Epigenome and phenome study reveals circulating markers pertinent to brain health

Characterising associations between the epigenome, proteome and phenome may provide insight into molecular regulation of biological pathways governing health. However, epigenetic signatures for many neurologically-associated plasma protein markers remain uncharacterised. Here, we report an epigenome and phenome-wide association study of the circulating proteome in relation to brain health. We perform epigenome-wide studies of 4,235 plasma proteins (n=778), identifying 2,895 CpG-protein associations (protein quantitative trait methylation loci; pQTMs) after stringent correction for multiple testing. These were independent of known genetic protein quantitative trait loci (pQTL) and common lifestyle effects, extending current knowledge by analysing a further 3,286 protein measurements with 2,854 novel pQTMs. We then perform a phenome-wide study of each protein in relation to neurological traits in 1,065 individuals, identifying 644 proteins related to cognitive, brain imaging phenotypes or APOE status. By integrating our pQTM dataset with our phenome association study, we uncovered 88 epigenetic associations for protein markers of neurological traits, 83 of which were previously unreported. These data are pertinent to understanding heterogeneity in brain health.


Introduction
Multiple layers of omics data indicate the biological pathways that underlie disease risk. Epigenetic modifications account for inter-individual variability in circulating protein levels [1][2][3] and have been linked to a range of neurological outcomes [4][5][6] and neurodegenerative conditions such as Alzheimer's disease 7 . Proteome-wide characterisation of blood protein signatures has also been facilitated at large-scale by SOMAscan ® protein measurements, with studies that have begun to map plasma protein signatures of cognitive decline and dementia risk [8][9][10] . There is, however, a need to further integrate omics datasets at large-scale to characterise and predict disposition to complex disease. 4 of neurological outcomes 20,21 and brain health is also thought to be influenced by peripheral factors 12,22 . Brain morphology alterations and disease pathogenesis can occur many years prior to overt neurological, cognitive and behavioural symptom presentation. Risk stratification and intervention for secondary prevention in advance of clinical diagnoses may therefore be feasible, with blood sampling that supersedes more invasive and expensive tests. Therefore, while datasets that allow for integration of novel protein markers, phenotypic outcomes and epigenetic signatures of protein levels from the blood are rarely-available, they hold substantial potential to advance the development of these endeavours.
Here, we conduct an integrated epigenome-wide and phenome-wide study of the plasma proteome ( Fig. 1). We characterise CpG-protein (pQTM) associations for 4,235 SOMAmer protein measurements in N=778 individuals from the Generation Scotland cohort (Supplementary Table 1).
Epigenetic pQTM datasets for the plasma proteome are provided, with 3,286 SOMAmers previously untested in protein EWAS, enabling future studies to extract phenome-specific pQTM signatures for traits of interest. We then characterise proteome signatures for a range of neurological traits (structural brain imaging measures, cognitive ability and APOE carrier status) in N=1,065 individuals from the same cohort where the pQTM data are a nested subset. By integrating these datasets, we probe the epigenetic signatures of proteins that are related to brain health and dementia risk in an ageing population.

Epigenome-wide studies of 4,235 plasma proteins
We conducted EWAS to test for pQTM associations between 772,619 CpG sites and 4,235 circulating protein levels (Supplementary Table 2) in N=778 individuals from Generation Scotland . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021 Table 5). Across all models, 2,451 pQTM associations remained consistently significant. Lambda values for the 153 unique proteins that had pQTMs in the fully-adjusted EWAS are presented in Supplementary Table 6.
Of the 2,895 pQTMs from the fully-adjusted models, 1,141 involved Pappalysin-1 (PAPPA) protein and cg07839457 (NLRC5 gene) was found to be the most frequently selected site, replicating the findings of a previous EWAS by Zaghlool et al 1 Table 5 and Fig. 2). A summary of CpGs selected in the fullyadjusted models, with EWAS catalogue 28 lookup of genome-wide significant phenotypic associations is presented in Supplementary Table 8. A summary of known pQTLs and whether these were available for adjustment in our models is provided in Supplementary Table 9.

Proteome associations with neurological phenotypes
We next conducted a proteome-wide association study on neurological phenotypes (protein PheWAS of brain imaging, cognitive and APOE e4 status, alongside age and sex; Fig. 3). A maximum sample of 1,065 individuals was available (mean age 59.9 years, SD 9.6 years, 59% Female; Table 1); all N=778 individuals from the pQTM study were included in these analyses.

Supplementary
At False Discovery Rate (FDR)-corrected P < 0.05, the levels of 798 plasma proteins were associated with age and 814 were associated with sex, with 394 proteins common to both phenotypes (Supplementary Table 10). When compared with work by Lehallier et al using SOMAmer measurements 29 , 340 of 476 age associations replicated and 372 of 450 sex associations replicated, with direction of change consistent across these associations.
There were 644 protein markers with FDR P < 0.05 that associated with neurological phenotypes.
These consisted of 54 brain imaging (Supplementary  10 , and the five novel relationships were for NEFL, ING4, MENT, PAF and TMCC3. The strongest association for APOE was observed with levels of LRRN1 (Beta = 0.57, SE = 0.03, FDR P = 1.73x10 -104 ); a selection of associations for APOE are presented in Fig. 3b.
Many of the 644 protein marker associations were independent and did not cross neurological modalities. However, levels of APOB and ING4 were associated with APOE haplotype status and cognitive scores. Notably, 25 unique proteins were associated (FDR P < 0.05) with both cognitive and imaging traits ( Fig. 3a; Supplementary Table 15). The direction of effect for all of the 25 common proteins was consistent with poorer brain health across both modalities. Integration of the neurological proteome with our pQTM dataset pQTM signatures were explored for the 644 protein markers with FDR P < 0.05 in the protein PheWAS of neurological outcomes. We examined whether the 644 proteins identified in the PheWAS had significant pQTM associations in our EWAS. Twenty-five of the 644 proteins had significant pQTMs. 88 loci were shared across these 25 proteins. Of the 88 pQTMs, 42 were trans . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint ( Fig. 4) and 46 were cis associations (Supplementary Table 16). Eighty-three of the 88 pQTMs were novel.
Several CpG sites were associated with multiple protein levels in the trans pQTMs (Fig. 5). DNAm at site cg06690548 in the SLC7A11 locus was associated with PSAT1, SCUBE1, ACY1 and ALDOB levels. The cg11294350 site in the CHPT1 genomic region was associated with HEXB and SMPD1 levels. pQTMs were reported between circulating RBL2 levels and sites cg01132052, cg0539861 and cg18404041, within the NEK4/ITIH3/ITIH1 gene region of chromosome 3. Finally, a large proportion of trans pQTMs revolved around NLCR5 inflammatory CpGs, with five protein markers associated with either the cg07839457 or cg16411857 sites in seven pQTM associations ( Supplementary Fig. 2). A further nine and 13 associations were identified for GBP1 and MX1, respectively. Two singular trans associations were found; one between cg02521229 and TREM2 and one between cg08228578 and PSAT1. Finally, the levels of CRTAM and SCUBE1 were also associated with DNAm at cg18544413 and cg06088069, respectively.

Discussion
We have conducted a large-scale study of epigenetic and phenotypic correlates of the circulating proteome in relation to brain health outcomes. We provide the first epigenome-wide characterisation of the full SOMAscan ® panel V.4, for 4,235 protein measurements, adding data for 3,286 SOMAmer protein measurements and over 2,854 novel pQTM associations to current knowledge. By characterising 644 proteins associated with neurological phenotypes and integrating them with our pQTM dataset, we uncover 88 epigenome-proteome-phenome signatures pertinent to brain health.
The directionality of pQTM relationships is unclear; however, these signatures may have predictive . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint value for risk stratification and identify biological effectors that may contribute to neurological phenotypes.
Though informative, protein biomarkers are only one component of phenotypic regulation.
Integration of our protein EWAS dataset with protein PheWAS associations in this study revealed novel pQTMs associated with neurological markers (Fig. 5). While this study is focused on blood samples -limiting interpretation to tissues such as the brain -many of the 88 pQTMs involved CpG sites and proteins that have been previously implicated in studies of neurological phenotypes. The NEK4/ITIH3/ITIH1 is a key locus implicated in schizophrenia and bipolar disorder by several largescale, genome-wide association studies (GWAS) [30][31][32][33] . Similarly, the RBL2 locus has been associated with intelligence, cognitive function and educational attainment in GWAS (n > 260,000 individuals) [34][35][36] . The HEXB and SMPD1 proteins associated with DNAm at cg11294350 (in the CHPT1 gene) are involved in neuronal sphingolipid degradation pathways in the brain and have been associated with the pathology and onset of a range of neurodegenerative conditions across the lifespan 37-40 . We make our proteome-wide epigenetic dataset publicly available, such that proteome-specific pQTM signatures can be extracted by studies for any desired trait, as we have demonstrated here for traits related to brain health.
DNAm at cg06690548 has previously been identified as a blood-based marker for Parkinson's disease risk (N > 900 cases and N > 900 controls); in this study, Vallegra et al found evidence that DNAm at this site was consistent with environmental exposure and causally implicated in Parkinson's disease risk 41 . Xc-is the cystine-glutamate antiporter encoded by SLC7A11, which facilitates glutamatergic transmission, oxidative stress defence and microglial immune response in the brain 4243 . Xc-is also a target for the environmental neurotoxin β -methylamino-L-alanine (BMAA), which has been linked to the onset of neurodegenerative process 41 . Given that the proteins implicated in the cg06690548 (SLC7A11) pQTMs were associated with declines in processing speed . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 0 and increased relative brain age, the CpG sites we identify in this study -such as cg06690548 -may be important plasma markers for brain health that merit further exploration.
In the fully-adjusted EWAS we corrected for smoking and BMI; the 88 identified pQTMs may therefore represent unmeasured biological or environmental factors associated with inter-individual variation in protein levels. The EWAS dataset we provide represents the largest set of pQTMs for circulating plasma proteins to date (4,235 protein levels, with 2,895 fully-adjusted pQTMs). Our Several of the markers for cognitive function were identified in previous work relating Olink proteins to cognitive function (such as PVR, VWC2 and CPM) 49 and work that characterised SOMAmer signatures of cognitive decline and incident Alzheimer's disease (such as SVEP1 and TREM2) 8 . C5 and C3 are potent mediators of widespread peripheral inflammation 52 and were associated with cognitive, but not brain morphology traits. The APOE carrier-related proteins were also largely distinct from cognitive and brain imaging proteomic signatures. The strong association between lower BIRC2 levels with e4 carrier status replicates key findings of a previous study of APOE haplotype 10 . Each measurement strategy may therefore be reflective of divergent biological pathways. Continued integration of omics with these phenotypic modalities in increasing sample populations will help to clarify these signals.
Notably, there were 25 proteins that were associated with both cognitive and imaging phenotypes; our results suggest that these proteins are reflective of a common, biological signature of poorer . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 1 outcomes across modalities. Higher levels of three (IGLON5, NCAN and GLIPR2) of the 25 proteins had were associated with increased brain health, and previous studies have linked IGLON5 and NCAN to neural maintenance pathways 50,51 . Higher levels of the remaining 22 markers were associated with poorer neurological health; RBL2, HEXB and SMPD1 are examples of these and each had novel pQTMs. These proteins are therefore strong candidates for exploration of pathways pertaining to brain health.
Our study has several limitations. First, many of the proteins identified in our protein PheWAS did not have epigenetic pQTMs; this may be due to 1) the presence of pathways relating to neurological disease that are not reflected by the plasma epigenome, 2) underpowered analyses, or 3) the presence of indirect pathways that are not captured by our linear mixed model approach. Second, though full replication of our results was not possible, we did see overlap with published findings (across 1,123 SOMAmer protein measurements) from Zaghlool et al 53 suggesting that pQTMs generalise to individuals of other European ancestries. Third, we observed a degree of inflation for the PAPPA and PRG3 proteins in particular, which had large signals across the epigenome. This was despite the fact that mixed effects model structure was used to account for unknown confounding by modelling intercorrelations between CpGs. One reason for this inflation may be a lack of adjustment for eosinophil cell proportions 53 , which were unavailable. Fourth, the extent of non-specific and crossaptamer binding with SOMAmer technology has not been fully resolved 54 . Fifth, although we regressed known pQTLs effects for 3,622 of the 4,235 proteins 55 , there are likely unknown genetic pQTLs that influenced pQTM associations. Within our neurological trans pQTMs, however, only one association for CD163 had a pQTL that was not possible to adjust for in our sample. There were overall 184 of 1,021 possible pQTLs 24 that were unavailable in the sample. Further characterisation of pQTLs and advances in multi-omic modelling techniques 2 will aid in the separation of genetic and environmental influences on epigenetic signatures. Finally, though we have incorporated a wide portfolio of neurological phenotypes, we recognise that the 14 we include in this study are not . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint extensive and results may vary depending on the type of cognitive or imaging measure studied. This is why increasing triangulation across modalities, as we have shown here, will be useful in identifying candidate markers of brain health.
In conclusion, by integrating epigenetic, proteomic and phenotypic data we have uncovered 2,854 novel pQTMs, 83 of which involve the levels of plasma proteins that were associated with neurological phenotypes. We also define proteomic signatures of cognitive ability, brain morphology and APOE carrier status. As these traits are pertinent to cognitive ageing and neurodegeneration, these data are likely to inform preventative approaches and risk stratification.

The Generation Scotland sample population
The Stratifying Resilience and Depression Longitudinally (STRADL) cohort used in this study is a There were N=1,065 individuals with proteomic data available and N=778 of these had DNAm data available. Supplementary Table 1 summarises the demographic characteristics across the two groups, with descriptive statistics for phenotypes.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

DNAm measurement
Measurements of blood DNAm in the STRADL subset of GS subset were processed using the same methodology as those collected in the wider Generation Scotland cohort. Two sets were generated using the Illumina EPIC array. Quality control details have been reported previously 59-61 . Briefly, outlying probes were removed based on visual inspection, bead count and detection P value. Samples were removed if there was a mismatch between predicted and actual sex, as well as those with outlying detection P values. All non-specific CpG and SNP probes (with allele frequency > 5%) were removed from the methylation file. After quality control, 793,706 and 773,860 CpG were available in sets 1 and 2, respectively. These sets were truncated to include a total of 772,619 common probes and were joint together for use in the EWAS, with 479 individuals included in set 1 and 299 individuals in set 2.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint is exclusive to measures such as memory and processing capability that are considered fluid. gf may therefore be of pertinence to delineate when assessing cognitive performance in ageing individuals.
The acquisition protocols for the brain imaging phenotypes are documented in a previous publications 574867 . The derived brain volume measures were recorded at two sites (Aberdeen and Edinburgh) and site was included as a covariate in all imaging analyses. Brain volume data included whole brain volume, global grey matter volume, cerebral total volume and total intracranial volume.
Intracranial volume was used as a nuisance covariate to adjust for head size in all tests of brain volume associations. The derived global white matter integrity measures included global fractional anisotropy (gFA) and global mean diffusivity (gMD). Please refer to the previous publications for details of the protocols applied to derive the brain volume measures from T1-weighted scans, and white matter integrity measures from diffusion tensor imaging (DTI) scans 4867 . White-matter hyperintensity ratings for each participant were derived through visual inspection of FLAIR scans, using the Fazekas scale 68 . White matter hyperintensities (WMH) were defined as punctuate, focal or diffuse lesions in the deep or periventricular white matter, basal ganglia or brainstem, visible as areas of hyperintensity on FLAIR images. Total score was on the scale of 0-6 and was obtained by summing deep and periventricular hyperintensity scores (each on the scale 0-3). Brain Age was estimated using the software package 'brainageR' (Version 2.1; DOI: 10.5281/zenodo.3476365, available at https://github.com/james-cole/brainageR), which uses machine learning and a large training set to predict age from whole-brain voxel-wise volumetric data derived from structural T1 images 13 . This estimate was residualised over chronological age to produce a measure of Relative Brain Age. Sex and scan site were also included as predictors in the model used to residualise to ensure that any bias related to either was excluded from the final measure.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Phenome-wide association analyses
Prior to running protein PheWAS analyses, protein levels were transformed by rank-based inverse normalisation and scaled to have a mean of zero and standard deviation of 1. Models were run using the lmekin function in the coxme R package 69 . This modelling strategy allows for mixed-effects linear model structure with adjustment for relatedness between individuals. Models were run in the maximum sample of 1,095 individuals, with the 4,235 protein levels as dependent variables and phenotypes as independent variables. A random intercept was fitted for each individual and a kinship matrix was included as a random effect to adjust for relatedness. Age, sex (male = 1, reference female = 0), numerical APOE variable (e2 = 0, e3 = 1, e4 = 2), cognitive and brain imaging phenotypes were included as outcomes. Diagnosis of depression (case = 1, reference control = 0) at the STRADL clinic visit in GS was included as a covariate in all models, due to known selection bias for depression phenotypes in STRADL 57 . Missing data were excluded from lmekin models.
Sensitivity analyses were run for cognitive and imaging phenotypes, with additional adjustments for . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Epigenome-wide association study of protein levels
There were 778 individuals with protein level and DNAm measurements. Prior to running the EWAS, protein levels were log transformed and regressed onto age, sex, study site, lag group and 20 genetic principal components (generated from multidimensional scaling of genotype data from the Illumina 610-Quadv1 array). Residuals from these models were then rank-based inverse normalised.
Methylation data were in M-value format and were pre-adjusted for age, sex, processing batch, methylation set, and depression status in the basic model 62  Omics-data-based complex trait analysis (OSCA) 70 Version 0.41 was used to run EWAS analyses.
Within OSCA, a genetic relatedness matrix (GRM) was constructed for the STRADL population and a threshold of 0.05 was used to identify 120 individuals likely to be related based on their genetic similarity. For this reason, the MOA method was used to calculate associations between individual CpG sites and protein levels, with the addition of the GRM as a fixed covariate to adjust for relatedness between individuals. MOA is a mixed-linear-model-based method that is designed to account for unobserved confounders and the correlation between distal probes that are likely to be introduced by such confounders, by modelling each CpG as a random effect 70 . CpG sites were the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint 1 8 independent variables and the 4,235 proteins were the dependent variables. A threshold for multiple testing correction was applied based on the significance level of P < 0.05 / 4,235 proteins / 772,619 CpG sites, which equalled P < 1.5x10 -11 . Five fully-adjusted models did not converge (NAGLU,

Data availability
Datasets generated in this study are made available in Supplementary Tables. The source datasets from the cohorts that were analysed during the current study are not publicly available due to them containing information that could compromise participant consent and confidentiality. Data can be obtained from the data owners. Instructions for accessing Generation Scotland data can be found here: https://www.ed.ac.uk/generation-scotland/for-researchers/access; the 'GS Access Request Form' can be downloaded from this site. Completed request forms must be sent to access@generationscotland.org to be approved by the Generation Scotland Access Committee.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021.  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint Fig. 2. Epigenome-wide association study of 4,235 plasma proteins a, Genomic locations of CpG sites associated with differential levels of circulating SOMAscan ® proteins. The 597 pQTMs that were significant (P = 1.5x10 -11 ) for 151 of the 153 proteins are shown here. The x-axis represents the chromosomal location of CpG sites. The y-axis represents the position of the gene encoding the associated protein. The 342 cis CpG sites (purple) identified by our EWAS on protein levels lay within 10Mb of the protein-coding gene, whereas the 255 trans CpG sites (green) lay outside of the protein gene. The PGR3 and PAPPA proteins had a combined pQTM total of 2,298 and are not shown in this plot.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint Fig. 3. Plasma protein associations with neurological phenotypes a, Standardised beta coefficients plotted from phenome-wide protein association studies (PheWAS) between 4,235 protein levels and 14 neurologically-relevant phenotypes in Generation Scotland (maximum n=1,095). Negative and positive direction of effects are shown in blue and red, respectively. Associations for 25 unique proteins that were associated (FDR P < 0.05) with both a cognitive and imaging modality and are indicated by an asterisk. b, Four of the 11 protein markers associated with APOE status with FDR P < 0.05. Positive associations (orange) indicate a relationship between APOB levels and the presence of one or more APOE e4 alleles. Negative associations (blue) indicate inverse relationships between protein levels and APOE e4 status. c, A selection of associations (FDR P < 0.05) for the RBL2 and TREM1 protein markers in relation to both cognitive and imaging modalities.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint

Fig. 4. pQTMs for protein markers of neurological phenotypes
Circular plot showing 42 trans pQTM associations between DNAm at CpG sites and the levels of 14 proteins that were associated with one of more of the neurological phenotypes (FDR P < 0.05). Chromosomal positions are given on the outermost circle. Cis pQTM associations are available in Supplementary Fig. 1. Full details of the 88 pQTMs are reported in Supplementary Table 16. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint Fig. 5. Candidate markers revealed through neurological pQTM associations a, Two trans associations between DNAm at cg11294350 in the CHPT1 gene and two proteins with lysosomal-associated function (HEXB and SMPD1) that were associated with relative brain age and processing speed. b, Three trans associations between the region ITIH3/ITIH1/NEK4 on chromosome 3 and the levels of RBL2, which was associated with reductions in both Global Grey Matter Volume and Cognitive Ability (g-score). c, Four trans associations revolving around the CpG site cg06690548 in the SLC7A11 gene, which encodes for a synaptic protein involved in glutamate transmission and oxidative stress. The seven trans associations for NLCR5 inflammatory-associated CpG sites are not shown, but can be viewed in Supplementary Fig. 2.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted September 6, 2021. ; https://doi.org/10.1101/2021.09.03.21263066 doi: medRxiv preprint