The 28-year-old female patient and her healthy parents provided written informed consent as approved by the Institutional Review Boards of Hanyang University Hospital (Seoul, Korea). Genomic DNA was isolated from peripheral blood leukocytes using the Wizard Genomic DNA Purification kit (Promega, Madison, WI, USA). The exomes of subjects were captured using the Agilent SureSelect all Exon 50Mb kit (Agilent, Santa Clara, CA, USA) and sequenced on an Illumina NextSeq500 machine (paired-end and 100-bp reads) (Illumina, San Diego, CA, USA). Reads were mapped to a custom GRCh37/hg19 build using the Burrows-Wheeler Aligner (BWA). Annotation was performed using an in-house custom-made script. We selected rare variants with allele frequency less than 0.01 identified in the NHLBI Exome Sequencing Project (http://evs.gs.washington.edu/EVS/), the 1000 Genomes Project (http://www.1000genomes.org/), and gnomAD (http://gnomad.broadinstitute.org/). All amino acid-altering de novo variants with filtering criteria were validated by Sanger sequencing on DNA samples from the family trio using primers designed by the authors.

Full-length cDNAs for RAPGEF2 and Bcl-2-associated X protein (BAX) were PCR-amplified from total human cDNAs and cloned into the pEGFP-C1 vector (Clontech, Mountain View, CA, USA) to generate pEGFP-RAPGEF2 and pEGFP-BAX. The E1357K mutation was introduced into pEGFP-RAPGEF2 by PCR-mediated mutagenesis to produce pEGFP-RAPGEF2-E1357K.

Primary human fibroblasts were established from punch biopsies on the forearm skin of the patient and a 43-year-old male control as described previously [23] and maintained in Dulbecco's modified Eagle's medium (DMEM) containing 20% heat-inactivated (30 min, 55℃) fetal bovine serum (FBS), 1% non-essential amino acids, and antibiotics. Passage-matched control and patient fibroblasts (prior to passage 10) were used in each experiment. For inhibition of HDAC6, human skin fibroblasts were treated with 1 µM tubastatin A (Sigma-Aldrich, St. Louis, MI, USA) overnight at 37℃. Human HeLa cells were maintained in DMEM supplemented with 10% heat-inactivated FBS and transfected using FuGENE HD transfection reagent (Promega, Madison, WI, USA).

Flies were maintained at 25℃ on standard food. Transgenic D42-GAL4 and D42-GAL4, UAS-mito-HA-GFP flies were obtained from the Bloomington Stock Center (Bloomington, IN, USA). The UAS-HA-RAPGEF2-WT and UAS-HA-RAPGEF2-E1357K transgenes were produced in the w¹¹¹⁸ background by standard germline transformation. The RNAi line PDZ-GEFKK¹⁰²⁶¹² (referred to here as to UAS-gef26^RNAi) was obtained from the Vienna Drosophila Resources Center (Vienna, Austria).

Cultured cells were fixed with 4% formaldehyde in PBS for 20 min at room temperature, permeabilized with 0.2% Triton X-100 in PBS for 10 min, and blocked with 1% BSA in PBS for 1 h. Samples were then incubated with primary antibodies for 1 h and sequentially incubated with fluorescently labeled secondary antibodies for 30 min at room temperature. Wandering third-instar Drosophila larvae were dissected in Ca²⁺-free HL3 saline, fixed with 4% formaldehyde in PBS for 20 min, and immunostained as previously described [24]. The following primary antibodies were used in this study: anti-mitochondria (1:1000; Millipore, Burlington, MA, USA), anti-acetylated α-tubulin (1:500; Sigma-Aldrich), anti-tyrosinated α-tubulin (1:500; Millipore, Bedford, MA, USA), anti-α-tubulin (1:500; Sigma-Aldrich), and anti-Futsch (1:50; DSHB, Iowa City, IA, USA). The FITC-, Cy3-, and Cy5-conjugated secondary antibodies were obtained from Jackson ImmunoResearch (1:200; West Grove, PA, USA).

Fluorescent images were acquired with an LSM 700 laser-scanning confocal microscope using a C Apo 40x W or Plan Apo 63x 1.4 NA objective (Carl Zeiss, Jena, Germany). The length of Mito-RFP-labeled mitochondria in fibroblasts was determined using ImageJ. For quantification of the number and area of mitochondria in Drosophila motor neurons, we acquired confocal images of A5 segment in dissected third instar larvae labeled with anti-GFP (for Mito-GFP) and anti-HRP antibodies. The number and area of Mito-GFP puncta normalized to the respective area were also determined using ImageJ as previously described [25].

A climbing test was used to assess the locomotor function of adult flies as previously described [24]. Briefly, 45 adult flies aged for 20 days were transferred into a glass graduated cylinder. Following 5-min acclimation, flies were gently tapped to the bottom and the distance climbed by individual flies in a 30 s period was measured.

Primary human skin fibroblasts were homogenized in ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Triton X-100, and protease inhibitors) and subjected to western blotting as previously described [23]. For some experiments, we separated the mitochondrial and cytosolic fractions from fibroblast lysates using the BioVision Mitochondria/Cytosol Fractionation kit (BioVision, Milpitas, CA, USA). The following primary antibodies were used: anti-acetylated α-tubulin (1:1000; Sigma-Aldrich), anti-tyrosinated α-tubulin (1:1000; Millipore), anti-α-tubulin (1:1000; Sigma-Aldrich), anti-BAX (1:1000; BD, Franklin Lakes, NJ, USA), and anti-GAPDH (1:1000; Santa Cruz Biotechnology, Dallas, TX, USA).

The mitochondrial membrane potential was assessed in live primary fibroblasts using 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide (JC-1; Sigma-Aldrich) as described [26]. Briefly, fibroblasts were washed and incubated with 5 µg/ml JC-1 dye for 20 min at 37℃. The cells were then rinsed with culture medium, and their images were obtained using the Applied Precision DeltaVision fluorescence microscopy system (GE Healthcare, Chicago, IL, USA). JC-1 accumulates as red fluorescent aggregates within polarized mitochondria but does as green fluorescent monomers within less polarized mitochondria.

Primary human skin fibroblasts were fixed in PBS containing 4% paraformaldehyde and 2.5% glutaraldehyde for 24 h and rinsed in PBS. The samples were then subjected to 70-nm sectioning after gradual dehydration in ethanol solutions and propylene oxide (Acros Organics, Morris Plains, NJ, USA), and stained with epoxy resins using standard procedures. Images were acquired with a Hitachi electron microscope (Hitachi, Tokyo, Japan) equipped with a ES500W digital camera (GATAN, Pleasanton, CA, USA).

Comparisons were made by one-way ANOVA analysis with a post-hoc Turkey test. Data are presented as mean±SEM.

While performing WES on sALS trios, we identified a 28-year-old female patient carrying a de novo variant in the RAPGEF2 gene (c.4069G>A, p.E1357K) (Fig. 1A). The de novo occurrence of this variant in the patient was confirmed by Sanger sequencing (Fig. 1B). In addition, we proved no additional pathogenic variants in the most common fALS genes including C9orf72, SOD1, FUS, TARDBP, and TBK1. She developed progressive weakness first in the lower limbs and then in the upper limbs from 27 years of age. The neurologic examination at our clinic showed clinical signs of upper motor neuron involvement, mild hand wasting, and fasciculation in the lower limbs. Needle electromyography (EMG) showed widespread active denervation potentials with fasciculation in both upper and lower limbs. In addition, we observed long-duration, large-amplitude polyphasic potentials during volitional contraction, suggesting that there is chronic re-innervation. Based on the revised El Escorial criteria [27], she was diagnosed with clinically probable ALS.

The RAPGEF2 gene encodes a Rap1-specific guanine nucleotide exchange factor (GEF) that harbors an N-terminal cyclic nucleotide monophosphate-binding (cNMP) domain, a Ras-GEFN domain, a PSD-95/Dlg/ZO-1 (PDZ) domain, a Ras association (RA) domain, and a RasGEF domain, which is followed by two C-terminal low-complexity domains (LCDs). Previous studies have shown that ALS-causing mutations are frequently found in the LCD domains of RNA-binding proteins, including TDP-43, FUS, hnRNP A1, and TIA1 [28,29,30]. The RAPGEF2-E1357 residue is highly conserved among species ranging from Drosophila to vertebrates and locates within the second LCD domain (Fig. 1C).

Disruptions in microtubule network assembly have been proposed as a critical component of ALS pathogenesis [22]. We therefore asked whether the E1357K variant of RAPGEF2 affects microtubule dynamics or organization. To address this, we visualized microtubule networks in control- and patient-derived skin fibroblasts using antibodies against α-tubulin (detecting both free α-tubulin and microtubules), acetylated α-tubulin (detecting long-lived, stable microtubules [31]), and tyrosinated α-tubulin (detecting both free tubulin and newly formed microtubules [32]). In control cells, all of these tubulin antibodies revealed the typical microtubule pattern of an aster-like distribution extending toward the cell periphery (Fig. 2A). The levels and distribution of α-tubulin and tyrosinated α-tubulin remained unchanged in patient fibroblasts (Fig. 2A). In sharp contrast, anti-acetylated α-tubulin signals were weaker in patient fibroblasts than in control fibroblasts (Fig. 2A). In addition, acetylated α-tubulin networks were restricted only in the perinuclear region. We confirmed the selective reduction of acetylated α-tubulin in patient cells by western blotting (Fig. 2B~D). These results suggest that the E1357K variant of RAPGEF2 affects the stability of the microtubule network.

To corroborate the above conclusion, we analyzed the effect of the RAPGEF2-E1357K variant on the level of α-tubulin acetylation in HeLa cells. Levels of anti-acetylated α-tubulin signal were significantly decreased in RAPGEF2-E1357K-transfected cells compared with untransfected control cells (Fig. 2E), confirming a deleterious effect of the RAPGEF2 variant on microtubule stability.

Since appropriate mitochondrial distribution critically depends on microtubule-based transport, we investigated whether alteration of the microtubule network in patient-derived fibroblasts is paralleled with abnormalities in mitochondrial morphology and distribution. Control and patient fibroblasts were transfected with a mitochondrial matrix-localized RFP (Mito-RFP) reporter construct. In control cells, Mito-RFP signals largely appeared as tubular networks extending throughout the cytoplasm (Fig. 3A). In contrast, the Mito-RFP-labeled mitochondria networks in patient-derived cells were fragmented and more restricted around the perinuclear area (Fig. 3A). To quantify mitochondrial fragmentation, we measured mitochondrial length. Patient fibroblasts showed a reduction of about 88% in average mitochondrial length compared with control cells (Fig. 3B). However, the size of cells was comparable between both genotypes (Fig. 3C).

Next, we investigated whether the decrease in the stability of microtubules is a causative mechanism of abnormal mitochondrial distribution in patient-derived fibroblasts. Acetylation of lysine 40 in α-tubulin, an indication of microtubule stabilization, is increased by inhibiting the catalytic activity of histone deacetylase 6 (HDAC6) [33], which is the major deacetylase of α-tubulin [34]. We examined the effect of an HDAC6-selective inhibitor (tubastatin A) on mitochondrial distribution in patient fibroblasts. Treatment of fibroblasts with 1 µM tubastatin A restored abnormalities in acetylated α-tubulin distribution and mitochondrial distribution and length (Fig. 3A~C). These results suggest that impaired microtubule stability is a causative mechanism accounting for abnormal mitochondrial distribution in patient-derived fibroblasts expressing the RAPGEF2-E1357K variant.

To validate the effects of the RAPGEF2-E1357K mutant on microtubule maintenance and mitochondrial distribution in motor axons, we generated Drosophila carrying UAS transgenes of HA-tagged wild-type RAPGEF2 (UAS-HA-RAPGEF2-WT) and its E1357K mutant (UAS-HA-RAPGEF2-E1357K). We targeted expression of these transgenes in larval motor neurons using the D42-GAL4 driver. First, we performed confocal analysis on third instar larvae with antibodies against acetylated α-tubulin and Futsch, a microtubule-associated protein stabilizing microtubules [35]. Both acetylated α-tubulin and Futsch staining in motor axons and neuromuscular junction (NMJ) synapses were not altered by HA-RAPGEF2-WT overexpression (Fig. 4A~H). However, levels of both markers were significantly reduced in motor axons and NMJ synapses in larvae overexpressing HA-RAPGEF2-E1357K (Fig. 4A~H). These results suggest that expression of the RAPGEF2-E1357K mutation deleteriously impacts microtubule stability through a toxic gain-of-function mechanism rather than a simple dose-dependent mechanism. Next, we compared the distribution of Mito-GFP-labeled mitochondria in the soma and at the terminals of motor neurons in control larvae and larvae expressing HA-RAPGEF2-WT or HA-RAPGEF2-E1357K. The levels and distribution of Mito-GFP in the soma of motor neurons were not significantly different between the three genotypes (Fig. 5A). However, mitochondria number and area in motor axons were significantly reduced in HA-RAPGEF2-E1357K-expressing larvae compared with control or HA-RAPGEF2-WT-expressing larvae (Fig. 5B~E). We also observed a similar reduction in the area of mitochondria in NMJ synapses of HA-RAPGEF2-E1357K-expressing larvae (Fig. 5C and F). The aggregated distribution of mitochondria precluded comparison of mitochondria number in NMJ synapses.

We have recently shown that loss of the Drosophila RAPGEF2 homolog (Gef26) causes an increase of axonal Futsch staining [24]. To confirm this and to preclude the possibility that the E1357K mutation exerts its deleterious effect on mitochondria distribution through a loss-of-function mechanism, we depleted expression of Gef26 in motor neurons using a transgenic RNA interference (RNAi) approach. We first confirmed an increase of both acetylated α-tubulin and Futsch staining in motor axons and NMJ synapses in Gef26-knockdowned larvae (Fig. 4A~H). We found no obvious changes in the levels and distribution of mitochondria in the soma and at the terminals of motor neurons (Fig. 5A~F). These results support the notion that the RAPGEF2-E1357K mutation deleteriously impact mitochondria distribution through a toxic gain-of-function mechanism.

To test if expression of the RAPGEF2-E1357K mutant induces motor dysfunction, we performed climbing assays on adult flies at 20 days of age. Compared with transgenic controls (D42-GAL4/+), flies expressing UAS-HA-RAPGEF2-WT in motor neurons under the control of D42-GAL4(D42-GAL4/UAS-HA-RAPGEF2-WT) displayed normal climbing ability (D42-GAL4/+: 20.56±0.22 cm; D42-GAL4/UAS-HA-RAPGEF2-WT: 20.04±0.29 cm; p>0.05; Fig. 5G and H). However, age-matched flies expressing the RAPGEF2-E1357K variant made relatively short climbs (16.98±0.47 cm; p<0.01 from D42-GAL4/+; Fig. 5G and H), demonstrating the toxic effect of the human RAPGEF2-E1357K variant in Drosophila motor neurons.

To investigate the effects of the RAPGEF2-E1357K variant on mitochondrial structures, we performed transmission electron microscopy (EM) for control and patient fibroblasts. Control mitochondria showed a typical crista structure with electron-dense deposits in the matrix (Fig. 6A). In contrast, patient fibroblasts displayed swollen and vacuolated mitochondria without lamella cristae and electron-dense deposits (Fig. 6A).

To investigate whether the ultrastructural mitochondrial abnormalities in patient fibroblasts are paralleled with mitochondrial dysfunction, we assessed mitochondrial membrane potential using JC-1, a cationic lipophilic dye. The JC-1 dye accumulates as red-fluorescent J-aggregates within mitochondria at high membrane potentials (energized mitochondria), while it accumulates as green fluorescent monomers within mitochondria at low membrane potentials (deenergized mitochondria) [36]. In live control fibroblasts loaded with JC-1, we observed strong red fluorescent signals but not green fluorescent signals (Fig. 6B), suggesting the majority of mitochondria are functional. In contrast, both red and green fluorescent signals were prominent in JC-1-loaded patient fibroblasts (Fig. 6B). These results suggest that mitochondria activity is lower in patient fibroblasts than in control fibroblasts.

Mitochondrial dysfunction is intimately linked to apoptotic cell death, which involves the mitochondrial recruitment of the proapoptotic regulator BAX from the cytosol [37]. We therefore examined whether mitochondrial dysfunction in patient-derived skin fibroblasts is paralleled with abnormal accumulation of BAX on the mitochondria. To this end, we transfected control and patient fibroblasts with a GFP-BAX construct and stained them with anti-GFP and anti-mitochondria. In control fibroblasts, BAX showed a diffuse cytoplasmic distribution with a minimal overlap with mitochondria (Fig. 7A). Notably, we observed a prominent overlap between BAX signals and mitochondria in patient fibroblasts (Fig. 7A). As an additional approach, we separated cytosolic and mitochondrial fractions from lysates of control and patient fibroblasts and performed Western blot analysis of these fractions using anti-BAX. This experiment demonstrated that BAX was more recruited to the mitochondria in patient fibroblasts than in control fibroblasts (Fig. 7B and C). In a control experiment, we confirmed that total BAX levels were comparable between control and patient derived fibroblasts (data not shown). Finally, we found that the abnormal accumulation of BAX at the mitochondria was rescued by inhibiting HDAC6 with 1 µM tubastatin A (Fig. 7A~C). The same dose of tubastatin A had no effect on the BAX distribution in control-derived fibroblasts (Fig. 7A~C).