A Zn‐dependent structural transition of SOD1 modulates its ability to undergo phase separation

Abstract

The misfolding and mutation of Cu/Zn superoxide dismutase (SOD1) is commonly associated with amyotrophic lateral sclerosis (ALS). SOD1 can accumulate within stress granules (SGs), a type of membraneless organelle, which is believed to form via liquid–liquid phase separation (LLPS). Using wild‐type, metal‐deficient, and different ALS disease mutants of SOD1 and computer simulations, we report here that the absence of Zn leads to structural disorder within two loop regions of SOD1, triggering SOD1 LLPS and amyloid formation. The addition of exogenous Zn to either metal‐free SOD1 or to the severe ALS mutation I113T leads to the stabilization of the loops and impairs SOD1 LLPS and aggregation. Moreover, partial Zn‐mediated inhibition of LLPS was observed for another severe ALS mutant, G85R, which shows perturbed Zn‐binding. By contrast, the ALS mutant G37R, which shows reduced Cu‐binding, does not undergo LLPS. In addition, SOD1 condensates induced by Zn‐depletion exhibit greater cellular toxicity than aggregates formed by prolonged incubation under aggregating conditions. Overall, our work establishes a role for Zn‐dependent modulation of SOD1 conformation and LLPS properties that may contribute to amyloid formation.

Analysis of amyotrophic lateral sclerosis (ALS) disease‐associated mutations in SOD1 reveals a role for Zn in regulating SOD1 conformation and liquid–liquid phase separation.

Introduction

Amyotrophic lateral sclerosis (ALS) is an adult‐onset progressive neurodegenerative disease that affects motor neurons of the brain and spinal cord, resulting in loss of control over voluntary muscle, paralysis, and death, often by respiratory failure (Rowland & Shneider, 2001). About 8–10% of ALS cases are familial (fALS; Rosen et al, 1993), while the rest occur sporadically (sALS). There are more than 140 point mutations of human Cu/Zn superoxide dismutase (SOD1), which are associated with fALS (Andersen, 2006). Cu/Zn SOD1 is a highly conserved 153 amino acids (aa) long metalloenzyme, encoded by the SOD1 gene located on chromosome 21 (Rosen et al, 1993). It is the primary scavenger of reactive oxygen species, catalyzing the conversion of superoxide radicals to hydrogen peroxide and water in a two‐step reaction.

Each SOD1 monomer consists of eight anti‐parallel β strands arranged in a β‐barrel structure. Extensive structural and biophysical experiments (Furukawa et al, 2016) have previously established that the immature, monomeric state of SOD1 contains two disordered regions (Fig 1A–C, Appendix Fig S1A–C). The first region corresponds to loop IV (aa:49–83) and is also referred to as the “Zn‐binding” loop. The second region corresponds to loop VII (aa:121–142) and is commonly referred to as the “electrostatic loop”. SOD1 matures through three sequential steps, namely zinc (Zn) binding, copper (Cu) insertion via Cu chaperone protein (CCS), and disulfide bond formation (Culotta et al, 1997; Furukawa et al, 2004; Zittin Potter et al, 2007). While Cu is responsible for the activity of the enzyme, Zn binding to loop IV decreases the disorder by guarding the electrostatic loop and structuring the Cu insertion site (Furukawa et al, 2016). In the early steps of SOD1 maturation and its intracellular transport to mitochondrial inter‐membrane space, the protein in its monomeric metal‐free state is highly disordered in solution, and prone to misfolding and subsequent aggregation (Mojumdar et al, 2017).

Figure 1. The structure of SOD1.

1. Human SOD1 amino acid sequence comprised 8‐beta sheets (highlighted in blue) connected by seven loop regions (in red); Zn‐binding loop IV and electrostatic loop VII highlighted in purple.
2. Cartoon representation of SOD1 monomer highlighting loops IV and VII (Coulombic surface coloring: red = negative, white = neutral, blue = positive).
3. Schematic representation of SOD1 variants used in the study. WTSOD1^2SH is the disulfide‐reduced holo form with both metal ion cofactors Zn and Cu. ApoSOD1^2SH is the disulfide‐reduced de‐metalated form of WTSOD1^2SH lacking both metal cofactors. ZnSOD1^2SH and CuSOD1^2SH are the only Zn‐ and Cu‐bound forms of the disulfide‐reduced monomer.

The assembly and aggregation of SOD1 ALS mutants have been shown in cytosolic SGs (Mateju et al, 2017; Ros et al, 2020; Samanta et al, 2021). SGs are membraneless organelles composed of proteins and nucleic acids, possessing liquid‐like properties (Bounedjah et al, 2012). They originate as a response to intracellular stress through a process called liquid–liquid phase separation (LLPS; Hyman & Simons, 2012; Brangwynne, 2013; Hyman et al, 2014; Brangwynne et al, 2015; Boeynaems et al, 2018; Dignon et al, 2020; Mohanty et al, 2022) and may revert to their nondroplet state once the stress is dissipated. The presence of SGs comprising of SOD1, TAR DNA‐binding protein 43 (TDP‐43), and FUS forms the central pathogenic hallmark of ALS (Pokrishevsky et al, 2012; Farrawell et al, 2015; Kang et al, 2018). Previous studies have demonstrated the phase separation of other intrinsically disordered proteins (IDPs) involved in the formation of SGs like FUS and TDP‐43. It is hence important to investigate whether SOD1, which is also a component of SGs, would phase separate under suitable conditions. It may also be noted that while a recent study reports that Zn promotes LLPS of tau protein (Singh et al, 2020), the phase behavior of metal‐containing proteins has been relatively unexplored.

In this study, we observed that disulfide‐reduced, metal‐free SOD1 (ApoSOD1^2SH) forms liquid droplets, which can be reversed by Zn addition, whereas the fully metalated WT protein (WTSOD1^2SH) does not undergo LLPS (Fig 2A). We then show that two severe ALS mutants, namely I113T SOD1^2SH and G85R SOD1^2SH with a low affinity towards Zn, also undergo LLPS, while another mutant G37R SOD1^2SH (with less disease severity and strong binding to Zn) does not. For ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH variants, LLPS is followed by aggregation—although, unlike the liquid droplets, the aggregated state could not be reversed by the addition of Zn. Interestingly, Cu, which is primarily responsible for dismutase activity, does not have any effect on LLPS and/or aggregation. Using Fluorescence Correlation Spectroscopy (FCS) and Fourier‐transform infrared spectroscopy (FTIR), we find that the LLPS and subsequent aggregation are regulated by the conformational transition between a disordered and a relatively compact folded state. The extent of intrinsic disorder in various SOD1 monomeric states was further characterized using all‐atom molecular dynamics (AAMD) simulations. In addition, coarse‐grained (CG), phase coexistence simulations were used to characterize the role of loop‐IV/VII regions in stabilizing the condensed phase of ApoSOD1^2SH. Using multiple toxicity and flow cytometric assays, we showed that the condensates promoted greater cytotoxicity than aggregates. Overall, the cofactor Zn, by virtue of its efficient shielding (Rotunno & Bosco, 2013) of loops IV and VII, acts as a switch to control conformational disorder and the phase separation propensity of SOD1.

Figure 2. ApoSOD1^2SH undergoes liquid–liquid phase separation.

1. SOD1^2SH undergoes LLPS on the removal of metal ion cofactors.
2. DIC microscopic image of WTSOD1^2SH diluted to 100 μM in HEPES buffer, pH 7.4 and 100 mM NaCl, and incubated 37°C, 180 RPM for 30 min shows no droplet formation; scale bar: 100 μm; inset shows Alexa Fluor 488 maleimide‐labeled WTSOD1^2SH; scale bar: 5 μm.
3. Solution turbidity plot (absorbance at 600 nm) shows that WTSOD1^2SH (at concentrations ranging from 25 to 200 μM does not undergo LLPS in the presence or absence of heparin, LLPS inducer; data are represented as mean ± SD (from three biological replicates).
4. DIC microscopic image of ApoSOD1^2SH under condensate inducing conditions (37°C, 180 RPM) at a concentration of 100 μM incubated 100 mM NaCl; scale bar:100 μm (inset on bottom left shows Alexa Fluor 488 maleimide‐labeled protein condensates; scale bar: 5 μm.
5. Fluorescent microscopic images of Alexa Fluor 488 maleimide‐labeled ApoSOD1^2SH condensates diluted in HEPES buffer, pH 7.4 and 100 mM NaCl and incubated (180 RPM, 37°C) in the absence (left) and presence (right) of heparin at 30 min time point; scale bar: 100 μm.
6. Solution turbidity measurements with ApoSOD1^2SH (absorbance at 600 nm) show that while in the absence of heparin the turbidity increases with time, the presence of heparin results in faster LLPS; data are represented as mean ± SD (from three biological replicates).
7. Comparison of ApoSOD1^2SH droplet numbers per microscopic field of view (view area 0.001963 mm²) calculated from five different images of droplets incubated with and without heparin at 2 h time point. Data are represented as mean ± SD. Statistical significance was established using an unpaired nonparametric t‐test (****P < 0.0001).
8. Liquid nature of ApoSOD1^2SH showing droplet fusion and surface wetting; scale bar: 5 μm.
9. Amplitude normalized FCS curves of ApoSOD1^2SH (in blue) of dilute phase (DP) and condensed phase (CP). Intense color indicates CP and light variant indicates DP. The increase in τ_D (as shown by the arrow) translates to a decrease in (D; please refer to equation (2) in Materials and Methods section).
10. Scheme depicting that presence of Zn and not Cu in a preincubation mixture inhibits LLPS.
11. DIC microscopic images of ApoSOD1^2SH in presence of Zn (top) showing no droplet formation and Cu (below) showing the presence of droplets; scale bar: 100 μm.
12. Solution turbidity plot (absorbance at 600 nm) of ApoSOD1^2SH subjected to increasing concentrations of Cu and Zn; turbidity decreases in a dose‐dependent manner for Zn while no significant change is observed in presence of Cu; data are represented as mean ± SD (from three biological replicates).

Source data are available online for this figure.

Results

SOD1 is different from other SG proteins that are implicated in ALS such as FUS and TDP‐43 as it lacks long intrinsically disordered regions (IDRs), which are known to drive the LLPS. It is also much smaller (153 aa) than these other proteins (FUS: 526 aa, TDP‐43: 414 aa) thereby reducing its ability to form comparable, multivalent interactions. As SOD1 has been found to be associated with SGs in ALS pathology, it is important to determine whether it can undergo LLPS without the presence of other SG proteins. To answer this question, we conducted several biophysical measurements to check whether the SOD1 phase separates in vitro. All experiments mentioned in this paper were carried out using reduced monomeric protein variants using 20 mM HEPES buffer, pH 7.4 in presence of 100 mM NaCl, and temperature 37°C, unless explicitly noted otherwise.

ApoSOD1^2SH undergoes liquid–liquid phase separation

Using phase contrast microscopy (Fig 2B and C, Appendix Fig S2A), we did not observe the formation of condensates for different concentrations (25, 50, 100, and 200 μM) of WTSOD1^2SH. We then used three co‐solvents: a neutral molecular crowder polyethylene glycol 8000 (PEG), a positively charged polymer (polyethylenimine, PEI), and a negatively charged polysaccharide (heparin) to check whether these could induce the formation of WTSOD1^2SH condensates. Neutral crowders like PEG have been found to facilitate condensate formation for other proteins, presumably by increasing their local concentrations (Annunziata et al, 2002; Wegmann et al, 2018; Park et al, 2020). The idea of PEI has originated from the fact that SOD1 has a net negative charge, and hence the use of cationic PEI (Pullara et al, 2022) would facilitate LLPS in SOD1. By contrast, heparin—a polyanionic glycosaminoglycan known to be an aggregation‐inducing agent (Vilasi et al, 2011; Maïza et al, 2018), has also been shown to drive the phase separation of low complexity regions (LCRs) or IDRs like Tau (Ambadipudi et al, 2017), C‐terminal low complexity domain of TDP43 (Babinchak et al, 2020) and SH₃‐PRM₅ (Ghosh et al, 2019). Notably, we did not observe any condensate formation for WTSOD1^2SH in the presence of different concentrations of any of these above crowding agents (Appendix Fig S2B–D).

Since immature SOD1^2SH exhibits a conformational disorder that is known to promote LLPS, we next chose the metal‐free Apo variant (ApoSOD1^2SH) for further investigation (See Materials and Methods). Interestingly, ApoSOD1^2SH formed spherical droplets in HEPES buffer at 37°C at concentrations greater than 2 μM (Fig 2D). Fluorescence imaging of Alexa‐488 Maleimide‐labeled protein (15 nM‐labeled protein in the presence of 100 μM of unlabelled protein) showed that the protein molecules were distributed throughout the liquid droplets (Fig 2D inset). While ApoSOD1^2SH formed droplets in the absence of co‐solvents (heparin, PEG8000, and PEI as mentioned above), the process was accelerated in the presence of all three (Fig 2E–G, Appendix Fig S2E–G). We used 7% heparin for all our studies described henceforth. With time, the droplets in proximity underwent fusion events forming larger droplets and surface wetting that could be monitored using fluorescence microscopy, exhibiting the liquid‐like nature of these SOD1 condensates (Fig 2H), like other proteins undergoing LLPS (Molliex et al, 2015; Kanaan et al, 2020; Ray et al, 2020; Bhandari et al, 2021; Krainer et al, 2021; Siegert et al, 2021).

Next, we used Fluorescence Correlation Spectroscopy (FCS) to monitor the change in the diffusional dynamics of protein variants outside (dilute phase) and inside (condensed phase) droplets. FCS probes the fluctuations in fluorescence intensity arising out of the diffusion of labeled protein molecules inside an approximate femtoliter confocal volume. Analyses of the correlation functions using a suitable model provide information about the diffusion coefficient (D) of the molecule. The advantage of FCS comes from its application to probe a selected region inside a microscopic image, and its ability to provide single‐molecule resolution using diffusing molecules. From the value of the diffusion coefficient (D), we calculated the hydrodynamic radius (r _H) of ApoSOD1^2SH outside the droplet to be around 2.80 nm, which is consistent with earlier reports (Vassall et al, 2011; Bunck et al, 2018). Inside the droplets, we observed a large (~ 2.5 fold) reduction in the value of D (and reduced mobility; Fig 2I). Interestingly, between the outside and inside droplets, we also observed a large (17‐fold) increase in the value of rotational correlation times (Appendix Fig S2H), which we measured by time‐resolved anisotropy measurements using time‐domain fluorescence lifetime imaging microscopy (TD‐FLIM). The restricted environment inside the droplet likely causes a decrease in rotational and translational motion of the protein molecules inside the droplets. It may be noted that a large decrease in D for the condensed phase compared with the dilute phase has been previously observed for different proteins including FUS (Murthy et al, 2019), TDP‐43 (Conicella et al, 2020), and Tau (Boyko et al, 2019). In addition, we measured the fluorescence lifetime values of the attached dye (Alexa 488 Maleimide) to probe its excited state properties, which increased from the dilute phase to the condensed phase (Appendix Fig S2I–K). Droplet formation was also found to be salt‐dependent as increasing salt concentration (from 100 to 600 μM) gives rise to larger droplets (Appendix Fig S2L and M). In addition, when the condensates were subjected to 1,6‐hexanediol treatment, an aliphatic alcohol that has been proposed to disrupt weak hydrophobic interactions (Kroschwald et al, 2017), there was a significant reduction in the number of droplets (Appendix Fig S2N and O). These results highlight the importance of weak hydrophobic interactions in stabilizing the ApoSOD1^2SH protein condensates.

Studies with ApoSOD1^2SH and ALS mutants suggest that the cofactor Zn modulates LLPS

Since WTSOD1^2SH does not form condensates, while the metal‐free ApoSOD1^2SH forms these readily, it is apparent that the metal cofactors may have a role to play in condensate formation. Since SOD1 contains two metals (Cu and Zn) as cofactors, we then wanted to test the effect of the exogenous addition of each of these on the LLPS propensity of ApoSOD1^2SH. We added Cu and Zn, respectively, to 100 μM protein solutions, which were then incubated for LLPS (preincubation condition, Fig 2J). We found that the addition of Zn at the time of incubation inhibited LLPS formation, while exogenous Cu addition did not have any effect on the condensed phase of ApoSOD1^2SH (Fig 2J–L). We then added Zn to preformed condensates of ApoSOD1^2SH (postincubation condition) to check whether Zn could dissolve them (Appendix Fig S2P). The addition of Zn to preformed condensates was observed to dissolve them immediately. This experiment shows that LLPS of ApoSOD1^2SH is reversible through Zn binding and that metal coordination can either inhibit (preincubation condition) or disrupt (postincubation) condensates.

After establishing that Zn has a major role in modulating the LLPS of ApoSOD1^2SH, we selected three ALS‐associated disease mutants for further studies. Two of these mutants, I113T SOD1^2SH and G85R SOD1^2SH are known to have compromised Zn binding in their native conditions (Hayward et al, 2002; Cao et al, 2008). The inefficiency in Zn binding for I113T SOD1^2SH has been attributed to the mutational stress introduced at the dimeric surface of the protein (Hennig et al, 2015). In G85R SOD1^2SH, the loss of Zn binding is likely due to the proximity of the mutation site to the Zn‐binding domain (residue 63–83). G85R SOD1^2SH was found to bind ~ 5% Zn relative to WTSOD1^2SH when expressed with the same metal ion supplementation (Hayward et al, 2002). In the third ALS mutant, G37R SOD1^2SH the mutation site is comparatively distant from the Zn‐binding site (23.7 Å). Although it has an intact Zn‐binding site, G37R SOD1^2SH shows Cu deficiency (Tokuda et al, 2018). We have shown previously that G37R SOD1^2SH and I113T SOD1^2SH behave similarly to the WTSOD1^2SH and ApoSOD1^2SH, respectively, in terms of their membrane binding and aggregation propensity (Sannigrahi et al, 2021). In addition, the G37RS OD1^2SH mutant possesses a less severe disease phenotype (average patient survival time after disease onset ~ 17 years) as compared to I113T SOD1^2SH and G85R SOD1^2SH mutations (average survival times of ~ 4.3 and 6 years, respectively; Wang et al, 2008).

Analytical size exclusion chromatography (SEC) of all three mutants (with or without metal cofactors) showed single eluting species (Appendix Fig S3A). Mass spectrometric analyses also suggested these to be in their monomeric states (Appendix Fig S3B). Since SOD1 has been found to form dimer by involving Cys57 and Cys146, we used nonreducing SDS gel electrophoresis (Appendix Fig S3C) and native gel electrophoresis (Appendix Fig S3D) for further size characterization. Both methods show a single protein band corresponding to the monomeric state indicating the absence of any multimeric species. We found that the two variants in their native forms, I113T SOD1^2SH and G85R SOD1^2SH formed droplets while G37R SOD1^2SH did not (Fig 3A, Appendix Fig S3E). We observed that the addition of Cu did not have any effect on the condensed phase in either I113T SOD1^2SH or G85R SOD1^2SH (Fig 3A, Appendix Fig S3F). By contrast, the addition of Zn inhibited LLPS in I113T SOD1^2SH in a dose‐dependent way (Fig 3A, Appendix Fig S3F). Interestingly, with G85R SOD1^2SH, the addition of Zn did not lead to complete inhibition of LLPS (Fig 3A, Appendix Fig S3F) when compared to that of I113T SOD1^2SH. FCS measurements confirmed that phase‐separated condensates of I113T SOD1^2SH and G85R SOD1^2SH were composed of slow diffusing species (Fig 3B). The fluorescence lifetime of the probe increased inside the condensates (Appendix Fig S3G–I). Further, when we removed the metal ions from G37R SOD1^2SH by dialyzing the protein in EDTA, we found that the metal‐free G37R SOD1^2SH mutant also readily underwent LLPS, further highlighting the central role of metal binding in this process (Fig 3C). The addition of Zn into metal‐free G37R SOD1^2SH inhibited LLPS, while the addition of Cu did not yield any change (Fig 3C). When Zn was added to preformed condensates (postincubation condition) to see whether the LLPS can be reversed (Appendix Fig S3J), I113T SOD1^2SH droplets disappear completely whereas G85R SOD1^2SH droplets show partial dissolution (Appendix Fig S3J). These experiments suggest that the LLPS of SOD1^2SH is effectively reversible through Zn coordination for ALS mutants as well unless the Zn binding is significantly compromised as in the case of G85R SOD1^2SH.

Figure 3. ALS mutants with compromised Zn binding undergo LLPS.

1. DIC microscopic images of I113T SOD1^2SH (top) and G85R SOD1^2SH (bottom) droplets incubated in the absence and presence of Zn and Cu, respectively (scale bar: 100 μm); insets on bottom left show Alexa Fluor 488 maleimide‐labeled protein condensates for I113T SOD1^2SH and G85R SOD1^2SH, respectively; scale bar: 5 μm. Droplet dissolution was observed for I113T SOD1^2SH on the addition of Zn while droplets persisted on the addition of Cu. Red arrowhead indicates fusion of liquid droplets.
2. Amplitude normalized FCS curves of I113T SOD1^2SH (in green) and G85R SOD1^2SH (in gray) of dilute phase (DP) and condensed phase (CP) showing an increase in diffusion time from DP to CP. The intense color indicates CP, and light variant indicates DP.
3. On metal cofactor removal, ApoG37R SOD1^2SH formed condensates, which ceased to exist when protein was incubated with Zn, while condensates persisted on Cu addition (left to right); scale bar: 100 μm.

Source data are available online for this figure.

LLPS of monomeric SOD1 is driven by an order‐to‐disorder transition in the absence of Zn

The results above lead to a natural question about the mechanism of LLPS in the absence of Zn or reduced Zn‐binding in the case of ALS mutants. NMR studies indicate that the conformational landscape of ApoSOD1^2SH exhibits two sparsely populated conformers that correspond to non‐native oligomers (Sekhar et al, 2015). These oligomers form via interactions between the native dimeric interface and disordered loop VII. The tendency to form non‐native oligomers may facilitate the formation of a liquid‐like, condensed phase under suitable conditions.

We used FCS to study the hydrodynamics of the different SOD1^2SH variants in the absence and presence of Zn (Fig 4A, Appendix Table S1). Both ApoSOD1^2SH and severe mutants (I113T SOD1^2SH and G85R SOD1^2SH) showed larger r _H when compared to G37R SOD1^2SH and WTSOD1^2SH (Appendix Table S1). The value of r _H obtained by FCS (this study) for ApoSOD1^2SH matched well with previous experimental data (Vassall et al, 2011; Bunck et al, 2018). To obtain further insights into the average conformational properties of SOD1, we plotted the literature values (Chakraborty et al, 2021) of r _H of different proteins in their compact folded and chemically unfolded states. The log–log plot of the r _H values of the folded proteins as a function of the number of residues could be fit to a straight line with a slope of 0.29 (Fig 4B). For chemically unfolded proteins, the slope of the log–log plot was found to be 0.59. Interestingly, when we plotted the r _H values of different SOD1^2SH variants in Fig 4B, we found that ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH were positioned near the line for the chemically unfolded proteins. When we plotted the r _H value of α‐synuclein, a well‐characterized intrinsically disordered protein, it was found positioned near the same line. WTSOD1^2SH and G37R SOD1^2SH variants, on the other hand, were found near the line obtained for the compact folded proteins. This suggests that ApoSOD1^2SH and severe variants (I113T SOD1^2SH and G85R SOD1^2SH) are more extended (or sample extended conformer more often) compared with G37R SOD1^2SH and WTSOD1^2SH. We also observed that r _H decreased substantially with the addition of Zn in ApoSOD1^2SH and severe variants I113T SOD1^2SH and G85R SOD1^2SH (Fig 4B, Appendix Fig S4A–C).

Figure 4. Zn drives disorder to order transition in SOD1.

1. Scheme shows Zn‐dependent disorder to order transition in monomeric SOD1.
2. Plot of the log_e of the hydrodynamic radius (r _H) versus the log_e of the number of residues in the polypeptide chain. The line fitted to these data for the native folded proteins (yellow dashed) has a slope of 0.29 ± 0.02 and a y‐axis intercept of 1.56 ± 0.1, while the other fitted to the chemically denatured protein (gray dashed) data has a slope of 0.57 ± 0.02 and a y‐axis intercept of 0.79 ± 0.07. Literature data have been used for folded and chemically denatured proteins while we employed FCS to calculate the r _H of all SOD1^2SH variants with (solid shapes) and without Zn (hollow shapes). The hollow gray circle of G85R SOD1^2SH is directly under hollow green I113T SOD1^2SH in the log–log plot.
3. Variation of diffusion coefficients of de‐metalated protein variants determined from FCS measurement with increasing Zn concentration (inset shows a comparison of binding constants between mutants and Zn; ApoSOD1^2SH and ApoG37R SOD1^2SH has a higher binding affinity to Zn than ApoI113T SOD1^2SH and ApoG85R SOD1^2SH). Data are represented as mean ± SD (from three biological replicates).
4. Disordered/ extended conformation content decreases in ApoSOD1^2SH and ApoI113T SOD1^2SH with the addition of Zn calculated from FTIR spectra; data are represented as mean ± SD (from three biological replicates).

Source data are available online for this figure.

Since the addition of Zn resulted in large compaction in ApoSOD1^2SH and severe variants, we used FCS data to determine their Zn‐binding affinity (Fig 4C). For the Zn‐binding study, all protein variants were in their de‐metalated Apo forms, which we prepared by dialyzing the protein samples with EDTA using previous literature. We found that while ApoSOD1^2SH and ApoG37R SOD1^2SH bind to Zn with the highest affinity (k_a of 1.28 × 10⁷ M⁻¹ and 1.54 × 10⁷ M⁻¹, respectively), the binding of ApoG85R SOD1^2SH is the weakest (k_a of 1.71 × 10⁶ M⁻¹). In comparison, ApoI113T SOD1^2SH showed an intermediate binding affinity (k_a of 0.75 × 10⁷ M⁻¹) towards Zn. We then used FTIR to determine how the secondary structure of the protein changed as a result of cofactor insertion. Using FTIR, we found that disorder/extended conformations were higher in metal‐free ApoSOD1^2SH and severe variants (I113T SOD1^2SH and G85R SOD1^2SH) when compared to WTSOD1^2SH and G37R SOD1^2SH (Appendix Fig S4D, Appendix Table S2). The disordered/extended components in ApoSOD1^2SH and I113T SOD1^2SH (Fig 4D, Appendix Fig S4E and F) decreased as Zn was added, while the effect of Zn was less visible in G85R SOD1^2SH (Appendix Fig S4G). No significant change in the secondary structure for any variant was observed in the presence of Cu (Fig 4D, Appendix Fig S4E–G). Together, FTIR and FCS data revealed that the addition of Zn in ApoSOD1^2SH and I113T SOD1^2SH resulted in a conformational transition of the Zn‐free protein from an extended (FCS data, greater r _H) and disordered (FTIR data) state to a compact (lower r _H) and less disordered state (FTIR data). For G85R SOD1^2SH, this transition is partial, presumably because of its weaker zinc binding compared with the I113T mutant.

Characterization of disorder in monomeric ApoSOD1^2SH using all‐atom simulations

To complement FITR and FCS experiments, we characterized the extent of conformational disorder for SOD1 monomer variants using microsecond timescale and atomistic MD simulations (Souza et al, 2019; Timr et al, 2020). Simulations were performed using the AMBER99SB‐disp force field (Robustelli et al, 2018) that was shown to be suitable for simulating both folded and disordered proteins (see Materials and Methods). The hydrodynamic radius r _H was calculated using the HullRad algorithm (Fleming & Fleming, 2018), which employs a convex hull method to estimate the hydrodynamic volume of the protein molecule. The mean r _H value determined from simulations ranges from 2.1 nm for ApoSOD1^2SH variants to 3.2 nm for unfolded SOD1^2SH (Fig 4B). The value of r _H for ApoSOD1^2SH from FCS measurements (2.80 ± 0.09 nm) lies within the range of simulation results, which indicates the partially unfolded nature of this variant. Secondary structure analysis of ApoSOD1^2SH variants using the DSSP algorithm (Kabsch & Sander, 1983) indicates the presence of coil‐like conformations in loop IV/VII while β‐sheet conformations were dominant in the expected β‐barrel regions (Appendix Fig S5A). By contrast, unfolded SOD1^2SH adopted coil‐like conformations with short stretches of α‐helix and β‐sheet conformations (Appendix Fig S5B). Distance root mean square deviation (dRMSD) analysis confirms that the β‐barrel fold is stable (dRMSD <= 0.3 nm) across all ApoSOD1^2SH variants (Appendix Fig S5C), which is consistent with previous NMR/SAXS experiments (Furukawa et al, 2016). Representative conformations of ApoSOD1^2SH and unfolded SOD1^2SH are shown in Appendix Fig S5D.

Consistent with secondary structure and dRMSD analysis, root mean‐square fluctuation (RMSF) of Cα atoms suggests that ApoSOD1^2SH variants exhibit high conformational flexibility (RMSF_max > 0.4 nm) in loop‐IV/VII regions (Fig 5A). To assess the effect of Zn binding on the conformational disorder of monomeric SOD1, we also simulated three additional SOD1 states: ApoSOD1^S‐S (disulfide present, without metal cofactors), ZnSOD1^2SH (Zn‐bound and disulfide‐reduced) and ZnSOD1^S‐S (Zn‐bound and disulfide present). The stable coordination of Zn to SOD1^2SH and SOD1^S‐S was confirmed based on distance analysis for Zn/Zn‐binding residues over the course of the trajectories (Appendix Fig S5E and F). We compared the conformational flexibility across different regions of the SOD1 by analyzing their RMSF profiles (Fig 5B). The RMSF profile of ApoSOD1^2SH shows considerable flexibility in loop IV and loop VII wherein RMSF for several residues exceed 0.3 nm (Fig 5B, Appendix Movie EV1). Notably, the presence of disulfide bond only reduced flexibility in the aa: 51–60 region (RMSF_max ~ 0.15 nm) within loop IV, which is involved in disulfide bond formation through C57 while loop VII remained flexible. Zn significantly reduced the flexibility of loop IV (RMSF_max ~ 0.15 nm) to which it directly binds and led to a reduction in the RMSF of loop‐VII residues (Appendix Movie EV2). In conclusion, the high flexibility of loop IV/VII observed in the absence of Zn in simulations is consistent with experimental observations and points towards an intrinsic propensity for structural disorder within the SOD1 monomer in the absence of Zn.

Figure 5. Characterization of SOD1^2SH variants and molecular interactions in the condensed phase using molecular dynamics simulations.

• A, B
  (A) Per‐residue RMSF profiles of ApoSOD1^2SH & its mutants computed from 5 μs simulations performed for each variant. (B) Per‐residue RMSF profiles of ApoSOD1^2SH, ApoSOD1^S‐S, ZnSOD1^S‐S, and ZnSOD1^2SH computed from 5 μs simulations performed for each variant. Error bars in A and B correspond to standard errors, which are estimated by blocks averaging over 10 (500 ns each) blocks.
• C
  Salt‐bridge analysis of ApoG85R SOD1^2SH (dist. cutoff < 0.5 nm = salt bridge).
• D
  Snapshots from the ApoG85R SOD1^2SH trajectory showing the formation of non‐native salt bridges, which are likely to be detrimental for Zn‐binding.
• E
  Density profile of ApoSOD1^2SH and its loop‐deletion variants computed from the CG slab simulations. The condensed phase is centered in the middle of the simulation box.
• F
  Slab snapshots from CG simulations of ApoSOD1^2SH and its variants at 275 K (blue beads represent loop IV, red beads represent loop VII, and orange beads represent the other domains).
• G
  Pairwise intermolecular contact formation (normalized by maximum probability) in the condensed phase as a function of residue index.
• H
  Saturation concentration measured in the CG phase coexistence simulations for ApoSOD1^2SH and its variants including ALS mutants at 275 K. The error bars correspond to SD, which are estimated from block averages over four blocks.

Source data are available online for this figure.

To understand the mechanism by which G85R substitution in ApoSOD1^2SH may lead to loss of Zn binding, we analyzed the interactions of R85 with neighboring residues over the course of the 6 μs run (Fig 5C and D). We observed that R85 formed salt bridges with key aspartate residues (Strange et al, 2006), which either stabilize Zn (D83/D124) or may play a role in stabilizing the native conformation of loop IV (D101). D83 directly binds to Zn whereas D124 coordinates two key histidine residues involved in Zn coordination. D101 natively forms a salt bridge with R79, which may help stabilize the conformation of loop IV. Overall, the simulation of ApoG85R SOD1^2SH suggests a mechanism wherein the formation of dynamic, non‐native salt bridges involving R85 may hinder its ability to bind Zn. The induction of non‐native, intramolecular salt bridges through post‐translational modifications (Skinner et al, 2017; Mohanty et al, 2019, 2021) has been previously shown to alter the binding affinity of protein–protein interactions and may therefore modulate ApoG85R SOD1^2SH affinity for Zn²⁺ through a similar mechanism.

CG simulations highlight the role of loop regions in ApoSOD1^2SH condensate formation

To determine which regions of ApoSOD1^2SH are involved in stabilizing the condensate, we performed coarse‐grained (CG), phase coexistence simulations (Fig 5E) using a protocol, which has been described previously (Mammen Regy et al, 2021). Briefly, ApoSOD1^2SH monomers were modeled at a single bead per amino acid resolution and formed a stable condensate at 275 K (Fig 5F—top, Appendix Movie EV3). To model ApoSOD1^2SH, the residues which form the β‐barrel structure were made rigid while those belonging to loop IV/VII were allowed to remain flexible. The analysis of intermolecular contact maps between SOD1 monomers in the condensed phase indicates that several regions of the β‐barrel and disordered loop IV (aa:60–75) collectively stabilize the condensate (Fig 5G). By contrast, fewer intermolecular contacts were observed for loop VII owing to its shorter length. The observed contacts based on residue type indicate the presence of electrostatic interactions between lysine and aspartate/glutamate residues (Appendix Fig S5G and H). In addition, a wide range of pairwise interactions involving residue types such as valine/isoleucine and serine/glycine also contributed towards phase separation. Taken together, the CG simulation suggests that ApoSOD1^2SH condensate is stabilized through numerous interactions involving both the β‐barrel and loop‐IV/VII regions.

The presence of long, disordered loops was shown to destabilize the β‐barrel structure of ApoSOD1^2SH (~ 3 kcal/mol) and promote aggregation (Yang et al, 2018). To assess the importance of disordered loop‐IV/VII regions in the formation of ApoSOD1^2SH condensate, we tested the effect of loop deletion on the saturation concentration (c_sat) in simulations (Fig 5E). In comparison with full‐length ApoSOD1^2SH, deletion of either loop IV/VII or both was found to substantially increase the concentration of the dilute phase (c_sat; Fig 5F and H). These observations collectively suggest that loop‐mediated interactions play an important role in the stabilization of the ApoSOD1^2SH condensed phase. Several ALS‐associated mutations are found to occur in the loop‐IV/VII regions, which do not impact the stability of the SOD1 but instead disrupt surface‐exposed hydrogen bonds (Byström et al, 2010). Many of these mutations alter the net charge of SOD1, which may impact the stability of the condensed phase by modulating electrostatic interactions. Hence, we tested the impact of these mutants and found that those which decreased the net negative charge of SOD1 (D76 → V/Y, D124V, D125H, N139K) reduced c_sat by up to two‐fold (Fig 5H, Appendix Fig S5I). By contrast, the neutral substitution—S134N had no impact on c_sat (Fig 5H). These observations indicate the potential sensitivity of the condensed phase towards mutations, which alter the net charge of SOD1. Altogether, CG simulations highlight the importance of loop‐IV/VII interactions in the condensed phase.

LLPS of SOD1 is followed by aggregation

Studies of RNA‐binding proteins (Anderson & Kedersha, 2006; Satoh et al, 2012; Wheeler et al, 2016) involved in SG formation such as FUS, hnRNPA1/2, and TDP‐43 indicate that LLPS of their disordered, low complexity domains (LCDs) gives rise to labile fibrils upon droplet maturation. Moreover, the liquid‐to‐solid phase transition, which leads to the formation of fibrils was shown to be enhanced by mutations associated with neurodegenerative diseases. Based on these observations, it has been proposed that LLPS may represent an intermediate state en route to toxic aggregate formation (Babinchak & Surewicz, 2020).

After establishing that ApoSOD1^2SH and severe ALS mutants (I113T SOD1^2SH and G85R SOD1^2SH) undergo a Zn‐dependent phase separation, we investigated the temporal maturation of ApoSOD1^2SH droplets using fluorescence microscopy in combination with FCS (Fig 6A and B). As the maturation progressed with time, we observed the presence of high‐intensity fluorescent aggregates, which started forming within the droplets (Fig 6C). FCS experiments were carried out at different regions and different time points to investigate the diffusional dynamics of droplets maturation. For the FCS experiments, we chose three regions, namely the diffused part of the image outside the droplets (region 1, Fig 6A–C, Appendix Fig S6A); the relatively low‐intensity portion inside the droplets (region 2, Fig 6A–C, Appendix Fig S6B); and the high‐intensity regions inside the droplets, which contained the aggregates (region 3, Fig 6A–C). The analyses of the correlation functions obtained at region 1 showed the presence of fast diffusing molecules (Fig 6D). The hydrodynamic properties (r _H of 2.80 ± 0.09 nm) of these fast‐moving molecules resembled that of labeled ApoSOD1^2SH, suggesting that region 1 predominantly contained monomeric protein.

Figure 6. Maturation of liquid droplets precedes aggregation.

1. Cartoon scheme showing the maturation of protein droplets with time (region1: diffused region outside droplet, region 2: low‐intensity portion inside the droplet, region 3: highly intense portion inside droplet appears during maturation) at different time points.
2. Diffusion coefficient of regions 1 and 2 (within droplet) was measured using point FCS as shown in the scheme.
3. Upper panel shows fluorescence images of the maturation of Alexa Fluor 488‐labeled ApoSOD1^2SH droplets with time, samples were incubated at 37°C up to ~ 25 h. Images are taken at indicated time point: 0* hour (the time after incubation required for droplet formation), 12 h, and > 24 h; images below show magnified scale bar: 5 μm. The middle panel shows magnified view of the marked region from corresponding images on the upper panel. Red arrowhead indicates matured (solid) portion inside the droplet. Lower panel shows the coexistence of liquid and solid phases inside droplet; scale bar: 10 μm. Intensity plot (bottom right) shows the distribution of high‐intensity region within matured droplet.
4. Diffusion coefficient obtained from point FCS measurement at different regions of the droplet during maturation course; 0* (x‐axis label) indicates the time after incubation required for droplet formation. Data are shown mean ± SD (from three biological replicates). Statistical significance was measured using paired t‐test; ns denotes nonsignificant; **P < 0.01; ***P < 0.001.
5. ThT fluorescence assay plot showing aggregation kinetics of all SOD1^2SH variants. ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH show high ThT fluorescence.
6. AFM micrographs of ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH show fibrillar aggregates after prolonged incubation at 37°C, 180 RPM in absence of Zn (left panel); AFM micrographs of ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH incubated with Zn (at 1:5 protein:Zn molar ratio) show the presence of oligomers for ApoSOD1^2SH, I113T SOD1^2SH and short fibrils for G85R SOD1^2SH (right panel); scale bar: 200 nm.
7. Cell proliferation of SHSY5Y human neuroblastoma cells assessed by MTT assay post 12 h treatment with 5 μM ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH condensates and aggregates (fibrils). Cell culture medium (DMEM) was used as negative control and untreated SHSY5Y cells were used as a positive control. Data are shown as mean ± SEM (from three biological replicates). P values were determined by two‐tailed unpaired t tests; *P < 0.05, **P < 0.01; ***P < 0.001, ****P < 0.0001.
8. Dot plots showing flow cytometric analysis of annexin V and propidium iodide (PI) staining of apoptotic SHSY5Y neuroblastoma cells following a 12 h treatment with 5 μM ApoSOD1^2SH condensates, I113T SOD1^2SH condensates, G85R SOD1^2SH condensates, ApoSOD1^2SH fibrils, I113T SOD1^2SH fibrils, and G85R SOD1^2SH fibrils, respectively (see Materials and Methods for details). High cytotoxicity was observed for SOD1^2SH condensates.

Source data are available online for this figure.

By contrast, analyses of the correlation functions obtained inside the droplets (region 2) were complex. Even at the early time points (for example, at 0* h in Fig 6D, which shows the data at the earliest time point when LLPS was observed) the values of D were found to be considerably less than what was observed for the monomeric protein in solution or for the molecules present at the diffused region (region 1) indicating a restrictive environment inside the droplets. As the maturation progresses, the values of D decreased further with time (Fig 6D). Interestingly, at 24 h and beyond, the correlation functions contained more than one component with a large fraction of an extremely low diffusing component (Fig 6D). The region containing the aggregates (region 3) did not yield any correlation functions as their diffusion behavior was too slow to measure.

Aggregation kinetics and formation of amyloid fibrils

ThT fluorescence assay was used to determine whether the aggregates obtained after long incubation would eventually lead to the amyloid formation or these are amorphous. We found that the ApoSOD1^2SH and two severe mutants (I113T SOD1^2SH and G85R SOD1^2SH) showed large ThT fluorescence indicating amyloid formation, while G37R SOD1^2SH did not aggregate (Fig 6E). The aggregation kinetics were sigmoidal with prominent lag phases. We used atomic force microscopy (AFM) experiments to complement ThT fluorescence data (Fig 6F, Appendix Fig S6C), which provided information about the height and morphology of the aggregates. AFM micrographs of ApoSOD1^2SH showed the presence of long curved fibrils with a height of 4 nm (Fig 6F, Appendix Fig S6C). I113T SOD1^2SH aggregates formed a fibrillar network of long dense fibrils with a height of 5 nm (Fig 6F, Appendix Fig S6C). AFM micrographs of G85R SOD1^2SH showed fibrillar aggregates with a height ranging between 4 and 8 nm.

We had observed earlier that LLPS was completely reversible when Zn was added either at the preincubation or postincubation conditions. Therefore, we chose to probe the effect of the metal cofactors on the formation and stability of amyloid aggregates under preincubation and postincubation conditions, respectively. When we added Zn at the beginning of incubation (preincubation), aggregation of ApoSOD1^2SH and I113T SOD1^2SH (Fig 6F, Appendix Fig S6D) was inhibited while the presence of Cu did not suppress aggregation (Appendix Fig S6E and F). AFM micrographs of ApoSOD1^2SH and I113T SOD1^2SH incubated in presence of Zn were observed to consist of oligomers alone (Fig 6F). Interestingly, short fibrils were observed for G85R SOD1^2SH incubated with Zn (Fig 6F), which is consistent with its significantly reduced affinity towards Zn. By contrast, when we added Zn at the stationary phase of the amyloid formation (144 h, postincubation), we found no significant change in ThT fluorescence (Appendix Fig S6G). These data in combination with the Zn‐induced dissolution of ApoSOD1^2SH droplets clearly suggest that while the state of droplet formation is reversible by the addition of Zn, the protein aggregates are stable and not altered by either of the cofactors.

SOD1 condensates show higher toxicity compared with aggregates

Since mutations in SOD1 are linked to motor neuronal death and ALS pathology, we studied the toxicity of condensates as opposed to the aggregates using three different assays. For these assays, we prepared condensates and aggregates of different SOD1 variants, and their formation was checked using phase contrast microscopy (for condensates) and AFM (for aggregates), respectively. WTSOD1^2SH and G37R SOD1^2SH were excluded from these assays as they did not form either condensates or aggregates. It may also be noted that nontoxic nature of WT‐SOD1 was established by previous results (Clement et al, 2003; Magrané et al, 2009; Zhu et al, 2018; Sannigrahi et al, 2021). Untreated neuroblastoma cells in DMEM media were used as a control in all the following assays. Cultured SHSY5Y human neuroblastoma cells were treated for 12 h with 5 μM SOD1^2SH condensates and aggregates (after long incubation of 144 h), respectively, and cell death was first probed by trypan blue exclusion assay. This assay is based on the principle that cells that lose their membrane integrity are permeable to trypan blue dye. We observed a significant increase in cell death in SHSY5Y cells treated with SOD1^2SH condensates (Appendix Fig S6H). We complemented our observations with an MTT assay, which assesses the metabolic activity of cells reflecting the cell proliferation of the sample. 5 μM mutant SOD1^2SH condensates and aggregates were added to the cell media of SHSY5Y cells in a 96‐well plate, respectively (see Materials and Methods). Condensates showed significantly higher toxicity than aggregates after incubation with MTT (Fig 6G). To validate these results, we used flow cytometry with annexin V and propidium iodide. 5 μM condensates and aggregates were added to the cell media of cultured SHSY5Y cells and incubated for 12 h at 37°C. Both apoptotic and necrotic cell populations were highest in cells treated with condensates. I113T SOD1^2SH condensate was observed to be the most toxic in terms of apoptotic and necrotic cell population (Fig 6H, Appendix Fig S6I). Complementing our findings, observations from recent studies (Bolognesi et al, 2019; Kanaan et al, 2020) on TDP‐43 C‐terminal domain and Tau, respectively, also indicate that condensates promote cellular toxicity as opposed to aggregates. Further studies are needed to determine the relative significance of condensate versus aggregate toxicity for proteins implicated in neurodegenerative diseases.

Discussion

The link between SOD1 stability, aggregation, and its relation to ALS disease characteristics has been extensively studied. Mutant SOD1 aggregates were found in inclusion bodies formed in the spinal cord of familial ALS patients (Shibata et al, 1996). Notably, aggregates isolated from spinal cord tissue of ALS‐SOD1 mouse models were found to be metal deficient (Shaw et al, 2008) implying that ApoSOD1 is aggregation‐prone. ALS mutants were shown to exert their destabilizing effect predominantly on ApoSOD1^2SH and could trigger unfolding at physiological temperature (Furukawa & O'Halloran, 2005). The extent of structural destabilization (decrease in T _m) induced by fALS‐associated mutations in monomeric ApoSOD1^2SH was shown to be positively correlated to aggregation propensity (Vassall et al, 2011). However, decreased thermal stability/increased aggregation propensity was uncorrelated with disease severity (Byström et al, 2010; Vassall et al, 2011). Based on these observations, it was suggested that ApoSOD1^2SH may play an important role in ALS, while additional factors, such as environmental changes and the presence of conformational heterogeneity may be involved, highlighting the mechanistic complexity of ALS onset and progression.

SOD1 has been reported to associate with and alter SG dynamics. T‐cell intracellular antigen‐1 (TIA‐1) is an RNA‐binding protein that promotes the formation of SGs and is critical to the eukaryotic cellular stress response. Overexpression of mutant SOD1 increases TIA‐1 positive inclusions in spinal cord tissue of ALS transgenic mouse and familial ALS patient (Lee et al, 2020). Mutant SOD1 also co‐localizes and interacts with another important SG marker G3BP1 (Gal et al, 2016). Since both G3BP1 and TIA‐1 are RNA‐binding proteins, their interactions with mutant SOD1 suggest a potential link between SOD1 mutations and RNA metabolism alterations.

Using both experiments and molecular simulations, we show here that while compact, WTSOD1^2SH does not undergo phase separation, partially disordered ApoSOD1^2SH or Zn‐compromised mutants engage in intermolecular interactions, which lead to condensate formation. Based on the data presented in this study, we suggest here a “cofactor‐regulated LLPS model of SOD1” (Fig 7), which shows that Zn (and not Cu) removal induces sampling of disordered states and the generation of LLPS. We also observe that further incubation of the condensates results in a liquid‐to‐solid‐like transition, which culminates in aggregation.

Figure 7. A schematic describing the mechanism by which Zinc regulates LLPS and aggregation of SOD1.

In the absence of Zn, SOD1 exhibits conformational disorder (instability) in the loop regions (IV/VII), which promotes LLPS and the subsequent maturation of condensates into aggregates. Upon binding to Zn, the loop regions adopt a rigid conformation, which inhibits LLPS.

Recently, Hatters and colleagues (Ruff et al, 2022) showed that mutations, which reduce the thermal stability of globular proteins, can trigger their phase separation in vivo and lead to the formation of unfolded protein deposits (UPODs). These deposits were found to be stabilized by homotypic interactions between hydrophobic core residues, which are exposed upon unfolding. Further, heterotypic interactions between unfolded proteins could promote co‐localization within condensates as observed for barnase (I25A, I196G) and SOD1 (A4V). Interestingly, Ebbinghaus and colleagues (Samanta et al, 2021) observed that the partitioning of SOD1 into SGs in response to heat shock treatment was more sensitive to the modulation of specific protein–protein interactions in the cytoplasm than on fold stability.

Here, we show that the loss of Zn, which promotes solvent exposure of disordered loop regions, is sufficient to trigger phase separation of SOD1 under physiological conditions. Our CG simulations of ApoSOD1^2SH suggest that interactions that involve the disordered loop‐IV/VII regions can promote its phase separation. Notably, Oliveberg and co‐workers (Yang et al, 2018) showed that the presence of disordered loops IV/VII destabilizes the ApoSOD1^2SH monomer by ~ 3 kcal/mol. The structural destabilization induced by the absence of Zn (either in ApoSOD1^2SH or in the Zn‐compromised ALS mutants) can result in local unfolding and/or more efficient sampling of the transient locally unfolded structures. These locally unfolded states have been shown to generate amyloid formation in globular folded proteins (Chiti & Dobson, 2009), and a similar mechanistic interpretation can also be applied in the present case of SOD1 phase separation, which eventually matures into the amyloid aggregation.

Our results correlate well with the severity of disease phenotypes for three ALS disease mutants under in vitro conditions. The less severe G37R SOD1^2SH mutant is compact with a lower tendency towards LLPS. By contrast, more severe I113T SOD1^2SH and G85R SOD1^2SH mutants, with reduced Zn affinity are more disordered, readily forming droplets and subsequent aggregates. Interestingly, unlike I113T SOD1^2SH (and ApoSOD1^2SH), the disease mutant G85R SOD1^2SH exhibits a significant reduction in Zn‐binding affinity due to non‐native salt bridge formation as substantiated by atomistic simulation data. A number of complementary assays presented in our study suggest that the cellular toxicity of SOD1 condensates is higher than large aggregates, which form after maturation. One limitation of these assays where condensates are added to the cells exogenously in the cell culture media is that it is difficult to ascertain whether the condensates remain in their initial state or transform to a different state, given their transient nature and uncharacterized material properties. Keeping this limitation in mind, we note the similarity of our data to the results of Lehner and colleagues (Bolognesi et al, 2019). They generated more than 50,000 mutants in the prion‐like domain of TDP‐43 and showed that LLPS‐promoting mutants exhibited increased toxicity in yeast cells while aggregate‐forming mutants reduced cellular toxicity. Detailed and careful investigations are needed to characterize the toxicity of the droplets with respect to their interconversion dynamics, material properties, and cellular uptake properties.

Several studies emphasize the toxicity induced by small oligomers as opposed to large aggregates in ALS and other neurodegenerative diseases (Vega et al, 2019; Chakraborty et al, 2021). Incubation of ApoSOD1^S‐S in vitro under physiological conditions was shown to form soluble, non‐native oligomers, which persisted even under reducing conditions (Redler et al, 2014). Based on NMR experiments, Kay and colleagues reported that the free energy landscape of immature ApoSOD1^2SH is rugged and comprises four transiently populated conformers in equilibrium with the native state (Sekhar et al, 2015). Two of the conformers exist as non‐native oligomers and are regarded as “off‐pathway” states to the mature, native SOD1 dimer. Key events during SOD1 maturation such as metalation and oxidation assist in eliminating the off‐pathway oligomers that correspond to high energy states in immature ApoSOD1^2SH (Culik et al, 2018). Non‐native, trimeric SOD1 oligomers, which are transiently formed, were found to strongly promote neuronal cell death (Proctor, 2016). Interestingly, it was observed that LLPS of Tau promoted the acquisition of pathogenic conformations and the formation of toxic oligomers (Kanaan et al, 2020). Similarly, we speculate that LLPS may facilitate the formation of toxic SOD1 oligomers and would explain the greater toxicity of SOD1 condensates compared with aggregates. A detailed understanding of how condensates metamorphose to inclusion bodies could open possibilities for developing therapeutic drugs targeting these non‐native oligomers instead of aggregate species.

Materials and Methods

Materials

For the purification of SOD1, LB media, isopropyl‐1‐thio‐β‐d‐galactopyranoside (IPTG), CuSO₄, Tris–HCl, NaCl, imidazole, dithiothreitol (DTT), sodium phosphate monobasic, sodium phosphate dibasic, sodium acetate, and EDTA from Sigma Aldrich, St. Louis, USA were used. Ni‐NTA resin from Thermo Fisher Scientific, MA, USA was used. For LLPS study, HEPES, heparin, polyethylene glycol (PEG) 8,000, and polyethylenimine (PEI) were purchased from Sigma Aldrich, MO, USA. For protein labelling, AlexaFluor488 maleimide and tris 2‐carboxyethyl phosphine (TCEP) from Thermo Fisher Scientific, MA, USA, and Sigma Aldrich, MO, USA, were used, respectively. All media, supplements, reagents, and kits for cell culture were obtained from Invitrogen, CA, USA, unless stated otherwise.

Recombinant SOD1^2SH purification

SOD1 plasmid was transformed into competent E. coli (BL21 DE3 strain) cells using the heat shock method. Transformed bacterial cells were grown in LB media at 37°C. The overexpression of SOD1 was induced with 1 ml of 1 M IPTG in 1 L culture after cells reached the log phase (at OD₆₀₀ ~ 0.6–0.8). 1 mM CuSO₄ was added to the culture for proper metal loading over the protein. The cells were allowed to grow for 3.5 h after which they were pelleted down by centrifugation at 4,000 g for 15 min at 4°C. Cells were re‐suspended in prechilled lysis buffer (20 mM Tris–HCl + 500 mM NaCl, pH 8.0). After thorough re‐suspension in lysis buffer, cells were sonicated (20 pulses, each of 30 s with an interval time of 1 min) to allow cell lysis. Unbroken cells and debris were removed by another centrifugation step at 16,100 g for 30 min. The supernatant obtained was carefully removed and allowed to bind to Ni‐NTA agarose resin. The Ni‐NTA column was washed using 50 ml wash buffer (20 mM Tris–HCl, 500 mM NaCl, and 50 mM imidazole, pH 8.0) followed by elution with 20 mM Tris–HCl, 500 mM NaCl, 500 mM imidazole, pH 8.0, and 1 mM DTT. Protein concentration of eluted fractions was determined by their absorbances at 280 nm. The eluted fractions were pooled and dialyzed using SnakeSkin Dialysis Tubing (10 KDa MWCO) in 20 mM Na‐phosphate buffer and 1 mM DTT, pH 7.4. The protein concentration of the postdialyzed fraction was estimated by recording absorbance at 280 nm using the monomeric molar extinction coefficient of 5,500 M⁻¹ cm⁻¹ (Wright et al, 2013).

Preparation of ApoSOD1^2SH

ApoSOD1 was prepared from the WTSOD1^2SH by metal removal following previously reported protocol (McCords & Fridovich, 1969). WTSOD1^2SH was subjected to overnight dialysis in 50 mM Na acetate, 10 mM EDTA, pH 3.8 to carry out the proper removal of metal ions. EDTA was removed by successive dialysis in 50 mM Na acetate pH 5.2 and in 20 mM Na‐phosphate, pH 7.4. This was followed by dialysis in Na‐phosphate buffer, pH 7.4.

Size exclusion chromatography and MALDI mass spectrometry

100 μM of freshly purified protein (G37R SOD1^2SH, ApoG37R SOD1^2SH, I113T SOD1^2SH, ApoI113T SOD1^2SH, G85R SOD1^2SH, and ApoG85R SOD1^2SH) were loaded on a gel filtration column (Bio‐Rad Enrich SEC 70 10 × 300) fitted to a Fast Protein Liquid Chromatography (FPLC) system (Bio‐Rad NGC chromatography system, Bio‐Rad Laboratories, USA) and absorbance change at 280 nm of the elution was monitored. The molecular weights of the eluted fractions of the proteins collected after the run were determined by Matrix Assisted Laser Desorption/Ionization (MALDI) mass spectrometer (4800 MALDI‐TOF/TOF Analyzer, Applied Biosystems/ MDS SCIEX, USA).

Native PAGE

Freshly purified SOD1 variants (G37R SOD1^2SH, ApoG37R SOD1^2SH, I113T SOD1^2SH, ApoI113T SOD1^2SH, G85R SOD1^2SH, and ApoG85R SOD1^2SH) were run on a 15% nondenaturing polyacrylamide gel. The wells were loaded with 20 μg protein. 20 μg Bovine Serum Albumin (BSA) and lysozyme were used as protein standards.

Liquid–liquid phase separation in vitro

To study LLPS and liquid droplet formation, 100 μM SOD1^2SH and its mutants, respectively, were incubated in 20 mM HEPES buffer (pH 7.4), ~ 7% (w/v) heparin, and 100 mM NaCl for 20 min at 37°C with constant shaking of 180 RPM. 10 μl of the reaction mixture was aliquoted and drop casted onto grease‐free glass slides with single concavity and covered with a 22 mm coverslip (Blue Star, India), sealed with commercially available nail polish. LLPS was also induced with neutral crowding agent PEG‐8000 from 5% to 20% (w/v) and cationic polymer PEI 5% and 10% (w/v) and 100 mM NaCl for 20 min at 37°C with constant shaking of 180 RPM. The slides were visualized with a 10× and 40× objective using Leica microsystems DMIL LED inverted fluorescence microscope (Wetzlar, Germany) in the DIC mode and fluorescence mode for imaging droplets composed of labeled protein. Droplet numbers from different fields were computed using ImageJ software (NIH, MD, USA). The optical density of the samples was measured at 600 nm with Hidex Sense Microplate Reader (Turku, Finland). The turbidity was measured as optical density from three independent measurements.

Fluorescent labelling of protein

All SOD1^2SH protein variants were labeled using a thiol active fluorescence dye AlexaFluor 488 maleimide (Invitrogen, MA, USA) following a previously established protocol (Kundu et al, 2017). The fluorescence dye was dissolved in DMSO and added to 2 mg/ml solution of protein under constant stirring. The molar ratio between the protein and dye was 1:10. The reaction mixture was incubated at 4°C for 5 h with vortexing after every 30 min. The labelling reaction was then quenched by adding excess β‐mercaptoethanol. Excess free dye from the reaction mixture was removed by extensive dialysis in Na‐phosphate (pH 7.4) buffer using SnakeSkin (Thermo Fisher Scientific, MA, USA) Dialysis Tubing (10 KDa MWCO) followed by column chromatography using a Sephadex G20 column (Bio‐rad Laboratories, CA, USA), which was pre‐equilibrated with 20 mM Na‐phosphate buffer (pH 7.4).

Lifetime and anisotropy of protein condensates

15 nM‐labeled ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH along with 100 μM unlabeled protein were subjected to LLPS conditions, and droplets were visualized under a confocal microscope for time‐domain FLIM experiments. The time‐resolved fluorescence measurements were made using a time‐correlated single photon counting (TCSPC) setup of Alba (ISS Inc., IL, USA). Measurement samples were excited using a 488 nm QuixX picosecond pulsed laser made by Omicron‐Laserage Laserprodukte GmbH (Germany). The repetition rate of the laser was set to be 20 MHz. The laser was linearly polarized in the vertical direction, and a linear polarization cleanup filter (DPM‐100‐VIS by Meadowlark Optics) was used for further improving the extinction ratio. For FLIM measurements, the fluorescence emission was detected by a single photon avalanche diode (SPAD) detector (SPD‐100‐CTC by Micro Photon Devices) after the 530/43‐nm band‐pass filter (Semrock, NY, USA). For the anisotropy measurements, the fluorescence emission was separated by the polarization beam splitter into two channels, which are parallel and perpendicular to the orientation of the linear polarization of the excitation; both parallel and perpendicular emissions were simultaneously detected by two SPAD detectors using the same 530/43‐nm band‐pass filters. Both the FLIM and the anisotropy data were analyzed using the ISS 64‐bit VistaVision software. For FLIM data analysis, the software allows the single or multiexponential curve fittings on a pixel‐by‐pixel basis using a weighted least‐squares numerical approach (Lakowicz, 1999; Gryczynski et al, 2009; Sun et al, 2011). The double‐exponential model was used for fitting the lifetime data of diffused and condensed ApoSOD1^2SH with the instrument response function, estimated by taking the first derivative of the rising of the decay. For time‐resolved anisotropy data analysis, the software performs the global fittings of both fluorescence lifetimes and rotation times. The steady‐state anisotropy (r) was calculated by

$$r=\left({\mathrm{I}}_{\mathrm{par}}-{\mathrm{I}}_{\mathrm{per}}\right)/\left({\mathrm{I}}_{\mathrm{par}}+{\text{2I}}_{\mathrm{per}}\right)$$ (1)

where I_par and I_per are the measured intensities in the parallel and perpendicular channels, respectively.

Fluorescence correlation spectroscopy

Fluorescence Correlation Spectroscopy (FCS) measures the fluctuations of fluorescence intensity in the confocal volume and yields the diffusion times of the fluorescent species. For nanomolar level binding study using FCS, 15 nM Alexa 488 labeled monomeric protein mixed with unlabeled protein keeping total unlabeled protein (demetallized) concentration at 100 nm in the presence of increasing concentration of Zn from 1:0 to 1:5 molar ratio and 10 mM TCEP. The FCS measurements were carried out using an ISS Alba FFS/FLIM confocal system (Champaign, IL, USA), coupled to a Nikon Ti2U microscope equipped with the Nikon CFI PlanApo 60X/ 1.2NA water immersion objective. The 48‐nm picosecond pulsed diode laser was used for the excitation of the FCS measurements. The fluorescence emission was collected using a pair of SPAD (Single Photon Avalanche Detector) detectors with the 50/50 beam splitter and the 530/43‐nm band‐pass filter. The use of two detectors enabled us to determine single‐color cross‐correlation functions to eliminate the artifacts given the detector after‐pulsing. The FCS correlation curves were fit to the 3D Gaussian 1‐component diffusion model, where the beam waists in the radial and axial dimensions were calibrated using a standard fluorescence dye (Rhodamine 6G, R6G) in water of known diffusion coefficient (2.8 × 10⁻⁶ cm²/s). Diffusion coefficients of different mutants are measured by measuring the diffusion time (${\tau }_D$) and using the relation (Chatterjee et al, 2019):

$$\frac{D_{\text{protein}}}{D_{\mathrm{R}6\mathrm{G}}}=\frac{{\mathrm{\tau }}_{D,\mathrm{R}6\mathrm{G}}}{{\mathrm{\tau }}_{D,\text{protein}}}$$ (2)

The hydrodynamic radii of each species were computed from their respective diffusion coefficients (D). For comparison between hydrodynamic radii of the wild‐type protein and its mutants, Alexa Fluor 488 labeled monomeric protein (15 nm) and 100 nm unlabeled protein in presence of 10 mM TCEP in 20 mM Na‐phosphate buffer (pH 7.4) was used. The FCS data obtained were normalized with respect to free dye following a previously published method (Chattopadhyay et al, 2005).

In the 3D Gaussian diffusion model involving a single type of diffusing molecules (the 3D Gaussian 1‐component model excluding the contributions of the triplet state), the correlation function G($\mathrm{\tau }$) can be defined by the following equation:

$$\mathrm{G}\left(\mathrm{\tau }\right)=1+\frac{1}{N}.\left(\frac{1}{1+\left(\frac{\mathrm{\tau }}{{\tau }_D}\right)}\right).\left(\frac{1}{\sqrt{1+{\mathrm{S}}^2\left(\frac{\tau }{{\tau }_{\mathrm{D}}}\right)}}\right)$$ (3)

where τ _D denotes the diffusion time of the diffusing molecules, N is the average number of molecules within the observation volume, and S is the structural parameter that defines the ratio between the radius and the height. The value of ${\mathrm{\tau }}_D$ obtained by fitting the correlation function is related to the diffusion coefficient (D) of a molecule by the following equation:

$${\mathrm{\tau }}_D={\mathrm{\omega }}^2/4D$$ (4)

where ω is the size of the observation volume. From here, the value of the hydrodynamic radius (r _H) of the protein molecule/ complex/ aggregate can be obtained from D using the Stokes−Einstein formula:

$$D=\mathrm{k}T/6\mathrm{\pi \eta }{r}_{\mathrm{H}}$$ (5)

where η is the viscosity, T is the absolute temperature and k is the Boltzmann constant.

All‐atom MD simulation protocol and analysis

Initial structures for all ApoSOD1^2SH monomers were taken from PDB entry 2C9V (Chain F). Structural mutations were made through rotamer substitutions using the Dunbrack rotamer library in UCSF Chimera (Pettersen et al, 2004). Initial conformation for the all‐atom simulation of unfolded SOD1 was taken from a 1 μs single‐chain simulation performed using the single bead per‐residue, HPS‐Urry model (Mammen Regy et al, 2021). The coarse‐grained structure was converted to an all‐atom model using MODELER (Eswar et al, 2008). All systems were modeled based on the AMBER99SB‐disp force field (Robustelli et al, 2018) along with a modified version of the TIP4P‐D water model (Piana et al, 2015). Force field parameters for Zn and Zn‐coordinated residues were obtained from previous studies (Macchiagodena et al, 2019, 2020). Energy minimization and equilibration were performed using GROMACS 2020 (Páll, 2020). The SOD1 monomer was placed into an octahedral box of 6.5‐nm length. Energy minimization of the protein was first performed in vacuum using the steepest descent algorithm. Following vacuum minimization, TIP4P‐D water molecules were added, and the solvated system was further minimized using the steepest descent algorithm. To mimic physiological salt concentration (0.1 M), Na⁺ and Cl⁻ ions were added along with additional Na⁺ counter ions to achieve electrical neutrality. NVT equilibration was performed using Noose‐Hoover thermostat (Τ_c = 1.0 ps) to stabilize the system temperature at 300 K. NPT equilibration was performed using the Berendsen barostat (Berendsen et al, 1984) with isotropic coupling (Τ_p = 5.0 ps) to achieve a system pressure of 1 bar. All production simulations were performed using OpenMM 7.5 (Eastman et al, 2017) in the canonical ensemble at 300 K using the Langevin middle integrator (Zhang et al, 2019) with a friction coefficient of 1 ps⁻¹. Masses of all hydrogen atoms were increased to 1.5 amu, which allowed for a simulation timestep of 4 fs. Constraints were applied to all hydrogen‐containing bonds using the SHAKE algorithm (Ryckaert et al, 1977). Short‐range nonbonded interactions were calculated based on a cutoff radius of 0.9 nm. Long‐range electrostatic interactions were treated using the PME method (Darden et al, 1993).

Hydrodynamic radii were calculated using the HullRad algorithm (Fleming & Fleming, 2018) and averaged over 500 ns intervals. All other analysis was carried out using analysis programs available within GROMACS. Secondary structure fractions were calculated based on the DSSP library (Kabsch & Sander, 1983) using gmx do dssp. Per‐residue root mean‐square fluctuation (RMSF) and distance root mean square deviation (dRMSD) were calculated using gmx rmsf and gmx rmsdist, respectively. Minimum distances in G85R ApoSOD1^2SH were measured between arginine Nη atoms and aspartate Oδ atoms using gmx mindist.

CG MD slab simulation protocol and analysis

Phase coexistence simulation of ApoSOD1^2SH and its variants were conducted using the HOOMD‐Blue 2.9.3 software package (Anderson et al, 2020) with features extended using azplugins (version 0.10.1; https://github.com/mphowardlab/azplugins), using a protocol proposed in our previous works (Dignon et al, 2018; Mammen Regy et al, 2021). We simulated “partially rigid/flexible” SOD1 chains wherein loop IV (aa: 49–81) and VII (aa: 124–139) of SOD1 are allowed to be flexible while the rest of the protein remains folded/rigid. Regions except Loop IV/VII were constrained as a rigid body using the hoomd.md.constrain.rigid function (Nguyen et al, 2011; Glaser et al, 2020), keeping the initial structure from PDB entry 2C9V. Partially folded SOD1 was modeled based on the HPS‐Urry model (Regy et al, 2021), in which solvent is treated implicitly. The initial slab geometry (14 × 14 × 280 nm) was prepared from the 100 chains of SOD1 sequence, then a 5 μs NVT simulation (with 10 fs time step) was conducted at 275 K using a Langevin thermostat with the residue friction factor, γ_i = m_i/τ _damp. Here m_i is the mass of each amino acid bead (g/mol), τ _damp is the damping parameter, which was set to 1,000 ps. The first microsecond of the trajectory was treated as equilibration time and skipped for the calculation of the density profile and contact maps.

Contact analysis of the coarse‐grained simulation was carried out with MDAnalysis (Michaud‐Agrawal et al, 2011). Two amino acids (i and j) were considered in contact in the simulations when their inter‐bead distance was less than 1.5σ_ij where σ_ij = (σ_i + σ_j)/2 is the average of the bead diameter of the respective amino acids i and j. The intermolecular contact probability (P = mean(n _ij) where n _ij = 1 if distance criteria are met, otherwise n _ij = 0) was computed by averaging the individual contact over the trajectory. The residue index contact map is plotted as − ln(P/P _max), where P _max is the maximum probability observed between any pairs of amino acids in the given sequence.

The concentrations c_sat (μM) and c_dense (μM) of vapor and dense phases along Z, respectively, were determined by centering the trajectory on the slab for each frame so that the dense phase with the largest number of chains is placed at z = 0 for all frames. This is done using clustering according to the center‐of‐mass distance between chain pairs. All the measured properties were divided into four independent blocks to estimate error bars.

Time‐based maturation study using fluorescence correlation spectroscopy

100 μM unlabeled ApoSOD1^2SH with 15 nM Alexa‐488 labeled ApoSOD1^2SH was incubated at 37°C, 180 RPM for maturation study for 28 h. 10 μl of the sample aliquoted from different time points 30 min postincubation (marked as 0*), 12 h, and ~ 25 h were drop casted in a depression slide (Blue Star, India) and mounted with 22 mm coverslip (Bluestar, India) and sealed with commercially available nail polish. The slide was visualized under the confocal microscope using TD‐FLIM (Nikon Eclipse Ti2, Japan) at 60× water emersion objective. FCS measurements were taken from regions with varying intensities within the droplet using point FCS mode. The FCS correlation curves were fit to the 3D Gaussian 1‐component diffusion model, where the beam waists in the radial and axial dimensions were calibrated using a standard fluorescence dye (Rhodamine 6 G, R6G) in water of known diffusion rate.

Thioflavin T (ThT) assay

For monitoring the effect of Zn on the aggregation kinetics of SOD1^2SH, all protein variants were incubated in the presence and absence of Zn under shaking at 180 RPM at 37°C for 350 h. The protein concentrations for the aggregate preparation were 100 μM in 20 mM HEPES buffer at pH 7.4. Aliquots of proteins were withdrawn at each time point in the aggregation pathway and diluted in 20 mM HEPES buffer at pH 7.4 to reach a volume of 500 μl. ThT was added to protein in a 1:10 molar ratio (protein: ThT). Steady‐state fluorescence measurements were recorded in a quartz cuvette of path length 1 cm using a Photon Technology International (PTI) fluorescence spectrometer with excitation wavelength of 450 nm, emission wavelength of 485 nm and an integration time of 0.1 s averaging over three times. For seeded aggregation, 10% (10 μM) seed of prepared condensates and aggregates were added to 100 μM protein, respectively, and studied for aggregation by detecting ThT fluorescence as mentioned previously.

Atomic force microscopy

Aggregating samples of SOD1^2SH and its mutants were aliquoted after prolonged incubation at 37°C and diluted 20 times with milliQ water. A 5 μl diluted sample was drop casted on freshly cleaved mica. The aggregates were rinsed with MilliQ water and then dried using a stream of nitrogen. Images were acquired at room temperature using a Bioscope Catalyst AFM (Bruker Corporation, MA, USA) with silicon probes. The standard tapping mode was used to image the morphology of aggregates. The nominal spring constant of the cantilever was kept at 20–80 N/m. The spring constant was calibrated by a thermal tuning method. A standard scan rate of 0.5 Hz with 512 samples per line 6 was used for imaging the samples. A single third‐order flattening of height images with a low pass filter was done followed by section analysis to determine the dimensions of aggregates.

Cell culture and cytotoxicity assay

Neuroblastoma cell lines SHSY5Y acquired from the national cell repository (National Centre for Cell Science, Pune, India) were authenticated using STR analysis. The cells tested negative for mycoplasma contamination as tested using PCR. Cells were maintained in Dulbecco's modified Eagle's media (DMEM), which in turn were supplemented with 10% heat‐inactivated fetal bovine serum (FBS), respectively, 4.5 g/L of glucose, 1.5 g/L sodium bicarbonate, 110 mg/L sodium pyruvate, 4 mM l‐glutamine, 50 units/ml penicillin G, and 50 μg/ml streptomycin in humidified air containing 5% CO₂ at 37°C. Sub‐culturing was done by allowing the passaging of cells as per ATCC recommendations (ATCC, VA, USA).

Trypan Blue exclusion assay was performed to determine the number of viable cells present in cell suspension after a 12 h treatment with 5 μM SOD1^2SH condensates, oligomer, and fibrils. It is based on the principle that viable cells have intact cell membranes that exclude certain dyes such as trypan blue. SHSY5Y cells were seeded in 6‐well plates at 1 × 10⁶ cells/well and incubated for 24 h at 37°C, 5% CO₂. 5 μM preformed ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH condensates, oligomers and fibrils were added to each well, respectively, and incubated for 12 h at 37°C, 5% CO₂. Following incubation, cells were trypsinized and 20 μl 0.4% trypan blue solution from Sigma Aldrich, MO, USA was added to 20 μl cell suspension. 20 μl of this mixture was added between the hemocytometer and cover slip. The loaded hemocytometer was examined under a light microscope at 10× magnification. The number of blue‐stained cells and total cells were counted. Percentage of viable cells was calculated from the formula:

$$\%\text{viable cells}=\left[1-\left(\text{number of blue cells}/\text{total number of cells}\right)\right]\times 100$$ (6)

MTT assay was employed to evaluate the cellular cytotoxicity on the addition of SOD1^2SH condensates, oligomers, and fibrils to determine which structures were the most toxic. SHSY5Y cells (4 × 10³ cells per well) were seeded in a 96‐well plate and incubated at 37°C, 5% CO₂ for 24 h. After 24 h, cells were treated with preformed ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH condensates, oligomers, and fibrils at a concentration of 5 μM for 12 h. After 12 h of incubation, cells were washed with PBS, and then following the manufacturer's protocol, 10 μl of 12 mM MTT stock solution from Invitrogen Vybrant MTT Cell Proliferation Assay kit was added to each well and kept in incubator at 37°C, 5% CO₂ for 4 h to form formazan salt. The formazan salt was solubilized by adding 50 μl DMSO to each well and mixing thoroughly, following an incubation of 10 min at 37°C. The absorbance was recorded at 540 nm using an ELISA reader (Emax, Molecular Device). DMEM was used as the blank or negative control and untreated cells were used as the positive control. Percentage cell viability was calculated from the formula:

$$\%\text{cell viability}=\left[\left({\text{absorbance}}_{\text{sample}}-{\text{absorbance}}_{\text{blank}}\right)/\left({\text{absorbance}}_{\text{control}}-{\text{absorbance}}_{\text{blank}}\right)\right]\times 100$$ (7)

For the FITC‐Annexin V/ Propidium Iodide early and late apoptosis analyses, cells were seeded in 6 wells plates at 1 × 10⁶ cells/ well. 24 h after seeding, the cells were subjected to preformed ApoSOD1^2SH, I113T SOD1^2SH, and G85R SOD1^2SH condensates, oligomers, and fibrils to understand which structures were the most toxic. The final concentration of the protein added to the cells was maintained at 5 μM. After the addition of the treatments, the cells were incubated for 12 h at 37°C, 5% CO₂. The percentage of apoptotic cells was determined using the Invitrogen Dead Cell Apoptosis Kit following instructions from the manufacturer. Apoptosis is a cellular process that entails a genetically programmed series of events leading to the death of a cell. During early apoptosis, the lipid phosphatidylserine (PS) is translocated to the outer leaflet of the plasma membrane from the cytoplasmic side. FITC‐conjugated Annexin V is a strong probe for the exposed PS and can be used for detecting early apoptosis in stressed cells. For the determination of late apoptosis and/or necrosis as a result of oligomer treatment, propidium iodide (PI) was added to the treated cells at a concentration of 2 μg/ml. PI labels the cellular DNA in late apoptotic/ necrotic cells where the cell membrane has been completely ruptured, and the nuclear membrane has been dilated to release nucleus contents. The BD LSRFortessa flow cytometer was employed for Fluorescence Activated Cell Sorting (FACS) to analyze the apoptotic, necrotic, and viable cell populations in treated SHSY5Y cells.

Fourier transform infrared spectroscopy (FTIR)

FTIR spectra of ApoSOD1^2SH and SOD1^2SH mutants in the presence of increasing concentrations of zinc and copper were acquired using Bruker 600 series FTIR spectrometer. Protein samples at a concentration of 20 μM were treated with increasing concentrations of zinc sulphate and copper sulphate, from molar ratios 1:1 to 1:5 (protein:metal ion), respectively, in 20 mM Na‐phosphate buffer at pH 7.4 and incubated for 15 min at room temperature before the measurements. All the FTIR measurements were carried out in Na‐phosphate buffer. The experiments were carried out in solution, and the buffer baseline was subtracted before recording each spectrum. The spectral readouts were obtained on absorbance mode with a path length of 0.01 mm following standard methodology. The deconvolution of raw spectra in the amide I region (1,700–1,600 cm⁻¹) was done using least‐squares iterative curve fitting to Gaussian/Lorentzian line shapes. Peak identifications were typically carried out by using double derivatives of the FTIR spectra, as described before (Goormaghtigh et al, 1990; Sarkar‐Banerjee et al, 2016; Mahapatra et al, 2019). We used OriginPro 9 software for the curve fitting, second derivative analysis, and other data fitting.

Author contributions

Krishnananda Chattopadhyay: Conceptualization; resources; supervision; funding acquisition; writing – review and editing. Bidisha Das: Formal analysis; validation; investigation; writing – original draft; writing – review and editing. Sumangal Roychowdhury: Data curation; formal analysis; validation; writing – original draft; writing – review and editing. Priyesh Mohanty: Formal analysis; methodology; writing – original draft; writing – review and editing. Azamat Rizuan: Formal analysis; methodology; writing – original draft; writing – review and editing. Jeetain Mittal: Conceptualization; supervision; writing – original draft; project administration; writing – review and editing. Joy Chakraborty: Visualization.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Movie EV1

Movie EV2

Movie EV3

Source Data for Appendix

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

The EMBO Journal (2023) 42: e111185

Data availability

This study includes no data deposited in external repositories.