Proteasomal activation ameliorates neuronal phenotypes linked to FBXO11-deficiency

Summary

Haploinsufficiency of FBXO11, encoding a ubiquitin ligase complex subunit, is associated with a variable neurodevelopmental disorder. So far, the underlying nervous system-related pathomechanisms are poorly understood, and specific therapies are lacking. Using a combined approach, we established an FBXO11-deficient human stem cell-based neuronal model using CRISPR-Cas9 and a Drosophila model using tissue-specific knockdown techniques. We performed transcriptomic analyses on iPSC-derived neurons and molecular phenotyping in both models. RNA sequencing revealed disrupted transcriptional networks related to processes important for neuronal development, such as differentiation, migration, and cell signaling. Consistently, we found that loss of FBXO11 leads to neuronal phenotypes such as impaired neuronal migration and abnormal proliferation/differentiation balance in human cultured neurons and impaired dendritic development and behavior in Drosophila. Interestingly, application of three different proteasome-activating substances could alleviate FBXO11-deficiency-associated phenotypes in both human neurons and flies. One of these substances is the long-approved drug Verapamil, opening the possibility of drug repurposing in the future. Our study shows the importance of FBXO11 for neurodevelopment and highlights the reversibility of related phenotypes, opening an avenue for potential development of therapeutic approaches through drug repurposing.

This work elucidates pathomechanisms underlying FBXO11-related neurodevelopmental disorder. Using the fruit fly and human neurons as model systems, we show that FBXO11 loss impairs important processes of neuronal development. Activation of the proteasome with small molecules ameliorates observed phenotypes, opening a new avenue for therapeutic development through drug repurposing.

Introduction

FBXO11 is a member of the F Box protein family and is a subunit of a nuclear E3 ubiquitin ligase complex, the SCF complex (SKP1-Cullin-F-Box), formed together with SKP1 and CUL1.^1,2,3 In this complex, FBXO11 is thought to be important for substrate recognition. Several substrates (e.g., BCL6, SNAIL, and BLIMP-1) that are ubiquitinated by the FBXO11-containing SCF complex and subsequently degraded have been reported.^{4,5,6,7,8,9,10,11,12,13,14} It has also been proposed that FBXO11 is involved in neddylation, another form of post-translational modification and a degradation signal.¹⁵ The function of FBXO11 has mainly been studied in the context of cancer. There, it has been associated, among others, with various B-cell malignancies,^14,16,17 glioblastoma,¹² gastric cancer,¹⁸ and pancreatic cancer.¹⁹ Furthermore, roles of FBXO11 in bone development,²⁰ erythropoiesis,¹¹ and regulation of immune responses, e.g., through regulation of MHC class II expression^4,5 have been discussed.

We and others have recently identified heterozygous, mostly de novo variants in FBXO11 as causative for a variable neurodevelopmental disorder (MIM #618089, intellectual developmental disorder with dysmorphic facies and behavioral anomalies (IDDFBA), also FBXO11-related neurodevelopmental disorder).^21,22,23 To date, more than 75 individuals with pathogenic or likely pathogenic variants in FBXO11 have been reported.^{21,22,23,24,25,26,27,28,29,30} The associated clinical spectrum is very variable, with intellectual disability as the core feature, and behavioral anomalies, hypotonia, and nonspecific facial dysmorphism as common features. Furthermore, variable other anomalies can occur.^21,23,24 The mutational spectrum encompasses likely gene-disrupting and missense variants, for which no genotype-phenotype correlations could be delineated regarding type and location of variants.²⁴ Functional studies have suggested a loss-of-function (LOF) effect also for missense variants, highlighting haploinsufficiency and LOF as the most likely disease mechanism for FBXO11-related NDDs.²⁴ Despite the recent association of FBXO11 LOF with neurodevelopmental phenotypes, little is known about its role in the nervous system.

To decipher the neurodevelopmental pathomechanisms associated with loss of FBXO11 function, we generated FBXO11-deficient human induced pluripotent stem cells (hIPSCs) and differentiated them to neural progenitor cells and neurons. We found that in cells lacking FBXO11, neuronal differentiation, proliferation, and migration are impaired. We could confirm alterations in neuronal function and morphology in an in vivo model for Fbxo11 deficiency in Drosophila melanogaster. Phenotypes associated with FBXO11 deficiency could be ameliorated with Verapamil and other proteasome-activating substances in both model systems, highlighting a potential reversibility of phenotypes and a possible mode of therapeutic intervention through drug repurposing.

Material and methods

Cell culture

All cells were handled under sterile conditions and cultured at 37°C and 5% CO₂ and 90% relative humidity.

A control human IPSC line generated from cord blood-derived CD34+ lymphocytes obtained from Thermo Scientific was used for all subsequent experiments (A-18944). This line was previously shown to be pluripotent and with differentiation potential into all three germ layers.³¹ hIPSCs were cultured in mTESR1 Plus medium (Stem Cell Technologies) and 1% Penicillin/Streptomycin (Thermo Scientific) and grown on Geltrex (Thermo Scientific) coated plates and media was renewed at least every other day. NPCs were maintained in DMEM/F12 with 0.5x N2, 0.5x B27 (Thermo Scientific), and 20 ng/μL FGF2 (Preprotech).

HEK293 cells were cultured in DMEM (Thermo Scientific) with 10% fetal bovine serum (FBS) (Thermo Scientific) and 1% Penicillin/Streptomycin (Thermo Scientific).

hIPSC culture and generation of knockout lines

For generation of knockout cells, the CRISPR-CAS9 system was used. A single guide RNA targeting exon 4 of FBXO11 (ACCTGCTGAACAGTATCTTC) was designed using CRISPOR (http://crispor.tefor.net/)³² and cloned into a pX330-GFP plasmid. The guide RNA was predicted to only contain homology to other regions in the genome with at least four mismatches. To confirm efficient cutting of the selected guide RNA, a T7 assay was performed in HEK293 cells.

For genome editing in hIPSCs, 2.5 μg of plasmid was nucleofected into the control hIPSC line with the nucleofector 2B with program B16 using the human stem cell nucleofector kit 2 (Lonza). Seventy-two hours post nucleofection, GFP+ cells were sorted individually into 96-well plates. After 7–10 days, colonies were expanded to two 48-well plates that were used for cryopreservation with BamBanker medium (Nippon Genetics) and genotyping. Following DNA extraction, exon 4 of FBXO11 was amplified (F: TTAAGTGTATTCGAGTATTACCTTTGG, R: AATTGTAACTACCAACTCACCGC) and analyzed using Sanger sequencing. Three clones each carrying either compound heterozygous (KO-1-3) or heterozygous (HET-1-3) out-of-frame indels and wild-type clones without variants (WT-1-3) were identified, thawed, expanded, and used for further studies.

Validation of hIPSC lines

All used and generated hIPSC lines were tested for their genetic integrity using conventional karyotyping with G-banding and chromosomal microarray testing using the CytoScan HD array (Affymetrix) to exclude complex chromosomal rearrangements and copy number changes. Additionally, variants introduced by CRISPR-CAS9 gene editing in eight potential exonic off-target regions with four mismatches compared with the guide RNA sequence were excluded through PCR amplification and Sanger sequencing of these regions. Potential exonic off-target regions and primer sequences can be found in Table S9. Cell lines were genotyped regularly to re-confirm their identity.

Differentiation of hIPSC into neural progenitor cells

Differentiation of hIPSC to neural progenitor cells (NPCs) was performed using a modified version of the dual-SMAD inhibition protocol.³³ For this, a modified version of the STEMdiff SMADi neural induction kit (StemCell Technologies) was used. In brief, hIPSC colonies were treated with BMP inhibitor Dorsomorphin (1 μM, Tocris), TGFβ inhibitor A83-01 (2 μM, Tocris) and 1x N2 (Thermo Scientific) in DMEM/F12 medium 1 day prior to harvesting. HIPSC colonies were then dissociated using Accutase (Thermo Scientific) and plated in Aggrewell800 plates (StemCell Technologies) in STEMdiff SMADi neural induction medium (StemCell Technologies) at 10,000 cells per Aggrewell (3 × 10⁶ cells per line) to form embryoid bodies (EBs) for 5 days. EBs were harvested and plated on poly-ornithine/laminin- (POL, Sigma-Aldrich) coated plates in the same medium and grown to form neuronal rosettes for 7 days. Rosettes were selected using the neural rosette selection medium (StemCell Technologies) and plated on fresh POL plates (passage P0 of NPCs). NPCs were maintained in DMEM/F12 with 0.5x N2, 0.5x B27 (Thermo Scientific), and 20 ng/μL FGF2 (Preprotech).

Differentiation of NPCs to neurons

For differentiation of NPCs to neurons, the STEMdiff forebrain neuron differentiation and maturation kits were used (StemCell Technologies). NPCs were passaged using Accutase and kept in STEMdiff forebrain neuron differentiation on POL plates for 1 week. Following passaging using Accutase, cells were seeded in STEMdiff forebrain neuron maturation medium on POL plates at a density of 0.5–1 × 10⁶ cells/mL in 24-well plates. For staining, cells were seeded on coverslips.

Immunofluorescence

hIPSCs, NPCs, or neurons were grown on coated cover slips (hIPSC: Geltrex, NPC and neurons: POL) and fixated with 4% paraformaldehyde in PBS for 10 min. Following permeabilization with 0.1% Triton X-100 in PBS for 10 min, cells were blocked with 5% normal goat serum. Staining was performed with primary antibodies at indicated dilutions against NANOG (1:200, #4903, Cell Signaling), OCT4 (1:200, #75463, Cell Signaling), PAX6 (1:500, AB2237, Millipore), Nestin (NES, 1:1.000, MAB5326, Millipore), MAP2 (1:200, MAB3418, Millipore), Ki-67(1:1.000, #9449, Cell Signaling), phospho-histone H3 (pHH3, 1:500, ab10543, Abcam), and TUBB3 (1:200, #5568, Cell Signaling). An overview of all primary antibodies used can be found in Table S11. Secondary antibodies used were Alexa Fluor 546 donkey anti-mouse (1:1.000, A10036, Thermo Scientific) and Alexa Fluor 488 goat anti-rabbit (1:1.000, A11008, Thermo Scientific). Nuclei were counterstained with DAPI (Serva). Images were acquired on a Zeiss AxioImager Z2 with an Apotome 3 with a 20×, 40×, or 63× objective and the Zen software v3.4. Images were analyzed using ImageJ³⁴ (v1.53) and CellProfiler³⁵ (v4.2.5). CellProfiler was used for quantification of PAX6-positive, Ki-67-postive, pHH3-positive NPCs and TUBB3-positive neurons from at least five different images taken at random positions of the slides for each line.

Protein expression analysis and western blotting

For expression analysis, cells were harvested in RIPA buffer (50 mM Tris pH 8.0, 150 mM NaCl, 1% Igepal CA-630, 0.1% SDS, 0.5% sodium-deoxycholate). Proteins were separated by SDS-PAGE on 4%–20% Mini-PROTEAN TGX precast protein gels (Bio-Rad) and blotted onto nitrocellulose membranes using the semi-dry Turbo blotting system (Bio-Rad). Blots were stained with antibodies against FBXO11 (1:2,500, NB100-59826, Novus Bio), MAP2 (1:1,000), GAPDH (1:5,000, #2118, Cell Signaling), H3 (1:10,000, #4499, Cell Signaling), c-Myc (1:5,000, M4439, Sigma-Aldrich), c-Myc (1:2,500, #2272, Cell Signaling), FLAG (1:5,000, F7425, Sigma-Aldrich), FLAG (1:1,000, 194502, Addgene, gift from Melina Fan; http://n2t.net/addgene:194502; RRID:AB_2924869 ³⁶), and HA (1:2,000, H3663, Sigma-Aldrich). An overview of all primary antibodies used can be found in Table S11. Secondary antibodies used were goat anti-rabbit HRP (1:15,000, 170–6515, Bio-Rad) and goat anti-mouse HRP (1:15,000, ab97023, Abcam). Blots were developed with SuperSignal West Femto Maximum Sensitivity substrate (Thermo Scientific) and imaged using the Chemidoc Imaging system (Bio-Rad). For image analysis the Image Lab software v6.0.0 (Bio-Rad) was used. Band intensities of FBXO11 and MAP2 were normalized to GAPDH and H3, respectively. Significance was calculated using a one-sample t test with a theoretical value set to 1.

Drosophila stocks and maintenance

All flies were raised on standard food containing cornmeal, sugar, yeast, and agar unless otherwise specified. Crosses were kept at 28°C unless otherwise specified. To induce tissue-specific knockdowns the UAS-Gal4 system was used.³⁷ The GAL4-driver lines BL#3954 (Act5C-GAL4/Tm3 Sb Tb, ubiquitous expression), BL#8765 (elav-GAL4/CyO, pan-neuronal expression), BL#8746 (477-GAL4; UAS-mCD8GFP for multiple dendrite neuron-specific expression and simultaneous membrane-bound GFP-expression), BL#8860 (ms1096-GAL4, wing-specific expression), the RNAi line BL#31484 (RNAi 3), and its respective control line (BL#36303, Control 3) were obtained from the Bloomington Drosophila Stock Center (BDSC, Bloomington, IN, USA). The RNAi lines VDRC #24041 (RNAi 1) and VDRC #24039 (RNAi 2) and the respective control line (VDRC #60000, Control 1/2) were obtained from the Vienna Drosophila Research Center (VDRC, Vienna, Austria). Mushroombody-specific driver (UAS-Dcr-2(X); 247-GAl4) and the pan-neuronal line UAS-Dcr2; elav-GAL4 used for NMJ analysis were a gift from Annette Schenck (Nijmegen, The Netherlands).

A fourth RNAi line was obtained from VDRC (VDRC #109559, RNAi 4 with Control 4 [VDRC #60100]) but excluded from further analysis due to insufficient knockdown efficiency found using qPCR (Figure S6A). An overview of all used Drosophila stocks can be found in Table S10.

Drosophila viability and wing analysis

For viability analysis in Drosophila upon ubiquitous knockdown, hatching flies were assessed for either carrying the Tm3 Sb Tb balancer allele or the Act5C-GAL4 (knockdown) allele. At least 200 hatching flies for each genotype from three independent experiments were scored. For the analysis of the wing phenotype, the right wing of the fly was removed and mounted on coverslips in mounting solution (ethanol: glycerol = 2 : 3). Images were taken on a Zeiss Discovery.V8 stereo microscope. Length and width of wings (as marked in Figure S9A) were measured using ImageJ (v1.53).

RNA isolation

For RNA extraction from neurons, cells were harvested in RLT buffer, and RNA was extracted with the RNeasy Mini Kit (Qiagen). DNAse digest was performed on column with the RNAse free DNAse kit (Qiagen). For RNA extraction from Drosophila, three L3 larvae or 10 fly heads were pooled for each sample. Total RNA was isolated according to instructions of the RNeasy Lipid Tissue Mini Kit (Qiagen), except for QIAzol being substituted by TRIzol (Thermo Scientific). For homogenization, QIAshredder columns (Qiagen) were utilized.

Expression analysis using quantitative RT-PCR

To synthesize cDNA, SuperScript II reverse transcriptase (Thermo Scientific) and random primers were used. Expression analysis in Drosophila was performed with exon-spanning primers for Fbxo11 and Tubulin (used as endogenous control) and the PowerUP SYBR Green Master Mix (Thermo Scientific) on a QuantStudio 12K Flex (Thermo Scientific). Expression of Fbxo11 upon ubiquitous knockdown was compared with control flies with the same genetic background. For the quantitative real-time PCR, quadruplicates were used. To analyze the data, RQ values were calculated with the ΔΔCt method.

Whole transcriptome sequencing and analysis

RNA sequencing (RNA-seq) was performed in-house at the NGS core unit of the Medical Faculty of the FAU Erlangen-Nürnberg. For library preparation, the TrueSeq Stranded mRNA kit (Illumina) was used. Libraries were subjected to single-end sequencing (101 base pairs [bp]) on a HiSeq2500 platform (Illumina). Reads were converted to .fastq format and demultiplexed using bcl2fastq v2.17. Adapter trimming and quality filtering was performed using cutadapt v1.18.³⁸ Quality control after trimming was done using fastqc v0.11.8. Processed reads of RNA-seq on human samples were mapped to the human reference genome hg19 with the Ensembl gene annotation GRCh37.87 using STAR aligner v2.7.5a.³⁹ Read counts were obtained using the subread featureCounts program v.2.01.⁴⁰ For Drosophila samples, processed reads were mapped to the Drosophila reference genome BDGP6.11, Ensembl Annotation v98 using STAR aligner v.2.6.1.³⁹ Mapped reads were counted for all non-overlapping exons. The sum of all exons determines the number of reads per gene and was obtained using samtools v1.8 and subreadv1.6.1.⁴⁰ The following analysis were performed using R v3.6.1. Differential gene expression (DGE) analysis was performed using DESeq2 v1.24.0⁴¹ and for noise removal, the apeglm package⁴² was used.

Gene ontology enrichment analysis was performed in Panther^43,44 with a Fisher’s exact test and Bonferroni correction for multiple testing. As a background, a list of all genes for which expression was detected in RNA-seq was used. To filter out very broad and general terms for plotting graphs, only biological processes that contain less than 4,000 genes were used.

To integrate RNA-seq data from human neurons and Drosophila heads, Drosophila genes were mapped to its human orthologs using DIOPT ortholog finder.^45,46 For DGE, only the best human matches were retained. For enrichment analysis, multiple matches were allowed in both orientations (multiple fly genes mapping to one human ortholog and vice versa).

For enrichment analysis of gene expression changes during differentiation from hIPSC to neurons, a publicly available dataset from the Lieber Brain Institute was used.⁴⁷ Read counts were processed for differential expression analysis as described above. Differential gene expression analysis was performed for hIPSC (class: RENEW) compared with neurons cultured with astrocytes (class: NEURONS_PLUS_ASTROCYTES). For enrichment analysis, a hypergeometric test was performed.

Neurosphere assay

Neurosphere assay was performed as described previously.⁴⁸ In brief, NPCs were dissociated in Accutase and 1 × 10⁶ cells were plated in one well of an ultra-low attachment six-well plate. After 48 h, neurospheres have formed, which were plated individually in Matrigel- (Corning) coated 96-well plates in Matrigel. Cells were allowed to migrate for 48 h from the neurosphere before imaging. Images were taken on a Nikon Ts2-FL microscope with an Axiocam 305 color camera (Zeiss). Images were analyzed in ImageJ and the area of the plated neurosphere (t = 0 h) and the area covered after 48 h (t = 48 h) were measured. At least five neurospheres were analyzed per line (three WT, three HET, and three KO lines) per experiment and the experiment was performed three times. Significance was calculated using a Student’s t test.

XTT assay

Proliferation of NPCs and differentiating and maturing neurons was tested with the CyQUANT XTT Cell Viability assay (Thermo Scientific). NPCs of all nine lines (three WT, three HET, three KO) were plated on POL-coated 96 well plates in triplicate. On the day following plating, cells were switched to STEMdiff neuron differentiation medium. On the day following plating (D20), and at D23, D26, and D28 (differentiation) or at D29, D35, and D56 (maturation) cells were treated with XTT reagent according to the manufacturer’s instructions and incubated for 4 h. Following incubation, absorbance was measured using plate reader at wavelength 450 nm and 660 nm. For quantification, specific absorbance was calculated using the following formula: (Abs450(sample) − Abs450(blank)) − Abs660(sample). The absorbance for each sample was normalized to D20 (differentiation) or D29 (maturation), respectively. The experiment was performed three times. Significance was calculated using a Student’s t test.

Climbing assay

Climbing assay⁴⁹ in Drosophila was performed as described elsewhere.⁵⁰ In brief, at least 20 vials with 10 flies (five male and five female) each for each genotype tested were collected 0–48 h post eclosion. Following a recovery period from CO₂ anesthesia of at least 24 h, flies were transferred to test vials, tapped to the bottom of the vial, and filmed for 30 s while walking up. From the film, the time for 70% of the flies in each vial to climb up 8.8 cm was measured. The analysis was performed blinded to the genotype of the flies. Results were confirmed in an independent experiment. Statistical analysis was performed using a Wilcoxon signed rank test.

Bang sensitivity

To test seizure susceptibility of the flies, a bang sensitivity assay was performed as described previously.⁵⁰ In brief, flies were collected under CO₂ anesthesia as described for the climbing assay. For testing, flies were transferred to a testing vial, vortexed for 10 s, and filmed until all flies recovered. Flies showing seizure-like behavior 5 s after vortexing were counted. Statistical analysis was performed using a Wilcoxon signed rank test.

Activity monitoring

To test spontaneous locomotor activity of the flies, an activity assay was performed using the monitor system from Trikinetics (Trikinetics Inc.) as described previously.⁵¹ The assay was performed at 25°C and 70% relative humidity and a 12-h light/dark cycle. The tested cohort consisted of 32 3- to 5-day-old single male flies per genotype. For data collection, male flies were transferred into monitor tubes with standard food. Their activity counts (infrared beam passes by the fly) were registered every minute over the course of 4 days. To analyze the data, the ShinyR-DAM tool (https://karolcichewicz.shinyapps.io/shinyr-dam/) was used.⁵² Significance of difference of daily activity was determined on daily activity averages using a Student’s t test.

Courtship conditioning paradigm

Courtship conditioning was performed as described previously.⁵⁰ The assay was performed at 25°C and 70% relative humidity and a 12-h light/dark cycle with replicates performed over the course of 4 days. Virgin males were trained with premated females. Learning and short-term memory were assessed directly after training or after 1 h, respectively. The courtship index (CI) was assessed as time spent courting in a 10-min window. Naive and trained males were compared to calculate the Learning Index (LI): LI = (CInaive-CItrained)/CInaive. For testing statistical significance, a custom R script was used comparing differences in learning indices with a randomization test with 10,000 bootstrap replicates.⁵³

Analysis of neuromuscular junctions

Analysis of type 1b neuromuscular junctions (NMJs) of muscle IV was performed as described previously⁵³ using the UAS-Dcr2; elav-GAL4 driver line crossed with Fbxo11 RNAi lines or respective controls. Images were taken using a Zeiss AxioImager Z2 with 10× and 63× objectives. Images were stacked and analyzed in ImageJ. Synaptic area and length, and numbers of synaptic branches, boutons, and active zones were determined, blinded to the genotype. At least 25 NMJs from at least five different larvae were analyzed.

Analysis of dendritic arborization neurons

To analyze dendritic arborization (da) in Drosophila larvae, Fbxo11 RNAi lines and respective controls were crossed to the 477-Gal4; UAS-mCD8GFP driver line. L3 male larvae were dissected from the ventral side and fixated in 4% paraformaldehyde for 30 min. Solitary type IV multidendritic sensory neurons of the larval body wall were stained with an anti-mouse CD8A antibody (1:500, MCD0800, Thermo Scientific) and an Alexa 488 anti-rat secondary antibody (1:1,000, A21208, Thermo Scientific). z stack images were acquired on a Zeiss LSM 800 confocal microscope with a 20× objective. Image analysis was performed in ImageJ (v1.53) and number and length of dendrites were quantified using the NeuronJ plugin. Significance was calculated using a Student’s t test.

Substance treatment

For rescue experiments, PD169316 (Sigma-Aldrich, stock solution: 10mM in DMSO), R-Verapamil (Sigma-Aldrich, stock solution: 10mM in DMSO), or Verapamil (Sigma-Aldrich, stock solution: 10 mM in DMSO) was used for the fly experiments; they were added to the fly food prior to pouring vials at 1-μM final concentration. A concentration of 1 μM was determined to be most effective before, as outlined in Figure S9. As a control, the same volume of DMSO was added to control vials. Flies were either kept on substance containing food from the time of egg laying for the entire cross until time of experiments (developmental supplementation) or after flies had hatched (adult supplementation). For the cellular rescue, substances were added to the medium and renewed every other day at final concentrations of 20 μM for PD169316, 15 μM for R-Verapamil, or 10 μM Verapamil with DMSO used as the respective control.

Quantification and statistical analysis

Unless otherwise noted, data were analyzed using a Student’s t test. For Drosophila climbing and bang sensitivity assay, a Wilcoxon signed rank test was performed to determine statistical significance. For quantification of western blotting and coIP, a one-sample t test with a theoretical value set to 1 was done to determine statistical significance. Graphs were plotted using R version 4.3.0.

Results

Generation of a human neuronal model for FBXO11 deficiency

As haploinsufficiency of FBXO11 has previously been shown to be the most likely disease mechanism for FBXO11-associated NDDs,²⁴ we generated hIPSC FBXO11-deficient cells using CRISPR-CAS9 to study the effects of loss of FBXO11 in a human neuronal system (Figure 1A). We identified three wild-type (WT), heterozygous (HET), and compound heterozygous (KO) LOF clones each, which were used for further analysis (Table S1). By creating both HET and KO LOF hIPSC lines, we can both model the heterozygous situation found in affected individuals and study the effects of a complete loss of FBXO11. Genomic integrity and lack of off-target effects for all hIPSC clones was confirmed using conventional karyotyping, chromosomal microarray testing, and Sanger sequencing of selected loci. Stemness of all analyzed clones was confirmed using immunofluorescence staining with the stem cell markers NANOG and OCT4 (Figure S1). Almost complete lack of FBXO11 expression in biallelic knockout clones and reduced expression in heterozygous clones was confirmed via western blot analysis (Figures 1B and 1C).

Figure 1.

Generation of a neuronal FBXO11 deficient cell model using CRISPR-CAS9

(A) Outline of generation process of FBXO11 heterozygous and complete knockout hIPSC lines is shown, including validation of mutated clones and isogenic controls.

(B) Representative western blot of all nine FBXO11 hIPSC lines is shown. Blots were stained with anti-FBXO11 and anti-GAPDH antibodies.

(C) Quantification of three independent western blot experiments of FBXO11 hIPSC lines confirmed loss of FBXO11 in HET and KO lines. Individual values are shown as dots and mean values are shown as bars with SEM. Mean of all three WT controls was set to 1. p values were calculated using a one-sample t test with a hypothetical control mean set to 1. For all HET and KO lines compared with the control mean, reduction was significant at p < 0.01.

(D) Schematic outline of differentiation protocol from hIPSCs to NPCs and neurons with timeline and culture media used. Below, experiments performed here are indicated at various timepoints.

(E) Representative images of immunofluorescence of FBXO11 WT, HET, and KO NPCs stained with antibodies against neural progenitor markers Nestin (NES, red) and PAX6 (green) confirm differentiation to NPCs. Images of all nine NPC lines can be found in Figure S2. Images were taken on an AxioImager Z2 with Apotome 3 with a 40× objective. Scale bar, 20 μm.

(F) Quantification of PAX6-positive cells among NPCs shows comparable levels of PAX6-positive cells for WT, KO, and HET cells. Quantification for individual lines can be found in Figure S3. For quantification, cells were analyzed using CellProfiler identifying DAPI-positive cells and PAX6 positive cells (fraction of PAX6-positive cells = PAX6 stained cells/DAPI-stained cells).

For generation of a neuronal model of FBXO11-deficiency, hIPSC lines were differentiated into neural progenitor cells (NPCs) and later into neurons using a modified dual-SMAD inhibition protocol (Figure 1D).³³ Successful differentiation to NPCs was confirmed with staining of neural progenitor markers PAX6 and Nestin (Figures 1E and S2) with a PAX6 positivity rate of above 80% for all lines (Figures 1F and S3). No differences in NPC generation efficiency were observed between WT and HET or KO FBXO11 lines, suggesting that initial differentiation efficiency to NPCs is not affected by FBXO11 loss (Figures 1E and 1F). NPCs were then further differentiated and matured into neurons for different assays. Successful neuronal differentiation was confirmed by staining 1-week-old (D28), 3-week-old (D42), and 5-week-old (D56) neurons with neuronal markers MAP2 and TUBB3 (Figure S4).

Loss of FBXO11 leads to deregulated gene expression in neurons

FBXO11 encodes a nuclear ubiquitin ligase, so we hypothesized that key substrate proteins of FBXO11 may be involved in transcriptional regulation, an important process in neurodevelopment. To identify potential regulatory targets and to assess the effects of a complete loss of FBXO11 on gene expression, we performed RNA-sequencing experiments on 2-week-old early neurons (D35) harboring a biallelic FBXO11 loss. Transcriptome analysis was performed in three WT lines and three KO lines. The two conditions were clearly separated along the first principal component (Figure 2A). In FBXO11 KO neurons, 723 genes were deregulated with an adjusted p value less than 0.05 compared with WT neurons (Figure S5A; Table S2). Of these, 590 were downregulated in KO neurons, and 133 were upregulated. To identify common themes among deregulated genes, we performed gene ontology (GO) term enrichment using Panther gene enrichment analysis. Among all differentially expressed genes, many processes involved in cell differentiation, migration, adhesion, cell signaling and communication, and locomotion/motility were significantly enriched, all of which are involved and necessary for proper neuronal development (Figure 2B; Table S3). When specifically focusing on nervous system specific terms, we found that generation of neurons (GO:0048699), neuron differentiation (GO:0030182), nervous system development (GO:0007399), neurogenesis (GO:0022008), and neuron development (GO:0048666) were highly enriched (Figure S5B). Processes enriched separately in down- and upregulated genes were similar and comparable to all deregulated genes, suggesting that a complex deregulation of genes may impact development of FBXO11-deficient neurons (Tables S4 and S5).

Figure 2.

Gene expression changes due to loss of FBXO11 in human neurons and fly heads

(A) Principal-component analysis (PCA) of three FBXO11 WT and three KO neuron samples showed clear separation of WT and KO samples along the first principal component.

(B) Enriched gene ontology (GO) terms among differentially expressed genes were grouped based on function and show a broad enrichment of biological processes involved in, e.g., development, signaling, and migration. Detailed results on individual enriched GO terms can be found in Figure S5.

(C) Integration of GO term analysis of FBXO11-deficient human neuron and Drosophila head transcriptome analysis. The top five biological processes enriched in GO term analysis of human neurons are shown in black. The enrichment of these processes in Fbxo11-deficient Drosophila heads are shown in green.

(D) Stacked bar chart grouping genes expressed in FBXO11 KO neurons based on their differential gene expression and colored by corresponding expression changes during neuronal differentiation in a publicly available dataset on gene expression during differentiation from hIPSC to neurons.⁴⁷ Increasing expression during differentiation is marked in green, and decreasing expression during differentiation is shown in pink. Unchanged expression is shown in gray. Number of genes with increasing expression during differentiation is increased for genes downregulated in FBXO11 KO neurons. down = downregulated, up = upregulated, not sig = expression not significantly changed, exp. = expression.

To validate transcriptomic changes and functional consequences of loss of FBXO11 in vivo, we utilized D. melanogaster as a model system. Ubiquitous knockdown of Fbxo11 to roughly 50% using the UAS-GAL4 system is partially lethal at the late pupal stage (Figures S6A and S6B). We therefore turned to tissue-specific knockdown of Fbxo11 in the nervous system to further elucidate its neuronal role. We performed RNA sequencing on heads of flies with a neuronal Fbxo11 knockdown by two independent RNAi lines (RNAi 1 [n = 3] and 2 [n = 2]) and the respective control line (Control 1/2 [n = 3]). This also revealed a broad deregulation of 1,624 differentially expressed genes when comparing controls with both knockdown lines (Figures S5C and S5D; Table S6). A total of 1,236 of these genes had at least one human ortholog. We found that of the biological processes enriched among the deregulated genes in the human dataset, roughly 50% were also deregulated among the differentially expressed fly genes with human orthologs including four of the top five biological processes (27 of 53 processes) (Figure 2C; Table S7). The overlap of enrichment was even more pronounced when specifically focusing on genes downregulated in human neurons (18 of 20 processes) but was not present for the upregulated genes (1 of 18) (Table S7). This suggests that loss of FBXO11 both in Drosophila heads and human IPSC-derived neurons results in conserved disruption of (neuro-) developmental processes.

To follow up on the possibility of disturbed neuronal differentiation upon FBXO11 knockout, we investigated whether genes showing expression changes during differentiation from hIPSCs to neurons may be particularly affected. When linking differentially expressed genes found in this study to gene expression changes during differentiation from hIPSC through NPCs and neurons from a publicly available dataset from the Lieber Brain Institute,⁴⁷ we found that genes downregulated in FBXO11 KO neurons were enriched for genes whose expression increases with differentiation (hypergeometric test, OE = 1.4, p = 2.6 × 10⁻¹⁵) (Figure 2D; Table S8). This suggests that reduced expression of differentiation-specific genes may impair neuronal differentiation in FBXO11 KO cells and is in line with the gene ontology enrichment data. In contrast, upregulated genes in FBXO11 KO neurons were not enriched for genes whose expression increases or decreases with differentiation. FBXO11 therefore seems to be specifically important for proper execution of expression programs characteristic for neuronal differentiation.

Loss of FBXO11 impairs neuronal migration and differentiation

Following up on abrogated cellular processes in FBXO11-deficient neuronal cells, we assessed potential defects in neuronal migration using a neurosphere assay in FBXO11 HET and KO cells and in control NPCs. At 48 h after plating of neurospheres, we found that WT NPCs had migrated significantly farther than NPCs with a heterozygous loss of FBXO11 (Figures 3A, 3B, and S7A). A complete loss of FBXO11 led to almost complete lack of migration from the neurospheres (Figures 3A and 3B). This suggests that FBXO11 expression in NPCs is crucial for their migration capacities.

Figure 3.

Loss of FBXO11 alters neuronal migration, proliferation, and differentiation

(A) Representative images of neurosphere assay on FBXO11 WT, HET, and KO NPCs imaged 48 h after plating. Inner circle represents initial neurosphere size, outer circle represents migration after 48 h. Images were taken on a Nikon Ts2-FL microscope. Scale bar, 100 μm.

(B) Quantification of migration as ratio between area occupied at 48 h and plating (0 h). Migration is impacted in HET and more severely in KO neurospheres. Quantification of all nine individual lines can be found in Figure S7A. At least 30 neurospheres per genotype (≥8 neurospheres per line) from three independent experiments were analyzed. Significance was calculated using a Student’s t test.

(C) Representative images of immunofluorescence of FBXO11 WT, HET, and KO NPCs stained with antibodies against proliferation markers Ki-67 (red) and mitotic marker pHH3 (green) are shown. Images were taken on an AxioImager Z2 with a 20× objective. Scale bar, 100 μm.

(D) Quantification of Ki67-positive cells among NPCs shows increased levels of Ki67-positive HET and KO cells. Quantification for individual lines can be found in Figure S7D. For quantification, cells from 15 images were analyzed using CellProfiler identifying DAPI-positive and Ki67 positive cells (fraction of Ki67-positive cells = Ki67 stained cells/DAPI-stained cells). Significance was calculated using a Student’s t test.

(E) Proliferation of differentiating neurons (D20-D28) was assessed using an XTT assay. Absorbance was normalized to the absorbance of D20 (NPC stage and first day of measurement). Plotted is the mean (circle) together with a trend line (colored and dashed) and the standard error (gray shading). Proliferation differences between WT vs. HET and HET vs. KO were significant for all three tested timepoints (D23, D26, D28, p < 0.01) and between WT vs. HET for two timepoints (D23 and D26, p < 0.01). The experiment was carried out three times with three technical replicates each. Significance was calculated using a Student’s t test.

(F) Representative western blot of 3-week-old neurons (D42) stained against neuronal marker MAP2, FBXO11, and H3 as a loading control.

(G) Quantification of MAP2 levels from western blot in (F) showed reduced MAP2 levels (normalized to loading control H3) for HET and KO neurons compared with WT. The experiment was performed three times. Mean expression of three WT controls were set to 1. Significance was calculated using a one-sample t test with a theoretical mean of 1.

(H) Representative images of immunofluorescence of FBXO11 WT, HET, and KO 1-week old neurons (D28) stained with antibodies against neuronal markers MAP2 (red) and TUBB3 (green) are shown here and images of all nine neuronal lines can be found in Figure S4. Images were taken on an AxioImager Z2 with a 40× objective. Scale bar, 40 μm.

(I) Quantification of TUBB3-positive cells among neurons showed reduced levels of TUBB3-positive KO cells. Quantification for individual lines can be found in Figure S7F. For quantification cells from at least 15 images were analyzed using CellProfiler identifying DAPI-positive cells (all cells) and TUBB3-positive cells (fraction of TUBB3-positive cells = TUBB3-stained cells/DAPI-stained cells). Significance was calculated using a Student’s t test. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Processes involved in neuronal differentiation and proliferation were particularly enriched among the differentially expressed genes. We first checked whether proliferation differences were already present in NPCs by staining with antibodies against Ki-67 and phospho-histone H3 (pHH3) as markers for proliferation and mitosis, respectively. We found that FBXO11 HET and KO NPCs are more proliferative, and more cells are undergoing mitosis compared with WT NPCs (Figures 3C, 3D, S7B–S7D, and S8). We additionally performed an XTT assay on differentiating neurons, which measures cell number and viability, and which can therefore be used to measure proliferation. Here, we found that during the first week of differentiation from NPCs to neurons (D21-D28), WT cells gradually proliferated less. In contrast, FBXO11 HET or KO cells continued to proliferate at higher rates for the first days and only showed reduced proliferation after 8 days (Figure 3E). For FBOX11 KO cells, proliferation was even higher compared with HET neurons. At later stages, following replating of neurons during neuronal maturation (D29-D56), cells lost most of their proliferative capabilities, and low proliferation rates in the XTT assay were comparable among WT, HET and KO neurons (Figure S7E).

To investigate whether this altered proliferation in NPCs and at early stages of differentiation also affects neuronal differentiation, we measured expression levels of neuronal marker MAP2 of 3-week-old neurons (D42, 2 weeks of maturation). We found that MAP2 levels were highest in WT neurons, lower in HET neurons and even lower in FBXO11 KO neurons (Figures 3F and 3G). Additionally, we found that at D28 (1 week of differentiation from NPCs) the number of neurons positively stained for neuronal marker TUBB3 was significantly reduced in FBXO11 KO neurons compared with WT neurons. The number of TUBB3-positive cells was not altered in FBXO11 HET cells (Figures 3H, 3I, S4, and S7F). These data hint at possibly slower or reduced differentiation of FBXO11-deficient neurons compared with WT.

Loss of Fbxo11 leads to neuronal defects in D. melanogaster

To validate neuronal defects upon loss of FBXO11 in a second model system, we evaluated neuronal morphology and behavior in Drosophila. We first analyzed basic locomotor behavior using the climbing assay. Pan-neuronal knockdown of Fbxo11 led to a mild impairment of basic evoked locomotor behavior in the climbing assay for two of three different RNAi lines (Figure 4A). In contrast, spontaneous locomotor activity as measured in an activity assay was unaltered (Figure S6C). Flies also did not show any seizure susceptibility in the bang sensitivity assay (Figure S6D). We furthermore tested complex learning and memory using the courtship conditioning paradigm but did not observe any learning or memory deficits upon mushroom body-specific knockdown using RNAi 1 (Figure S6E). Synaptic formation at larval body wall NMJs was also not altered upon pan-neuronal knockdown of Fbxo11 (Figure S6F). To investigate any defects in dendrite formation, we analyzed the class IV dendritic arborization (da) neurons. These large sensory neurons in the larval body wall present an established model to study dendritic arborization.⁵⁴ We found that dendritic trees of da neurons were less complex upon specific Fbxo11 knockdown using the 477-Gal4 driver in da neurons. Both, the number of dendritic branches as well as the total length of dendrites was significantly reduced for all three tested RNAi lines (Figures 4B–4D). This highlights that also in Drosophila, Fbxo11 is indispensable for proper neuronal development and fly behavior.

Figure 4.

Deficiency of Fbxo11 leads to impaired behavior and dendritic branching in Drosophila melanogaster

(A) Climbing assay upon pan-neuronal knockdown of Fbxo11 showed impaired locomotor ability for two of the three RNAi lines tested. Individual data points are shown as circles and summarized data are shown as boxplots. At least 200 flies in batches of 10 (n = 20) were analyzed per condition. Significance was calculated using a Wilcoxon signed rank test.

(B) Representative image of traced da neurons from control and knockdown larvae upon da neuron-specific knockdown (477-Gal4; UAS-mCD8GFP driver line) is shown. Images were acquired using a Zeiss LSM 710 confocal microscope with a 20× objective. Scale bar, 100 μm.

(C) Quantification of total dendrite length of da neurons.

(D) Quantification of number of branches from da neurons. Tracings of da neurons were performed in ImageJ using the NeuronJ plugin. At least 10 da neurons from five different larvae from two independent crosses were analyzed for each line. Statistical significance was calculated using a Student’s t test.

FBXO11-deficiency-related phenotypes can be alleviated with proteasome activating substances

As FBXO11 is important for proper ubiquitination and degradation of target proteins, we wanted to test whether chemical activation of the proteasome may be able to rescue FBXO11/Fbxo11-deficiency-associated phenotypes. Several small molecules have been described to increase proteasomal activity.⁵⁵ From this study, we have chosen two promising substances, PD169316, a potent experimental substance, and Verapamil, a long-approved drug used to treat for example cardiac arrhythmias (Figure 5A). As the R-enantiomer of Verapamil was shown to be the proteasome-activating version of Verapamil, we included R-Verapamil as a third substance to test on FBXO11-deficient flies and human neurons (Figure 5A).

Figure 5.

Rescue of FBXO11-deficiency-associated phenotypes with proteasome-activating substances

(A) Formulas of tested substances PD169316, R-Verapamil, and Verapamil are shown.

(B) Scoring scheme for rescue experiments corresponding to the level of completeness of the rescue. For dark-filled boxes, rescue resulted in almost complete normalization to control levels under the DMSO (75%–100%). For boxes filled with light shades of respective color, rescue levels reached 50%–75%. Different tested substances were supplemented to the fly food or the cell culture medium and are color-coded as follows: black – DMSO control, green – PD169316, red – R-Verapamil, purple – Verapamil. For fly experiments, all substances were used at 1 μM, for cell-based experiments, different concentrations were used (PD169316: 20 μM, R-Verapamil: 15 μM, Verapamil: 10 μM).

(C) Climbing assay deficit upon pan-neuronal Fbxo11 knockdown with RNAi 1 could partially be rescued with proteasome-activating substances supplemented to the fly food. Improvement of phenotypes was seen when adding substances at time of egg laying (developmental supp.) or after flies hatched (adult supp.). At least 200 flies in batches of 10 (n = 20) were analyzed per condition.

(D) Total dendrite length of da neurons increased upon addition of proteasome-activating substances to the fly food at time of egg laying. At least five different neurons from five different larvae were analyzed per condition.

(E and F) NPCs (E) or D28 neurons (F) were treated with proteasome-activating substances for 2 days (NPCs) or 1 week (neurons) before staining with Ki-67 antibody to assess proliferation. For quantification, cells from at least 15 images were analyzed using CellProfiler identifying DAPI-positive and Ki67-positive cells (% of Ki67-positive cells = Ki67-stained cells/DAPI-stained cells). For all plots, individual data points are shown as circles and summarized data are shown as boxplots. Statistical significance was calculated using either a Wilcoxon signed rank test (climbing assay) or a Student’s t test (da neuron assays and cell-based experiments) with correction for multiple testing. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

We first tested the substances on an easily scorable morphological phenotype in Drosophila. Wing-specific knockdown of Fbxo11 with three different RNAi lines consistently led to smaller wing size with both decreased wing length and width (Figure S9 [black boxes]). We added PD169316, Verapamil, or R-Verapamil in different concentrations to the fly food, assessed wing size and determined the most efficient concentration to be 1 μM (Figures S9A–S9D), which we used for all assays in the fly. When supplementing the fly food with either substance from the time of egg laying, we found that wing length and width significantly increased for hatching flies of all three tested lines (Figures S9A, S9B, S9E, and S9F), indicating that Fbxo11-deficiency-associated wing phenotypes can partially be rescued. We next performed rescue experiments for neuronal phenotypes. For this, we tested one RNAi line corresponding to the two control lines each, where we have observed a phenotype in the climbing assay and the dendritic arborization neuron analysis (RNAi line 1 for climbing assay and RNAi lines 1 and 3 for dendritic arborization neurons). We found that the mild locomotor phenotype observed upon knockdown with RNAi line 1 in the climbing assay could be rescued by supplementation of PD169316, Verapamil, or R-Verapamil from the time of egg laying through to hatching of the flies (Figure 5C). Interestingly, a rescue was also possible when supplementing fly food with either substance only after flies had hatched (Figure 4C). We also tested reversibility of the dendritic arborization neuron phenotype upon knockdown with two of three RNAi lines (RNAi 1 and 3) and found that supplementation of any of the three substances also lead to more complex dendritic trees with longer total dendrite length and more branches (Figures 5D and S9G).

We then tested whether phenotypes in our human cell-based neuronal model are also amendable to proteasomal activation. We therefore examined whether proliferation and differentiation differences in NPCs and differentiating neurons could be alleviated with any of the three substances tested in flies. When examining proliferation in NPCs and in differentiating neurons, we found that increased proliferation in HET and KO NPCs and early differentiating neurons (D28) observed with increased Ki-67 staining could be reduced closer to wild-type levels by at least 50% with all three substances tested (Figures 5E and 5F). In contrast, the reduced/slowed differentiation in HET and KO neurons, as assessed by reduced number of TUBB3-positive differentiating neurons, could not be improved by proteasomal activation. There was no change in TUBB3 staining upon treatment with any of the three substances in FBXO11 HET or KO neurons (Figure S9H).

Our results indicate that impairment of proteasomal function seems to contribute to FBXO11-deficiency-related pathology and that increasing proteasomal activity can alleviate some of the associated phenotypes.

Discussion

In this study, we combined transcriptomic analysis with phenotyping and rescue experiments in both a FBXO11-deficient human neuronal cell-based system and in Fbxo11-deficient flies as an in vivo model to capture different aspects of neuronal development and dysfunction. This study was to our knowledge the first investigation of the function of FBXO11 in a neuronal context, improving the understanding of the pathomechanisms underlying FBXO11-associated NDD. Our experiments unveiled a conserved disruption of various processes important during neuronal development upon FBXO11 deficiency, including neuronal migration, proliferation, and differentiation. Due to their importance during development and the fact that disruptions of these processes have been implicated in various NDDs,^56,57,58 they are very likely to contribute to the pathomechanisms of FBXO11-related NDD.

Several clinical phenotypes observed in affected individuals with FBXO11-haploinsufficiency may be attributed to these abrogated processes. For instance, based on our neurosphere assay results, it may be expected that neuronal migration in brains of individuals harboring heterozygous FBXO11 LOF is still possible but reduced. Accordingly, affected individuals do not display severe neuronal migration disorders, but several of the reported affected individuals with FBXO11 variants display various magnetic resonance imaging abnormalities including defects of the corpus callosum,^21,23,24 a brain abnormality that is frequently found in individuals with variants in genes related to neuronal migration.⁵⁸ Additionally, around 30% of individuals with pathogenic FBXO11 variants present with seizures or electroencephalogram abnormalities, and many known genes involved in the etiology of epilepsies are also involved in neuronal migration.^59,60,61 We additionally found evidence for increased proliferation of differentiating NPCs early during development and for a possibly slowed differentiation of neurons in culture. This is in line with our results from transcriptomic analysis showing an enrichment of differentiation-specific genes among genes downregulated in FBXO11 KO neurons. We therefore provide evidence for an altered proliferation/differentiation balance in differentiating neurons. Such defects are also commonly found in cell-culture-based models for autism spectrum disorder-associated genes,⁵⁶ associated with or without differences in head circumference in affected individuals.^62,63,64 Especially rather subtle changes in the proliferation/differentiation balance in cell culture may result in disturbed neuronal networks without consistent effects on head circumference.⁵⁶ For FBXO11 defects, head circumference is rather variable among affected individuals.²⁴ While in the majority of cases, head circumference is in the normal range, around 20% of individuals are microcephalic and 5% are macrocephalic. Therefore, it currently remains unclear if and how those differences are linked to FBXO11 defects. The search for additional modifiers of such phenotypic variability could represent an interesting avenue for future studies. Additionally, a more in-depth analysis of synaptic and dendritic structures may provide further insights into the underlying disease mechanism.

As part of the SCF-ubiquitin ligase complex, FBXO11 is involved in substrate recognition and thereby in ubiquitination and subsequent degradation of target proteins. Potential neuronal substrate proteins of FBXO11 have barely been elucidated so far, but present valid candidates that could link FBXO11 to different developmental processes and furthermore link the observed neuronal phenotypes to proteasomal dysfunction.

The importance of proteasomal dysfunction resulting from FBXO11 deficiency was highlighted by the amelioration of associated phenotypes through application of small molecules PD169316, R-Verapamil, and the approved drug Verapamil, previously identified to act as proteasome activators through different indirect mechanisms.⁵⁵ Partial rescue of phenotypes was possible both in vitro in a human neuronal cell culture model and in vivo in Drosophila. Interestingly, the behavioral phenotype in Drosophila did not only improve, when treating flies with proteasome activators throughout development, but also when only treated for 2 days after hatching, though effects of longer treatment times in adult flies have yet to be determined. This indicates that Fbxo11-associated phenotypes may be reversible not only in early (embryonic) development, which may be of particular relevance for a potential treatment with proteasomal activators in affected individuals with FBXO11-associated NDD in the future. The therapeutic potential of proteasomal activation has been investigated before, for instance in neurodegenerative disorders,⁶⁵ where accumulation of (ubiquitinated) proteins is thought to be one of the hallmarks of disease.^66,67 Activation of the proteasome may help reduce proteotoxic load in those conditions.^68,69 Additionally, proteasomal activation has also been shown to delay aging⁷⁰ and result in an increased lifespan in different model organisms.^71,72 The effects of proteasomal activation are likely not evenly distributed across all cellular proteins but likely largely dependent on specific protein structure, organization, folding or aggregations.^65,73 Therefore, specific effects in pathological situations are plausible, even though the proteasome is activated in general.^65,74 At this point it is difficult to predict which proteins may be particularly sensitive to enhancement of proteasomal activity and how proteasomal activation may influence observed gene expression changes. In FBXO11-deficient neurons, we were only able to rescue the proliferation, but not the differentiation phenotype. Different FBXO11 substrates might be involved in those separate processes. As neuronal substrates of FBXO11 are not elucidated yet, it also remains unclear how and which substrates might be particularly affected upon FBXO11 deficiency, which could also partially explain that some phenotypes may be more susceptible to proteasomal activation.

Regarding potential adverse effects, we did not observe any toxic effects upon proteasomal activation with any of the three substances at the tested concentrations in either cells or flies. In line, no cytotoxic short-term effects have been observed in cell cultures upon proteasomal activation through various means, suggesting that this may be well tolerated on a cellular level.⁷⁵ However, on an organismal level, long-term studies are still lacking.⁷⁵ The rescue potential of Verapamil might be of particular interest, as it is a long-approved drug that is widely used for the treatment of high blood pressure, angina pectoris, and arrhythmias and also used as a preventive treatment for cluster headaches,^76,77 and may show potential to be repurposed for rare disorders. Verapamil functions as a calcium channel blocker, and is also a known P-glycoprotein inhibitor.⁷⁸ Only recently, it has been shown that the inhibitory function on channels is primarily carried out by the S-enantiomer,⁷⁹ while R-Verapamil is a potent activator of the proteasome.⁵⁵ Interestingly, in our study, rescue potential was comparable between the racemic mixture Verapamil and the R-enantiomer, suggesting that Verapamil itself may be suitable as a potential therapeutic option. We therefore provide two preclinical model systems, in which we have shown the potential to ameliorate and reverse phenotypes associated with FBXO11-deficiency.

In summary, we could show that FBXO11 is also required for proper neuronal development and its (heterozygous) loss results in impairment of various processes such as in neuronal migration, proliferation, and differentiation, contributing to the pathomechanisms underlying FBXO11-associated neurodevelopmental disorders. Treatment with Verapamil ameliorated associated phenotypes in preclinical fly and human cell-based models, thus providing first evidence for a potential targeted therapy for FBXO11-related NDD using repurposing of an approved drug.

Data and code availability

RNA-seq data are deposited in NCBI’s Gene Expression Omnibus database (GSE287638 (human) and GSE287637 (Drosophila)). All data reported in this paper will be shared by the lead contact upon request.

This paper does not report original code.

Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank James Terhune for technical assistance. We thank Felix Engel for help with the confocal microscope, which was supported by a grant from the German Research Foundation (DFG, INST 410/91-1 FUGG). Transcriptome sequencing was performed at the Core Unit NGS of the Faculty of Medicine of the Friedrich-Alexander Universität Erlangen-Nürnberg. Fly stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) and the Vienna Drosophila Research Center (VDRC, www.vdrc.at) were used in this study. This work was supported by an ELAN-Fonds grant from the interdisciplinary center for clinical studies (IZKF) of the Friedrich-Alexander-Universität Erlangen-Nürnberg to A.G. (18-08-09-1-Gregor) and from a Marie-Skłodowska-Curie Actions Fellowship from the European Commission to A.G. (837547). This work was also supported by the RTG2162 “Neurodevelopment and Vulnerability of the Central Nervous System” of the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation—270949263/GRK2162).

Author contributions

Conceptualization: A.G., methodology: A.G., S.T., and B.W.; formal analysis: A.G., P.K., and S.U.; investigation: A.G., L.D., A.B.E., and M.K.; resources: A.G., B.W., and C.Z.; writing – original draft: A.G.; writing – review and editing: L.D., A.B.E., P.K., S.U., M.K., S.T., B.W., and C.Z.; supervision: B.W. and C.Z.; funding acquisition: A.G. and C.Z.

Declaration of interests

The authors declare no competing interests.

Web resources

• CRISPOR: http://crispor.tefor.net/

• Panther Gene Enrichment: https://www.pantherdb.org/

• ShinyR-DAM tool: https://karolcichewicz.shinyapps.io/shinyr-dam/

• OMIM: https://www.omim.org/