Chapter 1 Flashcards
1
Q
How prevalent is PD (worldwide and in Canada)? How has the incidence changed over time?
A
Worldwide: >6M as of 2016; predicted >9M as of 2020; projected >12M by 2040
2.5-fold increase in incidence (new cases) between 1990-2016
Canada: affects 1 in 500
CA has highest age-standardized worldwide prevalence of PD (2016)
2
Q
When was PD first described, and what were its key features at the time?
A
1817, James Parkinson’s “An Essay on the Shaking Palsy” (6 case reports)
Resting tremor, flexed posture, festination (gait disturbance); progressive
3
Q
What are the 3 “classical” motor symptoms of PD?
A
Bradykinesia (slowness of movement), resting tremor, and rigidity
4
Q
List 3-4 motor and nonmotor symptoms associated with PD.
A
Motor: resting tremor, gait disturbance, bradykinesia, rigidity
Nonmotor: hyposomia (decreased sense of smell), constipation, RBD, depression, anxiety, cognitive decline
5
Q
Describe how possible, probable, and definite PD are diagnosed, according to standard criteria such as UK Brain Bank, Gelb, and MDS. What are the major updates between 90’s and 2010’s criteria?
A
UK Brain Bank (1988):
“Parkinsonism” = Bradykinesia and at least one of rigidity, resting tremor, postural instability
“Parkinson’s disease” = 3 or more of unilateral onset, resting tremor, asymmetrical response to L-dopa, L-dopa-induced chorea (uncontrolled movement), L-dopa response for 5+ years, progression for 10+ years
Calne (1992): “Clinically possible,” “clinically probable,” and “clinically definite” PD
Gelb (1999):
“Possible” = Resting tremor or bradykinesia, and either rigidity or asymmetric onset (duration <3 years); response to L-dopa or dopamine agonist
“Probable” = 3 of resting tremor, bradykinesia, rigidity, or asymmetric onset (duration >3 years); response to L-dopa or dopamine agonist
“Definite”: Upon autopsy, neuronal/glial death in SN, LBs in SN or locus ceruleus (noradrenergic nucleus)
MDS (2015):
“Parkinsonism” = bradykinesia with resting tremor and/or rigidity
“Clinically probable” = Lack of exclusion criteria; “red flags” can be counterbalanced
“Clinically established” = Lack of exclusion criteria and “red flags”; >2 supportive criteria (response to DA treatment, unilateral resting tremor, olfactory/cardiac/imaging test result)
6
Q
Give examples of research scenarios where better sensitivity or specificity would be desired in diagnosing PD.
A
Sensitivity (catching all possible cases): research into etiology, particularly early-stage
Specificity (correct diagnosis of PD): clinical trials, patient treatment
7
Q
Describe typical PD progression using Hoehn and Yahr stages.
A
Stage I (year 0-3): Unilateral impairment, most of function retained
Stage II (year 3-6): Bilaterial/midline impairment, balance retained
Stage III (year 6-7): Balance impairment begins (loss of righting reflex). Mild-moderate disability.
Stage IV (years 7-9): Disabling and severe PD
Stage V (years 9-14): Confined to bed/wheelchair
8
Q
Describe typical PD progression through the prodromal, early, and late stages of PD.
A
Prodromal: Pathology in brainstem, olfactory bulb, and peripheral NS; nonmotor symptoms (RBD, constipation, depression/anxiety, urinary dysfunction)
Early: Tremor, mild slowness; some memory problems
Manifest: Falls, sometimes dementia, L-dopa side effects
9
Q
List 4-5 clincial features used to classify subtypes of PD.
A
Age of onset, rate of progression, motor/nonmotor symptoms, response to L-dopa
10
Q
What are some problems associated with clinical subtyping in PD, and why is the field searching for more biological subtyping?
A
Symptoms can change over time, so individuals may resemble a different clinical subtype than they did before. Clinical subtypes also have no difference in pathology. Genetic subtypes are being identified now…more consistent, may be able to define mechanisms/treatments
11
Q
Name 2 hallmark characteristics of PD pathology.
A
LBs in brainstem, DAn loss in SN
12
Q
What is Braak staging/how does aSyn pathology progress over the course of PD?
A
Lewy bodies begin in olfactory bulb and move to brainstem, limbic system, then neocortex
13
Q
How do “brain-first” and “body-first” PD differ? Describe early indicators, pathological progression, post mortem pathology, and supporting evidence.
A
“Brain-first”: classic PD pathology of LBs in brainstem and DAn loss in SN. aSyn pathology progresses according to Braak staging. RBD is seen after onset of motor symptoms.
“Body-first”: LBs begin in PNS/ENS, travel through vagus nerve to brain. RBD is seen prior to onset of motor symptoms. Evidence: protective vagotomy; animal studies showing gut-brain spread of aSyn; LBs in GI nerves prior to PD diagnosis
14
Q
How does PD medication work, and what are some medications commonly used to treat PD?
A
Replaces DA lost from striatum.
Early stage: L-dopa (supply DA), DA agonist (activate DAR), COMT (increases DA bioavailability), MAO-B inhibitors (prevent mitochondrial breakdown of DA)
15
Q
When was a genetic connection to PD first discovered?
A
In 1997, A53T aSyn mutation (autosomal dominant, 85% penetrant) discovered in Greek and Italian families
16
Q
List 7 genes where mutations are strongly implicated in familial PD.
A
SNCA, LRRK2, GBA, DJ-1, VPS35, PRKN, PINK1
17
Q
What approaches were/are used to identify rare variants associated with PD?
A
Linkage studies, genome sequencing, exome sequencing
18
Q
What are some of the main challenges associated with identifying and replicating genes linked to familial PD?
A
Rare mutations, globally dispersed, high cost of sequencing
19
Q
Which 4 genes are implicated in PD through both rare variants/monogenic disease and common variants/sporadic disease?
A
SNCA, LRRK2, GBA, VPS13C
20
Q
Compare and contrast the symptoms and onset of monogenic and sporadic PD.
A
Monogenic: highly penetrant, earlier onset (30s to 50s), atypical symptoms.
Sporadic: later onset (60s to 80s), slower progression, typical symptoms/disease course (for instance, cognitive decline is more rare and observed late stage)
21
Q
List 4 ways in which SNCA variants are implicated in PD.
A
Rare point mutations in coding regions of the gene, variations in REP1 dinucleotide repeat length, SNPs, whole gene duplication/triplication
22
Q
Name 5 genes whose SNPs associated with sporadic PD risk in the first GWAS analyses conducted in 2009.
A
SNCA, MAPT, PARK16, LRRK2, BST1
23
Q
Summarize the results of the largest PD GWAS to date (2023). How many loci were implicated and what % of PD heritability do they explain?
A
2019 meta-analysis of 17 European ancestry PD GWAS. 90 loci implicated (including SNCA, GBA, LRRK2), explaining 22% of PD heritability. Enriched for lysosomal genes, genes expressed in brain; SNPs also associated with brain volume, educational attainment, and smoking.
24
Q
What are 4 major biological pathways implicated in genetic susceptibility for PD?
A
Lysosome, endosome, immune, mitochondrial
25
Name 4 reasons why there is thought to be "missing heritability" in sporadic PD not captured by current GWAS approaches.
1. Lack of power to detect rare variants with small effect sizes in GWAS
2. Limited ability of microarray platform and imputation to capture structural variants
3. GxG, GxE interaction effects may be present
4. Epigenetic contribution to heritability not fully understood
26
Describe two ways common genetic variation can be used to estimate sporadic PD risk.
1. Polygenic risk score for PD, based on 90 Nalls loci (top decile has 6-fold increased PD risk)
2. PD composite score based on genotype, asomia (loss of smell), age, sex, family history (90% specificity, 80% sensitivity)
27
Besides risk of developing PD, what other two aspects of PD can be estimated by GWAS?
Age of onset, progression (early stage, limited data so far)
28
Name 3 strategies for inferring causality from PD GWAS.
1. Mendelian randomization (estimates the probability that an exposure/trait has a causal relationship with PD phenotype due to influence of genetic variation on the exposure/trait)
2. Colocalization (estimates probability that two SNPs/traits affect PD at all, at the same locus, or at different loci)
3. Linkage score disequilibrium regression (tests shared genetic etiology between traits; for instance, PD and smoking, education, brain volume in Nalls 2019)
29
How can understanding the genetic etiology of PD inform precision medicine approaches? Provide two real-life examples.
Personalized targeting of the pathway that lead to PD in an individual. Examples: Increasing GCase production in GBA-mutant individuals, reducing SNCA expression in SNCA mutant individuals.
30
Name some limitations of our current understanding of the genetic architecture of PD. What knowledge gaps still exist?
Genetic diversity beyond European-ancestry populations
GxG, GxE interactions
Epigenetic-genetic interactions
Rare/structural variants
31
List 4 risk factors and 4 protective factors for PD.
Risk factors: pesticide exposure, TBI, dairy intake, melanoma
Protective factors: smoking, caffeine, physical activity, ibuprofen
32
Name 5 limitations/common pitfalls in studying and interpreting the epidemiology of PD. What are the limitations specifically with respect to studies of PD and pesticide exposure?
General epidemiological limitations:
1. Applies to only sPD
2. Uncertainty in diagnosis of early/mid-stage PD
3. Recall bias
4. "Normative" control populations?
5. Accuracy and completeness in measurement of exposures and other confounders
Pesticide exposure: Inconsistent timing, duration, and dose of exposures; self-reported rather than quantitative data
33
How was a link between pesticide exposure and PD first established?
In 1983, heroin users of a MPTP-contaminated supply developed parkinsonism. Similar structure to paraquat herbicide. Studies in the 1980s-2000s revealed increased PD prevalence in communities with pesticide and heavy metal-contaminated well water.
34
What are the mechanisms of action for MPTP, paraquat, and rotenone in causing parkinsonism?
All are mitochondrial toxins, impairing activity of mitochondrial complex I (first enzyme in respiratory chain).
MPTP: immediate cell loss in SN & sustained loss over time, without LBs (rodent & monkey studies)
paraquat: SN cell loss, microglial reactivity, upregulation of aSyn, Ub/protesome dysfunction, with LBs (rodent intraperitoneal injection)
rotenone: SN cell loss, upregulation of aSyn, Ub/protesome dysfunction, with LBs
35
How does organochlorine insecticide exposure relate to PD pathology?
Ochl concentration in postmortem brain correlates with LB pathology in caudate/putamen, and amount of DA
36
List 2-3 examples of gene-pesticide interactions in PD risk.
LRRK2 G2019S influences paraquat inflammatory response/microglial reactivity in mice
GSTP1 genotype and herbicide exposure interact to modify PD age of onset
ABCB1 genotype and Ochl exposure interact to modify PD risk
37
Describe how genetics and exposure influence PD risk and progression through the life course.
Baseline PD risk due to genetics.
Exposure to pollutants/toxicants in early or mid-life may initiate aSyn accumulation in olfactory system or GI tract, with response to exposure depending on genotype.
Further exposures and aging may exacerbate PD progression.
38
Summarize the hypothesized contributions of early life exposures, injuries, and/or infections to developing PD later in life.
"Two-hit" hypothesis: one exposure or injury early in life may increase inflammation, while a second exposure, injury, or effects of aging might exacerbate this and initiate neurodegeneration
39
What are the mechanisms of action for smoking and coffee drinking in protecting against PD?
Smoking activates nicotinic ACh receptors, which stimulate DA production and Ca2+ release, have antioxidant activity, and enhance neuronal excitability.
Coffee activates D2 receptors in striatum and blocks A2A receptors. May interact with genotype and estrogen (stronger protective effect dependent on genotype and in males). Also increases SCFA production in gut.
40
What clues can be used to ascertain whether PD occurring in families is driven more by genetics or environment/exposures
If genetically driven, age of onset will be earlier and similar across generations. If environmentally driven, time of onset cross-sectionally will be more similar within a generation, not across generations.
41
Describe 3-4 mechanisms behind brain injury and risk for PD.
Breakdown of BBB, increase in microglial activation/release of inflammatory cytokines, resulting in immediate and sustained cell loss. Mitochondrial disruption, increased glutamate release
42
Is alcohol consumption a risk factor or protective factor for PD?
Mild- to moderate-drinking is protective for PD because of increased urate with alcohol consumption (natural antioxidant, byproduct of purine metabolism). Heavy drinking increases risk for PD.
43
Which substances/behaviours act on PD risk by modifying urate levels?
Alcohol, fructose consumption, and exercise increase urate levels; dairy consumption decreases urate levels
44
Describe the evidence for the protective role of exercise in PD and the mechanisms of action.
In prospective studies, playing sports in college or level of daily physical activity in adulthood is associated with reduced PD risk. Exercise can also ameliorate motor symptoms in individuals with PD, and in mice, treadmill exercise reduces cortical aSyn aggregation.
Mechanisms: BDNF release, DA release, synaptic plasticity, stabilized antioxidant response, increased urate, upregulation of PGC1a (energy metabolism regulator)
45
How is "epigenetics" defined, and how has this definition changed over time (1950s-present)?
1957: Waddington's "epigenetic landscape" (effect of gene regulation on phenotype during development and differentiation)
1996: Riggs, "mitotically heritable" factors other than DNA sequence
2007: Bird, "structural adaptation" of chromosomal regions to register or perpetuate altered activity states
2017: Greally, "cellular reprogramming"
2019: NIH Epigenomics Roadmap, "heritable changes in gene activity and expression, and stable alterations in transcriptional potential"
46
List 5-10 examples of epigenetic marks.
DNAm, DNAhm, miRNA, lncRNA, histone variants, histone acetylation, histone methylation, histone Ub/SUMOylation, RNA modifications, adenine methylation...
47
How are epigenetic states (DNAm, DNAhm, histone PTMs) inherited mitotically and/or meiotically? What is the evidence for inheritance across generations?
Mitotic inheritance:
DNAm - DNMT1 recognizes hemi-methylated DNA and copies DNAm from parent strand to daughter strand.
DNAhm - Some mC is oxidized to hmC during S-phase (perhaps those not bound by MeCp2), so overall hmC level is maintained through mitotic divisions
Histone PTMs - repressive PTMs are mitotically inherited (parental histones recruit HMTs), while active PTMs are re-established after mitosis
Meiotic inheritance:
DNAm - all except imprinted DNAm is erased during gamete formation, again after fertilization. In agouti mice, incomplete DNAm erasure at IAP retrotransposon results in meiotic transmission of agouti phenotype (yellow fur, obesity, diabetes). In humans, evidence is more intergenerational than transgenerational...
DNAhm - increases in zygote via TET3 action
Histone PTMs - histone methylation thought to be erased in zygote and re-established during development, similar to DNAm...in amoebas and worms, transcriptional states are inherited via HMTs
48
What is polycredodism and how is it relevant to studying epigenetics?
A series of cell fate decisions in response to perturbation, that create a population of different cell comprising a tissue (cell type heterogeneity/variation in cell type proportion)
49
What is biological embedding and how is it relevant to studying epigenetics? Provide an example.
Effect of life experience on biology, which can impact later life health and wellbeing. Example: licking and grooming maternal behaviour in rats alters GR promoter DNAm in hippocampus, leading to higher GR expression in adulthood and impacting stress response.
50
List 5 examples of major challenges in interpreting epigenome-wide association studies.
1. Cohort selection (representative?)
2. Statistical approaches to high-dimensional data analysis
3. Accounting for technical artefacts
4. Accounting for genetic variation and cell type proportion
5. Reverse causation (transcription --> DNAm)
51
Where is the methyl group in 5mC found?
5th carbon in pyrimidine ring of cytosine
52
How many CpGs/CpHs are in the human genome, and how many are methylated?
~5% of cytosines are CpG (28M), 80% methylated
~95% of cytosines are CpH (~530M), 2-6% methylated
53
What is a CpG island, and where are CGIs found (in general, and when methylated)?
>500bp stretch of DNA with >50% CpGs. Evolutionarily conserved, found at gene promoters. TET is present to maintain low DNAm.
Methylated CGIs at imprinted genes, inactive Xchr, and genes with germ cell-specific expression.
54
How is DNAm deposited and removed?
DNMT1: Recognizes hemi-methylated DNA, deposits DNAm on daughter strand during replication.
DNMT3A/B: De novo DNAm, active during early development/after fertilization.
DNMT3L: enhances activity of other DNMTs (no catalytic domain)
TET: oxidation of 5mC --> 5hmC --> 5fC --> 5cac
TDG/BER: excision of 5fC/5cac and replacement w/C
55
Describe the function of DNAm depending on CpG density and its position in a gene.
CGIs/promoters: transcriptional repression (prevents transcription complex binding, recruits repressive proteins)
CpG shores: tissue-specific gene expression
CpG open seas/enhancers: tissue-specific/dynamic gene expression control (i.e., response to environment)
Gene bodies: may be initiated by transcription; splicing regulation; suppression of repeat transcription
56
What are the relative abundances of mCG, hmCG, mCH, and C in adult mouse cortex?
94% unmodified C
3% mCG
1% hmCG
1% mCH
5fC, 5cac are removed by DNA repair and are transient - low abundance...
57
Describe the tissue distribution of DNAhm.
Enriched in brain (17-30% of modC in adult mouse frontal cortex) and in ES cells
58
What are the functions of DNAhm generally and in the brain?
Keep promoters from accumulating mC (transcriptional activation); recruitment of splicing regulators to DNA; learning and memory, neuronal activity-related gene expression
59
How is DNAhm maintained during cell divisions?
mC --> hmC during S-phase (higher hmC on parental DNA strand)
60
How do DNAm and DNAhm change during development and aging?
Development: after fertilization, DNAm is erased and DNAhm deposited (TET3). After implantation, DNAm is re-established, forming tissue specific gene expression. Germline stem cells are passively/actively demethylated.
Aging: Passive/active loss of DNAm, gain of DNAhm (brain region-specific) due to reduced fidelity of DNA(h)m machinery. "Epigenetic drift": stochastic or environment-related.
- Predictable change in DNAm w/aging = clocks
61
What is the function and distribution of CpH methylation?
Found in ES cells and neurons (50% of adult mouse and human neuronal modC). Thought to be transcriptionally repressive through MeCP2 recruitment; depleted in bodies, upstream, downstream of transcribed genes. Transcription induces H3K36me3, which prevents DNMT3A from binding and catalyzing mCpH at gene bodies.
62
Describe how histones and DNA are arranged in a nucleosome.
Nucleosome contains 146bp of DNA wrapped around a histone octamer, with 2 each of H2A, H2B, H3, H4 (2 H2A-H2B dimers and 1 H3-H4 tetramer). H1 is the "linker" histone between nucleosomes.
63
How many nucleosomes are in a typical cell, and how do they compress DNA?
Histones have a positive charge, attracting negatively charged DNA. Each cell has ~30M nucleosomes, packing 2m of DNA into 5-10uM diameter
64
What is the "histone code", and why is it controversial? What are some functions of histone modifications beyond transcription regulation?
The "histone code" hypothesis postulates that transcription is regulated by a combination of different histone marks that are "read" by chromatin remodelers; the balance between recruiting and repelling other proteins and DNA tunes transcription.
Controversy: Reverse causation - histone modifications can be added during transcription or after transcription.
65
What types of post-translational modifications are deposited on histone tails, and what effects do these have on chromatin? List 3-5 examples.
Acetylation: negative charge, repels DNA
Methylation: promotes or represses transcription via recruitment of methyl-binding protein complexes
Ubiquitination: DNA replication, DNA repair, transcription regulation
Phosphorylation: DNA repair
SUMOylation: recruits HDACs (transcription regulation)
66
Describe the difference between euchromatin and heterochromatin, and give examples of histone marks found in each context.
Euchromatin: open, transcriptionally active. Marked by H3/H4ac, H3K4me1
Heterochromatin: closed, transcriptionally repressed. Marked by H3K27me3 (facultative), H3K9me3 (constitutive)
67
What are the six H3 modifications subject to profiling by IHEC, and where are they found?
H3K4me1 - enhancers and downstream of TSS
H3K4me3 - TSS of active/bivalent genes
H3K27me3 - TSS of repressed/bivalent genes
H3K36me3 - gene bodies of active genes
H3K27ac - enhancers and TSS of active genes
H3K9me3 - constitutive heterochromatin
68
How stable are histone modifications, and how are they inherited mitotically and/or meiotically?
Less stable than DNAm; more involved in dynamic transcription regulation.
During mitosis, repressive modifications are copied from parent strand to daughter strand (new histones) via HMT enzymes. Active modifications are not inherited mitotically, but re-established after mitosis via "seeding" of H4K16ac and recruitment of further HATs.
Most histone methylation is erased during fertilization/gametogenesis (histones are removed from sperm) and re-established in the same manner as DNAm.
69
What is a typical consequence of deregulation of histone methylase/demethylase enzymes?
Aberrant epigenetic/transcriptional profiles are associated with tumorigenesis and metastasis
70
Where is H3K4me1 found, how is it deposited/removed, and what is its function?
Found at enhancers of expressed or poised genes, and downstream of the TSS.
Deposited by MLL-family HMTs, removed by KDM family of histone demethylases.
"Fine-tunes" gene transcription at enhancers via recruitment of chromatin remodelers.
71
Where is H3K4me3 found, how is it deposited/removed, and what is its function?
At TSS of actively transcribed genes.
Deposited by MLL-family HMTs during RNAPII initiation, removed by KDM-family histone demethylases.
Recruits transcription factors.
72
Where is H3K27ac found, how is it deposited/removed, and what is its function?
At enhancers and TSS of actively transcribed genes.
Deposited by p300/CREB HATs, removed by HDACs.
Is sufficient for transcription initiation and deposition of H3K4me3.
73
Describe the interplay between histone modifications and DNA modifications.
H3K4m inhibits DNMT3L, and H3K4me3 recruits TET and transcription initiation complex; H3K4m is correlated with DNAhm, anti-correlated with DNAm.
DNAm and DNAhm both recruit HDACs to chromatin through MeCP2/MBDs, and are anti-correlated with H3K27ac.
74
Describe several low- and high-dimensional approaches to profiling histone PTMs in cells/tissues, with advantages and drawbacks of each (focus on ChIP-seq and its technical considerations).
Western blot: Cheap and convenient. Bulk measure rather than locus-specific.
Mass spectrometry: highly sensitive, but bulk measure; tells you the overall abundance of a variety of modifications rather than their location.
ChIP-seq: provides locus-specific information. Enrichment-based, semi-quantitative. Can be costly and depends on sequencing depth (for instance, broad marks require higher depth).
ChIP-seq and WB are highly sensitive to the antibody used, and appropriate controls must also be assayed (e.g. unmodified histone) to obtain an accurate measurement of the modified histone.
75
Provide 3 examples of small noncoding RNAs, describing their structure, function, and any known relevance to PD.
1. miRNA: 22 nt long. Interferes with translation by binding to or degrading mRNA (RISC complex). miRNA dysregulation is associated with oxidative stress, aSyn aggregation, and cell death in PD brain.
2. piRNA: 26-31 nt long. expressed in germ cells; TE silencing. Deregulated in PD brain (de-repression of SINEs/LINEs).
3. tRNA: 15-50bp long. Adapter for translation. Mutations found in PD and ALS.
3.
76
Describe the structure and function of lncRNAs.
>200bp. Transcription regulation: can bind to DNA/RNA, recruit epigenetic and transcriptional modifiers. Can also regulate splicing and can be precursors to or neutralize miRNAs. Typically highly expressed in brain.
77
Describe the interplay between lncRNAs and DNA modifications in gene regulation at the H19/IGF2 ICR.
H19: maternally expressed lncRNA (growth restriction of placenta)
IGF2: paternally expressed insulin growth factor (growth-promoting)
Typically, IGF2 is silenced maternally and H19 is silenced paternally (via DNAm at the ICR upstream of H19). CTCF binding prevents distal enhancer activity for IGF2 transcription, stimulating H19 transcription and silencing IGF2 transcription. When ICR is methylated, CTCF cannot bind; H19 is silenced and IGF2 is transcribed.
78
Describe the interplay between lncRNAs, DNA modifications, and histone modifications in gene regulation during X chromosome inactivation.
Xist: lncRNA expressed from the inactive Xchr. Binds to the inactive Xchr, recruiting PRC2, which catalyzes H3K27me3 (producing facultative heterochromatin).
Tsix: lncRNA expressed from the active Xchr, which recruits DNMT to the Xist promoter, silencing it by DNAm.
79
What is known about the role of lncRNAs in PD?
Some lncRNAs were reported to be DE in PD prior to clinical onset (postmortem Braak stage in brainstem only) and may be biomarkers. UCLH1 antisense RNA is downregulated in PD, reducing level of UCLH1 and leading to Ub-protease dysfunction.
80
Define broad-sense and narrow-sense heritability.
Broad-sense: proportion of phenotypic variation attributed to total genetic effects
Narrow-sense: proportion of phenotypic variation attributed to additive genetic effects
81
How much of variation in DNAm on average is due to genotype?
20-80%, varying with tissue and relatedness between individuals.
13% monocytes
20% heritability in MZ twins
20-60% in HapMap LCLs
65% of tested CpGs in at least one of adipose, naive T cells, whole blood, and muscle
up to 80% within a family
82
What is the heritability of DNAm in MZ twins?
19% (narrow-sense: additive heritability)
83
Describe the contribution of cis- and trans-mQTLs to DNAm variation. How much does each contribute, and what are the mechanisms?
cis-mQTLs: 93% of all mQTLs are within 1 Mb of their target CpG, with most being within 50kb. In areas with repressive histone modifications in fetal brain, which become open during adulthood (flexibly regulated): insulators, enhancers. Tend to be either tissue-specific or consistent across tissues/life stages.
SNP influences DNAm through TF binding: either presence/absence of TF is correlated with DNAm state, or TF recruits DNMT/TET enzymes.
trans-mQTLs: ~7% of mQTLs, often in telomeres.
SNP may affect DNAm indirectly through an eQTL (expression modifies DNAm), cis-mQTL (cis-mQTL modifies activity of a protein which modifies distal DNAm), by recruiting TFs that loop DNA, or by altering the coding region of a TF which binds distal sequences.
84
How are mQTLs detected?
Typically by linear regression (DNAm ~ SNP genotype + covariates). Can also be detected by non-parametric methods, or machine learning.
85
How concordant are mQTLs across age, tissue, and ethnicity?
Tissue: Between brain and blood, 18-30% concordant (higher than expected by chance). Between blood and saliva, up to 73% concordant.
Lifecourse: some childhood mQTLs are retained into adulthood (40-50%), while some are lost due to stochasticity and environmental influences. Majority of adult mQTLs are present at earlier ages.
Ethnicity: 21-70% shared between White and African infants (27K; depending on which group was baseline)
86
How can Mendelian randomization (MR) and reverse MR be used to integrate mQTLs into EWAS?
If a CpG has an associated mQTL, genetic data can be used to infer whether the relationship between CpG methylation and a trait is causal.
Reverse MR: infer whether the trait causes CpG methylation.
87
Generally speaking, is genetic variation in complex traits thought to be mediated by DNAm? Why or why not?
Although many GWAS loci are enriched for mQTLs, the majority of genetic variation influencing complex traits is not known to be associated with DNAm. This might partly be due to insufficient characterization of GxG, GxE effects on phenotype...
88
How do CNVs directly/indirectly affect DNAm, and how does this apply to SNCA?
The CNV region might encode epigenetic and transcriptional modifiers, or cell signaling proteins, which could affect gene regulation and cellular phenotype. Altered dosage of these affects DNAm.
For example, SNCA multiplications lead to overexpression of SNCA protein, which might impact DNAm through interactions with HATs, DNA, or cell signaling pathways.
89
What is known about the relative contributions of SNP genotype, environment, and their additive/interaction effects to DNAm?
Genotype and environment jointly influence DNAm in neonatal cord blood and cord tissue studies.
Teh: 25% G, 75% GxE top-ranked for VMRs (but inadequate MTC, small magnitude)
Czamara: ~30% each G, GxE, G+E; very few E
90
Why is it not straightforward to interpret environmental/exposure effects on the epigenome?
Exposures may impact the epigenome differently depending on life stage, tissue/cell type, genotype, timing and duration of exposure
91
Describe how diet influenced the epigenome in the agouti mouse model.
Coat colour and obesity phenotype is controlled by methylation of an IAP retrotransposon. Eating a methyl donor-rich diet during pregnancy maintains IAP DNAm, so that offspring have brown coat and normal weight.
92
Summarize 3-4 mechanisms by which smoking alters DNA methylation.
1. Activates nicotinic ACh receptors, resulting in altered downstream signaling cascades
2. Generates DNA damage and recruits DNMTs to DSB repair sites
3. Alters TF expression and activity
4. Hypoxia --> upregulation of SAM metabolism
93
Relate the concepts of DNAm changes with age and DNAm changes associated with environments/exposures.
Age-associated DNAm changes can be predictable (clocks) or stochastic ("epigenetic drift"). The stochastic changes are a combination of random chance/infidelity of DNAm machinery and of exposures/experiences. This is why the epigenome of MZ twins becomes more discordant with age.
94
Provide two examples of how lifestyle changes reversed abberant DNAm patterns.
DNAm changes in glucose uptake genes associated with obesity are reversed after bariatric surgery.
DNAm in skeletal muscle, adipose tissue, and blood is altered with exercise.
95
Summarize what is known about the existence of hmQTLs and speculate on mechanisms.
A small amount of hmQTLs (23) were detected in fetal brain w/450K array. SNPs may influence TET binding or activity, or might have same mechanisms of action as mQTLs (i.e., since mC and hmC are related, mQTLs which alter mC level may also impact hmC level. Or may influence MeCP2 binding, which has affinity for both mC and hmC)
96
What are the main mechanisms by which brain DNAhm levels are influenced by environment?
Neuronal activity-related changes in gene expression, plasticity, learning, and memory associate with altered DNAhm. Exercise, caloric restriction, and chronic stress alter DNAhm levels in mice.
97
Why is CpH methylation proposed to be more environmentally sensitive than CpG methylation?
DNMT enzymes recognize CpG, but not CpH, sites. Thus, CpH methylation is more likely to be lost with mitotic divisions. CpHs are also more sparsely distributed, so repression is less stable.
98
Describe the 3 main chemical steps in bisulfite conversion, and what happens to an unmethylated, methylated, or hydroxymethylated cytosine during each step.
1. Sulfonation (add bisulfite ion)
- SO3- (bisulfite) added to nucleotide
2. Deamination (activity of bisulfite ion)
- NH3 (ammonia) group lost from C
- methylated, hydroxymethylated C retain NH3 group (reaction is extremely slow)
3. Desulfonation (removal of bisulfite ion)
- removal of SO3- (bisulfite) from nucleotide
- C --> uracil
- methylated, hydroxymethylated C become methyl-sulfonate and are read as C
99
What is the chemical agent used to oxidize DNA, and what does it do to 5hmC? How is 5hmC affected when bisulfite conversion is subsequently applied?
Potassium perruthenate (KO4Ru) oxidizes 5hmC --> 5fC. 5fC is deaminated during bisulfite conversion and read as uracil.
100
Describe the 3 main chemical steps in TAB conversion, and what happens to an unmethylated, methylated, or hydroxymethylated cytosine during each step.
1. Glucosylation (beta-glucosyltransferase)
- 5hmC --> 5gmC (beta-glucosyl-5-hydroxymethycytosine)
2. Mouse recombinant Tet1 oxidation
- 5mC --> 5cac
- 5gmC is resistant
3. Bisulfite conversion
- 5mC --> 5cac --> U
- C --> U
- 5gmC is resistant (read as C in PCR)
101
List 3-4 alternatives to bisulfite-based detection of cytosine modifications and discuss their benefits and drawbacks.
Enzymatic digestion (followed by PCR, array, or NGS): prone to false positives
LC-MS: Very sensitive, but can only detect global DNAm levels.
Antibody-based methods: MeDIP/hMeDIP. Depends on the antibody quality. Regional enrichment, not site-specific resolution.
Capture (MBD-based): better genomic coverage overall and coverage of CpG-sparse regions than MeDIP
102
Compare and contrast the advantages and disadvantages of microarray vs next-generation sequencing approaches for profiling genome-wide DNA methylation.
Microarrays: cost effective for population studies, obtain the same sites every time (very comparable between experiments) with good quality data at all sites provided good quality sample.
Sequencing: good for experimental/model organism studies, higher genomic coverage can be achieved. More costly and sequencing depth/coverage varies between experiments, samples, and CpGs.
103
Describe 5 steps of sample preparation prior to array runs.
1. Bisulfite conversion
2. Denature and neutralize bcDNA
3. Incubate DNA overnight (isothermal amplification)
4. Enzymatic fragmentation of DNA
5. Precipitate and resuspend DNA
104
Describe hybridization of genomic DNA to the arrays, single-base extension, and methylation detection in 4 steps.
1. Hybridize fragmented and resuspended DNA to BeadChips (incubate overnight)
2. Wash away unhybridized DNA
3. Flurophore staining and extension (in capillary flow chamber)
4. Imaging (laser excitation) on HiScan/iScan system
105
What are the characteristics of type I and II probes?
Type I probes are either complementary to the methylated or unmethylated allele. Methylation is determined based on whether a single base extension has or has not occurred (presence/absence of colour). Good for CpG-dense regions (e.g. islands); comprise all of the 27K array.
Type II probes were added to the 450K array and can bind either the methylated or unmethylated allele, with methylation determined based on color (red or green). Better for CpG-sparse regions.
106
Compare and contrast the number of probes, probe chemistry, and genomic coverage on the 27K, 450K, EPIC, EPIC v2, and mouse arrays.
27K: ~25K probes mostly targeting CGIs, cancer genes, imprinted genes, and miRNAs. All type I probes.
450K: ~485K probes (3k CpH), including type II probes (producing different distributions: peak-based correction methods suggested). 41% of probes in promoters, plus improved coverage of genomic elements relative to 27K including CpG shores, shelves, and open seas, 5'/3' UTRs, gene bodies, MHC regions, and some enhancers.
EPIC: ~860K probes (~3k CpH), covering additional ~413K from 450K. Improved coverage of enhancers, gene bodies, and other regulatory elements.
EPIC v2: ~930K probes (~3K CpH), with 77% of EPIC probes and ~200K new probes covering more intergenic regions, open chromatin, and enhancers. Optimized for low-input DNA (~1ng). Fewer probes where DNAm is biased by genetic ancestry (stating broader applicability for diverse populations).
Mouse: ~280K CpGs, ~900 CpHs. ~70% are CpG open seas, gene bodies; ~10% CGIs. Includes ~1.3K genotyping probes for mouse strain identity.
107
Compare and contrast the number of probes, probe chemistry, and genomic coverage on the 27K, 450K, EPIC, EPIC v2, and mouse arrays.
27K: ~25K probes mostly targeting CGIs, cancer genes, imprinted genes, and miRNAs. All type I probes.
450K: ~485K probes (3k CpH), including type II probes (producing different distributions: peak-based correction methods suggested). 41% of probes in promoters, plus improved coverage of genomic elements relative to 27K including CpG shores, shelves, and open seas, 5'/3' UTRs, gene bodies, MHC regions, and some enhancers.
EPIC: ~860K probes (~3k CpH), covering additional ~413K from 450K. Improved coverage of enhancers, gene bodies, and other regulatory elements.
EPIC v2: ~930K probes (~3K CpH), with 77% of EPIC probes and ~200K new probes covering more intergenic regions, open chromatin, and enhancers. Optimized for low-input DNA (~1ng). Fewer probes where DNAm is biased by genetic ancestry (stating broader applicability for diverse populations).
Mouse: ~280
108
What types of control probes are included on the human and mouse arrays?
450K: 65 SNP probes
EPIC: 59 SNP probes; ~600 ctrl probes
Mouse: ~1.3K strain probes, ~600 ctrl probes
109
Describe the 9 steps involved in RRBS library preparation.
1. MspI digestion (~300bp, CG-rich fragments)
2. End-repair and A-tailing (overhang to ligate adapters)
3. Ligation of methylated adapters (Cs protected from BS conversion)
4. Size selection of MspI fragments (agarose gel or magnetic beads; larger DNA fragments bind a lower bead conc, then wash with a higher conc to get smaller ones)
5. Optional sample pooling
6. Bisulfite conversion
7. PCR amplification
8. Check library size (PCR, DNA HS chip) and quantify concentration (Qubit, nanodrop, qPCR (KAPA)-based)
9. Sequencing
110
How is a bisulfite conversion check implemented during RRBS library prep?
Phage lambda DNA (unmethylated) spiked in during MspI digestion step
111
How does size selection with magnetic beads work during RRBS library preparation?
The concentration of beads and duration of binding time affects which fragments of DNA bind. Larger DNA fragments bind a low concentration, whereas smaller ones bind a high concentration. Multiple washes can be used to capture or discard varying fragment sizes.
112
How are adapter dimers checked for during RRBS library prep?
PCR (smearing of bands), DNA HS chip
113
Describe how NGS cluster generation works in 5 steps.
Cluster generation = isothermal amplification of DNA fragments in flow cell.
1. DNA hybridizes to an oligo complementary to the adapter.
2. Polymerase copies the ssDNA, making it dsDNA.
3. DNA is denatured and bridge amplification occurs. Adapter hybridizes to a second oligo, and DNA fragment bends. Polymerase makes a dsDNA bridge.
4. dsDNA bridge is denatured, creating 2 copies of the DNA molecule tethered to the flow cell.
5. Reverse DNA strands are cleaved off, and 3' ends blocked to prevent spurious priming
114
Describe how NGS sequencing by synthesis works in 8 steps.
1. Sequencing primers added to the flow cell
2. Fluorescently tagged nucleotides are added and incorporated one by one to DNA strands
3. A laser excites the clusters and records the color (nucleotide identity).
This occurs simultaneously across many fragments in the flow cell ("massively parallel").
4. Product is washed away, and index sequencing begins. Index read primer added, index read generated, and index product washed off.
5. De-protection of 3' ends of template (to allow sequencing of 2nd index)
6. Folding of template and binding to 2nd oligo
7. Sequencing of 2nd index
8. Forward strand cleaved and washed away; process repeats for reverse strand
115
Compare and contrast WGBS, RRBS, MeDIP, and Methyl capture-seq, summarizing the differences in inputs, library preparation, and outputs; describe the advantages and disadvantages of each approach.
WGBS: Whole genome is fragmented via sonification. Typically covers ~75% of genome as compared with RRBS/MeDIP, which cover ~9% of genome. Requires high input (~5ug) of DNA.
RRBS: MspI-digested fragments (CG-rich). Can be used with low input (as low as 10 ng).
MeDIP: ~100bp resolution (regional); moderate input requirement; antibody-dependent
MBD/capture: ~100bp resolution (regional); moderate input requirement; 14% coverage
116
When is pyrosequencing appropriate or inappropriate?
When the target is not in a repetitive area of DNA and is within a short region (PCR amplicon up to 115bp)
117
Describe preparation of samples for bisulfite pyrosequencing in 7 steps.
1. Bisulfite convert template DNA
2. PCR amplify with one biotinylated primer
3. Bind amplicons to streptavidin beads
4. Denature dsDNA
5. Wash beads with vacuum filtration (remove non-biotinylated strand)
6. Anneal sequencing primers and ssDNA
7. Prepare enzymes (ATP sulfurylase, luciferase, DNA pol), subtstrate (luciferin), nucleotides
118
How is sequencing carried out during bisufite pyrosequencing?
- dNTP incorporation releases pyrophosphate molecule
- pyrophophosphate is converted to ATP by ATP sulfurylase
- ATP is used by luciferase to convert luciferin to oxyluciferin, releasing light
119
List 5 challenges and issues with hmC estimation by subtraction and separate analysis of mC/hmC signal in EWAS.
1. mC and hmC are technically and biologically dependent, but we treat them as independent datasets
2. When a CpG has changes in both mC and hmC and these are analyzed separately, both might not be detected due to multiple test correction
3. mC and hmC have different beta distributions
4. Negative/zero estimates of hmC might be unreliable
5. DNA damage in oxBS samples results in lower intensity measurements on arrays, making signal more noisy/less accurate
120
Why are negative hmC values obtained when subtracting BS - oxBS signal, and how can this be addressed during data preprocessing?
When true hmC level is 0 or near 0, technical discrepancy may result in BS beta value < oxBS beta value. Ways to address this:
1. 95% quantile threshold: a 95% confidence interval for the level of technical noise in hmC, determined by the negative values obtained after subtraction
2. Maximum likelihood: assumes true hmC value is 0, and that the mC value is the average of the BS/oxBS measurements
121
List 3-5 methods which can be used to calculate hmC levels from paired oxBS/BS signals in microarray data and/or in bisulfite sequencing data.
1. 95% quantile threshold: a 95% confidence interval for the level of technical noise in hmC, determined by the negative values obtained after subtraction
2. Maximum likelihood (OxyBS, oxBS-MLE): calculates probability of mC, hmC, or C using methylated/unmethylated intensities, and constraining hmC > 0
3. oxBSQC: calculates hmC by subtraction, then calculates a cutoff value for hmC detectability, using coefficient of variation to ID probes with "stable" hmC levels. Further filters probes based on out-of-band signal (from opposite flurophore) due to incomplete DNA hybridization
4. CHYME: Bayesian hierarchical model which estimates mC, hmC, and C proportion at each site, taking into account DNA damage, covariate effects, and repeated measures. Uses BS signal intensities
122
How is an hmC cutoff determined using the oxBSQC package?
The coefficient of variation in hmC among all probes is calculated, and the Youden index is used to determine an optimal median hmC cutoff where probes have a "stable" hmC level
123
How is extra information borrowed/adjusted for when estimating hmC levels using the CHYME approach?
Assumes the data is nonrandom, and estimates the data distribution and its variability in paramters. Each hmC/mC measurement at a CpG acts as a prior for another measurement, so lowly covered/poor quality probes can "borrow" information.
124
Summarize what is known about DNAm changes at the SNCA gene in familial and sporadic PD.
sPD: Decreased DNAm at SNCA intron 1, increased SNCA mRNA expression in putamen, cortex, and SN.
familial: patient-derived SNCA triplication iPSCs have decreased DNAm of SNCA intron 1 and increased protein expression.
125
What is known about the causality between SNCA intron 1 DNAm and SNCA mRNA expression?
In patient post-mortem brain, SNCA intron 1 methylation and SNCA mRNA/protein levels are correlated, but we don't know which comes first.
Experimental manipulation:
In patient-derived triplication iPSCs, forcibly methylating SNCA intron 1 with dCas9-DNMT3A is associated with decreased protein expression
In SK-N-SH cells, 5-aza treatment increases SNCA mRNA/protein expression and is associated with demethylation of intron 1; in vitro methylation of SNCA intron 1 reporter construct decreases reporter activity in HeLa cells
126
Describe 3-5 mechanisms by which nuclear aSyn can alter the epigenome.
1. Altered cell signaling cascades, indirectly affecting gene expression
2. RA-dependent translocation to nucleus and binding directly to DNA; affects nuclear receptor activation, transcription of RA-dependent genes such as DLG1, ATP13A2, PINK1, EHMT2 (H3K9m), PER1, NOTCH1...
3. Interaction with histones
4. Interaction with p300 HAT (inhibitory)
5. Sequesteration of DNMT1 from nucleus --> cytoplasm
6. DNA nicks/DNA damage
127
Summarize the results of DNAm studies conducted on patient-derived dopaminergic neuron iPSCs (sporadic and LRRK2 mutant PD).
450K/RNA-seq: Both LRRK2 G2019S (6) and sPD-derived DAn (4) had genome-wide DNAm and expression changes compared to controls (4), not seen in skin cells.
WGBS: increased genome-wide DNAm in intergenic regions (DAn iPSC from 1 LRRK2 G2019S, 1 sPD)
128
Summarize mQTL findings from PD GWAS, and SMR findings from PD EWAS.
IPDGC 2011: 5 PD SNPs, 3 mQTL in brain (PARK16, STX1B, GPNMD)
Nalls 2014: 26 PD SNPs, 6 mQTL in brain
Vallerga 2020: CpG on chr4 associated with PD in EWAS meta-analysis was associated with SLC7A11 expression, with no evidence for a mediation by genotype (genotype did not drive association between CpG and PD or CpG and expression)
129
Illustrate how PD GWAS can be used to infer effects of DNA methylation on expression and splicing, in the example of Kia et al. 2021.
PD GWAS results were used to determine whether DNAm mediated the relationship between PD and expression or PD and splicing (conditional analysis; this was true for 3 genes).
1. PD GWAS data and public eQTL data were used to predict PD-associated DEGs and splicing (colocalization: relationship between SNP, trait, and eQTL)
2. Predicted PD-DEGs and PD EWAS data were used to predict whether predicted PD-associated gene expression/splicing changes were conditional on DNAm
130
Describe 3 ways by which pesticide exposure impacts cell function and DNAm.
1. Oxidative stress and ROS generation
2. Altered SAM/SAH ratio (may impact DNMT activity)
3. Immune activation/immunotoxicity
131
Summarize 2 findings from in vitro and in vivo studies on the effects of pesticide exposure on DNAm and 1 finding on histone acetylation in rodent brain tissues/neural cell lines.
1. Changes in global DNAm (increase or decrease depending on pesticide, exposure, duration, brain region: paraquat = increased DNAm in mouse striatum, organochlorines = decreased DNAm in mouse hippocampal cell culture)
2. Reduction in DNAm at SNCA intron 1 (mouse SN)
3. Reduced HDAC expression in SH-SY5Y cells, associated with increased pan-histone acetylation and histone expression
132
Summarize findings on the effects of pesticide exposure on the epigenome in human post-mortem brain.
Pesticides crossed the BBB, altering DNAm at genes expressed in glial cells and related to cell death, neurological disease, inflammatory genes (PD patient postmortem TC)
133
Summarize findings on the effects of pesticide exposure on blood DNAm in human populations.
van der Plaat 2018: non-PD (n=1500)
- DNAm associations with exposure (overall occupational) were dependent on sex (females only), airway obstruction (obstructed individuals only), and smoking (pesticide-smoking interaction); no change in cell type composition or epigenetic age
Paul 2018: PEG (PD and control), blood and saliva
- ambient OP exposure associated with reduced global DNAm, altered DNAm of nicotinic ACh receptor genes (PD and non-PD)
Go 2020: 20 plantation workers with PD (blood/brain)
- 10 years plantation work associated with differential DNAm of genes related to mitochondrial and neuronal function (blood)
134
List 3-5 factors interacting with pesticide exposure and DNAm changes, and describe supporting evidence.
1. Smoking: ambient OP exposure alters DNAm at nicotinic AChR genes; occupational pesticide exposure and smoking interact to influence DNAm
2. Sex: female-specific association of occupational pesticide exposure and DNAm
3. Airway obstruction: obstruction-specific association of occupational pesticide exposure and DNAm
4. PD: a subset of blood-specific DNAm changes with ambient OP exposure occur only in PD patients (7/70, 10% in PEG)
5. Cell type: genes expressed in glia have altered DNAm in brains of plantation workers
135
Based on our current understanding, are DNAm alterations related to pesticide exposure dependent on or independent of PD status?
In PEG ambient OP study, associations were mainly independent of PD status (63/70 DMPs in blood found in patients and controls when analysis was stratified)
136
Briefly summarize our state of knowledge on the contributions of heavy metals, pollution, and other environmental toxins to DNA methylation and the relevance of this for PD. What tissues are affected, and by which mechanisms are DNAm patterns altered?
- Heavy metals and air pollution alter local and global DNAm patterns in multiple tissues (human blood and lung, rodent liver, brain, and reproductive organs)
- PD patients (SGPD & PEG) have higher DNAm-predicted lead exposure than controls
- Heavy metals can deplete SAM and reduce DNMT activity/increase TET activity in short term; increased compensatory DNMT activity long-term
137
Name 3-5 mechanisms by which exercise impacts DNAm in general (pan-tissue relevance)
1. Increase in creatine kinase levels (phosphocreatine energy reservoir is produced)
2. Cytokine release
3. Altered expression of energy metabolism genes
4. Oxidative DNA damage
5. Hormone/growth factor secretion (e.g. BDNF)
138
Describe what is known about the effects of exercise on DNAm and epigenetic modifiers in blood, skeletal muscle, and adipose tissues in human studies, focusing on intervention-based designs (acute, 3 months, 6 months)
Acute (1 bout of exercise):
- increases DNMT3A/B expression in blood
- decreases DNMT3B expression in muscle
- reduces HDAC levels in muscle
1-3 months:
- altered global blood DNAm in elderly
- altered site-specific blood DNAm in young men
- decreased global blood H4ac in SCZ population
- altered site-specific skeletal muscle DNAm and gene expression
6 months:
- altered site-specific blood DNAm in elderly adults with MCI (amyloid biology, protein trafficking, lipoprotein regulation genes)
- altered adipose global and site-specific DNAm in men (women not studied)
- counteracts tumor suppressor DNAm changes, possibly buffering against cancer
139
What do we know about the long-term stability of exercise-induced DNAm changes in humans?
DNAm changes in skeletal muscle observed immediately after acute exercise are stable if exercise is continued for a 7 week on, 7 week off, 7 week on period, and are associated with gene expression at the end of the 22 weeks
140
Summarize the effects of exercise on epigenetic modifiers in rodent brain studies. How does exercise impact the aging neuroepigenome? How do the effects of exercise differ from the effects of environmental enrichment?
In the rodent hippocampus, exercise:
- Decreases Dnmt1 and Dnmt3b protein levels (rats)
- Increases Tet1 and Tet2 mRNA levels, global DNAhm (mice)
- Prevents decline in Tet expression with aging (rats)
- Increases phospho-MeCp2, Bdnf expression; decreases Bdnf promoter DNAm (rats)
EE is associated with no change in Tet1/2/3 mRNA levels, and decreased global DNAhm in mouse hippocampus.
141
Briefly summarize some of the key findings from human and macaque social epigenetics studies, providing 3 examples.
Lam PNAS 2012: PBMC DNAm levels were related to cortisol, stress, and early life SES
Smith 2017: Monocyte DNAm levels at stress and inflammation-related genes were related to neighbourhood social environment (quality, safety, social cohesion)
Larger body of research on biological embedding of negative environments (maternal deprivation, childhood trauma, violence, crime) associated with DNAm and health outcomes.
Tung 2012: Macaque PBMC DNAm levels, epigenetic age, and gene expression were related to social rank in females; rank could be predicted from gene expression
142
What mechanisms are hypothesized to connect social epigenetics to brain gene regulation? Summarize some evidence from animal studies.
Stress/immune pathways altered in response to social environments are associated with release of pro/anti-inflammatory cytokines, which can impact microglial activity and brain inflammation. In rats and mice, social isolation is associated with microglial activation and changes to global DNAm, H3K4m, H3K9m, and HDAC activity. In macaques, maternal deprivation is associated with altered cortical DNAhm.
143
Describe impacts of housing mice in enriched environment on cyotsine modifications in the hippocampus (CpG/CpH methylation, DNAhm), and chromatin modifications in the cortex. How does housing mice in EE affect the aging neuroepigenome?
- EE increases CpG/CpH methylation and DNAhm differences between dorsal and ventral DG in mice, particularly at NeuroD1 binding sites (neurogenesis TF)
- EE is associated with global hippocampal DNAhm loss in mice (6-week, 18-month), particularly at axon guidance genes; estimated to be dependent on Tet1 activity (as opposed to expression)
- EE increases cortical enhancer chromatin accessibility, and alters promoter (H3K4me3, H3K27ac) and gene body (H3K79me2) histone PTMs
- Promoter- and enhancer-related cortical EE changes associated with mRNA and protein expression of synaptic (e.g., glutamatergic) and ECM proteins
- 3 months of EE counteracts 30% of age-related changes in hippocampal DNAm (14m mice)
- EE prevents age-related DNAm loss at hippocampal MeCP2 binding sites (neuroplasticity, neurogenesis genes)
144
What evidence supports global loss of DNAm in PD brain, and what are two proposed mechanisms for this?
2011: global DNAm loss in PD cortex (5mC antibody)
2013: site-specific analysis revealed more DNAm loss than gain in PD cortex (450K)
2019: IHC study in neocortex revealed no change in global DNAm
2021: global DNAm loss in SN of MPTP-treated mice, in MPTP-treated SH-SY5Y cells, and PD patient plasma
Possible mechanisms = aSyn sequesteration of DNMT1, miRNA interference of DNMT1 translation in PD
145
Which gene has the most evidence of replication for differential DNAm in PD brain?
CYP2E1: lower DNAm and higher expression in PD cortex & putamen
146
Which brain region(s) show the most DNAm changes in PD and why?
Young 2019: DMV (part of brainstem) > CG > SN
Proposed that DMV is a site of early PD pathology and connected to the vagus nerve, with CG developing pathology later. What about SN? Authors also suggested differing amounts of cell loss (e.g. greatest in SN) likely contributed to the differences.
147
What is the evidence for altered SNCA intron 1 DNAm in PD brain, which CpGs does it affect, and what are predicted to be the consequences of this altered DNAm?
Decreased DNAm at intron 1 in PD SN, putamen, and cortex (12 sPD patients), mostly at TFBS, suggesting that DNAm of SNCA intron 1 TFBS might regulate its transcription. This might be specific to PD rather than aging, as another study found no change in SNCA intron 1 DNAm w/age.
148
What are the proposed roles of Tet2 and DNAhm in PD neuropathology?
Global DNAhm is increased in cerebellar white matter from PD patients (IHC), and PD patient-derived PFC neurons have increased TET2 expression and enhancer DNAhm. In mouse experiments, Tet2 was shown to mediate inflammation-related DAn loss (Tet2 inactivation prevented this). Proposed to be involved in transcription of immune response genes.
149
Discuss the overall strengths and weaknesses of our current understanding of DNAm changes in PD brain.
PD brain EWAS are typically done using smaller sample sizes (i.e. n=10-40 individuals) and often don't have correction for cell type. Post-mortem brain studies also only assess late stage PD.
Strengths include assessing multiple brain regions (Young 2019) or tissues (Masliah 2013) from the same individuals, and comparing results to cell and animal models (Desplats 2011, Marshall 2020, Zhang 2021).
150
How concordant are PD-related DNAm changes in brain and blood, and what is the strength of evidence supporting this?
In Masliah 2013 study (5 PD, 6 ctrl), 30% of total gene-annotated differentially methylated CpGs in brain and blood were concordant between these tissues
151
Summarize 2-3 of the most interesting genes differentially methylated in both tissues in the Masliah 2013 blood/brain study.
GSTTP1: polymorphisms associated with response to paraquat
MIR886: small vault RNA upregulated in early PD and throughout Braak stages 1-5 (vault ribonucleoprotein complex, nucleocytoplasmic transport)
MAPT
152
As PD blood EWAS statistical approaches progressed over the 2010s and into 2020, what were some key advances made regarding identification of confounders and contributors to PD-associated blood DNAm patterns?
Cell type: Chuang 2017 study (PEG) was the first to report EWAS results with and without adjustment. Without adjustment, mostly immune pathways; with adjustment, Wnt signaling and neuron differentiation.
PD-associated cell type shifts confirmed by Henderson-Smith 2019 and Vallerga 2020. Adjustment for age and sex, but no smoking/pesticides.
Medication use, and disease progression: Henderson-Smith 2019 (medication-stratified analysis of progression). Adjusted for age, sex, smoking, cell type.
Genotype: Vallerga 2020 used SMR to investigate whether relationship between DNAm and expression in PD was genetically influenced. Adjusted for age and cell type.
153
How is PD ascertained in peripheral tissue-based studies on living individuals?
Parkinsonism or PD is clinically diagnosed (typically by a neurologist using standard criteria such as UK Brain Bank, Gelb, or Calne). In early stages this includes some uncertainty in the diagnosis (sensitivity > specificity)
154
How do PD-associated blood DNAm patterns change over time?
Over a 2-year period (starting with Hoehn & Yahr stage <3 at baseline), Henderson-Smith et al. found 138 DM-CpGs, or 200 DM-CpGs when stratified by medication use (levodopa and entacopone, COMT inhibitor)
155
How does L-dopa impact blood/brain DNAm, and how do these findings inform interpretation of PD EWAS?
Conversion of levodopa --> dopamine consumes methyl groups from SAM.
Blood: L-dopa treatment increases DNAm at SNCA gene in sPD patients (n=1k, validated in cell culture).
Brain: In 6-OHDA rat model, L-dopa is associated with striatal DNAm changes and increased expression of Tet3 and IEGs. Treating with L-dopa and methionine concurrently reduces LID and DNAm alterations associated with L-dopa treatment alone, suggesting some benefit for methyl donor supplementation alongside L-dopa treatment.
Interpretation: L-dopa might both disturb and restore DNAm patterns in PD, by regulating normal DA levels/DAn functioning while simultaneously reducing overall methyl donor availability.
156
How does immune cell type composition change in PD? Name 4 characteristic cell type shifts.
1. Increased granulocytes (granule secretion to kill pathogens, part of innate immunity; Horvath & Ritz 2015, Henderson-Smith 2019, Vallerga 2020)
Decreased overall lymphocytes, predictive of developing PD later. Genotype underpinning this was suggested to be causal of PD based on MR (Jensen, 2021).
More specifically:
2. Lower antibody-producing B cells in PD (Horvath & Ritz 2015, Henderson-Smith 2019, Vallerga 2020)
3. Lower CD4+ T helper cells in PD (produce cytokines for cytotoxic T cell growth; Horvath & Ritz 2015, Vallerga 2020)
4. Lower NK cells in PD (receptor-mediated apoptosis or lysis of pathogens based on MHC recognition; Henderson-Smith 2019, Vallerga 2020)
157
Summarize findings on epigenetic age acceleration in blood and brain of PD/RBD patients.
Blood:
- EAA/IEAA increase in PD reported in PEG1
- Horvath, Skin&Blood EAA increase in RBD (n=23 male); no association with synucleinopathy conversion (n=8) or disease duration
Brain:
- cortical age acceleration associated with LB pathology in neocortex and with parkinsonism (DLFC, n = 721)
- cortical age was not associated with PD in sex-stratified analysis (patient-derived parietal cortex neurons, n = 100), but was associated with parkinsonism, dementia, cognitive decline, and AD pathology
158
How is epigenetic age acceleration in PD affected by cell type and demographic factors?
In PEG, EAA with PD was reported regardless of correction for cell type, smoking, pesticides, or family history of PD, but was dependent on sex (men had higher EAA and IEAA).
159
Summarize the main findings from the Vallerga et al. study, the largest PD EWAS to date, and discuss its benefits and drawbacks. How did the authors determine that DNAm was likely to affect SLC7A11 expression via environment and not genotype?
EWAS in SGPD (~1600): 2 probes with MOA method, none with MOMENT method
EWAS in PEG (~400): no hits
EWAS meta-analysis in SGPD + PEG (~2000): 2 probes, 1 from SGPD MOA EWAS
One of the two probes had an associated mQTL. SMR was used to determine whether DNAm at that probe, expression at the nearest gene (SLC7A11), and genotype of the associated mQTL were related; found evidence for association between DNAm and expression, independent of mQTL.
~260k probes associated with PD in SGPD explained 2.8% of variance in case-control status in PEG (AUC 0.7).
Strengths: large sample size, meta-analysis, discovery and replication analysis
Weaknesses: poorly phenotyped (even age is predicted since not recorded for all participants), varying disease duration
*note genomic inflation factor was calculated and was near 1, so authors argue there was unlikely to be confounding
160
Summarize findings of PD-associated DNAm changes in saliva in the Chuang 2017 study.
n = 300 with saliva. 5 DM-CpGs relating to mitochondrial function, neuron projection, cytoskeletal organization, immune response, iron handling. No adjustment for cell type
161
How does non-CpG methylation change in PD?
One study reported some mitochondrial CpH changes in SN of PD patients...
162
Name 4 advantages of applying multi-omics integration to human complex disease research.
1. Prioritize functionally relevant loci for follow up
2. Develop better predictive biomarkers by incorporating crosstalk between omics layers
3. Develop better treatments/personalized medicine strategies
4. Aid in determining causality (through genetic data)
163
Summarize and evaluate the main findings from two studies integrating blood DNAm and expression data in PD.
Wang et al., Clin Epigenetics 2019: PEG DNAm data and another publicly available mRNA expression dataset.
Used average DNAm within a gene feature or gene, Pearson correlation with expression (no adjustment for covariates). 52 genes had correlated DNAm and expr, mostly negative; random forest classifier for PD based on 85 correlated genes had an AUC of 0.74.
Henderson et al., 2021: 30 individuals with matched blood/brain tissue (15 PD, 15 control). Batch and cell type but not sex or age corrected for (showed they didn't influence models). Site-specific analysis with limma and DMR analysis with bumphunter; 31 DMRs, including VTRNA2-1.
eQTM analysis with MatrixEQTL (DNAm ~ expression): found 19 cis- and 43 trans-eQTMs, in genes relevant to NDDs (AD/HD).
Used FEM to find 6 modules: immune, Ub ligase, cell signaling, transcription.
164
What is the correlation between mRNA and protein expression in PD brain?
In a 2022 study in PD midbrain (19 PD, 12 control), only 1 gene passed thresholds in mRNA (~600 DEGs, padj < 0.1) and proteomics analyses (~20 DE proteins, padj < 0.1) when separate hits were overlapped, although top hits were correlated in both datasets.
In a 2015 PFC study, 8 genes (5% of total) were DE in mRNA and proteomics analyses, fewer than expected by chance (overlapped hits from separate analyses). Hypothesized something may be interfering with translation in PD...
165
Describe how PD GWAS and brain DNAm data can be used to jointly impute associations, with specific examples from [[Rawlik et al. 2016]] and [[Kia et al. i2021]].
Rawlik 2016: Imputed PD-associated DNAm levels in 4 brain regions using PD GWAS data and healthy control brain DNAm/genotype data. Found 6 CpGs DM in 1 or more brain regions each, within chr17 region containing MAPT and other PD-associated SNPs, one of which was validated by 450K array in SN tissue.
Kia 2021: Sought to determine mechanism of action of PD GWAS variants on gene expression and splicing, and whether these were mediated by DNAm. Annotated PD-SNPs to eQTLs and sQTLs to predict differentially expressed/spliced genes under genetic influence (TWAS, colocalization). Then, used conditional analysis with PD-associated CpGs to find evidence of DNAm and eQTLs/sQTLs jointly influencing phenotype.
166
Summarize approaches and results from [[Multi-omics in PD midbrain Caldi Gomes 2022]] (integration of mRNA, miRNA, and proteomics in advanced PD).
RNA-seq (total and small RNA) and MS profiling in midbrain of 19 PD, 12 control. Found ~600 DEGs related to immune, stress, inflammation, apoptosis, metabolism, cytoskeleton, and development. 1/20 proteins (CH13L1) was also differentially expressed on mRNA level. Not many DE miRNAs.
167
How is eQTL/eQTM analysis performed with MatrixEQTL?
Additive linear modelling or ANOVA is used to test relationship between DNAm and expression/genotype and expression
168
Compare and contrast WGCNA, FEM, and SMITE approaches to multi-omic integration, focusing on DNAm/expression integration.
WGCNA: Constructs networks of co-expressed/co-methylated genes, with modules constructed via hierarchical clustering.
FEM: Uses promoter (or 1st exon/TSS1500) DNAm levels and mRNA expression levels to score genes, assuming a negative correlation. Gene scores are annotated to a PPI, with enriched modules selected via spin-glass algorithm.
SMITE: A more flexible version of FEM, where different gene features can be assessed, which can have a bidirectional, positive, or negative relationship with expression. Users can add in other omics, custom files to segment the features (e.g. H3K4me1 ChIP-seq for enhancers), and can weight features differentially.
169
Explain how colocalization, TWAS, conditional analysis, and Mendelian randomization can be used for causal inference.
Colocalization: probability that multiple traits are related to a phenotype, either jointly or opposingly.
TWAS: using multiple datasets to impute relationships between different omics layers (e.g., PD GWAS + eQTL data = imputing PD-DEGs)
Conditional analysis: like mediation analysis in theory; the relationship between a trait and phenotype is conditional on another trait (e.g., relationship between genotype and expression is conditional on DNAm)
MR: probability that a trait and phenotype are causally related through genetic variation. Assumes genotype and trait are related and there is no other confounding (not typically met).
170
Provide two examples of how machine learning can be used in multi-omic integration.
Hierarchical clustering: can input multiple omics into iCluster tool to find related subtypes .
Feature selection & classification: DIABLO finds features that can classify or predict phenotypic groups in a sample across multiple datasets.
ChromHMM: HMM used to "learn" and infer hidden chromatin states from multiple ChIP-seq datasets.
171
Discuss scenarios when supervised vs unsupervised integration methods would be desired, and explain the pros and cons of each approach.
Unsupervised: exploratory analysis, hypothesis-generating.
Supervised: classification analysis or developing predictive biomarkers. can be biased/return what you "want" to see; should be validated with testing/training data sets.
172
How can ultra-high dimensionality be addressed in multi-omics integration?
Feature selection, e.g. in machine learning-based approaches, so that not all data is used for integration.
Or correlate/score omics and then test those correlations. This can help address the MTC burden seen when datasets are analyzed individually and then combined.
In some cases, e.g. QTL analysis, both data types can be jointly modelled together.
(Mouse project likely suffered from this, but there was also weak DNAm signal, so wasn't sure about trusting scoring/enrichment-based methods...chose to be more stringent).
173
How can prior information be incorporated into multi-omics integration?
Bayesian approaches, e.g. "integrative regression", Bayesian clustering, or colocalization (uses prior information about a SNP from GWAS summary statistics when estimating SNP-trait relationship)
