Altered oligomeric states in pathogenic ALS2 variants associated with juvenile motor neuron diseases cause loss of ALS2-mediated endosomal function

Familial amyotrophic lateral sclerosis type 2 (ALS2) is a juvenile autosomal recessive motor neuron disease caused by the mutations in the ALS2 gene. The ALS2 gene product, ALS2/alsin, forms a homophilic oligomer and acts as a guanine nucleotide–exchange factor (GEF) for the small GTPase Rab5. This oligomerization is crucial for both Rab5 activation and ALS2-mediated endosome fusion and maturation in cells. Recently, we have shown that pathogenic missense ALS2 mutants retaining the Rab5 GEF activity fail to properly localize to endosomes via Rac1-stimulated macropinocytosis. However, the molecular mechanisms underlying dysregulated distribution of ALS2 variants remain poorly understood. Therefore, we sought to clarify the relationship between intracellular localization and oligomeric states of pathogenic ALS2 variants. Upon Rac family small GTPase 1 (Rac1) activation, all mutants tested moved from the cytosol to membrane ruffles but not to macropinosomes and/or endosomes. Furthermore, most WT ALS2 complexes were tetramers. Importantly, the sizes of an ALS2 complex carrying missense mutations in the N terminus of the regulator of chromosome condensation 1–like domain (RLD) or in-frame deletion in the pleckstrin homology domain were shifted toward higher molecular weight, whereas the C-terminal vacuolar protein sorting 9 (VPS9) domain missense mutant existed as a smaller dimeric or trimeric smaller form. Furthermore, in silico mutagenesis analyses using the RLD protein structure in conjunction with a cycloheximide chase assay in vitro disclosed that these missense mutations led to a decrease in protein stability. Collectively, disorganized higher structures of ALS2 variants might explain their impaired endosomal localization and the stability, leading to loss of the ALS2 function.

Familial amyotrophic lateral sclerosis type 2 (ALS2) is a juvenile autosomal recessive motor neuron disease caused by the mutations in the ALS2 gene. The ALS2 gene product, ALS2/ alsin, forms a homophilic oligomer and acts as a guanine nucleotide-exchange factor (GEF) for the small GTPase Rab5. This oligomerization is crucial for both Rab5 activation and ALS2-mediated endosome fusion and maturation in cells. Recently, we have shown that pathogenic missense ALS2 mutants retaining the Rab5 GEF activity fail to properly localize to endosomes via Rac1-stimulated macropinocytosis. However, the molecular mechanisms underlying dysregulated distribution of ALS2 variants remain poorly understood. Therefore, we sought to clarify the relationship between intracellular localization and oligomeric states of pathogenic ALS2 variants. Upon Rac family small GTPase 1 (Rac1) activation, all mutants tested moved from the cytosol to membrane ruffles but not to macropinosomes and/or endosomes. Furthermore, most WT ALS2 complexes were tetramers. Importantly, the sizes of an ALS2 complex carrying missense mutations in the N terminus of the regulator of chromosome condensation 1-like domain (RLD) or in-frame deletion in the pleckstrin homology domain were shifted toward higher molecular weight, whereas the C-terminal vacuolar protein sorting 9 (VPS9) domain missense mutant existed as a smaller dimeric or trimeric smaller form. Furthermore, in silico mutagenesis analyses using the RLD protein structure in conjunction with a cycloheximide chase assay in vitro disclosed that these missense mutations led to a decrease in protein stability. Collectively, disorganized higher structures of ALS2 variants might explain their impaired endosomal localization and the stability, leading to loss of the ALS2 function.
Mutations in the ALS2 gene account for a number of autosomal recessive juvenile-onset motor neuron diseases (MNDs), 2 including ALS type 2 (ALS2), juvenile primary lateral sclerosis, and infantile-onset ascending hereditary spastic paralysis (IAHSP). These MNDs are characterized by a selective and progressive degeneration of motor neurons, particularly the upper motor neurons (1)(2)(3), indicating that the ALS2 gene and its gene product, ALS2 (aka alsin), play crucial roles in the maintenance and/or survival of motor neurons (4,5).
ALS2 is a 184-kDa protein of 1657 amino acid residues (aa) that comprises three predicted guanine nucleotide-exchange factor (GEF) domains as follows: the N-terminal regulator of chromosome condensation 1-like domain (RLD) (6); the central Dbl-homology and pleckstrin-homology (DH/PH) domain (7); and the C-terminal vacuolar protein sorting 9 (VPS9) domain ( Fig. S1) (8). In addition, eight consecutive membrane occupation and recognition nexus (MORN) motifs (9) that are related to membrane binding are noted in the region between DH/PH and VPS9 domains. Indeed, it has been shown that ALS2 acts as a GEF for the small GTPase Rab5 via the C-terminal MORN/VPS9 domain and modulates endosome fusion and trafficking by activating Rab5 (10). Furthermore, the activation of Rac family small GTPase 1 (Rac1) results in a drastic redistribution of ALS2 from the cytosol to membrane compartments, thereby inducing endosome enlargement in an ALS2mediated Rab5 GEF activity-dependent manner (11).
Thus far, a total of 40 mutations in the ALS2 gene has been reported in patients with MNDs (1,2,4,5,. As most of them are frameshift or nonsense mutations, the resulting ALS2 variants either are unstable and thus liable to be degraded via the ubiquitin-proteasome system (UPS) (33) or are nonfunctional due to lack of the C-terminal Rab5 GEF domain (33). Recently, several missense or in-frame deletion mutations in the ALS2 gene, which are predicted to code for mutant proteins preserving the C-terminal Rab5 GEF domain (i.e. ALS2 G49R (13), ALS2 F65S (12), ALS2 S100I (14), ALS2 A120del (15), ALS2 C157Y (16), ALS2 P192L (18), ALS2 G540E (20), ALS2 A861_T904del (14), ALS2 S1116_T1170del (29), and ALS2 R1611W (15)), were identified (Fig. S1). It has been reported that ALS2 C157Y is preferentially degraded by the UPS when compared with WT ALS2 (16). In addition, we have demonstrated that, although both ALS2 C157Y and ALS2 G540E variants retain the Rab5 GEF activity in vitro, they fail to localize to macropinosomes and early endosomes upon Rac1 activation when ectopically expressed in cells (34). Thus, it is possible that such abnormal cellular behaviors of missense ALS2 variants are also implicated in loss of the ALS2 function. However, due to a limited number of pathogenic variants analyzed thus far, it is still too early to conclude as such, let alone underlying molecular mechanism for ALS2linked MNDs.
Previously, we have shown that ALS2 interacts with itself through the C-terminal region, forming a homophilic oligomer (35). Importantly, this homo-oligomerization is not only crucial for the Rab5 GEF activity in vitro but also for the ALS2-mediated endosome fusion in cells (11,35). These results lend credence to the hypothesis that structural abnormalities in pathogenic missense and/or in-frame deletion ALS2 variants are closely linked to loss of its cellular function, and therefore that formation of the proper ALS2 complex might be a prerequisite for Rac1-induced endosomal localization of ALS2 and the following activation of endosomal Rab5.
To prove this hypothesis, we here sought to clarify the relationship between intracellular localization and oligomeric states of several missense and in-frame deletion pathogenic ALS2 variants by conducting a series of immunocytochemical and biochemical analyses. In addition, to further gain insight into the structural basis of missense mutations, we generated a protein structure of N-terminal RLD harboring the majority of the missense pathogenic mutations, and we performed an in silico mutagenesis analyses to evaluate the potential effects of those mutations on the structure. We here demonstrated evidence showing that the altered oligomerization and/or destabilization of ALS2 was implicated in their impaired endosomal localization. Our data may provide important clues to understand the molecular mechanisms underlying the pathogenesis of ALS2-linked MNDs.

Pathogenic ALS2 variants localize to Rac1-induced membrane ruffles but not to macropinosomes and early endosomes
Previously, we have demonstrated that the majority of the ectopically expressed WT ALS2 protein (ALS2 WT ) is scattered throughout the cytoplasm in HeLa and COS-7 cells (10). Upon Rac1 activation, ALS2 WT is recruited to membrane ruffles and then relocalized to macropinosomes and early endosomes via Rac1-stimulated macropinocytosis (11). By contrast, two pathogenic ALS2 variants, ALS2 C157Y and ALS2 G540E , fail to be localized to macropinosomes and early endosomes (34).
To clarify the intracellular behavior of other missense as well as small in-frame deletion ALS2 variants, which have been newly identified but yet to be tested, we chose three representative pathogenic ALS2 variants: ALS2 S100I , the N-terminal RLD missense variant (14); ALS2 A861_T904del , the central PH in-frame deletion variant (14); and ALS2 R1611W , the C-terminal VPS9 missense variant (15), in addition to ALS2 WT and normal variant: ALS2 P132L . Using these variants, we investigated the changes in their intracellular distributions by the Rac1 activation. All variants were distributed throughout the cytoplasm in the absence of active Rac1 (Rac1 Q61L ), similar to ALS2 WT (Fig.  1A) and ALS2 P132L (Fig. S2A). Upon Rac1 activation, ALS2 WT was relocalized to the surface of large vesicular compartments, whereas the majority of variant proteins remained in the cytoplasm (Fig. 1A). Notably, despite that pathogenic ALS2 variants failed to localize to vesicular compartments, a significant portion of them was recruited to the Rac1-induced F-actinpositive membrane ruffles (Fig. 1A), reproducing the phenotypes of two previously examined ALS2 variants: ALS2 C157Y and ALS2 G540E (Fig. S2B).
Because the activation of Rac1 induced the formation of large vesicular compartments both in ALS2 WT and ALS2 variantoverexpressing cells, we next investigated whether these vesicular compartments represented either macropinosomes or early endosomes and whether ALS2 was closely localized to these compartments. ALS2 WT was not only present on the surface of the dextran-positive macropinosomes ( Fig. 2A) but was also localized to EEA1-positive early endosomes (Fig. 1B). Line analysis clearly showed that the signals for engulfed dextran in macropinosomes were surrounded by those of ALS2 WT (Fig.  2B). In contrast, all three pathogenic ALS variants examined, ALS2 S100I , ALS2 A861_T904del , and ALS2 R1611W , were overlapped neither with dextran-positive macropinosomes nor with EEA1positive endosomes (Figs. 1B and 2, A and B). Furthermore, coexpression of active Rac1 with ALS2 WT , but not with ALS2 variants, led to an enlargement of EEA1-positive endosomal compartments (Fig. 1B), consistent with our previous observations (34).
These results indicate that pathogenic ALS2 variants, which retain the C-terminal Rab5 GEF domain, lose their capability to internalize via Rac1-stimulated macropinocytosis. In other words, pathogenic ALS2 variants are defective in the localization to macropinosomes and/or early endosomes, even though they do not seem to hinder macropinocytosis itself and can be localized to membrane ruffles in cells.

Pathogenic ALS2 variants interact with themselves
We have previously reported that the ALS2 molecules selfinteract through the N-and/or C-terminal regions (11,35). It has also been shown that the pathogenic ALS2 variants, ALS2 C157Y and ALS2 G540E , can interact with themselves (34). To confirm whether ALS2 S100I , ALS2 A861_T904del , or ALS2 R1611W self-interact, we conducted coimmunoprecipitation experi-ments. In all the cases examined; ALS2 WT , ALS2 G540E , ALS2 S100I , ALS2 A861_T904del , and ALS2 R1611W , the HA-tagged ALS2 molecules were efficiently coimmunoprecipitated with FLAG-tagged ones (Fig. 3). The results demonstrate that like ALS2 WT , missense or small in-frame deletion pathogenic ALS2 variants can interact with each other, indicating that pathogenic ALS2 variants could either homodimerize or homo-oligomerize in cells.

WT ALS2 forms a tetramer complex in mammalian cells
We have previously shown that both purified FLAG-tagged ALS2 and its artificial truncated mutant lacking both the N-terminal RLD and DH/PH domain of ALS2 (ALS2-MORN/VPS9) can form stable homophilic oligomers, presumably octamer, with no other detectable stable binding partners (36) and that this homo-oligomerization is crucial for the activation of Rab5 (35). Because the detergents used for the purification of the ALS2 proteins were different between the GEF assay, in which Tween 20 was used, and gel filtration, in which IGEPAL CA-630 was used, we tested the effects of differences in detergents as well as in the length of the storage period of purified ALS2 on the Rab5 GEF activity and oligomeric states. Surprisingly, we found that ALS2 in the solution containing Tween 20 kept stable activities for an approximate 1-year period when stored at 4°C, whereas those in the solution containing IGE-PAL CA-630 gradually lost their catalytic activity over a month with concomitant increase in sizes of the oligomeric complex (data not shown) (36). Thus, we suspected that an octameric form of the ALS2 complex in vitro (35) might not reflect the native form.
To determine the bona fide ALS2 complex in cells, we prepared fresh cell lysates from FLAG-ALS2 WT -expressing COS-7 cells without any purifications, used them for gel-filtration chromatography, and detected ALS2 by Western blotting using ALS2-specific antibody. As a result, ALS2 WT (ϳ184 kDa) was eluted in fractions with an apparent peak molecular mass of ϳ700 kDa (Fig. 4, A and B). Importantly, this elution pattern was well matched to that of endogenous ALS2 in nontransfected cells (Fig. S3). Thus, it is reasonable that ALS2 forms a homophilic tetramer rather than an octamer in native conditions. Nonetheless, because the endogenous ALS2 complexes are slightly larger than those of overexpressed ALS2, it is still possible that the ALS2 complex contains some additional interactors in vivo.

Pathogenic mutations alter the oligomeric states of ALS2
larger complex. Because the C-terminal region of ALS2 between MORN and VPS9, which covered the Rab5 GEF catalytic domain (10,35), directly mediated self-interaction (35), we next conducted gel-filtration analysis of cell lysate prepared from FLAG-ALS2-MORN/VPS9-expressing (Fig. S1) COS-7 cells as well as the recombinant purified ALS2-MORN/VPS9 protein prepared from bacterial cultures. Both FLAG-ALS2-MORN/VPS9 and the recombinantly purified ALS2-MORN/ VPS9 (molecular mass was ϳ70 kDa for each) were eluted in fractions with an apparent peak molecular mass of ϳ160 kDa (Fig. S5), indicative of a dimeric complex of the C-terminal MORN/VPS9 region of ALS2.
Taken together, it is predicted that the dimeric ALS2 complex, which interacts through the C-terminal MORN/VPS9 region in an antiparallel fashion (35), can further interact with each other in a DH/PH domain-dependent manner, forming a tetrameric ALS2 complex in cells. It is also suggested that the RLD of ALS2 is not essential to form a proper tetrameric complex, even though it retains the ability to interact with the C-terminal region of ALS2 (11).

Pathogenic ALS2 mutations alter the oligomeric states of the ALS2 complex
To define the oligomeric states of normal ALS2 variants (ALS2 P132L and ALS2 E1173K ; rs# cluster ID/NCBI Assay IDs rs41308810/ss65624394 and rs41309046/ss65624421, respectively) (dbSNP, LOCUSID 57679), Rab5 GEF-defective artificial mutant (ALS2 P1603A ), and Rab5 GEF-defective/endosomal localization-defective artificial mutant (ALS2 L1617A ) (10), we performed gel-filtration analysis under the same experimental conditions. As a result, the elution patterns for ALS2 P132L , ALS2 E1173K , and ALS2 P1603A were mostly similar to those of ALS2 WT , except there was a minor peak at an apparent molecular mass of ϳ1000 kDa in these variants or mutants (Fig. 4, A   Figure 2. Pathogenic ALS2 variants do not redistribute to macropinosomes via Rac1-stimulated macropinocytosis. A, HeLa cells were transfected with pCIneoFLAG-ALS2 WT , pCIneoFLAG-ALS2 S100I , pCIneoFLAG-ALS2 A861_T904del , or pCIneoFLAG-ALS2 R1611W together with or without pCMV10 -2xHA-Rac1 Q61L . The cells were treated with tetramethylrhodamine-conjugated dextran (70 kDa) (red). After fixation, the cells were stained with anti-ALS2-RLD antibody (green). Higher magnification images of the boxed regions are shown on the right panels. Arrows and arrowheads indicate ALS2-enwrapped dextran-positive and ALS2-negative dextran-positive macropinosomes, respectively. Scale bars, 10 m. B, quantitative line analysis of the fluorescence intensities for ALS2 (green) and dextran (red) on dextran-positive macropinosomes. In each image, the asterisk indicates dextran-positive macropinosome of interest, and the end of the transverse bar with the number corresponds to the left end of the x axis of the graph. Asterisks and arrowheads noted in the graphs indicate the positions of dextran-positive macropinosome of interest and ALS2-localizing macropinosomal membranes, respectively.

Pathogenic mutations alter the oligomeric states of ALS2
and B). Interestingly, the elution pattern for ALS2 L1617A was different from others, in which ALS2 L1617A was eluted in a wide range of fractions with dual major peaks (ϳ400 and ϳ700 kDa) and a single minor peak (ϳ1000 kDa) (Fig. 4, A and B).
Next, we investigated the oligomeric states for the missense and small in-frame deletion pathogenic ALS2 variants (ALS2 G49R , ALS2 S100I , ALS2 C157Y , ALS2 G540E , ALS2 A861_T904del , and ALS2 R1611W ). Unlike ALS2 WT , the distributions of all pathogenic variants except for ALS2 R1611W were shifted toward the high-molecular weight fractions with two apparent peaks of ϳ700 and ϳ1000 kDa (Fig. 4, C and D). In contrast, ALS2 R1611W was broadly eluted in fractions with apparent molecular masses ranging from ϳ400 to ϳ700 kDa and an additional single minor peak of ϳ1000 kDa (Fig. 4, C and D), which was similar to those of ALS2 L1617A mutant.

ALS2-RLD WT and mutant RLDs were localized to phalloidin-positive F-actin
We have previously demonstrated that the ectopically-expressed ALS2-DH/PH/MORN/VPS9 fragment lacking the N-terminal RLD is instantly localized to early endosomes and induces their enlargements (10). In addition, the activation of Rac1 results in relocalization of full-length ALS2 from the cytoplasm to membrane ruffles and then to macropinosomes and/or early endosomes via macropinocytosis and heterotypic vesicle fusions (11). Thus, we consider that although the RLD of ALS2 acts as a suppressor for the recruitment of ALS2 to membranous compartments in a dormant state (10), it may rather function as a recruiter of ALS2 to membrane ruffles upon Rac1 activation. Taking into account the fact that 7 of 10 missense or in-frame deletion mutations that have been reported thus far are densely localized to the N-terminal RLD of ALS2, the RLD might have crucial roles in the intracellular behavior of ALS2.
To clarify the molecular tropism and function of the RLD, we investigated the localization of the RLD fragment of ALS2 (ALS2-RLD) in cells. Although all the mutants examined, ALS2-RLD S100I , ALS2-RLD C157Y , and ALS2-RLD G540E , were primarily distributed throughout the cytoplasm, a portion of them were colocalized with phalloidin-positive F-actin on membrane ruffles under Rac1-activating conditions, as were ALS2-RLD WT and ALS2-RLD P132L (Fig. 5, A and B). It was also noted that all the missense pathogenic ALS2 mutants could be recruited to membrane ruffles but failed to be relocalized to macropinosomes and/or early endosomes by the Rac1 activation (Figs. 1 and 2). The results combined suggest that the RLD carrying missense pathogenic mutations still retains the ability to mediate the localization of ALS2 to membrane ruffles but not to macropinosomes and/or endosomes.

Pathogenic mutations alter the oligomeric states of ALS2
fusion (37). We have previously demonstrated that two missense pathogenic ALS2 variants, ALS2 C157Y and ALS2 G540E , show a lower affinity to specific phosphoinositide phosphates (PIPs), particularly to PI(3)P, PI(4,5)P 2 , and PI(3,4,5)P 2 (34), suggestive of the regulatory mechanism for macropinosomal localization of ALS2. To clarify the affinities of ALS2-RLD Figure 4. Gel-filtration analysis of WT ALS2 and its variants. A, Western blot analysis of the gel-fractionated FLAG-tagged ALS2 and its variants and/or mutants. Lysates from COS-7 cells that were transfected with pCIneoFLAG-ALS2 WT , pCIneoFLAG-ALS2 P132L , pCIneoFLAG-ALS2 E1173K , pCIneoFLAG-ALS2 P1603A , or pCIneoFLAG-ALS2 L1617A and were applied to the Superose6 10/300 gel-filtration column. Cell lysates and the fractions were analyzed by Western blotting using anti-ALS2-MORN/VPS9 antibody. Arrows indicate the positions of molecular mass markers. B, quantitative analysis of signal intensities of ALS2 molecules. Each fraction was expressed as a mean percentage of the total amount of signals representing ALS2 across the entire fractions (n ϭ 4). C, Western blot analysis of the gel-fractionated FLAG-tagged pathogenic ALS2 variants. Lysates from COS-7 cells that were transfected with pCIneoFLAG-ALS2 G49R , pCIneoFLAG-ALS2 S100I , pCIneoFLAG-ALS2 C157Y , pCIneoFLAG-ALS2 G540E , pCIneoFLAG-ALS2 A861_T904del , or pCIneoFLAG-ALS2 R1611W and were applied to the Superose6 10/300 gel-filtration column. Cell lysates and the fractions were analyzed as shown in A. Arrows indicate the positions of molecular mass markers. D, quantitative analysis of signal intensities of ALS2 molecules. Each fraction is expressed as a mean percentage of the total signal intensity representing ALS2 across the entire fractions (n ϭ 4; except for G49R mutant (n ϭ 1)). For comparison, data for ALS2 WT that are the same as those shown in B are included.

Pathogenic mutations alter the oligomeric states of ALS2
toward PIP molecules, we performed protein-lipid overlay assay using the ALS2-RLD fragments as probes. All the ALS-RLD peptides, ALS2-RLD WT , normal variant ALS2-RLD P132L , and pathogenic missense ALS2-RLD mutants, bound to specific PIPs with different degrees of affinity (Fig. 5, D and E). However, unlike the cases of full-length ALS2 (34), the differences in the patterns of lipid affinity between WT and mutated RLD were not statistically significant (Fig. 5, D and E). The results indicate that missense mutations themselves, at least in the N-terminal RLD, do not alter their affinity toward PIPs, consistent with the unaltered intercellular distribution of ALS2-RLD mutants observed (Fig. 5, A and B).

ALS2-RLD WT can form multiple types of high-molecular weight protein complexes
We have previously reported that the N-terminal RLD can interact with the C-terminal MORN/VPS9 region of ALS2 (11). Furthermore, coimmunoprecipitation experiments demonstrated that the RLD self-interacted (data not shown). However, the oligomeric structure of the RLD complex is still undetermined. To address this issue, we conducted gel-filtration analysis of fresh cell lysates prepared from FLAG-ALS2-RLD WTexpressing COS-7 cells and the recombinant-purified ALS2-RLD WT prepared from bacterial cultures. FLAG-ALS2-RLD WT (ϳ73 kDa) was eluted in a wide range of fractions with three apparent peaks of ϳ150, ϳ700, and ϳ1000 kDa with the highest at ϳ150 kDa (Fig. 6, A and B). Interestingly, purified ALS2-RLD WT was eluted in fractions with two apparent peaks ranging from 400 to 700 and ϳ1000 kDa with no observable signals representing low-molecular mass complexes (Fig. S6). Thus, it is likely that ALS2-RLD WT can exist as a dimeric form in cells, but it easily forms multiple high-molecular mass complexes, probably tetramer to octamer (400 -700 kDa) or much larger complexes (ϳ1000 kDa) of ALS2-RLD in vitro, suggesting that the ALS2-RLD fragment may be intrinsically aggregation-prone.

Some pathogenic mutations in the RLD alter the oligomeric states of the ALS2-RLD fragment
Next, we investigated the oligomeric states of ALS2-RLD carrying normal variation ALS-RLD P132L or missense pathogenic mutations ALS2-RLD S100I , ALS2-RLD C157Y , and ALS2-RLD G540E . Signal distributions for some of these mutants were slightly shifted toward high-molecular weight fractions with concomitant decrease in the signal intensity of ϳ150 kDa dimeric fractions (Fig. 6, A and B). Indeed, quantitative analysis revealed that although the total signal intensities in high-molecular weight (HMW) fractions (fractions 1-8) for these mutants were comparable with that for ALS2-RLD WT , tetra-octameric fractions (fractions 9 -16) for ALS2-RLD S100I , ALS2-RLD P132L , and ALS2-RLD C157Y were significantly higher than that for ALS2-RLD WT with a concomitant decrease in their dimeric fractions (fractions 17-23) (Fig. 6C). These data indicate that some pathogenic missense mutations as well as normal variation in the N-terminal RLD can alter the oligomeric states of the ALS2-RLD fragment.

RLD of ALS2 is predicted to form a seven-bladed ␤-propeller
To clarify the structural and functional implications of missense mutations in the RLD of ALS2, we adopted the optimal target-template alignment approach and built a structural model for the ALS2-RLD fragment. First, we performed a BLASTP search against the PBD database using ALS2-RLD sequence as a query. The third RLD of human HERC2 (PDB code 3KCI) was selected as a template. Alignment of amino acid sequences between target (ALS2-RLD) and template (PDB code 3KCI) revealed that the N-terminal RLD of ALS2 encoded a seven-bladed ␤-propeller (Figs. 7 and 8 and Fig. S7). It was noted that among seven blades, two were predicted to be discontinuous, in which blade five was interrupted by 306 aa of intercalated disordered sequence (residues 219 -524) (Figs. 7 and 8), and blade 1 consisted of two separate sequences; both ends correspond to N-and C-terminal regions of the RLD (residues 1-39 and 664 -690, respectively) (Fig. 7). The three-dimensional ribbon representation of the RLD structure model in conjunction with the surface property prediction revealed that the "top face" of the propeller showed a hydrophilic nature, whereas the "bottom face" with an open central hole formed a hydrophobic cluster (Fig. 8, A and B, and Fig. S8), suggesting that the bottom face is implicated in the protein interactions. These results are fully consistent with the previously reported model, which was built based upon the human RCC1 sequence as a template (62).

Pathogenic mutations in the RLD reduce protein stability
To predict the effects of mutations and/or variations in the structure of the RLD, the position of the residues for six pathogenic missense mutations (Gly-49, Phe-65, Ser-100, Cys-157, Pro-192, and Gly-540), a single-residue deletion (Ala-120), and a variant in healthy individuals (Pro-132) was plotted on the three-dimensional ribbon representation of the RLD structure model (Fig. 8A). All residues except for Pro-192 were predicted to be exposed to the protein surface close to the top face of blades 2-5 (Fig. 8A). It is also noted that all mutations are localized within the seven-bladed propeller and not to the intercalated disordered sequence, suggesting that the structural distortion in the propeller is implicated in dysfunctional RLD of ALS2.
Next, to evaluate the effect of each mutation on the protein stability, we performed molecular dynamics simulations using CHARMm force field and calculated the difference in the folding free energy of the mutant and WT structures. The simulations showed that all pathogenic mutations but P192L reduced the protein stability (mutation energy Ͼ0.5 kcal/mol) ( Table 1). In particular, the S100I, C157Y, and G540E mutations strongly affected the protein stability (15.90, 4.49, and 10.09 kcal/mol, respectively). In contrast, the values of the mutation energy for P132L and P192L were 0.52 and 0.41 kcal/mol, respectively, both of which were lower than or very close to the cutoff value (0.5 kcal/mol), indicating that the protein structure is minimally affected by these variations. It is noted that homozygous P192L mutation was originally identified in Japanese patients with sporadic juvenile ALS (18) and that pathogenicity of the P192L mutation has yet to be validated. Therefore, it is still Pathogenic mutations alter the oligomeric states of ALS2 possible that like P132L, P192L merely represents a rare normal variant of ALS2.
As the residues at positions 100, 132, 157, and 540 were exposed to the protein surface, we further assessed the surface hydrophobicity. However, there were no significant differences in the hydrophobicity around these four residues between variants and WT (Fig. S9), indicating that the local surface hydrophobicity might not be responsible for the functional alteration in the pathogenic ALS2 variants. Taken together, the results imply that missense pathogenic mutations in the RLD of ALS2 destabilize the normal protein structure, namely the architecture for seven-bladed ␤-propeller.

Some pathogenic mutations decrease the stability of the fulllength ALS2 protein in NSC-34 cells
It has previously been shown that pathogenic frameshift ALS2 mutants and ALS2 C157Y were both unstable and thus rapidly degraded in cells (16,33). To examine whether other pathogenic missense mutations as well as normal variation affected the stability in ALS2, we performed the cycloheximide (CHX) chase assay using the NSC-34 motor neuron-like hybrid cell line (Fig. 9, A and B). The levels of ALS2 in all constructs exam-ined were significantly decreased by 8 h after the CHX treatment (Table S1). For the normal ALS2 variant, ALS2 P132L showed comparable stability with ALS2 WT . By contrast, 4 h after the CHX treatment, the levels of all pathogenic variants, ALS2 S100I , ALS2 C157Y , and ALS2 G540E , except for ALS2 G49R , were significantly lower than that of ALS2 WT (Fig. 9B), indicating that some, but not all, pathogenic variants were unstable and thus degraded faster than ALS2 WT . It was remarkable that these in vitro results were consistent with those for in silico predictions, in which three mutations, S100I, C157Y, and G540E, exhibited higher mutation energy among others (Table 1).

Pathogenic mutations alter the oligomeric states of ALS2
distribution of ALS2 variants are still unclear. To address this issue, we sought to clarify the relationship between intracellular localization and oligomeric states of several missense and inframe deletion pathogenic ALS2 variants. Here, we demonstrated evidence showing that ALS2 formed a homophilic tetramer in native conditions, and pathogenic mutations might destabilize the protein structure thereby altering the oligomeric states and intracellular behavior of ALS2.
It has been widely accepted that complex formation is one of the fundamentally important molecular features for GEFs in their catalytic exchanging reaction of GDP for GTP on-target small GTPases in response to upstream signals. Among the VPS9 domain-containing Rab5 GEFs, RABGEF1/Rabex-5, one of the well characterized Rab5 GEFs, has been shown to form a heteromeric complex with Rabaptin-5, thereby enhancing the RABGEF1-mediated Rab5 GEF activity and facilitating endosome fusions (38). However, RIN2 (Rab interactor 2), a member of the VPS9 domain containing RIN family proteins, is known to form a homophilic tetramer through the interaction of two anti-parallel RIN2 dimers to each other (39,40). It has also been reported that the activation of small GTPase Rab7 is mediated

Table 1 Protein stability for the wildtype and mutated RLDs of ALS2
When the values of difference in the free energy between the mutated and wildtype structures are Ͼ0.5, they are defined as "destabilizing."

Residue Mutation
Mutation energy Pathogenic mutations alter the oligomeric states of ALS2 Figure 9. Analysis of the stability for ALS2 and its variants by the cycloheximide chase assay. A, Western blot analysis of FLAG-tagged ALS2 and its variants. NSC-34 cells that were transfected with pCIneoFLAG-ALS2 WT , pCIneoFLAG-ALS2 P132L , pCIneoFLAG-ALS2 G49R , pCIneoFLAG-ALS2 S100I , pCIneoFLAG-ALS2 C157Y , or pCIneoFLAG-ALS2 G540E were treated with CHX and harvested at different time points (0, 2, 4, and 8 h). Protein samples were analyzed by Western blotting using anti-ALS2-MORN/VPS9 and anti-␣-tubulin antibodies. Arrowheads and arrows indicate ALS2 (184 kDa) and its degraded product, respectively. B, quantitative analysis of signal intensities for ALS2. Signal intensities were normalized by the level of ␣-tubulin. Value in each variant is presented as a mean (ϮS.D.) percentage relative to the respective signal intensity of ALS2 at 0 h (n ϭ 5). Individual data points are also shown. Statistical significance was evaluated by two-way ANOVA with Bonferroni's multiple comparison test (*, p Ͻ 0.05; **, p Ͻ 0.01).

Pathogenic mutations alter the oligomeric states of ALS2
by the Mon1-Ccz1 heteromeric GEF complex (41). Furthermore, several Rho GEFs, including RasGRF1 (42), RasGRF2 (42), ␤1PIX (43,44), Dbl (45), p115RhoGEF (46), PDZ-RhoGEF (46), and LARG (46), have been shown to form either dimer or oligomer, and such complex formation plays a crucial role in the GEF activities as well as their cellular functions. Consistent with these findings, we here demonstrated that ALS2 could form a tetramer in cells. Because there are no observable stable interactors other than ALS2 itself, the majority of ALS2 might exist as a homophilic tetramer in cells. Notably, the architecture of the ALS2 complex is well matched to that of the RIN2 complex, in which ALS2 tetramers are predicted to be assembled through the DH/PH domain-dependent self-interaction of two dimeric ALS2 complexes, each of which is interacted via the C-terminal MORN/VPS9 region in an antiparallel fashion (35,39). Thus, it is reasonable to assume that tetrameric structure of the ALS2 complex might be crucial not only for the Rab5 activation but also for the intracellular localization as well as the functions of ALS2 (Fig. 10, A and B). Previously, we have demonstrated that two pathogenic missense ALS2 variants, ALS2 C157Y and ALS2 G540E , which retain the Rab5 GEF activity in vitro, can be recruited to membrane ruffles but fail to be relocalized to macropinosomes and/or early endosomes upon Rac1 activation, despite that macropinocytosis itself is activated (34). In this study, we further tested a single normal ALS2 variant and five additional missense or in-frame deletion pathogenic ALS2 variants, and we confirmed that all the pathogenic variants exhibited the same abnormal intracellular behaviors, irrespective of their residue positions of mutation, whereas normal ALS2 variant was normally relocalized to endosomes upon Rac1 activation. Then, we extended the analysis for the genotype (mutation)-phenotype (intracellular behaviors)-structural (complex) relationship (Table S2). Remarkably, the sizes of the ALS2 complex carrying missense mutations in the RLD or in-frame deletion in the central PH domain were shifted toward a higher molecular weight, indicating that these mutants formed extremely large ALS2 complexes in addition to normal tetramer. In contrast, the C-terminal VPS9 domain missense variants, ALS2 R1611W and ALS2 L1617A , were rather present as dimeric or trimeric smaller forms in cells. More importantly, one of the artificial Rab5 GEF-defective mutants, ALS2 P1603A , not only primarily existed as a tetramer but also showed the endosomal localization in the presence of active Rac1, although it could not induce the fusion and enlargement of endosomes due to lack of the Rab5 GEF activity (34). These results suggest that the normal tetrameric structure of ALS2 might be requisite for the Rac1-induced relocalization of ALS2 from membrane ruffles to macropinosomes and/or endosome via macropinocytosis. Thus, a disorganized higher structure of the ALS2 complex is strongly implicated in the loss of cellular functions of ALS2 due to dysregulated intracellular behaviors of ALS2 and not just simply due to loss of Rab5 GEF activity.
One of the questions arising from this study includes the residue position-dependent differences in the oligomeric structures of ALS2. All the missense pathogenic mutations except for R1611W lead to form the larger complexes than normal tetramer, whereas ALS2 carrying the pathogenic mutation R1611W or the artificial mutation L1617A both exist as dimeric and/or trimeric smaller complexes. We have previously shown that the exon 25-deleted variant of ALS2 lacking the region spanning 1280 -1335 aa completely lose the C-terminal regionmediated self-interaction of ALS2 (Table S2) (35). Because the C-terminal MORN/VPS9 region normally mediates the dimerization of ALS2, these particular missense mutations within the VPS9 domain may cause loss of proper dimerization of the C-terminal regions. Even though ALS2 loses the C-terminally-mediated dimerization property, ALS2 itself still can form some complex through the self-interactions between DH/PH and MORN/VPS9 and/or the RLD and MORN/VPS9 region (11,35), resulting in the unusually smaller ALS2 complexes. By contrast, missense and small in-frame mutations in the RLD or PH domain may deregulate the normal assemble for the tetrameric complex of ALS2, thereby forming abnormally larger complexes of ALS2.
Another question arising from this study is how the structural distortion of the ALS2 complex leads to dysregulated intracellular ALS2 localization. Previously, we have demonstrated that two missense pathogenic full-length ALS2 variants, ALS2 C157Y and ALS2 G540E , show a lower affinity to specific PIPs than does WT ALS2 (34). However, we showed here that the RLD fragment carrying the same missense mutations exhibited not only a similar pattern of affinities toward PIPs to those of the wildtype, but also the altered higher complex structure that was not associated with mutation itself. Importantly, in silico structure modeling demonstrated that missense pathogenic mutations in the RLD destabilized the normal protein structure, namely the architecture for a seven-bladed ␤-propeller. Indeed, the CHX chase assay in a motor neuron-like cell line revealed that some missense pathogenic mutations in the RLD destabilized the ALS2 protein. Because it is speculated that the RLD acts as a critical determinant for the intracellular localization of ALS2 (10), the structural distortion in the RLD might cause abnormal protein interactions between ALS2 and other regulators, thereby losing the normal intracellular behavior, in particular, the internalization of ALS2 via macropinocytosis (Fig. 10C).
Another possible explanation for the RLD-mediated regulation of intracellular ALS2 localization may be associated with the disordered region of the RLD. 306 aa of the intercalated disordered region (residues 219 -524), which resides on the bottom surface of blade 5 of ␤-propeller, are predicted to contain several residues modified by phosphorylation, acetylation, and ubiquitylation (47). Furthermore, a database search identified two residues, Ser-483 and Ser-492, representing highly conserved phosphorylation sites (https://www.phosphosite. org/uniprotAccAction?idϭQ96Q42) 3 (47). In fact, it has been reported that three residues in this region of ALS2, i.e. Ser-277, Ser-492, and Thr-510, are phosphorylated (48). We have previously demonstrated that the artificial ALS2 mutant carrying a single amino acid substitution on Thr-510 (ALS2 T510A ) is properly relocalized to macropinosomes upon Rac1 activation (34), indicating the phosphorylation of Thr-510 may not have a major impact on the ALS2 localization. Currently, effects of other phosphorylated residues in the disordered region of the RLD remain to be investigated. Thus, to define the molecular mechanisms underlying the RLD-mediated intracellular localization of ALS2, future detailed studies will be required.
Recent molecular and cell biological studies have gradually been uncovering the molecular functions of ALS2 as well as their regulatory mechanisms: 1) oligomerization of the C-terminal MORN/VPS9 region of ALS2 is necessary for the activation of Rab5 (35); 2) ALS2-mediated Rab5 activation facilitates the fusion between macropinosome and endosomes and/or autophagosomes in cells (49,50); 3) the intact DH/PH domain of ALS2 is indispensable for the redistribution of ALS2 to macropinosomes (11); and 4) active Rac1 induces the localization of ALS2 to membrane ruffles and then to macropinosome and/or endosomes via macropinocytosis (11). In this study, we have further revealed some additional regulatory mechanisms: 5) a tetrameric structure of ALS2 plays an essential role in Rac1-dependent endosomal localization of ALS2, and 6) mutations in the RLD distort the native tetrameric structure of ALS2, thereby losing the ability for the macropinocytosis-mediated internalization of ALS2. Such accumulating evidence allows us  2). Then, ALS2-inherent Rab5 GEF activity can facilitate the fusion of endosomal compartments (indicated as 3). ALS2 WT , normal variant ALS2 P132L , and artificial mutant ALS2 P1603A , all of which can form a tetramer, are redistributed from the cytosol to endosomal compartments upon Rac1 activation and following macropinocytosis, whereas all pathogenic ALS2 mutants and artificial mutant ALS2 L1617A , all of which exist as the structurally-distorted complexes, fail to be loaded onto nascent macropinosomes on macropinocytosis and remain in ruffle membranes. It is noted that endosomal ALS2 P1603A cannot mediate endosome fusion due to the defective Rab5 GEF activity.

Pathogenic mutations alter the oligomeric states of ALS2
to speculate that, under normal conditions, ALS2 acts as follows: Rac1 activation induces the conformational changes in the RLD by certain modifications on some residues, which in turn unlocked the RLD-mediated inhibitory switch; therefore, ALS2 can be recruited to membrane ruffles followed by internalization to macropinosomes via macropinocytosis. Ultimately, ALS2-inherent Rab5 GEF activity can facilitate the heterotypic as well as homotypic fusion of endosomal compartments. In the course of this dynamic signaling pathway, pathogenic ALS2 variants may become unresponsive to Rac1 signaling and/or fail to be loaded onto nascent macropinosomes on macropinocytosis due to the distorted complex structure, and thereby ALS2-mediated signaling might be disrupted (Fig. 10). However, further detailed studies are warranted to obtain direct evidence supporting these notions.
ALS2 is originally identified as a causative gene for autosomal recessive juvenile MNDs. Thus, loss of the ALS2 functions is believed to cause motor neuron degeneration. Concerning its lost functions, it has been shown that ALS2 is implicated in a variety of cellular processes and functions, including endosome fusion (10), macropinocytosis (11,34,49), autophagy (51), amphisome formation (50), neurite extension (49), receptor trafficking (52), mitochondrial morphology in neurons (53), and oxidative stress (53,54). In addition, ALS2 is known to interact with many factors, including Rab5 (10), ALS2 C-terminal-like (ALS2CL) (55), glutamate receptor-interacting protein (GRIP) (52), and Rab17 (56). Most recently, NIMA-related expressed kinase 1 (NEK1), one of the important risk factors in sporadic as well as familial ALS, directly interact with ALS2 (57). Furthermore, molecular function of C9ORF72, a protein product of the most common causative ALS gene, is tightly associated with vesicle trafficking in the endolysosomal pathway (58,59), where ALS2 functions in cells. Deciphering such intricate molecular systems regulated by ALS2 along with other interactors and/or regulators is still a formidable task. However, our data might provide important clues to understand not only the normal function of ALS2 but also molecular mechanisms underlying the pathogenesis of ALS2-linked MNDs.

Dextran uptake assay and line plot profile
HeLa cells were transfected with pCIneoFLAG-ALS2 WT , pCIneoFLAG-ALS2 S100I , pCIneoFLAG-ALS2 A861_T904del , or pCIneoFLAG-ALS2 R1611W together with or without pCMV10-2xHA-Rac1 Q61L as described previously (34). Twenty four hours after transfection, the cells were washed with PBS(Ϫ) and cultured in serum-free Dulbecco's modified Eagle's medium containing 2 mg/ml tetramethylrhodamine-conjugated dextran MW 70,000, lysine-fixable (Invitrogen) for 20 min. After washing with PBS(Ϫ), the cells were fixed and observed by Zeiss LSM-700 laser-scanning confocal microscope (Carl Zeiss). Captured cell images were further subjected to the line-analysis Pathogenic mutations alter the oligomeric states of ALS2 by drawing two cross-sectional lines (each 16 m in length) per macropinosome using RBG profile plot in ImageJ software.

Preparation and purification of FLAG-tagged ALS2 protein
COS-7 cells were transfected with pCIneoFLAG-ALS2 constructs. Twenty four hours after transfection, the cells were washed with PBS(Ϫ) and lysed in buffer A (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Tween 20, and Complete Protease Inhibitor (Roche Diagnostics)). After centrifugation at 23,000 ϫ g for 20 min, supernatants were recovered and subjected to gel-filtration chromatography. Resulting supernatants were further subjected to immunoprecipitation with anti-FLAG M2 affinity gel (Sigma). The affinity gels were washed three times with buffer B (20 mM Tris-HCl (pH 7.5), 150 mM NaCl containing 1% Tween 20). Subsequently, the affinity gels were treated with 100 g/ml 3xFLAG peptide (Sigma) for 30 min at room temperature, and FLAG-tagged ALS2 was eluted. After centrifugation, eluates were subjected to silver staining using Sil-Best Stain One (Nacalai Tesque) according to the manufacturer's instructions.

Coimmunoprecipitation
Coimmunoprecipitation experiments were carried out as described previously (34). In brief, COS-7 cells were transfected with pCIneoFLAG-ALS2 and pICneoHA-ALS2 constructs. Twenty four hours after transfection, the cells were lysed in buffer A (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1% Tween 20, and Complete protease inhibitor (Roche Diagnostics)). Supernatants were recovered by centrifugation at 23,000 ϫ g for 20 min. Immunoprecipitation (IP) was performed by using anti-FLAG M2 affinity gel (Sigma). The FLAG M2 affinity gels were washed with buffer B (20 mM Tris-HCl (pH 7.5), 150 mM NaCl containing 1% Tween 20). IP products were subjected to Western blot analysis with either the anti-FLAG M2 or anti-HA antibody.

Gel-filtration chromatography
COS-7 cells were transfected with pCIneoFLAG-ALS2 constructs. Twenty four hours after transfection, the cells were washed with PBS(Ϫ) and lysed in buffer D (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% IGEPAL CA-630 (Sigma), and Complete protease inhibitor (Roche Diagnostics)). Supernatants were recovered by centrifugation at 23,000 ϫ g for 20 min and loaded onto the column (Superose 6 10/300 GE Healthcare), and the eluates were fractionated and collected. Elution was carried out at 4°C at a flow rate of 0.3 ml/min with a fraction volume of 0.5 ml. Fractionated samples were analyzed by Western blotting using the anti-ALS2-MORN/VPS9 or anti-ALS2-RLD antibody. The elution profile of the column was calibrated with the size standards (IgM, 975 kDa; thyroglobulin, 669 kDa; ferritin, 440 kDa; aldolase, 158 kDa; and ovalbumin, 43 kDa).

Quantification of signal intensities on Western blotting for gel-filtration chromatography
Signal intensities of Western blots were quantified by using CS Analyzer Version 3.0 (ATTO), and each fraction was expressed as a percentage of the total signal intensities from protein bands of interest within the entire fractions.

Protein-lipid overlay assay
PIPs overlay assay of FLAG-fused proteins was conducted according to the manufacturer's instructions. In brief, PIP Strips TM (100 pmol per spot) (Echelon Biosciences) were incubated with 0.2 g/ml purified FLAG-ALS2-RLD WT , FLAG-ALS2-RLD S100I , FLAG-ALS2 P132L , FLAG-ALS2-RLD C157Y , or FLAG-ALS2-RLD G540E in TBST containing 3% (w/v) of fatty acid-free BSA (Sigma) overnight at 4°C. FLAG peptides purified from the cell lysate of empty vector-transfected cells were used as a control. PIP strips TM (Echelon Biosciences, Inc.) were washed and incubated with anti-ALS2-RLD antibody (1:5000) for 2 h at room temperature. After washing with TBST, membranes were incubated with HRP-conjugated anti-rabbit secondary antibody (1:5000) (GE Healthcare) in TBST for 2 h at room temperature. After washing with TBST, immunoreactive protein spots were visualized by Immobilon Western HRP (Merck Millipore) and detected by Ez-Capture MG (ATTO). Signal intensities of PIP strips TM were quantified by using CS Analyzer Version 3.0 (ATTO). Each value was expressed as a

Pathogenic mutations alter the oligomeric states of ALS2
percentage of the total signal intensities from all the spots within the entire strip.

Homology modeling
Amino acid sequence of the RLD of human ALS2 (ALS2-RLD) (GenBank TM accession no. BAB69014) was obtained from NCBI GenBank TM database and used as a target for homology modeling. After BLASTP searches against Protein Data Bank (PDB) sequence entries using ALS2-RLD sequence as a query, the third RLD of human HERC2 (PDB code 3KCI) was selected as the best template for homology modeling (Bit score: 115.2 and Identity: 32%). The 3KCI structure was optimized for inserting missing atoms in incomplete residues, modeling missing loop regions, and deleting alternative conformations by the Prepare Protein protocol. The template and target amino acid sequences were first aligned. Then, the quality of the alignment was improved by removing gaps, particularly 306 aa of the intercalated disordered sequence (residues 219 -524), and realigning the sequence, and we generated 50 structure models of the RLD of ALS2 using a Build Homology Models protocol/ MODELLER (60) in Discovery Studio version 4.1 from BIOVIA (Accelrys Inc. San Diego). The Probability Density Function total energy and Discrete Optimized Potential Energy score were used to select the best model. The loops in the resulting model were refined using the protocol Loop Refinement. The Ramachandran plots and Verify 3D in Discovery Studio were also used for the model evaluation. A CHARMm force field implemented in Discovery Studio was typed for the simulation.

Structural analysis of RCC1-like domain of ALS2
The structural models of ALS2 mutants were built based on optimized models using the Build Mutants protocol of the Macromolecular tool. To optimize the conformation of both the mutated and neighboring residues surrounding mutations, the cut radius (4.5Å) was used in this simulation. Protein stabilities for mutants were calculated using a Calculate Mutation Energy (Stability) protocol. All molecular modeling, optimization, and visualization were performed in Discovery Studio with default parameter setting. The surface hydrophobicity (61) of a residue was calculated in a 3-amino acid window and was drawn using Discovery Studio.

Statistical analysis
Statistical analyses were conducted using Prism 7 (Graph-Pad). Statistical significances for signal intensities on Western blotting for gel-filtration chromatography or PIPs strip membrane for protein lipid overlay assay were evaluated by one-way ANOVA with Bonferroni's multiple comparison tests. Signal intensities on Western blotting for the CHX chase assay were evaluated by two-way ANOVA with Bonferroni's multiple comparison tests. A p value Ͻ0.05 was considered as reaching statistical significance.