Novel Rare SORL1 Variants in Early-Onset Dementia

INTRODUCTION

Pathogenic variants in the genes coding for amyloid precursor protein (APP), presenilin-1 (PSEN1), and presenilin-2 (PSEN2) have been shown to be the cause of autosomal dominant early-onset Alzheimer’s disease (EOAD) in 5–10%of the cases [1]. In the pathophysiology of AD, the key elements are the accumulation of extracellular senile plaques consisting of aggregates of amyloid-β (Aβ) peptides [2, 3] and the intracellular neurofibrillary tangles of hyperphosphorylated tau protein [4]. Aβ is a product of cleavage of amyloid-β protein precursor (AβPP), and presenilins are part of the γ-secretase that does the cleavage. This process, called the amyloid cascade hypothesis [2], leads to neuronal degeneration over the years or decades and clinically leads to memory impairment, cognitive dysfunction, and dementia.

The ɛ4 allele of the apolipoprotein E (APOE ɛ4) gene is the main genetic risk factor for late-onset AD (LOAD) [5]. One allele increases the risk of AD by 3-fold and two alleles by 12-fold [6]. In recent years, genome-wide association studies (GWAS) have found over 50 other genetic loci which show association with LOAD [7, 8], revealing that LOAD is highly polygenic. The functionality of most of the associated genes can be categorized within three major pathways: inflammatory response, lipid metabolism, and endocytosis. Understanding the role of these risk factors is under active investigation. Rare variants widespread in the genome could explain the missing genetic components for complex neurodegenerative diseases. The common variants found by GWAS do not explain all of the heritability of AD indicating that there are still more associations to be found, especially rare variants (minor allele frequency (MAF) < 0.01) [9].

Rare variants associated with AD have been found especially in genes like TREM2, ABI3, and SORL1 [10–14]. A rare variant of TREM2, p.Arg47His, was first found to be strongly associated with AD in two separate studies [11, 12]. Furthermore, an association of the TREM2 p.Arg47His variant with EOAD has also been reported [15]. One of the key features of the AD neuropathology is local chronic neuroinflammation around the senile plaques [16, 17]. The Aβ deposits trigger an innate immune activation, resulting in the release of inflammatory cytokines and free radicals from the microglia, the resident macrophages of the brain. Several of the AD risk genes are highly expressed in the microglia. Examples of immune-related genes include TREM2, CD33, ABCA7, CR1, PLCG2, and MS4A [18]. The TREM2 p.Arg47His variant increases the risk of AD comparable to the risk increase of one APOE ɛ4 allele [19]. Other rare variants of TREM2 with a significant risk of AD are, e.g., p.Arg62His and p.His157Tyr [10, 19]. A rare variant of PLCG2, p.Pro522Arg, reduces the risk of multiple neurodegenerative diseases, including AD [10, 20].

The sortilin related receptor 1 (SORL1, also known as sorLA or LR11) gene is located in 11q23-q24. It encodes an intracellular transmembrane protein expressed predominantly in neurons. One of its functions is to traffic other proteins to their correct locations [21, 22]. The SORL1 expression has been shown to be decreased in the cortex and hippocampus of sporadic AD patients [23]. Later, SORL1 has been proposed to be responsible for intracellular trafficking of AβPP between Golgi and endosomes/lysosomes [24, 25]. Overexpression of SORL1 has been shown to reduce the Aβ production in the cells [24, 25] and, on the other hand, loss of SORL1 increases the Aβ formation [24]. Later it was elucidated that the SORL1 protein reduces the amyloidogenic processes in two ways. First of all, it directs internalized AβPP to trans-Golgi network and slows down its exit from the Golgi, thus preventing its cleavage by secretases [25, 26]. Secondly, it targets Aβ for degradation in the lysosomes [27].

The SORL1 gene has been associated with both EOAD and LOAD [13, 28–31]. Many potentially disease-causing rare variants have been found, but their impact on AD pathogenesis is still uncertain. In this study, the aim was to research the presence of rare variants in SORL1 in a Finnish cohort of sporadic and familial EOAD patients.

MATERIALS AND METHODS

The cohort consists of 115 (59%women) EOAD patients (age at onset (AAO) before 65 years) (see Table 1). All the patients were examined and diagnosed by experienced neurologists specialized in memory disorders at the memory outpatient clinic of Oulu University Hospital in Northern Finland. The diagnoses were made according to the current diagnostic criteria for AD [32]. All the patients underwent clinical and neurological examinations, routine screening laboratory tests, a neuropsychological examination, and magnetic resonance imaging (MRI) of the brain or rarely computed tomography (CT) of the brain if MRI was contraindicated. When needed, cerebrospinal fluid analyses of the biomarkers Aβ₄₂, tau, and phospho-tau and/or functional neuroimaging by fluorodeoxyglucose positron emission tomography (FDG-PET) were performed for the patients to confirm the diagnosis. The mean age at onset was 58.3 years (range 46–65 years). A positive family history of dementia (at least one first-degree relative with dementia) was present in 33%(n = 38) of the patients, 32%(n = 37) had a negative family history, while in 35%(n = 40) of the patients, family history was unknown. The study was performed according to the principles of the Declaration of Helsinki. Written informed consent was obtained from all the patients or their caregivers. The research protocol was approved by the Ethics Committee of the Oulu University Hospital.

Genomic DNA was extracted from the peripheral blood samples by standard methods. The APOE genotypes were determined by HhaI digestion [33] and confirmed by examining the alleles rs7412 and rs429358 in the exome data.

Whole-exome sequencing (WES) data were generated at the Broad Institute, the Baylor College of Medicine’s Human Genome Sequencing Center, and Washington University’s McDonnell Genome Institute [34]. Multiallelic sites were converted into biallelic variants and then left-aligned using bcftools 1.10. Variants with variant quality score log-odds (VQSLOD) less than –2 or inbreeding coefficient less than –0.3 were excluded. All variants included in the final analysis had genotype read depth at least 10 and genotype quality at least 25. PLINK 1.9 [35] was used to estimate relatedness between individuals. All pairs of individuals had PI_HAT statistic less than 0.185.

The SORL1 variants were annotated with the Variant Effect Predictor (VEP) tool release 97 in Ensemble database [36]. We extracted all rare SORL1 variants with global MAF less than 1%in the Genome Aggregation Database (gnomAD) v2.1.1 non-neuro dataset [37]. The non-neuro dataset is a subset of gnomAD, which contains samples only from individuals who have been ascertained for not having a neurological condition in a neurological case/control study. The variants which were not annotated as missense, stop-gain, or splice region variants were excluded from the study. All included variants were verified by Sanger sequencing (ABI3500xL Genetic Analyzer, Applied Biosystems). The pathogenicity of the variants was evaluated using PolyPhen-2 (HumVar) [38], MutationTaster [39], SIFT 4G [40], and Combined Annotation Dependent Depletion (CADD) GRCh37-v1.6 [41] in silico prediction programs. Odds ratios were calculated between the cohort samples and the Finnish gnomAD non-neuro samples. Odds ratios and p values were calculated with R package exact2x2 using two-tailed Fisher’s exact test with matching confidence intervals as detailed by Fay [42]. Analyses of p values were not adjusted for multiple testing.

RESULTS

We detected 46 SORL1 variants in our EOAD cohort: one stop-gain, 12 missense, 10 synonymous, and 23 intronic variants. Nine variants had global MAF less than 0.01 in gnomAD: one stop-gain (p.Gln290^*) and eight missense variants (p.Lys80Arg, p.Asp734Asn, p.Ala789Val, p.Args866Gln, p.Phe1099Leu, p.Arg1470Leu, p.Asn1809Ser, and His1813Gln). These rare variants were detected in 13 EOAD patients (11.3%) (Table 2). The clinical characteristics of the carriers are shown in Table 3. Nine of them had at least one APOE ɛ4 allele (three homozygotes). One patient carried two rare SORL1 variants, p.Arg866Gln and p.Phe1099Leu. Other patients carried only one rare variant. Four of the rare variants identified in our EOAD cohort have not been previously reported in AD patients: the protein-truncating variant p.Gln290^* and three missense variants: p.Lys80Arg, p.Ala789Val, and p.Arg866Gln. Summary of all SORL1 variants found in our EOAD cohort can be found in Supplementary Table 1. Known pathogenic variants in APP, PSEN1, and PSEN2 were absent in the patients with SORL1 variants.

DISCUSSION

In this study, we report the allele frequencies and clinical manifestations of SORL1 variants in a Finnish cohort of early-onset AD patients. We detected one nonsense variant (p.Gln290^*) and two missense variants (p.Ala789Val and p.Arg866Gln) which are overexpressed in our cohort. The nonsense variant p.Gln290^* in the VPS10p domain of SORL1 protein is a novel finding. The allele frequency for p.Gln290^* could not be found in the Finnish SISu database [43], the Exome Aggregation Consortium (ExAC) database [44], or gnomAD. It is expected to disrupt the SORL1 protein function completely, as it is located so near to the start of the protein structure. The patient carrying this variant had early-onset dementia of AD type; however, he also had features of vascular dementia. Pathogenic APP, PSEN1, and PSEN2 variants were excluded. There was a strong family history of dementia in this family and the mother and five siblings of the proband had also dementia. Unfortunately, since no material for genetic testing was available from the relatives of this carrier, we are unable to confirm the segregation in the family. The p.Gln290^* meets the American College of Medical Genetics and Genomics (ACMG) criteria for classifying as pathogenic based on 1 very strong and 1 moderate and 1 supporting criterion [45]. These criteria are PVS1 (null variant in a gene where loss-of-function (LoF) is a known mechanism of disease), PM2 (absent from controls), and PP3 (multiple lines of computation evidence support a deleterious effect on the gene or gene product). Protein-truncating SORL1 variants have been previously shown to be pathogenic [46].

Some SORL1 variants seem to be specific to the Finnish population. This is the first study reporting the p.Lys80Arg SORL1 variant in two EOAD patients. In gnomAD non-neuro dataset this variant has been found seven times and only in the Finnish population. Even so, all four in silico tools categorize this variant as a polymorphism. The same applies to the variant p.Ala789Val, which appears 25 times only in the Finnish population in gnomAD non-neuro dataset. The p.Arg866Gln SORL1 variant was found in one of our EOAD patients. This variant is rare, appearing three times in the Finnish population and one time in the non-Finnish European population in gnomAD (non-neuro). All four in silico prediction tools classify this variant as damaging.

Four variants found in our EOAD cohort have been previously reported. We found the p.Asp734Asn SORL1 variant in two EOAD patients (2/115, MAF = 0.00870). The MAF of this variant in our cohort does not differ from the MAF (0.00715) of the Finnish gnomAD (non-neuro) samples. Previously, Verheijen et al. have reported the p.Asp734Asn variant in four EOAD patients (4/1255, MAF = 0.00159) and three controls (3/1938, MAF = 0.00077) [31]. Fernandez et al. found this variant in familial LOAD cases (5/875, MAF = 0.00286) but not in controls (n = 763) or sporadic EOAD (n = 217) or LOAD patients (n = 134) [47]. All four prediction tools predict this change to be damaging, but the current evidence suggests that this variant is not pathogenic. Variants p.Phe1099Leu, p.Asn1809Ser, and p.His1813Gln have been previously identified both in AD patients and controls [31, 48]. These variants are predicted to be damaging at most by one of the four prediction tools. They are unlikely to be pathogenic. Furthermore, variant p.Asn1809Ser has been reported not to segregate with AD in a family [49]. One affected family member did not carry the variant and one healthy family member was a carrier [49]. p.Arg1470Leu has been previously reported in one Swedish female control, who had no dementia at age of 61 years [31]. This is another Finnish-dominated variant; of the 9 occurrences reported in gnomAD (non-neuro), 8 are reports in the Finnish population. It is enriched in our patients (OR = 9.7), but it does not reach statistical significance.

Association between SORL1 and sporadic and late-onset AD has been shown in several association studies [50–59], meta-analyses [28–30], and GWAS [60–62]. There is also growing evidence that SORL1 variants explain part of the EOAD cases. Pottier et al. detected two nonsense and five missense SORL1 variants in 29 autosomal dominant EOAD families in which APP, PSEN1, and PSEN2 variants had earlier been excluded [14]. Co-segregation analysis was possible only in one family in which p.Gly511Arg SORL1 variant was also carried by the index case’s mother. Later it was shown that this variant leads to loss of Aβ binding [27]. Rare missense variants of SORL1 have been reported to be enriched (OR = 5.03) in EOAD patients [13]. Verheijen et al. sequenced the coding region of SORL1 of 1255 EOAD patients [31]. They detected 84 non-synonymous rare variants of which 36 were only detected in patients. Rare non-synonymous variants were enriched in patients by 1.5-fold. Similar findings have also been reported in other studies [49, 64]. Protein-truncating variants due to frameshift and nonsense variants have been found only in patients. Still, relatively little is known about individual variants, their effect on AβPP metabolism and pathogenicity.

The strength of this study is a clinically well-characterized cohort from a defined geographic area and genetically rather solid population. A limitation is the lack of neuropathological confirmation of the diagnoses. A few patients did not have a pure AD. This raises the question of whether the SORL1 variants were responsible for AD or other dementia characteristics. The gnomAD non-neuro dataset contains controls without neurological conditions, but because some of them have been examined before 65 years of age, they might have developed dementia linked to SORL1 later in life. The small size of the cohort is another limitation, and the statistical significance of p values disappears after Bonferroni correction for multiple comparisons (p < 0.00625). Despite the rather small number of cases, we identified several rare variants in our cohort.

In conclusion, we found a new pathogenic SORL1 variant p.Gln290^* from an EOAD patient with a strong family history of dementia. In addition, many SORL1 rare variants are more common in Finnish EOAD patients than in controls. The three other identified variants seem to be specific to the Finnish population, but several other previously identified rare variants were also detected. This study strengthens the earlier findings, that rare variants of SORL1 are associated with EOAD and that they play a role in neurodegeneration.