Dear Sirs,

AP4 is an adaptor protein crucial for vesicle and endosomal trafficking from the trans-Golgi network consisting of four subunits, namely AP4B1, AP4M1, AP4E1 and AP4S1 [1] [2]. Biallelic variations in AP4B1 (MIM 607245) have been primarily described in a syndromic form of autosomal recessive spastic paraplegia known as “SPG47” (MIM 614066) [3]. SPG47 is a severe childhood-onset genetic condition characterized by global developmental delay, intellectual disability, spastic diplegia, microcephaly, seizure, cerebellar ataxia and brain MRI abnormalities, such as thin corpus callosum, white matter loss and ventriculomegaly [4] [5]. Although late-onset cerebellar ataxia, along with other related neurodegenerative diseases, has recently been associated with high rate of hypomorphic variants in various severe disease-causing genes, involvement of AP4B1 in such moderate phenotypes has never been established thus far [6]. Here, we describe three previously unreported individuals displaying sporadic form of adult-onset cerebellar ataxia resulting from compound heterozygous variants in AP4B1, including two novel hypomorphic missense variants highly suspected to contribute to this milder neurological disorder.

Methods

Recruitment of individuals and data collection

AP4B1 variants were identified in affected individuals either by clinical exome sequencing (CES) conducted in Montpellier molecular genetics laboratory and targeting approximately 6700 OMIM genes, or genome sequencing (GS) as part of the 2025 French Genomic Medicine Plan. Each individual had previous exclusion of Friedreich ataxia, CANVAS, FXTAS and SCA27B diagnoses. Medical and molecular data were systematically collected using a standardized form sent to collaborators. Clinical examination, brain MRI, electroneuromyogram, fundus and genetic testing were performed in routine care.

Molecular analyses

Next-generation sequencing (NGS) was performed in a total of 772 probands with cerebellar ataxia using either Illumina NextSeq 550 (CES, n = 644) or NovaSeq 6000 (GS, n = 128) in accordance with local quality standards and following manufacturer’s protocol. AP4B1 variants were confirmed in probands using Sanger sequencing and familial segregation was conducted in available relatives. ACMG/AMP guidelines were used for variant interpretation and compliance with HGVS nomenclature has been verified using batch validation tool of VariantValidator software (https://variantvalidator.org/service/validate/batch/) (Supplemental Table 1).

Assessment of previously reported AP4B1 missense variants

Documented missense variants in AP4B1 were retrieved by searching for articles indexed in PubMed using the terms “AP4B1”, “hereditary spastic paraplegia type 47” or “SPG47”. These variants were individually assessed using MobiDetails software (https://mobidetails.iurc.montp.inserm.fr/MD/). Positional tolerance to missense variants was analyzed using MetaDome (https://stuart.radboudumc.nl/metadome/) while three-dimensional modeling and calculation of the Grantham score were conducted using MIZTLI (https://miztli.biokerden.eu/). Further conservation analyses were carried out utilizing three distinct computational tools, namely PhyloP, PhastCons 46-Way and GERP, all accessible through UCSC software (https://genome-euro.ucsc.edu/index.html).

Results

AP4B1 variants (NM_001253852.3, GRCh37/hg19) were detected in three unrelated individuals presenting with sporadic adult-onset ataxia using NGS (Fig. 1). These affected individuals were born to unrelated healthy parents, and their family history was unremarkable. None of them exhibited neurodevelopmental delay, intellectual disability nor facial features. No hearing impairment was observed and fundus examination, along with electroneuromyogram, yielded normal results across this series. Main clinical and molecular data are summarized in Table 1.

Fig. 1
figure 1

Molecular spectrum associated with AP4B1-related adult-onset ataxia in this series. A Family pedigrees are depicted for each affected individual, in addition with the confirmation of AP4B1 compound heterozygous variants in probands using Sanger sequencing. B Linear representation of AP4B1 (NM_001253852.3), featuring the localization of reported variants. Loss-of-function and novel hypomorphic missense variants are highlighted in red and orange, respectively. Variant annotation was conducted using GRCh37/hg19 genome assembly

Full size image
Table 1 Detailed clinical and molecular data from affected individuals.
Full size table

Individual 1

At the age of 43, Individual 1 manifested gait ataxia alongside dysarthria. Last examination revealed a scale for the assessment and rating of ataxia (SARA) score of 6.5/40, demonstrating relative stability over the preceding two years. Mild dysmetria was noted during the heel-to-shin test, in association with unilateral dysdiadochokinesis, saccadic pursuit and binocular diplopia. Upper limb reflexes were brisk and bilateral Hoffmann sign was present. However, neither Babinski sign, lower limb spasticity, nor sensory-motor deficits were observed. Individual 1 also reported symptoms compatible with overactive bladder and brain MRI exhibited moderate vermian atrophy, with no other morphological abnormalities (Fig. 2). The c.470-2A > C p.(?) variant is inherited from the maternal branch and has previously been documented as likely pathogenic within the ClinVar database (RCV002773396). In silico predictions using SpliceAI and SPiP tools support the notion of a deleterious effect induced by this variant, which entails an alteration of the consensus splice acceptor site, thereby leading to a highly suspected exon skipping phenomenon. Moreover, this splicing variant exhibits a notably low frequency within healthy population databases, with no reported homozygous individuals. These collective findings consequently warrant the classification of this variant as pathogenic (class 5). The c.320G > A p.(Arg107Gln) variant is paternally inherited and annotated as variant of uncertain significance (VUS) in the ClinVar database (RCV003077695.1). In silico assessments using REVEL, SIFT, ESM1b, Polyphen 2 and ClinPred tools suggest a deleterious impact of this variant. Situated within the HEAT repeat domain of the adaptin N-terminal region, this missense variant is further suspected to induce alterations in the three-dimensional conformation of AP4B1. Indeed, although both arginine and glutamine belong to the polar amino acid family, arginine is positively charged unlike glutamine, yet highly conserved at the 107th position across vertebrates. Finally, this variant displays a very low frequency in healthy population databases, with no reported instances of homozygosity, and is found in trans with the maternally inherited variant. Consequently, this missense variant has been categorized as likely pathogenic (class 4).

Fig. 2
figure 2

Brain MRI of affected individuals. Sagittal T1-weighted images show mild cerebellar atrophy in all three affected individuals. No morphological abnormality of corpus callosum, ventriculomegaly nor white matter loss, ordinarily associated with SPG47, were detected in this series

Full size image

Individual 2

At the age of 21, Individual 2 manifested gait ataxia followed by the graduate onset of dysarthria. The most recent assessment indicated a SARA score of 7/40, showing relative stability across the last five years. Moderate dysmetria was observed during the heel-to-shin test, along with spontaneous multidirectional nystagmus and binocular diplopia. Clinical examination revealed myokymia, brisk reflexes in both upper and lower limbs, bilateral Hoffmann sign and impaired vibratory sensation at the ankles. No Babinski sign, lower limb spasticity, impaired tactile sensation, nor motor deficits were discerned, and Individual 2 did not report any symptoms of sphincter disturbance. Similar to Individual 1, brain MRI showed mild vermian atrophy without any other notable morphological findings (Fig. 2). The c.1114 + 1G > A p.(?) variant, maternally inherited, remains absent from the ClinVar database. SpliceAI and SPiP in silico tools predict its deleterious effect, inducing an alteration in the consensus splice donor-site and resulting in a highly suspected frameshift due to intron retention. Additionally, this splicing variant remains absent from healthy population databases. Taken together, these findings prompt the classification of this variant as pathogenic (class 5). The c.1558C > T p.(Arg520Cys) variant is inherited from the paternal lineage and exhibits conflicting pathogenicity classification within the ClinVar database, ranging from VUS to likely pathogenic (RCV001862068.14). In silico predictions encompassing REVEL, SIFT, ESM1b, Polyphen 2 and ClinPred tools indicate a deleterious effect of this variant. Additionally, arginine being hydrophilic, contrasts with cysteine’s hydrophobic nature and further remains strongly conserved at the 520th position across vertebrates. Computational analyses using SpliceAI and SPiP further suggest that this missense variant might reveal a cryptic splice acceptor-site within AP4B1’s exon 9, while it displays an extremely low frequency in healthy population databases with no reported instances of homozygosity. Considering these findings, this variant has been classified as likely pathogenic (class 4).

Individual 3

At the age of 38, Individual 3 presented with gait ataxia accompanied by progressive dysarthria. Last examination using SARA yielded a score of 15/40. Dysmetria and hypermetria were observed in both upper and lower limbs, alongside unilateral dysdiadochokinesis. Clinical evaluation also identified horizontal gaze-driven nystagmus, brisk reflexes and unilateral Babinski sign. No Hoffmann sign, lower limb spasticity nor sensory-motor deficits were evident. Individual 3 further reported symptoms evocative of overactive bladder and brain MRI showed isolated mild vermian atrophy, consistent with the findings in Individuals 2 and 3 (Fig. 2). The c.865_866del p.(Ser289Profs*2) variant is not documented within the ClinVar database. This frameshift variant is anticipated to result in AP4B1 loss-of-function and remains absent from healthy population databases. Therefore, these findings collectively justify the classification of this variant as pathogenic (class 5). The c.1558C > T p.(Arg520Cys) variant, shared with Individual 2, was unphased due to the deceased status of both parents and the lack of available siblings. However, it has still been categorized as likely pathogenic (class 4) in light of segregation data from Individual 2 helping assuming its trans location.

Discussion

Hypomorphic pathogenic variants in genes responsible for severe neurodegenerative disorders are increasingly being recognized as a new cause of attenuated neurological phenotypes, predominantly manifesting as late-onset cerebellar ataxia with or without spastic paraplegia [6],[7,8,9,10,11] (Table 2). According to Muller’s morphs classification, a hypomorph variant, sometimes referred as “leaky mutation”, is characterized by a reduction in gene activity, rather than a complete loss-of-function. Typically exhibiting recessive behavior in haplo-sufficient genes, hypomorphs can lead to reduced transcription, aberrant mRNA splicing, impaired translation or altered protein functional activity [12].

Table 2 Summarized list of hypomorphic variants associated with cerebellar ataxia in the literature.
Full size table

Affected individuals we are documenting display comparable forms of sporadic adult-onset ataxia, primarily characterized by gait ataxia coupled with progressive dysarthria. This cerebellar ataxia is of moderate severity and slowly evolving, while being associated with mild cerebellar atrophy without any other observable morphological abnormalities of the brain. Pangenomic analyses conducted on these three mildly affected individuals identified a likely pathogenic missense variant in AP4B1 found in trans with either a canonical splicing site variant (Individual 1 and 2) or a frameshift variant (Individual 3) that we strongly suspect to act as hypomorphic with subsequent residual activity of AP4B1, as evidenced by the attenuated phenotype observed across this series. In Individual 1, we hypothesize that the hypomorphic nature of the c.320G > A p.(Arg107Gln) may result from the in silico-predicted neutrality of this position toward missenses (MetaDome score: 0.84), and the relatively low evolutionary distance between arginine and glutamine. Conversely, AP4B1 residual activity stemming from the c.1558C > T p.(Arg520Cys) variant identified in both Individual 2 and 3 might be attributed to the potential unmasking of a cryptic splicing acceptor-site within AP4B1’s exon 9 (SpliceAI acceptor gain: 0.32), therefore leading to a partial splicing impairment given the variant’s location outside of the canonical acceptor site.

Although SPG47 predominantly results from biallelic truncating variants, five pathogenic missenses have already been documented in infants displaying a classical phenotype of SPG47, in contrast to the individuals we are currently describing [4], [5], [13], [14], [15]. Noteworthily, exhaustive computational reassessment of these previously reported missense variants did not reveal significant discrepancies when compared to the hypomorphic variants at study, as detailed in Supplemental Table 2. Specifically, these documented missense variants do not affect more evolutionarily conserved nucleotides, while their respective in silico predictions do not appear more deleterious than the hypomorphic variants under investigation. Moreover, the non-canonical transcript of AP4B1 (ENST00000432415.6) is characterized by the alternative splicing of exons 3 to 5, whereas the hypomorphic variants at study are respectively located in conserved exons 2 and 9, thereby weakening the plausibility that such moderate phenotype arises from of a rescue mechanism mediated by an unaffected alternative transcript of AP4B1.

Hence, exercising caution is imperative when solely relying on computational tools to validate the hypomorphic nature of a specific variant since only in vitro functional studies encompassing mRNA and protein expression analyses might definitively ascertain the reduction of AP4B1 expression, rather than a complete loss-of-function. Additional investigations into cellular processes, such as vesicle trafficking and autophagy, further hold the potential to provide supplementary evidence supporting the hypomorphic nature of these two missense variants. Lastly, the integration of omics approaches combining transcriptomics, proteomics and epigenomics technologies may also be employed to uncover additional dysregulated pathways as a result of AP4B1 impaired expression.

In summary, this report highlights a novel phenotypic expression of AP4B1-related disorder manifesting as an isolated form of autosomal recessive adult-onset ataxia, and resulting from two previously unreported hypomorphic missense variants, namely c.320G > A p.(Arg107Gln) and c.1558C > T p.(Arg520Cys). From a broader standpoint, we believe this study will contribute to refine the genetic landscape currently involved in sporadic late-onset ataxia, while assisting molecular geneticists in the interpretation of pangenomic analyses conducted on individuals displaying mild neurological phenotypes.