Missense variants in the middle domain of DNM1L in cases of infantile encephalopathy alter peroxisomes and mitochondria when assayed in Drosophila

ABSTRACT
Defects in organelle dynamics underlie a number of human degenerative disorders, and whole exome sequencing (WES) is a powerful tool for studying genetic changes that affect the cellular machinery. WES may uncover variants of unknown significance (VUS) that require functional validation. Previously, a pathogenic de novo variant in the middle domain of DNM1L (p.A395D) was identified in a single patient with a lethal defect of mitochondrial and peroxisomal fission. We identified two additional patients with infantile encephalopathy and partially overlapping clinical features, each with a novel VUS in the middle domain of DNM1L (p.G350R and p.E379K). To evaluate pathogenicity, we generated transgenic Drosophila expressing wild-type or variant DNM1L. We find that human wild-type DNM1L rescues the lethality as well as specific phenotypes associated with the loss of Drp1 in Drosophila. The p.A395D variant as well as the novel variant p.G350R neither rescue lethality nor other phenotypes. Moreover, over-expression of p.A395D and p.G350R in Drosophila neurons, salivary gland and muscle strikingly altered peroxisomal and mitochondrial morphology. In contrast, the other novel variant (p.E379K) rescued lethality and did not affect organelle morphology, although it was associated with a subtle mitochondrial trafficking defect in an in vivo assay. Interestingly, the patient with the p.E379K variant also has a de novo VUS in pyruvate dehydrogenase 1 (PDHA1) affecting the same amino acid (G150) as another case of PDHA1 deficiency suggesting the PDHA1 variant may be pathogenic. In summary, detailed clinical evaluation and WES with functional studies in Drosophila can distinguish different functional consequences of newly described DNM1L alleles.

INTRODUCTION
Mitochondrial diseases are a clinically and genetically heterogeneous group of disorders characterized by defects in mitochondrial function. Diagnosis of mitochondrial disease can be difficult because of phenotypic variability including lactic acidosis, epilepsy, muscle weakness, deafness, optic atrophy and encephalopathy, each of which may be variably present.(1-5) These disorders may be inherited in a mitochondrial, X- linked, recessive or dominant manner, or arise de novo. Because over 100 genes in both mitochondrial and nuclear DNA have been associated with mitochondrial disease, the identification of the responsible gene can be challenging.(6, 7) In addition, the interplay between mitochondria and other organelles makes analysis of cases of possible mitochondrial diseases more challenging. For example, a number of mitochondrial phenotypes can be observed in patients with peroxisomal disorders such as peroxisomal biogenesis disorders (PBD).(8)Given the phenotypic and genetic heterogeneity of mitochondrial and peroxisomal diseases, whole exome sequencing (WES) has emerged as a diagnostic modality for the diagnosis of mitochondrial disorders.(7, 9) WES is a powerful tool for the diagnosis of genetic disease.(10-14) However, WES may uncover novel sequence alterations and variants of unknown significance (VUS)(15) which cannot be categorized as benign or pathogenic due to lack of functional evidence. Even with careful phenotyping, identification of the responsible gene can be even more challenging in patients with two or more potential causative genes. In one series, approximately 5% of individuals referred for WES had two molecular diagnoses.(13) In a series of 53 patients with clinically- diagnosed mitochondrial disease, several patients had two or more VUS that could not be categorized based on lack of functional data.(7)

One method to determine whether a variant is pathogenic is through functional validation in model organisms(16) such as Drosophila.(17) Drosophila offer the advantage of diversity and availability of many reagents for genetic manipulation, short generation time compared to mammalian models, and conservation of many human genes.(18) Drosophila functional studies of human VUS can shed light on the functional significance of a single VUS or more than one VUS giving rise to a blended phenotype.(19)DNM1L (Dynamin 1-like, synonyms Drp1) is a member of the dynamin superfamily of GTPases and mediates mitochondrial and peroxisomal fission.(20-22) One patient has been previously described with a lethal encephalopathy due to defective mitochondrial and peroxisomal fission (MIM #614388). This individual had poor feeding, poor growth, lactic acidosis, seizures, hypotonia, nystagmus, and an abnormal gyral pattern on MRI, and passed away at 37 days of life.(23) Analysis of very long chain fatty acids revealed elevated cerotic acid, suggesting peroxisomal dysfunction. The patient’s fibroblasts exhibited a decreased number of peroxisomes and dysmorphic mitochondria. Sequencing of DNM1L revealed a c.1184C>A, p. A395D variant.Mice which lack Dlp1, the homologue of DNM1L, die at embryonic day 11.5, indicating a crucial role for this gene in mammalian development and these mice display abnormal mitochondrial and peroxisomal morphology.(24) Drosophila Drp1 mutants were identified in a screen for synaptic transmission mutants.(25) Fly Drp1 mutants display altered cellular distribution of mitochondria in the nervous system leading to a near absence of mitochondria from synapses and exhibit defects in mitochondrial morphology and synaptic transmission.(26, 27) Given the evidence across species, DNM1L is clearly a candidate gene for encephalopathy, but thus far, only one case has been reported.(23) Here, we report two additional heterozygous missense variants in DNM1L and we use Drosophila to understand the function of these alleles.

RESULTS
We clinically identified two patients with lactic acidosis, poor feeding, poor growth, developmental delay, and hypotonia. We initially identified Patient 1, a 14 month-old male with global developmental delay (GDD), hypotonia and status epilepticus. He was born at term and had normal development until 5 months of age, when he developed seizures and developmental regression. MRI of the brain revealed a progressive volume loss and demyelination (Figure 1). At 13 months, MRI showed cerebral volume loss and thinning of the corpus callosum (Figure 1B versus control in Figure 1A). By 2 years of age, he had evidence of T2 hyperintense regions in the cortex as well as progressive volume loss (Figure 1D versus control in Figure 1C). Though serum lactate levels were initially normal, they were elevated at 4 years of life. A muscle biopsy was performed and respiratory chain enzyme activities were nominally reduced for mitochondrial complexes I, III and IV but did not meet modified Walker criteria (Figure 1E).(28) Electron microscopy of the muscle revealed mitochondrial pleomorphism as well as some lipid accumulation in the muscle (Supplemental Figure 1A-C). Plasma very long chain fatty acid levels were normal (Supplemental Figure 1D- E). The patient passed away at age 5 due to severe status epilepticus with respiratory failure. WES revealed a VUS in the DNM1L gene: c.1048G>A, p.G350R. His father did not exhibit this change and variant analysis of Sanger reads from the maternal blood suggested a low-level (6-8%) maternal mosaicism (Figure 1F). In addition this patient had a variant in ALG13 inherited from the mother, but this variant seemed unlikely to be pathogenic given the phenotype and an N-glycan analysis was done and was normal. A full list of candidates is shown in Supplemental Table 1.

Subsequently, Patient 2 was identified at 4 days of life for a persistent lactic acidosis. She was born at 37 3/7 weeks and pregnancy was complicated by intrauterine growth restriction and hydrocephalus. At delivery, she had poor tone and apnea. She was intubated and found to have metabolic acidosis with an elevated lactate. She was placed on thiamine and a ketogenic diet. MRI of the brain showed microcephaly, absence of the corpus callosum and diffuse volume loss with enlarged ventricles (Figure 1G-J). At 3 months of age, a ventriculoperitoneal shunt was placed for hydrocephalus. She had diffuse hypotonia, global developmental delay, and poor growth. Her parents declined muscle biopsy to evaluate electron transport chain activity. There was persistent elevation of lactate in whole blood (Figure 1K). The patient passed away at 10 months of age due to pneumonia. WES revealed two de novo changes in mitochondria-related genes, namely a VUS in the PDHA1 gene (c.448G>A, p.G150R), known to be associated with pyruvate dehydrogenase E1 deficiency as well as a VUS in the DNM1L gene (c.1135G>A, p.E379K) (Figure 1L). This raised the possibility that one or both of these variants were contributing to the patient’s phenotype. A full list of candidates is shown in Supplemental Table 1.The clinical characteristics of the previously-reported patient with the p.A395D variant were compared to the patients reported in this study (Table 1) (23). All three cases share the features of hypotonia, poor feeding, developmental delay, and shortened life span. However, while in the previous case, cerotic acid was elevated, Patient 1 did not exhibit elevations in VLCFAs, whereas Patient 2 did not have a plasma VLCFA analysis. Both the previous case and Patient 2 showed congenital lactic acidosis, while Patient 1 did not have lactic acidosis until 4 years of life. Given the variability of the phenotypes, it was not clear whether the phenotypes seen in Patient 1 and Patient 2 were due to the pathogenic variants in DNM1L.

DNM1L encodes a GTPase with an N-terminal GTPase domain, a C-terminal GED domain and a middle (M) domain (Figure 2A). The middle domain of DNM1L has previously been shown to be important for the tetramerization of DNM1L protein, as missense variations of conserved residues including p.A395 and p.G350, lead to elongated mitochondria in HeLa cells(21). Interestingly, the p.G350D variant was selected for structure-function studies prior to the identification of our patient due to conservation of the amino acid at that position.(21) However, Patient 1 had a different missense substitution affecting this amino acid (p.G350R). The glycine at this position is highly conserved and in yeast this residue (G385) is required for the formation of mitochondrial fission complexes, and self assembly is defective in G385D point mutants (29, 30).All three variants in the DNM1L gene are located in the middle domain of the protein in a highly conserved region, although the E379K is not a conserved amino acid (Figure 2A). Missense substitutions in this region have been shown to exhibit dominant- negative effects in vitro due to the middle domain’s role in oligomerization(21). This region is also of interest because it shares homology with dynamins (Supplemental Figure 2A). Though it shares significant homology with other dynamin genes in the genome (DNM1, DNM2 and DNM3), DNM1L is distinct because of its functional role in organelle fission. Variants within the middle domain of dynamin proteins can result in very different phenotypes. For example, the DNM2 gene is associated with both centronuclear myopathy (MIM 160150) and Charcot-Marie Tooth disease type 2M (MIM 606482). Interestingly, missense variants in the middle domain of DNM2 are thought to be associated with centronuclear myopathy rather than CMT. However, we recently reported two cases of CMT rather than centronuclear myopathy with middle domain variants(19) (Supplemental Figure 2A), one of which had been observed previously(31). In addition, the allele frequencies of heterozygous missense variants in the ExAC and EVS databases suggest the possibility of selection against missense variants in the DNM1L middle domain (Supplemental Figure 2B). Importantly, neither the p.G350R nor the p.E379K variant was observed in ExAC. Based on the observations that missense variants affecting the DNM1L middle domain are rare, and that pathogenic variants in very similar proteins, namely dynamins, underlie a spectrum of neurologic disease, we hypothesized that these variants were pathogenic in our patients.

We therefore undertook a functional study of DNM1L in Drosophila melanogaster. Drosophila Drp1 mutants are lethal with defects in mitochondrial trafficking to synapses, mitochondrial morphology and synaptic transmission.(27) The Drp11 and Drp12 alleles are Ethyl-Methane Sulfonate (EMS) induced point mutations, which are lethal. Transheterozygous Drp11 /Drp12 are larval lethal with mitochondrial trafficking defects.(27, 32) We generated transgenic flies carrying the human DNM1L gene with and without the three human variants. Because Drosophila Drp1 is the closest homolog of DNM1L, we first crossed the transgenes into Drp1 backgrounds to determine if the human reference sequence DNM1L(Ref) construct was able to rescue a Drp1 fly mutant. By expressing human DNM1L(Ref) ubiquitously with Da-Gal4, we rescued the lethality of Drp1 (Drp11 /Drp12) mutants (Figure 2B). However, the DNM1L(A395D) from the previously reported case as well as DNM1L(G350R) observed in WES from Patient 1 did not rescue lethality (Figure 2B). In addition, both of these alleles exhibited some toxicity on a sensitized (Drp11/+) and in a wild-type Drosophila background (Supplemental Figure 2C). In contrast, the DNM1L(E379K) variant was able to rescue lethality (Figure 2B) and did not exhibit toxicity with over-expression (Supplemental Figure 2C).

The DNM1L protein is part of the machinery that allows both mitochondria and peroxisomes to undergo fission. We therefore examined peroxisomal morphology in third instar larval salivary glands by over-expressing the DNM1L constructs in the presence of a peroxisomal GFP-SKL marker (ActinGal4>UAS-GFP-SKL) and anti-Pex3 antibody(33) (Figure 3). Given studies showing p.A395D overexpression can recapitulate peroxisomal phenotypes resulting from DNM1L loss,(21) we determined if the VUS in our cases had similar effects. Overexpression of DNM1L(Ref) has no effect on peroxisomal morphology, as the salivary gland peroxisomes are approximately 0.3 µm2 (Figure 3 A-A’’, 3E). In contrast, expression of the DNM1L(A395D) and DNM1L(G350R) both led to dramatic increase in peroxisomal size and altered cellular distribution (Figure 3B-C’’, 3E). However, the DNM1L(E379K) had no effect on peroxisomal size (Figure 3D-D’’, 3E). Increased peroxisomal size with DNM1L(A395D) and DNM1L(G350R) was associated with a decreased number of total peroxisomes per cell (Figure 3F). The results were similar on a Drp1/+ (sensitized) background (data not shown). Therefore, the p.G350R variant in Patient 1 exhibits a strong dominant negative effect on peroxisomal morphology similar to p.A395D. However, the p.E379K variant in Patient 2 did not cause any obvious peroxisomal effects.Next, we examined mitochondria in muscle of third-instar larvae in a sensitized genetic background (Drp11/+) by driving expression with MEF2-Gal4. Again, we observed a remarkable alteration in morphology of muscle mitochondria with DNM1L(A395D) and DNM1L(G350R), but not DNM1L(E379K) compared to DNM1L(Ref) (Figure 4A-E). The mitochondrial distribution in muscle was also altered. Mitochondria are normally seen intercalating between muscle fibers (Figure 4F-F’).

In contrast, there was a paucity of mitochondria between sarcomeres in muscle and reduced mitochondrial numbers and size in both the Drp11/+;DNM1L(A395D) and Drp11/+;DNM1L(G350R) larvae, but not the Drp11/+; DNM1L(E379K) larvae when compared to Drp11/+;DNM1L(Ref) (Figure 4F-M). We also observed strong effects on mitochondrial morphology with DNM1L(A395D) and DNM1L(G350R) when overexpressed in a wild-type background, suggesting this is a dominant-negative effect (data not shown).Drosophila Drp1 mutants exhibit altered distribution of mitochondria in the nervous system.(27, 32) We examined this phenotype for the human VUS on a sensitized background (Drp11/+). The results were similar to overexpression on a wild-type background (data not shown). We again noted altered mitochondrial trafficking in the ventral nerve cord, axons and synaptic boutons of Drp11/+;DNM1L(A395D) and Drp11/+;DNM1L(G350R) larvae (Figure 5A-C). In addition, while Drp11/+;DNM1L(E379K) larvae appeared to have normal mitochondrial trafficking in the VNC and in the axon, we observed a clear trafficking defect at the level of the bouton in the Drp11/+;DNM1L(E379K) larvae which was statistically significant and consistent with that seen with the other two variants (Figure 5D,F). Therefore, the trafficking experiments suggest that while the DNM1L(E379K) variant does not exhibit a strong dominant-negative effects on organelle morphology in our assay the DNM1L(E379K) larvae do exhibit trafficking defects.

DISCUSSION
Here we report two cases with encephalopathy and missense mutations in DNM1L, the gene underlying the lethal encephalopathy phenotype (MIM 614388) noted in one previous case. Because of the known role for DNM1L in peroxisomal and mitochondrial fission and the Drp1 mutant phenotypes in Drosophila we studied the effect of these variants on organelle fission and trafficking. Our Drosophila studies suggest a dominant negative effect on peroxisomal and mitochondrial morphology for the p.A395D allele reported by Waterham(23) and the p.G350R variant observed in Patient 1. The E379K allele observed in Patient 2 was able to rescue lethality of Drosophila Drp1 mutants and did not cause obvious organelle morphology defects. This finding is relevant to the clinical phenotype of Patient 2 as she had two de novo variants, one in DNM1L and one in PDHA1. She exhibited a severe cortical atrophy, dilated ventricles and an incomplete corpus callosum, similar to those seen in other cases of female PDHA1 deficiency.(34) Moreover the PDHA1 missense allele affects the same amino acid as an allele from that study, glycine at position 150 (34). The patient in that report exhibited an overall similar brain phenotype to Patient 2 but with later onset of lactic acidemia, a lower lactate level and less severe brain abnormalities(34). The difference in severity and phenotype could be explained by differences between the amino acid change (G150R versus G150E), differences in the inherited genetic background of Patient 2, or differences related to the additional de novo event in Patient 2 in DNM1L. The p.E379K allele did not produce the strong-dominant negative effects of the other alleles but it did exhibit an abnormal phenotype in an assay for the trafficking of mitochondria to synaptic boutons. In any case, the fly studies allow us to distinguish this range of possibilities all suggesting a minimal effect of p.E379K, from the strong dominant-negative effects seen in the other alleles.

Another interesting feature of the Drosophila functional analysis relates to the similarity between the p.A395D allele and the p.G350R allele in peroxisomal and mitochondrial morphology. However, the p.A395D appeared to have greater toxicity when compared to p.G350R. The slight difference in severity might relate to the observation that Patient 1 exhibited normal plasma VLCFA levels compared to the abnormalities reported for p.A395D. Prior to this study, peroxisomal fission had not been studied in Drosophila, but whether distinct amino acids in the middle domain of DNM1L have distinct roles in mitochondrial versus peroxisomal fission is a hypothesis that can be explored through further studies of human DNM1L middle domain variants.
In conclusion, WES is a powerful diagnostic tool for infantile mitochondrial and peroxisomal phenotypes. In addition, patients like Patient 2 who have two de novo variants present a diagnostic challenge. Functional exploration of human gene variants in Drosophila is informative in these cases. Our data suggest that DNM1L variants may need to be considered in a range of encephalopathies. Our data also suggest that studying organelle dynamics in Drosophila can aid in determining the pathogenicity of variants linked to organelle dysfunction and rare disease phenotypes.

Both patients were enrolled in IRB-approved human studies at University of Texas Houston (Patient 1) and Baylor College of Medicine (Patient 2), as part of the Biochemical and Cell Biology Correlates of Peroxisomal Disorders study. Clinical case histories presented represent standard clinical care including radiologic, biochemical and molecular testing. Plasma sample from Patient 1 were sent for VLCFA levels on a clinical basis as described.(35)
Both patients underwent WES through the Whole Genome Laboratory (https://www.bcm.edu/research/medical-genetics-labs/index.cfm?PMID=21319) using methods described.(36) Produced sequence reads were aligned to the GRCh37 (hg19) human genome reference assembly using the HGSC Mercury analysis pipeline (http://www.tinyurl.com/HGSC-Mercury/). Variants were determined and called using the Atlas2 suite to produce a variant call file.(37) For the population comparisons we utilized data from the Exome Aggregation Consortium (ExAC), Cambridge, MA (URL: http://exac.broadinstitute.org) [November 2015] and Exome Variant Server, NHLBI GO Exome Sequencing Project (ESP), Seattle, WA (URL: http://evs.gs.washington.edu/EVS/) [November 2015]. Parental studies for DNM1L and PDHA1 as well as other variants noted in Supplemental Table 1 were performed by Sanger confirmation in proband and sequencing in parental blood DNA samples. The % mosaicism for the maternal sample for Patient 1 was determined by examination of the Sanger traces.

We generated human DNM1L with and without the three variants of interest in a series of constructs which were codon-optimized for Drosophila expression (GeneArtTM). We then subcloned these constructs in the pUAST-attB vector and generated transgenic flies by injecting prepared DNA into embryos.(38). We targeted the VK00033 site for site- specific integration (y[1] w[1118]; PBac[y[+]-attP-3B]VK00033).(39, 40)The Drp11 and Drp12 alleles were those reported.(27) Transgenic DNM1L constructs were crossed into these genetic backgrounds.Two peroxisomal reporters were used in third instar larval salivary gland, a UAS-GFP- SKL construct generated by subcloning a c-terminal SKL tagged GFP into the UAS vector and a transgenic insertion on 2nd chromosome was recombined with Actin-GAL4 (y1 w*; P{Act5C-GAL4}25FO1/CyO, y+). Pex3 staining was performed as described.(33) Confocal images were quantified using Image J software.Mitochondrial trafficking and quantification at the third instar neuromuscular junction was assayed as described.(32)
We thank the families for their participation in the research. This work was funded by the National Institutes of Neurological Disorders and Stroke (K08NS076547 to MFW), the Simmons Family Foundation Collaborative Award (to MFW and HJB), the Clayton Murphy Peroxisomal Disorders Research Funds, and the Baylor College of Medicine Medical Genetics Training Grant T32-GM07526-37 (LR). The authors would like to thank the Exome Aggregation Consortium and the groups that provided exome variant data for comparison. The authors would also like to thank the NHLBI GO Exome Sequencing Project and its ongoing studies which produced and provided exome variant calls for comparison: the Lung GO Sequencing Project (HL-102923), the WHI Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the G150 Seattle GO Sequencing Project (HL-102926) and the Heart GO Sequencing Project (HL- 103010).