5. Discussion
NDDs are multifactorial, complex, diseases in which genetic factors do not fully explain disease onset and progression. There is mounting evidence that environmental factors and epigenetics contribute to NDD pathogenesis. Environmental effects on gene expression, however, are mediated at least in part by various epigenetic mechanisms. DNA methylation is now recognized as a reliable biomarker in several diseases, including cancer, neurological disorders, and autoimmune disorders
[48][49]. We previously showed that global DNA methylation is reduced in neurodegenerative and cerebrovascular diseases
[35]. In the current study, we examined whether those findings could be replicated in a different, larger cohort of healthy and NDD subjects, that further included patients with other NDDs such as Huntington’s disease and multiple sclerosis.
The most common forms of senile dementia observed in the elderly are AD, VaD, and mixed dementia; together, they represent a continuum of pathologies with considerable overlap in terms of prevalence with age, symptomatology, etiology, risk factors, and comorbidity
[50]. In this study, the NDD-D group, therefore, included patients with AD, VaD, and mixed dementia. In our previous study, we found no correlation between 5mC levels and psychometric parameters (Mini-Mental State Examination, MMSE) in healthy subjects nor in any patients in the NDD (PD, AD, and VaD) group; the only significant positive correlation was between age and 5mC levels in patients with PD (
p = 0.0385)
[35]. We did not collect Unified Parkinson’s Disease Rating Scale (UPDRS) data scores from patients in the present study. We did, however, use the MMSE for screening all patients for cognitive impairment. We found no correlation between 5mC, SIRT activity and BDNF mRNA levels versus MMSE values in healthy subjects nor in patients in the NDD group (data not shown). Consistent with our previous data, however
[35], global 5mC levels decreased substantially more in the NDD group than in the non-NDD group; this reduction in global 5mC levels was also observed in other NDDs such as Huntington’s disease and multiple sclerosis. Among the different NDDs, 5mC levels were slightly lower in NDD-D patients than in the NDD-PD and NDD-O groups, but these differences were not statistically significant.
We previously detected a slightly significant (
p = 0.0385) correlation between DNA methylation and age in patients with PD, but not in subjects with AD or VaD
[35]. In the current study, however, there was no correlation between age and DNA methylation in healthy and patients with NDDs. Global DNA methylation changes with age, and lower 5mC levels are found in brain and blood samples from animal models and patients with NDDs
[51][52][53][54][55][56][57][58][59][60]. In our previous study, the NDD (AD, PD, VaD) cohort included 101 patients; in terms of the age-range, 100% of NDD patients were older than 60 years old, with a positive correlation between age and 5mC levels found in PD patients only
[35]. In the current study, the lower number of patients with NDDs (
n = 35) and an age-range where 80% of NDD patients were older than 60 years old, led to no correlation between 5mC levels and age. In our study, since the majority of samples were obtained from individuals older than 60 years, patient age could therefore explain the disparity between our findings and those from other authors
[54][55]. The expression of DNMTs also decreases with age
[61]; DNMT1, DNMT3a, and DNMT3a2 levels are reduced in the frontal cortex and hippocampus in older human and mouse brains
[62]. DNMTs are closely linked to memory and cognitive functions
[63] and DNMT activity is required for the formation of associative memory and induction of long-term potentiation
[64][65]. The loss of DNMT activity during certain periods of development significantly impacts cognitive function
[66], suggesting that DNA methylation is important in regulating age-associated cognition. In the AD post mortem brain, DNMT1 expression and global 5mC and 5hmC levels are reduced within neurons in the entorhinal cortex layer II and hippocampus
[57][58][62].
DNMT3a expression was furthermore reduced in buffy coat samples from patients with dementia
[35][62]. In the present study, we confirmed that
DNMT3a expression was lower in the NDD-D group than in healthy patients. Several studies, however, did not find significant differences between healthy and AD brain samples
[60][62][67], or increased global DNA methylation levels in different regions of the brain in patients with AD
[60]. These differences may be due to the analyses of different brain regions (e.g., whole brain, hippocampus, entorhinal cortex) in separate studies, or to the heterogeneity of pathological diagnoses in analyzed samples, since phenotypic heterogeneity in AD may influence DNA methylation levels. Those authors conducted experiments on serum samples but also on leukocyte samples; this sampling heterogeneity may further explain the variability in their data. The number of samples, in several cases, did not yield conclusive data
[60][62][67]. Depending on the brain region, 5mC and 5hmC expression differ; both epigenetic marks are lower in astrocytes from patients with AD than healthy subjects
[67][68]. In the brain of late-onset AD patients with Braak stage IV-VI pathology, obtained post mortem, there are no differences in 5mC or 5hmC levels in AD-resistant calretinin interneurons or microglia, nor any differences near β-amyloid plaque regions of interest, nor in plaque-free zones
[67][68]. There were, however, high 5mC and 5hmC levels in neurofibrillary tangles.
DNA methylation-based age predictors are referred to as “epigenetic clocks”
[69], with compelling evidence linking epigenetic age acceleration to common diseases
[54]. In the present study, we did not find any correlation between 5mC levels and age in buffy coat samples from patients with NDDs, nor in subjects with no NDDs; the patient median age was 53 years, ranging from 20 to 86 years old. There was also no correlation between gender or APOE genotype and 5mC expression. However, global methylation levels increase in AD patients harboring the
APOE4 genotype
[27][52]. The same authors report higher 5mC levels in whole blood from AD subjects, and a correlation between global methylation levels and psychometric parameters
[27][52]. In those studies, global methylation was measured with a chemiluminescence substrate in whole blood samples, which contain different cell types with different methylation profiles
[70]. In our study, however, we used an ELISA-like colorimetric assay to measure 5mC levels in buffy coat samples; this, along with the method of methylation quantification, could explain the disparities between the two studies.
Chromatin remodeling and histone post-translational modifications play important roles in NDDs. HDACs participate in transcriptional repression, leading to the generation of a compact chromatin structure. SIRT expression changes with aging and age-related NDDs
[4][71][72]. Here, SIRTs promote lifespan and healthy aging by delaying the onset of neurodegenerative processes, and are new targets for treating neurodegenerative disorders
[30][32][33][36]. Modulation of SIRT1 levels and/or activity is beneficial in various models of AD
[36]; SIRT1 protects against β-amyloid plaque formation and ameliorates learning and memory deficits in animal models of AD
[72]. SIRT1 deacetylates and reduces the levels of pathogenic p-tau proteins; SIRT1 silencing increases tau levels
[36]. SIRT1 also regulates key PD-linked processes such as autophagy, apoptosis, mitochondrial dysfunction, oxidative stress and neuroinflammation
[29]. Furthermore, SIRT1 overexpression blocks α-synuclein aggregation in in vivo and in vitro models of PD
[73]. These findings are consistent with our present data, which show reduced SIRT1 expression in NDD-D and NDD-PD patients. SIRT2 is a highly conserved lysine deacetylase involved in aging, energy production and lifespan extension. SIRT2 levels increase with age and SIRT2 mediates processes involved in PD pathogenesis, including α-synuclein aggregation, microtubule dysfunction, oxidative stress, inflammation and autophagy
[74]. High levels of SIRT2 are found in AD, PD and other neurodegenerative disorders, suggesting that it may therefore promote neurodegeneration
[15]. SIRT2 may cause dopaminergic neuronal death
[74]; in in vitro and in vivo models of PD, pharmacologic or genetic inhibition of SIRT2 protects against α-synuclein toxicity
[36][75].
SIRT2 variants influence biochemical, hematological, metabolic and cardiovascular phenotypes, and modestly affect pharmacoepigenetic outcome in AD
[15]. However, SIRT2 may also be protective against neuronal injury
[74]. In our study, we observed a small, non-significant reduction in
SIRT2 mRNA in samples from the NDD-D group; SIRT2 expression in patients from the NDD-PD group were unchanged. The present data suggest that SIRT1 expression is a better biomarker than SIRT2 for diagnosing patients with NDDs. Measurements of SIRT activity, however, consider global SIRT activity rather than just SIRT1 and SIRT2. Since reductions in SIRT activity were much higher than changes in SIRT1 expression, we cannot exclude the possibility that the expression of other SIRTs are also reduced in NDDs. To this, SIRT3 is implicated in the pathogenesis of AD, PD, amyotrophic lateral sclerosis, and Huntington’s disease
[76]. SIRT3 mRNA and protein levels are reduced in the cerebral cortex of patients with AD and in the cortex of APP/PS1 double transgenic mice
[76]. SIRT3-5 are active in mitochondria. In our study, we analyzed SIRT activity in nuclear protein extracts from buffy coat samples; given that SIRT3 is active in mitochondria, it may not be responsible for the decrease in SIRT activity. Furthermore, SIRT6 contributes to telomere maintenance, DNA repair, genome integrity, energy metabolism and inflammation, promotes longevity
[77][78], regulates tau stability and phosphorylation
[79], and is absent in patients with AD
[78]. SIRT7, the least characterized SIRT, may be functionally significant in neural pathways and diseases
[77]. Therefore, we cannot rule out the possibility that SIRT6 or SIRT7 may also be regulated in patients from the NDD-D and NDD-PD groups.
NRG1 signaling influences cognitive function and neuropathology in AD
[80]. NRG1 attenuates deficits in spatial memory in AD transgenic mice in the Morris water-maze task and ameliorates neuropathology
[80][81]. Concordantly, our data showed reduced
NRG1 mRNA levels in buffy coat samples obtained from patients diagnosed with various types of dementia. NRG1 further protects the mouse cerebellum against lipopolysaccharide-induced oxidative stress and neuroinflammation
[43]. In samples from NDD-PD patients, our study showed that NRG1 expression decreased, similar to published data
[42]; NRG1 is neuroprotective against 6-hydroxydopamine-induced toxicity in vivo
[82]. MAPT expression is low in brain samples of patients with AD
[43][44] and increases in PD
[45]. Our data showed reduced MAPT expression in buffy coat samples of NDD-D patients.
Changes in the levels and activity of neurotrophic factors such as BDNF occur in several types of NDDs, including AD and PD
[46][83][84][85][86][87][88]. BDNF levels are reduced in serum and brain samples from mouse models of tauopathy
[83][84][85]. Intracerebroventricular administration of an adeno-associated virus carrying the gene encoding BDNF into mice produces stable BDNF expression, restores BDNF levels, prevents neuronal loss, alleviates synaptic degeneration, and attenuates behavioral deficits
[89]. However, BDNF expression does not affect tau phosphorylation
[84]. Reduced
BDNF mRNA levels are found in the hippocampus and frontal cortex of patients with AD
[84][85], suggesting that BDNF depletion or deficiency may contribute to the cognitive deficits in these patients. Low
BDNF expression is also found in the plasma of patients with mild cognitive impairment (MCI) and AD
[85][89]; the serum from patients with AD show significantly lower BDNF levels than those with MCI, confirming a connection between BDNF and AD; however, detection of BDNF is only possible in late stages of the disease
[86]. Other types of dementia (frontotemporal, Lewy Body, or vascular) are associated with low BDNF levels, both in the systemic circulation and the central nervous system
[83][84][85]. In our study, BDNF expression was almost non-existent in buffy coat samples obtained from patients suffering from various types of dementia, including AD and VaD.
Among individuals with PD, several pre-clinical and clinical studies report alterations in BDNF expression, implicating this neurotrophin in PD pathogenesis
[87][89][90]. BDNF is crucial for dopaminergic neuron viability and maturation
[82][87]. BDNF deficiency in the substantia nigra pars compacta is associated with the loss of dopamine-containing neurons, and patients with PD exhibit lower
BDNF mRNA levels in the substantia nigra pars compacta than in healthy controls
[83][89]. Neurons with low BDNF levels may be highly vulnerable to injury
[88]. Inhibition of local BDNF production with an antisense oligonucleotide causes a significant loss of dopaminergic neurons in the rat substantia nigra pars compacta, showing that BDNF is important in neuronal survival
[89][90]. In the present study, BDNF expression was dramatically reduced in buffy coat samples obtained from patients with PD. While pharmacologic treatment with levodopa increases BDNF expression
[91], non-pharmacologic interventions such as cognitive rehabilitation speech therapy and physiotherapy may also positively affect BDNF levels
[92].
With respect to existing diagnostic tools/algorithms for neurodegenerative diseases, while pathological analysis is regarded the gold standard in a wide range of disorders, it cannot be used to diagnose NDDs prior to the patient’s death. Other methods, such as positron emission tomography (PET) scanning or novel biomarkers (genomics and proteomics), may provide solutions and are being included into revised and improved diagnostic criteria
[93]. The International Group of Alzheimer´s Precision Medicine Initiative, for example, was formed to assess the current state of the art for blood-based AD biomarkers. To date, 19 blood-based biomarkers have been chosen for further study towards the diagnosis of Alzheimer’s disease
[93]. Recently, Huang Y et al. created Epigenome-Wide Association Studies (EWAS) plus, a computational technique that employs a supervised machine learning strategy, to expand the coverage of multiple EWASs to the entire genome rather than only about 2% of all CpG sites in the genome
[94].
Blood DNA analysis is a non-invasive and inexpensive method for liquid biopsies, with diagnostic potential. Finding new non-invasive biomarkers for NDD diagnosis would be beneficial in treating these patients. Methylation levels in brain and blood samples from patients with PD are concordant
[53]. In our current study, the area under the ROC curve (AUC) from the global DNA methylation assay in buffy coat samples from patients with NDDs was 0.66, ranking as “sufficient”
[95]; the Youden index was 0.28 (40% specificity and 87.5% sensitivity). As higher AUC values correlate to better biomarker diagnostic strength, our data revealed that SIRT activity and
BDNF expression are more reliable biomarkers; both had AUC values > 0.8, with a higher Youden index J. For patients with NDD-D, we calculated AUC values of 1.00 for SIRT activity (Youden index 1.00; 100% specificity, 100% sensitivity) and 0.95 for
BDNF expression (Youden index 1.00; 100% specificity, 100% sensitivity). For patients with NDD-PD, we calculated AUC values of 0.84 for SIRT activity (Youden index 0.66; 81.25% specificity, 85% sensitivity) and 0.95 for
BDNF expression (Youden index 0.66; 100% specificity and 60% sensitivity). Together, these findings show that global DNA methylation represents the biomarker with less diagnostic power than SIRT activity and
BDNF expression for diagnosing and monitoring disease activity and treatment intervention in patients with dementia than in individuals with PD. Nonetheless, since DNA methylation levels, SIRT activity and
BDNF expression all significantly decline in patients with dementia or PD, analyzing these three epibiomarkers may be useful in the diagnosis of NDDs. We propose combining the three markers to increase the efficacy of NDD diagnosis. Epigenetic modifications are reversible, and measuring DNA methylation levels, SIRT activity and BDNF expression may help clinicians monitor patient treatment responses.