In this initial study the main goal was to see whether we could identify a set of DML that could distinguish between individuals with Parkinson’s Disease who had high or low levels of two PD associated factors, i.e., plantation work years and detectable OGC pesticide exposures, both previously identified as significant risk factors for PD in the same longitudinal cohort of Japanese-American men [15] [25]. Our results indicated that in both the brain and blood tissue specimens, there were significant differences in DNA methylation related to plantation work and OGC exposures. To examine whether these DML stratified PD cases based on exposure conditions in brain and blood, we used PCA and hierarchical clustering analyses. The results indicated that cases with high and low exposure levels can be distinguished based on brain and blood-specific DNA methylation profiles.
Many of the genes for the DML identified in this study participate in key mechanisms involved in neurodegeneration and PD, i.e., apoptosis, autophagy, mitochondrial dysfunction and mitophagy, oxidative damage and repair, ubiquitin/proteasome system dysfunction, neuroinflammation, axonal development and degeneration, as well as epigenetic processes. The Ingenuity Pathways Analyses revealed that the top affected function is in neurodevelopment, and the IPA® ‘neurological disorders’ pathway contains many of these DML. In particular, we identified DML in PD susceptibility loci, such as those in Parkin RBR E3 ubiquitin protein ligase (PARK2) and MAPT that were differentially methylated in the comparisons of high and low exposure groups. This is not unexpected given their importance in key pathways related to the hypothesized mechanisms through which exposure to agrichemicals are believed to cause neurodegeneration and death of dopaminergic neurons [26] [27] [28], i.e., via neuroinflammation, mitochondrial dysfunction, apoptosis, and ubiquitination/proteasome dysfunction. Below we describe some of these key mechanisms and genes with DML in our study that are related to them.
Mitochondria Dysfunction and Neuroinflammation
Many pesticides affect protein complexes in the mitochondrial electron transport chain, and experimental animal models for PD have been developed using compounds, such as 1-Methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), 6-Hydroxydopamine (6-OHDA), salsolinol, rotenone, and paraquat. Studies using these animal models have shown that promotion of mitochondrial dysfunction occurs leading to subsequent death of dopaminergic neurons; and all of these neurotoxic models exhibited involvement of activated glial cells [29] [30] [31]. For example, it has been shown for MPTP, that a single exposure can induce prolonged and vast microglia activation and increased levels of pro-inflammatory cytokines, Tumor Necrosis Factor- α (TNF-α), and Interleukins IL-1α, IL-1β, and IL-6 [32] [33]. These cytokines lead to up-regulation of Reactive Oxygen Species (ROS), Nitric Oxide (NO), and superoxide radicals, to form highly oxidizing peroxynitrite species and activation of Dihydronicotinamide-adenine dinucleotide phosphate (NADPH) oxidase in DA neurons that promotes further oxidative damage and their eventual death through activation of cell death processes in these oxidative stressed neurons (see [34] [90] for review). The brain and in particular the substantia nigra is highly sensitive to oxidative stress due to its high oxygen consumption, low antioxidant defenses, high number of oxidizable species, e.g., polyunsaturated fatty acids, iron, and dopaminergic neurons, and its pro-oxidative state relative to other brain regions [35]. The brain tissues used in our study was primarily of glial origin, so the results from these animal models implicating activated glia in the neuropathological processes in PD, mirrors our results that focused on OGC and Plantation Work exposures in PD cases.
Hence mitochondria are central players in apoptotic pathways [36] [37] and are key to the survival of neurons, and are featured in our study in several concordant DML shown in Tables 6 and 7. Key DML identified in our analyses related to mitochondrial dysfunction were MAPT, MAP1B, SGK1, KCCN3, PACS2, DNAJC15, C1QTNF7, BCL2, CORO7, and TMCO3.
MAPT has been linked to PD and epigenetic studies on PD subjects have reported differences in DNA methylation at the MAPT gene [4]. MAP1B, as MAPT, is a microtubule associated protein involved in microtubule assembly. MAPT’s subunit LC1 has microtubule stabilizing activity and it appears to inhibit Leucine Rich Repeat Kinase 2 (LRRK2) activity, interestingly, mutations in LRRK2 are the most common cause of autosomal dominant and sporadic PD [38]. SGK1 is also linked to MAPT and it codes for a serine/threonine protein kinase that plays an important role in cellular stress response [39]. It phosphorylates MAPT and mediates microtubule depolymerization and neurite formation [40]. Mutations in microtubule associated proteins that lead to hyperphosphorylation and aggregation of tau in brain are linked to a family of neurodegenerative disorders, tauopathies, which leads to destabilization of tau-microtubule interactions leading to instability, axon transport defects, mitochondrial dysfunction, neuroinflammation, and ultimately neuronal death. Mitochondrial dysfunction is a major source of ROS, though more recently microglial cells have also been identified as ROS generators in tauopathies and other neurodegenerative diseases such as PD [41].
Neuroinflammation is closely associated with neuronal degeneration and cell death through biological mechanisms, such as elevated oxidative stress and glial (astrocyte and microglia) cell activation [42] [43]. A DML in this neuroinflammation pathway was found for MPPED1 which is indirectly connected to SPP1, a PD gene [44] known to be differentially expressed in PD cerebrospinal fluid and brain tissue. SPP1 (OPN), osteopontin, is a glycosylated phosphoprotein expressed in neuronal cell bodies. Osteopontin expression increases after neuronal damage, employing the role of glial cell attractant in this neurodegenerative process [45]. In neurodegenerative disorders, it can be toxic to neurons and cause cell death in some instances, but is neuroprotective in others [46]. A SNP of the osteopontin gene was shown to be associated with Lewy Body Disease [47]. SPP1’s ties to PD is due to its anti-apoptotic and anti-inflammatory properties and its upregulation in activated microglia which are responding to neuroinflammatory signals [46] [48].
KCNN3, a potassium intermediate small conductance calcium-activated channel (SK) protein, is an integral membrane protein, and it is thought to regulate neuronal excitability by contributing to the slow component of synaptic AHP [49, 50]. A rare intronic SNP for this gene, rs116286121, was found in meta analyses in the PDGENE database [51] to be a significant susceptibility gene for PD with an OR of 1.47 [44]. SK channels have been linked to mitochondrial dysfunction caused by agrichemicals, since it has been shown that activation of SK potassium channels prevents rotenone-induced neuronal cell death and neuronal network degradation, by inhibiting mitochondrial complex I activity [52] [53]. KCCN3 modulates electrophysiological properties of the dopaminergic cells of the substantia nigra by regulating the frequency and precision of pacemaker spiking, whereby dysregulation could ultimately lead to altered cell survival signaling pathways [54] [52] [55].
PACS2 is a multifunctional sorting protein that controls the endoplasmic reticulum (ER)-mitochondria communication, including the apposition of mitochondria with the ER and ER homeostasis [56] [57]. Both TMCO3 and PACS2 are involved in communication between the ER and mitochondria [58] through mitochondria-associated membrane (MAM), and hence are important in maintaining normal mitochondrial function. TMCO3 is a Na(+)/H(+) antiporter, i.e., it takes part in the catalysis of the transfer of Na(+)/H(+) from one side of a membrane to the other and is expressed in brain tissues. Most relevant to our exposures’ neurotoxic mechanisms, it Interacts with STX17, which has been shown to be involved in autophagy [59] [60] via a subdomain of the endoplasmic reticulum (ER), called MAM, which forms contacts with mitochondria and determines mitochondria metabolism via the transfer of lipids and Ca2+ signals between the ER and mitochondria [58] [61].
BCL2 suppresses apoptosis in a variety of cell systems, including neural cells, by controlling mitochondrial membrane permeability and is a major regulator of neural plasticity and cellular resilience [62] [63]. BCL2 expression levels were found significantly decreased in brain samples of rotenone treated mice and also in peripheral blood mononuclear cells (PBMC) when comparing PD cases to controls. Both BCL2 and C1QTNF7 play important roles in mitochondrial membrane permeability, apoptosis and cell death. As a member of the complement family of proteins, C1QTNF7 plays a prominent role as a mediator of inflammation in the removal of immune complexes and apoptotic cells, and is a regulator of the immune response, including T-cell responses [64] [65] [66]. C1QTNF7 is also part of the interaction network of the BCL-2 athanogene 3 (BAG3), which plays multiple roles in physiological and pathologic processes, including antiapoptotic activity, signal transduction, virus infection, cell adhesion and migration, and initiation of autophagy [67] [68].
CORO7, one of the Coronins, plays a role in Golgi complex morphology and function and interacts with clathrin adaptor AP-1 and is required for the maintenance of Golgi morphology and protein export from the Golgi. It is expressed in high levels in the healthy brain [69, 70]. It has also been shown to be part of the Mitochondrial Protein Import Superpath, which is important for mitochondrial biogenesis and function [71].
Ubiquitin Proteasome System (UPS) Dysfunction
Since many neurodegenerative diseases, such as Parkinson’s Disease, involve the accumulation of aberrant and damaged proteins, e.g., a-synuclein and Lewy bodies, and since the Ubiquitin Proteasome System (UPS) is crucial for the degradation of these proteins and hence maintenance of protein homeostasis and normal cell function, reactive gliosis which affects the efficiency of the Ubiquitin Proteasome System (UPS) [72], by promoting proteasome inhibition and neuroinflammation [73], plays a critical role in these diseases. The UPS is a major protein complex involved in the degradation of oxidized proteins, which have been associated with aging and neurodegenerative diseases and is disrupted by organochlorines [74]. It has been shown that mutations in the parkin protein gene PARK2 are associated with a genetic form of familial PD. Parkin acts as a Ubiquitin ligase in association with proteasomal degradation, and mutations and post-translational modification of this protein causes loss of function of E3 ligase that leads to UPS impairment and the loss of the neuroprotective effects of parkin [75] [76]. Lewy bodies contain not only α-synuclein but also parkin and ubiquitin. Hence in PD, inhibition of the UPS system may contribute to glial dysfunction and subsequent neuronal dysfunction by loss of efficiency in degrading neurotoxic proteins such as α-synuclein (see [73] for review).
The key DML identified in this study linked to the UPS mechanism are DNAJC15 and PEX19. DNAJC15 forms a stable subcomplex with a component of the mitochondrial import motor and so participates in the import of proteins into mitochondria, and thus assists in the regulation of the mitochondrial respiratory chain. As a member of the family of J proteins, it also participates, in conjunction with Hsp70 chaperone proteins, in cellular processes, such as folding of proteins, prevention of protein aggregation, disaggregation of proteins and protein transport [40] [41] [42]. PEX19 acts as key component of peroxisomes [77] by acting as a chaperone for insertion of peroxisomal membrane proteins (PMPs). The loss of PEX19 results in the absence of detectable peroxisomal structures, destabilization of many integral PMPs, and the mis-localization of other PMPs to the mitochondrion [77].
In eukaryotic cells, mitochondria and peroxisomes are the main ROS contributors [78], [79]. Both are equipped with their own ROS scavenging repertoire of enzymes and hence are key organelles in maintaining cellular ROS homeostasis. Peroxisomes contain several antioxidant systems, which are important for ROS homeostasis, e.g., the β-oxidation pathway, which directly produces H2O2, which is detoxified by catalase activity and is of central importance for redox balance of the organelle [80] [81]. When peroxisomes are either damaged by excessive ROS production these are marked by ubiquitination, which in mammalian cells is a common signal that triggers autophagy, and pexophagy. Thus, peroxisomal (together with mitochondrial) activity in maintaining the cellular redox state is important for cell survival and health [81] [82] [83].
A DML was found also in AP2A2; which is involved in pathways resulting in importation of aberrant proteins and toxins into mitochondria and neurons. AP2A2 is a clathrin associated adaptor protein that is involved in clathrin-mediated endocytosis [84] found to be overexpressed in Frontal Temporal Lobular Dementia [85] [45] and to be important for internalization of alpha synuclein, toxins and neuropathogens into neurons and cells [86] [87].
Excitotoxicity
Several DML in our study are in genes related to excitotoxicity, neuronal damage and neurodegeneration leading to PD. For example AIG1, which is down-regulated in expression studies of PD brains (Nextbio Disease Atlas) [88]. This protein plays an important role in regulating neuron excitability, axonal growth, synaptogenesis and neuronal survival [89] [28]. Another DML is in GRIN2A; coding for an N-methyl-D-aspartate (NMDA) receptor involved in long-term potentiation, and thought to underlie certain kinds of memory and learning processes. NMDA receptors however play also a critical role in excitatory synaptic transmission and plasticity. Lastly, one of the top DML discovered in blood, related to the Plantation Work exposure, is in Ectonucleoside triphosphate diphosphohydrolase 8 (ENTPD8) gene, which codes for an extracellular enzyme that has phosphohydrolytic activity on ATP and consequent effects on P2-receptor signaling. NTPDase/CD39 ectoenzymes are distributed in the nervous system ubiquitously and they are directly involved in the control of P2 receptor function in nervous tissues [90 ].
Aberrant Neuronal Development
Two DML tied to neurodevelopmental pathways are associated with the Wingless-Type MMTV Integration Site Family, Member 16 (WNT16) and EPHA4. WNT16 is a member of the WNT gene family that is highly expressed in healthy brain [26] [23]. Wnt ligands modulate expression of target genes that regulate cell proliferation, differentiation, and migration during development of the nervous system [91] [24]. EPHA4 belongs to the ephrin receptor subfamily of the protein-tyrosine kinases. EPH and EPH-related receptors have been implicated in mediating developmental events, particularly in the nervous system [92]. During development the Eph/ephrin system plays a role in the spatial organization of different cell populations, axon guidance, formation of synaptic connections between neurons, and blood vessel remodeling [93] [94] [95]. Single nucleotide polymorphisms in various Eph receptors and ephrins have been implicated as modifiers in the pathogenesis of amyotrophic lateral sclerosis as well as Parkinson's disease [96].
Concordant DML as Biomarkers for Exposure-Related PD
The studies of Masliah et al [10] and Davies et al [97] indicate that use of concordant DML in brain and blood may be important for identification of biomarkers for PD, since these DML may represent the effects of previous environmental exposures that have been flushed from the body, and/or transgenerational effects where environmentally induced changes in epigenetics are passed through the germline [98]. Of the 6 concordant genes with exposure-related DMLs identified in our study (DML with p-values <0.001 and methylation change in the same direction) between brain and blood for plantation work, only one gene DNAJC15 had two concordant DML that were not directly SNP-related and one additional DML with a SNP at the CpG, while the other 5 concordant genes all had a DML that contains a common SNP at the CpG that can possibly explain the methylation differences. The value of SNP-related DMLs are however becoming increasingly significant, since the growing number of genome-wide epigenetic (EWAS, like ours here), and mQTL (methylation quantitative trait loci) studies have allowed researchers to combine these data with the GWAS studies to now assign possible function to the many significant SNP signals that have been found to increase risk to diseases. An example is the 5-hydroxytryptamine receptor 2A (5HT2A) genomic variant (T102C) that increases the methylation and decreases expression of the 5HT2A and is hypothesized to be involved in the expression of the schizophrenia phenotype [99]. The proportion of DML found to be related to cis-acting genetic effects vary by phenotype and could be quite high, for example a recent study by Chen et al [100] found that cis-acting genetic effects could possibly account for as much as 50% of the methylation-expression/phenotype correlation. Hence our 6 concordant DM loci with SNPs could still be significant biomarkers of exposure-related PD, exposure alone, or of populations susceptible to the toxic effects of the neurotoxin exposure, such as experienced by the Japanese male population working on the sugarcane and pineapple plantations in Hawaii.
Importance of Glial Origin of Methylation Signals
Although our brain temporal tissue used for this study was not based on a single cell type, cell composition analyses indicated our methylation signals were primarily glia-derived. Notably, the glial signature was apparent also among the concordant DML in blood as well. The importance of glia to PD pathology has recently become apparent, with the discovery of increased density of activated microglia and astroglia in PD brains and in the brains of genetic, infectious, and environmentally induced animal models of PD. Chronic presence of activated microglia and astroglia is one of the most common features of PD neuropathology and accompanying neurodegeneration [101] [102].
Study Limitations
While our results suggest DNA methylation may provide a biomarker of environmental exposures related to PD, there are limitations to our study. The key limitation of our study is the low number of PD cases analyzed in the high exposure groups, which was due to limited brain autopsy material availability and matched OGC pesticide exposure data. In the design of the study on plantation work and OGC pesticide exposures, a cases-only design was adopted, so as to minimize the identification of genes related to PD susceptibility and pathology only. Another limitation of our study design is the analysis of DNA methylation in bulk brain tissue and blood specimens without focusing on isolated cell types. Previous studies have looked at pesticides dieldrin and paraquat in cultured cells, and noted increased histone acetylation in treated cells [103, 104], however we only investigated a single epigenetic mechanism, DNA methylation, in this study. Moreover, we sampled from only one portion of the brain (temporal), and did not sample from the most affected areas for PD, nor did we compare our results to other relatively unaffected areas of PD brains. Given that our results from the DNA methylation cell type analysis in brain tissue revealed primarily a glial methylation, we assume the results in this study derive largely from this cell type. A further limitation is that the time of exposure for the pesticides in each case cannot be pinpointed to a particular time period in an individual’s life span, and therefore cannot be directly attributed solely to their plantation work history. However, a major strength of this study is that we examined matched blood and brain tissue specimens, taken from the same subjects.