TNF receptor–associated factor 6 interacts with ALS-linked misfolded superoxide dismutase 1 and promotes aggregation

Amyotrophic lateral sclerosis (ALS) is a fatal disease, characterized by the selective loss of motor neurons leading to paralysis. Mutations in the gene encoding superoxide dismutase 1 (SOD1) are the second most common cause of familial ALS, and considerable evidence suggests that these mutations result in an increase in toxicity due to protein misfolding. We previously demonstrated in the SOD1G93A rat model that misfolded SOD1 exists as distinct conformers and forms deposits on mitochondrial subpopulations. Here, using SOD1G93A rats and conformation-restricted antibodies specific for misfolded SOD1 (B8H10 and AMF7-63), we identified the interactomes of the mitochondrial pools of misfolded SOD1. This strategy identified binding proteins that uniquely interacted with either AMF7-63 or B8H10-reactive SOD1 conformers as well as a high proportion of interactors common to both conformers. Of this latter set, we identified the E3 ubiquitin ligase TNF receptor–associated factor 6 (TRAF6) as a SOD1 interactor, and we determined that exposure of the SOD1 functional loops facilitates this interaction. Of note, this conformational change was not universally fulfilled by all SOD1 variants and differentiated TRAF6 interacting from TRAF6 noninteracting SOD1 variants. Functionally, TRAF6 stimulated polyubiquitination and aggregation of the interacting SOD1 variants. TRAF6 E3 ubiquitin ligase activity was required for the former but was dispensable for the latter, indicating that TRAF6-mediated polyubiquitination and aggregation of the SOD1 variants are independent events. We propose that the interaction between misfolded SOD1 and TRAF6 may be relevant to the etiology of ALS.

genetic cause of ALS to date, accounting for about 12% of familial cases (fALS) and 1% of sporadic ALS cases (2). Over 180 different mutations in SOD1 have thus far been reported as pathogenic (RRID:SCR_018075). These mostly comprise missense mutations in coding regions, but rare nonsense mutations, deletions, insertions, and mutations in noncoding regions have also been reported (3). Mutations in the coding region are found in all five exons of SOD1. It is now agreed that SOD1-mediated ALS rises from a toxic gain-of-function due to mutation-induced changes in the folding of the SOD1 protein, collectively referred to as misfolded SOD1.
Misfolded SOD1 exposes certain structural features, normally buried in natively folded SOD1, that have been exploited for the generation of antibodies that detect misfolded SOD1 on a conformation-restricted basis (4,5). Among those, the mouse monoclonal antibody B8H10 recognizes an epitope within the metal-binding loop (loop IV, within aa. , the exposure of which is associated with a deficiency in metal co-factor binding (6 -8). The rabbit monoclonal AMF7-63 is the high-affinity descendant of the mouse monoclonal DSE2 and was raised against residues in the electrostatic loop (loop VII, aa. 125-142), the exposure of which is associated with increased conformational flexibility in this region (9,10). Strategies to target and thus eliminate misfolded SOD1 in transgenic rodent models, using either active or passive immunization approaches, show benefit in delaying disease phenotypes (6,8,(11)(12)(13)(14)(15). Whether or not a common target exists for misfolded SOD1 remains unknown. Using B8H10 and AMF7-63, we have previously demonstrated in the SOD1 G93A rat model of ALS that different conformers of misfolded SOD1 deposit on the surface of spinal cord mitochondria in an age-dependent manner, supporting the view that distinct misfolded SOD1 conformers exist (10, 16). Interestingly, the accumulation of both conformers positively correlates with mitochondrial damage (10, 16,17).
Here, we report on the identification of the E3 ubiquitin ligase TNF receptor-associated factor 6 (TRAF6) as a novel binding partner of misfolded SOD1. In cellulo, we show that the interaction with TRAF6 is unique to mutant but not WT SOD1, and we map the interaction to the C terminus of TRAF6. However, not all SOD1 variants interact equally with TRAF6. We show that TRAF6-interacting and noninteracting SOD1 variants are conformationally distinct in terms of their exposure of misfolding-associated epitopes recognized by B8H10 and AMF7-63. We demonstrate that TRAF6 stimulates mutant SOD1 polyubiquitination and aggregation but that these variably depend on TRAF6 E3 ubiquitin ligase activity and are therefore independent events. Specifically, siRNA-mediated reduction of TRAF6 levels reduces mutant SOD1 aggregation in a cellular model. Importantly, we uncover that TRAF6 localization is not restricted to the cytoplasm, with a pool of TRAF6 being constitutively localized to spinal cord mitochondria. Finally, TRAF6 is expressed in the cell types harboring misfolded SOD1 in vivo, establishing relevance to misfolded SOD1-linked pathology in ALS. Taken together, these results suggest that misfolded SOD1 interacts with a mitochondriallylocalized pool of TRAF6, and this interaction may underlie misfolded SOD1 accumulation.

Identification of novel binding partners of mitochondriaassociated misfolded SOD1
In mutant SOD1 rodent models of ALS, a portion of misfolded SOD1 conformers is localized to the surface of spinal cord mitochondria (10, 16 -18). To identify the interactomes of specific misfolded SOD1 conformers at the mitochondria, we performed immunoprecipitation coupled with MS (IP-MS) of spinal cord mitochondria isolated from symptomatic SOD1 G93A rats (Fig. 1A). Briefly, misfolded SOD1 was immunoprecipitated with anti-B8H10 or anti-AMF7-63, and eluates were analyzed by MS. As published previously, both B8H10 and AMF7-63 specifically detect misfolded forms of SOD1 but are unreactive for natively-folded SOD1 WT (8 -10, 16, 17). From all identified prey proteins, only those with a maximum peptide score of Ն40 in three of three biological replicates and not detected in IgG controls were considered as potential binding partners. Identified proteins are listed in Table S1. Human SOD1 was detected in all samples and served as an internal control. Also, known binding partners of SOD1 (e.g. the copper chaperone for SOD1 (CCS) (19)) and misfolded SOD1 (e.g. voltage-gated anion channel 1 (VDAC1) (20)) were detected, which additionally supports the quality and validity of our dataset. Of the 52 proteins identified, three proteins (6%) interacted uniquely with B8H10-reactive SOD1 conformers and seven uniquely interacted with AMF7-63-reactive conformers (13%) ( Fig. 1B and Table S2). Thus, the majority of interactors (81%) were common to B8H10-and AMF7-63-reactive conformers of misfolded SOD1.

Presence of ubiquitin in AMF7-63 and B8H10 immunoprecipitates
Ubiquitin is highly conserved and encoded by four different genes: UBB, UBC, UBA52, and RPS27A. Although UBB and UBC encode polyubiquitin precursors made up of multiple

Misfolded SOD1 binds the ubiquitin ligase TRAF6
head-to-tail repeats of ubiquitin, UBA52 and RPS27A are ubiquitin hybrid genes encoding a fusion protein composed of a single N-terminal ubiquitin moiety fused to a C-terminal extension. In reviewing our misfolded SOD1 interactomes, we noticed that ubiquitin-60S ribosomal protein L40 (RL40), which is encoded by UBA52, was a binding partner detected in both B8H10 and AMF7-63 immunoprecipitates (Table S1 and Fig. S1A). To elucidate whether the peptides listed under RL40 correspond to the N-terminal ubiquitin moiety or the C-termi-nal extension of the protein, we generated a sequence alignment for rat RL40 with ubiquitin (single moiety) (Fig. S1B). All three peptides listed under the identifier RL40 mapped to the N-terminal ubiquitin moiety and not the C-terminal extension of the protein. This indicates that the detected peptides are not unique for RL40, but they generally correspond to ubiquitin and may thus derive from RL40, RS27A, UBB, or UBC proteins. To illustrate this, we also aligned rat RS27A (Fig. S1B). To investigate whether ubiquitin is indeed associated with B8H10- Figure 1. TRAF6 is a novel binding partner of B8H10 and AMF7-63-reactive misfolded SOD1 at the mitochondria. A, workflow for immunoprecipitationbased MS proteomics for the identification of binding partners of mitochondria-associated misfolded SOD1 in the SOD1 G93A rat model. Misfolded SOD1 was immunoprecipitated with the misfolded SOD1-specific conformation-restricted antibodies B8H10 and AMF7-63 from SOD1 G93A rat spinal cord mitochondria and binding partners identified by nanoLC-MS/MS. B, Venn diagram of identified misfolded SOD1-binding partners common to both B8H10 and AMF7-63 conformers or selective for either. C, novel binding-partner TRAF6 (UniProt ID: B5DF45) was identified based on six unique peptides covering 17% of the protein sequence. D and E, Flag-tagged SOD1 (WT or mutants) and Myc-tagged TRAF6 (WT or deletion mutants) were co-expressed in 293FT cells. Co-immunoprecipitations were performed on Flag-SOD1 or reciprocally on Myc-TRAF6, and co-precipitation of either binding partner was analyzed by immunoblotting. Immunoblotting for Flag-SOD1 (bait) is to demonstrate equal IP efficiency across conditions. Mock refers to transfection with an equivalent amount of empty vector. Note, the slower migrating band in TRAF6 lanes is auto-ubiquitinated TRAF6. Whole-cell lysates (WCL) were loaded to demonstrate equal plasmid expression. Actin serves as loading control. Data are representative of at least three independent experiments. and AMF7-63-reactive misfolded SOD1, we precipitated misfolded SOD1 from SOD1 G93A rat spinal mitochondria and immunoblotted for ubiquitin (Fig. S1C). In both B8H10-and AMF7-63-immunoprecipitates, ubiquitin was detected as a high-molecular-weight smear refusing entry into the separating gel, a pattern typically observed with intensely polyubiquitin-modified proteins. Taken together, these data suggest that B8H10 and AMF7-63 conformers are modified with ubiquitin (or polyubiquitin) or associate with stable binding partners that are themselves ubiquitinated.

Identification of the E3 ubiquitin ligase TRAF6 as a novel binding partner of misfolded SOD1
In our dataset, we detected six unique peptides for an E3 ubiquitin ligase: TNF receptor-associated factor 6 (TRAF6). TRAF6 peptides, covering 17% of the protein sequence, were detected in both B8H10 and AMF7-63 immunoprecipitates (Table S1, Fig. 1C, and Fig. S2).
TRAF6 is a cytoplasmic-localized RING-type ubiquitin ligase classically known for its role in the canonical NF-B-signaling pathway, where it participates in the transduction of signals downstream of a multitude of immunoregulatory receptors (23)(24)(25)(26) and functions as a hub protein in the cross-talk of immune-regulatory pathways with the autophagy machinery (27)(28)(29)(30). TRAF6 is not strictly localized to the cytoplasm but translocates to mitochondria in certain contexts (31)(32)(33)(34)(35). Importantly, TRAF6 has been previously implicated in neurodegenerative diseases, including Parkinson's disease (36 -38), Huntington's disease (39), and Alzheimer's disease (40 -42), where it stimulates the ubiquitination of disease-associated proteins and exerts a regulatory role on their turnover and aggregation (43). The detection of TRAF6 as a binding partner of misfolded SOD1 was additionally intriguing, because physiological interactions between E3 ubiquitin ligases and their substrates are generally known to be weak, transient, and not readily detected by IP-MS-based techniques (44 -46). We speculated that the interaction is unnaturally stable and of possible pathological relevance. Thus, we elected to more fully investigate the interaction between mutant/misfolded SOD1 and TRAF6.

TRAF6 interacts with mutant but not WT SOD1
To characterize the interaction between mutant SOD1 and TRAF6, we established an in cellulo system based on transient co-expression of Flag-tagged SOD1 and Myc-tagged TRAF6 in human 293FT cells. To elucidate whether this system can recapitulate our in vivo results, we performed reciprocal co-immunoprecipitations. SOD1 G93A and SOD1 G85R , but not SOD1 WT , co-precipitated TRAF6 WT (Fig. 1D, lanes 5-7). Reversely, TRAF6 WT co-precipitated mutant SOD1, but not SOD1 WT (Fig. 1E, lanes 5-7), thus confirming our IP-MS results. Note, the slower-migrating TRAF6 band is the auto-ubiquitinated form, consistent with previous work (47). In addition, these data indicate that TRAF6 selectively interacts with mutant but not WT SOD1, indicating that the interaction with TRAF6 is unique to nonnative SOD1 conformers.

Mutant SOD1 binds the TRAF6 C terminus
TRAF6 is a multidomain protein and is highly conserved across mammalian species (48,49). The N terminus features a really interesting new gene domain (RING) with a series of zinc finger motifs, which are connected via an ␣-helical coiled-coil motif to a C-terminal meprin and TRAF homology (MATH) domain (Fig. 2, A and B). The RING domain and adjacent zinc finger motifs are involved in E2 ubiquitin conjugase binding and are required for TRAF6 function as an E3 ubiquitin ligase (50). The MATH domain is an eight-stranded anti-parallel ␤-sandwich with which TRAF6 engages in a multitude of protein-protein interactions (32,(51)(52)(53). To determine the domain of TRAF6 that interacts with mutant SOD1, two TRAF6 deletion constructs were generated: TRAF6 ⌬C (aa.  comprising the N-terminal RING domain and zinc finger motifs, and TRAF6 ⌬N (aa. 289 -522) comprising the central coiled-coil motif and C-terminal MATH domain (Fig. 2, A and  B). SOD1 G93A and SOD1 G85R efficiently co-precipitated TRAF6 ⌬N but not TRAF6 ⌬C (Fig. 2C, compare lanes 14 and 15 to lanes 11 and 12). SOD1 WT did not interact with TRAF6 WT nor the two TRAF6 deletion mutants (Fig. 2C, lanes 7, 10, and 13). We conclude that the C terminus of TRAF6 is necessary and sufficient for mutant SOD1 binding.
Consistent with the view that more than one misfolded SOD1 conformer exists (10, 54), these data suggest that many SOD1 mutant proteins adopt a misfolded conformation(s) that is (are) favorable for TRAF6 binding, but that these are not conformationally identical, which may account for the observed variable interaction with TRAF6. Interestingly, one ALS-causing mutant, SOD1 V148I , did not interact with TRAF6, consistent with previous data indicating that this mutant is WT-like in terms of stability and structure (55)(56)(57). Strikingly, the substitution at Val-148 with glycine is sufficient for mutant SOD1 to adopt a TRAF6-interacting conformation.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
Finally, the variation in mutant SOD1 binding was observed for TRAF6 WT but not for TRAF6 ⌬N . SOD1 A4V and SOD1 V148G co-precipitated TRAF6 WT to variable amounts but TRAF6 ⌬N to similar amounts (Fig. 3C, compare lanes 6 and 7 to lanes 9 and 10). Although the TRAF6 C terminus is required for mutant SOD1 binding, these data suggest that the N terminus of TRAF6 modulates the interaction with mutant SOD1.

SOD1 misfolding-associated functional loop exposure is required for TRAF6 interaction
Next, we sought to investigate the relevance of SOD1 misfolding-associated exposure of the metal-binding loop (loop IV) and electrostatic loop (loop VII) to TRAF6 binding. For this, Flag-SOD1 variants expressed in 293FT cells were evaluated and the C-terminal MATH domain. B, protein structure models for TRAF6 WT , TRAF6 ⌬C , and TRAF6 ⌬N . C, Flag-tagged SOD1 (WT or mutants) and Myc-tagged TRAF6 (WT or deletion mutants) were co-expressed in 293FT cells. Co-immunoprecipitations were performed on Flag-SOD1, and Myc-TRAF6 co-precipitation was assessed by immunoblotting. Immunoblotting for Flag-SOD1 (bait) is to demonstrate equal IP efficiency across conditions. Mock refers to transfection with an equivalent amount of empty vector. WCL were loaded to demonstrate equal plasmid expression. Actin serves as loading control. Data are representative of at least three independent experiments.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
for reactivity to the conformation-restricted misfolded SOD1 antibodies B8H10 (epitope in loop IV) or AMF7-63 (epitope in loop VII) by immunoprecipitation (Fig. 4, A and B). We incorporated the C-terminal truncation mutant SOD1 G127X as a negative control for AMF7-63 immunoprecipitations, as this mutant lacks the electrostatic loop (loop VII). Although most SOD1 variants immunoprecipitated with both B8H10 and AMF7-63, SOD1 V148I was not immunoprecipitated with either antibody (Fig. 4B, compare lanes 3-9 to lane 10). SOD1 G127X was not immunoprecipitated with AMF7-63 (as expected), but was with B8H10 (Fig. 4B, lane 11). As expected, neither antibody immunoprecipitated SOD1 WT (Fig. 4B, lane 2). To summarize, all TRAF6-interacting SOD1 variants were reactive for either B8H10 and/or AMF7-63. In contrast, SOD1 WT and SOD1 V148I , which both do not interact with TRAF6, were unreactive for both antibodies suggesting that SOD1 misfoldingassociated exposure of the functional loops or monomerization/ unfolding of SOD1 are conformational requirements for TRAF6

Figure 3. SOD1 mutation-dependent variation in the interaction with TRAF6 in cellulo.
A-C, Flag-tagged SOD1 (WT or mutants) and Myc-tagged TRAF6 (WT or ⌬N) were co-expressed in 293FT cells. Co-immunoprecipitations were performed on Flag-SOD1, and Myc-TRAF6 co-precipitation was analyzed by immunoblot. Immunoprecipitates were immunoblotted for Flag-SOD1 to demonstrate equal IP efficiency across conditions. Mock refers to transfection with an equivalent amount of empty vector. WCLs were loaded to demonstrate equal plasmid expression. Actin serves as loading control. B, densitometric analysis of the co-immunoprecipitated amount of Myc-TRAF6 relative to the amount of immunoprecipitated Flag-SOD1. Plotted values are the mean Ϯ S.D. of four independent experiments. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; and ****, p Ͻ 0.0001.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
interaction. Finally, data obtained with SOD1 G127X indicate that exposure of the metal-binding loop is sufficient for TRAF6 binding.

TRAF6 stimulates mutant SOD1 ubiquitination
TRAF6 is reported to stimulate the ubiquitination and aggregation of mutant/misfolded proteins in the context of other neurodegenerative diseases (36,39,42). To determine whether mutant SOD1 is also a substrate of TRAF6, we performed in cellulo ubiquitination assays (Fig. 5A). For this, we co-expressed Flag-SOD1 and Myc-TRAF6 WT or E3 ubiquitin ligase-inactive mutant TRAF6 C70A (58) with HA-ubiquitin, immunoprecipitated HA-ubiquitin from whole-cell lysates, and immunoblotted for Flag-SOD1. In this assay, we selected a SOD1 variant that demonstrates robust interaction with TRAF6 and is dually reactive for B8H10 and AMF7-63 (SOD1 A4V ), a SOD1 variant that exhibits a slightly less robust interaction with TRAF6 and is also reactive for B8H10 and AMF7-63) (SOD1 V148G ), and a variant that does not interact with TRAF6 and does not display B8H10-or AMF7-63-reactive SOD1 epitopes (SOD1 V148I ). TRAF6 strongly stimulated the polyubiquitination of SOD1 A4V and SOD1 V148G , but not SOD1 V148I (Fig. 5A, compare lanes 5-7). The polyubiquitin signature on SOD1 A4V and SOD1 V148G was absent in cells co-expressing TRAF6 C70A (Fig. 5A, compare lanes 5 and 6 to 8 and 9). These data indicate that TRAF6 can stimulate the polyubiquitination of mutant SOD1 in an E3 ubiquitin ligase activity-dependent manner. Moreover, this effect is selective to TRAF6-interacting SOD1 variants, and the level of detected polyubiquitin-modified mutant SOD1 positively correlates with the observed differences in the TRAF6 -mutant SOD1 interaction.
Next, to determine the fate of mutant SOD1 with respect to TRAF6-mediated polyubiquitination, we performed ubiquitination assays with single-lysine ubiquitin mutants to elucidate the polyubiquitin chain linkages with which mutant SOD1 is modified (Fig. 5B). Single-lysine ubiquitin mutants have only one of seven lysines available (all other lysines are mutated to arginine) and thus can only be used to catalyze polyubiquitin chains linked through one particular lysine residue. We performed these assays with Flag-SOD1 A4V as we determined that this mutant had the strongest interaction with TRAF6 and was the most ubiquitinated. Our data suggest that TRAF6 promotes the polyubiquitination of mutant SOD1, with Lys-6, Lys-27, and Lys-29 linkages prevailing (Fig. 5B, lanes 3, 4, and 6).

TRAF6 is a modifier of mutant SOD1 aggregation
To address whether mutant SOD1 aggregation could be causally linked to the interaction with TRAF6, we performed filter retardation assays using 293FT cells expressing Flag-SOD1 (WT or mutants), either following siRNA-mediated knockdown of TRAF6 (Fig. 6A) or co-expression of Myc-TRAF6 (WT, ⌬N, or C70A) (Fig. 6, C and E). Cell lysates were filtered through cellulose acetate membrane filters, which trap protein aggregates with a diameter of Ն200 nm, and immunoblotted for Flag-SOD1. Western blottings of cell lysates were included to verify the efficiency of TRAF6 knockdown, to demonstrate equal plasmid expression across conditions, and to provide an indirect loading control for the filter membranes ( Fig. 6, B, D, and F). Note, cell viability was observed to be equivalent across conditions. Flag-SOD1 variants tested in the assays were as follows: SOD1 A4V (high-aggregation propensity and strong TRAF6 interaction); SOD1 G93A and SOD1 E100G (high-aggregation propensity and less strong TRAF6 interaction). In experiments with TRAF6 knockdown, all SOD1 variants aggregated less in cells treated with two independent siRNAs for TRAF6 compared with siControl-treated cells (Fig.  6A, compare rows 2 and 3 to row 1). TRAF6 protein levels were effectively reduced with both siRNAs (Fig. 6B). In contrast, coexpression of TRAF6 WT or TRAF6 ⌬N stimulated the aggregation of mutant SOD1 (Fig. 6C, columns 2-4 compare row 1 to rows 2 and 3), but not SOD1 WT (Fig. 6C, column 5). Taken together, these data indicate that lowering TRAF6 levels alleviates mutant SOD1 aggregation, whereas the interaction of mutant SOD1 with the TRAF6 C terminus is sufficient to exacerbate mutant SOD1 aggregation. This is consistent with the nondegradative ubiquitin linkage types observed.
Our results with the TRAF6 N terminus harboring the RING domain, which is required for ubiquitin ligase activity, suggested that the E3 ligase activity may not be implicated in SOD1 aggregation. To directly test this, we co-expressed mutant SOD1 with ubiquitin ligase-inactive TRAF6 C70A . TRAF6 C70A stimulated SOD1 A4V aggregation comparable with TRAF6 WT (Fig. 6E, columns 2-4 compare row 2 to row 3), confirming that TRAF6 ubiquitin ligase activity is indeed not required for . Exposure of epitopes associated with SOD1 misfolding discerns TRAF6-interacting from noninteracting mutants. A, schematic representation of Flag-SOD1 constructs, with the location of mutation sites relative to B8H10 (loop IV) and AMF7-63 (loop VII) antibody epitopes. SOD1 G127X is a truncation mutant with a nonnative C terminus and inherently lacks the AMF7-63 epitope. B, Flag-tagged SOD1 (WT or mutants) was expressed in 293FT cells. Misfolded SOD1 was immunoprecipitated with the B8H10 or AMF7-63 antibody, and precipitates were analyzed by Western blotting. Mock refers to transfection with an equivalent amount of empty vector. WCLs were loaded to demonstrate equal plasmid expression. Actin serves as loading control. Data are representative of three independent experiments.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
TRAF6-mediated mutant SOD1 aggregation. A similar result was obtained for SOD1 G37R (data not shown). Finally, we examined whether TRAF6 C70A -stimulated aggregation of mutant SOD1 correlates with a bimolecular interaction between TRAF6 and mutant SOD1. We observed that mutant SOD1 interacted comparably with TRAF6 C70A and TRAF6 WT (Fig.  S3, compare lanes 7 and 8 to 10 and 11). Taken together, these data indicate that TRAF6 ubiquitin ligase activity is not required for TRAF6 interaction with mutant SOD1 nor TRAF6's capacity to stimulate mutant SOD1 aggregation. We conclude that the interaction of mutant SOD1 with TRAF6 is sufficient to modulate mutant SOD1 aggregation.

Mitochondrial recruitment of TRAF6 in SOD1 G93A rats
A pool of mutant SOD1 exists in an aggregated state on the surface of spinal cord mitochondria in SOD1 G85R mice and SOD1 G93A rats (18). In addition, we have demonstrated that B8H10 and AMF7-63-reactive misfolded SOD1 accumulate in mitochondrial fractions isolated from SOD1 G93A rat spinal cords and correlate with disease progression (10). We speculated that these observations could be linked to misfolded SOD1 interacting with TRAF6. TRAF6 is known as a cytoplasmic-localized protein, but its capacity to translocate to mitochondria in certain contexts has been previously reported (31,33,34). Because we detected TRAF6 as a binding partner of misfolded SOD1 at the mitochondria, we reasoned that TRAF6 must be recruited. To experimentally address this, we examined TRAF6 levels in spinal cord lysates and purified floated mitochondria from SOD1 G93A rats (Fig. S4). We collected samples at two different disease stages: 10 weeks (pre-symptomatic) and 14 weeks (pre-symptomatic/prior to disease onset). These two age groups were chosen for comparison because misfolded SOD1 is not readily detectable in spinal cord mitochondria in 10-week-old SOD1 G93A rats but is at 14 weeks (10). Age- Figure 5. TRAF6-mediated polyubiquitination of mutant SOD1 depends on its ubiquitin ligase activity and is selective to TRAF6-interacting SOD1 variants. A, in cellulo ubiquitination assays with mutant SOD1 as substrate and TRAF6 as catalyzing E3 ubiquitin ligase. Flag-SOD1 A4V and SOD1 V148G (TRAF6interacting) or Flag-SOD1 V148I (TRAF6-noninteracting) was co-expressed with Myc-TRAF6 WT or ubiquitin ligase-inactive TRAF6 C70A in the presence of HAubiquitin WT in 293FT cells. Immunoprecipitations were performed on HA-ubiquitin, and immunoprecipitates were immunoblotted for Flag-SOD1 and HAubiquitin to demonstrate equal IP efficiency across conditions. B, in cellulo ubiquitin-linkage assays were performed similar to A, but in the presence of HA-ubiquitin WT or single-lysine mutants. TRAF6 mediates mutant SOD1 polyubiquitination with primarily Lys-6, Lys-27, and Lys-29 linkages. Note that no polyubiquitinated mutant SOD1 was recovered in the 4% stacking gel (A) and therefore was removed prior to transfer in subsequent experiments. Mock refers to transfection with an equivalent amount of empty vector. WCLs were loaded to demonstrate plasmid expression. Actin serves as loading control. Data are representative of three independent experiments.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
matched nontransgenic and SOD1 WT -expressing rats were included as controls. Collectively, no differences in TRAF6 protein levels were observed in SOD1 G93A rat spinal cord lysates compared with controls and between-age groups (Fig. S4A) nor was there a difference observed in the level of TRAF6 in mitochondrial fractions compared with controls and between-age groups (Fig. S4B). Surprisingly, however, we observed TRAF6 in mitochondrial fractions from nontransgenic animals, indicating that a pool of TRAF6 is constitutively localized to spinal cord mitochondria. These data suggest that misfolded SOD1 interacts with a mitochondrial-localized pool of TRAF6 rather than requiring TRAF6 to be recruited.
Finally, we noticed that the TRAF6 (EP591Y) antibody we used to detect TRAF6 at 60 kDa, corresponding to its reported size, additionally detected two faster-migrating bands at ϳ48 and ϳ35 kDa (both unreported previously) enriched in spinal cord mitochondrial fractions compared with lysates (Figs. S4 and S5C). Using a range of tools and approaches, we validated that all three detected bands are specific to this antibody (Fig.  S5).

Effect of mutant SOD1 on TRAF6-stimulated NF-B activation
Because the constitutive localization of TRAF6 at the mitochondria is unreported, the function of TRAF6 as a resident mitochondrial protein is unclear. Thus, we further investigated the functional repercussions of mutant/misfolded SOD1 binding to TRAF6 using a proof-of-principle approach. TRAF6 plays a central role in cytoplasmic signal transduction from a multitude of immune receptors in response to stimuli (induced NF-B signaling) but is also crucial for cellular homeostasis Figure 6. TRAF6 modifies mutant SOD1 aggregation but independently of its ubiquitin ligase activity. A, C, and E, filter retardation assays in which protein aggregates Ն200 nm in diameter are captured on cellulose acetate membrane filters. Flag-SOD1 (WT or mutants) was either co-expressed with Myc-TRAF6 WT , TRAF6 ⌬N , or ubiquitin ligase-inactive TRAF6 C70A , or Flag-SOD1 variants were expressed in 293FT cells that received siTRAF6 or siControl. Filter membranes were immunoblotted for Flag-SOD1. B, D, and F, Western blotting of cell lysates to demonstrate equal plasmid expression and to verify TRAF6 knockdown across conditions. Immunoblots also serve as loading controls for the filter membranes. Mock refers to transfection with an equivalent amount of empty vector. Actin serves as loading control. All data are representative of three independent experiments.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
through maintaining constitutive NF-B activity (basal NF-B signaling) (59). Constitutive NF-B activity in the mature CNS is restricted to neurons and is required for neuronal survival (60,61). Glial cells have very low levels of constitutive NF-B activity, but activity is strongly induced in inflammatory contexts (62). An up-regulation of NF-B activity is seen at late stages of disease in mutant SOD1 transgenic mice and primarily in glial cells, which is associated with neuroinflammatory processes (63,64). Thus, we reasoned that mutant SOD1 binding to TRAF6 may have repercussions on TRAF6's capacity to transduce signals in the NF-B pathway. To experimentally address this, we performed dual NF-B luciferase reporter assays in cells expressing Myc-TRAF6 and an escalating dose of Flag-SOD1 WT or Flag-SOD1 A4V , together with a firefly luciferase NF-B reporter and a Renilla luciferase internal control (Fig.  S6). The amount of transfected Myc-TRAF6 plasmid was previously determined by dose titration and was chosen to yield ϳ50% NF-B activation relative to maximal TRAF6-stimulated activation, thus allowing for the detection of decreased or increased NF-B activity. SOD1 A4V was chosen because it most strongly interacted with TRAF6 in other experiments, and SOD1 WT served as TRAF6-noninteracting control. Flag-p65/ RelA was incorporated as positive control. The sole expression of Myc-TRAF6 increased NF-B activity by 7-fold relative to basal level (as expected), but mutant SOD1 did not have a significant effect on this across the entire titration range. This suggests that in this cellular context, mutant SOD1 binding does not adversely affect TRAF6 in activating NF-B and that the signaling transduction cascade remained intact.

TRAF6 is expressed in the cell types affected in ALS
We and others have previously demonstrated in mutant SOD1 transgenic rodent models that misfolded SOD1 is primarily detected in motor neurons at disease end-stage and rarely in glial cells (8,10,16,(65)(66)(67). To assess whether TRAF6 is expressed in these cell types, we attempted to detect TRAF6 expression by immunohistochemistry in SOD1 G93A rat spinal cords. Unfortunately, none of the four commercial antibodies we tried gave specific labeling in rats, mice, or in cultured cells. As an alternative, we immunoblotted whole-cell lysates prepared from human iPSC-derived motor neurons (Fig. 7A and   Fig. S7) and mouse primary cortical neurons and astrocytes (Fig. 7B). TRAF6 was detected in all cell types, but we observed a difference in the ratio of the 60-kDa TRAF6 and the 48-kDa band in mouse cortical astrocytes compared with neurons. Densitometry revealed that 60-kDa TRAF6 was 3-fold more abundant in neurons than in astrocytes, whereas the 48-kDa band was 0.7-fold less abundant in neurons than in astrocytes (Fig. 7C). Moreover, in neurons the 60-kDa TRAF6 was 6.5-fold more abundant than the 48-kDa band and 1.4-fold in astrocytes. These data indicate that TRAF6 is expressed in the cell types that primarily harbor misfolded SOD1 in vivo. Thus, we explored the interaction between misfolded SOD1 and TRAF6 in post-mortem tissues from SOD1 mutation-carriers affected by ALS. Although misfolded SOD1 was detected in the motor cortex of three independent SOD1 A4V cases (data not shown), we were unable to detect TRAF6 in misfolded SOD1 immunoprecipitates, possibly because of the low overall abundance of TRAF6 in CNS tissues (68).

Discussion
In this study, we conducted IP-MS proteomics using the SOD1 G93A rat model to identify the interactomes of misfolded SOD1 conformers at the mitochondria. Using antibodies reported to recognize distinct misfolded SOD1 conformers, we found a high degree of overlap for interactors (81%). Gene ontology analyses revealed a convergence on cellular responses related to unfolded proteins and protein ubiquitination. It is noteworthy that a number of interactors identified here are chaperone family members, which is in agreement with studies performed by others in different SOD1 models (69 -72). In addition, we confirmed the previously reported interaction between human misfolded SOD1 and VDAC1 in rat spinal cord mitochondria (20). We also report that both misfolded SOD1 conformers interact with CCS, which may be expected given the known role of CCS in SOD1 maturation (73). However, the CCS-SOD1 heterodimer is an obligatorily transient complex (74). Thus, the detection of CCS in complex with misfolded SOD1 proteins here could reflect a sustained interaction that disrupts the available pool of CCS, thus impeding SOD1 maturation, as suggested by others (74). This could effectively contribute to a feed-forward loop of SOD1 misfolding. However,

Misfolded SOD1 binds the ubiquitin ligase TRAF6
CCS has been previously ruled out as a modifier of the development of ALS-related phenotypes in transgenic SOD1 mice (75). Perhaps this concept should be revisited, given the recent data in monozygotic siblings discordant for ALS where CCS is found to be down-regulated in the affected twin (76).
Interestingly, a few binding partners on our proteomics list were also selectively detected in B8H10 but not AMF7-63 IPs and vice versa (Table S2). Although the significance of these conformer-selective binding partners remains to be investigated, these data support the possibility that different misfolded conformers of SOD1 may have one or more distinct interactome-dependent pathomechanisms. One binding partner that was uniquely identified in B8H10 immunoprecipitates was 2-oxoglutarate dehydrogenase (OGDH), which has been described as an endogenous protein processed for mitochondrial antigen presentation in certain conditions (77,78). This raises the question whether misfolded SOD1 is itself presented on mitochondrial-derived vesicles, similar to OGDH, or whether misfolded SOD1 may interfere with OGDH antigen presentation.
Another binding partner uniquely identified in B8H10 immunoprecipitates was ATP synthase subunit ␤ (ATP5B), which has been previously reported as an interactor of apoSOD1 (metal co-factor deficient) in rat spinal cords (70). Consistent with this, B8H10 immunoprecipitates the metaldeficient mutants H46R, G85R, and G127X comparably in our in cellulo experiments, and we have previously shown that B8H10 preferentially recognizes apoSOD1 (10).
Interestingly, another ATP synthase subunit, ATP synthase F 0 complex subunit ␤1 (ATP5F1), was uniquely recovered in AMF7-63 immunoprecipitates. Thus, the misfolded SOD1 conformers examined here each bound a different portion of the F 1 F 0 -ATP synthetase complex that drives mitochondrial ATP production. Defective mitochondrial ATP production has been previously reported in mutant SOD1-mediated disease and could positively contribute to compromised mitochondrial axonal trafficking (79).
We identified the E3 ubiquitin ligase TRAF6 as a novel binding partner of mitochondria-associated misfolded SOD1. We showed that SOD1 misfolding-associated exposure of the metal-binding loop and electrostatic loop in mutant SOD1 is a contributing factor for interaction with TRAF6. To this end, Garg et al. 5 have performed in silico molecular dynamics simulations to further investigate the biophysical nature of the intermolecular interaction between mutant/misfolded SOD1 and TRAF6. It is our view, which is supported by this in silico work and our experimental evidence with SOD1 G127X , that the electrostatic loop is not required for TRAF6 binding, because it is absent in this truncation mutant. This leads us to conclude that the interface for SOD1 binding to TRAF6 is formed primarily by residues located in the metal-binding loop, with a secondary contribution of residues in the electrostatic loop, as well as a potential co-option of the native SOD1 dimer interface. Interestingly, the conformational criterion for TRAF6 interaction is not universally fulfilled by all SOD1 variants. SOD1 V148I , which is WT-like with respect to stability and structure (55)(56)(57), does not adopt a conformation that is characterized by the exposure of the B8H10 or AMF7-63 epitopes and does not interact with TRAF6 in cellulo. These data are reminiscent of the absence of an interaction of SOD1 V148I with Derlin-1, a functional component of endoplasmic reticulum-associated degradation (80). Indeed, all TRAF6-interacting SOD1 variants also interact with Derlin-1, suggesting that these mutants may collectively adopt a misfolded conformation that exposes a similar or possibly identical binding site for both interacting proteins. These findings certainly strengthen the notion that different SOD1 mutations may lead to motor neuron toxicity via distinct mechanisms. This possibility could be a contributor to the heterogeneity in disease onset observed for genetic subtypes of SOD1-ALS (81,82).
As shown previously in the context of other neurodegenerative diseases, TRAF6 stimulates the ubiquitination and aggregation of mutant/misfolded proteins (36,39,42,43). Here, TRAF6 stimulates the polyubiquitination of mutant SOD1 in an E3 ubiquitin ligase activity-dependent manner in cellulo. The polyubiquitin chain linkages we detected on mutant SOD1 were primarily Lys-6-, Lys-27-, and Lys-29 -linked, which is reminiscent of the type of chain linkages that have been found on other mutant/misfolded proteins when TRAF6 is co-expressed in cultured cells (36,39). TRAF6 also stimulated the aggregation of mutant SOD1, but surprisingly, the TRAF6 N terminus and E3 ubiquitin ligase activity were dispensable for this. Therefore, we conclude that these two events are independent, contrary to the currently proposed concept for the modification and autophagic degradation of mutant/misfolded protein aggregates (43). Moreover, polyubiquitinated mutant SOD1 was readily detectable in our assays in the absence of proteasome or autophagy inhibitors. This indicates that TRAF6-stimulated polyubiquitination and/or aggregation of mutant SOD1 does not prompt mutant SOD1 turnover but instead leads to its cellular accumulation. In fact, we observed in multiple experiments (e.g. Figs. 2C (2nd blot from bottom, compare lanes 1 and 7), 3A (2nd blot from bottom, compare lanes 1 and 4, 2 and 5), and 6D (2nd blot from bottom, compare lanes [13][14][15]) that the co-expression of TRAF6 resulted in increased mutant SOD1 levels in whole-cell lysates. This was observed with TRAF6 WT and TRAF6 C70A , but not TRAF6 ⌬N or TRAF6 ⌬C , and was observed for TRAF6-interacting SOD1 variants as well as for TRAF6 noninteracting SOD1 WT and SOD1 V148I . Collectively, this suggests that TRAF6 negatively affects mutant SOD1 turnover but that this effect is not dependent on an intermolecular interaction between the two proteins and does not require TRAF6 E3 ligase activity. In line with this conclusion, we observed a reduction in the total level of mutant SOD1 in TRAF6-depleted cells (Fig. 6B), suggesting that lowering TRAF6 levels alleviates the aggregation of mutant SOD1 and prompts mutant SOD1 turnover. Taken together, this alludes to a scenario in which TRAF6 may act as a molecular scaffold serving a mechanical function in the context of mutant/misfolded SOD1 accumulation/aggregation at the mitochondria. That TRAF6 has E3 ubiquitin ligase-independent roles as a scaffolding protein has been previously demonstrated in the context of IL-1 signaling (83). However, given the role of TRAF6 in NF-B signaling, which is intricately connected with regulating autophagy flux, the possibility remains that mutant SOD1 aggregation is indirectly and dynamically modulated in response to effects of TRAF6 on the autophagy machinery (28,29,84).
In addition to being localized to the cytoplasm, we found that a pool of TRAF6 is constitutively localized to mitochondria in the spinal cord. We propose that this mitochondrial TRAF6 interacts with and scaffolds the accumulation of misfolded SOD1. The TRAF6 (EP591Y) antibody we used for immunoblotting specifically detected TRAF6 at 60 kDa in spinal cord mitochondrial fractions and two other bands at 48 and 35 kDa, which were enriched in mitochondrial fractions compared with whole-tissue lysates. This raises the possibility that there is more than one form of TRAF6 and that these forms may be differentially targeted to the cytoplasmic or mitochondrial compartments. How such short forms of TRAF6 are generated and what their function is at the mitochondria remains to be investigated.
TRAF6 is a multifaceted protein; it is a central player at the interface of inflammatory signaling and the autophagy machinery (27)(28)(29)(30) and plays a role in mitophagy (34,35) and RNA metabolism (85,86), all of which are recurrent pathomechanistic themes in ALS, independent of genetic etiology. As such, we speculate that TRAF6 dysfunction due to mutant/misfolded SOD1 binding could affect cellular integrity at various levels and in different cellular compartments. In fact, TRAF6 could participate in a variety of pathomechanisms underlying a broad spectrum of genetic subtypes of ALS, given that TRAF6 is reported to interact with many proteins already implicated in ALS (e.g. heterogeneous nuclear ribonucleoprotein A1 (86), Matrin-3 (86), Profilin-1 (86), Sqstm1/p62 (87), and TBK1 (88)). Our initial examination of the impact of mutant SOD1/ TRAF6 on NF-B signaling did not yield any differences. However, it is possible that the portion of TRAF6 bound to mutant SOD1 may be small compared with the total TRAF6 pool, which may explain our experimental outcome. Further investigations into the role of TRAF6 in ALS could yield important and unexpected insights, which may help explain the heterogeneity and apparent complexity of the disease.

Animals
Transgenic rats expressing human SOD1 WT and SOD1 G93A have been previously described (89,90). SOD1 G93A rats were bred and monitored biweekly as previously published (16). SOD1 G93A rats followed the expected disease course and reached end-stage at around 110 days. SOD1 G93A rats at 10 and 14 weeks were pre-symptomatic, defined as no loss of body weight and intact hindlimb reflex. Animals described as symptomatic displayed bilateral hindlimb paralysis with no phenotypic involvement of the forelimbs. SOD1 WT -expressing and nontransgenic control animals were age-matched. Animals of both sexes were used. Breeding, housing, and all manipulations were carried out in strict accordance with approved protocols from the Centre de Recherche du Centre Hospitalier de Université de Montréal (CRCHUM) Institutional Committee for the Protection of Animals (CIPA) and the Canadian Council on Animal Care (CCAC).

Preparation of mitochondria and lysates from rat spinal cords
Rat spinal cords were homogenized in mitochondria isolation buffer I (10 mM Tris, pH 7.5, 1 mM EDTA, 210 mM D-mannitol, 70 mM D-sucrose) with protease inhibitors (10 g/ml leupeptin, 10 g/ml pepstatin A, 10 g/ml chymotrypsin). Mitochondria were isolated by sedimentation according to previously published protocols (91). For Fig. S4, spinal cords were homogenized in mitochondria isolation buffer II (20 mM HEPES-NaOH, pH 7.4, 1 mM EDTA, 250 mM D-sucrose), and mitochondria were isolated by buoyant density centrifugation on OptiPrep (Sigma, D1556) gradients as published previously (17). For tissue lysate preparations, an aliquot of homogenate was retained, adjusted to 1% v/v SDS and 1% v/v Nonidet P-40, and incubated for 10 min on ice and 10 min at ambient temperature. Extracts were mildly sonicated three times for 30 s with 1 min icing between pulses with a Branson 2510R-DTH bath sonicator and cleared by centrifugation at 13,500 ϫ g for 10 min at 4°C.

Protein extraction, quantification, and Western blotting
Unless otherwise stated, whole-cell lysates were prepared by extraction in RIPA buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 1% v/v Triton X-100, 0.1% v/v SDS, 1% v/v sodium deoxycholate) with protease inhibitors (10 g/ml leupeptin, 10 g/ml pepstatin A, 10 g/ml chymotrypsin) for 10 min on ice and 10 min at ambient temperature. Supernatants were recovered by centrifugation at 13,500 ϫ g for 5 min at 4°C. Protein quantifications were performed with a BCA protein assay (Thermo Fisher Scientific, 23227). Samples were boiled in 1ϫ Laemmli buffer (60 mM Tris-HCl, pH 6.8, 2% w/v SDS, 10% v/v glycerol, 0.025% w/v bromphenol blue, 100 mM DTT) at 98°C for 5 min and separated on 12.5% Tris-glycine polyacrylamide gels. Gels were run at 100 V in running buffer (25 mM Tris, 192 mM glycine, 0.1% w/v SDS) and wet-transferred at 60 V for 1 h at 4°C in transfer Misfolded SOD1 binds the ubiquitin ligase TRAF6 buffer (25 mM Tris, 190 mM glycine, 20% v/v methanol) onto 0.45-m nitrocellulose membranes (Bio-Rad, 1620155). Membranes were blocked in 5% milk/PBS-T (5% w/v Carnation instant skim milk powder in PBS, pH 7.4, 0.1% v/v Tween 20) for at least 30 min at ambient temperature. Primary antibodies were immunoblotted in 5% milk/PBS-T for 2 h at ambient temperature or overnight at 4°C. For antibody pre-absorption with blocking peptide (Fig. S5), anti-TRAF6 (EP591Y) was diluted at 0.412 g/ml in 5% milk/PBS-T, and blocking peptide was added at 2-fold excess of the antibody, and the solution was rotated end-over-end at ambient temperature for 1 h. HRP-coupled secondary antibodies were bound for 2 h at ambient temperature. After primary and secondary antibody incubation, the membranes were washed three times with PBS-T. For all experiments, signals were captured on CL-XPosure TM radiography films (Thermo Fisher Scientific, 34090) using ECL Western blotting substrate (Pierce, 32106), with the exception of Fig. 3A, which was captured on the Bio-Rad ChemiDoc system.

Misfolded SOD1 co-immunoprecipitation from mitochondria and MS
Rabbit anti-AMF7-63 (concentration 1.54 g/l) was coupled to protein A Dynabeads (Invitrogen, 10002D) at full saturation according to bead-binding capacity given by the manufacturer (6.15 l per 40 l of beads). Mouse anti-B8H10 (concentration not provided by the manufacturer) was coupled to protein G Dynabeads (Invitrogen, 10004D) at an equivalent volume to AMF7-63. Beads coupled to isotype-matched ChromPure whole molecule IgGs (Jackson ImmunoResearch, 015-000-003 or 011-000-003) at full saturation were used as control. B8H10 and AMF7-63-coupled beads were cross-linked with BS 3 (Thermo Fisher Scientific, 21580), according to the manufacturer's instructions. Beads were then tumbled with 150 g of spinal cord mitochondria isolated from symptomatic SOD1 G93A rats at 1 g/l final protein concentration in coimmunoprecipitation (IP) buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% v/v Nonidet P-40) overnight at 4°C. The antigen-bound beads were washed three times with co-IP buffer and snap-frozen in liquid nitrogen. For MS analysis, all samples were reduced on-bead with 5 mM DTT for 10 min at 95°C, alkylated with 5 mM iodoacetamide for 1 h at ambient temperature, and trypsin-digested on-column for 18 h at 37°C. Tryptic peptides were subjected to nanoLC-MS/MS analysis using a NanoAcquity UPLC (Waters) coupled to ESI-LTQ-XL-ETD mass spectrometer (Thermo Fisher Scientific), as published previously (92). The .raw files were used to generate .mgf files using ProteoWizard (version 3.0.18250) and submitted to the Matrix Science Mascot search engine (version 2.6.2) to search against a rat Uniprot database (2019-02 release, number of entries 36,064), with human SOD1 sequence added (UniProt ID: P00441), and a decoy database (same database but randomized). Searches were performed with a specified trypsin enzymatic cleavage with one missed cleavage allowed and a 0.6-Da mass tolerance for precursor and fragment ions. Allowed modifications included carbamidomethyl at cysteines (ϩ57 Da) as fixed and oxidation at methionines (ϩ16 Da) as variable. Peptides showing Mascot scores Ն40 show an overall false discovery rate less than 2.5% (decoy/target hits) and were determined as positive hits. The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD01527 (93).
Transfections 400 ng per expression plasmid or respective empty vector (referred to as mock) were transfected into 293FT cells with Lipofectamine LTX with Plus reagent (Invitrogen, 15338030) according to the manufacturer's instructions in Opti-MEM Reduced Serum Medium (Thermo Fisher Scientific, 22600050). Normal growth medium was replaced after 3 h, and cells were collected after 24 h. For siRNA delivery (150 pmol), 293FT cells were transfected with Lipofectamine 2000 (Invitrogen, 11668027) according to the manufacturer's instructions. Normal growth medium was replaced after 5 h, and cells were collected after 72 h. If cDNA expression plasmids were introduced into siRNA-treated cells, this was 24 h prior to cell collection. For NF-B luciferase assays, 293FT cells were co-transfected with pGL3-P2(2X)TK (25 ng, firefly luciferase NF-B reporter), pRL-null (50 ng, Renilla luciferase internal control), p65-Flag (0.5 ng, positive control) or Myc-TRAF6 (20 ng), and an escalating dose of Flag-SOD1 WT or Flag-SOD1 A4V (5-80 ng) with Lipofectamine LTX. Transfections were balanced with pCI-neo-Flag empty vector.

In cellulo protein-protein interaction assays
293FT cells were transfected with the indicated plasmid combinations. Cell pellets were resuspended in PBS (137 mM NaCl, 2.7 mM KCl, 8 mM Na 2 HPO 4 , 1.5 mM KH 2 PO 4 , pH7.4) with protease inhibitors, and lysates were prepared by freezethaw extraction (three cycles of freezing on dry ice and snap thawing in a 37°C water bath). The supernatants were recovered after centrifugation at 13,500 ϫ g for 5 min at 4°C. Lysates were adjusted with 2ϫ IP-buffer (100 mM Tris-HCl, pH 7.4, 300 mM NaCl, 2 mM EDTA, 1% v/v Nonidet P-40, with protease inhibitors) to 1.5ϫ final concentration and tumbled overnight at 4°C with protein G Dynabeads coupled to mouse anti-Flag or rabbit anti-Myc. Antigen-bound beads were washed three times with 1ϫ IP-buffer and eluted by boiling at 98°C for 5 min in 2.5ϫ Laemmli buffer. The eluates were analyzed by Western blotting as described.

In cellulo ubiquitination assays
293FT cells were transfected with the indicated plasmid combinations. Whole-cell lysates were prepared by extraction in RIPA buffer for 15 min on ice and were sonicated three times for 30 s with 1 min icing between pulses with a Branson 2510R-DTH bath sonicator. Extracts were cleared by centrifugation at 13,500 ϫ g for 10 min at 4°C. Protein G Dynabeads were coupled to rat anti-HA, and 400 g of cell lysate were immunoprecipitated as described.

Filter retardation assays
293FT cells were transfected with the indicated plasmid combinations as described. Filter retardation assays were per-formed along previously published methods with slight modifications (99). Cells were lysed by freeze-thaw extraction in PBS with protease inhibitors. The DNA condensate was manually removed with a pipette tip, and lysates were additionally digested with 0.35 Kunitz units of DNase I (Roche Applied Science, 11284932001) per l of extraction buffer for 30 min at 37°C. Extracts were cleared by low-speed centrifugation at 800 ϫ g for 10 min at 4°C. Supernatants were recovered and quantified, and samples of 60 g in a total volume of 150 l of PBS were loaded onto cellulose acetate membrane filters with 0.2-m pore size (Whatman, 10404131) using the Bio-Dot microfiltration apparatus (Bio-Rad, 1706545). After 1 h of passive filtration by gravity flow, the wells were vacated and washed three times with 200 l of PBS by gentle vacuum application. Filters were removed from the apparatus, blocked in 5% milk/PBS-T for 1 h, and immunoblotted as described.

NF-B luciferase assays
NF-B luciferase assays were performed with the Dual-Luciferase reporter assay system (Promega, E1960) according to the manufacturer's instructions and as published previously (100). Briefly, 293FT cells were transfected with the indicated plasmid combinations in quadruplicate conditions. Luciferase luminescence was measured with a BioTek Synergy HT multimode microplate reader equipped with an automated substrate injection system. Data are expressed as the ratio of mean luminescence Ϯ S.D. of firefly luciferase/Renilla luciferase. For SDS-PAGE, quadruplicate samples were pooled, and 30 l were loaded.

qRT-PCR
Human iPSC-derived motor neurons were differentiated for 14 days, and total RNA was extracted with the miRNeasy mini kit (Qiagen, 217004) according to the manufacturer's instructions. Reverse transcriptions were performed on 400 ng of total RNA extract in a 4-l total volume, using random primers (Invitrogen, 48190011), and Moloney murine leukemia virus reverse transcriptase (Invitrogen, 28025013), according to the manufacturer's instructions. Reactions were conducted in singleplex, in a 10-l total volume containing 1ϫ TaqMan Fast Advanced Master Mix (Applied Biosystems, 4444556), 1ϫ Taq-Man primers/probe set (Thermo Fisher Scientific), and 1 l of diluted cDNA. qRT-PCRs were performed on a QuantStudio 3 real-time PCR system (Thermo Fisher Scientific). Probes were from Applied Biosystems: ACTB (Hs01060665_g1); CHAT (Hs00758143_m1); GAPDH (Hs02786624_g1); HB9 (Hs00907365_m1); and SOX1 (Hs01057642_s1). The data were analyzed using the ⌬CT method (101).

Software and statistics
Enrichr was used to perform gene ontology (BP_2018, MF_2018, and CC_2018) and pathway enrichment (KEGG_ 2019_Human) analyses (21,22). Protein sequence alignments were generated with MultAlin (102). Three-dimensional protein structure models were generated using iTasser (Iterative Threading Assembly Refinement) (103). The full-length model was constrained with Protein Data Bank code 3HCS.A, without alignment. No constraints were used for the deletion models.

Misfolded SOD1 binds the ubiquitin ligase TRAF6
The highest C score models were chosen for display. Densitometry was performed in Adobe Photoshop CS4, and graphs were generated with GraphPad Prism 8. Statistics were performed in GraphPad Prism 8 (one-way analysis of variance with Dunnett's multiple comparisons or t test, where appropriate).