AP2A2 mutation and defective endocytosis in a Malian family with hereditary spastic paraplegia

Abstract

Hereditary spastic paraplegia (HSP) comprises a large group of neurogenetic disorders characterized by progressive lower extremity spasticity. Neurological evaluation and genetic testing were completed in a Malian family with early-onset HSP. Three children with unaffected consanguineous parents presented with symptoms consistent with childhood-onset complicated HSP. Neurological evaluation found lower limb weakness, spasticity, dysarthria, seizures, and intellectual disability. Brain MRI showed corpus callosum thinning with cortical and spinal cord atrophy, and an EEG detected slow background in the index patient. Whole exome sequencing identified a homozygous missense variant in the adaptor protein (AP) complex 2 alpha-2 subunit (AP2A2) gene. Western blot analysis showed reduced levels of AP2A2 in patient-iPSC derived neuronal cells. Endocytosis of transferrin receptor (TfR) was decreased in patient-derived neurons. In addition, we observed increased axon initial segment length in patient-derived neurons. Xenopus tropicalis tadpoles with ap2a2 knockout showed cerebral edema and progressive seizures. Immunoprecipitation of the mutant human AP-2-appendage alpha-C construct showed defective binding to accessory proteins. We report AP2A2 as a novel genetic entity associated with HSP and provide functional data in patient-derived neuron cells and a frog model. These findings expand our understanding of the mechanism of HSP and improve the genetic diagnosis of this condition.

1. Introduction

Hereditary spastic paraplegia (HSP) is a heterogeneous group of neurological disorders characterized by slowly progressive spasticity, weakness, and hyperreflexia in the lower extremities, resulting from corticospinal tract degeneration (Panza, 2022). The estimated worldwide prevalence of HSP is 1.8 cases per 100,000 people for both autosomal dominant and recessive forms, with a prevalence ranging from 0.5–5.5 and less than 5.3 cases per 100,000 people, respectively (Ruano et al., 2014). The classification of HSPs is based on clinical features, inheritance pattern, and underlying molecular mechanisms (Sireesha et al., 2021). Clinically, they are classified as pure or complex depending on the absence or presence of additional findings. HSP is pure or uncomplicated when spasticity is the main clinical manifestation along with weakness and urinary disturbance. Complex or complicated HSP denotes the presence of additional neurologic and non-neurologic features including cerebellar dysfunction, axonal or demyelinating peripheral neuropathy, intellectual disability, epilepsy, extrapyramidal features, brain and spine abnormalities, ophthalmologic manifestations, dysmorphic features, and/or orthopedic anomalies (Panza, 2022; Sireesha et al., 2021; Blackstone, 2018).

In addition to their broad phenotypic spectrum, HSPs are among the most genetically diverse neurodegenerative disorders associated with over 80 genetic loci and over 50 mutant genes (Sireesha et al., 2021; Blackstone, 2018). These diseases can be inherited through autosomal dominant, autosomal recessive, X-linked recessive, or mitochondrial inheritance (Blackstone, 2018). While dominant HSPs are more prevalent in northern Europe and North America, recessive cases are mostly seen in North Africa, the Middle East, and Mediterranean regions, which have higher rates of consanguineous marriage (Ruano et al., 2014; Sireesha et al., 2021; Blackstone, 2018; Erichsen et al., 2009). Complicated forms of HSP most often present with autosomal recessive inheritance (Meyyazhagan et al., 2022; Sireesha et al., 2021; Koh et al., 2018). In some populations, SPG4 and SPG11 types are the most common inherited forms of HSP, and constitute 45% of pure form (SPG4) and 40% of the complicated form (SPG11) respectively (Solowska et al., 2015; Pérez-Brangulí et al., 2014). Despite the wide genetic heterogeneity, many suspected familial cases remain without an identified cause.

Most genetically confirmed African cases of HSP have been reported in North Africa (Mahungu et al., 2022; Coutinho et al., 1999; Lesca et al., 2003; Boukhris et al., 2008; Bouslam et al., 2005; Stevanin et al., 2008; Boukhris et al., 2009). However, several novel variants in known HSP genes including SPG10, SPG11, SPG35 and a novel gene causing SPG43 have been reported in sub-Saharan Africa (West Africa), particularly in the Malian population (Mahungu et al., 2022; Guinto et al., 2017; Landouré et al., 2020; Landouré et al., 2019; Landouré et al., 2013; Meilleur et al., 2012; Diarra et al., 2022). In this study, we report AP2A2 as a novel complicated HSP gene in a consanguineous Malian family with three patients who presented with progressive lower limb spasticity, seizures, and intellectual disability. AP2A2 encodes one of two isoforms of the α subunit of adaptor protein-2 (AP-2), a heterotetrameric complex involved in clathrin-mediated endocytosis (CME) (McMahon et al., 2011; Dell’Angelica et al., 2019; Ball et al., 1995). We did clinical, genetic, cellular studies in patients, unaffected family members, and unrelated healthy control individuals, and animal studies. The results of our studies show that the variant decreases the levels of AP2A2 and CME specifically in neurons. Moreover, the mutation impairs the ability of AP2A2 to interact with accessory proteins involved in clathrin-dependent endocytosis, and ap2a2 knockout in Xenopus tropicalis causes cerebral edema and progressive seizures. These findings establish defective CME as a novel cause for HSP.

2. Materials and Methods

2.1. Patients

Patients were recruited from our neurogenetics study in Mali. Patients were classified based on previously published diagnostic criteria (Nagase et al., 1988; Schüle et al., 2006) and genetic variants were classified according to American College of Medical Genetics and Genomics (ACMG) recommendations (Richards et al., 2015). Subjects completed extensive neurological examinations to confirm the clinical phenotype of HSP in the Neurology Department of the Teaching Hospital of Point “G” in Bamako. All patients with progressive spastic paraplegia were included in this study. All available unaffected family members were enrolled as controls for evaluating disease segregation. Patients were classified using the following diagnostic criteria: 1) pure form spastic paraplegia, 2) spastic tetraparesis with earlier and greater severity in the lower limbs, or 3) spastic paraplegia as an early prominent sign (within the first 3 years of the disease) of a degenerative disease affecting several parts of the nervous system and 4) exclusion of other causes of the presenting symptoms. The pure form of HSP was diagnosed by the presence of lower extremity hyperreflexia, clonus, and plantar extensor response with or without urinary incontinence and weakness. The complex form of HSP was diagnosed when additional features including neurological findings such as intellectual disability, seizures, peripheral neuropathy, cerebellar ataxia or non-neurological findings of foot deformity, vision or hearing impairment were present. Socio-demographic features such as sex, age at examination and at onset, ethnicity, region or country of origin, profession, and disease features such as disease onset, walking difficulty, lower extremity spasticity, progressive weakness, foot deformities, and bladder urgency were collected from patients and relatives. Family history was also recorded by assembling a family pedigree to establish the disease transmission pattern. Laboratory testing was done for human T-cell leukemia virus type 1 (HTLV-1) by serology and PCR, human immunodeficiency virus (HIV), vitamin B-12 levels, very long chain fatty acid, and brain and spinal cord MRI imaging were done to support the diagnosis and exclude acquired and non-genetic causes of spastic paraplegia such as infection, inflammation, tumor, trauma and leukodystrophies in the nervous system. The HTLV-1 testing was done at the National Institutes of Health (NIH) by polymerase chain reaction (PCR). Other neurological evaluations included electroencephalography (EEG) to evaluate for epileptiform activity when patients presented with seizure.

The study was approved by the ethics committee of the Faculty of Medicine and Dentistry (N^o 2018/182/CE/Faculté de Médecine et d’Odonto-Stomatologie (FMOS), Bamako, Mali). Patients and all available relatives were examined by a group of experienced neurologists after giving informed and written consent for the collection of questionnaires, family history for pedigree, medical imaging, biological samples, subsequent testing, and the use of the photographs in this report before enrollment in the study, according to the Declaration of Helsinki. For minors and intellectually disabled patients, consent was obtained from a parent or guardian.

2.2. Genetic testing and screening

Candidate gene testing including ALSIN and ERLIN2 were performed in our laboratory followed by an HSP targeting next-generation DNA sequencing gene panel including 58 HSP genes, SPG4 deletion, and mtDNA mutations in a CLIA-certified laboratory (Medical Neurogenetics, Atlanta, GA). Whole exome sequencing (WES) was done on all available affected individuals, the unaffected mother, and two unaffected siblings. Additional information on genetic testing methods can be found in the supplementary material. Minor allele frequency (MAF) was calculated using dbSNP, gnomAD/ExAC Brower, 1000 Genomes, HGMD, and Clinvar. Computational pathogenicity was assessed by in silico approaches using algorithms scores from SIFT, Polyphen-2, CADD, and MutationTaster (MT) for variant prediction. Identified variants were classified as pathogenic, likely pathogenic, uncertain significance, likely benign, or benign according to American College of Medical Genetics and Genomics (ACMG) recommendations. This recommendation classifies variants using the five criteria of population data, computational and prediction data, functional data, segregation data, and de novo data. Candidate gene function and protein expression were evaluated in GeneCards, the Human Protein Atlas, and e!Ensembl. Conservation of variants of interest from mammals to birds was checked using Uniprot. The variant of interest was screened in 303 ethnically matched Malian unaffected controls, HSP cohort databases from Miami (United States) and Paris (France), and submitted to GeneMatcher for phenotype match screening. Phenotype similarity screening of candidate genes was done manually by comparison to related gene family members.

2.3. Cell lines

Dermal skin fibroblasts were obtained by three-mm punch biopsy of the anterior forearm. Induced pluripotent stem cells were generated using the CytoTune Sendai reprogramming Kit and cultured in E8 media (Invitrogen). Stem cells were differentiated into cortical neuron-like cells (iNeurons) by stable insertion of an inducible transcription factor cassette expressing neurogenin-2 (NGN2) into the CLYBL safe harbor locus as previously described (Fernandopulle et al., 2018). Dermal fibroblasts were expanded in Dulbecco’s Modified Eagle’s Medium (DMEM, Cat# 112-319-101, Quality Biological) supplemented with 10% v/v fetal bovine serum (FBS, Cat# 35–011-CV, Corning), 50 U/mL penicillin, 50 μg/mL streptomycin (Cat# 30002-CL, Corning), passaged at a 1:3 splitting ratio using trypsin-EDTA (0.25%), and grown at 37°C with 5% CO₂. iPSCs were generated from dermal fibroblasts, and a second control iPSC line (WTC11) was obtained from the Coriell Institute.

Differentiation of iPSC was initiated on Matrigel coated dishes using DMEM/F12 medium containing N2 supplement, non-essential amino acids, L-glutamine, 10 μM ROCK inhibitor, and 2 μg/mL doxycycline. For microscopic characterization, the cells were dissociated with Accutase 2 days after differentiation and plated on poly-d-lysine/laminin coated dishes. iPSC-derived neurons (iNeurons) were maintained in BrainPhys^™ Neuronal medium (Cat# 05790, STEMCELL technologies) supplemented with BDNF (Cat# 450–02, Peprotech), recombinant human NT-3 (Cat# 450–03, Peprotech), laminin (Cat# 23017015, Thermo Fisher Scientific), B-27 (Cat# 17504044, Thermo Fisher Scientific), and doxycycline. Cell lysis and western blotting methods can be found in the supplementary material.

2.4. Transferrin uptake and imaging

Clathrin-mediated endocytosis was measured by assaying transferrin uptake in patient, unaffected mother, and wild type control fibroblasts and their respective iPSC-derived cortical-like neurons (iNeurons). 20,000 fibroblasts and iPSC-derived neurons (Day 7) were cultured on coverslips. The following day, the cells were incubated with uptake medium (DMEM with 1 % BSA for fibroblasts, and BrainPhys^™ with 1 % BSA for neurons) for 30 min at 37°C. The medium was replaced with prewarmed uptake medium containing 25 μg/mL Alexa 647-conjugated human transferrin (Cat# T23366, Thermo Fisher) for 5 min at 37°C. The cells were immediately placed on ice and washed twice with ice-cold PBS for 5 min, and then fixed in 4 % v/v paraformaldehyde for 30 min at RT. For transferrin uptake assays cells were washed three times with PBS, then mounted with ProLong^™ Gold Antifade Mountant with DAPI, and examined by confocal fluorescence microscope using a Zeiss LSM 780 inverted confocal microscope (Carl Zeiss) fitted with a Plan-Apochromat 63X objective (NA=1.4). Maximum intensity projections were generated from the Z-stack images, and composite images were prepared using ImageJ/Fiji and quantified with ImageJ software. After background subtraction, the mean fluorescence intensity of Tf-Alexa 647 was measured. Data are presented as mean values +/− SEM. N=the number of independent experiments. n=the total number of analyzed cells. One-way ANOVA were used for statistical testing.

Fibroblasts and 7-day-old iNeurons were initially seeded on 24-well chambered slides at 20,000 cells/well. For fixation, slides were washed with phosphate-buffered saline (PBS) and treated for 30 min with 4% v/v paraformaldehyde (PFA) with 4% w/v sucrose in PBS at room temperature (RT), then incubated with 0.1M glycine in PBS for 5 min, and subsequently washed three times with PBS. Cells were permeabilized in PBS supplemented with 0.1% w/v saponin and 1% w/v bovine serum album (permeabilization buffer) for 30 min at RT, and then incubated in permeabilization buffer for 30 min at 37°C with the following diluted primary antibodies: rabbit anti-AP2A2 (Cat# LS-C482433, LSBio, 1:50), goat anti-AnkG (Cat# sc31778, Santa Cruz Biotechnology, 1:50), chicken anti-MAP2 (Cat# ab5392, Abcam, 1:500), mouse anti-β-tubulin, (Cat# 801201, BioLegend, 1:1000). After three washes with permeabilization buffer, cells were further incubated with secondary antibodies diluted in antibody permeabilization buffer for 30 min at 37°C in the dark, then washed three times in PBST and once with PBS. Finally, coverslips were mounted on glass slides using ProLong^™ Gold Antifade Mountant with DAPI (Cat# P36931, Thermo Fisher), and confocal images were captured using a Zeiss LSM 780 inverted confocal microscope (Carl Zeiss) fitted with a Plan-Apochromat 63X objective (NA=1.4). Maximum intensity projections were generated from the Z-stack images, and composite images were prepared using ImageJ/Fiji (https://fiji.sc/).

2.5. Axon initial segment analysis

The axon initial segment (AIS) and the soma/dendrites were analyzed by immunofluorescent staining of ankyrin G and MAP2, respectively, using goat anti-AnkG (Cat# sc31778, SantaCruz biotechnology, 1:50) and chicken anti-MAP2 (Cat# ab5392, Abcam, 1:500). DAPI was used for nuclear staining. The length of the AIS was measured using ImageJ (https://imagej.net/ij/).

2.6. Protein-protein binding assay

Purified human AP-2 αC appendage domain was used for protein-protein binding assays. GST-fusion proteins were expressed in DH5α competent cells at 25°C overnight. Expression was induced by adding 0.2 mM IPTG (Isopropyl ß-D-1-thiogalactopyranoside) in LB medium containing 100 μg/mL ampicillin. The cells were lysed by three freeze/thaw cycles, sonicated five cycles for 10 seconds at 4°C, and insoluble material was centrifuged at 35,000 g at 4°c for 20 minutes. Protein was loaded on a glutathione-Sepharose column for purification following the manufacturer’s protocol for spin purification of GST-tagged proteins (Thermo Scientific, #16108). Protein was concentrated and purified using centrifugal filter units Amicon Ultracel-30k (Millipore, # UFC903008). The final yield of GST purified protein was assessed to be >95% as shown by Coomassie blue staining on 10% SDS-PAGE analysis. Brain extract was prepared by homogenizing human control cervical spinal cord (National Disease Research Interchange, NDRI) in buffer A [150 mM NaCl, 20 mM HEPES pH 7.4, 4 mM DTT with Roche protease inhibitor tablet] as previously described (Wigge et al., 1997, Owen et al., 1998). Interaction assays were performed by incubating 20 μg of purified GST fusion protein with 0.5 mL of 1 mg/mL human spinal cord lysate containing 0.1% TritonX-100 with 20 mL of 50% slurry of glutathione-Sepharose beads in buffer A for 1 hour at 4°C with shaking. Beads were washed three times for 5 minutes with buffer A. After washing, bound proteins were analyzed on 10% SDS-Tris glycine gel by immunoblotting analysis with the following cargo accessory antibodies: anti-EPS15 (HPA008451,1:500, Sigma-Aldrich), anti-AP180 (sc58229, 1:200, mouse, Santa Cruz biotechnology), anti-amphiphysin (#610714, 1:1000, mouse, BD Transduction laboratories), anti-auxilin or anti-DNAJC6 (HPA031182, 1:100 rabbit, sigma-Aldrich, Prestige antibodies powered by Atlas antibodies), anti-epsin (HPA06220, 1:200, rabbit, Sigma-Aldrich) and anti-clathrin heavy chain (610499, 1;500, mouse, BD Transduction laboratories). Methodology for cDNA cloning can be found in the supplementary material.

2.7. Animal model system: genome editing in Xenopus tropicalis

The benefit of Xenopus tropicalis (X. tropicalis, frog) tadpoles has been widely reported (McQueen et al., 2017; Wen et al., 2017; Boskovski et al., 2013). CRISPR-Cas9-mediated genome editing in X. tropicalis tadpoles was used as previously described (McQueen et al. 2017; Bhattacharya et al., 2015). sgRNAs were injected at 400 pg/embryo along with 1.6 ng of Cas9 protein (CP03, PNA Bio) into single-cell fertilized embryos according to standard methods (Ran et al., 2013). Un-injected control (n=120), Ap2a2-exon5 (n=72) and ap2a2-exon9 (n=75) tadpoles were placed in individual wells of a 48-well plate at stage 42 and observed from stage 42 to stage 48 of development. Positive controls for seizure activity included both targeting the neurod2 gene as previously reported in early infantile epileptic encephalopathy in Xenopus tadpoles (Sega et al., 2019), and treatment with the powerful, non-fatal chemoconvulsant pentylenetetrazol (PTZ) (Hewapathirane et al., 2008; Bell et al., 2011). Additional methodology on the Xenopus model can be found in the supplementary material.

3. Results

3.1. Clinical review details

We identified a consanguineous Malian family of Soninké ethnicity with nine siblings in which three individuals exhibit symptoms of HSP. The index, a 11-year-old male (V.12), his 25-year-old sister (V.7), and a brother who died at 9 years of age (V.9) presented with a similar phenotype. Individual V.9 was reported by his family to be affected and died before evaluation. The disease distribution in this family is consistent with a recessive inheritance pattern (Figure 1A). Neurological evaluation of 16 unaffected relatives in this family including parents was normal.

Fig. 1. Characterization of a family with complicated hereditary spastic paraplegia.

(A) Pedigree of the family showing consanguinity between unaffected parents with a recessive inheritance pattern. The arrow indicates the proband. Ages at examination are shown at the top of each symbol, and asterisks (*) indicate those seen in clinic for history and examination. Individual V.9 died at 9 years of age before an evaluation could be performed. Generation is indicated on the left side in roman numerals and subject number is shown at the bottom of each symbol, with segregation of the AP2A2 variant shown in the nuclear family. Muscle atrophy (blue arrow) and skeletal deformities in patient#1 (11-year-old boy) with severe lower limb atrophy (B) and “club foot” deformities (C) before Achille’s tendon lengthening surgery (single green arrow). Patient#2 (25-year-old female) with equinus foot deformities as shown by double green arrows (D). (E) sagittal brain MRI in both affected individuals showing thinning of the corpus callosum and atrophy of the medulla (red arrows) and spinal cord (red circle) with normal brain structure in the unaffected mother. Coronal images show moderate widening of cerebral sulci in both patients (purple arrows) alongside normal imaging in the unaffected mother.

The index patient and his older sister (V.12 and V.7, respectively) presented with early-onset of gross motor developmental delay at 8 months of age. They both had delays in crawling and walking, and did not achieve any physical milestones in the first years of life. Both patients walked with difficulty by 24 months of age; distance walking abilities were also delayed and running and climbing were challenging. Gait difficulties began to progressively worsen with the onset of lower extremity spasticity, leading to frequent falls in the first five years of life for both patients. The parents noticed bilateral leg muscle atrophy and skeletal foot deformities. The feet were turned inward, resulting in a clubfoot deformity in the lower extremities at age 9 in the affected older sister. Both affected siblings failed their primary education due to learning disabilities. They developed generalized epilepsy with one to two tonic clonic seizures per year beginning at ages 9 (V.12) and 15 (V.7). Seizures subsequently became recurrent and increased in frequency from about one episode every 6 months to once a month, necessitating antiepileptic treatment. Gait symptoms worsened, leading patient V.7 to become wheelchair-bound at age 22, and the proband V.12 to require physical assistance for walking. Neurological examination of the two affected individuals found spasticity, diffusely brisk reflexes, bilateral extensor plantar responses (positive Babinski signs), and bilateral ankle clonus in the lower limbs. There was distal quadriceps muscle weakness in both patients with MRC scores of 2/5 (V.7) and 3+/5 (V.12), severe leg and calf muscle atrophy in the proband (Figure 1B), and clubfoot and pes cavus deformities (Figure 1C–D) in the proband and older sister, respectively. Due to the severity of the foot deformities, both patients underwent reconstructive surgical treatment that consisted of lengthening the Achilles tendon near the heel and releasing tissues elsewhere. Tendon elongation in the proband resulted in near-normal walking in the first six months after corrective surgery, foot orthosis bracing, and physical therapy. However, after six months, the proband had decreased ambulation. The older sister’s (V.12) Achilles tendon elongation surgery was done 10 years before evaluation and a similar evolution was reported. Evaluation of the upper extremities showed hyperreflexia and normal strength. There were no visual, sensory, or hearing manifestations. In addition, both patients had moderate intellectual disability, and failed calculation testing. These symptoms were more marked in the older sister, who presented with additional behavioral changes including verbal aggressiveness, inappropriate talk, and both temporal and spatial disorientation. Both patients had moderate (V.12) to severe (V.7) spastic dysarthria on initial examination. Evidence of cerebellar ataxia was not detected. Additional manifestations were consistent with severely abnormal brain activity with recurrent generalized tonic-clonic seizures in both patients that were only partially responsive to treatment with phenobarbital. Electroencephalogram (EEG) was done in both patients and showed a slow background in the proband and a normal pattern in the older sister (Supplemental Figure 1A–B). Brain MRI of both patients showed thinning of the corpus callosum (TCC), particularly in the body and isthmus regions, with atrophy of the spinal cord and medulla (Figure 1E). Cerebral sulci were very wide in the older sister and moderately wide in the proband. A brain and spinal cord MRI of the unaffected mother did not reveal any abnormalities including structural anomalies of the corpus callosum or spinal cord. Vitamin B-12 and very long chain fatty acid levels were within normal range, and serologies of HTLV-1 (a cause of tropical spastic paraplegia) and HIV-1/2 were negative. HTLV-1 serology was confirmed negative by PCR. None of the other 19 family members enrolled as controls had similar symptoms or signs. These features are consistent with a complicated form of HSP with childhood onset and autosomal recessive inheritance. A detailed description of clinical phenotype and imaging findings is shown in Table I.

Table 1.

Clinical and laboratory findings in two patients with mutation in AP2A2

	Patient V.12	Patient V.7 (Proband)
Age seen (year) and sex	25, F	11, M
Onset at age	8 months	7–8 months
Parental consanguinity	First degree	First degree
First symptom	Motor development delay/walking difficulty	Motor development delay/walking difficulty
Leg spasticity and hyperreflexia	Present	Present
Babinski sign and ankle clonus	Bilateral	Bilateral
Lower limb weakness and atrophy	2/5, severe	3+/5, severe
Skeletal deformity	Equine feet	Clubfoot
Intellectual disability	Severe	Moderate
Additional manifestations	Epilepsy and dysarthria	Epilepsy and dysarthria
Epilepsy age (year) and types	15, Tonic-clonic generalized seizures	9, Tonic-clonic generalized seizures
Anti-epileptic treatment responsive	Partially responsive	Partially responsive
Wheelchair-bound age	22	Ambulant with assistance
Brain MRI findings	TCC, SM, atrophy of SC and CC	TCC, SM, atrophy of SC and CC
EEG findings	Normal	Slow background
HTLV-1 serology and PCR	Negative	Negative
Vitamin B12 and VLCFA	Normal	Normal

MRI: Magnetic resonance imaging, TCC: Thin corpus callosum, SM: Small medulla, SC: Spinal cord, CC: Cerebral cortex, EEG: Electroencephalogram, HTLV-1: Human T Lymphoma virus 1, PCR: Polymerase chain reaction, VLCFA: Very long chain fatty acids.

3.2. Genotyping and variant identification

Candidate gene testing excluded mutations in ERLIN2 and ALS2 genes, two known causes of HSP. An HSP next-generation sequencing panel did not identify any other known HSP-causing mutation (Supplemental Table 1). As the family pedigree suggested a recessive inheritance pattern, homozygosity mapping (HM) was completed in 9 individuals including both patients and the following unaffected family members: maternal grandmother, mother, three siblings, maternal great aunt, and paternal aunt (III.6, IV.5, VI.8, V.I10, VI.11, VI.3, III.9, respectively). Eight of the shared homozygosity regions for both patients were identified on 6 chromosomes (Supplemental Figure 2A, Supplemental Table 2). The two affected siblings had the highest proportion of the genome with regions of homozygosity in the family (V.7 – 6.4%, V.12 – 13.5%). Whole exome sequencing (WES) was done in all patients, the unaffected mother, and two healthy siblings (V.7, V.12, IV.5, V.10, V.11), and filtering was performed using both VarSifter 1.7 and IVA 5.6 pipelines. Exome sequencing data in both patients showed a homozygous missense mutation in the coding region of AP2A2 at position c.2767A>G resulting in a substitution of the protein at residue 923 in exon 22 (g.1,010,572; transcript ID ENST00000448903.7, c.2767A>G; p.Ser923Gly). Integrated Genome Viewer (IGV) was used to visualize the variant as well as the sequence coverage quality for the AP2A2 coding region (Supplemental Figure 2B). The Ser923Gly variant is not present in public SNP and local databases including 1000 Genomes, gnomAD/ExAC browser: 0.000000, dbSNP (rs1167095747), ClinVar and HGMD, OMIM, or GeneCards. Four predictive computational in silico programs (CADD, MutationTaster, SIFT, and PolyPhen-2) were used to predict pathogenicity. The variant is predicted to be pathogenic with CADD (30, deleterious), MutationTaster (56, deleterious), SIFT (0.20, likely deleterious) and PolyPhen (0.89, possible damaging) (Supplemental Table 3). AP2A2 protein is expressed in most tissues with the highest expression in the spinal cord, and both fetal and adult brain (#KIAA0899, e!Ensembl and NCBI). DNA sequencing confirmed the mutation segregation with the disease status in the family. Patients have the mutation in a homozygous state (Supplemental Figure 2C), while the unaffected mother (VI.5), maternal grandmother (III.6) and two unaffected siblings (V.10 and V.11) were heterozygous for the mutation. An unaffected sibling sister (V.8) was a normal homozygous reference.

Sequencing of 303 ethnically matched Malian controls did not show the sequence variant. We subsequently sequenced the full coding region of AP2A2 in 23 Malian patients with diverse forms of HSP and none had the c.2767A>G variant or other mutations in this gene. Mutations in AP2A2 were not detected in an HSP database in Miami (United States) and in 56 families with recessive HSP (41 from Sudan, 10 of Algerian origin, and 5 of European origin) from a database in Paris (France). The Paris database also included 10 dominant cases of spastic paraplegia and 4 cases of isolated spastic paraplegia of European origin. To date, no confirmed match was found following our submission of the AP2A2 gene variant to GeneMatcher and no other HSP-causing mutation was found in AP2A2. The c.2767A>G variant in AP2A2 is classified as a variant of uncertain significance according to the ACMG recommendations.

3.3. Characteristics of AP2A2

AP2A2 is a protein-coding gene located on chromosome 11p15.5 (Fig. 2A). The ~104-kDa AP2A2 protein (also known as αC) is one of two isoforms of the α subunit of the heterotetrameric (α-β2-μ2-σ2) adaptor protein-2 (AP-2) complex (Fig. 2A–B) (McMahon et al., 2011). The other isoform is the ~108-kDa AP2A1 or αA (Figure 2B) and is encoded by a different gene on chromosome 19q13.33. These isoforms share approximatively 84% amino-acid sequence identity (Robinson, 1989). Alternative splicing of the AP2A1 isoform (αA^L) in the brain generates a protein with a 22 amino-acid deletion in the hinge region (αA^S) and a similar molecular mass as AP2A2 (Dell’Angelica et al., 2019; Ball et al., 1995; Robinson, 1987). A third αA isoform (αA^S) is a short transcript of αA^L, encoding a protein with a deletion of 22 amino acids and the same molecular weight as AP2A2. αC (AP2A2) and αA^S (AP2A1) are expressed in most tissues, with highest levels of αC (AP2A2) In the spinal cord and fetal and adult brain (Nagase et al., 1988). αA^L (AP2A1) on the other hand, is brain specific (Ball et al., 1995; Scorilas et al., 2002).

Fig. 2. AP2A2 cytogenetic localization, coding regions, complex structure, and protein conservation.

(A) AP2A2 is found on chromosome 11p15.5 with 22 exons (grey rectangles). The N-terminal domain is shown in green (residues 27– 680) from trunk to hinge region, and appendage domain (residues 701–939 including β-sandwich (yellow) and the C-terminal appendage platform domain (light blue). The variant of interest Ser923 location is marked by a red dashed line and circle in both the exon and domain diagrams. (B) The schematic illustration of the AP-2 complex with the four subunits colored in blue (α-subunit), purple (β2- subunit), red (μ2-subunit) and gray (σ2-subunit). Regions labeled on diagram indicate binding sites as well as sites involved in clathrin-mediated endocytosis. (C) Protein sequence alignment of AP2A2 in various species (amino acid numbers in relation to the human sequence: 902–939). The AP2A2 mutation causes an amino acid substitution at residue Ser923 (black asterisk) with conserved residues indicated in red.

AP-2 is a member of the adaptor protein (AP) family of complexes that also includes AP-1, AP-3, AP-4 and AP-5. These complexes are components of protein coats involved in intracellular protein trafficking. AP-2 is specifically involved in clathrin-mediated endocytosis (CME) of cargo proteins from the plasma membrane. The AP-2 complex consists of a “core” domain comprising the N-terminal “trunk” domains of the α and β2 subunits and the full-length μ2 and σ2 subunits. The C-terminal parts of the α and β2 subunits are structured as long, disordered “hinge” domains followed by “appendage” or “ear” domains (Traub et al., 1999) (Figure 2A–B). The appendage domain comprises an N-terminal β-sandwich subdomain and a C-terminal platform subdomain (Figure 2A–B). The core is involved in interactions with plasma membrane phosphatidylinositol-4,5-bisphosphate (PIP2) and cargo proteins, the hinge domains interact with clathrin, and the appendage domains interact with accessory proteins (McMahon et al., 2011; Collins et al., 2002; Jackson et al., 2010) (Fig. 2B). The Ser923Gly mutation is found within the platform subdomain in a location known to interact with endocytic accessory proteins (Figure 2B). This residue is conserved across species from mammals to birds as well as in AP2A1, suggesting an essential role (Figure 2C).

Mutations in AP2A2 have not previously been linked to disease, but mutations in two of the four subunits of the AP-2 complex, AP2S1 and AP2M1, are associated with two Mendelian disorders: familial hypocalciuric hypercalcemia type 3 (FHH3), and developmental and epileptic encephalopathies (DEE), respectively. Mutations in subunits of two other adaptor complexes, AP-4 and AP-5, are associated with complicated forms of HSP, namely, SPG47, SPG50, SPG51 and SPG52 for AP-4, and SPG48 for AP-5. Patients in this study shared a similar inheritance mode and phenotype with AP-4- and AP-5-disease causing mutations, including motor and behavior deficits as well as thinning of the corpus callosum. However, our patients presented with epilepsy and intellectual decline and did not manifest symptoms like those in familial hypocalciuric hypercalcemia type 3 (FHH3). Moreover, our patients had an earlier age of seizure presentation.

3.4. The Ser923Gly mutation reduces AP2A2 levels and endocytosis in neurons

To examine the effect of the Ser923Gly mutation on the properties of the AP-2 complex, fibroblasts from skin biopsies were collected from patients and related controls. iPS cells with normal karyotype were subsequently generated from fibroblasts and differentiated to cortical neuron-like cells (iNeurons) (Supplemental Figure 3C–D). Most familiar α-adaptin antibodies cross-react with all three isoforms of the α-adaptin proteins. It is important to distinguish the specific α-adaptin isoforms (AP2A1 and AP2A2) and the use of specific anti-α-adaptin antibody isoforms is crucial. Herein, we used specific commercially available antibodies for AP2A2 (LSC156387) and AP2A1 (610501, BD transduction) as reported in a previous study (Sukanya et al., 2022). Also, a published specific rabbit polyclonal AP2A2 antibody (C619–656) produced by the laboratory of Dr. Margaret Robinson was used for additional western blot and immunofluorescence staining (Ball et al., 1995). Fibroblasts from a control, the unaffected mother and both patients (Patient#1: older sister-V.7 and patient#2: Proband-V.12) showed similar AP2A2 levels by western blotting (Figure 3A, Supplemental Figure 3A) and similar immunofluorescence (IF) staining patterns (Figure 3B). Moreover, no change was observed in the levels of the paralogous AP2A1 protein (Supplemental Figure 3B). We next evaluated these fibroblasts for transferrin uptake, which is dependent on CME. We observed that fibroblasts from the control, unaffected mother and both patients exhibited similar levels of transferrin uptake (Figure 3C–D).

Fig. 3. The Ser923Gly substitution in AP2A2 does not affect the protein level and transferrin uptake in fibroblasts.

(A) Western blot and (B) immunofluorescence microscopy analyses shows no change in the protein level and staining pattern for AP2A2 in fibroblasts. (C, D) Tf-Alexa 647 uptake for 5 min showing similar uptake in patient relative to control fibroblasts. Tf fluorescence intensity quantification analysis per cell normalized to control in fibroblasts from immunofluorescence staining. N=3 biological replicates, n=50 cells, with the mean of each replicate shown by the larger colored circles, statistical significance was analyzed by one-way ANOVA with Dunnett’s multiple comparison test; error bars show SEM; *p<0.05, **p<0.01. ns (not significant). Tf (transferrin). Scale bars for B and C are 10 μm, and corresponding zoomed inset image scale bars are 5 μm.

In contrast to fibroblasts, iNeurons showed significantly decreased AP2A2 levels in both patients relative to the mother and control (Figure 4A–B, Supplemental Figure 4A). Western blot analysis using an antibody recognizing both AP2A2 and AP2A1 showed the presence of several immunoreactive species with different molecular weights, with the patient iNeurons exhibiting fewer high molecular weight bands relative to the mother and control iNeurons (Figure 4C–D; Supplemental Figure 4B–C). These patterns suggested a possible post-translational modification (PTM) defect in AP2A2/AP2A1. Treatment of control and mother neuronal lysates with lambda phosphatase caused disappearance of the two higher bands (Figure 4E, Supplemental Figure 4D), indicating that these correspond to phosphorylated forms of the proteins and that the patient mutation reduces phosphorylation.

Fig. 4. The Ser923Gly substitution in AP2A2 results in decreased protein level and transferrin uptake in iPSC-derived cortical neurons.

(A) Anti-AP2A2 western blot analysis in iPSC-derived neurons (iNeurons) shows a similar protein level in the control and unaffected mother, while patient #1 and patient#2 iNeurons have reduced AP2A2 protein levels. (B) Quantification of anti-AP2A2 blot signal intensity normalized to control in iNeurons. N=3 biological replicates, analyzed by one-way ANOVA with Bonferroni’s multiple comparison test. (C, D) Anti-AP-2 α-adaptin western blot analysis in iNeurons in 4–20% (C) and 6% gels (D), showing an alteration in the expression pattern in patient iNeurons cells. Patient samples show reduction of upper molecular weight bands. (E) Anti-AP2 α-adaptin western blot analysis before (P−) and after phosphatase treatment (P+) in control and unaffected mother iNeurons showing a shift in the upper molecular weight pattern of AP-2 alpha expression following treatment. (F) Anti-AP2A2 immunofluorescence in 7-day old iNeurons from control, unaffected mother, and both patients. (G, H) Patient iNeuron immunofluorescence staining shows reduced transferrin uptake of Tf-Alexa 647 in 7-day old iNeurons following 5 min of incubation, and Tf fluorescence intensity quantification per cell normalized to control in iNeurons. N=3 replicates, with the mean of each replicate shown by the larger colored circles, n=50 cells. (I) Axonal initial segment analysis (AIS) in iNeurons using anti-AnkG immunofluorescence. (J) Quantification of AIS length per cell normalized to control. N=3 replicates, with the mean of each replicate shown by the larger colored circles, n=50 cells; statistical significance was analyzed by one-way ANOVA with Dunnett’s multiple comparison test; error bars show SEM, *p<0.05, **p<0.01, ***p<0.005, ****p<0.0001. ns (not significant). Tf (transferrin), AIS (axon initial segment), IF (immunofluorescence), iNeurons (iPSC-derived cortical neurons). Scale bars:10 μm.

Immunofluorescence microscopy showed a similar pattern of AP2A2 immunofluorescent staining in the control, mother and patient iNeurons (Figure 4F). However, day-in-vitro 7 iNeurons from both patients showed reduced transferrin endocytosis relative to the control and mother iNeurons (Figure 4G–H). In addition, patient iNeurons exhibited a longer axon initial segment (AIS) relative to control and mother iNeurons, as detected by staining for the AIS marker ankyrin G (AnkG) (Figure 4I–J). This phenotype may arise from decreased endocytosis of AIS plasma membrane proteins, as previously shown for neurofascin (Torii et al., 2020). The somatodendritic staining for MAP2 was unaltered in the patient cells, suggesting that neuronal polarity was not affected (Figure 4I).

Taken together, the cellular studies show that the Ser923Gly mutation reduced AP2A2 levels and transferrin endocytosis in iNeurons, but not in fibroblasts. In addition, the Ser923Gly mutation increased the length of the AIS in iNeurons, a phenomenon that may be related to decreased endocytosis.

3.5. CRISPR-Cas9 mediated knockout in X. tropicalis

Based on the above results, we hypothesized that the p.Ser923Gly variant caused a loss of function of the AP2A2 protein. To further test this hypothesis, we used a X. tropicalis model in which we assessed the in-vivo effect of ap2a2 knock-out using CRISPR-Cas9. Un-injected controls showed normal behavior throughout developmental stages 43–48 (Figure 5A and Supplemental Video 1). However, ap2a2-exon1/CRISPR tadpoles displayed head deformities (cerebral edema) and ap2a2-exon4/CRISPR showed mixed phenotypes including head deformities and abnormal swimming with knock-out scores of 43 and 33, respectively. Some tadpoles with ap2a2-exon5 and ap2a2-exon9/CRISPR treatment had abnormal swimming behavior at stage 43 with ventral side up, compared to wild-type tadpoles that swim dorsal side up. At stage 45, ap2a2-exon5 and ap2a2-exon9 tadpoles initiate spontaneously seizing with C-shaped convulsions where the body axis is changed to a “C” curve and the head curves toward the tail in a “C” shape. This phenotype is also detected in PTZ treated and neurod2-CRISPR X. tropicalis tadpoles (Figure 5B and Supplemental Video 2+3). Ap2a2-exon1 and ap2a2-exon4 tadpoles did not age to a maturation stage needed to observe convulsions. Both ap2a2-exon5 and ap2a2-exon9 CRISPRs independently cause spontaneous seizures in frog tadpoles, with C-shaped convulsions followed by periods of post-ictal paralysis. Seizure frequency increases with continued development (Figure 5C). At stage 45, ap2a2-exon5 CRISPR tadpoles displayed seizure-like behavior for 3.528 s/min (n=72), while ap2a2-exon9 displayed seizure-like behavior for 4.135 s/min (n=75), and un-injected controls displayed seizure-like behavior for 0.173 s/min (n=98). Seizure behavior became more pronounced by stage 46, with ap2a2-exon5 CRISPR tadpoles seizing for an average of 6.194 s/min (n=72), ap2a2-exon9 tadpoles seized for an average of 9.627 s/min (n=75) for and un-injected controls displayed seizure-like behavior for 0.049 s/min (n=102).

Fig. 5. CRISPR-Cas9 mediated knockout induces seizure-like behavior in Xenopus tropicalis tadpoles and analysis of wildtype AP2A2-ear and Ser923 AP2A2-ear binding to EPS15, AP180, amphiphysins, auxilin, epsin and clathrin.

(A) Un-injected control (UIC) tadpole shows normal swimming behavior with straight tail (top image) and (B) while ap2a2-exon9 exhibit seizure-like behavior (c-shaped convulsion), with tail bending towards the head (bottom image). (C) Quantification of seizure-like behavior in seizures per minute, showing increase in frequency of seizures episodes in tadpoles injected with both non-overlapping CRISPRs (ap2a2-exon5 and ap2a2-exon9) over the course of development. Data are a compilation of 3 independent experiments with the mean of each replicate shown by the larger colored circles. Analyzed by one-way ANOVA with Bonferroni’s multiple comparison test; error bars show SEM, **p<0.01, ****p<0.0001, ns (not significant). Con = uninjected control, Ex5 = ap2a2-exon5 guide, and Ex9 = ap2a2-exon9 guide injected. (D) The interaction of wildtype AP2A2-ear and Ser923Gly AP2A2-ear with candidate cargo proteins is shown following incubation of purified AP2A2 protein fragments incubated with total brain lysate. Western blot shows AP2A2 wildtype and mutant protein complexes, beads alone, and total brain input samples probed with antibodies for EPS15, AP180, Amphiphysin, auxilin, Epsin, and clathrin heavy chain. (E) Quantification from immunoblotting accessory protein pull down assay signal intensity normalized to wildtype AP2A2-ear. Two biological replicates were used to quantify each accessory protein signal mean intensity showing that the presence of Ser923Gly impaired the cargo protein interaction by 80% for AP180, 70% for auxilin, and modest reductions of 30% and 20% for Eps15 and amphiphysin respectively. Equal binding was observed to Epsin and only one pull down was performed. No interaction was observed for clathrin heavy chain to wildtype AP2A2-ear and Ser923Gly AP2A2-ear. Error bars show SEM.

By stage 47, ap2a2-exon5 CRISPR tadpoles were seizing-like for an average of 15.972 s/min (n=72); ap2a2-exon9 tadpoles seized for an average of 15.947 s/min (n=75), while un-injected controls displayed seizure-like behavior for an average of 0.308 s/min (n=120). At stage 48, the average time spent seizing for ap2a2-exon5 CRISPR tadpoles was 35.708 s/min (n=72); ap2a2-exon9 tadpoles seized for an average of 29.627 s/min (n=75), while un-injected controls had seizure-like behavior for an average of 0.387 s/min (n=119).

The editing efficiency of the targeting sgRNAs ap2a2-exon5 (n=10 animals) and ap2a2-exon9 (n=10 animals) was analyzed using ICE (Intersection Control Evaluation) Synthego software. Ap2a2-exon5 was found to have 86% indel (82% out-of-frame indel, 4% in frame deletions), 3% unedited and 7% “other” (defined as greater than 700bp deletion, or 500bp insertion) with a knockout-score of 82 (supplemental Figure 5A). However, ap2a2-exon 9 analysis shows 69% indel (63% out-of-frame deletions and 6% in-frame deletions), 26% unedited, 5% “other” with a knockout-score of 63 (Supplemental Figure 5B). Ap2a2-exon5 and ap2a2-exon9 sequences were compared with un-injected control sequences (Supplemental Figure 5C–D). Significant disruption of the ap2a2 gene by CRISPR-Cas9 in Xenopus tropicalis leads to spontaneous seizure-like behavior in tadpoles that progresses with developmental time, further supporting the role of AP2A2 in seizure behavior in our patients.

3.6. AP2A2 Ser923Gly mutation causes a defect in protein-protein interactions

Binding studies of wildtype and mutant Ser923Gly AP2A2 protein were performed to evaluate their interactions with known AP-2 accessory proteins, including EPS15, epsin, AP180, amphiphysin and auxilin (Figure 5D). The purity of glutathione S-transferase eluted purified AP2A2 appendage domain wildtype and mutant proteins was assessed to be >95% as shown by Coomassie blue staining (Supplemental Figure 6). We observed that the mutant Ser923Gly protein showed reduction of binding with all the accessory proteins tested except for epsin (Figure 5E, Supplemental Figure 7).

3.7. Proposed mechanism for the effect of Ser923Gly on AP-2 function

In this study we used genetic testing, cell line characterization, and both cellular and in vitro functional assays to show that the AP2A2 c.2767A>G (p.Ser923Gly) mutation contributes to cause neuronal specific functional defects. The Ser923Gly mutation decreases the levels of AP2A2 and results in an alteration of AP2A2 migration pattern in iNeurons. Moreover, mutant iNeurons have reduced transferrin uptake and increased axonal initial segment length. Lastly, purified Ser923Gly mutant protein showed decreased ability to bind accessory proteins involved in clathrin-mediated endocytosis. We propose that the AP2A2 Ser923Gly mutation results in destabilization of the AP-2 complex and reduced ability of the mutant protein to bind accessory cargo proteins, with consequent defects in endocytosis (Figure 6).

Fig. 6. Proposed pathogenic mechanism of the AP2A2 p. Ser923Gly mutation resulting in a defect in cargo membrane protein selection and hereditary spastic paraplegia gene illustration.

(A) Model shows the wildtype AP-2 complex (top) and mutant (bottom with red star) in cargo selection and vesicle assembly by clathrin for endocytosis vesicle formation. From top to bottom, the four subunits are colored (blue: αC; purple: β2; red: μ2; and gray: σ2) in green: cargo, in yellow: accessory protein, and in teal for clathrin cage assembly. Bottom, the clathrinmediated endocytic pits carrying the mutant Ser923Gly of AP2A2 with attenuated numbers of cargo membrane proteins. (B) An illustration of the genes mutated in patients with hereditary spastic paraplegia and evidence of thinning of the corpus callosum on MRI imaging.

4. Discussion

HSPs are a heterogenous group of neurogenetic disorders that have been clinically characterized in several populations throughout Europe, Asia, North America, and North Africa (Blackstone, 2018, Erichsen et al., 2009; Mahungu et al., 2022; Coutinho et al., 1999; Lesca et al., 2003; Boukhris et al., 2008; Bouslam et al., 2005; Stevanin et al., 2008; Boukhris et al., 2009; Do et al., 2022; Bouty et al., 2019; Ortega et al., 2019). However, there are only six reports in sub-Saharan Africa (Mahungu et al., 2022). The rarity of HSP in this region may be due to the high prevalence of other diseases including infection, cancer, and trauma of the brain and spinal cord that could overshadow genetic causes. Exclusion of acquired or non-genetic causes of spastic paraplegia is necessary to establish the clinical diagnosis of HSP. In several sub-Saharan African countries, there are limited resources that may prevent a complete diagnostic assessment of possible HSP cases including limited availability of MRI, CT scans, equipped laboratories, genetic testing facilities, and the expertise and infrastructure to establish the genetic and molecular basis of these diseases. Consequently, there is limited epidemiological data on HSP in general and few genetically confirmed causes in sub-Saharan Africa, in particular (Mahungu et al., 2022; Diarra et al., 2022). In this study, we conducted a family case study to summarize the socio-demographic data, clinical characterization, and molecular phenotyping in a consanguineous Malian family with three patients seen at the Neurology Department of the Teaching Hospital of Point “G”, Bamako, Mali. The recessive inheritance in this family is likely due to the high prevalence of intra-ethnic and consanguineous marriage in Mali (Diarra et al., 2022; Sangaré et al., 2014 ; Landouré et al., 2017) and other populations (Boukhris et al., 2009; Yalcouyé et al., 2022). HSPs may occur at any age, from childhood to late adulthood (Ruano et al., 2014; Sireesha et al., 2021; Koh et al., 2018). The age at onset in this family was in the first year of life (8-months of age) and patients were first seen in our clinic in their teenage years, which may be a result of initially seeking the care of traditional healers known to be nearby and affordable. The delay may also be related to the deficit of specialists who are frequently located in big cities and generalists who may misdiagnose HSP as cerebral palsy or spinal cord trauma.

Major clinical features in both patients included spasticity, brisk reflexes, ankle clonus, and extensor plantar reflexes with normal upper limbs reflexes. These features were found to be present in the majority of HSP patients reported in the literature (Siressha et al., 2021; Blackstone, 2018; Diarra et al., 2022). Complicated forms of HSPs are typically associated with recessive inheritance compared to the dominant form of disease (Sireesha et al., 2021; Koh et al., 2018; Chrestian et al., 2017). A recessive inheritance was also found in our family with a complicated HSP presentation. Complicated cases of HSP can present with additional neurological symptoms including intellectual disability, epilepsy, dysarthria, brain and spinal cord MRI abnormalities and non-neurological signs such as orthopedic anomalies including a “club foot” deformity. The older patient (V.7) with active epilepsy had more intellectual disability and abnormal brain MRI findings compared to the proband who presented with slowing on EEG. These findings have also been reported previously in HSP studies (Helbig et al., 2016; Hardies et al., 2015). The older sister became wheelchair dependent at age 22, and the index (V.12) required assistance for walking at age 18 (V.12). Several large studies have reported that complicated symptoms are linked to disease severity (Do et al., 2022; Schüle et al., 2016; Chrestian et al., 2017), and our study would also suggest that these features are important in the clinical diagnosis and prognosis of HSP.

To date, up to 50 genes along with 85 different spastic gait loci have been associated with HSP, but many other clinically entities remain with no molecular diagnosis (Sireesha et al., 2021). In our family we completed genetic characterization and functional studies for deep phenotyping. A targeted gene panel did not identify the disease-causing variant in our HSP family, which may be due to the previously reported genetic diversity in Africa. The use of next-generation targeted gene panel testing is widely used in the diagnosis of HSP and identifies a variant in 25% to 52.5% of cases (Burguez et al., 2017). Panels screen for mutations within genes known to be associated with HSP, and do not detect variants in undiscovered HSP genes or deep intronic mutations. The use of such panels has limitations particularly in the African population where known genetic variation is expanding and population data is limited. Altogether, WES or WGS approaches should be considered for the initial evaluation of HSP, especially for poorly characterized genetic populations. Exome sequencing was completed along with shared homozygosity analysis in two affected and available unaffected family members to identify a rare missense homozygous variant in the novel HSP gene: Adaptor Related Protein Complex-2 Subunit Alpha-2 (AP2A2). The variant at codon Ser923Gly segregated with the disease in the family and was classified as “likely pathogenic” according to ACMG criteria since it satisfied the following categories including: being absent in public and local SNP databases, prediction to be pathogenic by in-silico prediction, and functional characterization suggesting a loss of the normal function of AP2A2. Moreover, we excluded a variant with common minor allelic frequency in gnomAD that was detected in one of the first 10 Malian controls screened for this variant.

In this study, we evaluated the molecular function of the pathogenic homozygous mutation identified in the AP2A2 gene, encoding the alpha-C subunit of AP-2 complex protein. Importantly, AP-2/AP2A2 function has previously been described to be critical in clathrin-dependent endocytosis in all known eukaryotic cells (Dell’Angelica et al., 2019; Traub et al., 1999; Owen et al., 1999; Moravec et al., 2012; Kadlecova et al., 2017). Since clathrin itself does not bind directly to the cargo or membrane, endocytosis depends on AP-2 and accessory protein recruitment to the plasma membrane. Clathrin coated vesicle formation is achieved through five steps: 1) nucleation with FCHO, EPS15, 2) cargo selection with AP-2, Epsin, AP180, 3) coat assembly with clathrin, 4) scission involving dynamin, amphiphysin, and 5) uncoating with HSC70, auxilin, GAK, for the early endosomal delivery of endocytic cargo (McMahon et al., 2011; Traub et al., 1999).

We hypothesize that the location of the mutant Ser923Gly in the AP-2 alpha appendage domain may be causative by destabilizing the AP-2 complex as suggested by the alteration of AP2A2 protein stability and endocytosis function in iNeurons. Endocytosis is fundamental to cellular function and accomplished by an invagination of the plasma membrane to form an intracellular bounded vesicle carrying plasma membrane and extracellular proteins for essential cellular processes including nutrient intake, cell polarity, and adhesion. In this study, the endocytosis function of cargo receptors was evaluated using the transferrin receptor as described previously (Owen et al., 1999; Wigge et al., 1997). Our results show similar AP2A2 protein levels and transferrin uptake in patient fibroblasts. In contrast, patient iNeurons had lower AP2A2 protein levels and a reduction in transferrin uptake after 7 days in culture. This result suggests neuronal cells have a cell type specific vulnerability, supported by the fact that AP2A2 protein is highly expressed in the fetal and adult brain and spinal cord (#KIAA0899, e!Ensembl And NCBI).

AP-2 is a member of family of coat proteins that also includes AP-1, AP-3, AP-4, and AP-5. Mutations in subunits of these coat complexes cause congenital disorders referred to as “coatopathies” (Dell’Angelica et al., 2019). Until now, the AP-2 complex subunit alpha-2 (AP2A2) was not associated with HSP or other diseases, but mutation in the subunit σ2 (AP2S1) was reported to cause familial hypocalciuric hypercalcemia type 3 (FHH3) (Nesbit et al., 2013), and mutation in μ2 (AP2M1) (Helbig et al., 2019) was reported to cause developmental and epileptic encephalopathy. Clinically, our patients presented with similar epileptic and intellectual decline manifestations as seen in AP2M1 patients, although AP2M1 patients typically have an earlier age at seizure onset than patients in this study. The difference in seizure onset could be that epilepsy is not the primary phenotype in our family, but an additional associated manifestation. The coat complexes AP-4 and AP-5 subunit members are associated with five different complicated autosomal recessive HSPs: SPG47, SPG50, SPG51, SPG52 (AP-4) (Hardies et al., 2015; Jamra et al., 2011; Bauer et al., 2012; Verkerk et al., 2009; Moreno-De-Luca et al., 2011), and SPG48 (AP-5) (Slabicki et al., 2010). In comparison to AP-2, AP-4 and AP-5 are part of non-clathrin coats and are associated with the trans-Golgi network (TGN) (Dell’Angelica et al., 1999; Hirst et al., 1999) and late endosomes (Hirst et al., 2011; Hirst et al., 2013), respectively. The two patients in this study share a similar inheritance mode and phenotypic mechanism to those with AP-4 and AP-5 mutations, including motor and behavior deficits and thinning of the corpus callosum. These similarities can be explained by the fact all of these complexes are involved in neuronal cargo sorting. However, the pathogenic mechanism is different and dependent on the specific compartment affected by the cargo delivery: AP-2 for early endosomes, AP-4 for the TGN, and AP-5 for late endosomes. Importantly, our results suggest a neuronal specific destabilizing effect for the AP-2 Ser923Gly mutation. Neurons are known to be highly polarized cells in which protein distribution within axons and dendrites is critical for neuronal function. The axon initial segment (AIS), a cellular region between the soma and the axon, serves an important role in action potential initiation and modulation (Kole et al., 2007; Kole et al., 2008), and is enriched in ion channels, cell adhesion molecules, and cytoskeletal proteins (Torii et al., 2020, Hedstrom et al., 2007). Our findings show that length of the AIS was increased on immunofluorescence staining in both patient iNeurons. This phenotype may arise from reduced endocytosis of AIS plasma membrane proteins, as previously shown for neurofascin (NF-1860) (Torii et al., 2020). This result may explain the predisposition for epilepsy in our patients due to the impaired organization of ion channels and in line with previous studies that have characterized AIS elongation and excitability (Kuba et al., 2010; Kuba et al., 2015). To the best of our knowledge, the association between mutations in adaptor proteins with the AIS has not previously been reported.

Further exploration using X. tropicalis as a model system supports the role of ap2a2 in maintaining a seizure-free state, as shown with the replication of patients’ seizure phenotype in tadpoles. CRISPR-Cas9 mediated knockout of ap2a2 suggests that normal function of AP2A2 is crucial to prevent spontaneous seizures. This hypothesis is further supported by the alteration of the AIS in Ser923Gly iNeurons, which may predispose patients for epilepsy.

We hypothesized that the mutation interfered with AP-2 interaction with accessory proteins. Our data indicates that a single binding site on the alpha-adaptin appendage domain is present in the Ser923 protein on the C-terminal domain and we assessed for accessory protein interaction by immunoprecipitation. Residue Ser923 has been shown to have variable effects on binding ranging from severe to moderate effects on target accessory proteins known to have direct binding site to the appendage domain of alpha-adaptin (Verkerk et al., 2009) including EPS15, epsin, AP180, amphiphysins, and auxilin. This residue is conserved from mammals to birds as well as between αA (AP2A1) and αC (AP2A2). This study identified that at least four proteins have a severely reduced interaction with the mutant AP-2 appendage domain. The equal binding of epsin seen here was also reported in a previous study (Owen et al., 1999). This finding was supported by the fact that the DPW motif present in Epsin binds differently than the DPF motifs present in the other ligands or that epsin may have more compatibility than the other adaptin accessory proteins tested here. The number of DPF/W sequence motifs was reported to be correlated with affinity for the appendage domain (Owen et al., 1999). This would suggest that EPS15 has stronger affinity for the alpha appendage domain than AP180, auxilin or amphiphysin, which is compatible with the previous observation that EPS15 remains bound to AP-2 complex when isolated from brain extract (Owen et al., 1999; Benmerah et al., 1995). The alpha appendage domain is involved in coordinating the recruitment of components involved in endocytosis in time and space, reinforcing its critical role in the formation of clathrin-coated vesicles.

5. Conclusions

In summary, AP-2 complexes serve important roles in the brain and spinal cord, and we hypothesize that the Ser923Gly variant impairs AP-2 function and underlies some forms of complicated HSP. A limitation of this study is the identification of HSP in a single family. Additional genetic testing of more affected individuals is needed, particularly in Africa (Choudhury et al., 2020). Although we have evaluated a defect in AP-2 biology using several approaches to model the mutation in AP2A2, it is possible that other shared variants in this consanguineous family are contributing to the clinical features. This study thus suggests a novel genetic entity of HSP in the Malian population with thorough molecular diagnostic testing. We report a complicated form of HSP with early onset, autosomal recessive inheritance, and impairment of AP2A2 protein in neuronal cells. These findings will expand the clinical spectrum and genetic landscape of HSP. As genetic testing becomes less costly and more accessible, the discovery of additional novel genetic entities may be found in other populations.

Supplementary Material

2

Supplementary material 5: Supplemental figure 1. Analysis of Electroencephalogram of patients. (S1A+B) Electroencephalogram (EEG) traces in patient #1, a 11-year-old boy showed slow background and normal EEG waves in 25-year female patient #2.

Supplemental figure 2. Homozygosity mapping (HM) image and variant visualization on Integrate Genome viewer (IVG). (S2A) Homozygosity mapping in this family showed a shared region in red brackets from 1–2, 411,738 was identified on chromosome 11. This homozygous region is shared only by both patients (clear box: Homozygous region) but not shared by the seven families’ members (black dot box: Heterozygous region). Patients labelled in red and family members in black in panel label. (S2B) Variant Ser923Gly on AP2A2 was visualized on Integrate Genome viewer (IGV) in full orange box for both patients at the top panel where the nucleotide A was substitute by G in homozygous state (Orange: G/G) and unaffected family’s members in heterozygous state: (Green/Orange: A/G). Nucleotide A (Green), G (Orange), T(Red) and C (Blue). (S2C) Electropherogram images of the novel AP2A2 sequence variant: c.2767A>G (pSer923Gly). DNA sequencing shown in both patients with abnormal homozygous (G/G), while the unaffected mother is heterozygous (A/G) and one unaffected sibling shows a normal homozygous pattern (A/A). Substituted variant position is denoted by an asterisk.

Supplemental figure 3. Alpha-adaptin full blot images in fibroblasts and iPSC karyotyping and characterization. (S3A+B) AP2A2 and AP2A1 protein levels in fibroblasts showing similar protein levels for both isoforms. GAPDH was used for loading control (36 kD) for AP2A2 and beta actin (42 kD) for AP2A1. Lane 1: Control, lane 2: unaffected mother, lane3: patient#1(V.7) and lane 4: patient#2 (V.12). (S3C) karyotype in patients and parent analyses were normal for all characterized iPSC clones. Unaffected mother (50-year-old Female, VI.5, XX), patient#1 (25-year female, V.7, XX) and patient#2 (11-year-boy, V.12, XY). (S3D) Immunofluorescence of differentiated iNeurons showing staining for the neuronal marker β-tubulin (red). Scale bar = 25μM.

Supplemental figure 4. Analysis of AP-2 alpha and AP2A2 protein levels in iNeurons and treated control iNeuron lysates with phosphatase. (S4A) AP2A2 protein levels in iNeurons with vinculin used as loading control (124 kD). (S4B) AP-2 alpha-adaptin protein levels in iNeurons 4–20% full blot showed decreased protein in both patients with HSP90 as loading control (124kD). (S4C) AP-2 alpha (panel C) and AP2A1 (panel C’) proteins level in iNeurons in 6% full blot and HSP90 as loading control (124 kD). L: ladder, lane 1: Control, lane 2: unaffected mother, lane 3: patient#1(V.7) and lane 4: patient#2 (V.12). (S4D) AP-2 alpha-adaptin protein levels from treated unaffected mother, wild-type control, and patient#1 iNeuron lysates with lambda phosphatase and HSP90 was used for loading control. Lane1+2: Unaffected mother and lane 3+4: wild-type control, lane 5+6: wild-type control, lane 7+8: patient #1, P− (before treatment), P+ (after treatment).

Supplemental figure 5. Editing efficiency analysis of ap2a2 sgRNA-CRISPR-Cas9 mediated knockout in X. tropicalis using ICE Synthego. (S5A) ap2a2-exon5-CRISPR-Cas9 showed 86% edited efficiency and (S5B) 69% for ap2a2-exon9-CRISPR-Cas9. Bar graphs show percentage of sequences with specific number of nucleotides gained or lost, divided into indel (Blue bar) and no edit (Red bar) and (n=10). (S5C) Electropherogram images of ap2a2-exon 5-CRISPR-Cas9 knockout (edited sample) showing disrupted sequence from the cut site compared to control sample (un-injected) and (S5D) electropherogram images from ap2a2-exon 9-CRISPR-Cas9 knockout (edited). Control samples (bottom panel) and edited samples (top panel) with cut site sequences underlined in black.

Supplemental figure 6. AP2A2-GST staining with Coomassie blue. The purity of glutathione S-transferase eluted purified AP2A2 appendage domain mutant (lanes 3+4) and wild-type (lanes 5+6) proteins were assessed to be >95% as shown by Coomassie blue staining for AP2PA2-GST (indicated by red arrow). Input lysates (lanes 1+2) and washes lysates (lanes 7+8) as negative control show multiple bands on Coomassie staining.

Supplemental figure 7. Immunoprecipitation (pull down) of wildtype AP2A2-ear and p.Ser923Gly AP2A2-ear. (S7A+B) Eps15 and AP180 pull down show reduced mutant binding and in (S7C+D) Amphiphysin and auxilin show less interaction with the mutant compared to wild type AP2A2-ear. However, (S7E) Epsin showed equal binding and (S7F) no binding was seen to clathrin. L: ladder, lane 1: Wild-type control alpha-C-ear, lane 2: mutant (p.Ser923Gly), lane 3: beads, lane 4: input (brain lysate).

3

Supplementary video 1: (Video 1) Un-injected wild type controls X. tropicalis tadpole showed normal behavior which swim dorsal side up while tadpoles with AP2A2 targeted reduction (videos 2 and 3) display spontaneously seizing phenotype with C-shaped convulsions (the body axis is changed to a “C” curve and the head curves toward the tail).

Highlights.

• AP2A2 mutation is a novel genetic entity associated with neurodegenerative disease

• Mutation causes cell type specific defects in adaptor protein levels and function

• Patient derived cortical neurons have alteration of axonal initial segment length

• ap2a2 knockout in Xenopus tropicalis recapitulates seizures detected in patients

Data availability

Data will be made available upon request.