Disulfide‐mediated oligomerization of mutant Cu/Zn‐superoxide dismutase associated with canine degenerative myelopathy

Abstract

A homozygous E40K mutation in the gene coding canine Cu/Zn‐superoxide dismutase (cSOD1) causes degenerative myelopathy (DM) in dogs. A pathological hallmark of DM with the cSOD1 mutation is the aggregation of mutant cSOD1 proteins in neurons. The amino acid substitution E40K disrupts a salt bridge between Glu40 and Lys91 and is considered to destabilize the native state of cSOD1; however, the mechanism by which mutant cSOD1 aggregates remains unclear. Here, we show that mutant cSOD1 losing a copper and zinc ion forms oligomers crosslinked via disulfide bonds. The E40K substitution was found to result in the increased solvent exposure of the Cys7 side chain, which then attacked the disulfide bond (Cys57–Cys146) in cSOD1 to form disulfide‐linked oligomers. We also successfully prevented the Cys7 exposure and thus the oligomerization of mutant cSOD1 by a fragment antibody that specifically recognizes the region around the mutation site. The fragment antibody covered the β‐plug region, reinforcing the interactions compromised by the E40K substitution and thus contributing to the maintenance of the structural integrity of the β‐barrel core of cSOD1. Taken together, we propose that the Cys7 exposure in cSOD1 upon the salt bridge disruption plays a central role in the aggregation mechanism of DM‐associated mutant cSOD1.

1. INTRODUCTION

A homozygous E40K mutation in the gene encoding canine Cu/Zn‐superoxide dismutase (cSOD1) is linked to degenerative myelopathy (DM), a progressive neurodegenerative disease in dogs that results in neuron loss, paralysis, and death several years after the onset of symptoms (Awano et al. 2009). While a pathomechanism of DM has not been fully understood, immunohistochemical analysis of spinal cords from DM‐affected dogs has revealed abnormal accumulation of cSOD1 in neurons (Awano et al. 2009). In addition, detergent‐insoluble cSOD1 in spinal cord has been reported to correlate with disease severity (Crisp et al. 2013). E40K‐mutant cSOD1 (cSOD1^E40K) is hence expected to gain toxicity upon aggregation; however, a mechanism by which the E40K mutation leads to the cSOD1 aggregation remains unknown.

The pathological role of mutant cSOD1 in DM has recently received increasing attention, as DM is likely to serve as a reliable model for amyotrophic lateral sclerosis (ALS), a fatal neurodegenerative disease in human (Nardone et al. 2016): numerous mutations in the gene coding human Cu/Zn‐superoxide dismutase (hSOD1) underlie a subset of familial ALS cases (SOD1‐ALS) (Rosen et al. 1993). Pathological and clinical differences exist between ALS and DM, such as the mode of inheritance, which is typically autosomal dominant in most SOD1‐ALS cases and mostly autosomal recessive in DM. Compared to DM, furthermore, ALS is characterized by more significant loss of motor neurons and less involvement of proprioceptive pathways (Crisp et al. 2013). Nonetheless, the presence of cytoplasmic inclusions containing mutant hSOD1/cSOD1 in motor neurons is a pathological hallmark common to both SOD1‐ALS and DM cases (Bruijn et al. 1998). The cytoplasmic inclusions have also been reproduced in rodent models overexpressing mutant hSOD1 (Turner and Talbot 2008), which might, however, be artificially induced due to the extremely high levels of mutant hSOD1. In this sense, DM‐affected dogs are a naturally occurring animal model for ALS with two copies of mutant cSOD1 gene, making the aggregation of cSOD1 in DM a more faithful representation of hSOD1 aggregation in ALS.

Both hSOD1 and cSOD1 (collectively referred to as SOD1) are homodimeric proteins in which each subunit binds a copper and zinc ion and also forms an intra‐subunit disulfide bond between Cys57 and Cys146 (Figure 1). The function of SOD1 as an enzyme is to disproportionate superoxide radicals to oxygen and hydrogen peroxide at the bound copper ion, while the zinc binding and disulfide formation play critical roles in stabilizing the enzymatically competent conformation of SOD1 (Sheng et al. 2014). hSOD1 has been shown to experience significant structural destabilization upon the metal dissociation and/or the disulfide reduction, which subsequently misfolds the protein into soluble oligomers and/or insoluble aggregates (Furukawa 2021; Shaw and Valentine ²⁰⁰⁷; Trist et al. ²⁰²¹). These oligomers and aggregates appear to describe abnormal conformations of hSOD1 proteins in the SOD1‐ALS patients as well as the rodent models (Anzai et al. 2017; Deng et al. ²⁰⁰⁶; Furukawa et al. ²⁰⁰⁶; Tokuda et al. ²⁰¹⁷), while their pathogenic roles need to be investigated in more detail.

FIGURE 1.

Three‐dimensional structure and amino acid sequence of cSOD1. (a) In the homodimer of cSOD1, a copper (cyan) and zinc (magenta) ion in each of the subunits are represented using a sphere model (PDB ID: 7WWT). The conserved disulfide bond between Cys57 and Cys146, as well as the free thiol group of Cys7, in the subunit displayed on the right are colored yellow. Additionally, Glu40 and Lys91 are shown in a stick model. (b) An amino acid sequence alignment of human and canine SOD1 is performed with CLUSTALW. The amino acid residues responsible for the binding of a copper and zinc ion and the formation of the disulfide bond are indicated as Cu, Zn, and S‐S, respectively. (c) The amino acid substitutions linked to ALS in hSOD1 that are in proximity to E40 and K91 are shown. These substitutions have been registered in the ALSoD database (https://alsod.ac.uk/).

cSOD1 has 82% identity in amino acids with hSOD1, and Glu40 is expected to form a salt bridge with Lys91 in both proteins (Figure 1a,b). The substitution E40K disrupts the salt bridge and is thus considered to destabilize and then misfold cSOD1; indeed, the cytoplasmic aggregation of cSOD1^E40K was significantly suppressed by the re‐introduction of a salt bridge by the additional substitution K91E (Hashimoto et al. 2023). Interestingly, only cSOD1^E40K but not hSOD1 with E40K was found to form cytoplasmic inclusions when expressed in cultured cells (Hashimoto et al. 2023), while several ALS‐linked substitutions in hSOD1 have been known in close proximity to E40 and K91 (Figure 1c). Compared to hSOD1, cSOD1 is considered to have an intrinsically more vulnerable structure to the E40K substitution, partly due to imperfect structural packing in the hydrophobic core of the central β‐barrel structure of cSOD1 (Hashimoto et al. 2023). Nonetheless, it remains unknown how the E40K substitution leads to the aggregation of cSOD1.

In this study, we have shown that cSOD1^E40K forms disulfide‐linked oligomers upon loss of a copper and zinc ion at physiological temperature. The E40K substitution was found to result in the increased exposure of the Cys7 side chain, allowing Cys7 in one subunit to attack the conserved disulfide bond between Cys57 and Cys146 in another subunit. Such disulfide shuffling between the subunits then led to the formation of the disulfide‐linked cSOD1 oligomers. Furthermore, the oligomerization of cSOD1^E40K was mitigated by a fragment antibody covering the region around the E40K substitution site. Based on the experimental results presented here, we propose that disruption of the salt bridge between Glu40 and Lys91 increases the solvent exposure of Cys7 through the destabilization of the β‐barrel core structure, thereby triggering the formation of disulfide‐shuffled cSOD1 oligomers.

2. RESULTS

2.1. cSOD1^E40K lacking metal ions forms disulfide‐crosslinked oligomers

To test the E40K‐induced aggregation of cSOD1, we purified recombinant cSOD1 proteins and incubated those at 37°C for 9 days in the presence of both copper and zinc ion (i.e., in the holo state); nonetheless, the Western blotting analysis of both cSOD1^E40K and wild‐type cSOD1 (cSOD1^WT) that were electrophoresed under non‐reducing conditions showed a single band corresponding to denatured cSOD1 monomer (Figure 2a). In contrast, in the absence of both copper and zinc ion (i.e., in the apo state), cSOD1^E40K but not cSOD1^WT showed smear in the high molecular weight region under the non‐reducing conditions that disappeared after treated with a reductant, β‐mercaptoethanol (Figure 2b,c). Furthermore, when alanine was substituted for Cys7, the only freely available Cys in cSOD1^E40K, the smear in the high molecular weight region was not observed, even in the absence of the reductant (Figure 2d). These results thus suggest that, at physiological temperature, apo‐cSOD1^E40K, but not cSOD1^WT, forms the disulfide‐crosslinked oligomers in which Cys7 plays a critical role.

FIGURE 2.

Formation of the disulfide‐crosslinked oligomers in the apo‐form of cSOD1^E40K. cSOD1 (WT, E40K, E40K/K91E, C7A/E40K) proteins either in (a) the holo form or (b–d) the apo form at the concentration of 50 μM were incubated at 37°C for 9 days. Followed by the electrophoretic separation in a 10–20% gradient polyacrylamide gel under non‐reducing (a, b, and left panel of d) or reducing (c and right panel of d, in the presence of β‐mercaptoethanol, β‐ME) conditions, cSOD1 proteins were detected by Western blotting with a polyclonal anti‐SOD1 antibody.

A salt bridge interaction was considered to be formed between Glu40 and Lys91 in cSOD1^WT but was lost in cSOD1^E40K. To investigate the influence of this salt bridge interaction on the oligomerization propensity of cSOD1, we reintroduced the interaction in cSOD1^E40K by engineering a K91E substitution (cSOD1^E40K/K91E). As shown in Figure 2b,c, the disulfide‐crosslinked oligomers were absent in apo‐cSOD1^E40K/K91E incubated at 37°C for 9 days. Therefore, the disruption of the salt bridge interaction between Glu40 and Lys91 is considered to be involved in the oligomerization of cSOD1^E40K in the apo state.

2.2. E40K substitution increases flexibility of Cys7 and triggers the formation of disulfide‐linked oligomers

To understand how the disruption of the salt bridge in cSOD1^E40K leads to the formation of the disulfide‐crosslinked oligomer, we first investigated effects of the salt‐bridge interaction on the structural stability of cSOD1 by differential scanning fluorometry (DSF). A melting temperature (T _m ) of cSOD1^WT in the apo state was 53°C, which was found to be decreased to 49°C by the E40K substitution and then restored to 53°C by re‐introducing the salt‐bridge interaction with the K91E substitution in cSOD1^E40K (i.e., cSOD1^E40K/K91E) (upper panel of Figure 3a). DSF monitors the fluorescence from a dye, SYPRO Orange, that becomes fluorescent when bound to hydrophobic regions in a protein exposed upon thermal denaturation. cSOD1 with the salt‐bridge interaction (cSOD1^WT and cSOD1^E40K/K91E) was thus anticipated to begin denaturation around the temperature higher than 49°C, which was dropped to 43°C in cSOD1^E40K (lower panel of Figure 3a). Therefore, the E40K substitution decreased the thermal stability of cSOD1 in the apo state, possibly leading to increased structural flexibility and eventually trigger the oligomerization at 37°C.

FIGURE 3.

Conformational destabilization of cSOD1 by the E40K substitution. (a) Thermal stability of cSOD1 was examined by DSF using a fluorescent dye, SYPRO Orange. The lower panel shows the temperature‐dependent changes in the fluorescence intensity (see Experimental procedures of Data S1), while the upper panel represents the normalized derivative of fluorescence intensity with respective to temperature. In both panels, the curves for cSOD1^WT, cSOD1^E40K, and cSOD1^E40K/K91E in the apo state were colored black, red, and blue, respectively. (b, c) SDS‐PAGE analysis of the modification of cSOD1 variants in the (b) holo and (c) apo state with mPEG‐MAL was shown. The cSOD1 samples at a concentration of 20 μM were incubated with 0.1 mM mPEG‐MAL at the temperature indicated on top of the gel and electrophoretically separated in 15% polyacrylamide gel.

For further elucidation of the oligomerization mechanism, we next noted the disulfide crosslinks in the oligomers of cSOD1^E40K. cSOD1 possesses a total of three cysteine residues (Cys7, Cys57, and Cys146), among which Cys57 and Cys146 form the intra‐subunit disulfide bond; therefore, only Cys7 can initiate the crosslinking reaction (Figure 1). Importantly, the disulfide‐crosslinked oligomers were observed as smear in the high‐molecular‐weight region (Figure 2a), suggesting that they would form not by simple oxidation at Cys7 but through the shuffling of the disulfide bond between cSOD1^E40K molecules. In the native structure of cSOD1, nonetheless, the thiol group of Cys7 is buried within the β‐barrel structure (Figure 1) and is thus considered to be difficult to participate in the disulfide shuffling. We thus investigated the availability of the thiol group at Cys7 in cSOD1 by using mPEG‐MAL, a bulky thiol‐specific modifier derived from polyethylene glycol with maleimide.

All three cSOD1 variants (cSOD1^WT, cSOD1^E40K, cSOD1^E40K/K91E) in the holo state were not modified with mPEG‐MAL at temperatures from 20 to 50°C (Figure 3b), consistent with the unavailability of Cys7 for modification due to its burial within the β‐barrel structure characterized by the extremely high thermostability (~90°C of T _m , Figure S1, Supporting Information). In contrast, in the apo state, nearly all of cSOD1 molecules were modified with mPEG‐MAL at 50°C, which was close to T _m obtained by DSF (Figure 3c). More notably, the modification with mPEG‐MAL was significant at 37°C in apo‐cSOD1^E40K but not in apo‐cSOD1^WT and apo‐cSOD1^E40K/K91E (Figure 3c), suggesting the increased structural fluctuation around Cys7 by the E40K substitution. Even at physiological temperature, in other words, the thiol/thiolate group of Cys7 in cSOD1^E40K but not cSOD1^WT is considered to become exposed and thereby available for shuffling the intra‐subunit disulfide bond. Since the salt‐bridging residues Glu40 and Lys91 are located in the loop regions (Loop III and Loop V, respectively) (Figure 1), the disruption of this salt bridge is likely to increase the flexibility of those loops. While it remains obscure how the side chain of Cys7 become more flexible upon losing the salt bridge interaction, we expected that reducing the flexibility of the loop regions could prevent the oligomerization of cSOD1^E40K.

2.3. Development of antibody fragment recognizing cSOD1^E40K

Recently, antibodies have been used to reduce the conformational flexibility of proteins and protein complexes by binding them tightly, thus facilitating their structural analysis (Maeda et al. 2018). Inspired by this use of antibodies to reduce conformational flexibility, we investigated whether the oligomerization of cSOD1^E40K could be prevented by specific complexation with an antibody. For this purpose, a monoclonal antibody specifically recognizing cSOD1^E40K was developed by immunizing mice with purified GST‐fused cSOD1^E40K proteins, preparing hybridoma cells from the isolated splenocytes, and then screening clones that bind cSOD1^E40K but not cSOD1^WT by ELISA. As shown in Figure 4a,b, the clone 22E1 was confirmed to bind cSOD1^E40K with significantly higher affinity than cSOD1^WT both in the holo and apo state.

FIGURE 4.

A monoclonal antibody 22E1 selectively binds cSOD1^E40K. (a, b) The binding affinities of 22E1 to cSOD1 variants in the holo (a) and apo (b) state were assessed by indirect ELISA (left panels). The total amounts of cSOD1 variants captured on the wells were quantified using a polyclonal anti‐SOD1 antibody (right panels). For holo‐cSOD1^E40K detected by 22E1, the substrate OPD exhibited rapid color development, necessitating a shortened development time of 5 min (left panel in (a)), compared to the standard 30 min for the other assay conditions. Each experiment was conducted in triplicate, with error bars representing standard error. (c) Double fluorescence immunohistochemical analysis was performed on the caudal cervical spinal cord sections of two Pembroke Welsh Corgis (PWCs) diagnosed with degenerative myelopathy (DM, E40K homozygotes, depicted in the upper and middle rows of panels) and one asymptomatic Beagle (wild‐type, WT homozygotes, shown in the lower row of panels). The sections were co‐stained with a monoclonal antibody 22E1 (green) and a polyclonal anti‐SOD1 antibody, SOD‐100 (red), along with nuclear counterstaining using DAPI (blue). Notably, SOD‐100 prominently labeled motoneurons in both DM and WT spinal cords (arrows in the central column of panels). However, 22E1 reactivity was observed in the motoneurons of DM cases (arrows in the left column of panels) but was absent in the WT case. Additionally, inclusions were identified within the motoneurons of the DM cases using SOD‐100 staining (arrowheads in the central column of panels).

To examine the detection of cSOD1^E40K in vivo with 22E1, furthermore, we performed co‐staining using immunohistochemistry with 22E1 and polyclonal anti‐SOD1 antibody (SOD‐100) on the spinal cord sections of dogs with and without the E40K mutation (Figure 4c). In the DM dogs with the E40K mutation, SOD‐100 detected the cytoplasmic inclusions where cSOD1^E40K accumulated, while diffuse cytoplasmic staining was observed in the wild‐type dogs without the mutation under identical experimental conditions (Figure 4c, a middle column). In contrast, 22E1 was found to diffusely stain the cytoplasm in the DM dogs and did not react with the inclusions detected by SOD‐100 (Figure 4c, a left column). No staining with 22E1 was also observed in the spinal cords of the wild‐type dogs. These immunohistochemical data suggest that 22E1 preferentially binds to a conformational epitope of cSOD1^E40K that is available prior to its aggregation into pathological inclusions.

2.4. Antibody‐mediated suppression of cSOD1^E40K oligomerization

To test if 22E1 has the ability to suppress the oligomerization of cSOD1^E40K in vitro, recombinant 22E1 was prepared in the Fv‐clasp format (22E1^Fv‐clasp). This format not only enables large‐scale protein production but also facilitates the structural analysis of the complex with an antigen (Arimori et al. 2017). 22E1^Fv‐clasp is composed of the V_H and V_L domain of 22E1, each of which are C‐terminally fused with the SARAH domain (22E1(V_H)‐SARAH and 22E1(V_L)‐SARAH, respectively). Through the SARAH domains, the interaction between the V_H and V_L domain is known to be stabilized (Arimori et al. 2017).

As noted before, the apo form of cSOD1^E40K appeared as a smear in the high‐molecular weight region of the Western blot under non‐reducing conditions, suggesting the formation of disulfide‐crosslinked oligomers (Figures 2 and 5a). Notably, incubation of apo cSOD1^E40K with an equimolar amount of 22E1^Fv‐clasp resulted in the disappearance of the smear (Figure 5a), suggesting the prevention of the disulfide‐crosslinked cSOD1^E40K oligomerization by 22E1^Fv‐clasp. In contrast, the addition of an unrelated Fv‐clasp molecule, 12CA5^Fv‐clasp—known for recognizing the HA tag (Arimori et al. 2017)—did not affect the oligomerization of cSOD1^E40K (Figure 5a). Furthermore, the modification of Cys7 in cSOD1^E40K by mPEG‐MAL at 37°C was almost completely suppressed in the presence of 22E1^Fv‐clasp but not 12CA5^Fv‐clasp (Figure 5b). Collectively, these results demonstrate the ability of 22E1^Fv‐clasp to reduce the flexibility of Cys7 and thereby mitigate the oligomerization of cSOD1^E40K in its apo state.

FIGURE 5.

Inhibition of cSOD1^E40K oligomerization by 22E1^Fv‐clasp. (a) cSOD1^E40K was prepared at the concentration of 50 μM in its apo state and incubated at 37°C for 9 days, both alone and with equimolar quantities of either 22E1^Fv‐clasp or 12CA5^Fv‐clasp. Followed by the electrophoretic separation in a 10–20% gradient polyacrylamide gel under non‐reducing (left panel) or reducing (right panel, in the presence of β‐mercaptoethanol, β‐ME) conditions, cSOD1 proteins were detected by Western blotting with a polyclonal anti‐SOD1 antibody. (b) Apo‐cSOD1^E40K in the concentration of 20 μM, both alone and with equimolar quantities of either 22E1^Fv‐clasp or 12CA5^Fv‐clasp, were incubated with 0.1 mM mPEG‐MAL at the temperature indicated on top of the gel and electrophoretically separated in a 15% polyacrylamide gel under reducing conditions.

2.5. 22E1^{Fv ‐clasp} forms a tight complex with cSOD1^E40K in solution

To elucidate the mechanism by which 22E1^Fv‐clasp prevented the oligomerization of cSOD1^E40K, the interactions between 22E1^Fv‐clasp and cSOD1 variants were analyzed by size exclusion chromatography (SEC). The elution profiles of cSOD1^E40K and 22E1^Fv‐clasp were first characterized by SDS‐PAGE after their separation by a gel filtration column, where fractions were collected every minute. The peak elution for both proteins was observed in the fraction denoted as “21 min,” corresponding to the eluate collected from 21 to 22 min (Figure 6a). We next examined the equimolar mixture of cSOD1^E40K and 22E1^Fv‐clasp by SEC; the SDS‐PAGE analysis of the eluted fractions indicated that both proteins co‐eluted in the “18 min” fraction (Figure 6b), which is earlier than that observed for each protein alone (i.e., 21 min), suggesting the complexation between cSOD1^E40K and 22E1^Fv‐clasp. Based upon these results, we assumed that each subunit of the cSOD1^E40K homodimer binds one 22E1^Fv‐clasp molecule; consistent with this assumption, indeed, a molar mass of 99.4 kDa (31.6 and 37.4 kDa for the calculated mass of a cSOD1^E40K homodimer and 22E1^Fv‐clasp, respectively) was obtained by SEC with multi‐angle light scattering (SEC‐MALS) (Figure S2a).

FIGURE 6.

SEC analysis of the complex formation between cSOD1^E40K and 22E1^Fv‐clasp. A protein sample in a concentration of 10 μM in the MN buffer was separated by a gel‐filtration column at a flow rate of 0.5 mL/min and fractionated every minute. Followed by the modification with iodoacetamide, the fractions were analyzed by SDS‐PAGE using a 15% polyacrylamide gel under non‐reducing conditions. The proteins examined are (a) cSOD1^WT, cSOD1^E40K, cSOD1^E40K/K91E, or 22E1^Fv‐clasp alone, and an equimolar mixture of (b) cSOD1^E40K and 22E1^Fv‐clasp, (c) cSOD1^WT and 22E1^Fv‐clasp, and (d) cSOD1^E40K/K91E and 22E1^Fv‐clasp. The time labeled at the top of the gel indicates when collection of the corresponding fraction started.

When cSOD1^WT was mixed with an equimolar amount of 22E1^Fv‐clasp, the fraction of 19 min contained both cSOD1^WT and 22E1^Fv‐clasp (Figure 6c), which is slightly later than that observed for the cSOD1^E40K and 22E1^Fv‐clasp complex. A significantly more amount of 22E1^Fv‐clasp than that of cSOD1^WT was also detected in the fraction of 21 min (Figure 6c), corresponding to the peak fraction of 22E1^Fv‐clasp alone. Additionally, cSOD1^WT was observed in the fraction of 20 min, which again corresponds to the peak fraction of cSOD1^WT alone (Figure 6a,c). These results are consistent with the idea that a cSOD1^WT homodimer would bind one molecule of 22E1^Fv‐clasp; indeed, our SEC‐MALS analysis characterized the elution peak at 19 min by a molar mass of 51.7 kDa, which is close to that of the cSOD1^WT homodimer (31.6 kDa) with one molecule of 22E1^Fv‐clasp (37.1 kDa) (Figure S2b).

While cSOD1^E40K/K91E is expected to behave like cSOD1^WT in the sense that the salt‐bridge interaction is expected between the position 40 and 91 in both proteins, the SDS‐PAGE analysis on the fractions separated by SEC did not support the complexation between 22E1^Fv‐clasp and cSOD1^E40K/K91E (Figure 6a,d). We thus supposed that Lys91 on cSOD1 plays a critical role in the interaction with 22E1^Fv‐clasp. To get more insight into the mechanism of the recognition of cSOD1^E40K by 22E1^Fv‐clasp, we further performed the crystal structural analysis on the 22E1^Fv‐clasp‐cSOD1^E40K complex.

2.6. Crystal structural analysis of the specific interaction between cSOD1^E40K and 22E1^{Fv ‐clasp}

Figure 7a shows the crystal structure of the complex between the holo‐form of cSOD1^E40K and 22E1^Fv‐clasp (PDB ID: 8ZD5). Each of the cSOD1^E40K subunits was found to bind one 22E1^Fv‐clasp molecule, and the two structural units, each of which are composed of the cSOD1^E40K subunit, 22E1(V_H)‐SARAH, and 22E1(V_L)‐SARAH, were indistinguishable in their binding sites and relative geometries. The structure of the cSOD1^E40K homodimer was almost completely overlapped with and without the binding of 22E1^Fv‐clasp (Figure S3), suggesting that 22E1^Fv‐clasp did not significantly affect the structure of cSOD1^E40K. As schematically depicted in Figure 7b, the interface between the cSOD1^E40K subunits and that between the V_H and V_L subunit of 22E1^Fv‐clasp covers 720 and 730 Ų, respectively. The cSOD1^E40K subunit was bound by both of the V_H and V_L subunit of 22E1^Fv‐clasp with the interface area of 230 and 340 Ų, respectively. 22E1^Fv‐clasp was found to utilize the complementarity‐determining regions (CDRs) of both the V_H and V_L subunit for the binding of cSOD1^E40K and also considered to recognize Loop I, Loop III (Lys40), and Loop V (Lys91) (Figure 7c).

FIGURE 7.

Crystal structure of the complex between cSOD1^E40K and 22E1^Fv‐clasp (PDB ID: 8ZD5). (a) An overall structure of the cSOD1^E40K‐22E1^Fv‐clasp complex is shown: cSOD1^E40K, 22E1(V_H)‐SARAH, and 22E1(V_L)‐SARAH in the complex is colored in gray, light green, and light pink, respectively: CDRs of 22E1(V_H)‐SARAH in dark green and those of 22E1(V_L)‐SARAH in dark pink. Copper and zinc ion in cSOD1^E40K are represented as spheres in cyan and magenta, respectively. Lys40 and Lys91 in cSOD1^E40K shown as a stick model are also indicated. (b) The interaction between the cSOD1^E40K subunit and 22E1^Fv‐clasp is schematically represented with the respective interface areas for each pair shown in the figure. (c) A close‐up view of the interface between cSOD1^E40K and 22E1^Fv‐clasp is shown with CDRs of 22E1(V_H)‐SARAH in dark green and those of 22E1(V_L)‐SARAH in dark pink. Together with Lys40 and Lys91 in a stick model, Loops I, III, and V of cSOD1^E40K are also indicated.

The E40K substitution significantly increased the binding affinity of cSOD1 to 22E1^Fv‐clasp; consistent with this, the Nε atom of Lys40 in cSOD1^E40K appeared to form the hydrogen‐bonding interaction with the carbonyl oxygen of the Asp31 main chain (3.08 Å) in 22E1(V_H)‐SARAH (Figure 8a). Besides, the Nε atom of Lys91 in cSOD1^E40K was found to be close enough to the carboxyl oxygen of the Asp52 side chain (2.70 Å) and also the carbonyl oxygen of the Tyr32 main chain (2.94 Å) in 22E1(V_H)‐SARAH (Figure 8a), both of which were confirmed in the electron density map. As represented by a close‐up view of the interaction site around Lys40 and Lys91 that is aligned with the corresponding region in cSOD1^WT and cSOD1^E40K (Figure 8c), the orientation of the Lys91 side chain was notably distinct among those structures. The Nε atom of Lys91 in cSOD1^WT formed the salt‐bridge interaction with the carboxy group of Glu40, while the salt‐bridge was lost by the E40K substitution, resulting in the increased conformational flexibility of the Lys91 side chain. In the complex with 22E1^Fv‐clasp, the Lys91 side chain was oriented outward, by which the Nε atom interacted with the carboxyl group of Asp52 and the carbonyl oxygen of Tyr32. No binding of cSOD1^E40K/K91E with 22E1^Fv‐clasp, as described above (Figure 6d), is also consistent with the critical interactions of Lys91 at the interface with 22E1^Fv‐clasp. Collectively, the interactions involving the Lys40 and Lys91 side chain in cSOD1^E40K would contribute to the specific recognition by 22E1^Fv‐clasp.

FIGURE 8.

Interactions between cSOD1^E40K and 22E1^Fv‐clasp in their complex. (a, b) Hydrogen‐bonding interactions of cSOD1^E40K in gray with the CDRs of (a) 22E1(V_H)‐SARAH in dark green and (b) 22E1(V_L)‐SARAH in dark pink are shown as dotted lines. (c) The orientation of the Lys91 side chain is compared among cSOD1^WT alone (sky blue), cSOD1^E40K alone (orange), and cSOD1^E40K complexed with 22E1^Fv‐clasp (gray). The hydrogen‐bonding interactions between Glu40 and Lys91 in cSOD1^WT are represented as dotted lines. 22E1(V_H)‐SARAH is shown in dark green. Note that the side chain of Lys91 in the crystal structure of cSOD1^E40K alone is represented even though its electron density is not well‐defined, suggesting that the side chain is not in a fixed orientation but is highly flexible.

While the Lys40 and Lys91 side chain are recognized by the V_H domain of 22E1^Fv‐clasp, the V_L domain also plays important roles in the binding of cSOD1^E40K. The guanidino group of Arg92 in 22E1(V_L)‐SARAH forms hydrogen bonding interactions with the carboxylate of Glu16 in cSOD1^E40K (Figure 8b). Hydrogen bonding interactions were also expected between the Nε atom of Lys30 in 22E1(V_L)‐SARAH and the carbonyl groups of the Gly11 and/or Gly13 main chain in cSOD1^E40K and between the hydroxy group of Tyr50 in 22E1(V_L)‐SARAH and the carbonyl group of the Leu38 main chain in cSOD1^E40K (Figure 8b). Gly11/13 and Glu16 in cSOD1^E40K are situated within a loop region and a β‐strand, respectively, both of which are in direct contact with the N‐terminal β‐strand containing Cys7; therefore, those interactions might be important for inhibiting the disulfide cross‐linked oligomerization of cSOD1^E40K by reducing the fluctuation of Cys7.

We also analyzed a crystal structure of the complex between 22E1^Fv‐clasp and apo‐cSOD1^E40K (PDB ID: 8ZD6). The main chain structure of the complex was found not to significantly depend upon whether cSOD1^E40K was in the apo or holo state, while the electron density was not obtained in many of the side chains (Figure S4). In one subunit of apo‐cSOD1^E40K, structures were not determined for two loop regions (Lys69‐Asp76 and Asp124‐Ser142) often called as the zinc‐binding and electrostatic loop, respectively. Furthermore, the electron density map did not clearly resolve the side chains of Lys40 and Lys91 in apo‐cSOD1^E40K, which are critical for the specific recognition by 22E1^Fv‐clasp. Nonetheless, the notable structural similarity in the 22E1^Fv‐clasp complex with the apo and holo form of cSOD1^E40K implies that the interactions involving Lys40 and Lys91 are likely retained in the complex between 22E1^Fv‐clasp and apo‐cSOD1^E40K.

The other interactions between 22E1(V_L)‐SARAH and holo‐cSOD1^E40K were also considered to be maintained in the apo‐cSOD1^E40K complex. While the side chain of Lys30 in 22E1(V_L)‐SARAH, which interacted with Gly11/13 in holo‐cSOD1^E40K, was not resolved in the electron density map, the hydrogen bonding interaction between Arg92 (22E1(V_L)‐SARAH) and Glu16 (apo‐cSOD1^E40K) and also the interaction between Tyr50 (22E1(V_L)‐SARAH) and Leu38 (apo‐cSOD1^E40K) were confirmed.

Taken together, 22E1^Fv‐clasp can recognize cSOD1^E40K both in the apo and holo state, with more structural fluctuation in the apo state. 22E1^Fv‐clasp can discern cSOD1^E40K from cSOD1^WT by recognizing the side chain of Lys40 and Lys91, both of which are not available in cSOD1^WT. 22E1^Fv‐clasp is also found to interact with Gly11/13 and Glu16 in cSOD1^E40K, which would suppress the fluctuation of Cys7. These interactions could hence provide a structural basis of how 22E1^Fv‐clasp prevents the oligomerization of cSOD1^E40K through the disulfide shuffling.

3. DISCUSSION

In this study, we have revealed that the E40K mutation in cSOD1 increases the availability of Cys7 for shuffling the disulfide bond in the apo state, leading to the formation of the disulfide‐crosslinked oligomers. Also, the oligomerization of cSOD1^E40K was found to be suppressed by the fragment antibody 22E1^Fv‐clasp that specifically recognized a folded form of cSOD1^E40K but not cSOD1^WT. The structural analysis of the complex between cSOD1^E40K and 22E1^Fv‐clasp has offered a mechanistic description for the reduced flexibility of Cys7 upon the specific interaction with 22E1^Fv‐clasp. Therefore, restricting the flexibility of Cys7 will be a crucial factor in preventing the pathological oligomerization implicated in DM.

The disulfide‐crosslinked oligomers of SOD1 have been identified in the spinal cords of affected ALS‐model mice overexpressing mutant hSOD1 (Anzai et al. 2017; Furukawa et al. ²⁰⁰⁶) but were not clearly detected in the spinal cord tissues of the ALS patients (Tokuda et al. 2017). This is probably due to the limited amounts of pathological hSOD1 species present exclusively in the motoneurons of ALS patients, where only one or two copies of the mutant SOD1 gene are expressed. A similar observation was made in DM; indeed, we were unable to detect disulfide‐crosslinked cSOD1^E40K oligomers in the spinal cord tissues of dogs affected with DM (data not shown). Nonetheless, the experiments in vitro with recombinant proteins have demonstrated that ALS‐causing mutations decrease the conformational stability of hSOD1 in the apo state (Furukawa and O'Halloran ²⁰⁰⁵; Rodriguez et al. ²⁰⁰⁵), leading to the emergence of a “misfolded dimer”—a state where hSOD1 retains a dimeric configuration but lacks most of its secondary structures (Anzai et al. 2017). This conformational disruption facilitates the exposure of free Cys residues (i.e., Cys6 and Cys111), which then actively participate in the shuffling of the conserved disulfide bond between Cys57 and Cys146. It is also noteworthy that while Cys111 is unique to hSOD1, Cys6, equivalent to Cys7 in cSOD1, plays a pivotal role in the formation of disulfide‐crosslinked oligomers (Toichi et al. 2013). Cys111 is highly solvent‐exposed in both apo and holo form of hSOD1; in contrast, the solvent accessibility of Cys6 is increased by losing the bound metal ions (Banci et al. 2009). It is hence conceivable that the dynamic roles of Cys6 (hSOD1) and Cys7 (cSOD1) represent a key commonality in the misfolding pathways of SOD1 implicated in both ALS and DM.

We showed here that the solvent accessibility of Cys7 in apo‐cSOD1 was significantly increased by the E40K mutation (Figure 3c). Glu40 in cSOD1 forms the salt‐bridge interactions with Lys91, and both of those residues are conserved in most of mammalian SOD1 proteins including hSOD1. Interestingly, equine SOD1 has Lys at position 40 but also substitutes Glu for the residue at position 91, reinforcing the importance of the salt bridge interactions between those corresponding sites for the structure of SOD1 (Draper et al. 2020). The residues at position 40 and 91 are a part of the Loop III (38–40) and V (90–94), respectively (Figure 1); these loops form a “plug” for the β‐barrel structure of SOD1, which is called as β‐plug (Hart et al. 1998). It has been shown that the electrostatic/hydrophobic interactions within the β‐plug significantly contribute to the dynamic properties of hSOD1. For instance, a NMR study has shown that the G37R mutation, which is a pathogenic variant associated with ALS and occurs in close proximity to the Loop III, does not substantially alter the overall architecture of apo‐hSOD1; however, it does influence the conformational features of the residues constituting the β‐plug (Banci et al. 2007). Also, increased flexibility of the region including the Loop III has been suggested in apo‐hSOD1 with another pathogenic mutation G93R, which is situated within the Loop V (Durazo et al. 2009). According to the molecular dynamics simulation, furthermore, ALS‐causing mutations occurring outside the β‐plug are expected to destabilize the β‐plug in apo‐hSOD1, resulting in a transient opening of the β‐barrel (Khare and Dokholyan 2006). Therefore, the E40K mutation in cSOD1 would compromise the conformational integrity of the β‐plug, which in turn perturbs the stability of the β‐barrel structure.

It is interesting to note that the side chain of Cys6 in hSOD1 and Cys7 in cSOD1 contribute to the hydrophobic core within the β‐barrel structure of SOD1 (Figure 1). Unlike Cys6 of hSOD1, which aligns almost perfectly with the side chain of Leu117, Cys7 in cSOD1 is in contact with a cavity due to a suboptimal interaction with the side chain of Met117 (Hashimoto et al. 2023). Consequently, the β‐barrel in cSOD1 is considered to be inherently less stable due to the presence of such a cavity. Also, this intrinsic instability of cSOD1 may render it more vulnerable to additional destabilizing influences from conformational alterations in the β‐plug, thereby increasing the likelihood of Cys7 to facilitate the disulfide shuffling.

Our present study is the first to elucidate the crystal structure of the SOD1‐antibody complex, revealing that the β‐plug of cSOD1^E40K is a critical determinant regulating the binding specificity of our 22E1 antibody. Notably, there are several precedents of monoclonal antibodies that specifically recognize non‐native forms of hSOD1; according to the epitope mapping, some of those precedents are expected to bind the β‐plug region of SOD1. For example, the D3H5 monoclonal antibody recognizing the ALS‐associated mutant but not wild‐type hSOD1 has been reported to bind the region between Glu24 and Ala55 (Gros‐Louis et al. ²⁰¹⁰; Rotunno and Bosco ²⁰¹³), encompassing Loop III that forms the β‐plug. C4F6 is another monoclonal antibody that selectively recognizes misfolded hSOD1, and the key residues of the epitope for the hSOD1‐C4F6 interaction include Asp92 located within Loop V constituting the β‐plug (Rotunno et al. 2014). As confirmed in the cSOD1^E40K‐22E1^Fv‐clasp complex, therefore, D3H5 and C4F6 might also bind mutant hSOD1 by covering the β‐plug region. Given that the β‐plug is considered to act as a shield to block solvent entry into the β‐barrel core (Deng et al. 1993; Hart et al. ¹⁹⁹⁸; Khare and Dokholyan ²⁰⁰⁶), those antibodies including our 22E1 could stabilize mutant SOD1 by preventing the entry of solvent molecules into the hydrophobic core. Additionally, the Loop V bridges the β5 and β6 strand, both of which constitute the edge strands of the β‐barrel structure in cSOD1. As reported previously (Khare and Dokholyan 2006; Richardson and Richardson ²⁰⁰²), the edge strands have fewer inter‐strand interactions and are generally considered to be the region triggering aberrant interactions that eventually lead to pathological aggregation. Stabilizing the region around the Loop V could thus inhibit the propensity of the edge strands, β5 and β6, to engage in abnormal interactions. Indeed, a passive immunization with D3H5 and C4F6 has been known to exhibit therapeutic benefits in ALS‐model mice, albeit with mechanisms that have not yet been characterized in detail (Gros‐Louis et al. ²⁰¹⁰; Urushitani et al. ²⁰⁰⁷).

Our 22E1 antibody is also found to overlay the region of Loop I (Gly11‐Val15) and have the hydrogen bonding interactions with Gly11/13 and Glu16 of cSOD1^E40K (Figures 7c and 8b). The flexibility of the Loop I connecting β1 and β2 strand would affect the availability of Cys7 in the β1 strand (Figure 1); therefore, Loop I and its neighboring structural elements are expected to be the important target site to inhibit the oligomerization through the disulfide shuffling. In that sense, the monoclonal antibody MS785, generated against the Cys6‐Gly16 peptide of hSOD1, might be able to cover and stabilize the Loop I (Fujisawa et al. 2012), possibly leading to the suppression of the disulfide shuffling. Notably, the Cys6‐Gly16 region in hSOD1 has been characterized as the binding site for Derlin‐1, a key component of endoplasmic reticulum (ER)—associated degradation (ERAD) machinery (Fujisawa et al. 2012; Nishitoh et al. ²⁰⁰⁸). Upon misfolding, that region in hSOD1 becomes available for the binding with Derlin‐1, which leads to the impairment of ERAD and the activation of the apoptosis signal‐regulating kinase 1 (ASK1)‐dependent cell death pathway (Nishitoh et al. 2008). Our 22E1 antibody, along with MS785, could hence suppress the mutant SOD1‐induced ER stress by hindering the binding to Derlin‐1; in addition to the suppression of the disulfide‐shuffling, 22E1 would offer therapeutic benefits for DM by attenuating the ER stress.

In summary, our findings here indicate that the E40K mutation in cSOD1 enhances the susceptibility of Cys7 to disulfide shuffling, resulting in the formation of disulfide‐crosslinked oligomers. Furthermore, the oligomerization process was found to be inhibited by the selective binding of the fragment antibody 22E1^Fv‐clasp to the β‐plug and Loop I of cSOD1. Given the structural homology between hSOD1 and cSOD1, we suggest that the β‐plug and Loop I in mutant SOD1 constitute crucial targets for the development of novel therapeutic interventions for canine DM and also human ALS.

4. MATERIALS AND METHODS

Preparation of recombinant cSOD1 proteins, Generation of the monoclonal antibody 22E1, Western blotting analysis of cSOD1 oligomerization in vitro, Modification of cSOD1 with mPEG‐Mal, Differential scanning fluorometry, Analysis of the complex formation by size‐exclusion chromatography, Enzyme‐linked immunosorbent assay (ELISA), Analysis of spinal cord tissues from dogs, and X‐ray data collection and structure calculation are described in the Experimental procedures section of Data S1.

4.1. Preparation of fragment antibodies in the Fv‐clasp format

The cDNAs coding the V_H and V_L domain of 22E1 were individually cloned into the modified pET‐15b plasmid between the NcoI and SalI site (Furukawa et al. 2010). Subsequently, a Gly‐Ser linker and the SARAH domain were fused to the C‐terminus of the V_H and V_L domain. In order to further introduce a disulfide bond between 22E1(V_H)‐SARAH and 22E1(V_L)‐SARAH, the serine residues (Ser112 in 22E1(V_H)‐SARAH and Ser147 in 22E1(V_L)‐SARAH) were replaced with Cys by the inverse PCR method. The amino acid numbering of 22E1(V_H)‐SARAH and 22E1(V_L)‐SARAH was performed as described in the previous paper (Li et al. 2019).

Expression of 22E1(V_H)‐SARAH and 22E1(V_L)‐SARAH in Escherichia coli BL21(DE3) were induced with 0.5 mM isopropyl β‐D‐thiogalactopyranoside at 37°C for 5 h. After a freeze–thaw cycle, the E. coli cells were lysed by sonication in phosphate‐buffered saline (PBS) containing 2% TritonX‐100. Inclusion bodies obtained as insoluble pellets by centrifugation were washed three times with PBS containing 2% TritonX‐100 and solubilized as a concentration of 100 μM in a buffer at pH 8.0 containing 50 mM Tris, 150 mM NaCl, 6M guanidine hydrochloride (GdnHCl), and 10 mM β‐mercaptoethanol (β‐ME). After incubation at 37°C for 30 min, 22E1(V_H)‐SARAH and 22E1(V_L)‐SARAH were then mixed in equal volumes and dialyzed at 4°C overnight against a buffer at pH 8.0 containing 50 mM Tris, 500 mM NaCl, and 0.1 mM GdnHCl. After removal of insoluble materials by centrifugation, the sample was dialyzed twice against a buffer at pH 8.0 containing 50 mM Tris and 500 mM NaCl. Ammonium sulfate (20% saturation) was added to the sample, and the supernatant obtained by centrifugation was loaded on HiTrap Phenyl FF (high sub) (Cytiva) equilibrated with a buffer at pH 8.0 containing 10 mM Tris and ammonium sulfate (20% saturation). The fraction eluted with 10 mM Tris at pH 8.0 was collected and then loaded on HiTrap Q FF (Cytiva) equilibrated with 10 mM Tris at pH 8.0. The fraction eluted with 10 mM Tris, 210 mM NaCl at pH 8.0 was collected, buffer‐exchanged to the MN buffer and purified by a gel‐filtration column (COSMOSIL 5Diol‐300‐II) equilibrated with the MN buffer. The concentration of 22E1^Fv‐clasp was determined spectroscopically from the absorbance at 280 nm with the extinction coefficient of 75,540 M⁻¹ cm⁻¹. Similarly, the V_H and V_L domain of 12CA5 that were C‐terminally fused with the SARAH domain were also prepared.

4.2. Crystallization of 22E1^{Fv ‐clasp} and cSOD1^E40K complex

The samples for the crystallization were prepared by mixing equimolar amounts of 22E1^Fv‐clasp and either holo‐ or apo‐cSOD1^E40K and purified by size exclusion chromatography using a gel filtration column (COSMOSIL‐5Diol‐300‐II, nacalai tesque) equilibrated in the MN buffer (with 1 mM EDTA for the preparation of the 22E1^Fv‐clasp/apo‐cSOD1^E40K complex). The fractions containing the complex between 22E1^Fv‐clasp and either holo‐ or apo‐cSOD1^E40K were concentrated to 10 mg/mL and used for the crystallization. Initial crystallization screening was carried out using automated crystallization screening robot (Kato et al. 2021). The crystals of the 22E1^Fv‐clasp/holo‐cSOD1^E40K complex were obtained by the sitting drop method in a solution containing 20% (w/v) polyethylene glycol 4000, 0.1M Tris pH 8.5, and 0.2M lithium sulfate at 20°C. The crystals of the 22E1^Fv‐clasp/apo‐cSOD1^E40K complex were obtained by the sitting drop method in a solution containing 2% (v/v) polyethylene glycol 400, 0.1M HEPES sodium pH 7.5, and 2.0M ammonium sulfate at 4°C.

AUTHOR CONTRIBUTIONS

Yuki Shino: Conceptualization; investigation; writing – original draft; writing – review and editing. Norifumi Muraki: Investigation. Yui Kobatake: Investigation; resources. Hiroaki Kamishina: Resources. Ryuichi Kato: Resources. Yoshiaki Furukawa: Conceptualization; writing – original draft; writing – review and editing; supervision; funding acquisition.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1. Supporting Information.

Shino Y, Muraki N, Kobatake Y, Kamishina H, Kato R, Furukawa Y. Disulfide‐mediated oligomerization of mutant Cu/Zn‐superoxide dismutase associated with canine degenerative myelopathy. Protein Science. 2024;33(12):e5210. 10.1002/pro.5210

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.