The genetic regulation of protein expression in cerebrospinal fluid

• DOI: 10.15252/emmm.202216359
• Publication Date: 2022-12-12T08:12:12Z
• Authors: Oskar Hansson; Atul Kumar; Shorena Janelidze; Erik Stomrud; Philip S Insel; Kaj Blennow; Henrik Zetterberg; Eric Fauman; Åsa K Hedman; Michael W Nagle; Christopher D Whelan; Denis Baird; Anders Mälarstig; Niklas Mattsson‐C...
• Keywords: Neurochemistry

Article12 December 2022Open Access Source DataTransparent process The genetic regulation of protein expression in cerebrospinal fluid Oskar Hansson Corresponding Author [email protected] orcid.org/0000-0001-8467-7286 Clinical Memory Research Unit, Faculty Medicine, Lund University, Lund, Sweden Clinic, Skåne University Hospital, Contribution: Conceptualization, Resources, Supervision, Funding acquisition, ​Investigation, Methodology, Project administration, Writing - review & editing Search for more papers by this author Atul Kumar orcid.org/0000-0002-2166-4805 Data curation, Software, Formal analysis, Visualization, Shorena Janelidze Erik Stomrud Philip S Insel Department Psychiatry and Behavioral Sciences, California, San Francisco, CA, USA Kaj Blennow orcid.org/0000-0002-1890-4193 Neurochemistry Laboratory, Sahlgrenska Mölndal, Neurochemistry, Institute Neuroscience Physiology, the Academy, Gothenburg, Henrik Zetterberg Neurodegenerative Disease, UCL Neurology, London, UK Dementia at UCL, Hong Kong Center Diseases, Kong, China Eric Fauman Internal Medicine Pfizer Worldwide Research, Development Medical, Cambridge, MA, Åsa K Hedman Stockholm, Medical Epidemiology Biostatistics, Karolinska Institutet, Michael W Nagle orcid.org/0000-0002-4677-7582 Neurogenomics, Genetics-Guided Discovery, Eisai, Inc, Christopher D Whelan Translational Biology, Biogen Development, Denis Baird Anders Mälarstig Niklas Mattsson-Carlgren orcid.org/0000-0002-8885-7724 Wallenberg Molecular Validation, original draft Information *,1,2, Kumar1, Janelidze1, Stomrud1,2, Insel1,3, Blennow4,5, Zetterberg4,5,6,7,8, Fauman9, Hedman10,11, Nagle12, Whelan13, Baird14, Mälarstig10,11 *,1,14,15 1Clinical 2Memory 3Department 4Clinical 5Department 6Department 7UK 8Hong 9Internal 10Pfizer 11Department 12Neurogenomics, 13Translational 14Department 15Wallenberg *Corresponding author. Tel: [phone]; E-mail: 072 5759329; EMBO Mol Med (2023)15:e16359https://doi.org/10.15252/emmm.202216359 PDFDownload PDF article text main figures.PDF PLUSDownload text, figures, expanded view figures appendix. Peer ReviewDownload a summary editorial decision including letters, reviewer comments responses to feedback. ToolsAdd favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures Info Abstract Studies (CSF) proteins may reveal pathways treatment neurological diseases. 398 CSF were measured 1,591 participants from BioFINDER study. Protein quantitative trait loci (pQTL) identified as associations between variants proteins, with 176 pQTLs 145 (P < 1.25 × 10−10, 117 cis-pQTLs 59 trans-pQTLs). Ventricular volume (measured brain magnetic resonance imaging) was confounder several pQTLs. plasma overall correlated, but CSF-specific also observed. Mendelian randomization analyses suggested causal roles example, ApoE, CD33, GRN Alzheimer's disease, MMP-10 preclinical SIGLEC9 amyotrophic lateral sclerosis, CD38, GPNMB, ADAM15 Parkinson's disease. levels GRN, MMP-10, GPNMB altered respectively. These findings point be explored novel therapies. finding that ventricular confounded has implications design future studies proteome. Synopsis can increase understanding disease mechanisms. This study (pQTLs) analyzed highly specific extension assays large human population. significant identified, most which had not been described previously proteins. When combining results external GWAS data sources experiments, potential diseases, others. pQTL imaging (MRI), ventricle possible some Introduction Cerebrospinal is produced within rich source biomarkers correlate pathologies across different diseases (Hansson, 2021). While control proteome yield insights into mechanisms treatments, only few such have performed date on larger sets (Kunkle et al, 2019; Sasayama Yang 2021), or focused disease-associated (Deming 2017; Maxwell 2018). In contrast, blood consistently shown associated levels, known (pQTL), using both aptamer-based (Sun 2018; Ferkingstad 2021) antibody-based (Folkersen 2020) are common explain up 30% variance. Identification makes it test if likely development (hence nominating them candidate drug targets). For needed rather than identify targets. largest previous used an approach (Yang Recent comparisons aptamer- highlight technology quantification could impact findings, methods being (Katz 2022). Therefore, we attempted discover deeply phenotyped cohort healthy controls patients biomarker capture wide range biological pathways. We mainly proximity (PEA), sensitive (Assarsson 2014; Katz 2022), together selected other assays, orthogonal validation. evaluated (MRI) (due dilution effects) pQTLs, our knowledge done before studies. overarching aim map decipher clinically relevant expressed subsequently secreted CSF, focus general particular. Results Genome-wide analysis reveals independent Fig 1. total, included (53% females), mean age 71.3 (standard deviation 7.1) years. Most cognitively unimpaired (CU) individuals (613 normal 210 subjective cognitive decline [SCD]), remaining 280 mild impairment (MCI) patients, 189 (AD) dementia 155 (PD) 30 Parkinsons' (PDD) 21 progressive supranuclear palsy (PSP) 24 Lewy bodies (DLB) 14 vascular (VaD) 5 frontotemporal lobe (FTD) 27 multiple system atrophy (MSA) 6 corticobasal syndrome (CBS) 11 unspecified parkinsonism, (demographics diagnosis Table EV1). Figure Study flowchart A schematic overview design. Download figure PowerPoint There available (Dataset EV1) genome-wide 10.53 million (with imputation INFO score ≥ 0.6; MAF 1%). After correction Bonferroni method (genome-wide significance P 10−8 divided number tested 1.253 10−10 CSF), found among (Figs 2 EV1), (N = 117) trans-pQTLs 59). 197 141 N 56 trans-pQTLs), 153 Dataset EV2 (together all least [P 10−8] facilitate meta-analyses). distributed assay platforms (Appendix S1). minor allele frequencies inversely correlated effect size (Fig 3A), strengths cis-pQTL decreased longer distance transcription start sites 3B). functional annotation 3C) enrichment (after over categories) intronic (49% vs. 36% reference panel), exonic (1.8% 1.0%), downstream (1.7% 1.1%) variants, while intergenic (35% 47%) ncRNA-intronic (8.9% 11.5%) significantly less common. 2. genomic Each represents protein. sizes bubbles proportional β-coefficient effects. An interactive version plot EV1. 3. mapping Relationships (β) frequency (MAF) indicated (only after included). cis-pQTLs, degree site (TSS). interquartile −37.3 0.12 Kb. annotated. Functional generated FUMA, correction. Enrichment consequences SNPs against European 1,000 genome panel. Asterixes indicate differences proportions versus panel (***P 0.001; *P 0.05). Click here expand figure. Interactive pQTLsThis Full functionality provided file "Fig EV1 CSF_pQTL_interactive.html", online. online adjusted diagnostic group, sensitivity repeated subgroup controls, similar S2). three measures PEA additional (Meso Scale Discovery CCL2 CCL4, ELISA CHI3L1), validated alternative EV2; EV2, lines marked yellow). Our annotate genes distance, through chromatin interaction yielded largely overlapping EV2). EV2. A–D. Between-assay correlations OLINK (proximity assay) methods, four where (panel A: CHI3L1, B: CCL2, C: D: NFL). manuscript. (using labels assays: MCP1 MIP1b YKL-40 rows same variant (rs2228467, trans-pQTL) assays. one (rs113341849) (rs879571071) LD (rs8064426, R2 0.200, D′ 0.687). (rs4950928), high (rs946262, 0.902, 1.0). these cases, similar, stable direction general, (Spearman Rho 0.69, 0.001, EV3). However, out 162 matching plasma, there 74 (46%) (44 43 trans-pQTLs, 39 putative genes) non-significant even level > 10−8) plasma. One example IL-6, (rs79103996) 2.9 10−29) no association 0.38), suggesting tissue-specificity IL-6 regulation. testing tissue two enriched cerebral cortex (OSTN, 34 DRAXIN, 1 cis-pQTL). EV3. plasmaThe shows relationship 5e-8) CSF. differed tissues, CXCL1, lower higher Genetic variant-to-protein specificity Among correction) 140 unique locus, predominantly acting cis 4A, see Appendix S3 pQTLs), trans-acting (for loci, strongest nominated carried further analyses), six CCL4 (two RBKS (all cis-pQTLs), each CCL24 CTSS LRPAP1 Conversely, 4B). notable exception region represented rs71635338 chromosome 3, (see below detailed analysis). noted rs429358 cis-pQTLs). defines APOE genotype, well-known related (AD), (lower Aβ42, T-tau P-tau (Chung 2018)). 4. Overview architecture Number per A), top B). Validation queried EBI-GWAS catalog (September 2022) study, Methods section. Novelty assessed depending whether any prior observed locus. result novelty assessment full hits EV4. Detailed CNS Datasets EV5 EV6, summary, pQTL-protein pairs 19 (10.9%) publications (the reported 82 (47.1%) serum (24.7%) protein, locus (15.5%) 3 (1.7%) completely (no tissue). Another recently published proteomics 971 samples, 275 applied panels overlapped 235 conducted head-to-head comparison results. Fifteen replicated (significant effects respect risk allele) EV4; EV7). Out 15, 12 cis-acting trans-acting. Further, 8 proxy match (up 10 kb) al (2021) directionality studies). (rs76904798) (2021). second (rs12126142) (PSME1) (IL6R). eight (including proteins) studies, reversed part Replication (2021)This current another recent publication EV4 pQTLs.html", Matching co-localization eQTLs To understand mechanism influence investigated mRNA gene expression. All eQTL meta-analysis (Sieberts GTEx database (restricting search tissues), P-value 0.05 (false discovery rate [FDR]-adjusted) denote statistical significance. comparing 53 showed