Decision letter: The metal cofactor zinc and interacting membranes modulate SOD1 conformation-aggregation landscape in an in vitro ALS model

Abstract

Article Figures and data Abstract eLife digest Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Aggregation of Cu–Zn superoxide dismutase (SOD1) is implicated in the motor neuron disease, amyotrophic lateral sclerosis (ALS). Although more than 140 disease mutations of SOD1 are available, their stability or aggregation behaviors in membrane environment are not correlated with disease pathophysiology. Here, we use multiple mutational variants of SOD1 to show that the absence of Zn, and not Cu, significantly impacts membrane attachment of SOD1 through two loop regions facilitating aggregation driven by lipid-induced conformational changes. These loop regions influence both the primary (through Cu intake) and the gain of function (through aggregation) of SOD1 presumably through a shared conformational landscape. Combining experimental and theoretical frameworks using representative ALS disease mutants, we develop a ‘co-factor derived membrane association model’ wherein mutational stress closer to the Zn (but not to the Cu) pocket is responsible for membrane association-mediated toxic aggregation and survival time scale after ALS diagnosis. eLife digest Amyotrophic lateral sclerosis, or ALS, is an incurable neurodegenerative disease in which a person slowly loses specialized nerve cells that control voluntary movement. It is not fully understood what causes this fatal disease. However, it is suspected that clumps, or aggregates, of a protein called SOD1 in nerve cells may play a crucial role. More than 140 mutations in the gene for SOD1 have been linked to ALS, with varying degrees of severity. But it is still unclear how these mutations cause SOD1 aggregation or how different mutations influence the survival rate of the disease. The protein SOD1 contains a copper ion and a zinc ion, and it is possible that mutations that affect how these two ions bind to SOD1 influences the severity of the disease. To investigate this, Sannigrahi, Chowdhury, Das et al. genetically engineered mutants of the SOD1 protein which each contain only one metal ion. Experiments on these mutated proteins showed that the copper ion is responsible for the protein’s role in neutralizing harmful reactive molecules, while the zinc ion stabilizes the protein against aggregation. Sannigrahi et al. found that when the zinc ion was removed, the SOD1 protein attached to a structure inside the cell called the mitochondria and formed toxic aggregates. Sannigrahi et al. then used these observations to build a computational model that incorporated different mutations that have been previously associated with ALS. The model suggests that mutations close to the site where zinc binds to the SOD1 protein increase disease severity and shorten survival time after diagnosis. This model was then experimentally validated using two disease variants of ALS that have mutations close to the sites where zinc or copper binds. These findings still need to be tested in animals and humans to see if these mechanisms hold true in a multicellular organism. This discovery could help design new ALS treatments that target the zinc binding site on SOD1 or disrupt the protein’s interactions with the mitochondria. Introduction The aggregation of SOD1 is believed to be one of the chief causative factors behind the lethal motor neuron disease, amyotrophic lateral sclerosis (ALS) (Shaw and Valentine, 2007). Although more than 140 SOD1 mutations have been reportedly associated with ALS, there is no correlation between the stability (and aggregation) of these mutations and their disease manifestations. SOD1 aggregation has been investigated extensively in vitro by altering the solution conditions, such as temperature, pH, the presence of metal chelators, and the reduction of disulfide linkages (Bush, 2002; Niwa, 2007; Rodriguez et al., 2005). The results of these studies clearly suggest that the aggregation of SOD1 is heterogeneous containing multiple steps, which is presumably the reason behind the lack of a structural understanding of aggregation processes (Pasinelli et al., 2004; Tomik et al., 2005). WT SOD1 contains Cu2+ (Cu) and Zn2+ (Zn) as cofactors. It has been established that Cu is responsible for the primary function of SOD1 (the dismutase activity), and cell membrane acts as a scaffold in the process of Cu transfer to apo-SOD1 (metal free non-functional protein) through a Cu delivery chaperone (CCS) (Culotta et al., 1997). Previous studies have found noticeable presence of SOD1 in human serum lipoproteins, mainly in LDL and HDL, hinting at a possible protective role of SOD1 against the lipid peroxidation (Mondola et al., 2016). It has also been noted that SOD1 has a physiological propensity to accumulate near the membranes (Ilieva et al., 2009) of different cellular compartments, including mitochondria, endoplasmic reticulum (ER), and Golgi apparatus (Manfredi and Kawamata, 2016). In addition, computational studies have shown that the electrostatic loop (loop VII, residues 121–142) and Zn-binding loop (loop IV, residues 58–83) promote membrane interaction of apo-SOD1 initiating the aggregation process (Chng and Strange, 2014). Membrane binding induced aggregation of SOD1 has also been shown experimentally both in vitro and inside cells (Hervias et al., 2006; Yamanaka et al., 2008; Choi et al., 2011). Inclusions of SOD1 have been detected in the inter-membrane space of mitochondria originating from the spinal cord (Mondola et al., 2016). The above results can be reconciled by suggesting that cell membrane can play crucial roles not only in shaping up the primary function of the protein, but also in defining its aggregation process of generating fibrillar and non-fibrillar aggregates (the gain of function), with the loops IV and VII contributing critically to both processes. We hypothesize that (1) the induction of metal cofactors for the stabilization of loop IV and VII, membrane interaction, and SOD1 aggregation would be some of the crucial elements in defining the overlapping folding-aggregation landscape of SOD1; (2) metal pocket perturbation by mutational stresses (as in disease variants) would modulate membrane association and facilitate aggregation, and (3) the difference in aggregate morphology as a result of differential membrane interaction may contribute to the variation in cellular toxicity observed in ALS. In this paper, we investigated the above hypothesis by studying how different structural elements (i.e. co-ordination of individual metals, membrane association, and the location of mutations) attenuate the toxic gain of function of SOD1. An effective understanding of the role of individual metals (Cu and Zn) would require studying SOD1 variants containing only one metal (Cu or Zn) in addition to a variant that contains none. We have therefore prepared an apo (metal free) protein, which serves the latter purpose. For the former, we have generated two single metal containing mutants of SOD1, viz. H121F (only Zn, no Cu) and H72F (only Cu, no Zn), which are situated near the key loop VII (H121F) and loop IV (H72F) at the protein structure (Figure 1a). Figure 1 with 1 supplement see all Download asset Open asset Statistical mechanical modeling of SOD1 folding mechanism. (a) Cartoon representation of SOD1 monomer highlighting the various structural elements. The positions of the mutation for H121F and H72F have been arrow marked. (b) Residue folding probability as a function of residue index for the different variants of SOD1 as predicted by the bWSME model. Note that WT represents the variant in which both the metallic cofactors are bound. (c) Average folding probabilities colored in the spectral scale going from 0 (dark blue) to 1 (dark red) as a function of the reaction coordinate, number of structured blocks. The vertical white dashed line signals the parts of the protein that fold first. For example, it can be seen that residues 1–40 fol d early in the apo SOD1 (dark red) when compared to WT where residues 40–80 fold first. Using computational analysis based on a statistical mechanical model and detailed in vitro experiments, we propose here a ‘Co-factor derived membrane association model’ of SOD1 aggregation and its possible implication in ALS. We demonstrate that differential metal binding and membrane assisted conformational changes can work in concert to attenuate the rate and propensity of aggregation. While apo (no metal) protein and H72F mutant (no Zn) experience strong membrane interaction, the WT (both metals) protein and H121F (no Cu) mutant do not show significant binding. We further find that membrane-induced aggregates of H72F and apo protein showed significantly higher toxicity in terms of cell death and model membrane deformation when compared to WT and H121F mutant. We finally check the validity of this model to ALS using computational and experimental studies. For the computational validation, we show, using 15 ALS disease mutants, that the distance between the mutation site and Zn correlates well with the membrane binding energy and patient survival time after disease diagnosis, while Cu site does not seem to have any prominent role. For the experimental study, we use two well-studied disease mutants (G37R where mutational site is close to the Cu pocket and I113T where mutational site is near Zn pocket) to show that the model accounts well for their membrane binding/aggregation, correlating well with their disease onset phenotypes. This model puts forward a mechanism that Zn pocket destabilization (either by metal content variation or by mutational stress near Zn center) is the driving force behind the toxic gain of function of SOD1 mediated by the process of membrane association. Results Statistical mechanical modeling of SOD1 folding mechanism hints at aggregation origins The large size of SOD1 (151 residues) precludes a detailed characterization of the conformational landscape, the role of ions in determining the stability-folding mechanism, and the effect of numerous mutations via all-atom simulation methods. To overcome this challenge and to obtain a simple physical picture of how the energetics of folding is governed by metal ions, we resort to constructing the folding landscape of reduced SOD1 variants through the statistical mechanical Wako–Saitô–Muñoz–Eaton (WSME) model (see Materials and methods for model description and parametrization) (Wako and Saitô, 1978; Muñoz and Eaton, 1999). Here, we employ the bWSME model where stretches of three consecutive residues are considered as a block (b) that reduces the total number of microstates from 42.7 million to just ~450,000 (Gopi et al., 2019). The model, however, incorporates contributions from van der Waals interactions, simplified solvation, Debye–Hückel electrostatics, excess conformational entropy for disordered residues, and restricted conformational freedom for proline residues (Naganathan, 2012; Rajasekaran et al., 2016). The predicted average folding path of SOD1 WT (with both Cu and Zn bound) highlights that the folding is initiated around the metal binding regions with early folding of the loop IV (nucleated by Zn) in the unfolded well and aided by flickering structure in the electrostatic loop (loop VII, nucleated by Cu). The rest of the structure coalesces around this initial folding site leading to the native state. This folding mechanism is very similar to that proposed earlier via detailed kinetic studies (Leinartaite et al., 2010). In the absence of metal ions, the apo variant folds through an alternate pathway wherein the folding is nucleated through the first three strands (residues 1–40) following which the rest of the structure folds, recapitulating the results of single-molecule experiments (Figure 1; Sen Mojumdar et al., 2017). It is important to note that the first three strands exhibit higher aggregation propensity as predicted from different computational servers (Figure 1—figure supplement 1–). Interestingly, the folding mechanism of the Zn-bound SOD1 (with no Cu bound) is similar to that of the WT hinting that Zn coordination promotes proper folding. On the other hand, the folding mechanism of the Cu-bound SOD1 (in the absence of Zn) is similar to that of apo-SOD1 with additional folding probability in the region around the electrostatic loop. Taken together, the statistical modeling highlights how the absence of metals and particularly the absence of Zn (or mutations that affect Zn binding and not Cu binding) alters the folding mechanism by populating partially structured states involving beta strands in the unfolded well thus possibly increasing the chances of aggregation. Importantly, the model provides multiple testable predictions on the differential roles of Zn and Cu, which we address below via experiments. Cu-deficient H121F behaves like WT SOD1, whereas zn-deficient H72F behaves like apo We have recently shown that the mutants H121F and H72F contain negligible Cu and Zn, respectively, while the apo protein is completely devoid of metal (Chowdhury et al., 2019). We validated this further using atomic absorption spectroscopy (Table 1) and activity measurements (Figure 2—figure supplement 1). Guanidinium-induced equilibrium unfolding transitions of H121F and H72F were found to be similar (Figure 2—figure supplement 2). We used steady-state tryptophan fluorescence, far UV CD, and FTIR spectroscopy to characterize these different proteins. SOD1 is a single tryptophan protein (Trp32), in which the tryptophan residue has been shown to be partially buried (Muneeswaran et al., 2014). The role of Trp32 within the sequence segment (Lomize et al., 2012) KVWGSIKGL (Gohil and Greenberg, 2009) of high aggregation propensity has been investigated before (Taylor et al., 2007). We found that the formation of apo form resulted in a large shift in Trp32 emission maximum (332 nm for WT protein and 350 nm for apo protein) (Figure 2a). In contrast, other two mono-metallated variants (H121F and H72F) exhibited fluorescence emission maxima at wavelengths, which were intermediate between the WT and apo proteins (342 nm for H121F and 345 nm for H72F) (Figure 2a). Next, we performed acrylamide-quenching experiments to measure the solvent surface exposure of Trp32 for all variants. The values (Table 2) of the Stern−Volmer constant (Ksv) were determined using a straight line fit, as shown in Figure 2—figure supplement 3. Ksv for WT (6.8 ± 0.1 M−1) was significantly lower than that of apo SOD1 (14.3 ± 0.1 M−1). Steady-state fluorescence maxima in combination with acrylamide quenching data suggested an appreciable conformational alteration in going from the WT to the apo form. Interestingly, Zn-starved H72F mutant showed higher Ksv compared to Cu-starved H121F mutant (Figure 2—figure supplement 3, Table 2). Figure 2 with 7 supplements see all Download asset Open asset Structural characterization of SOD1 mutants and membrane association. (a) Steady-state tryptophan fluorescence spectra of WT, apo, and other two metal mutants (H121F and H72F). The WT displays an emission maximum at 332 nm, whereas the apo variant shows a red-shifted spectrum with the emission maximum at 350 nm. On the other hand, H121F and H72F show emission maxima at intermediate wavelengths. Deconvoluted FTIR spectral signatures of (b) WT and (c) apo. Red contour (~1637 cm−1) indicates beta sheet; blue color contour stands for disorder (1644 cm−1) and loops and turns (~1667 cm−1); green contour represents alpha helical character. All these secondary signatures were obtained by considering the amide-I spectra, which arises due to carbonyl frequency (C = O). (d) Percentage of different secondary structural components in WT, apo, H121F, and H72F are shown in this figure. n.s denotes nonsignificant change, while ** stands for significant change with p-value<0.01. Error bars indicate the standard deviation of the data, which were obtained from triplicate experiments. Here, D + T/L stands for Disorder +Turns/Loops. (e) The membrane association of the apo protein as suggested by the OPM calculations. The membrane association of apo protein through the stretches 45–70 and 128–142 has been evaluated from the calculations. The residues which are involved in binding with membrane (Thr54,58, Ala54, Gly56,61, Pro62, Asn53, Glu49, Lys136, Glu132,133) are mentioned. (f) A schematic representation regarding the membrane binding experiments through FCS which suggests that with increasing concentration of DPPC small unilamellar vesicles (SUVs), the alexa labeled free monomeric protein populations (fast component of diffusion model) decreases with concomitant increase in the membrane bound labeled protein that is the slow component. (g) The correlation functions of alexa 488 maleimide labeled apo SOD1 in the absence (black) and presence of DPPC SUVs (red) where DPPC concentration was kept 500 nM. The green correlation curve corresponds to an intermediate DPPC concentration (100 nM). The inset shows the residual distributions of the correlation curves. (h) The hydrodynamic radii of free alexa 488-apo SOD1 and membrane bound labeled apo SOD1 were plotted against the concentrations of added DPPC SUVs. The average hydrodynamic radius of fast component that is free monomeric apo SOD1 (rH1) was found to be 13.5 Å, whereas the average radius for slow membrane bound protein molecule (rH2) was found to be 170 Å. The change of rH1 and rH2 with increasing DPPC SUV concentration remains invariant. (i) Percentage populations of membrane bound alexa-labeled protein variants were plotted against the concentrations of DPPC SUVs added to evaluate the binding affinities of the protein variants towards membranes. (j) Deconvoluted FTIR spectra of apo in membrane (DPPC SUV) bound condition. (k) Percentage of different secondary structural components in WT, apo, H121F, and H72F in the presence of DPPC SUVs are shown in this figure. n.s denotes nonsignificant change, while *** stands for significant change with p-value<0.001. Figure 2—source data 1 Structural characterization and membrane binding of SOD1 protein variants. https://cdn.elifesciences.org/articles/61453/elife-61453-fig2-data1-v2.xlsx Download elife-61453-fig2-data1-v2.xlsx Table 1 Metal contents (Cu and Zn) in WT and other mutants (H121F, H72F, and apo) as obtained from atomic absorption spectroscopy. Protein formsCu contentZn contentWT4.5 µM3.9 µMH121F<1 µM3.8 µMH72F4.1 µM<1.2 µMapo<0.5 µM<0.25 µM Table 2 The values of Stern–Volmer quenching constants for Trp 32 residue of all the protein variants in the absence (Ksv,M−1) and presence of DPPC SUVs (Ksvm, M−1) as obtained from acrylamide quenching for WT SOD1 and all the mutants including apo SOD1. ProteinsKsvKsvmKsv/KsvmWT SOD16.8 ± 0.15.7 ± 0.21.19H121F8.0 ± 0.16.3 ± 0.11.26H72F12.7 ± 0.37.0 ± 0.21.82apo SOD114.3 ± 0.17.6 ± 0.21.88 Far-UV circular dichroism (CD) spectra for WT and apo protein were in line with earlier observations (Figure 2—figure supplement 4; Banci et al., 2007). Specifically, we found a slight broadening in the far UV CD spectrum of the apo protein when compared to the WT. In agreement with steady-state fluorescence data, the far UV-CD spectrum of the Zn-deficient H72F protein was found to be similar to the apo variant, while the WT- and Cu-deficient H121F variant displayed similar spectra. We then used FT-IR spectroscopy to complement far-UV CD results and to obtain a preliminary estimate of the secondary structure contents of the protein variants, using amide-I FTIR spectral region. The carbonyl (C = 0) stretching vibrations at amide-I region provides information related to the secondary structure (beta sheet 1633–1638, alpha helix 1649–1656, disorder and turns and loops 1644 and 1665–1672 cm−1). The analyses of the FT-IR data were carried out using published method using two steps (Yang et al., 2015; Kong and Yu, 2007; Bandyopadhyay et al., 2021). First, the peak positions were assigned using the double derivatives of the FT-IR data for all protein variants (Figure 2—figure supplement 5 shows the representative double derivative plot of WT SOD1 in the absence of lipid). The peak positions were selected from the minima of the secondary derivatives of the FT-IR absorbance data. In the second step, selected peak positions thus determined were used for the fitting of the FT-IR raw data using Gaussian distributions analyses. Analysis of the secondary structure of WT protein (Figure 2b) showed the presence of 10% alpha helix, 38% beta sheet, and 52% turns and loops including disordered stretches. The percentage of the secondary structure determined from the FT-IR analysis was found to be consistent with the data obtained from the crystal structure (PDB 4BCY with 11% alpha helix, 40% beta sheet, and 49% turns and loops), thus validating our method (Danielsson et al., 2013). FT-IR data showed a decrease in beta sheet content (from 38% to 31%) as apo protein (Figure 2c) formed. In contrast, the behavior of H72F mutant (Figure 2—figure supplement 6a, beta sheet content of 32%) was found to be similar to the apo protein, while H121F mutant (Figure 2—figure supplement 6b, beta sheet content of 37%) remained similar to WT protein. The percentage of secondary structure elements of all protein variants are shown in Figure 2d. Zn-deficient SOD1 shows higher membrane association compared to the cu-deficient and WT proteins To obtain a preliminary understanding of the possible membrane binding sites of SOD1, we resorted to computational techniques using ‘Orientation of protein in membrane’ tool (Lomize et al., 2012), which predicted weak interaction of WT on membrane surface (Figure 2—figure supplement 7a). In contrast, the same calculation predicted higher binding affinity of apo protein with the membrane (Figure 2e). When we used ITASSER-modeled structures, the computed values of ΔGtransfer (free energy change of protein transfer from bulk to the membrane) was found to be substantially higher for the apo (−2.6 kcal mol−1) when compared to the WT protein (−1.2 kcal mol−1). When we used the crystal structure of the WT protein, ΔGtransfer calculation modeling yielded similar results for the WT protein (−0.9 kcal mol−1). To probe protein-lipid binding constants experimentally (Ka, M−1), we used fluorescence correlation spectroscopy (FCS). FCS monitors diffusional and conformational dynamics of fluorescently labeled biomolecules at single-molecule resolution (Chattopadhyay et al., 2002). For FCS experiments, we labeled the cysteine residues of all the SOD1 variants using Alexa-488-maleimide. Figure 2f shows a schematic diagram of how the labeled proteins and protein-lipids complex would behave inside the confocal volume. Using FCS we determined the correlation functions using 50 nM Alexa488Maleimide protein in the presence of increasing concentration of DPPC small unilamellar vesicles (SUVs) (Figure 2g showed the typical correlation functions of alexa labeled apo SOD1 in absence and presence of 100 nM and 500 nM DPPC SUVs). We fit the correlation functions using a two component diffusion model and the goodness of the fit was established using the randomness of the residual distribution. In this model, the fast and slow diffusing components corresponded to the free (with rH113.5 Å) and lipid bound protein (rH2170 Å) respectively (Figure 2h). With increasing DPPC SUV concentration, the percentage of slow component increased (Figure 2f), which occurred at the expense of the fast component, and a sigmoidal fit of either of these components yield the values of Ka, which showed that the binding affinities followed the trend: apo ≥H72F>H121F>WT (Figure 2i, Table 3). Since FCS experiments required the use of labeled proteins in which the presence of bulky fluorescence dye can potentially influence the results, we complemented FCS binding data by measuring the tryptophan fluorescence of the SOD1 variants with increasing concentrations of DPPC SUVs. From the gradual enhancement of tryptophan fluorescence due to lipid binding, we calculated the binding affinities of all the protein variants towards membrane which showed comparable binding constants as obtained from our FCS experiments (Figure 2—figure supplement 7b,c). We then measured the Stern–Volmer constants using acrylamide quenching experiments of Trp32 fluorescence with protein variants in the absence (Ksv) and presence of (Ksvm) membrane. The parameter Ksv/Ksvm was found maximum for the apo protein, and minimum for WT (Figure 2—figure supplement 7d, Table 2). H121F and H72F variants behaved like WT and apo protein, respectively. As observed by FT-IR, DPPC binding resulted in no or minimum change in conformation for WT and H121F proteins (Figure 2—figure supplement 7e,f), while a large decrease in beta sheet content with simultaneous rise in non-beta content, specifically alpha helical content, was observed for the apo protein and H72F mutant (Figure 2j,k Figure 2—figure supplement 7g). Table 3 Binding constants (Ka,M−1) of the protein variants with model DPPC SUVs as obtained from the FCS study for WT SOD1 and all the mutants. SystemsAsssociation constants (Ka,M−1)WT SOD1 + DPPC SUV(4.1 ± 0.1) × 106H121F + DPPC SUV(5.2 ± 0.2) × 106H72F + DPPC SUV(9.6 ± 0.4) × 107apo + DPPC SUV(9.8 ± 0.1) × 107G37R-DPPC SUV(2.2 ± 0.3) × 106I113T-DPPC SUV(8.8 ± 0.2) × 106 Lipid vesicles accelerate aggregation kinetics of apo and zn-deficient mutants Aggregation kinetics of WT, apo, and the mutant SOD1 in their TCEP reduced states were studied systematically both in the absence and in the presence of DPPC. A typical protein membrane ratio of 1:2 was maintained for all measurements involving membranes. For the initial assessment of the aggregation kinetics, the fluorescence intensity enhancement of amyloid marker Thioflavin T (ThT) was monitored. ThT is known to bind to protein aggregates with cross beta structure giving rise to a large increase in its fluorescence intensity. From the ThT fluorescence assay, we found that the WT protein does not aggregate, both in the absence or in the presence of membrane (Figure 3a,b). For the H121F variant in the absence of membrane, we found a small and slow enhancement of ThT fluorescence and the profile remained unchanged when we added the membrane (Figure 3a,c). In contrast, for apo and H72F variants, ThT assay showed large fluorescence increase and the kinetics followed typical sigmoidal patterns. The addition of membrane increased the rate of aggregation for both variants and a large decrease in the lag times. When compared between the apo and H72F variants, we found that the rate of aggregation is higher (i.e. with less lag time) for the apo protein (Figure 3a,b,c Table 4). Figure 3 with 3 supplements see all Download asset Open asset Aggregation of WT SOD1 and its mutants and membrane effects. (a) ThT fluorescence (at 484 nm) for the protein variants under reducing conditions to monitor the kinetics of aggregation. (b) Time-dependent increase in ThT fluorescence intensity of WT and apo both in the absence and in the presence of membrane (DPPC SUVs were used here as membrane). (c) Same as (b) but for H72F and H121F. Atomic force microscopy (AFM) images of the aggregates of apo SOD1 in the absence (d) and presence (e) of DPPC SUVs. These AFM images were taken at the plateau of the ThT aggregation curves. AFM images of apoagg showed linear fibrillar aggregates with an average size 1.8–2 μm. In contrast in the presence of membrane (apoaggm), we found network of small fibrils, which were connected by DPPC vesicles (as drawn in the inset of e). It may be noted that the size and height (average diameter is 70 nm and average height is 7 nm) of the connecting spherical objects are similar to (f) AFM micrograph of the control DPPC SUVs, which showed distinct membrane structures with an average size of 70–90 nm. Figure 3—source data 1 Aggregation and effect of membrane curvature and composition on the aggregation behavior of SOD1 protein variants. https://cdn.elifesciences.org/articles/61453/elife-61453-fig3-data1-v2.xlsx Download elife-61453-fig3-data1-v2.xlsx Table 4 Log-phase mid-points of different protein variants obtained from ThT assay. SystemsLog-phase mid-point (h)WT SOD1Not detectableWT SOD1 + DPPC SUVNot detectableH121FNot detectableH121F + DPPC SUVNot detectableH72F167.2H72F + DPPC SUV90.8apo SOD1112.8apo + DPPC SUV55.9 We then imaged using atomic force microscopy (AFM) the aggregates collected from the plateau regions of the aggregation kinetics (at a time point when the fluorescence of ThT was maximum [saturated] and did not change). Protein (P) aggregates will be denoted by Pagg, and Paggm to indicate if they are formed in the absence or presence of membranes respectively. For example, the aggregates of WT in the absence and presence of membranes would be denoted by WTagg, and WTaggm, respectively. AFM imaging also showed that in the absence of membrane, WT and H121F did not form aggregates, fibrillar, or otherwise (Figure 3—figure supplement 1), while large fibrillar aggregates were found to form with apo (Figure 3d) and H72F mutant (Figure 3—figure supplement 1). The average size of the fibrillar apoagg was found to be 1.8–2 μm with an average height of 20 nm. H72Fagg showed similar morphology (Figure 3—figure supplement 1). Significant morphological differences were noticed for the aggregates of apo and H72F variants, in the absence (Figure 3d,e, Figure 3—figure supplement 1) and the presence of the membrane. The apoaggm appeared to exhibit network of thin aggregates (the average size was found to be 700–800 nm with an average height of 6–8 nm)