The role of NOTCH3 variants in Alzheimer's disease and subcortical vascular dementia in the Chinese population

Abstract Aims NOTCH3 gene mutations predominantly cause cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy, a common etiology of subcortical vascular dementia (SVaD). Besides, there may be a pathogenic link between NOTCH3 variants and Alzheimer's disease (AD). We aimed to study the role of NOTCH3 variants in AD and SVaD patients. Methods We recruited 763 patients with dementia (667 AD and 96 SVaD) and 365 healthy controls from the Southern Han Chinese population. Targeted capture sequencing was performed on NOTCH3 coding and adjacent intron regions to detect the pathogenic variants in AD and SVaD. The relationship between common or rare NOTCH3 variants and AD was further analyzed using Plink1.9. Results Five known pathogenic variants (p.R182C, p.C201S, p.R544C, p.R607C, and p.R1006C) and two novel likely pathogenic variants (p.C201F and p.C1061F) were detected in 16 SVaD patients. Additionally, no pathogenic or likely pathogenic variants were found in AD patients. NOTCH3 was not associated with AD in either single‐variant association analysis or gene‐based association analysis. Conclusion Our findings broaden the mutational spectrum of NOTCH3 and validate the pathogenic role of NOTCH3 mutations in SVaD, but do not support the notion that NOTCH3 variation influences the risk of AD.


| INTRODUC TI ON
Dementia is an acquired cognitive impairment syndrome, along with a decline in occupational and social functioning. As the most populous country in the world, China has the largest population with dementia. 1 Alzheimer's disease (AD) is the most common dementia type, accounting for 2/3 of the dementia cases worldwide, and vascular dementia (VaD) is the second most common type. 2,3 VaD is mainly divided into multi-infarct dementia, strategic infarct dementia, and subcortical VaD (SVaD) subtypes according to different pathogenesis. Compared with multi-infarct dementia and strategic infarct dementia, SVaD often has an insidious onset, representing a more homogenous group. 4 Both AD and SVaD present as cognitive decline in adults with insidious onset. Although the pathogenesis and pathology are different, AD and SVaD share many risk factors, including advancing age, and genetic and vascular risk factors.
Vascular dysfunction also plays an important role in the pathogenesis of AD. 5 Some pathology studies have shown that most patients with dementia had mixed pathologies, most commonly AD and vascular disease. 6 NOTCH3 encodes a transmembrane receptor mainly expressed in vascular smooth muscle cells (VSMC) and pericytes. The NOTCH3 protein is composed of an extracellular domain (ECD), a single transmembrane domain, and a noncovalently bound intracellular domain (ICD), and the ECD contains 34 tandem epidermal growth factorlike repeat (EGFr) domains and three NOTCH Lin repeats. A pathogenic mutation in NOTCH3 was found to cause cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), which is clinically characterized by migraine, recurrent subcortical strokes, and vascular cognitive decline or VaD in adults. 7 CADASIL has been confirmed to be a common form of hereditary subcortical vascular cognitive impairment. 8 Currently, more than 200 pathogenic mutations in the NOTCH3 have been reported to cause CADASIL, and the majority are missense mutations distributing in exons 2-24, changing the number of cysteines in the EGFr domains, resulting in an odd number of cysteines in the ECD of the receptor, causing incorrect protein folding and aggregation. 9,10 Besides, an increasing number of NOTCH3 missense mutations that did not affect the number of cysteines in EGFr domains were identified in suspected CADASIL patients, but the pathogenic role of these mutations was controversial because of undefined mechanisms. 9,11 Although CADASIL mainly manifests as vascular cognitive impairment, it may also present as an AD phenotype, with one Turkish patient clinically diagnosed with AD being found to carry a pathogenic mutation in the NOTCH3. 12 Besides, NOTCH3 was found to be associated with AD in a British and North American cohort, 13

| Bioinformatics processing and variants analysis
The raw fastq sequences were aligned to the human genome v19 reference sequence with BWA. 16 Variants were called with GATK 17 and annotated with Annovar. 18  and Reve. 19 All the pathogenic or likely pathogenic mutations in the NOTCH3 gene were verified by Sanger sequencing. In the analysis of the association between dementia and NOTCH3, participants with a pathogenic or likely pathogenic mutation in NOTCH3 were excluded.
The flow chart of this study is shown in Figure 1. Due to the low coverage of exon 24, we performed Sanger sequencing on exon 24 for all subjects. Variants with minor allele frequency <0.01 in the gno-mAD East Asian population, ExAC East Asian population, and 1000 Genomes database were defined as rare variants, and the remaining variants were common variants.

| Statistical analysis
For common variants, Plink1.9 and logistic regression analysis (gender + age + APOE ε4 as covariates) were used to study the asso-

| Quality control and variants identification of NOTCH3
The average coverage rate of the target region of NOTCH3 was 98.47%, and the average sequencing depth was 138.96×. Most of the encoding sequences have a coverage depth of at least 10×, except for parts of exons 1, 24, and 33. After filtering out variants of F I G U R E 1 Flow chart of NOTCH3 variants analysis in Chinese patients with AD and SVaD low quality (coverage < 10× or variants supporting depth accounts for the total depth <25%), 344 variants were identified by targeted sequencing (Table S1).
As exon 24 of NOTCH3 is a hot spot region of pathogenic mutations, we performed Sanger sequencing of this exon in each sample to avoid false negatives. Furthermore, two rare variants were found in ten subjects by complementary Sanger sequencing, including p.P1354L in an AD patient and p.G1347R in nine subjects (seven AD, one SVaD, and one control). In the end, a total of 345 variants were identified by the targeted and Sanger sequencing.

| Phenotypes of the patients with a pathogenic or likely pathogenic mutation in NOTCH3
Five known CADASIL pathogenic mutations in the NOTCH3 gene The clinical and imaging data of 16 unrelated patients (one males and six females) with a pathogenic or likely pathogenic mutation are summarized in Table 3

| Phenotypes of participants with a cysteine-sparing missense mutation in EGFr domains of NOTCH3
Although most of the pathogenic mutations were located in the 34 EGFr domains and caused CADASIL by affecting the number of cysteines in EGFr domains, a few cysteine-sparing mutations in EGFr domains were also identified in suspected CADASIL patients.
As shown in Table 4, 33 cysteine-sparing NOTCH3 missense variants encoding for EGFr domains were found in our cohort, and all were  Figure   S1A-D) underwent a skin biopsy, but no GOM was observed in the subcutaneous arterioles by electron microscopy ( Figure S1E).

| Cysteine-sparing NOTCH3 mutations in SVaD patients without cysteine-altering NOTCH3 variant
Among the 80 SVaD patients without the cysteine-altering NOTCH3 variant, six carried a cysteine-sparing NOTCH3 mutation. Meanwhile, among the 365 healthy controls, 26 presented the cysteine-sparing NOTCH3 mutations (Table S2). Therefore, the cysteine-sparing NOTCH3 variants were not enriched in the SVaD patients without cysteine-altering NOTCH3 variant when compared to controls.

| Association analysis between NOTCH3 variants and AD
In the single-variant-based analysis, 53 common variants met the criteria for further association analysis: genotyping rate >80%   The most common pathogenic mutation in our cohort was p.R544C, followed by p.R607C. Nevertheless, a recent study investigating NOTCH3 variants in 261 Chinese patients with clinically suspected CADASIL showed that p.R607C and p.R544C were the first and second most common variants, respectively. 30 The discrepancy in the mutational spectrum may be explained by the phenotypes difference of different mutations. Since patients with p.R607C tend to present a more classical CADASIL phenotype than those with p.R544C, and in our study we have only included patients with dementia caused by SVaD and AD. The seven probands carrying the p.R544C mutation started with cognitive decline after the age of 60, except for patient 6, and none of them experienced a migraine.
These results were consistent with previous findings indicating that patients with the p.R544C mutation predominantly present a lateonset disease with a mild and atypical phenotype, such as the rare occurrence of migraine and a low frequency of WMH in the temporal pole 33 ; however, the specific mechanism behind these observations has not been identified. Some studies speculate that the different effects of the mutations on NOTCH3 signaling may be involved in modifying the CADASIL phenotype. 34 The majority of the mutations in our cohort were in exon 11 of the NOTCH3 gene, in accordance with previous studies indicating that exon 11 is a hot region in the Chinese population, 30 while exon 4 is currently the most common mutated exon in the literature. 9,35 Our findings showed that SVaD patients with NOTCH3 mutations had an earlier age of onset and a higher frequency of cognitive impairment in their family history than those without NOTCH3 mutations. However, a previous study did not find differences in the clinical features of patients with and without NOTCH3 variants that presented subcortical vascular impairment. We speculated that the differences observed were mainly caused by cysteine-altering AD and VaD are the two most common types of dementia, 37 and CADASIL is a common cause of VaD. Although the pathogenesis was different, AD and CADASIL had many similarities in pathology. Both AD and CADASIL are characterized by the aggregation of abnormal proteins, and in both cases, the deposited proteins were cleaved by gamma-secretase. 38 Although recent studies showed that variations in the NOTCH3 gene were correlated with AD in the western population, we found that neither common variants nor rare missense variants in NOTCH3 were associated with AD in the Chinese population.
The possible reasons are as follows: First, the relatively small sample size in our research (only 365 controls) may lead to a negative result in association analysis, particularly for rare variants; second, the inclusion and exclusion criteria varied among different studies; third, racial differences may account for the inconsistencies in the association analysis between NOTCH3 gene variants and AD; and finally, our targeted capture sequencing mainly focused on variants in the coding regions and adjacent intron regions of NOTCH3, while most known AD risk loci are in non-coding regions. Therefore, further investigation is necessary to study the relationship between NOTCH3 and AD in a larger cohort, including variants in both the coding and non-coding regions.
In conclusion, this is the first systematic study of NOTCH3 variants in a large Chinese cohort of AD and SVaD, and two novel likely pathogenic mutations, p.C201F and p.C1061F, were identified.
Pathogenic mutations in NOTCH3 are relatively common in SVaD patients. In addition, we believe that six cysteine-sparing mutations previously reported in CADASIL patients were rare polymorphisms rather than CADASIL causative mutations, suggesting that the pathogenicity of cysteine-sparing mutations in NOTCH3 should be interpreted prudently. Our association study showed that NOTCH3 was not related to AD in the Chinese population whether from the perspective of common variants or the perspective of gene-based rare mutations; more studies are necessary to further elucidate this finding.