Mutation in ATG5 reduces autophagy and leads to ataxia with developmental delay

1Macroautophagy, referred to hereafter as autophagy, is a cellular process by which proteins and organelles are degraded and recycled through sequestration within autophagosomes and delivery to lysosomes. The autophagy pathway is highly conserved and required for organismal development and function. Defects in autophagy are associated with diseases including cancer, metabolic disruption, and neurodegenerative disorder. Patients with mutations in any of the non-redundant core autophagy-related (ATG) genes have not previously been reported.

2Ataxia is a neurodegenerative disease caused by disruption of the cerebellum and Purkinje cells, which results in the lack of coordinated muscle movements. Large phenotype diversity is present in individuals with ataxia, including age of onset, rate of progression, and other accompanying neurological and non-neurological features, with corresponding genotypic heterogeneity. Even within the more defined phenotype of childhood ataxia with developmental delay, there are a large number of associated genes, such that similar phenotypic features alone are often insufficient information for an accurate diagnosis. Identification of genetic causes of childhood ataxia is important for understanding disease pathogenesis and for possible future treatment development.

3Whole exome sequencing has been successfully utilized to identify known and novel genetic mutations responsible for ataxia. Identification of candidate genes can be further verified through additional molecular analysis and utilization of specific and general animal models. Here we identified a novel mutation in a core autophagy gene, ATG5, in two children with ataxia, and demonstrate a reduction in autophagic response, also reproducing the phenotype in yeast and fly models.

4Two Turkish siblings presented with ataxia and developmental delay in childhood, as previously described. We performed linkage analysis on both affected siblings, their unaffected siblings, and their unaffected mother, using a model of remote parental consanguinity and identified a single broad (>14 Mbp) peak with LOD score 3.16 on chromosome 6q21, between 102 and 116 Mb (Figure 1). Whole exome sequencing identified a homozygous missense mutation, hg19 chr6:106,727,648 T>A, corresponding to E122D in ATG5 (Figure 2A) as the only damaging mutation within the genetically identified chromosomal linkage interval. The mutation was Sanger verified and found absent from variant databases and from Turkish controls.

5ATG5 plays a role in elongation of the phagophore and its subsequent maturation into the complete autophagosome. ATG12 is a ubiquitin-like protein that covalently binds ATG5, and this conjugate noncovalently binds ATG16L1. Crystal structure of the resulting ATG12–ATG5-ATG16L1 complex indicated that E122 is located in the vicinity of the ATG12–ATG5 interaction surface (Figure 2B); hence, we predicted that the mutation in ATG5 could affect the conjugation of ATG12. Comparison of protein isolated from control lymphoblastoid cell lines (LCL) and of affected subjects revealed a severe reduction of the ATG12–ATG5 conjugate in the mutant cells under basal conditions (Figure 3A), suggesting that the E122D mutation may have impaired autophagy by inhibiting conjugation between ATG12 and ATG5.

6The ATG12–ATG5-ATG16L1 complex functions in part as an E3 ligase to facilitate the conjugation of LC3 to phosphatidylethanolamine, generating LC3-II. Compared to control cells, LCLs from patients with the E122D mutation exposed to bafilomycin A1 showed a substantial reduction in LC3-II accumulation under basal conditions (Figure 3B), suggesting a possible decrease in E3 activity and subsequent attenuation of basal autophagic flux. The patient LCLs were also unable to upregulate their autophagic flux in response to Torin 1 (Figure 3C), which is a strong inducer of autophagy. ATG5E122D LCLs also showed elevated levels of SQSTM1/p62, an autophagy receptor and substrate, further indicating disruption of basal autophagy (Figure 3C).

7To examine the effect of the ATG5E122D mutation on formation of the ATG12–ATG5-ATG16L1 complex, we expressed the recombinant human proteins in insect Hi5 cells and analyzed the complexes by affinity isolation. We could detect the ATG12–ATG5 complex when both wild-type proteins were co-expressed, but we could only detect a minimal amount of the ATG12–ATG5E122D complex (Figure 4A). Although overexpression of human ATG5WT in HEK293 cells or Drosophila tissues resulted in efficient covalent conjugation with overexpressed human ATG12 (Figure 4B and C) or endogenous Drosophila Atg12 (Figure 4D), mutant ATG5E122D was dramatically impaired in this process (Figure 4B, C and D). Interestingly, expression levels of ATG5WT and ATG5E122D monomers were comparable to each other, indicating that the mutation affects the conjugation process, rather than the stability of proteins. This was consistent with the structural location of ATG5 E122 adjacent to the surface that interacts with ATG12 (Figure 2B). To confirm that the mutation does not overtly alter the structure of ATG5 or binding to ATG16L1, we analyzed formation of the noncovalent ATG5-ATG16L1 complex using constructs containing a TEV protease site. Both wild-type and mutant ATG5 protein were efficiently co-precipitated with ATG16L1 (Figure 4E). Indeed, the co-crystal structure of a human ATG5E122D-ATG16L1 complex (Figure 2 and Table 1) superimposes well with the previously determined structure of the WT proteins (Figure 2D), with the major obvious difference being replacement of the side-chain (Figure 2E and F). Thus, it appears that the E122D mutation interferes with the ATG12–ATG5 conjugation process, but not with ATG5 folding or binding of ATG16L1.

8ATG5 is a highly conserved protein, and sequence alignment demonstrated that E122 corresponds to yeast E141 (Figure 2A and C). We extended our analysis of the effect of the mutation on autophagy activity, by taking advantage of the yeast system. To test whether autophagy was affected by the Atg5 mutation in yeast, we initially relied on the GFP-Atg8 processing assay. During autophagy a population of Atg8 is continuously transported to the vacuole inside of autophagosomes. Tagging the N terminus of Atg8 with GFP makes it possible to monitor autophagy flux because Atg8 is rapidly degraded inside the vacuole whereas GFP is relatively resistant to vacuolar hydrolases; the generation of free GFP is an indication of autophagic activity. We observed a consistent decrease in autophagy activity with the Atg5E141D mutant relative to Atg5WT following autophagy induction by starvation (Figure 5A). Atg8, or GFP-Atg8, does not measure autophagic cargo per se, and the amount of GFP-Atg8 processing only corresponds to the inner surface of the autophagosome. To corroborate the effects observed through the GFP-Atg8 processing assay we examined autophagy using the quantitative Pho8∆60 assay. Pho8∆60 is an altered form of a phosphatase that is only delivered to the vacuole via autophagy; subsequent proteolytic processing generates an active form of the hydrolase. After 4 and 6 hr of starvation, yeast cells expressing the plasmid-based Atg5E141D mutant showed a significant decrease in autophagy levels compared to cells expressing Atg5WT (Figure 5B), and similar results were obtained when the WT and mutant ATG5 genes were integrated back into the chromosomal ATG5 locus (Figure 5C).

9To determine the reason for reduced autophagic activity we tested the effects of the Atg5E141D mutant on Atg8 lipidation. As shown by the ratio of Atg8–PE:total Atg8, cells expressing Atg5E141D displayed a decrease in Atg8–PE conjugation at 30 and 60 min of starvation compared to cells expressing Atg5WT (Figure 5D). We extended this analysis using the in vivo reconstitution of Atg8–PE conjugation as described previously. In brief, we examined Atg8 lipidation in a multiple-knockout (MKO) strain in which 23 ATG genes are deleted, when expressing only the E1, E2 and E3-like conjugation enzymes of the autophagy machinery. Atg8∆R that lacks the C-terminal arginine was used in the assay to bypass the initial activation step initiated by Atg4; due to the absence of Atg4, there is no cleavage of Atg8–PE from the membrane, resulting in stabilization of this form of the protein. We found that the Atg5E141D mutant was significantly defective in Atg8–PE conjugation compared to the cells with Atg5WT when Atg16 was not present (Figure 5E). Atg16 is not required mechanistically for Atg8 conjugation, but its presence increases the efficiency of this process and may dictate the site of conjugation. Thus, the presence of Atg16 may partially mask the Atg8 lipidation defects of the Atg5E141D mutant, and this may explain why the E122D/E141D mutation induces a hypomorphic rather than a complete null phenotype.

10To further characterize the effect of the E122D mutation on the development of ataxia, we generated Drosophila melanogaster knockouts for Atg5 (Figure 6A), and reconstituted the Atg5-null mutant flies with transgenes expressing wild-type (WT) or E122D human ATG5 (Figure 6B-D). Unlike mouse models, Atg5-null flies are viable, although they exhibit severe mobility defects after adult eclosion as demonstrated by a negative geotaxis assay (Figure 6E and I, and Video 1), similar to Atg7 null mutant flies. These mobility defects were substantially restored by expression of ATG5WT (Figure 6F and I, and Video 2), suggesting that the molecular function of ATG5 is conserved between human and Drosophila. However, Atg5-null mutant flies expressing ATG5E122D were still defective in mobility although slightly better than Atg5-null controls (Figure 6G–I, and Videos 3 and 4), demonstrating again that ATG5 activity is compromised but not eliminated by the E122D mutation. ATG5E122D was also inferior to ATG5WT in suppressing Ref(2)P (fly p62/SQSTM1) accumulation (Figure 6J and K) and cell death (Figure 6L and M) in the brain of Atg5-null mutant flies.

11In summary, we demonstrate that the homozygous E122D mutation of ATG5, a unique mutation found in two human subjects with ataxia, results in reduced conjugation to ATG12 and in an overall decrease in autophagy activity. The homologous mutation in yeast also interferes with autophagy, and the ataxia phenotype was replicated in a fly model. Based on these results we propose that this ATG5 mutation, and the consequent disruption in autophagy activity, is the cause of the ataxic phenotype and disturbance of the cerebellum in the affected siblings. This hypothesis is in agreement with previously characterized mouse models, in which neuron-specific knockout of Atg5 results in ataxia-like phenotypes. By contrast, mice with complete knockout of Atg5 die shortly after birth, demonstrating that autophagy is essential for mammalian survival. Our results indicate that E122D is a partial loss-of-function allele that impairs but does not completely abolish ATG5 activity. Although the overall structure of the E122D mutant ATG5 superimposes well with the wild-type protein, the mutation causes a striking decrease in the level of ATG12–ATG5 conjugate that is formed when the C terminus of ATG12 is covalently linked to Lys130 of ATG5. We speculate that the E122D mutation causes subtle changes in the conformational dynamics that propagate to Lys130, which is less than 10 Å away, resulting in less ATG12–ATG5, which in turn leads to reduced LC3/Atg8 conjugation.

12In neurons, which are among the cells most dependent on autophagy for tissue homeostasis, the residual function of the E122D allele is inadequate, resulting in predominantly neurological symptoms in the two patients. Since homozygous mutations with complete loss-of-function have not been reported, we predict that individuals carrying such mutations, similar to Atg5-null mice, might not be viable.

13Autophagy is quickly gaining importance for its roles in preventing neurodegeneration. WDR45 is a redundant, non-core autophagy gene, one of four mammalian homologs to Atg18, and mutations in WDR45 cause SENDA, static encephalopathy of childhood with neurodegeneration in adulthood. Autophagy appears to be critical in ataxia, whether mutant proteins evade autophagy processes or normal autophagy is disrupted. Several ataxias are attributed to intranuclear or cytoplasmic aggregation of mutant proteins within the cell. These protein aggregates, in humans and in mouse models, not only evade autophagic sequestration but may even inhibit autophagy, or lead to reduction in autophagy available for other proteins due to saturation. Assessment of autophagy in patient cells may be used to refine and identify the genetic cause of a patient’s ataxia.

14Further discovery of the role of autophagy in neurodegenerative diseases should be used to investigate therapies targeted at the autophagy process. Many drugs enhance autophagy and their effects on a multitude of neurodegenerative diseases have been studied. Recently more studies have been conducted assessing the value of autophagy enhancers in ataxia models and patients. Induction of autophagy through administration of Temsirolimus, a rapamycin ester, and lentiviral overexpression of BECN1 in SCA3 model mice increase autophagy and the clearance of mutant protein aggregates, and reduce the ataxic phenotype. In a single patient trial, trehalose treatment of patient fibroblasts increased autophagy and alleviated cellular pathogenic features by improving mitochondrial morphology and reducing free radicals in the cell. Trehalose also showed success in trials involving models of SCA17. Lithium, another inducer of autophagy, improved symptoms in a SCA1 mouse model, but did not slow or reduce symptoms in a treatment trial in SCA2 patients. An autophagy enhancer may be an appropriate treatment to test in the presented subjects, as autophagic flux is attenuated, but not completely abrogated, by the ATG5E122D/E122D mutation.

15This study’s finding of the pathogenic human E122D mutation in ATG5, a gene encoding part of the autophagy-controlling core machinery, is important and novel, but consistent with reports of neurodegenerative disorders in other autophagy-related genes. Our results suggest that other mutations in this and other ATG genes, which impair but do not completely abolish autophagy, may result in similar forms of ataxia, intellectual disability and developmental delay. This study exemplifies the utility of exome sequencing in the identification of rare disease-causing variants, and supports the role of impaired autophagy in neurodegenerative disease. In addition, we demonstrate the utility of a combined genetic, biochemical and cell biological analysis in multiple model systems to elucidate the underlying pathogenic mechanism of rare human diseases.

16Study protocols including written informed consents have been approved by the University of Michigan Institutional Review Board and the Boğaziçi University Institutional Review Board for Research with Human Participants. Two Turkish brothers, ages 5 and 7 in 2004, presented with ataxia and developmental delay, as previously described. Parents were initially reported to be unrelated, but recently suggested they might be remotely related. Both patients were delayed in walking, had truncal ataxia and dysmetria, nystagmus, and lower IQ (68 and 70). MRI revealed cerebellar hypoplasia. Follow-up examinations showed no progression of symptoms.

17DNA was isolated from peripheral whole blood using the Qiagen (Germantown, MD) Gentra Puregene isolation kit. Linkage analysis was performed using the genotype data generated with Illumina HumanOmniExpress-24 chip for the mother and the four sibs. The Allegro module of easyLINKAGE software was used, assuming autosomal recessive inheritance and parents as third cousins. No deletion or duplications common to just the two affected brothers were detected using cnvPartition plug-in in Illumina Genome Studio v.1.02 software.

18Exome sequencing was performed independently twice on one subject. Capture for whole exome sequencing was performed with NimbleGen SeqCap EZ Exome Library v1.0 kit (Roche, Indianapolis, IN). Captured regions were sequenced with Illumina HiSeq2000 instruments. Variants were filtered to remove common variants based on 1000 Genomes, Exome Sequencing Project, and Exome Aggregation Consortium databases, variants outside of identified linkage regions, variants not expected to change protein coding, and variants not following a recessive model of inheritance.

19PCR followed by Sanger sequencing was performed to validate the variant identified through exome sequencing and test for segregation within the family. The variant of interest was further examined in two separate collections of a total of 500 Turkish samples, and found absent.

20Lymphoblastoid cell lines (LCL) of both subjects were generated from heparinized whole blood samples and cultured as described. As they are made in house and cultured briefly, mycoplasma contamination risk is minimized.

21We used a baculovirus/insect cell expression system to examine formation of the human ATG12–ATG5 conjugate in a heterologous system described previously. Hi5 insect cells (Invitrogen, Carlsbad, CA) were infected with baculoviruses expressing human ATG7, ATG10, a GST-tagged version of ATG12 (residues 53–140, corresponding to the ubiquitin-like domain) and a His-tagged version of either ATG5WT or ATG5E122D. Three days post infection, lysates were subjected to glutathione affinity chromatography, and the GST-ATG12–His-ATG5 conjugate was detected by SDS-PAGE followed by Coomassie Blue staining. To confirm that the baculoviruses produce protein, Hi5 cells were coinfected with baculoviruses expressing the His-tagged WT and mutant versions of ATG5 and the N-terminal domain of ATG16L1 (residues 1–69, as an MBP fusion). At three days post infection, lysates were subjected to nickel affinity purification. The ATG5-ATG16L1 complex formation was detected by SDS-PAGE and Coomassie Blue staining.

22The complex containing ATG5E122D and the N-terminal domain of ATG16L1 (residues 1–69) was expressed in Hi5 insect cells, and purified by nickel affinity, ion exchange, and size exclusion chromatography into a final buffer of 20 mM Tris, pH 8.5, 50 mM NaCl, 10 mM DTT. The complex was concentrated to 18.5 mg/ml, aliquoted, flash-frozen and stored at -80°C until further use. Crystals were grown by the hanging drop vapor diffusion method by mixing purified protein 1:1 with reservoir solutions of 37.5 mM MES, pH 5.2–5.8, 0.2 M sodium tartrate, and 11–13% polyethylene glycol 3350. Final crystals were obtained by micro-seeding with reservoir solution of 40 mM MES, pH 5.5, 0.2 M sodium tartrate, 8.5% PEG3350, 10 mM DTT. Crystals were cryoprotected in reservoir solution supplemented with 25% xylitol, and flash frozen in liquid nitrogen prior to data collection. Diffraction data were processed with XDS. The structure was determined by molecular replacement using Phaser with the structure of the WT ATG5-ATG16L1 (1–69) (PDB: 4TQ0) complex as a search model. Model construction and rebuilding were performed using Coot. The structure was refined using Phenix. Diffraction data and refinement statistics are provided in Table 1.

23Cells or tissues were lysed in cell lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 2.5 mM sodium pyrophosphate, 1 mM beta-glycerophosphate, 1 mM Na3VO4, 1% Triton X-100) or RIPA buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% sodium deoxycholate, 1% NP-40, 0.1% SDS) containing protease inhibitor cocktail (Roche). After being clarified with centrifugation, lysates were boiled in SDS sample buffer, separated by SDS-PAGE, transferred to polyvinylidene difluoride membranes and probed with the indicated antibodies. ATG5 (12994), LC3 (3868) and SQSTM1/p62 (5114) antibodies were purchased from Cell Signaling Technology. Hemagglutinin (HA, 3F10) antibody was from Roche. Actin (JLA20) and tubulin (T5168) antibodies were from Developmental Studies Hybridoma Bank and Sigma, respectively. Ref(2)P antibody was previously described.

24Wild-type human ATG5-coding sequence was from Addgene #24922 (deposited by Dr. Toren Finkel). The E122D mutation was introduced into ATG5 by PCR-based site-directed mutagenesis. ATG5WT and ATG5E122D were cloned into the plasmid pLU-CMV-Flag. The HA-ATG12-expressing plasmid was from Addgene #22950 (deposited by Dr. Noboru Mizushima). HEK293 cells (the 293 A substrain from Invitrogen, tested negative for mycoplasma by PCR) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen) containing 10% fetal bovine serum (FBS) and penicillin/streptomycin at 37°C in 5% CO2. For transient expression of proteins, HEK293 cells were transfected with purified plasmid constructs and polyethylenimine (PEI, Sigma) as previously described. Cells were harvested 24 hr after transfection for immunoblot analyses.

25Saccharomyces cerevisiae strain WLY176 was used to generate an ATG5 knockout strain (atg5∆) as previously described. The MKO strain YCY137 (SEY6210 atg1∆, 2∆, 4∆, 5∆, 6∆, 7∆, 8∆, 9∆, 10∆, 11∆, 12∆, 13∆, 14∆, 16∆, 17∆, 18∆, 19∆, 20∆, 21∆, 23∆, 24∆, 27∆, 29∆), was used for in vivo reconstitution of Atg8 conjugation. Site-directed mutagenesis was performed to generate ATG5 amplicons with the E141D mutation as previously described. A pRS406 empty plasmid was digested with Spel and SalI, and then ligated with a DNA fragment encoding either wild-type or mutant Atg5-PA. atg5∆ was transformed with an empty pRS406 vector, or plasmids encoding Atg5-PA WT or Atg5-PA E141D. Wild-type WLY176 colonies were transformed with empty pRS406 vector as a control. Colonies were grown on SMD-URA medium and starved in nitrogen-deficient medium. Pho8Δ60 and western blot analyses were performed as described previously. Quantification was performed using ImageJ software. The pATG8∆R-ATG7-ATG10(414), pATG5(WT)-HA-ATG12(416) and pATG5(WT)-HA-ATG12-ATG16(416) plasmids were described previously. The pATG5(E141D)-HA-ATG12(416) and pATG5(E141D)-HA-ATG12-ATG16(416) plasmids were made by site-directed mutagenesis based on the wild-type constructs.

26Atg5-null Drosophila flies (Atg55cc5) were generated by CRISPR-Cas9-mediated genome editing, using a double gRNA approach, both targeting the same gene, as described. The Atg55cc5 mutant was recovered by screening viable candidate lines for accumulation of the specific autophagy cargo Ref(2)P using western blots, followed by PCR and sequencing. Atg55cc5 mutants have a deletion in X:7,322,242–7,323,717 residues (Drosophila melanogaster R6.06), which deletes five out of six exons of the Atg5 gene, eliminating more than 85% of protein-coding sequences including the translation start site (Figure 6A). The PhiC31 integrase-mediated site-specific transformation method was used to express human ATG5WT and ATG5E122D from an identical genomic locus. In brief, flag-tagged ATG5WT and ATG5E122D were cloned into a pUAST-attB vector and fully sequenced. pUAST-attB-ATG5WT and pUAST-attB-ATG5E122D were microinjected into y1 M{vas-int.Dm}ZH-2A w*; M{3xP3-RFP.attP}ZH-51D flies and stable transformants were isolated by the presence of the mini-white+ marker (Figure 6B). The UAS-ATG5WT or UAS-ATG5E122D transgenes were crossed with a double balancer strain (Bl/CyO; TM2/TM6B) and then with +/CyO; Tub-Gal4/TM2 to be constructed as stable Tub>ATG5 lines (UAS-ATG5/UAS-ATG5; Tub-Gal4/TM6B). The Tub>ATG5 male flies were crossed with Atg55cc5/FM7 female flies to generate Atg5-null flies expressing human ATG5 transgenes. Climbing assays and TUNEL staining were performed as previously described.