Common Variants Near ZIC1 and ZIC4 in Autopsy-Confirmed Multiple System Atrophy

Patient Recruitment

Ethical approval had been obtained from all responsible ethics committees. All participants had given written consent.

Neuropathologists at each recruitment site (Table 1) based the definite neuropathological diagnosis of MSA on histopathological criteria, taking into account glial cytoplasmic inclusions immunoreactive for α-synuclein in characteristic anatomical distribution as a defining feature of MSA.²³ Age, sex, disease history (including disease onset and duration), and neuropathological findings were recorded in a standardized manner for all cases.

Controls were ethnically matched to cases and either derived from biobanks KORA-gen²⁴ or popGen²⁵ (Europe sites) or from a North American site (Alzheimer’s Disease Genetics Consortium).²⁶ The Alzheimer’s Disease Genetics Consortium assembled and genotyped DNA from subjects enrolled in the 29 NIA-Alzheimer’s Disease Centers located across the United States. For this study, the Alzheimer’s Disease Genetics Consortium provided a subset of mostly clinical, cognitively normal controls. Patients and controls were of North-Western European and African American ancestry.

DNA Extraction

We isolated DNA from 30 mg frozen cerebellar cortex using QIAamp DNA Mini Kit (Qiagen, Venlo, the Netherlands). DNA extraction was performed at German Center for Neurodegenerative Diseases (DZNE, Munich, Germany). DNA was stored at −80°C until use. DNA concentration was measured using a NanoDrop Spectro-photometer. DNA quality was determined by gel electrophoresis.

Genotyping

All samples were genotyped on Infinium Global Screening Arrays (Illumina, San Diego, CA, USA). The cases were genotyped at the Institute of Clinical Molecular Biology, Kiel University, Germany. The samples were genotyped in one batch on array version 2.0 for cases and version 1.0 for controls. Genotypes were called using Illumina Genome studio according to the manufacturer’s instructions using in-house cluster files.

Quality Control and Imputation

We used PLINK (v. 1.9) [1] and R (v. 3.6.3)²⁷ for all analyses. Only variants successfully genotyped in both the patient and the control populations were included in the subsequent analyses. Variants with multicharacter allele codes, insertions, deletions, duplicated markers, and all A/T and G/C variants were excluded. We excluded all samples discordant between reported and genotypic sex. Missing sex was imputed, and samples with ambiguously imputed sex were discarded. After a first step of filtering out samples and variants with call rate of less than 85%, we excluded variants with an individual call rate of less than 98% in a second filtering step. Next, we removed variants with a minor allele frequency <0.01, a significant deviation from Hardy–Weinberg equilibrium (P < 1 × 10⁻⁶) in controls, or informative missingness (P < 1 × 10⁻⁵). Subsequently, we excluded individuals with a variant call rate of <98% or an outlying heterozygosity rate (mean ± 3 standard deviations). We used a pruned dataset containing only markers in low linkage-disequilibrium regions (pairwise r² < 0.2) to test for duplicated individuals and cryptic relatedness (Pihat > 0.125) using pairwise genome-wide estimates of the proportion of identity by descent. For each detected sample pair we excluded the individual with a lower call rate. Ethnical outliers were identified by a principal-component analysis (PCA) together with the publically available 1000 Genomes data with known ethnicities.²⁸ Because the study population has genetically a mainly European ancestry, as ascertained by the PCA, we determined a European center and excluded samples more than 1.5 times the maximal European Euclidean distance away from this center. After a first association analysis of genotyped single nucleotide polymorphisms (SNPs) only, we inspected visually the cluster plots of all variants with a P value <1 × 10⁻⁵ and discarded variants without adequate cluster separation. Imputation was carried out on the quality-assured dataset using the TOPMed Imputation Server, which employs Eagle2 for phasing and minimac4 for the imputation of genotypes.^29,30 The most likely genotype is used in downstream analyses. Variants were again filtered for minor allele frequency and deviation from Hardy–Weinberg equilibrium in controls with the same thresholds as before. In addition, SNPs with an imputation quality score R² < 0.7 were excluded, leaving 8,131,900 variants for analyses. As a final step of the quality-control procedure, we used the R package PCAmatchR to ethnically match cases to controls with a 1:4 ratio to overcome possible difficulties with population stratification, leading to 3240 individuals for the analyses.³¹

Association Analysis

We used logistic regression to test the additive genetic model of each marker for association with disease status. Following scree plot analysis, we incorporated the first two dimensions of the PCA and sex as covariates. We used a genome-wide significance threshold of P < 5 × 10⁻⁸ and the threshold of P < 5 × 10⁻⁶ for suggestive association. Conditional analyses, including, in turn, each SNP with a suggestive association as additional covariate, were conducted to identify adjacent independent signals. Furthermore, we tested for clumps of correlated SNPs, ie, to assess how many independent loci had been associated, and determined the number of variants supporting the lead SNP at each locus, ie, variants with P values less than the clumping threshold of 5 × 10⁻⁵ are in linkage disequilibrium (r² ≥ 0.4) and not farther than 250 kb away from the respective SNP. Visualization of the results was carried out with R and LocusZoom³² for regional plots. Variant positions in this article are reported on human genome version 38 (GRCh38/hg38).

Immunohistochemistry on MSA Patients’ Brain

Formalin-fixed and paraffin-embedded (FFPE) tissues from patients with MSA and controls without neurological or psychiatric diseases were obtained from the Neurobiobank Munich (Germany). All autopsy cases of the Neurobiobank Munich were collected on the basis of an informed consent according to the guidelines of the ethics commission of the Ludwig-Maximilians-University (Munich, Germany; #345–13). MSA cases had been diagnosed according to established histopathological diagnostic criteria.^10,23

For ZIC4 immunohistochemistry, 5-μm-thick sections of FFPE tissues of the frontal cortex and the cerebellar hemisphere, including the dentate nucleus, were prepared. After deparaffinization, heat-induced epitope retrieval was performed in Tris/EDTA, pH 9, at 95°C for 30 minutes. For blocking of endogenous peroxidase and unspecific protein binding, the sections were incubated with 5% H₂O₂ in methanol for 20 minutes and I-Block reagent (Applied Biosystems, Waltham, MA, USA) for 15 minutes. Subsequently, ZIC4 primary antibody (rabbit, polyclonal; Merck/Sigma-Aldrich, Darmstadt, Germany) was applied overnight at 4°C at a dilution of 1:100. Signal detection was performed using the DCS ChromoLine DAB kit (DCS, Hamburg, Germany) according to the manufacturer’s instructions. Sections were counterstained for 1 minute with Mayer’s hemalum solution (Waldeck, Münster, Germany).

To determine the fractions of ZIC4-positive neurons of all neurons in the dentate nucleus, we scanned stained slides using a slide scanner (Axio Scan. Z1; Zeiss, Oberkochen, Germany) and visualized using the free ZEN lite software (v. 3.3; Zeiss). For statistical evaluation of the data, Student t test was used, and statistical significance was defined as P < 0.05.

Patient Sample

From the initial sample of 731 cases, 13 cases had to be excluded because of insufficient tissue quality. After thorough quality control and filtering, 648 cases and 2592 controls covering 8,131,900 variants were included in the association analysis (Fig. 1). The number of excluded samples and variants in each step of the quality-control procedure is shown in Tables S1 and S2.

Association Results

We performed logistic regression incorporating sex and determined the first two dimensions of PCA as covariates using the scree plot method. The genomic inflation factor of λ = 1.01 (unimputed λ = 1.01; Fig. S1) indicates that no significant population stratification was present (Fig. 2A). We did not identify any disease-associated variants with a P value less than the genome-wide significance threshold of P < 5 × 10⁻⁸, but suggestive associations with P < 5 × 10⁻⁶ at 10 different loci (Fig. 2B) with the leading SNP at each locus shown in Table 2. Conditional analyses, including, in turn, any SNP with P < 5 × 10⁻⁶, excluded the presence of multiple independent signals at each locus. All variants with suggestive associations are listed in Table S3. The most noteworthy hits were rs16859966 on chromosome 3 (P = 8.6 × 10⁻⁷; odds ratio [OR], 1.58; 95% confidence interval [CI]: 1.32–1.89), rs7013955 on chromosome 8 (P = 3.7 × 10⁻⁶; OR, 1.8; 95% CI: 1.40–2.31), and rs116607983 on chromosome 4 (P = 4.0 × 10⁻⁶; OR, 2.93; 95% CI: 1.86–4.63), which were supported by at least one additional genotype, as well as several imputed SNPs with P values less than the clumping threshold of 5 × 10⁻⁵ as discovered in the clumping analysis (Table 2). The genes closest to the chromosome 3 locus are the Long Intergenic Non-Protein Coding RNA 2032 (LINC02032) approximately 100 kb downstream and the zinc-finger proteins of cerebellum 1 and 4 genes (ZIC1, ZIC4), located roughly 600 kb upstream (Fig. 2E). The top SNP rs7013955 on chromosome 8 maps to the lysyl oxidaselike 2 gene (LOXL2; Fig. 2D). The association signal around SNP rs116607983 on chromosome 4 is located in a region devoid of protein-coding genes approximately 2000 kb to either side (Fig. 2E). A fourth locus on chromosome 5 (rs2279135) was also supported by multiple clumped SNPs, but all SNPs, including the lead SNP, were imputed (Table 2). Several variants clumped at the chromosome 5 locus were located in the ARHGEF37 gene, coding for Rho Guanine Nucleotide Exchange Factor 37 (Fig. 2F). None of the identified SNPs is an expression quantitative trait locus in brain tissues according the Genotype Tissue Expression project.³³ At four of the six remaining loci with variants exhibiting suggestive associations, at most two supporting SNPs were present, which were all imputed; in the other two loci, no supporting SNPs could be found in the clumping analysis (Table 2, Fig. S2). We did not investigate these loci further because it is unlikely that they represent valid associations. No significant associations with Bonferroni-adjusted P values were detected with previously reported Parkinson’s disease associations from a meta-analysis of 17 datasets from a Parkinson’s disease GWAS (Table S4).³⁴

ZIC4 Immunohistochemistry on MSA Patients’ Brain

ZIC4 and ZIC1 are known to play a critical role in the embryonal development of the cerebellum. Heterozygous deletions comprising the ZIC1 and ZIC4 locus have been associated with the Dandy–Walker malformation, a rare congenital condition characterized by a hypoplastic cerebellar vermis and an enlarged fourth ventricle.^35,36 In mice, deletions of ZIC1 and ZIC4 lead to a striking phenotype similar to the Dandy–Walker malformation with cerebellar hypoplasia and foliation defects.^35,36 In addition, paraneoplastic autoantibodies against ZIC4 protein are linked to severe cerebellar dys-function and degeneration.^37,38

Because cerebellar degeneration and corresponding symptoms are also a central hallmark of MSA, we decided to follow up on a potential role of ZIC4 in MSA patient brains by performing immunohistochemical stainings. For ZIC1, no primary antibody was appropriately sensitive and specific on human tissue in our hands. Thus, FFPE tissues of the cerebellum and, for comparison, the frontal cortex of patients with MSA (n = 10 SND, n = 14 OPCA/mixed phenotype) and healthy controls (b = 5) were stained with antibodies raised against ZIC4.

Nuclear and cytoplasmic staining of frontal cortex neurons was observed in all brains examined without differences between healthy controls and patients with MSA (Fig. 3A–C). In the cerebellar dentate nucleus, we found strong expression of ZIC4 in a subset of neurons in healthy controls, as well as patients with MSA with predominant SND (Fig. 3D,E,G,H). In contrast, patients with MSA with mixed subtype or OPCA showed reduced numbers of ZIC4-positive neurons, which were furthermore only weakly stained (Fig. 3F,I). Quantification of the proportions of ZIC4-positive neurons among the total number of dentate nucleus neurons depicted relatively constant proportions in healthy controls and patients with MSA-SND (33.2% ± 0.0% vs 32.6% 0.0%), whereas in patients with MSA-OPCA or MSA-mixed phenotype, we found significantly lower percentages of ZIC4-positive neurons (15.5% ± 0.1%) (Fig. 3J).