dianzani

, Muzio

(2016) Mitochondrial dysfunction in cancer and neurodegenerative diseases: Spotlight on fatty acid oxidation and lipoperoxidation products. Antioxidants (Basel) 5, pii:E7.

65.

Cook

, Vitale

, Ravishankar

, Matigian

, Sutherland

, Shan

, Sutharsan

, Perry

, Silburn

, Mellick

, Whitelaw

, Wells

, Mackay-Sim

, Wood

(2011) NRF2 activation restores disease related metabolic deficiencies in olfactory neurosphere-derived cells from patients with sporadic Parkinson’s disease. PLoS One 6, e21907.