Abstract
Background The gut microbiome and its metabolites can impact brain health and are altered in Parkinson’s disease (PD) patients. It has been recently demonstrated that PD patients have reduced fecal levels of the potent epigenetic modulator butyrate and its bacterial producers. Here, we investigate whether the changes in the gut microbiome and associated metabolites are linked to PD symptoms and epigenetic markers in leucocytes and neurons.
Methods Stool, whole blood samples, and clinical data were collected from 55 PD patients and 55 controls. We performed DNA methylation analysis on whole blood samples and analyzed the results in relation to fecal short-chain fatty acid concentrations and microbiota composition. In another cohort, prefrontal cortex neurons were isolated from control and PD brains. We identified the genome-wide DNA methylation by targeted bisulfite sequencing.
Results We show that lower fecal butyrate and reduced Roseburia, Romboutsia, and Prevotella counts are linked to depressive symptoms in PD patients. Genes containing butyrate-associated methylation sites include PD risk genes and significantly overlap with sites epigenetically altered in PD blood leucocytes, predominantly neutrophils, and in brain neurons, relative to controls. Moreover, butyrate-associated methylated-DNA (mDNA) regions in PD overlap with those altered in gastrointestinal, autoimmune, and psychiatric diseases.
Conclusions Decreased levels of bacterially produced butyrate are linked to epigenetic changes in leucocytes and neurons from PD patients, and to the severity of their depressive symptoms. PD shares common butyrate-dependent epigenetic changes with certain gastrointestinal and psychiatric disorders, which could be relevant for their epidemiological linkage.
Background
Parkinson’s disease (PD) is a neurodegenerative disorder that is typically characterized by motor impairments due to the death of dopaminergic (DA) neurons located within the substantia nigra pars compacta (SNpc) [1]. A hallmark of PD is the aggregation of misfolded α-synuclein protein in both the central and peripheral nervous system, forming aggregates or Lewy bodies [2]. The discovery of α-synuclein aggregates in the enteric nervous system, coupled with the early gastrointestinal (GI) symptoms of PD (e.g. constipation), has led to the hypothesis that the pathogenesis of PD may originate in the GI tract, or at least outside of the central nervous system [3]. Additionally, it is hypothesized that α-synuclein aggregates can be transported in a prion-like manner from the enteric nervous system to the central nervous system through the vagus nerve [4]. This hypothesis is supported by the decreased risk for PD amongst patients who have undergone a vagotomy [5]. Further intertwining the GI system with PD is the finding that the presence of gut microbiome is required for mice that overexpress α-synuclein to develop motor deficits [6]. These same mice displayed increased motor deficits after receiving a fecal transplant from a PD patient compared to receiving transplant from a healthy donor [6]. Moreover, the composition and function of the gut microbiome in PD patients has been shown to be significantly different from healthy controls in multiple studies [7–9]. While research has consistently shown a difference in the PD microbiome compared to the healthy microbiome, there is no consensus on the exact differences or what effect this dysbiosis might have on the PD patient [9].
Among the common differences seen in the PD microbiome across multiple studies is a decrease in bacteria which produce short-chain fatty acids (SCFAs) [10]. SCFAs are saturated fatty acids produced via the fermentation of dietary fiber by certain colonic bacteria [11]. The deficiency of SCFAs has been implicated in multiple diseases such as autoimmune disorders, cancer, metabolic syndromes, and neurological disorders [12]. Fecal samples from PD patients have been shown to harbor significantly lower concentrations of the SCFAs acetate, propionate, and butyrate when compared to healthy controls [13]. Butyrate, specifically, has been implicated as an important bacterial metabolite due to its role as a strong endogenous histone deacetylase (HDAC) inhibitor indirectly affecting DNA-methylation [14], allowing it to epigenetically alter the gene expression of multiple cell types [15]. Several studies have demonstrated the ability of butyrate to reduce the inflammatory properties of both innate and adaptive immune cells through inhibiting reactive oxygen species release, inflammatory cytokine production, and inducing activated immune cell apoptotic mechanisms [16–18]. In addition to its immune modulating properties, butyrate has been shown in multiple in-vivo studies to influence the CNS through decreasing blood-brain barrier permeability, decreasing microglial activation, and relieving anxiety and depression, which are both common prodromal PD symptoms. Butyrate has also shown neuroprotective effects in PD mouse models [19, 20]. However, the possible impact of the low butyrate level observed in PD patients on epigenome status as well as on clinical symptoms has yet to be elucidated.
To test whether butyrate impacts epigenetic markers in the blood and brain of PD patients, and whether this is associated with symptom severity, we performed DNA methylation profiling in whole blood samples and neuronal tissue from two cohorts of PD patients and controls (Fig 1), and related the findings to fecal gut microbiome and metabolite data and clinical symptoms.
Methods
Human samples and metadata
Blood samples, clinical data, microbiome count data, inflammatory and permeability markers, and stool SCFA levels used in this study are from the Helsinki Parkinson microbiome cohort, and methodology of sampling, sample processing, and analysis have been described previously [21, 22]. The study was approved by the ethics committee of the Hospital District of Helsinki and Uusimaa. All participants gave informed consent. Human prefrontal cortex tissue for this study was obtained from the Parkinson’s UK Brain Bank, NIH NeuroBioBank, and Michigan Brain Bank, with approval from the ethics committee of the Van Andel Research Institute (IRB #15025). Methodology of sampling, sample processing, and analysis have been described previously [23]. We used the differential methylated cytosines data (supplementary file 5) in our analysis.
Statistical Analyses
We performed statistical analyses with R [24] (v3.6.1) with packages including metagenomeseq [25] (v 1.27.3) for differential microbial data comparisons; ChAMP [26] (v 2.20.1), and minfi [27] (v 1.31.1) for EPIC array analysis; Tensor Composition Analysis (TCA) [28] (v 1.1.0) for epigenome estimation for immune cell types; limma [29] (v 3.41.17) for robust linear regression; ggplot2 [30] (v 3.3.1) for data visualization. All the code used for this work is publicly available: Github: https://github.com/AojiXie/PD_microbiome_DNA_methylation
Genome-wide DNA methylation profiling
Whole-genome DNA methylation profiling for each sample was performed on Illumina Methylation EPIC BeadChip microarrays at Van Andel Institute Genomic core. Bisulfite converted DNA samples (n = 136, including replicates) were randomized across arrays (8 samples/array). Data generated from the microarrays were preprocessed with Minfi (v 1.31.1). Normalization was performed with noob [31]. We confirmed that the sex of the individuals matched that inferred from the DNA methylome (minfi getSex() function).
The portions of immune cell types (CD8+T cell, CD4+T cell, B cell, NK, monocyte, and neutrophil) were estimated by CIBERSORT [32] (Fig S2a) using whole blood-specific markers as reference [33]. The filtering method in ChAMP was used to filter probes: Probes that overlapped SNPs [34] (minor allele frequency > 0.05) on the CpG or single-base extension (95,485 probes), probes that aligned to multiple locations (42,558 probes), probes with a beadcount < 3 in at least 5% of samples (3380 probes), MultiHit Start [35] (11 probes), probes located on X,Y chromosome (16,541 probes), NoCG Start (2953), and those that failed detectability (p > 0.01) (7,951 probes) were excluded. After processing, 739,597 probes were left. Champ.svd() in ChAMP was used to test the batch effects. Those batch effects (array and slide positions) were corrected by ComBat[36]. After batch effects correction, the M value was ready for the subsequent statistical analysis. Cell-type-specific resolution epigenetics were performed using tensor composition analysis [28].
Statistical analysis for differentially methylated sites
DNA methylation analysis involved robust linear regression models with empirical Bayes from the limma (v 3.41.17) statistical package [29]. P-values were adjusted with a Benjamini-Hochberg correction for multiple testing and those with FDR q < 0.05 were deemed significant.
Model
1: In whole blood methylation epigenome
Variable selection: Cell type percentages were used as covariates because the whole blood contains several immune cell types in varying proportions. This variation might affect the interpretations of DNA methylation levels based on whole blood DNA[37]. BMI and smoking history are used as covariates because their associations with whole blood DNA methylation have been observed [38, 39].
M value ∼ Butyrate + Age + Sex + Smoking history + BMI + CD4+T cell + CD8+T cell + B cell + Monocyte + Neutrophil
2: In cell specific epigenome
M value ∼ Butyrate + Age + Sex + Smoking history + BMI
Pathway enrichment analysis
To identify proximal interactions with gene targets, we used the GREAT (v4.0.4) software [40]. Gene annotation was performed for the gene targets of the significant cytosine sites in our analysis and for the background, consisting of gene targets for all cytosines included in our analysis. The background was 18,455 genes. Pathway analysis of methylated cytosines altered in PD and correlated with butyrate level was performed using g:Profiler [41], with networks determined by EnrichmentMap and clustered by AutoAnnotate in Cytoscape (v3.7.1) [42].
Since enhancer elements dynamically regulate gene expression through three-dimensional physical interactions, we analyzed chromatin interaction data to reveal the gene targets of enhancers relevant to the differential methylation sites affected by butyrate. For this analysis, we used promoter-centric chromatin interactions identified in blood cell types [43]. Gene enrichment set test of different blood cell types was performed by GESA [44, 45] (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp) Reactome gene sets. Benjamini-Hochberg FDR q < 0.01 with minimal gene set less than 100 were used as the significant threshold.
Genetic-epigenetic correlation
Genetic-epigenetic analysis was performed using LD Score (LDSC) software [46, 47], to estimate the correlations between butyrate-associated mDNA regions and the GWAS summary statistics of other diseases. To construct butyrate-associated mDNA regions for linkage disequilibrium (LD) score regression analysis, SNPs within +-5000bp of EPIC chip array sites were included, and the p-values of methylation cytosines in butyrate linear model were assigned to those SNPs. If a SNP was within +-5000bp of more than one methylated cytosine, the smallest p-value was selected. The summary statistic of a 2019 PD GWAS study [48] was used in this analysis. For other diseases, we used the summary statistics which are provided in the LD Hub interface [49]. P < 0.05 was used as the significance threshold.
The common SNPs (with p < 0.05) between butyrate-associated mDNA regions, PD GWAS summary statistics, and other diseases, respectively, were extracted. GREAT [40] was used to get the gene annotation of those common SNPs with Association rule: Basal + extension: 5000 bp upstream, 1000 bp downstream, 600000 bp max extension, curated regulatory domains included. Gene set enrichment analysis was performed by GSEA using Reactome gene sets [44, 45] (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp). Benjamini-Hochberg FDR q < 0.01 with minimal gene set less than 100 were used as the significant threshold.
Results
Butyrate-producing microbes are altered in PD and correlate with depressive symptoms
First, we re-analyzed previously published [21, 22] raw data on SCFA levels and 16S rRNA gene amplicon counts from stool samples of PD patients and healthy controls. The metagenomeSeq was used for differential microbial analysis [25]. Unlike RNAseq studies, most OTUs are rare (absent from a large number of samples) because of insufficient sequencing depth (under sampling) or some organisms only being present in few samples. This sparsity can lead to strong biases when sequence read counts are tested for significant differences. Zero counts in samples with low coverage are misinterpreted as absent taxonomic features. The advantages of metagenomeSeq is the zero-inflated Gaussian mixture (ZIG) model which removes testing biases resulting from under sampling and the cumulative-sum scaling (CSS) normalization method to avoid biases of uneven sequencing depth [50]. We can confirm significantly reduced butyrate levels (Benjamini-Hochberg FDR q < 0.05, robust linear regression) (Fig S1a) and differential abundances of Bifidobacterium, Butyricicoccus, Clostridium_XlVa, Lactobacillus, Prevotella, and Roseburia in PD patients (Benjamini-Hochberg FDR q < 0.05, metagenomeSeq ZIG model) (Fig S1b) [21]. Important new findings within the PD group are links between depressive symptoms as measured using the Geriatric Depression Scale (GDS-15) and lower fecal butyrate levels (Benjamini-Hochberg FDR q < 0.05, robust linear regression) (Fig 2a) as well as lower counts of Prevotella, Romboutsia, and Roseburia and higher counts of Deltaproteobacteria_unclassified (Benjamini-Hochberg FDR q < 0.05, metagenomeSeq ZIG model) (Fig. S1c, 2b). In line with these findings, we confirm a positive correlation of Romboutsia and Roseburia with butyrate levels in PD patients (Fig.S1c) [22]. In addition, no other symptoms including gastrointestinal (Rome-III questionnaire; Wexner score), motor symptoms (the Unified Parkinson’s Disease Rating Scale, UPDRS), nor non-motor symptoms (Non-Motor Symptoms Questionnaire, NMSQ; Non-Motor Symptoms Scale, NMSS) were linked with butyrate in PD patients.
Fecal butyrate levels are associated with epigenetic alterations in leucocytes and neurons of PD patients
We identified 3195 CpG sites that correlated significantly with stool butyrate levels in PD patients (Benjamini-Hochberg FDR q < 0.05, robust linear regression) (Fig. 3a) and 2950 CpG sites that were significantly changed in PD relative to control (Benjamini-Hochberg FDR q < 0.05, robust linear regression). In a previous study, we identified genes epigenetically altered in cortical neurons from PD patients [23]. Genes containing the butyrate-associated methylated cytosines in blood cells significantly overlapped with those genes altered in PD patients’ blood cells (relative to controls) and with those genes altered in PD patients’ prefrontal cortex neurons (relative to controls) (genes containing modified cytosines at Benjamini-Hochberg FDR q < 0.05, Fisher’s exact test p < 0.05) (Fig. 3b, c). Pathways involving those genes epigenetically altered in PD or with butyrate were identified and included neurodevelopment, cell development, synaptic transmission, metabolism, and signal transduction (Fig. 3d).
Epigenetic links between butyrate and leucocytes differ between cell types
To test if leucocyte epigenetic changes are associated with fecal butyrate, we used cell type-specific resolution epigenetics, TCA [28], to analyze the epigenome for each cell type (Fig S2). To test the epigenetic links of butyrate on different immune cell types, we analyzed the significant methylation sites altered by butyrate in the epigenome of neutrophils, monocytes, CD8+ T cells, CD4+ T cells, and B cells. All cell types were linked differently to butyrate, with neutrophils and monocytes containing the largest number of genes that were epigenetically linked to butyrate (Fig. 4a). Levels of inflammatory stool cytokines were derived from a previously published study of the same samples [22]. We also analyzed the methylated cytosines that were correlated with blood inflammatory cytokine levels. There was a large overlap between the genes epigenetically altered by butyrate and the genes epigenetically associated with levels of inflammatory cytokines (TNF, IL6, CXCL8, IL4, IL1B, IL10, IFNg, IL13, IL12p70, IL2 and LBP) in both monocytes and neutrophils (Fisher’s exact test, p < 0.05) (Fig. 4b). Pathway analysis shows that epigenetic alterations of three recently identified PD polygenic risk pathways [51] (neutrophil degranulation, metabolism of lipids, and innate immune system) in the monocyte and neutrophil epigenome were linked to butyrate levels (Fig. 4c). We also analyzed the effects of other SCFAs, and our results showed that the genes epigenetically altered in neutrophils were most strongly associated with butyrate compared to other SCFAs (Fig. S3a) and that there were no significant PD polygenic risk pathways altered by other SCFAs in neutrophils (Fig. S3b). These results indicate that neutrophil and monocyte functions are specifically altered in PD patients and that they might be epigenetically altered by butyrate.
Epigenetic-genetic correlation between diseases and butyrate-associated methylated-DNA (mDNA) regions
To investigate the epigenetic-genetic correlation, we ran a linear regression model to compare the results of methylation analysis to the GWAS of the odds-risk ratios for PD and other diseases, including gastrointestinal (ulcerative colitis, Crohn’s disease), autoimmune (rheumatoid arthritis), neurodegenerative (Alzheimer’s disease), and psychiatric diseases (bipolar disorder, schizophrenia, and major depression) and our DNA methylation study, adjusting for LD scores. We found that butyrate-associated mDNA regions are most strongly related to GWAS loci linked to PD and inflammatory bowel diseases. Further significant links were found with the GWAS loci of rheumatoid arthritis and bipolar disorder (linear regression, p < 0.05) (Fig. 5a). In contrast, Alzheimer’s disease, schizophrenia, or major depression GWAS loci were not linked to butyrate-associated mDNA regions. Pathway analysis was performed on the common genomic regions (butyrate-associated mDNA regions, GWAS loci of PD and each of the other diseases, respectively). All of the six PD polygenic risk pathways without known PD risk loci [51] were significantly altered in ulcerative colitis, Crohn’s disease, rheumatoid arthritis, and bipolar disorder (FDR q < 0.05) (Fig. 5b). The top 15 significant pathways are shown in Fig. S4. These results indicate that butyrate-associated epigenetic changes may contribute to observed epidemiologic links between PD and gastrointestinal, autoimmune, as well as certain psychiatric diseases.
Discussion
Several studies have found alterations of microbial composition and metabolites in PD, but the mechanisms linking these changes to PD and its symptoms are not well understood [9]. The bacterial metabolite butyrate is a strong HDAC inhibitor, with known epigenetic effects [52]. Several studies have shown the ability of butyrate to affect the innate and adaptive immune system as well as to alter the blood-brain barrier permeability [19, 53]. This study is the first to examine the possible role of epigenetic changes as a link between gut microbiota, its metabolites, and the pathophysiology of neural and immune cells in PD.
In this study, genome-wide DNA methylation profiling revealed significant epigenetic changes in the leucocytes and neurons of PD patients that overlap with genes the methylation of which is linked to fecal butyrate levels, including loci in PD risk genes [23]. Furthermore, butyrate levels correlated most strongly with abundance of bacteria belonging to the genera Romboutsia and Roseburia, known SCFA producers.
A major finding of our study is that the epigenome of specific immune cell types in PD is differentially linked to fecal butyrate levels. In particular, neutrophil and monocyte epigenomes show the strongest links to butyrate. We have also shown that the neutrophil epigenome is linked less to other SCFAs. A recent study [51] reported that neutrophil degranulation is potentially linked to PD risk. Our gene set enrichment analysis suggests that butyrate-linked epigenetic changes may impact this pathway. Several studies have reported an increase of neutrophils in PD patients, including studies using flow cytometric and epigenetic profiling approaches [54], which is consistent with our cell deconvolution results from the methylation profiling. However, our study for the first time decomposed the estimated epigenome of each cell type and investigated the epigenetic change from different immune cell types in PD and in connection to SCFAs. Our results suggest that butyrate may impact PD through epigenetic effects on innate immune cells and PD-related genes.
The exact dynamics that link fecal butyrate levels with blood leucocyte and brain epigenetic are not known. While a significant proportion of microbial-released butyrate is rapidly taken up and consumed locally in the gut, butyrate can cross the epithelial barrier and enter the circulation via the hepatic portal vein [55]. Microbiota-derived butyrate impacts histone acetylation in multiple tissues [56]. While concentrations in the portal vein are still considerable, concentrations in peripheral blood appear to be relatively low [55]. Thus, leucocytes get exposed to butyrate mostly in the gut wall and portal vein, while the impact of butyrate in systemic venous blood can be expected to be less and influenced by liver function. Epigenetic changes in blood leucocytes may impact inflammation systemically and in the brain [19]. At physiological concentrations, butyrate’s impact on brain metabolism and hippocampal neurogenesis has been shown in pigs [57]. While our findings support the importance of epigenetic mechanisms, their relative impact on the physiological effects of butyrate in the brain as compared to other mechanisms remains to be established.
Interestingly, our results suggest that patterns of butyrate-related epigenetic changes in PD are most similar to those found in inflammatory bowel disease and clearly less similar to those found in Alzheimer’s disease. Bowel diseases, in particular inflammatory bowel diseases [58], have been linked to an increased risk of PD and Alzheimer’s disease, but associations are stronger for PD [59]. Our results suggest that microbiome-related epigenetic modulation could be a mechanism linking gastrointestinal disorders and PD. Also, bipolar disorder [60] has been linked to an increased PD risk, and our results support a role for common epigenetic mechanisms in this context. In contrast, we could not find significant overlap with epigenetic patterns found in schizophrenia [61] and depression [62] which points to a lesser impact of epigenetics linking these disorders to PD. Interestingly, some overlap was seen with epigenetic patterns of rheumatoid arthritis, which reportedly is associated with a decreased PD risk [63–65]. Rheumatoid arthritis has also been linked to the gut microbiome, but changes have been somewhat opposite to those seen in PD, e.g. increase of Prevotella abundance in arthritis but decrease in PD [66].
In this study, we partly reanalyzed microbiome and metabolite data with methods not used in the previous publications [21, 22]. We observed that fecal bacterial butyrate is inversely correlated with depressive symptoms (GDS-15) in PD patients. Although there is no correlation between bacterial butyrate and other non-motor symptoms-related scales which include depressive related items (UPDRS I, NMSS and NMSQ), GDS-15 is more specific for assessing depressive symptoms. While we were able to reproduce PD-related microbiota alterations and decreased butyrate levels identified using earlier methods, we gained important new insights. Fecal butyrate and counts of Prevotella, Romboutsia, and Roseburia were negatively correlated with depressive symptoms in PD patients, potentially implicating bacterial metabolites in this important non-motor PD symptom.
Conclusions
In sum, combining metabolite, microbiome, clinical data, and DNA methylation profiling, our study for the first time reveals a possible link between gut microbiome metabolite production and epigenetic changes implicating immune and neural pathways in PD patients with potential impact on depressive symptoms. Furthermore, our results point to microbiota-dependent epigenetic modulation as a potential pathway linking inflammatory bowel diseases and PD. Further research of altered bacterial metabolism and its impact on host physiology may reveal new biomarkers and therapeutic targets for PD.
Data Availability
Microbiota data are available at the European Nucleotide Archive (accession number PRJEB27564) (https://www.ebi.ac.uk/ena/browser/view/PRJEB27564). Other data and files utilized in this study are available from the corresponding authors upon reasonable request.
Ethics approval and consent to participate
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the ethics committees of the Van Andel Research Institute and the Hospital District of Helsinki and Uusimaa. All participants gave written informed consent.
Consent for publication
Not applicable
Availability of data and materials
Microbiota data are available at the European Nucleotide Archive (accession number PRJEB27564) (https://www.ebi.ac.uk/ena/browser/view/PRJEB27564). Other data and files utilized in this study are available from the corresponding authors upon reasonable request.
List of abbreviations
- PD
- Parkinson’s disease
- SCFAs
- Short-chain fatty acids
- BMI
- Body mass index
- TCA
- Tensor composition analysis
- LD
- linkage disequilibrium
- UPDRS
- Unified Parkinson’s Disease Rating Scale
- NMSQ
- Non-Motor Symptoms Questionnaire
- NMSS
- Non-Motor Symptoms Scale
- GDS-15
- Geriatric Depression Scale
- GSEA
- Gene set enrichment analysis
Competing interests
PB has received commercial support as a consultant from Axial Therapeutics, Calico, CuraSen, Fujifilm-Cellular Dynamics Inc., IOS Press Partners, LifeSci Capital LLC, Lundbeck A/S, Idorsia and Living Cell Technologies LTD. He has received commercial support for grants/research from Lundbeck A/S and Roche. He has ownership interests in Acousort AB and Axial Therapeutics. S.G. receives commercial support as a consultant from Coleman Research and Biogen.
VTEA, PABP, LP, PA and FS have patents issued (FI127671B & US10139408B2) and pending (US16/186,663 & EP3149205) that are assigned to NeuroBiome Ltd.
FS is founder and CEO of NeuroInnovation Oy and NeuroBiome Ltd., is a member of the scientific advisory board and has received consulting fees and stock options from Axial Biotherapeutics.
MGT is an advisor to INmune Bio, Longevity Biotech, Prevail Therapeutics, and Weston Garfield Foundation. MGT has patents issued (US Pat. Nos. 7144987B1 and 7244823B2) and pending (US20150239951, WO2019067789, 62/901698, see efiling Ack37193677, 62/905747, see efilingAck37274773) for co-invention of DN-TNFs.
Funding
This work was supported by a Farmer Family Foundation grant award and a Gibby & Friends vs. Parky Award to PB, with LB, JAP, and VL. FS received funding from the Michael J. Fox Foundation for Parkinson’s Research, the Academy of Finland (295724, 310835), the Hospital District of Helsinki and Uusimaa (UAK1014004, UAK1014005, TYH2018224), the Finnish Medical Foundation, and the Finnish Parkinson Foundation
Authors’ contributions
Conceptualization, AX, EE, PL, JG, LLM, and VL; data curation, EE, VTEA, MCH, KR, LP, PABP, LP, MGT, PA and FS; formal analysis, AX; funding acquisition, JAP, PB, LB, FS and VL; investigation, PB, LB, FS and VL; methodology, AX, LB and VL; visualization, AX; writing— original draft, AX; writing—review and editing, AX, EE, SG, PB, LB, PABP, VTEA, MCH, KR, MGT and FS. The authors read and approved the final manuscript.
Authors’ information
Center for Neurodegenerative Science, Van Andel Institute, Grand Rapids, MI, 49503 USA
Aoji Xie, Elizabeth Ensink, Peipei Li, Juozas Gordevičius, Lee L. Marshall, Sonia George, Patrik Brundin, Lena Brundin and Viviane Labrie
Center for Epigenetics, Van Andel Institute, Grand Rapids, MI, 49503 USA
J. Andrew Pospisilik
Department of Neurology, Helsinki University Hospital, and Clinicum, University of Helsinki, Haartmaninkatu 4, 00290 Helsinki, Finland
Velma T. E. Aho and Pedro A. B. Pereira and Filip Scheperjans
Institute of Biotechnology, DNA Sequencing and Genomics Laboratory, University of Helsinki, Viikinkaari 5D, 00014 Helsinki, Finland
Velma T. E. Aho, Pedro A. B. Pereira, Lars Paulin, Petri Auvinen
Nell Hodgson Woodruff School of Nursing, Emory University, 1520 Clifton Rd, Atlanta, GA 30322, USA.
Madelyn C. Houser
Department of Physiology, Emory University School of Medicine, 615 Michael St, Atlanta, GA 30322, USA.
Madelyn C. Houser and Malú G. Tansey
Faculty of Chemistry, Biotechnology and Food Science (KBM), Norwegian University of Life Sciences, 1433 Oslo, Ås, Norway
Knut Rudi
Department of Neuroscience and Neurology, Center for Translational Research in Neurodegenerative Disease, University of Florida College of Medicine, 1149 Newell Dr., Gainesville, FL 32611, USA.
Malú G. Tansey
Division of Psychiatry and Behavioral Medicine, College of Human Medicine, Michigan State University, Grand Rapids, MI, 49503 USA
Patrik Brundin, Lena Brundin and Viviane Labrie
Supplementary Files
Supplementary Figures
Acknowledgements
We dedicate this paper in memorial of our research adviser, mentor, and friend, Viviane Labrie. We thank the Van Andel Institute Genomics and Bioinformatics and Biostatistics Cores.
Footnotes
↵* Filip Scheperjans and Viviane Labrie should be considered joint senior author.