Specific Post-Translational Modifications of VDAC3 in ALS-SOD1 Model Cells Identified by High-Resolution Mass Spectrometry

Damage induced by oxidative stress is a key driver of the selective motor neuron death in amyotrophic lateral sclerosis (ALS). Mitochondria are among the main producers of ROS, but they also suffer particularly from their harmful effects. Voltage-dependent anion-selective channels (VDACs) are the most represented proteins of the outer mitochondrial membrane where they form pores controlling the permeation of metabolites responsible for mitochondrial functions. For these reasons, VDACs contribute to mitochondrial quality control and the entire energy metabolism of the cell. In this work we assessed in an ALS cell model whether disease-related oxidative stress induces post-translational modifications (PTMs) in VDAC3, a member of the VDAC family of outer mitochondrial membrane channel proteins, known for its role in redox signaling. At this end, protein samples enriched in VDACs were prepared from mitochondria of an ALS model cell line, NSC34 expressing human SOD1G93A, and analyzed by nUHPLC/High-Resolution nESI-MS/MS. Specific over-oxidation, deamidation, succination events were found in VDAC3 from ALS-related NSC34-SOD1G93A but not in non-ALS cell lines. Additionally, we report evidence that some PTMs may affect VDAC3 functionality. In particular, deamidation of Asn215 alone alters single channel behavior in artificial membranes. Overall, our results suggest modifications of VDAC3 that can impact its protective role against ROS, which is particularly important in the ALS context. Data are available via ProteomeXchange with identifier PXD036728.


Introduction
Amyotrophic lateral sclerosis (ALS) is an adult neurodegenerative disease characterized by the progressive loss of upper and lower motor neurons in the spinal cord and brainstem [1]. From its onset, the disease progresses rapidly, leading the patient to death [2]. To date, there are no effective treatments to counteract the outcome of the disease.
About 10% of ALS cases are associated with genetic defects (familial ALS, fALS), while the most common forms are sporadic (sALS, about 90-95% of cases). Although various environmental factors have been indicated as being responsible for sALS, the importance of the genetic component has also been demonstrated for these more common forms of the disease [3].
In this work, by combining HRMS analysis with "in-solution" digestion of an enriched fraction of VDACs, in VDAC3 purified from NSC34-SOD1G93A model ALS-SOD1 cells, we identified specific irreversible PTMs that can destabilize the channel by impacting its function.

Results
This work follows our previous study on NSC34 model SOD1-ALS cells. In that work, we had identified specific post-translational modifications in VDAC1 that can lead to important changes in the channel structure and, thus, in the bioenergetic metabolism of ALS motor neurons [20]. Considering that changes found are most likely consequential to the high levels of oxidative stress typical of motor neurons affected by ALS, in this work we wanted to extend our analysis also to the VDAC3 isoform purified from ALS-related NSC34-SOD1G93A cells and, as a control, from NSC34-SOD1WT or NSC34 cells. Recently, we established the important role played by VDAC3, but not VDAC1, in the cellular response to ROS-induced oxidative stress, showing in particular that VDAC3 cysteines are essential for the ability of the protein to control mitochondrial ROS homeostasis [24]. In particular, by using HRMS analysis we searched oxidized or succinate cysteines, oxidized or dioxidized methionines, ubiquitinated lysines, phosphorylated serines/threonines/tyrosines, citrullinated arginines, cysteinylated and deamidated asparagines and glutamines. Protein reduction and alkylation were performed before the purification of VDACs from mitochondria, to rule out the possibility of non-specific and/or unwanted oxidation occurring during the purification protocol. Hydroxyapatite (HTP) eluates were digested in-solution using trypsin and chymotrypsin, and then the highly complex enzyme peptide mixtures were analyzed in triplicate (technical replicates) by liquid chromatography-high resolution mass spectrometry.
The Mus musculus VDAC3 sequence (SwissProt Acc. N. Q60931) includes two methionines at positions 26 and 155, and six cysteines at positions 2, 8, 36, 65, 122, and 229. The numbering adopted here starts from Met 1 , which is actually absent in the mature protein as confirmed by our MS data ( Figure 1) and similar to rat and human VDAC3 isoforms [25,26].
Although, the role of VDAC1 in neurodegeneration is well known [12,21,22], the involvement of the other two isoforms in these pathways remains poorly defined. This is probably due to the greater relative abundance of VDAC1 compared with the other isoforms. In particular, although VDAC3 is the least abundant and least studied isoform, important functional roles, as an oxidative stress sensor or tumor suppressor, have been proposed for this isoform in cellular and animal models [16,23].
In this work, by combining HRMS analysis with "in-solution" digestion of an enriched fraction of VDACs, in VDAC3 purified from NSC34-SOD1G93A model ALS-SOD1 cells, we identified specific irreversible PTMs that can destabilize the channel by impacting its function.

Results
This work follows our previous study on NSC34 model SOD1-ALS cells. In that work, we had identified specific post-translational modifications in VDAC1 that can lead to important changes in the channel structure and, thus, in the bioenergetic metabolism of ALS motor neurons [20]. Considering that changes found are most likely consequential to the high levels of oxidative stress typical of motor neurons affected by ALS, in this work we wanted to extend our analysis also to the VDAC3 isoform purified from ALS-related NSC34-SOD1G93A cells and, as a control, from NSC34-SOD1WT or NSC34 cells. Recently, we established the important role played by VDAC3, but not VDAC1, in the cellular response to ROS-induced oxidative stress, showing in particular that VDAC3 cysteines are essential for the ability of the protein to control mitochondrial ROS homeostasis [24]. In particular, by using HRMS analysis we searched oxidized or succinate cysteines, oxidized or dioxidized methionines, ubiquitinated lysines, phosphorylated serines/threonines/tyrosines, citrullinated arginines, cysteinylated and deamidated asparagines and glutamines. Protein reduction and alkylation were performed before the purification of VDACs from mitochondria, to rule out the possibility of non-specific and/or unwanted oxidation occurring during the purification protocol. Hydroxyapatite (HTP) eluates were digested in-solution using trypsin and chymotrypsin, and then the highly complex enzyme peptide mixtures were analyzed in triplicate (technical replicates) by liquid chromatography-high resolution mass spectrometry.
The Mus musculus VDAC3 sequence (SwissProt Acc. N. Q60931) includes two methionines at positions 26 and 155, and six cysteines at positions 2, 8, 36, 65, 122, and 229. The numbering adopted here starts from Met 1 , which is actually absent in the mature protein as confirmed by our MS data ( Figure 1) and similar to rat and human VDAC3 isoforms [25,26].  Figure 1. Sequence coverage map of VDAC3 from NSC34, NSC34-SOD1WT and NSC34-SOD1G93A cell lines obtained by tryptic and chymotryptic digestion. The solid line indicates the sequence that was obtained from tryptic peptides; dotted lines: sequence obtained from chymotryptic peptides. Unique tryptic (indicated in bold and blue) and chymotryptic (indicated in bold and black) peptides originating from missed cleavages were used to distinguish and cover the sequences shared by isoforms. Sequences shared by multiple isoforms are shown in red. Sequence numbering considered the starting methionine residue, which is eliminated during protein maturation.

of 22
The results obtained from tryptic and chymotryptic fragments analyses yielded 98.9% protein coverage (279 of 282 amino acid residues) ( Figure 1), with the exception of the dipeptide Arg119-Arg120 and Tyr225. Although some tryptic peptides are shared among the three VDAC isoforms, due to the identification of unique peptides originating from missed cleavages, the obtained sequence coverage unambiguously identifies the mouse VDAC3 protein (Figure 1). MS analysis performed on samples of VDAC3 prepared by the same procedure but purified from another set of cell cultures of NSC34, NSC34-SOD1WT, and NSC34-SOD1G93A (biological replicates) confirmed the results obtained and the reproducibility of the experimental data produced (data not shown).

Mass Spectrometry Analysis of VDAC3 from NSC34 Cell Line
In the nanoRP-UHPLC/High-Resolution ESI-MS/MS analysis of the enzyme digests, Met 26 and Met 155 were identified in the normal form (Table S1, fragment 2, and Table S2, fragments 4, 21), and also in the sulfoxide form (Table S3, fragments 1, 5, 6, and  Table S4, fragments 1, 2). Figure S1A,B shows the full scan and fragment ion mass spectra of the doubly charged molecular ion of the peptide GYGFGMVK with Met 26 modified as methionine sulfoxide ( Figure S1A), and the triply charged molecular ion of the peptide DCFSLGSNVDIDFSGPTIYGWAVLAFEGWLAGYQMSFDTAK with Met 155 as methionine sulfoxide ( Figure S1B). This oxidation was confirmed by the presence in the MS/MS spectra of the characteristic neutral loss of 64 Da corresponding to the ejection of methanesulfenic acid from the side chain of MetO [27].
Although from these data it was not possible to obtain a precise determination of the relative amount of Met and Met sulfoxide, by comparing the absolute intensities of the multiply charged molecular ions of the respective peptides, a rough estimate of their relative abundance could be achieved.

Mass Spectrometry Analysis of VDAC3 from NSC34-SOD1WT Cell Line
The PTMs identified on purified VDAC3 from NSC34-SOD1WT control cells, which express the wt form of human SOD1, are similar to those of VDAC3 from parental NSC34 control cells and similar to what has been shown previously for the VDAC1 isoform [20]. Specifically, in addition to the peptides containing methionines 26 and 155 in normal form (Table S7, fragments 2, 11, and Table S8, fragments 6, 7), peptides with these residues oxidized to methionine sulfoxide (Table S9, fragment 5, and Table S10, fragments 1, 2, and Figure S4A,B) were also detected. The MetO/Met ratio was approximately 12:1 for Met 26 ( Table 2 and Table S12) and 7:1 for Met 155 (Table 1 and Table S11), respectively.
Furthermore, unlike in NSC34 cells, we identified Cys 2 in both reduced and acetylated form (Table S7, fragment 1, and Table S8, fragments 1-3) ( Figure S6), and in both non-acetylated and oxidized to sulfonic acid (Table S9,   The inset shows the full scan mass spectrum of the molecular ion. Fragment ions originating from the neutral loss of H2O are indicated by an asterisk. The fragment ion originating from the neutral loss of NH3 is indicated by two asterisks.

Mass Spectrometry Analysis of VDAC3 from NSC34-SOD1G93A Cell Line
The same HRMS investigative approach was used to analyze PTMs of VDAC3 extracted from the ALS model NSC34-SOD1G93A cells, stably expressing the human SOD1G93A mutant. We identified Met 26 and Met 155 in the fully oxidized sulfoxide form

Other Post-Translational Modifications of VDAC3 2.4.1. Succination, Ubiquitin and Ubiquitination, Phosphorylation, Citrullination, Cysteinylation, and Dioxidation
In addition to oxidations, other types of PTMs were sought in VDAC3 purified from all NSC34 cell lines analyzed. Specifically, we ascertained that in each of them, Cys 65 is succinated (Tables S19A-C and S20A-C) and in the NSC34-SOD1G93A ALS model cell line, this modification was interestingly present in greater amounts than in control cells (Table 3 and Figure S11A-C). Table 3. Ox/Red ratio (average and standard deviation) of the absolute intensities of the molecular ions of tryptic peptides containing succinated and non-succinated cysteines found in the analysis of VDAC3 from NSC34, NSC34-SOD1WT, NSC34-SOD1G93A cell lines reduced with DTT, carboxyamidomethylated and digested in-solution.

Peptide
Position Furthermore, no ubiquitin and ubiquitination of lysine residues, phosphorylation of serine, threonine or tyrosine, citrullination of arginine, cysteinylation, and dioxidation of methionines to sulfone were found.

Identification of Deamidation Sites
In this study, deamidation of asparagine and glutamines were identified exclusively in VDAC3 from the NSC34-SOD1G93A cell line, although deamidated glutamines were found in low amounts and the relative deamidate/normal ratio was, thus, not determined. Interestingly, we found that Asn 215 was converted to aspartate in significant amounts (deam/normal ratio 0.1) only in ALS cell model (Table 4 and Table S21, Table S22, fragments 4; Figure 3). Table 4. Ox/Red ratio (average and standard deviation) of the absolute intensities of the molecular ions of tryptic peptides containing deamidated and non-deamidated asparagine found in the analysis of VDAC3 from NSC34-SOD1G93A cell line reduced with DTT, carboxyamidomethylated and digested in-solution.  In contrast, asparagines 167, 238 and 239 were deamidated at minor level (deam/normal ratio 0.01) (Tables 4 and S21, Table S22, fragments 3 and 5; Figure S12BA-C), and asparagines 76, 106 and 168 were converted to aspartate only in trace amounts (deam/norm ratio 0.002-0.003) (Tables 4 and S21, Table S22, fragments 1, 2, 3; Figure  S12D-F).
In addition to Asn/Asp deamidation events, we analyzed VDAC3 for the presence of succinimide. Usually, succinimide intermediates are formed in the cell as a result of spontaneous deamidation of asparagine and dehydration of aspartic acid [28], and its accumulation represents a sign of aging or a response to cellular stress condition [29]. We identified succinimide at positions 167 and 168 of VDAC3 from all three cell lines studied (Tables S23-S25; Figure S13A-G), although with a different succ/norm ratio, and at position 238 only in the NSC34-SOD1WT cell line (Table S24, fragment 2; Figure S13H). It is worth noting that in the control NSC34 and NSC34-SOD1WT cell lines, asparagine-derived succinimide intermediates were visible in trace amounts (suc/norm ratio ranging from 0.002:1 to 0.004:1) (Tables 5, S26 and S27), whereas in the ALS model cell line NSC34-SOD1G93A they were observed in significantly larger amounts (Tables 5 and S28). In contrast, asparagines 167, 238 and 239 were deamidated at minor level (deam/normal ratio 0.01) ( Table 4 and Table S21, Table S22, fragments 3 and 5; Figure S12A-C), and asparagines 76, 106 and 168 were converted to aspartate only in trace amounts (deam/norm ratio 0.002-0.003) ( Table 4 and Table S21, Table S22, fragments 1, 2, 3; Figure S12D-F).
In addition to Asn/Asp deamidation events, we analyzed VDAC3 for the presence of succinimide. Usually, succinimide intermediates are formed in the cell as a result of spontaneous deamidation of asparagine and dehydration of aspartic acid [28], and its accumulation represents a sign of aging or a response to cellular stress condition [29]. We identified succinimide at positions 167 and 168 of VDAC3 from all three cell lines studied (Tables S23-S25; Figure S13A-G), although with a different succ/norm ratio, and at position 238 only in the NSC34-SOD1WT cell line (Table S24, fragment 2; Figure S13H). It is worth noting that in the control NSC34 and NSC34-SOD1WT cell lines, asparaginederived succinimide intermediates were visible in trace amounts (suc/norm ratio ranging from 0.002:1 to 0.004:1) ( Table 5, Tables S26 and S27), whereas in the ALS model cell line NSC34-SOD1G93A they were observed in significantly larger amounts (Table 5 and  Table S28).
Furthermore, the tryptic peptide Trp 75 -Lys 90 containing Asp 76 as a succinimide intermediate and Asp 79 in the form of aspartate methyl ester, derived from the hydrolysis of succinimide to L-IsoAsp followed by enzymatic methylation [30], was detected only in the NSC34 and NSC34-SOD1WT cell lines (Table 6 and Figure S14A,B).

Bioinformatic Prediction of VDAC3 N215D Mutant Structure
To evaluate the stability of VDAC3 N215D mutant, a computational structural prediction analysis was performed. The human VDAC3 (hVDAC3) structure was rebuilt by homology modelling starting from the VDAC1 PDB file (the crystallographic structure of VDAC3 has not yet been determined), while the VDAC3 N215D mutation was inserted using, as reference, the VDAC3 PDB file previous generated. Specifically, we used a simulation of the structure of human VDAC3 as the reference structure given the high percentage of identity (98.58%) and similarity (99%) with the mouse ( Figure S15).
To assess the effect of the N215D mutation on channel conformation and stability, the tertiary structure model of human VDAC3 (hVDAC3) taken as a reference (Q9Y277.1) was modified by replacing Asn 215 with aspartate, thus mimicking deamidation at the same position ( Figure 4A,B).  Ramachandran plots (RP) were also generated to evaluate the predict conformation of the mutated VDAC3 channel by determining the orientation of each amino acid residue relative to the reference wild-type (wt) structure ( Figure 4C).
The RP for VDAC3 wt indicates that almost all amino acid residues are distributed in favored regions (95.01%) (Table S29 and Figure S16). The remaining residues are either arranged in allowed regions (4.27%) but with a very high degree of stericity, or distributed in outlier (not allowed) conformations (0.72%).
The three amino acids located in regions are disallowed in the RP of VDAC3 N215D are Gly 23 , Gly 38 , Lys 53 . These amino acids are found spread out in the channel and away from the Asn 215 residue ( Figure S17). Therefore, it is unlikely that these residues can influence the structure of the channel or have an effect on Asn 215 , also because two of the three residues are glycine.
No major differences were observed in amino acid arrangement by comparing the RP of VDAC3 wt with that of VDAC3 N215D ( Figure 4C). The percentages of each individual region of both RP are shown in Table S29.
Overall, the single deamidation event on Asn 215 does not seem to affect the channel stability or produce a substantial modification in the VDAC3 channel structure. This result is not surprising, considering that the mutated residue resides not in the pore wall or in the alpha-helical stretch but in a loop.

Electrophysiological Properties of VDAC3 N215D Mutant
Considering the specific PTMs found in VDAC3 from the NSC34 SOD1-G93A cell line, we wanted to investigate their possible impact on channel functionality. Since it is impossible to produce recombinant proteins with specific oxidated or succinated residues, we focused on evaluating the effects of the most significant deamidation event found in VDAC3, namely the Asn/Asp substitution at the residue 215 residue. For this purpose, we first expressed, purified and refolded the mVDAC3 N215D mutant. Then, using a Planar Lipid Bilayer (PLB) instrument, the electrophysiological properties of the VDAC3 mutant were thoroughly investigated after reconstitution in a lipid bilayer. Measurements of single channel conductance, registered in 1 M KCl upon non-reducing conditions, revealed a slight increase in the pore diameter of VDAC3 mutant (approx. 600 pS) compared to the wild-type protein, mVDAC3 wt (approx. 450 pS). Histograms of the amplitude values as a function of the number of events indicate that N215D mutation also affects electrophysiological stability of the channel ( Figure 5). As shown in Figure 5B,D, the presence of Asp 215 residue caused a different distribution of low-and high-conducting states of VDAC3. Under applied voltage (+10 mV), the mutant protein widened the distribution of the amplitude values corresponding to low-conducting state/s (L), concurrently doubling the conductance in the low-conductance state to approx. 3 pA and abolishing the main peak of VDAC3 wild type (VDAC3 wt) high conducting state (H). These results suggest that VDAC3 N215D undergoes more rapid fluctuations between the high and the low conducting state/s than the wild-type protein. The effect of asparagine deamination on the voltage dependence of VDAC3 was monitored by triangular voltage ramps in the presence or absence of DTT.
As already reported elsewhere [31,32], in non-reducing conditions, the VDAC3 was found to be insensitive to the applied voltage as the channel current steadily increased and decreased in the range of ±50 mV ( Figure 6A) without any variation in the slope of the current vs. voltage (I-V) plot ( Figure 6B). As already reported elsewhere [31,32], in non-reducing conditions, the VDAC3 was found to be insensitive to the applied voltage as the channel current steadily increased and decreased in the range of ±50 mV ( Figure 6A) without any variation in the slope of the current vs. voltage (I-V) plot ( Figure 6B).  Figure 6C). The I-V plot shown in Figure 6D further emphasizes the current transitions corresponding to substantial changes in the slope of the curve. The graph of the normalized average conductance G/G 0 plotted as a function of membrane voltage (Vm) ( Figure 6E) depicts an almost flat line for wild-type VDAC3.
Vice versa, the G/G 0 ratio of VDAC3 N215D does not remain constant and looks more like the characteristic bell-shaped curve of VDAC channels, with lower conductance values at higher membrane potentials.  Voltage dependence analysis was also performed after pre-incubating VDAC3 wt and N215D with 5 mM DTT. As shown in Figure 7A,B, treatment of the protein with a reductant completely changes the current profile of VDAC3 in response to a voltage ramp from 0 to ± 50 mV: channel closure is readily observed at both positive and negative potentials. Under these experimental conditions, however, there are no appreciable differences in the voltage response of VDAC3 N215D ( Figure 7C,D) compared with the wild-type protein, as demonstrated by the current vs. voltage (I-V) plot of the mutant that almost completely overlaps those of VDAC3 wt ( Figure 7E). The only remark is that negative voltages starting from-40 mV close the VDAC3 N215D channel more steadily than the wild-type one ( Figure 7B,D).

Discussion
Multiple dysregulated mechanisms have been implicated in the pathogenesis of ALS, as of several neurodegenerative diseases. In particular, a mechanistic interplay between protein misfolding, oxidative stress and mitochondrial dysfunction in ALS has been highlighted. VDAC channel, the main protein at the outer mitochondrial membrane, is known to interact with SOD1 mutants linked to ALS. Recently, a specific range of non-enzymatic PTMs were identified in VDAC1 from NSC34-SOD1G93A cells, suggesting the appearance of important structural changes in the channel well correlated with defective mitochondrial metabolism in ALS motor neurons.
Unlike enzymatic PTMs, non-enzymatic PTMs mostly include irreversible chemical modifications on proteins, mediated by reactive compounds as ROS [33,34]. Despite these PTMs represent signs of aging or endogenous cell chemical damage, their role in neurodegenerative contexts has not yet been properly investigated.
Considering all of the above, in this work we investigated PTMs of the VDAC3 isoform from the NSC34 SOD1-G93A ALS cell model. Although the high sequence similarity between VDAC1 and VDAC3 is indicative of a similar pore-forming structure, a physiological unique role of redox sensor has been suggested for the VDAC3 isoform [23,35].
A function also supported by previous interactomics data shows the involvement of VDAC3 in important cellular signaling pathways [36].
Additionally, of interest are scientific validations of interactions between the SOD1 enzyme (whose mutants cause ALS1) and various VDAC3 interactors (PRDX5, TBP1, 37LRP, EF1D) reported at https://www.uniprot.org, accessed on 1 December 2021. Taken together, this information, combined with the already known presence of specific PTMs in VDAC1 from NSC34 SOD1G93A, suggests a role of the VDAC3 isoform in the pathogenesis of ALS.
In this work, to identify PTMs in VDAC3, we used high-resolution mass spectrometry [19,20,25,26], an extremely refined tool that provides particularly useful information for the interpretation of the biological functions of the proteins studied. Moreover, thanks to an "in-solution" digestion of the HTP eluate, we overcame the problems associated with the purification of membrane proteins, such as their very high hydrophobicity and low solubility. In addition, the difficulty in isolating individual VDAC isoforms resulted in the need to analyze them as components of a complex mixture, with HRMS investigation of the VDAC3 protein being particularly difficult because this isoform is generally the least expressed among VDACs. More importantly, these refined technical procedures avoided the dangers of unwanted oxidation due the sample preparation step as in [20].
In this study we obtained an excellent sequence coverage (about 99%), with the only uncovered moieties corresponding to Arg 119 -Arg 120 and Tyr 225 . Furthermore, the oxidation state of methionine and cysteine residues was precisely determined. Interestingly, we found in the VDAC3 sequence both Met 26 and Met 155 fully oxidized to methionine sulfoxide exclusively in NSC34-SOD1G93A cells, while in non-ALS control cell lines they were found to be partially oxidized. Met 26 localizes in the cytosolic rim of VDAC3, at the beginning of the first beta strand and facing partly inward in the pore. A position that could expose this residue to the effect of excess of ROS accumulated in the compartments as a result of pathology. As VDAC1 from the same ALS cell model [20], oxidative modifications of Met 155 could be indeed explained by its localization in β-sheet 10, facing the lipid bilayer. Thus, it is possible that OMM lipids peroxidized by ALS pathogenesis [37] can modify susceptible residues oriented towards the membrane.
We found no relevant differences in the over-oxidation of cysteines between VDAC3 purified from ALS model or non-ALS NSC34 cells. Indeed, Cys 36 , Cys 65 and Cys 229 , all of which faced the IMM, were detected either in tri-oxidized and reduced (carboxyamidomethylated) forms in both ALS and control cells, with a similar Ox/Red ratio for homologous cysteines. This most likely means that the level of cysteine oxidation has already reached the maximum stage compatible with the functionality of the protein.
Furthermore, Cys 8 and Cys 122 , respectively located in the N-terminal alpha helix and within an IMM-facing loop, are always in the carboxyamidomethylated form, thus suggesting their potential availability to form disulfide bridges.
In addition, Cys 2 was found to be both reduced and acetylated (the "normal" situation) and, unexpectedly, oxidized to sulfonic acid and was not acetylated, neither in cells expressing human SOD1 wt nor the mutated form. The presence of a cysteine at the Nterminus is a unique feature of the VDAC3 isoform. Evidently, just the peculiar positioning of this cysteine at the beginning of the N-terminal alpha helix could already promote its over-oxidation in cells that over-express SOD1wt. Indeed, high levels of native SOD1 are known to result in misfolding and aggregation of SOD1 itself, correlating with forms of sALS [5].
The oxidative stress condition, typical of all forms of ALS, as well as the results obtained in the previous work on VDAC1 [20], prompted us to search the sequence of VDAC3 for asparagine and glutamine deamidation events in addition to oxidations. Asparagine and glutamine deamidation are spontaneous, non-enzymatic and post-biosynthetic modifications affecting structure, stability, folding, and aggregation of proteins [38][39][40][41]. Besides playing a role in cataract formation, its importance has also been established in Alzheimer's, Parkinson's, and ALS as other degenerative diseases [20,[42][43][44]. Moreover, being a modification that proteins accumulate with cellular aging and following pathological conditions, it could be useful as a biomarker of disease [45,46]. The deamidation mechanism involves the formation of an unstable cyclic succinimidyl intermediate, which is hydrolyzed to yield Asp or isoAsp, with an approximate ratio of 3:1 [47,48]. Isoaspartic acid is an isomer of aspartic acid with the C β incorporated into the backbone. These isopeptide bonds are known to impair protein structure/function or enhance the aggregation process, contributing to protein aging and folding disorders typical of neurodegenerative diseases. In this regard, it has been reported that in erythrocytes of ALS patients there is an accumulation of L-isoaspartyl residues [30]. Consistent with the known data, our analysis showed a higher amount of succinimidyl intermediate in the VDAC3 of the NSC34-SOD1G93A cell line than in the other two NSC34 non-SLA cell lines. Additionally, we detected the presence of asparagine 79 as aspartate methyl ester in the NSC34 cells and in those expressing the SOD1 wt, but not in the cell line expressing the mutated SOD1 G93A. This can be explained considering that the free α-carbonyl group of the dangerous L-isoAsp is methylated by the enzyme L-isoaspartyl/(D-asp) methyltransferase (PIMT) and Sadenosylmethionine (AdoMet) as a methyl donor. Subsequently, hydrolysis of the methyl ester leads to the reformation of the intermediate L-succinimide which is then partly hydrolyzed to L-Asp, thus completing the physiological repair reaction that counteracts the accumulation of dysfunctional proteins [49][50][51] In addition, only in NSC34-SOD1G93A cells did we detect very high levels (10% of molecules) of deamidated Asn 215 (Table 4 and Figure 4C). Considering the potential impact of this modification on protein structure/function, we assessed the stability of the VDAC3 N215D mutant by performing a molecular modeling analysis. For this purpose, we replaced Asn 215 with an aspartate in the structure model of human VDAC3. The results show no significant changes in the pore structure of the VDAC3 N125D mutant. In both Ramachandran plots of VDAC3 wt and N215D there is a predominance of residues located in the β region, suggesting for both structures the typical β-barrel motif that characterizes all isoforms of VDAC ( Figure S15). Notably, the Asn/Asp substitution would not seem to result in any change in structure, and this could be due to the position that residue 215 has within the structure. Indeed, this asparagine is located in the center of a loop connecting two β-strands ( Figure 4C), with the side chain protruding toward the cytosol, with no apparent involvement in the formation of the pore wall. In addition, results on mVDAC3 N215D were produced using the model of human VDAC3, reconstructed by homology to the crystallographic structure of human VDAC1, as the reference structure. This model, although predictively very similar, may, in reality, have some differences.
Then, considering the little information provided by homology analysis, we wanted to further investigate whether deamidation of Asn215 was capable of producing effects on protein function. Therefore, we produced, purified and folded the mVDAC3 N215D mutant protein and analyzed its electrophysiological characteristics at PLB. Through this analysis, we ascertained that the N215D mutation results in a significative increase in the fluctuation of pore conductance under non-reducing conditions. In contrast, the voltage dependence analysis performed under reducing conditions indicates that the VDAC3 mutant has a behavior mostly overlapping with the wt protein. Furthermore, VDAC3 N215D shows voltage dependence already in non-reducing conditions, at a variance of the wt VDAC3, indicating that the residue may be involved in the voltage-sensor structure. On the other hand, in reducing conditions, no special difference between the wt and pathological forms are strikingly apparent.
As previously pointed out, Asn 215 localizes in a loop formed by six amino acids (212-217) and is exposed toward the cytosolic environment, i.e., in a water-exposed context where ROS can make their effect felt faster. Although the other asparagines of VDAC3 would also appear to be mostly exposed toward the cytosolic environment, they exhibit lower levels of deamidation than Asn 215 . This difference might depend on the nearby amino acids as well as on the specific interactors with which each residue is physiologically engaged [35]. Specifically, Asn 215 might be in a suitable position to establish interactions with cytosolic molecules important for VDAC3 function and whole organelle homeostasis.
It is noteworthy that the higher level of succinimide intermediate was detected for Asn 167 (much less for Asn 168 ) of VDAC3 from ALS model cells compared with the protein purified from control lines ( Table 5). This finding could depend on the fact that the aspartyl and iso-aspartyl residues produced by deamidation of Asn 167 dehydrate faster (or hydrate more slowly) in the reaction that still reforms the intermediate succinimide. The reduced levels of Asp produced by deamidation from Asn 167 could be explained as follows, as the positioning in the pore of this residue could also favor the formation of the intermediate. Indeed, Asn 167 localizes in the central portion of the 11th beta strand, protruding toward the membrane lipid environment, preceded by two polar amino acids (serine and glutamine) ( Figure 4C). The latter condition is favorable for the deamidation reaction. Furthermore, the exposure of Asn 167 toward the lipid bilayer allows us to hypothesize that OMM lipids peroxidized by ALS pathogenesis [36] may modify sensitive membrane-oriented residues, as previously reported for VDAC1 from the same cell type [20].
Another interesting result, only discovered in the VDAC3 from NSC34-SOD1G93A cells, concerns the much higher levels of Cys 65 succination compared with the control NSC34 cell lines. Succination, or the formation of S-(2-succino) cysteine (2SC), is a nonenzymatic and irreversible modification that physiologically occurs by fumarate adduction to the sulfhydryl group of Cys only, through a Michael-type reaction [52]. Succination preferentially targets cysteine residues with low pKa values (up to 3-4), such as those located in protein active sites or placed in a suitable biological context [53]. For this reason, 2SC is a recognized biomarker of metabolic stress and, more specifically, mitochondrial stress. In addition, this PTM increases with aging, when there is an excess of nutrients, a high ATP/ADP and/or NADH/NAD+ ratio, or, more generally, when there is an abnormal increase in Krebs cycle intermediates. It is, therefore, associated with mitochondrial dysfunction and increased ROS, resulting in neuronal damage [54].
It is, thus, not surprising that we found high levels of succination in VDAC3 of the ALS model cell line. Specifically, succinated Cys 65 found in VDAC3 is located on the edge of the pore looking toward the IMM, usually in itself a highly oxidizing environment, but even more exacerbated in the ALS pathological context due to the impairment of antioxidant defenses. It is noteworthy that several proteins known to interact with VDAC3 have been found succinated in other cell types [35,55]. Interestingly, among the interactors of VDAC3, succination of Protein Disulfide Isomerase was reported, thus providing a link between mitochondrial stress and endoplasmic reticulum stress [56]. Succination of α-and β-tubulins is also known [57] as a mechanism that can inhibit their polymerization and, thus, alter mitochondrial dynamics. The latter aspect has an important consequence for ALS, and neurodegeneration in general, as neurons are cells incredibly dependent on their cytoskeletal transport mechanism. Indeed, in the cell body and axons of mature neurons, the cytoskeleton enables fundamental processes, such as movement of every component, delivery of new proteins and organelles to distal sites, as well as removals of aged ones [58].

Chemicals
All chemicals were of the highest purity commercially available and were used without further purification. Ammonium bicarbonate, calcium chloride, phosphate-buffered saline (PBS), Tris-HCl, Triton X-100, sucrose, mannitol, ethylene glycol tetraacetic acid (EGTA), ethylenediaminetetraacetic acid (EDTA), formic acid (FA), dithiothreitol (DTT) and iodoacetamide (IAA) were obtained from Sigma-Aldrich (Milan, Italy). High-glucose DMEM (Dulbecco's Modified Eagle Medium) and fetal bovine serum (FBS) were obtained from Gibco-Thermo Fisher Scientific (Milan, Italy). DMEM F12 and tetracycline-free FBS were obtained from Euro Clone. G418 and Doxycycline were obtained from Carlo Erba and Sigma-Aldrich, respectively. Trypsin/EDTA (for cell cultures) and penicillin-streptomycin (P/S) were purchased from Invitrogen. All other stock solutions for cell cultures were from Euroclone (Milan, Italy). Modified porcine trypsin and chymotrypsin were purchased from Promega (Milan, Italy). Water and acetonitrile (OPTIMA ® LC/MS grade) for LC/MS analyses were provided from Fisher Scientific (Milan, Italy).

Extraction of Mitochondrial Proteins from NSC34 Cells under Reducing Conditions
Extraction of mitochondrial proteins from NSC34 cells under reducing conditions was performed as described in [20].

Liquid Chromatography and Tandem Mass Spectrometry (LC-MS/MS) Analysis and Database Search
Mass spectrometry data were acquired and processed as described in [20].

Identification of Deamidation Sites on VDAC3
A freely available command-line script for Python 2.x (https://github.com/dblyon/ deamidation), accessed on 1 December 2021 which uses the MaxQuant "evidence.txt" file, was used to estimate the percentage of deamidation in VDAC 3 for each cell line, as in [20].

Expression, Purification and Refolding of Recombinant VDAC3 Proteins
The sequence encoding mouse VDAC3 (mVDAC3) was cloned in pET21a vector (Novagen) as reported in [23]. Mutagenesis of mVDAC3 was achieved using a specific couple of primers designed to replace Asn 215 residue with Asp according to the protocol already described in [61]. The single mutation was confirmed by DNA sequencing.

Lipid Bilayer Experiments
Planar lipid bilayer experiments were performed as described previously [62,63]. Artificial membranes made of 1% (w/v) diphytanoyl phosphatidylcholine (DiphPC) (Avanti Polar Lipids, Alabaster, AL, USA) in n-decane were painted on a 200 µm hole in a Derlin cuvette (Warner Instruments, Hamden, CT, USA). All the experiments were carried out at RT. Membrane capacitances of 100-150 pF were established for appropriate lipid bilayers. Mutant or native mVDAC3 were added from the protein stock solution of 1 mg/mL to the cis side of the cuvette filled with symmetrical 1 M KCl/10 mM HEPES pH 7.0. The single channel conductance of the pores was measured upon application of a fixed membrane potential (+10 mV) [58]. The voltage dependence was calculated by applying a triangular voltage ramp from 0 to ±50 mV of 100 ms duration, with a frequency of 10 mHz. At least three independent experiments were performed for each protein. A Bilayer Clamp amplifier (Warner Instruments) at 100 ms/point and filtered at 300 Hz was used for data acquisition. Analyses were performed with the pClamp software (Ver-10; Molecular Devices, San Jose, CA, USA).

Modelling and Bioinformatics Analysis
The structures of VDAC3 N215D shown in this work were obtained computationally with MODELLER software v9.24 [64] using the structure of human VDAC1 WT as a template. N215D mutations were introduced by substitution of the selected amino acid residue in the FASTA sequence. The same software was used for evaluation of the energetic score associated with each structure. Graphical representation was obtained by using VMD-Visual Molecular Dynamic software (available at: https://www.ks.uiuc.edu/ Research/vmd/), accessed on 1 January 2022. The root means square deviation (RMSD) analysis and the Ramachandran plots (RPs) were obtained both through the free online software at https://zlab.umassmed.edu/bu/rama/, accessed on 1 January 2022 and by using specific tools in the VMD software.

Conclusions
In this work, combining HRMS analysis with "in-solution" digestion of an enriched VDACs fraction, we found specific and irreversible PTMs in VDAC3 purified from the ALS model cell line. In particular, in VDAC3 we identified channel oxidation, deamidation and succination events and provided experimental evidence of functional changes of deamidated VDAC3. The impact of the most abundant deamidation event was verified by reconstituting its activity in a functional assay.
Overall, our data complete the picture we have drawn with previous results about VDAC1 from the same cell type. Specifically, the post-translational modifications identified and discussed here may represent sufficient conditions to alter the physiological pool of interactors of VDAC3 and modify its specific ability to buffer ROS, possibly impacting IMM redox signaling in ALS.