Two Rare Variants in PLAU and BACE1 Genes—Do They Contribute to Semantic Dementia Clinical Phenotype?

We have performed whole-genome sequencing to identify the genetic variants potentially contributing to the early-onset semantic dementia phenotype in a patient with family history of dementia and episodic memory deficit accompanied with profound semantic loss. Only very rare variants of unknown significance (VUS) have been identified: a nonsense variant c.366C>A/p.Cys122* in plasminogen activator, urokinase (PLAU) and a missense variant c.944C>T/p.Thr315Met in β-site APP-cleaving enzyme 1 (BACE1)—along with known disease-modifying variants of moderate penetrance. Patient-derived fibroblasts showed reduced PLAU and elevated BACE1 mRNA and protein levels compared to control fibroblasts. Successful rescue of PLAU mRNA levels by nonsense-mediated mRNA decay (NMD) inhibitor (puromycin) confirmed NMD as the underlying mechanism. This is the first report of the PLAU variant with the confirmed haploinsufficiency, associated with semantic dementia phenotype. Our results suggest that rare variants in the PLAU and BACE1 genes should be considered in future studies on early-onset dementias.


Introduction
Clinical diagnosis of dementia is often difficult and inconclusive due to overlapping clinical presentations and neuropathology [1,2]. Moreover, the implementation of nextgeneration sequencing methods revealed a complex genomic nature of disease phenotypes, challenged the classical definition of genetic causality, and the concept of strictly monogenic disorders, shifting the pathogenicity model from monogenic to oligogenic inheritance [3,4].
Here we present a patient with the phenotype of early-onset semantic dementia. The diagnosis of atypical Alzheimer's disease (AD) was also considered due to episodic memory deficit accompanying profound semantic loss. Whole-genome sequencing (WGS) analysis did not reveal any pathogenic/likely pathogenic variants, as defined by The American College of Medical Genetics and Genomics (ACMG) criteria. Thus, we have selected the most promising variants of unknown significance (VUS) in PLAU (Plasminogen Activator, Urokinase) and BACE1 (β-site APP-cleaving enzyme 1) genes for which we performed functional in vitro analysis and discussed their putative contribution to the observed clinical phenotype.

Whole-Genome Sequencing
Genomic DNA was extracted from peripheral blood leukocytes using a standard salting-out procedure [5]. Then WGS was performed (Novogene, Beijing, China) according to the following protocol: sequencing libraries were generated using NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, Hitchin, UK) following the manufacturers' recommendations. Genomic DNA was randomly fragmented to a size of 350 bp by Bioruptor, DNA fragments were size-selected with sample purification beads. The selected fragments were end-polished, A-tailed, and ligated with the full-length adapter. After the treatments, the fragments were filtered with beads again. Finally, the libraries were analyzed for size distribution by Agilent2100 Bioanalyzer and quantified using real-time PCR. Libraries were sequenced by Illumina high-throughput HiSeq X Ten sequencer with paired-end sequencing strategy.

WGS Data Preprocessing
Raw read files were processed with Intelliseq Flow Annotation Pipeline (https:// intelliseq.com/, accessed on 31 July 2021) based on Cromwell (https://cromwell.readthedocs. io/en/stable/, accessed on 31 July 2021). Within the pipeline, fastq file quality was assessed with FastQC. Further, fastq files were then aligned to Broad Institute Hg38 Human Reference Genome with GATK 4.0.3. Duplicate reads were removed with Picard and base quality Phred scores were recalibrated using GATK's covariance recalibration. Variants were called using GATK best practices. Identified variants were assessed using the Intelliseq Flow annotation pipeline that implemented ACMG recommendations.

Quality Assessment
Mean sequencing depth for the whole genome was 24.95 and the mean genotype quality (GQ) score for the whole-genome vcf was 83.75. 99.9% of genotypes had an estimated probability of error <10%. Additional quality analysis was conducted for each of the genes included in the panel (for both selected rare and all 368,460 variants identified in the panel genes) and is available as Supplementary Table S1.

Rare Variant Analysis
Whole-genome sequencing identified 4,827,282 SNVs and small indels in the sample. Further filtering was based on Phred quality scores, allele frequency in the ExAC (Exome Aggregation Consortium) database (<5% for variants in genes associated with the disease phenotype), association with Human Phenotype Ontology (HPO) terms, and predicted pathogenicity. Variants with the following impact on protein and mRNA were retained: missense, nonsense, frameshift, and splice site variants. Common and low impact variants were then filtered out (with a max frequency threshold of 0.05 and minimal SnpEff predicted impact on gene product set as MODERATE). Low-quality multiallelic variants (QUAL < 300) were removed.
Annotated variants for genes from the defined gene list (Supplementary Table S2) were then classified according to the American College of Medical Genetics and Genomics (ACMG) criteria and prioritized. Selected variants were manually evaluated for quality in IGV. PLAU and BACE1 variants were confirmed with Sanger sequencing (ABI 3130 Genetic Analyzer, Applied Biosystems, Foster City, CA, USA) (Supplementary Figures S2 and S3).

Fibroblast Cultures and Inhibition of Nonsense-Mediated Decay (NMD)
Primary skin fibroblasts (obtained from the patient and age-matched, unrelated neurologically healthy subjects-Supplementary Table S3) were collected and cultured as previously described [11]. For NMD analysis primary fibroblast cultures were treated with an NMD inhibitor, puromycin (100 µg/mL or water) (Sigma Aldrich, Saint Louis, MO, USA) for 8 h [12,13].

mRNA Expression Analysis-Real Time PCR
RNA was extracted and reverse transcribed according to standard protocol with QIAzol Lysis Reagent (Qiagen, Manchester, UK) and QuantiTect ® Reverse Transcription kit (Qiagen, Manchester, UK), respectively. Quantitative real-time PCR analysis was done with RT HS-PCR Mix SYBR (A&A BIOTECHNOLOGY, Gdańsk, Poland) (primers are listed in Supplementary Table S4), using a StepOne Plus system (Applied Biosystems, Foster City, CA, USA). Changes in gene expression were determined with the ∆Ct method using GAPDH levels for normalization. Similar results were obtained with PPIB as a normalizing gene.

Statistical Analysis
For each experiment, the relative values obtained from different biological replicates (n ≥ 3) were used to calculate mean, standard deviations (SD) and statistical significance in a two-tailed t-test (GraphPad Prism 6.0). p < 0.05 was considered significant. For quantitative densitometric analysis of WB results, ImageJ software was used [14]. The intensity value of each protein band was normalized to the respective GAPDH value.
Extended methods can be found in Supplementary Materials.

Case Description
A 69-year-old patient, with university education and professional experience in forestry and agriculture, was referred to our neurology outpatient clinic with a five year history of cognitive dysfunction. When he was 64 (age at onset), he started experiencing problems with planning chemical fertilizers' application while running a large agricultural company. His insight was initially preserved as he employed a new worker to help him with the tasks he could no longer perform on his own. Progressive cognitive decline was evident in the report provided by his family. He was no longer able to recognize fruit types, he could not name trees or birds, despite his previous vast knowledge in this area. He became very effusive, also to strangers. The patient had a strong family history of dementia ( Figure 1A). One of the proband's paternal aunts was diagnosed with an unspecified late-onset dementia that also progressed slowly. One of the proband's brother presented with a mild cognitive impairment with predominant semantic deficit. The results of his neuropsychological examination are presented in a supplementary file (Supplementary  Table S5). Both the proband's father and the above-mentioned brother had marked cognitive rigidity ( Figure 1A).
At the time of the first clinical evaluation the general neurological examination did not reveal any motor or sensory problems nor frontal release signs. There was no evidence of depression or psychotic symptoms. His speech was fluent and effortless but empty. He did not demonstrate any insight.
The neuropsychological assessment planned and executed by a neuropsychologist showed dementia (Addenbrooke's Cognitive Examination III: 59/100) (Supplementary  Table S6), which was consistent with the patient's inability to perform complex daily life activities. His speech was severely anomic. Phonology, syntax and prosody were not affected. Phonemic fluency was initially preserved in contrast to very deficient semantic fluency (Supplementary Table S7). He had a widespread semantic deficit, which affected both verbal and visual semantics. He was unable to recognize many very famous faces. His spatial skills were very well preserved which contrasted with a very poor semantic memory as demonstrated by impaired drawing to command but good copying ( Figure 1C). He had episodic memory impairment and executive dysfunction, while the calculation was very well preserved [15]. Praxis was unaffected in copying tasks that did not require reference to semantics.   . Drawing, copying, and calculation tasks demonstrating the dissociation between impaired semantics and well preserved visuospatial processing (inability to retrieve specific features of an object with good visuoperceptual and visuoconstructive functions) and calculation. I A. Drawing a bicycle to a verbal command at the age of 71. I B. Copy of a bicycle drawing at the age of 71. II. The calculation task was performed correctly at the age of 71, as well as at the age of 72. III A. Drawing a bicycle to a verbal command at the age of 72. III B. Copy of a bicycle drawing at the age of 72.
The pattern of progression in neuropsychological assessment, as well as behavioral deterioration over the next years was typical for semantic dementia/semantic variant of a primary progressive aphasia [16,17]. He remained a good swimmer and dancer during the next few years. He continued to enjoy numerical paper and pencil games, especially Sudoku. However, he could no longer handle computer games. At the age of 71, he could not understand Easter celebrations nor understand the word "glass". Face recognition deficits progressed. At the age of 72 semantic deficit progressed significantly e.g., he was unable to recognize snow and tried to eat a dishwasher tablet. Instead of using his wife's first name, he called her "my girl". Parsimony appeared. He also required constant supervision due to compulsive behavior and disinhibition, e.g., during a walk he stole chips from a child who was passing by. He was also very agitated and had transient psychotic symptoms. At this time neuropsychological assessment had to be shortened due to impatience and deficient verbal comprehension. As he required 24h supervision and nursing care due to severe dementia and incontinence, he was institutionalized at the age of 73 and deceased in September 2021 at the age of 76.
MRI performed at the age of 67 and 69 revealed mild bilateral temporal atrophy that became more pronounced on the right side at the age of 72 ( Figure 1B, left panel) Single-photon emission computed tomography (SPECT) was conducted when he was 70 and showed an anterior pattern of hypoperfusion with bilateral frontal and temporal involvement with more pronounced deficits in the superior frontal gyrus and the temporal lobe on the right side ( Figure 1B, right panel).

Genetic Analyses
Initial genetic analysis of the patient's DNA has excluded mutations in MAPT, PGRN, and C9orf72 which are the most common genetic causes of the FTLD spectrum [18] and subsequently whole-genome sequencing was performed. Neither pathogenic or likely pathogenic variants (P/LP variants) nor high-confidence structural variants have been identified according to the ACMG criteria for pathogenicity [19]. However, very rare variants of unknown significance (VUS) have been found in PLAU, BACE1, and POLG genes along with common established risk factors in APOE and MAPT that could contribute to the observed phenotype (Table 1, and Supplementary Table S8).  Figure S2A) has been classified as a high confidence loss of function variant with the ACMG score of 0.6 (Supplementary Table S7). The PLAU c.366A has been previously reported only in the ALFA (for Alzheimer and Families) project (MAF = 0.00005; 1/21336), a research platform to identify early pathophysiological features of AD [20]. PLAU c.366C>A was absent from gnomAD (11.10.2021) and in Polish WGS samples (n = 287, an in-house database): sportsmen (n = 102); patients with Tourette syndrome (n = 129), and healthy controls (n = 56).
PLAU position c.366 is highly evolutionarily conserved according to PhastCons (1.0; high: >0.93) and GERP = 5.68 (high: >3.26) and C>A substitution introduces a stop codon in exon 5 (out of 11). As a consequence, a fraction of transcripts is predicted to undergo nonsense-mediated mRNA decay (NMD) which was confirmed by the successful rescue of PLAU mRNA levels ( Figure 2A) and PLAU mutant allele (Supplementary Figure S2B) by an NMD inhibitor, puromycin [12,13]. Accordingly, the PLAU mRNA and protein levels were decreased in the patient's fibroblasts compared to neurologically healthy, age-matched control individuals (Figure 2A,B).

BACE1 (β-Site APP-Cleaving Enzyme 1, Beta-Secretase 1)
BACE1 (# 604252, OMIM) is a transmembrane aspartic protease that catalyzes the first step in the formation of amyloid-beta (Ab) peptides from amyloid precursor protein (APP) [21]. Amyloid-beta peptides build amyloid-beta plaques that accumulate in the brains of AD patients [21]. Since BACE1 levels/activity are elevated in AD brains and fluids, the inhibition of this β-secretase has been extensively tested in clinical trials as a therapeutic strategy for AD [21]. c.944C>T variant (NM_012104.6 ENST00000313005.11, Table 1, Supplementary Figure S3) was predicted to be pathogenic by six programs: DANN, EIGEN, FATHMM-MKL, LIST-S2, M-CAP, and MutationTaster, and to affect splicing (Human Splicing Finder) (Supplementary Table S9). There was no evidence of additional/aberrant mRNA isoforms (Supplementary Figure S4). However, we have found elevated BACE1 mRNA and protein levels in patient's fibroblasts compared to controls ( Figure 2C,D).
Other identified variants potentially contributing to the clinical phenotype are presented in Supplementary Table S9 of PLAU mRNA levels ( Figure 2A) and PLAU mutant allele (Supplementary Figure S2B) by an NMD inhibitor, puromycin [12,13]. Accordingly, the PLAU mRNA and protein levels were decreased in the patient's fibroblasts compared to neurologically healthy, agematched control individuals (Figure 2A,B).

Discussion
Overlapping features of dementia phenotypes pose a considerable challenge for clinicians, especially semantic dementia, which in most cases is considered to be sporadic [22].
In our patient prominent deficits in both verbal and visual semantics, face recognition and episodic memory were present, consistent with the clinical presentation of a right temporal variant of frontotemporal dementia (rtvFTD) [23]. However, the changes observed in neuroimaging were bilateral with only slight right-sided predominance, so his clinicalradiological presentation did not fully correspond to the rtvFTD. The pattern of semantic impairment in this patient showed that limiting SD to a semantic variant of a primary progressive aphasia (svPPA) according to Gorno-Tempini et al. criteria [24] obscures the global semantic deficit. Although some patients may be classified as svPPA or rtvFTD our case presented with a mixed pattern of deficits, corresponding to the previous Lund-Manchester criteria of SD [25].
The disease duration was relatively long in the proband (12 years) and his affected relatives. Of note, long disease duration is typical for SD [26], unlike other early-onset dementias, which have a rather fast slope of progression.
The possible hereditary nature of semantic dementia has been recently shown in rtvFTD [27]. The potential association of psychotic symptoms in one of the proband's brothers with FTD cannot be excluded, although in FTD psychotic symptoms are usually associated with C9orf72 mutation [28].
While the bioinformatic analysis of WGS results revealed no pathogenic or likely pathogenic variants, in vitro functional analyses suggested that the very rare variants of unknown significance in PLAU (c.366C>A; p.Cys122*), and BACE1 (c.944C>T; p.Thr315Met) could contribute to the disease phenotype.
To date, only heterozygous tandem duplications of PLAU have been found to cause the disease phenotype of autosomal dominant Quebec platelet disorder (QPD, # 601,709 OMIM). QPD is a bleeding disorder due to a gain-of-function defect in fibrinolysis with significantly increased PLAU levels in patient's platelets, who showed delayed onset bleeding after vascular damage (e.g., surgery). To our knowledge, p.Cys122* (Table 1), is the first PLAU variant with the confirmed haploinsufficiency detected in a patient presenting semantic dementia phenotype. The loss-of-function (LOF) variants of the PLAU gene have not been previously reported in humans, and the consequences of PLAU deficiency have been studied only in mouse models. Mice lacking PLAU showed a reduced spontaneous exploration of the surrounding environment, impaired post-injury recovery [29,30], and had elevated amyloid-beta (Ab42 and Ab40) levels in plasma [31]. PLAU deficiency in mice also led to delayed wound healing with an abnormal angiogenic pattern [32]. On the other hand, Ab aggregates induced PLAU expression leading to increased plasmin, which in turn, degraded both aggregated and non-aggregated forms of Ab suggesting a negative feedback loop [33].
Interestingly, genotypes C/T and T/T of a missense PLAU variant in exon 6 (c.422T>C, Leu141Pro, rs2227564) have been previously associated with elevated plasma Ab42 levels and the increased LOAD risk (late-onset AD) [31,34], although the data remained controversial [35,36] (see also meta-analysis on http://www.alzgene.org, accessed on 5 October 2021). While the Pro141Leu variant seems to bind fibrin aggregates more efficiently [37] the exact molecular mechanism underlying its action is not known. However, our patient carried a genotype C/C at the position c.422. Moreover, other rare missense and nonsense single nucleotide variants (SNVs) in PLAU (p.T86A, p.H149Y, and p.C151F) have been identified in patients with multiple sclerosis (MS), an inflammatory disease characterized by myelin loss and neuronal dysfunction [38]. Finally, PLAU has also been identified as a frailty biomarker in aging and age-related diseases [39]. It can be speculated that the decreased PLAU levels due to p.Cys122* variant could lead to the dysfunction of fibrinolysis cascade and inefficient blood clots dissolution. However, our patient did not have any history of thrombotic disorders.
While patients with SD showed predominantly TDP-43 type C brain pathology, other pathology types, i.e., FTLD-TDP types A and B, Tau-positive (FTLD-tau) and Alzheimer's disease (AD) pathology have been reported in 17% up to 32% of SD cases [22,40]. A recent neuropathological report suggests that rtvFTD in particular may be caused by other pathologies apart from FTLD-TDP type C [41]. It highlights the distinction between svPPA and rtvFTD. Unfortunately, in our case the diagnosis of SD (rtvFTD) was not supported by a definite exclusion of Alzheimer's disease pathology through cerebrospinal fluid biomarker testing, amyloid-PET (Positron Emission Tomography) or neuropathological examination. For this reason, it cannot be excluded that the patient had amyloid-beta neuropathology. That speculation is strengthened by the fact that he showed several features of atypical AD. It is worth noting that the increased BACE1 mRNA/protein levels (Table 1, Figure 2C,D) could possibly increase the amyloid-beta load and in this respect act in synergy with PLAU deficiency [31]. It is also worth mentioning that the patient's other brother was diagnosed with paranoid schizophrenia (Figure 1A), in which BACE1 could play an important role [42]. To date, only the impact of common BACE1 polymorphisms on AD risk has been studied, yielding discordant results (see also meta-analysis on http://www.alzgene.org, accessed on 5 October 2021).
In addition, we report a POLG variant (mtDNA polymerase gamma; NM_002693.3 ENST00000268124.9, c.3436C>T, p.Arg1146Cys; Table 1), as it has been described previously in autosomal-dominant progressive external ophthalmoplegia type 1 (adPEO1) [43]. The pathogenic status of POLG p.Arg1146Cys variant is controversial and our patient has not shown any symptoms of adPEO1. However, since mitochondrial impairment is a common mechanism of many neurodegenerative disorders, it could be speculated that variants interfering with the mitochondrial function could aggravate the phenotype. This is also suggested by the fact that another POLG mutation (p.Y955C) has been associated previously with Alzheimer's pathology [44].
Finally, known moderate-impact variants such as APOE (ε2, ε4), which translates into a threefold greater risk of developing AD by age of 75 y. [45], as well as MAPT H1/H1 haplotype, previously associated with an increased AD risk [46] (Table 1), may contribute to the observed clinical phenotype.
Although our functional analyses suggested that PLAU and BACE1 variants had a biological impact, further functional and segregation studies are needed to clarify their pathogenic status. Overall, it could be envisioned that more complex patterns of inheritance including oligogenic inheritance may account for part of early-onset dementia cases. However, the discrimination between causative and modifying variants, and insignificant ones is one of the greatest challenges of medical genetics. The concept of oligogenic inheritance should influence not only the approach to gene identification but also genetic testing and counseling [47].

Conclusions
Our findings suggest that future genetic analyses of early-onset dementia cases, especially those with slow disease progression, should take into account very rare variants in the PLAU and BACE1 genes, that both are responsible for the regulation of amyloid-beta levels.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/genes12111806/s1, Table S1: Quality analysis conducted for each gene included in the panel; Table S2: List of genes selected for variant analysis; Table S3: Control fibroblast lines used  for the study; Table S4: Primers used for the study; Table S5: Neuropsychological assessment of the proband's brother; Table S6: Addenbrooke's Cognitive Examination-III of the proband; Table S7: Detailed neuropsychological assessment of the proband; Table S8: Variants identified in the proband according to selected criteria; Table S9: Human Splicing Finder Analysis of BACE1 variant; Figure S1: Family history of the patient; Figure S2A: PLAU-Sanger sequencing of patient's DNA; Figure S2B: PLAU-Sanger sequencing of patient's cDNA (with and without puromycin treatment); Figure S3A: BACE1-Sanger sequencing of patient's DNA; Figure S3B: BACE1-Sanger sequencing of patient's cDNA, Figure S4. Agarose gel electrophoresis of the PCR products for patient's and control cDNAs. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
The data presented in this study are sequencing data of human samples thus cannot be openly shared due to personal data protection.