Two affected members of a German family with a dementia syndrome were examined and diagnosed at a specialized center for neuropsychiatric disorders. The family history was consistent with autosomal dominant inheritance. The index patient’s legal guardians provided written informed consent for blood sample donation according to the guidelines of the German Genetics Diagnostics Act. The study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the ethics committee of the Justus-Liebig-University of Giessen (AZ24/14erw).

DNA was extracted from peripheral blood samples using standard procedures. Point mutations in the dementia-associated genes MAPT, GRN, TARDBP, PSEN1, PSEN2, and APP had been excluded previously. Fragment-length analysis at the corresponding loci for FTD/amyotrophic lateral sclerosis (C9orf72), HD (HTT), and SCA17 (TBP) was performed by polymerase chain reaction (PCR) using a specific fluorescent-labeled primer in combination with an unlabeled reverse primer. Primer sequences are shown in Supplementary Table 1A. Fragments were run on a SeqStudio Genetic Analyzer (Applied Biosystems, Austin, TX, United States) and repeat sizes were determined using GeneMapper v.5 software (Applied Biosystems). DNA was analyzed for variants in STUB1 (ENSG00000103266) by amplification and sequencing of the seven coding exons, the flanking sequence of intron 1, and the complete intronic sequences (IVS 2-6) of transcript ENST00000219548.9. Primer sequences are shown in Supplementary Table 1B.

Subsequently, whole exome sequencing was performed on the DNA of the index patient, examining all coding regions including flanking intronic sequences. SureSelect XT Human All Exon V7 kit (Agilent, Santa Clara, CA, United States) was used for enrichment. Quality control of the prepared library was performed using a Qubit 3.0 fluorometer (ThermoFisher Scientific, Waltham, MA, United States). Sequencing was performed on an Illumina NovaSeq platform (Illumina, San Diego, CA, United States). The mean coverage of exons was at least > 20× with > 98.5% targeted bases covered. GRCh37 was used as the reference genome. Filter criteria for pathogenic variants were minor allele frequency (MAF) < 0.005 in the gnomAD v2.1.1 database (Karczewski et al., 2020) and Combined Annotation Dependent Depletion score ≥ 25 (Kircher et al., 2014; Rentzsch et al., 2021).

Impact of the observed splice donor variant was analyzed by in silico analysis using the programs MutationTaster2 (Schwarz et al., 2014), NetGene2 (Brunak et al., 1991), and Alternative Splice Site Predictor (ASSP) (Wang and Marín, 2006). The detected variant was subsequently classified according to American College of Medical Genetics (ACMG) guidelines (Richards et al., 2015) and the recommendations for canonical splice site variants (Abou Tayoun et al., 2018).

Total RNA was extracted from PAX-conserved (PreAnalytiX, Hombrechtikon, Switzerland) whole blood of the index patient and a healthy control using the Monarch total RNA miniprep kit (New England Biolabs, Ipswich, MA, United States). Additional DNase I digestion (Invitrogen, Waltham, MA, United States) was performed to remove traces of genomic DNA. RNA was transcribed into cDNA using random primers and SuperScript IV reverse transcriptase (Invitrogen). Reverse transcription PCR (RT-PCR) was performed using primers STUB1_cEx2F (5’-CGCTGGTGGCCGTGTATTAC-3’), and STUB1_cEx4R (5’-GTGCTTGGCCTCAATGCAGG-3’). To confirm the integrity of the RNA and the transcribed cDNA, a 210 bp fragment was amplified with primers TBP_Ex5F (5’-TTGCTGCGGTAATCATGAGG-3’) and TBP_Ex6R (5’-GAAACTTCACATCACAGCTCCC-3’) derived from the TBP gene as an internal control.

Products were separated on 1.5% agarose gels. Different sized RT-PCR fragments were excised from the gel, purified, and either sequenced directly or cloned into vector pCR4-TOPO using the TOPO TA Cloning Kit (ThermoFisher Scientific). At least 25 clones of each fragment were sequenced.

Two minigenes were produced: one carried the splice donor variant c.524+1G>A (pcDNA3.1-STUB1-Mut) and the other corresponded to the wild-type sequence (pcDNA3.1-STUB1-WT). Both minigenes contained the genomic sequences of partial intron 1 to intron 4 of STUB1. In brief, primers STUB1_IVS1F_BamHI (5’-TATGGATCCGTACTCCACTGTGCACAGATCC-3’) and STUB1_Int4R_EcoRI (5’-TATGAATTCCACCCCACCTCCCTC CTCCA-3’) were used to amplify 995 bp fragments from the genomic DNA of the patient and the control. PCR products were restricted with BamHI and EcoRI (New England Biolabs), and subsequently cloned in pcDNA3.1+ (ThermoFisher Scientific). Competent OneShot Top-10 E. coli cells (ThermoFisher Scientific) were transformed with the recombinant plasmids. Integrity of the selected clones was confirmed by sequencing.

Chinese hamster ovary (CHO)-K1 cells (American Type Culture Collection, Manassas, VA, United States) were transiently transfected with 0.2 μg of either pcDNA3.1-STUB1-Mut, pcDNA3.1-STUB1-WT, or pcDNA3.1+ using FuGENE (Promega, Madison, WI, United States). Untransfected cells served as a control. Cells were cultured in F-12K medium (ThermoFisher) supplemented with 10% fetal bovine serum (Sigma-Aldrich, St. Louis, MO, United States) and 1% penicillin/streptomycin (Gibco, Waltham, MA, United States). Cells were harvested after 28 h of incubation. Total RNA was extracted and digested with DNase I (Invitrogen) to remove endogenous genomic DNA and recombinant plasmids. RNA was analyzed by RT-PCR using primers STUB1_cEx2F and STUB1_cEx4R as mentioned above.

The male index patient (II-4, Figure 1A) had normal early motor and cognitive development. He finished school after 9th class and then completed a three-year apprenticeship as a locksmith. First signs and symptoms were observed at age 34 years. According to his wife, he had shown word finding difficulties as well as decreased frustration tolerance and social withdrawal, with the latter initially suspicious of depression. Soon after, deficits were noted at his workplace, such as general slowing and clumsiness. He first presented at age 37 years with normal neurological findings, yet accompanied by shallow affect. Cognitive abilities, however, decreased rapidly, leading to early retirement 1 year after initial presentation. At age 39, tests such as Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) battery (which includes Mini-Mental State Examination, Modified Boston Naming Test, verbal fluency, learning of word lists, vocabulary test, construction ability and a test of visuomotor tracking) (Chandler et al., 2005), the Clock-test (Shulman et al., 1993), the Porteus Maze Test (Porteus, 1950), and the Executive Interview (EXIT25) (Royall et al., 1992) revealed a range of cognitive symptoms involving memory, language, praxis, psychomotor speed, attention, and executive functions suggestive of FTD.

Whereas magnetic resonance imaging (MRI) had only shown mild cerebellar atrophy, repeated fluorodeoxyglucose positron emission tomography (PET) scans were largely unremarkable. A dopamine transporter scan using single-photon emission computed tomography was performed to exclude Lewy body dementia (Supplementary Figure 1). Autoimmune or paraneoplastic etiologies were excluded by cerebrospinal fluid (CSF) analysis.

For 6 years, the patient was followed up as an outpatient until he moved into a nursing home at age 44 because of severe disability, including incontinence. At present (age 51), the patient is fully dependent for locomotion and feeding. He is aphasic and agnostic and shows pronounced apraxia as well as generalized spasticity. He exhibits oral motor and vocal tics as well as disinhibited behavior.

The eldest sister of the index patient (II-1, Figure 1A) was initially reported to be healthy but she attended a school for the learning disabled. In her late 40s, she developed personality changes such as inappropriate behavior and aphasia. The index patient also had monozygotic twin sisters (II-2, II-3 of Figure 1A), one of whom has suffered from epilepsy since the age of 18. She was found to have a benign retroorbital tumor that had caused visual disturbances and was surgically removed. The sisters refused genetic testing.

The patient’s deceased mother (I-6, Figure 1A) initially developed memory and behavioral problems in her early 40s. Subsequently, a dementia such as Alzheimer’s disease was assumed, but no precise etiological classification was made. Because of rapid dementia progression, she moved to a nursing home at age 45. She eventually developed spasticity in all four limbs. In consequence, she was severely impaired in all activities of daily living, becoming bedridden before the age of 50. She died at age 79.

Notably, of her nine siblings, five had a dementia syndrome (I-1, I-2, I-9, I-10, I-11 of Figure 1A). Her husband, father of the index patient (I-5, Figure 1A) was not known to have any behavioral or neurological disorder. He died of an unspecified lung condition at age 70. The index patient’s two children, aged 25 and 21, are both healthy.

Genes most frequently associated with early onset dementia, including PSEN1, PSEN2, APP, MAPT, PRNP, c9orf72, were excluded in the index patient and his affected mother (II-4 and I-6 of Figure 1A). Furthermore, repeat length was in the normal range at C9orf72 (FTD/amyotrophic lateral sclerosis) and HTT (HD). Remarkably, repeat expansions in the incomplete penetrance range at locus SCA17 (TBP) were detected. The affected mother (I-6, Figure 1A) carried 37/43 CAG/CAA repeats, while her son carried two moderately expanded TBP alleles of 41 and 43 CAG/CAA repeats. The expanded alleles showed CAA interruptions in the typical architecture (CAG₃, CAA₃, CAG_x1, CAA, CAG, CAA, CAG_x2, CAA, CAG), indicating stable transmission (data not shown).

Sanger sequencing of STUB1 coding and intronic regions revealed a heterozygous G>A transition at the first base of intron 3 (c.524+1G>A, Figure 1B). The same variant was detected in the affected mother (I-6 of Figure 1A) of the index patient. Analysis tools including MutationTaster2, NetGene2, and ASSP predicted the variant to disrupt the highly conserved splice donor site of intron 3. Interestingly, the wild-type (WT) splice donor of intron 3 was among the three splice sites in STUB1 with the highest score, i.e., the most conservation (Table 1). In addition to omission of the splice donor by c.524+1G>A, one cryptic splice donor was postulated in exon 3, and four cryptic splice donors and one cryptic splice acceptor were identified in intron 3 by ASSP in silico analysis (Table 1).

Variant c.524+1G>A was described only once in 244920 exomes (rs1457745122), resulting in a MAF(A) of 0.000004. No further data or clinical information on the database entry was available. According to ACMG guidelines, variant c.524+1G>A was initially classified as likely pathogenic/class 4; after experimental verification of its pathogenic impact it was reclassified as pathogenic/class 5 (Richards et al., 2015; Abou Tayoun et al., 2018).

To investigate whether an additional pathogenic variant contributed to the severe dementia syndrome of the index patient and his mother, whole exome sequencing was performed on DNA of the index patient. An analysis of all known dementia- and ataxia-related genes revealed no evidence of any further disease-related pathogenic variant.

Reverse transcription PCR analysis on cDNA of the index patient (II-4 of Figure 1A) and comparison with a cDNA generated from WT whole RNA revealed impaired splicing in the index patient (Figure 2A). While skipping of exon 3 was excluded in the patient, at least three larger fragments were detected (Figure 2A, lane 2) that were not present in the correctly spliced WT control (Figure 2A, lane 3). To verify these additional bands, they were excised from the gel and sequenced directly. Fragments are, from top to bottom (Figure 2A, lane 2; Figure 2B): (a) the unspliced (usp) fragment, (b) without intron 2 but with persisting intron 3 (V3), and (c) without intron 2 but use of a non-predicted cryptic splice donor at position c.524+86 (V1) in intron 3 (Figures 2B–D). The 443 bp fragment (WT, Figure 2A, lane 2), which corresponds to the fragment detected in the WT (Figure 2A, lane 3), is the correctly spliced product of exon 2, exon 3, and exon 4. In the index patient, it was transcribed and processed from his WT allele (Figure 2A, lane 2).

Due to the small size differences in the aberrant spliced fragments of the index patient, direct sequencing was challenging. To avoid missing additional utilized cryptic splice sites, fragments were cloned and multiple clones were re-sequenced. This essentially confirmed the results of the direct sequencing approach. However, one out of 100 clones analyzed used an ultra-rare splice site at position c.524+99 (V2), which was predicted by ASSP analysis (Figures 2C,E and Table 1).

Impaired splicing by variant c.524+1G>A had two main consequences. First, complete retention of intron 3 in the transcripts did not lead to a frameshift, but rather addition of 189 bp of in-frame coding sequence. This corresponds to an insertion of 63 amino acids (p.Arg175_Glu176ins63), which disrupted the coiled-coil domain located at aa 100–127 and aa 170–190 essential for dimerization of CHIP. The same applies to the ultra-rare cryptic splice site at position c.524+99 (p.Arg175_Glu176ins33), which led to the in-frame insertion of an additional 33 aa (Figure 2C). Second, the use of the cryptic splice site at position c.524+86 in intron 3 caused a frameshift (p.Arg175fs*93), resulting in a truncated protein and/or degradation of aberrant RNA via nonsense mediated decay.

To rule out the possibility that the patient’s RNA was degraded and the impaired splicing described above was artificial, two STUB1 minigenes containing the genomic sequence of parts of intron 1 to intron 4 were constructed: pcDNA3.1-STUB1-WT (corresponding to the wild-type sequence) and pcDNA3.1-STUB1-Mut (carrying the variant c.524+1G>A). CHO cells were transiently transfected with one of each. CHO cells were selected used since the homology of the coding regions of STUB1 between Cricetulus griseus and Homo sapiens is only 89%, allowing reliable assignment of the detected fragments.

Whole RNA was extracted and analyzed by RT-PCR (Figure 3A). RT-PCR on whole RNA derived from cells transfected with pcDNA3.1-STUB1-WT (Figure 3A, lane 5) revealed fragments that correspond (a) to an unspliced (usp) fragment, (b) a splice-intermediate fragment without intron 2 (V3 of Figures 3A,B), and (c) a WT-fragment of 443 bp (WT of Figure 3A, lane 5; Figure 3B). Sequencing confirmed that it was the correctly spliced human mRNA (exon 2, exon 3, and exon 4). An additional faint fragment of 530 bp was detected, which corresponds to a splice intermediate lacking intron 3 but containing intron 2 (Figure 3A, lane 5, asterisk).

In contrast, RT-PCR on RNA extracted from pcDNA3.1-STUB1-Mut transfected cells (Figure 3A, lane 6) revealed the fragment containing intron 3 (V3) as was seen in the index patient’s RNA (Figure 2A). The use of the cryptic splice site c.524+86 was barely detectable in the minigene assay. The corresponding 529 bp fragment was only rarely identified (Figure 3A, lane 6, circle). However, no WT fragment of 443 bp was detected in pcDNA3.1-STUB1-Mut cells. Furthermore, no 277 bp fragment was detected that would result from skipping of exon 3 (Figure 3A, lane 6).