Abstract
FZR1, which encodes the Cdh1 subunit of the Anaphase Promoting Complex, plays an important role in neurodevelopment by regulating cell cycle and by its multiple post-mitotic functions in neurons. In this study, evaluation of 250 unrelated patients with developmental epileptic encephalopathies (DEE) and a connection on GeneMatcher led to the identification of three de novo missense variants in FZR1. Two variants led to the same amino acid change. All individuals had a DEE with childhood onset generalized epilepsy, intellectual disability, mild ataxia and normal head circumference. Two individuals were diagnosed with the DEE subtype Myoclonic Atonic Epilepsy (MAE). We provide gene burden testing using two independent statistical tests to support FZR1 association with DEE. Further, we provide functional evidence that the missense variants are loss-of-function (LOF) alleles using Drosophila neurodevelopment assays. Using three fly mutant alleles of the Drosophila homolog fzr and overexpression studies, we show that patient variants do not support proper neurodevelopment. With the recent report of a patient with neonatal-onset DEE with microcephaly who also carries a de novo FZR1 missense variant, our study consolidates the relationship between FZR1 and DEE, and expands the associated phenotype. We conclude that heterozygous LOF of FZR1 leads to DEE associated with a spectrum of neonatal to childhood onset seizure types, developmental delay and mild ataxia. Microcephaly can be present but is not an essential feature of FZR1-encephalopathy. In summary, our approach of targeted sequencing using novel gene candidates and functional testing in Drosophila will help solve undiagnosed MAE/DEE cases.
Introduction
Developmental and Epileptic Encephalopathies (DEEs) are a heterogeneous group of disabling disorders, characterized by a combination of severe epilepsy and neurodevelopmental problems1. Seizures in DEE patients usually have an early age-of-onset and are often refractory to anti-epileptic drugs. The majority of DEE cases are thought to have a genetic basis, usually in the form of rare de novo variants with a highly disruptive effect on gene expression or protein function. Pathogenic variants in more than a hundred genes have been associated with a DEE phenotype so far2-4. Identifying the causal variant in a DEE patient can bring an end to an often long and stressful diagnostic odyssey and may lead to improved or targeted treatment for a subset of the cases2; 5; 6.
We set out to identify novel variants in DEE patients by screening a large cohort of patients with Myoclonic Atonic Epilepsy (MAE), a sub-type of DEE, as part of the efforts of the EuroEPINOMICS-RES consortium7. MAE is characterized by generalized seizure types including myoclonic, atonic, myoclonic atonic, absence and tonic-clonic seizures. The seizure onset is typically between 7 months and 6 years8. Using a combination approach of trio whole-exome sequencing (WES) and subsequent screening with a targeted gene panel including candidate genes for MAE, we identified two individuals diagnosed with MAE who each carried a unique de novo missense variant in FZR1. A third individual with a DEE with childhood onset generalized epilepsy and a de novo missense variant was identified through GeneMatcher.
FZR1 (fizzy and cell division cycle 20 related 1, MIM:603619) encodes Cdh1, one of the two regulatory subunits of the Anaphase Promoting Complex (APC) that confers the substrate-specificity on this E3 ubiquitin ligase complex9. APC was initially identified to play a key role in mitotic cell cycle progression10. When Cdh1 associates with APC (Cdh1-APC), it controls mitotic exit leading to G1 arrest11. Cdh1/Fzr1 knockout in mice leads to embryonic lethality, suggesting that this gene plays critical roles in development in vivo in mammals12. These Fzr1 knockout mice, which die as embryos, have impaired cortical neurogenesis due to delay of mitotic exit, leading to reduced cortical size and thickness13. Cdh1-APC also has a prominent function in post-mitotic cells in the nervous system14. It controls neuronal survival15; 16, axonal growth17, and synapse formation and function18. Similar to its mammalian counterpart, the Drosophila homolog of FZR1, fzr (fizzy-related), has been studied in the context neurodevelopment in addition to its role in cell-cycle regulation as part of the APC17; 19-21. fzr is necessary for photoreceptor patterning and regulates glial cell migration from the brain into the eye imaginal disc17. Through a forward genetic screen, we previously isolated two fzr alleles which displayed abnormal retina pattern and defects in electroretinograms (ERG)22. These phenotypes have been reported in hypomorphic mutants of fzr which were first identified through their rough eyes, leading to one of the gene’s synonyms ‘retina aberrant in pattern (rap)’23. These previously generated fzr alleles provide genetic tools to test the functionality of the human FZR1 variants identified in DEE patients.
During the course of our study, a different rare variant in FZR1 was reported in an individual with microcephaly and DEE24. This study showed that the missense variant found in their patient led to decreased stability of the FZR1 protein in patient leukocytes and in HEK293T cells as well as accumulation of proteins targeted by Cdh1-APC24. The mutant protein was also unable to rescue the aberrant cell cycle distribution of primary cortical progenitors from Cdh1/Fzr1 knockout mice. Based on these data, the authors argued that this variant potentially contributes to microcephaly in the patient.
Here, we present three individuals with DEE and a novel de novo missense variant in FZR1. These cases consolidate the role of FZR1 in DEE and expand the phenotypic spectrum of FZR1-related encephalopathy to DEE with childhood onset generalized epilepsy, mild ataxia and normal head circumference. To determine the functional consequence of the missense variants, we assessed the ability of the variant proteins to rescue the eye phenotype of the fly fzr mutant alleles isolated from a forward genetic screen using mosaic analysis. Furthermore, we also generated a new loss-of-function allele of fzr using the CRIMIC technology25 to study the impact of the patients’ variants during development. Through these assays in Drosophila, we provide functional evidence for a LOF mechanism of the pathogenic variants found in our patients, further supporting the involvement of FZR1 in MAE.
Subjects and methods
Identification of patients with de novo FZR1 variants
Patient 1 was recruited to the project on MAE of the EuroEPINOMICS-RES consortium7. WES was performed on an initial cohort of 39 parent-offspring trios (including Patient 1) with MAE at the Wellcome Trust Sanger Institute (Hinxton, Cambridgeshire). WES and subsequent data analyses were performed as previously described26. The GenomeComb program was used for annotation and filtering of the data27. Coding variants that lead to a missense change, stop gain or stop loss, frameshift or essential splicing change were retained for further filtering. Variants that were present in the ExAC28 and Genome Aggregation Database (gnomad, v2.1.1)29 with a frequency > 0.01 or > 0 for homozygous and X-chromosome or heterozygous variants respectively were excluded. We also excluded variants in genes that are not expressed in the brain in the Genotype-Tissue Expression project database (GTEx V8)30. Pathogenicity of missense variants was predicted using the Combined Annotation Dependent Depletion scores (CADD, v1.6)31 and missense variants with a CADD-score < 20 were excluded. All remaining candidate variants following a recessive, de novo or X-linked inheritance were validated using Sanger sequencing.
Patient 2 was part of a follow-up cohort consisting of 211 DEE patients, including 89 probands with MAE, and 122 probands with Dravet or Dravet-like syndrome. The cohort was screened using a targeted gene panel using Molecular Inversion Probes and consisted of the coding regions of 12 known and 40 candidate genes (at time of screening) for MAE, including FZR1 (see supplemental Table 1). Candidate genes included in this panel were selected from a list of genes with single de novo hits identified in the EuroEPINOMICS-RES WES data. Data analysis and variant filtering was performed as previously described32. Segregation analysis was performed using Sanger sequencing for all nonsynonymous, frameshift, and splice-site variants that were not present in the ExAC set of ∼61,000 WES28. Paternity and maternity were confirmed using the Powerplex®16S system (Promega).
Patient 3 was identified through GeneMatcher33. Trio-WES was performed in a clinical diagnostic setting using DNA from the patient and both healthy parents. Trio-exome data were filtered for non-synonymous de novo variants absent in the general population (gnomAD), and for rare biallelic variants with minor allele frequency (MAF) <0.1% and absence of homozygous carriers in the aforementioned database. The functional impact of the identified missense variants was predicted using the CADD, v1.631 SIFT34 and Polyphen35 pathogenicity prediction tools.
The Ethical Committee of the University of Antwerp, Belgium gave ethical approval for the study. Parents of each patient signed an informed consent form for participation in the study.
Gene burden analysis
A gene burden analysis was performed using denovolyzeR36 for the two de novo missense variants identified in the combined research cohort of 250 patients examined in our study. In parallel, we determined the number of rare, potential damaging variants in FZR1 in the control population set of gnomAD28 based on CADD31, SIFT34 and Polyphen35 scores using the Variant Effect Predictor tool from Ensembl37. We filtered missense variants and loss-of-function variants that had a general population frequency of less than or equal to 0.01, had a CADD score greater than or equal to 20, and were predicted to be Pathogenic/Likely pathogenic by SIFT and Polyphen. We performed a chi-square test to evaluate the allele-count burden as well as number of unique variant burden in the DEE patient cohorts compared to the gnomAD (control) population.
Evaluation of relative positions of affected residues in the 3D structural model of human Cdh1
To examine the localization of the residues predicted to be altered by the patients’ missense variants, we used the 3D structural model of Cdh1 generated from an electron microscopy reconstruction of the Homo sapiens APC (PDB ID:4ui9)9. We used PyMOL (Molecular Graphics System, Version 2.0 Schrödinger, LLC). to map the location of the residues affected. We also predicted the location of the Drosophila FzrA variant based on the homology of human FZR1 and Drosophila Fzr proteins (Figure S2A).
Functional assessment of DEE-associated missense variants in Drosophila melanogaster
The molecular lesions of two EMS-induced variants identified from our previous forward genetic screen22, fzrA and fzrB, were identified using Sanger sequencing of PCR amplicons of genomic DNA38. ERG analysis on mosaic adult eye were performed as described earlier39. We used CRISPR/Cas9 technology to insert a T2A-GAL4 gene trap cassette based on CRISPR mediated Integration Cassette (CRIMIC) technology into the intron of fzr at ChrX:4,676,058 (Drosophila melanogaster reference genome, Assembly 6) and generated the fzrCR00643-TG4.2 allele (referred in the text as fzrT2A-GAL4). Expression of GAL4 from the fzrT2A-GAL4 allele was examined using a UAS-nls::RFP reporter [Bloomington Drosophila Stock Center (BDSC) stock number: 8546). Complementation tests were performed using the fzrie28 allele19 (a kind gift from Dr. Christian Klambt) and Dp(1;3)DC472 P[acman] clone inserted in the third chromosome at the VK33 docking site40-42. To perform rescue experiments of fzr mutants, we subcloned the full length fzr cDNA from a pUASt-fzr vector19. Drosophila fzr cDNA carrying the variants at positions corresponding to the human variants were generated by site-directed mutagenesis using QuickChange II kit (Agilent). As patient 2 and patient 3 carry a variant affecting the same nucleotide and leading to the same protein change, only the C>G variant of patient 2 was modeled. Wild-type (wt) and mutant fzr expressing UAS constructs were integrated into VK33 docking site on the fly third chromosome via phiC31 transgenics40. Transgenic lines were recovered and established based on standard crossing schemes43. Rescue experiments for eye phenotypes of fzrB mutants were performed using the MARCM (Mosaic analysis with a repressible cell marker) technique44. MARCM-based rescue was achieved by crossing flies of the following genotype: y w fzrB, FRT19A/FM7c, Kr-GAL4, UAS-GFP (BDSC 52385) and tub-GAL80, ey-FLP, FRT19A/Y; Act-GAL4, UAS-GFP/CyO (FRT19A tester line). The FRT19A tester line was generated by crossing the following stocks tub-Gal80, ey-FLP, FRT19A (BDSC 42717) and FRT19A; Act5C-GAL4, UAS-GFP/CyO (BDSC 42726). Rescue of the lethality of the fzrT2A-GAL4 allele was assessed based on the T2A-GAL4 strategy25; 45, and achieved by setting the following cross: fzrT2A-GAL4 /FM7, Act-GFP crossed to UAS-fzr (wt or variant) and evaluation of non-GFP male larvae at late third instar stage. Overexpression studies in the developing eye imaginal disc were carried out using ey3.5-GAL4 (y1 w1118 ; P{ey3.5-GAL4.Exel}2) (BDSC: 8220)17.
Electroretinogram analysis of adult mosaic eyes
y w fzrB, FRT19A/FM7c, Kr-GAL4, UAS-GFP and y w fzrA, FRT19A/FM7c, Kr-GAL4, UAS-GFP (BDSC 52384) females were crossed with y w FRT19A; ey-FLP (BDSC 5579). The resulting females carried mosaic eyes with mutants identifiable with the lack of red pigmentation. ERG analysis on mosaics were performed as described earlier22; 39.
Immunofluorescence analysis of Drosophila tissue
Immunofluorescence analysis was performed as described earlier on embryos and third instar larval eye imaginal discs17; 19. The following antibodies from Developmental Studies Hybridoma Bank (University of Iowa) or commercial sources were used: mouse anti-Repo (1:50, 8D12)46, rat anti-ELAV (1:50, 7E8A10)47, rabbit anti-HRP (1:50, Jackson ImmunoResearch, #323-005-021), chicken anti-GFP (1:1000, Abcam ab13970), rabbit anti-mCherry (1:500, Abcam ab167453), Alexa488 conjugated goat anti-Chicken (1:200, ThermoFisher A-11039), Alexa647 conjugated goat anti-mouse (1:250, Abcam ab150115), Alexa647 conjugated goat anti-rat (1:250, Abcam ab150159), Alexa594 conjugated goat anti-rabbit (1:1000, ThermoFisher A-11012), Alexa405 conjugated goat anti-rabbit (1:100, ThermoFisher A-31556). Samples were imaged using a LSM 500 confocal microscope (Zeiss) to generate a Z-stack of different focal planes. The confocal Z-stack images were used to generate maximum intensity Z-projection using ImageJ48.
Mammalian overexpression vectors, cell culture, Immunofluorescence, and western blot
Human FZR1 cDNA (NM_001136198.1) in the donor vector pDONR221 (HsCD00042756;) was purchased from Harvard cDNA clone repository. The coding region was cloned into the pEzy-eGFP (Addgene #18671) destination vector in frame with an N-terminal eGFP tag using Gateway™ LR Clonase™ II Enzyme mix (ThermoFisher 11791020) following the manufacturer’s recommended protocol. The variants observed in the DEE patients were introduced into the pEzy-eGFP-FZR1 using Quickchange II site directed mutagenesis kit (Agilent).
Human embryonic kidney cells (HEK293) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM; Thermofisher) supplemented with 10% fetal bovine serum. Cells were transfected with overexpression vectors using Lipofectamine 3000 reagent (ThermoFisher) with OptiMEM media (Thermofisher) using the manufacturer’s recommended protocol. Cells were fixed using 4% Paraformaldehyde in PBS at room temperature. Immunofluorescence staining was performed using chicken anti-GFP antibody (1:1000, Abcam ab13970) as described earlier49. Nucleus and actin-cytoskeleton were visualized using DAPI (4’,6-diamidino-2-phenylindole, ThermoFisher D1306) and Alexa647 conjugated Phalloidin (Cell signaling, #8940) respectively. Cells were imaged using Zeiss LSM 500 confocal microscope. A 2-micron optical section with the nucleus in focus were obtained. Protein isolation and western blot analysis was performed 36 hours post transfection as described earlier49. Chicken anti-GFP (1:1000, Abcam ab13970) and IRDye® 800CW donkey anti-chicken Secondary Antibody (LiCOR P/N: 926-32218) were used for detecting N-terminal GFP tagged FZR1 protein.
Primers used in the study
Results
Clinical case reports
A summary of the clinical history of the three patients described above, and the patient described by Rodriguez et al.24 can be found in Table 1. In compliance with the privacy rules of MedRxiv, the details of patient’s symptoms, genetic diagnosis and treatment information are available to readers upon a reasonable request to the corresponding author.
WES on patient 1 and unaffected parents yielded a de novo missense variant in FZR1 [c.559G>A, p.(D187N), Table 1] located at the same residue as the previously reported de novo missense variant in a patient with microcephaly, psychomotor retardation, and epilepsy [p.(D187G)]24. No other variants passed the filtering steps as described in the methods section. A relative with mild learning disabilities but no seizures did not carry the variant. The variant is not present in the gnomAD database and has a CADD score of 29.731, SIFT-score of 0.01,34 and PolyPhen score of 0.93635, all indicating a deleterious or probably damaging effect. According to the gnomAD database29, FZR1 has a pLI score of 1, o/e score of 0.04, and a z-score of 3.64 for missense variants, showing that this gene is intolerant to both LOF (nonsense, frameshift, core splicing) and missense variants.
Following the identification of a de novo FZR1 missense variant in Patient 1, a follow-up panel-screening resulted in the identification of another de novo FZR1 missense variant [c.999C>G, p.(N333K), Table 1] in Patient 2. This variant had a CADD score of 23.6, SIFT-score of 0.01 and PolyPhen score of 0.814, all predicting a deleterious or possibly damaging effect. This variant was also absent from the gnomAD database. Previous genetic tests on this patient included karyotype analysis and SCN1A and PCDH19 screening, all of which were negative. Array-based comparative genomic hybridization (array-CGH) showed two duplications on Chr11p11.2. These copy number variations were inherited from the healthy mother and did not include disease-associated genes. Therefore, they deemed non-contributing to the patient phenotype.
Patient 3 was identified through Genematcher, and carries a de novo missense variant at the same nucleotide position as patient 2, leading to an identical amino acid change [c.999C>A, p.(N333K), table 1], with identical CADD, SIFT and PolyPhen scores. Previous chromosome analysis and array-CGH were normal. Filtering of exome data retained homozygous variants in two additional genes, not yet reported in the context of rare Mendelian disorders. The missense variant c.2105_2106delinsTC, p.(S702F) in PTPN21 (NM_007039.4) had a SIFT-score of 0.000 and PolyPhen score of 0.996, predicting a probably damaging effect. This multinucleotide variant is present twice heterozygous in gnomAD. PTPN21 (Tyrosine-protein phosphatase non-receptor type 21, MIM: 603271) is an oncogenic protein known to be upregulated in several types of cancer cells50 and functions as a key regulator of inflammation51. Two non-synonymous single-nucleotide polymorphisms in PTPN21 showed association to schizophrenia in a GWAS study52. In neurons, PTPN21 controls the activity of KIF1C, a fast organelle transporter implicated in the transport of dense core vesicles and the delivery of integrins to cell adhesions53. It was further shown to positively influence cortical neuronal survival and to enhance neuritic length54. The second homozygous variant identified in patient 3 was the truncating variant c.577C>T, p.(R193*) in the last exon of TPD52L2 (NM_199360.3). TPD52L2 (Tumor protein D52-like 2, MIM: 603747) is an ubiquitously expressed tumor protein shown to be involved in multiple membrane trafficking pathways, and to affect cell proliferation, adhesion and invasion55,56. As the de novo FZR1 missense variant resulted in the same amino-acid substitution as identified in patient 2 who has a very similar phenotype, we concluded that the FZR1 variant is most likely to underlie the neurological disease of patient 3. We can however not totally exclude the possibility that the variants in PTPN21 and TPD52L2 contribute to the phenotype of the patient 3.
Having found these cases, we tested the association between rare variants in FZR1 and DEE using two different gene burden analysis tools. First, we used the tool denovolyzeR36 which evaluates the overrepresentation of de novo variants that affect the protein sequence (missense and loss of function variants) in a specific gene. We found that identification of two de novo variants in FZR1 amongst the 250 patients in our research cohort is a 212-fold enrichment (Poisson distribution P-value of 4.41e-05) over the expected number of de novo variants, suggesting a strong association of FZR1 de novo missense variants with the disorder. Second, we compared the rate of rare, damaging variants in our research cohort (250 patients) with that of general population in gnomAD (141,456 samples)29 in FZR1. For this, we re-analyzed both the WES dataset and the targeted panel sequencing cohort of our patient population for the presence of damaging, rare variants regardless of inheritance, and did not identify any additional qualifying variants. Once again, we found a significant enrichment in the number of damaging variants in FZR1 within our patient cohort (2/250 patients) compared to that of gnomAD (total allele count = 97, unique variants = 57; Chi Square test P-value for unique variants = 1.613e-05; Chi Square test P-value for total allele count = 0.001574). These two burden analyses show that, even though only two patients with FZR1 variants were identified amongst 250 DEE patients within our research cohort, there are strong statistical data to support an association between rare FZR1-missense variants and DEE.
DEE-associated FZR1 variants lead to reduced57 protein abundance
FZR1 and its homologs in other species are highly conserved with Drosophila Fzr protein showing over 70.3% identity with human FZR1. The amino acids that are impacted by the missense variants that we identified in the patients are conserved in Drosophila Fzr (Figure 1A, S2A). To test the effect of the patient variants on protein stability and localization, we expressed FZR1 with an N-terminal eGFP tag in HEK293 cells. FZR1 expression was detected in the nucleus, with a weaker diffuse signal in the cytoplasm (Figure 1B). Variant FZR1 proteins showed a similar pattern of localization. However, western blot of total protein from cells transfected with wt or mutant eGFP-FZR1 showed a clear difference in FZR1 abundance, with both patient variants leading to a reduction in protein level of approximately 40% (p<0.01) (Figure 1C-D).
To further examine the effect of patient variants on human FZR1 function, we mapped the affected residues in a 3D protein structure of the APC9. Variants found in all three patients affect residues within the WD40 domain of the FZR1 protein (Figure 1E, S2B). WD40 domains, also known as beta transducing repeats, enable protein-protein interactions58, and de novo variants in another WD40 repeat containing protein encoding gene WDR37 has also been recently associated with epilepsy59. The active site of a WD40 domain is often found in the central cleft of the propeller where the loops connect the successive beta sheets. As both variants are present in those loops (Figure 1D), they may interfere with the protein-protein interaction sites and potentially affect the substrate binding capacity of FZR1 in addition to reducing overall protein stability.
Genetic characterization of fzr alleles in Drosophila
We decided to examine the functional effect of the DEE-associated variants that we identified using Drosophila as a model organism. For this study we used two EMS (ethylmethanesulfonate)-induced alleles of fzr (fzrA and fzrB) that we previously isolated. These alleles produce rough eye phenotypes in adult mosaic animals and show defects in electroretinograms22, suggesting that fzr is involved in the development and function of the fly visual system. We determined the molecular lesions in these lines using Sanger sequencing of PCR fragments of genomic DNA. The molecular lesions in the fzrA allele is a canonical splice-site mutation (fzr-RA:c.592+1G>A) and the fzrB allele is a missense mutation (fzr-PA:p.G316N, Figure 2A). In the protein structure model, this fly mutation is found close to the p.N333K residue affected in Patient 2 and 3 (Figure 1F).
We created an additional fly mutant by using CRISPR/Cas9 and homology directed repair to introduce an artificial exon containing a splice acceptor-T2A-GAL4-polyA cassette between the first and second exons of Drosophila fzr. This allele is predicted to generate a strong LOF allele that also produces a GAL4 in the same spatial and temporal pattern reflecting the endogenous fzr expression pattern25. This CRIMIC insertion (fzrCR00643-TG4.2, henceforth called fzrT2A-GAL4) leads to precocious termination of transcription of all fzr isoforms because of the presence of a polyA termination signal as well as an interruption of translation of fzr mRNA due to the viral T2A peptide sequence. T2A also allows re-initiation of translation to express the GAL4 when and where endogenous fzr is expressed (Figure 2A). The fzrT2A-GAL4 allele permits expression of transgenes including fzr cDNA in the native pattern of fzr in the fzr-mutant background by crossing this line to a UAS-fzr transgenic fly (Figure 2A)25, simplifying the functional studies of the variants of interest.
We found that the fzrT2A-GAL4 allele is recessive lethal and fails to complement the lethality of a previously reported null allele of fzr (fzrie28), suggesting that fzrT2A-GAL4 is a strong LOF allele (Figure 2B). Complementation test between the fzrT2A-GAL4 and the EMS-induced mutations (fzrA and fzrB ) reaffirmed that these mutations are allelic and are also LOF mutations (Figure 2B). Finally, we examined the rescue of the recessive lethality of fzrT2A-GAL4 and the EMS alleles using a duplication containing the fzr locus inserted on the third chromosome [Dp(1;3)DC472, Dup]. Each allele was individually rescued by this duplication and the rescued animals did not exhibit any morphological defects, confirming that the phenotypes observed in these alleles are due to LOF of fzr (Figure S3C). To examine the expression pattern of fzr in vivo, we crossed the fzrT2A-GAL4 allele to a UAS-nls::RFP transgenic line and explored the fluorescent pattern. We found that fzr is broadly expressed in larval brains and eye imaginal discs, consistent with their developmental roles in these tissues (Figure S3D-E).
The role of fzr in neurodevelopment in embryos and larva
To further characterize the phenotypes of the fzr alleles we generated, we first examined the effect of the Drosophila EMS mutants in developing larval eye imaginal disc and adult retina. We tracked the stereotypical pattern of larval photoreceptor precursors in homozygous mutant cells induced through the expression of eyeless (ey) enhancer driven Flippase57 (ey-FLP) that is generated using the MARCM (Mosaic Analysis with a Repressible Cell Marker) technique60. In this experiment, Flippase mediates the recombination of two FRT (Flippase Recognition Target) sites located at the base of the X-chromosome (FRT19A). When this occurs during mitosis and one of the two sister chromatid carries the frz mutant allele, homozygous mutant and homozygous wild-type cells are generated from heterozygous cells (Figure 3A). Simultaneously, it reverses the repression of GAL4 by GAL80 in homozygous mutant cells, allowing the GAL4 to drive the expression of a GFP reporter and simultaneously an optional UAS-fzr cDNA transgene (wt or variant, Figure 3B). Consequently, mutant cells are marked by the GFP reporter and if included, transgenes are expressed within these mutant cells to rescue the phenotype caused by fzr LOF.
In the larval eye disc, the developing photoreceptors were marked by staining for ELAV (Embryonic lethal abnormal vision), a pan-neuronal nuclear marker61. In the homozygous mutant cells identified using GFP reporter, we found that the larval photoreceptor patterns marked by ELAV were severely affected (Figure 3C). This phenotype in the fzrB allele is consistent with previous reports of fzr function in the eye17. In adults the pattern of the compound eye was also significantly affected in animals in which mutant mosaic clones were generated (Figure 3D).
To further assess the function of photoreceptors, we performed ERG recordings on 3-4-day old flies. Adult flies with fzr mosaic eyes showed a significant reduction in the depolarization amplitude of the ERG response to light (Figure S2A-B), indicating a defect in phototransduction. These flies also showed severe loss of the ON/OFF transients (Figure S2A-B), suggesting a defect in synaptic transmission.
Next, we examined the embryonic neurogenesis phenotype of the fzrT2A-GAL4 mutants by assessing the morphology of the nervous system through immunofluorescence staining and confocal microscopy using an antibody that recognizes neuronal/photoreceptor membranes (anti-HRP antibody). Hemizygous fzrT2A-GAL4 mutant males showed a clear defect in neuronal patterning in the central nervous system of fly embryos, displaying defective neuromere patterns in the ventral nerve chord (Figure S2C). This is consistent with the role of fzr in neuronal development and phenotypes reported for previously identified fzr alleles19.
DEE-associated variants fail to support Drosophila neurodevelopment in rescue experiments in vivo
fzrB mosaic flies showed severe morphological defects in larval eye imaginal discs as evidenced from aberrant anti-HRP staining pattern in the mutant mosaic clones (Figure 3E). We examined the ability of the fzr cDNA with wt sequence or cDNAs with variants corresponding to those of the human patients (Figure 1A) to rescue this phenotype MARCM)62. We observed that overexpression of the wt fly fzr cDNA in the mutant mosaic clones rescued the HRP pattern defects in the eye discs (Figure 3E-c), but the variant cDNAs failed to display a similar rescue (Figure 3E-d,e).
fzr mutants in Drosophila have also been reported to exhibit defects in glial cell migration17. Since glia are known to play important roles in epilepsy63, we examined the impact of the two patient variants on this phenotype, again using the MARCM system in the eye imaginal disc. In a control animal (mosaic animals without fzr mutant tissue, Figure 3E-a), glial cells only migrate to areas where photoreceptors have initiated their differentiation program indicated by HRP-positive cells (arrowhead, Figure 3E-a). However, in mosaic cells that are defective for fzr, we observed that glial cells migrate beyond the differentiated zone and prematurely enter areas where immature photoreceptors are present (HRP negative region, Figure 3E-b). We observed that wt fzr cDNA was able to suppress this phenotype (Figure 3E-c) but the fzr cDNAs carrying the patient’s variants were not able to rescue the abnormal migration of glial cells (Figure 3E-d,e). This suggests that neuron-glia communication mediated by fzr are also affected by the patient variants.
Since we noticed that overexpression of fzr wild-type cDNA in the eye clones occasionally led to aberrant development patterns, we examined if overexpression of fzr cDNA in wild-type background can disrupt the normal photoreceptor development pattern. Indeed, when wt fzr cDNA was overexpressed in the developing eye using ey-GAL4, we observed a severe reduction in the size of adult retina and loss of photoreceptor pattern (Figure 4B), similar to what has been reported earlier 21. In this assay, we observed that fzr cDNA with patient variants also showed caused aberrant photoreceptor patterns (Figure 4C-D). However, the reduction in the size of the adult retina due to the fzr variant overexpression was not as severe as the fzr wt cDNA and the differences between the overexpression of wt and variants were statistically significant (Figure 4E). These results further support that FZR1 variants associated with DEE are loss of function alleles.
We next examined if wild-type UAS-fzr cDNA can rescue the lethality observed in fzrT2A-GAL4 flies. We used GAL4 expressed from the T2A-GAL4 allele to drive the expression of the UAS-fzr. Expression of wild-type fzr cDNA was able to allow a significant fraction (∼20-23%) of fzr-T2A- GAL4 hemizygous males to live to third instar larval stages whereas they otherwise die as embryo or first instar larvae (Figure 5A). In contrast, the overexpression of fzr cDNA carrying either of the two patient variants fails to rescue this early lethality (Figure 5A). Finally, we examined whether the patient fzr cDNA can rescue the central nervous system development defects in the fzrT2A-GAL4 mutant embryos. The wild-type fzr cDNA was able to restore this defect (Figure 5D), but the two variant cDNAs did not rescue the lesions (Figure 5E-F). Together, the MARCM-based rescue experiments in the developing eye, overexpression analysis in the developing eye, and rescue studies of embryonic neuronal developmental defects all suggest the variants found in three DEE patient act as strong hypomorphic alleles, impacting neuronal development in multiple contexts.
Discussion
In this study, we describe three individuals with DEE with childhood onset generalized epilepsy carrying a de novo missense variant in FZR1. One missense variant affects the same amino-acid residue as identified in a single previously published patient with DEE24. We further provide statistical and functional support for pathogenicity of these variants.
All four individuals carrying pathogenic de novo FZR1 variants described so far suffer from neurodevelopmental delay and epilepsy. While the previously reported individual had neonatal onset treatment-resistant multifocal seizures, severe ID and prenatal microcephaly24, our study shows that the phenotypical spectrum also includes childhood onset generalized seizure syndromes associated with moderate to severe ID, mild ataxia and normal head circumference. This variability could be due to strength of alleles or genetic backgrounds, which will require further investigation. Of note, two patients in our study were diagnosed with MAE. A normal development prior to seizure onset is historically considered a diagnostic criterium for the diagnosis of MAE, but the phenotypic boundaries of this syndrome remain debated64. In a recent study on the genetic etiology of MAE, more than 20% of patients did have developmental delay prior to seizure onset8. This feature is indeed inherent to the concept of DEE, which acknowledges that the neurodevelopmental impairment of these patients is not solely related to frequent epileptic activity but is also a direct result of the underlying gene dysfunction. Of note, the <5 years-old patient 3 in our study did not have (myoclonic) atonic seizures at time of inclusion in this study. As early disease history is very similar to patient 2, carrying a variant leading to the same amino-acid substitution, further clinical evolution will tell whether other seizure types will still occur. Both patients also had similar signs of delayed myelination on early brain MRI.
There is a significant enrichment of both de novo and of (predicted) deleterious variants in our patient cohort of DEE patients compared to a control population such as gnomAD. Interestingly, all four DEE associated variants described so far affect one of two different residues of FZR1. FZR1 variants found in DEE patients affect residues that are likely to be important for the substrate recognition of Cdh1-APC9. Variants in this region may lead to altered substrate recognition and therefore lead to a diminished function of Cdh1-APC. Furthermore, we have shown that our patient variants lead to lower protein levels, which could indicate a reduction in the stability of mutant FZR1.
Using our Drosophila functional assays, we show that both p.D187N and p.N333K variants lead to functional deficits in Fzr (Cdh1) protein, especially in the developing nervous system. The DEE-associated variants fail to rescue photoreceptor pattern and glial cell migration that are phenotypes observed in previous fzr alleles in Drosophila and the patient variants behave as partial loss of function alleles. This is also observed in an overexpression assay in the developing eye, as we find that the variants display retention of some function compared to the reference allele. The difference in the functionality between the variants and the wt Fzr is dramatic in the embryonic CNS development as observed using the fzrT2A-GAL4 allele, which might provide a more sensitive readout of functional differences. As a regulatory subunit of the E3 ubiquitin ligase complex APC, FZR1 (Cdh1) is involved in the turnover of many substrates65. One substrate of Cdh1-APC is FMRP (Fragile X mental Retardation Protein) encoded by the FMR1 (MIM:309550) gene66. The interaction of FMRP with Cdh1-APC was shown to regulate metabotropic glutamate receptor (mGluR)-dependent synaptic plasticity66 and the formation of stress granules and protein synthesis-dependent synaptic plasticity67. Triplet repeat expansion in FMR1 causes Fragile X syndrome, a neurodevelopmental disorder often accompanied by seizures68. HECW2 (HECT, C2 And WW Domain Containing E3 Ubiquitin Protein Ligase 2, also known as NEDL2 MIM:617245) is another substrate of Cdh1-APC69 with a link to neurodevelopmental disorders, as de novo missense variants lead to ID, seizures and absent language70. Cdh1-APC-mediated degradation of HECW2 during mitotic exit is important for the regulation of metaphase to anaphase transition69. Abnormality of these or other substrates of Cdh1-APC may underlie the DEE and other phenotypes seen in our patients, which will require further molecular studies.
In summary, our work provides genetic, statistical and functional support for the role of FZR1 in DEE, and we expand the phenotypic spectrum of FZR1-related encephalopathy to include individuals with DEE with childhood onset generalized epilepsy and normal head circumference. The presence of early neurodevelopmental delay, even prior to seizure onset, is in line with the deleterious impact of patient variants on early neurodevelopment in a Drosophila model. The three molecularly defined fzr alleles (fzrA, fzrB, fzrT2A-GAL4) are of great value to the study of the role of fzr in Drosophila, and due to the extensive conservation of residues, the study of FZR1 in human diseases. The fly mutant, overexpression lines, and assays that we have developed in this study will be an important asset for future studies on the role of FZR1 in neurodevelopment and the impact of human disease variants on its interaction with substrates such as FMRP and HECW2 in the development of DEE.
Data Availability
All relevant data and supporting information are contained within the manuscript and supporting data files.
https://gnomad.broadinstitute.org/
Supplemental Information
Supplemental data include three figures and one table.
Declarations of Interests
The authors declare no competing interests.
Web resources
gnomAD https://gnomad.broadinstitute.org/
denovolyzeR https://denovolyzer.org/
Variant Effect Predictor https://www.ensembl.org/info/docs/tools/vep/index.html
OMIM http://www.omim.org/.
Data and Code Availability
All WES data generated in the EuroEPINOMICS-RES project was deposited in the European Genome-phenome Archive, accession numbers EGAS00001000190, EGAS00001000386, and EGAS00001000048.
Acknowledgements
N.S. is supported by the UA-Bijzonder Onderzoeksfonds (BOF)-DOCPRO4 (FFB180186).
S.Y. is supported by the following grants from the National Institutes of Health (U54NS093793, R01DC014932, R24OD022005) and through funds provided by the Nancy Chang Ph.D., Award for Research Excellence, Baylor College of Medicine, and the Jan and Dan Duncan Neurological Research Institute at Texas Children’s Hospital. S.W. is supported by the Fonds Wetenschappelijk onderzoek (FWO 1861419N). H.M. is supported by NIH (R01 NS069605). We would like to thank Dr. Michael Wangler for guidance on the bioinformatic analysis of the variants and comments and discussion on the manuscript.