Relationship between mutant SOD1 maturation and inclusion formation in cell models

Abstract

A common property of Cu/Zn superoxide dismutase 1 (SOD1), harboring mutations associated with amyotrophic lateral sclerosis (ALS), is a high propensity to misfold and form abnormal aggregates. The aggregation of mutant SOD1 has been demonstrated in vitro, with purified proteins, in mouse models, in human tissues, and in cultured cell models. In vitro translation studies have determined that SOD1 with ALS mutations is slower to mature, and thus perhaps vulnerable to off-pathway folding that could generate aggregates. The aggregation of mutant SOD1 in living cells can be monitored by tagging the protein with fluorescent fluorophores. In the present study, we have taken advantage of the Dendra2 fluorophore technology in which excitation can be used to switch the output color from green to red, thereby clearly creating a time-stamp that distinguishes pre-existing and newly made proteins. In cells that transiently over-express the Ala 4 to Val variant of SOD1-Dendra2, we observed that newly made mutant SOD1 was rapidly captured by pathologic intracellular inclusions. In cell models of mutant SOD1 aggregation over-expressing untagged A4V-SOD1, we observed that immature forms of the protein, lacking a Cu co-factor and a normal intramolecular disulfide, persist for extended periods. Our findings fit with a model in which immature forms of mutant A4V-SOD1, including newly-made protein, are prone to misfolding and aggregation.

Graphical abstract

Mutations in Cu/Zn superoxide dismutase 1 (SOD1) may cause amyotrophic lateral sclerosis by misfolding and aggregating. Here, we demonstrate that newly-made mutant SOD1 is rapidly incorporated into growing pathologic intracellular inclusions. Further, we link incomplete post-translational maturation (metal binding and intramolecular disulfide oxidation) of mutant SOD1 to aggregation. These findings suggest that promoting the maturation of mutant SOD1 could produce therapeutic benefits.

Mutations in Cu/Zn superoxide dismutase 1 (SOD1) cause a familial form of amyotrophic lateral sclerosis that is clinically identical to the more common sporadic form of the disease (Ravits et al. 2013; Ravits and La Spada 2009). To date, over 160 different mutations in this enzyme have been reported in ALS patients {http://alsod.iop.kcl.ac.uk/Als/Index.aspx}. Some of these 160 mutations are isolated, apparently de novo, mutations without family history, and in these cases establishing these mutations as causal is difficult. Still, there are a large number of mutations that occur in a sufficient frequency in families to establish causality. The most studied SOD1 mutations are those that show clear family history, and these mutations produce differing effects on enzymatic activity and protein stability, with some mutants possessing little or no activity while others retain significant specific activity (Borchelt et al. 1994; Hayward et al. 2002; Rodriguez et al. 2002). Our laboratory has studied more than 40 different ALS variants of SOD1 in cell over-expression models and found that all exhibit an increased propensity to misfold and form abnormal detergent insoluble aggregates (Prudencio et al. 2009). In vitro, wild-type (WT) human SOD1 can be induced to form fibrillar aggregates (Chattopadhyay et al. 2008; Chattopadhyay et al. 2015). WT SOD1 also aggregates when highly over-expressed in transgenic mice (Graffmo et al. 2013). However, in cultured cell models, WT SOD1 has been much more resistant to aggregation than mutant SOD1 (Prudencio and Borchelt 2011). A pathologic feature of SOD1-linked fALS that is commonly found is the accumulation of SOD1 immuno-reactive inclusions in surviving spinal motor neurons (Jonsson et al. 2008; Sasaki et al. 1998; Shaw et al. 1997; Shibata et al. 1996).

The role of SOD1 aggregation in the pathogenesis of disease is not entirely clear. In the G37R, G93A, H46R/H48Q, and L126Z mouse models our laboratory has shown that large sedimentable, detergent-insoluble, aggregates accumulate to the highest levels as symptoms emerge (Karch et al. 2009; Wang et al. 2005). Additionally, in transgenic mice that express the ALS variant G85R-SOD1 fused to yellow fluorescent protein (G85R-SOD1:YFP), inclusion-like structures are seen only in mice that are symptomatic and become abundant in motor neuron cell bodies as the mice reach endstage paralysis (Wang et al. 2009). More recently, studies of prion-like transmission of motor neuron disease in mice that express G85R-SOD1:YFP have shown that mutant SOD1 inclusion pathology begins to become abundant in motor neurons just as motor deficits begin to appear (Ayers et al. 2016). A similar timing of mutant SOD1 aggregation was reported in prion-like transmission studies with mice that express untagged G85R-SOD1 (Bidhendi et al. 2016). The timelines for the formation of similar aggregate structures in SOD1-ALS patients is unknown, but SOD1 antibody immunoreactive inclusions and detergent-insoluble aggregates of mutant SOD1 have been described in ALS patient spinal cords (Jonsson et al. 2008; Kerman et al. 2010; Shibata et al. 1994; Watanabe et al. 2001).

The normal structure of the 153 amino acid Cu/Zn SOD1 protein is a homodimeric enzyme that matures to native quaternary structure through post-translational modification (Fridovich 1974). These modifications include the formation of a single intra-subunit disulfide bond between residues 57 and 146 of the protein, and the binding of Zn and Cu ions (Ogihara et al. 1996; Parge et al. 1992). The order in which these modifications are acquired is not completely understood. When expressed and isolated from yeast, human SOD1 encoding a mutation at cysteine 57 (C57S), which abrogates the formation of the normal disulfide bond, purifies with 4 equivalents of Zn per dimer (Sea et al. 2015). Crystal structures of this experimental variant show that it achieves a high degree of native dimeric structure (Sea et al. 2015). Thus, it seems likely that the binding of Zn is the earliest post-translational event, and that such binding is sufficient to stabilize the structure of individual SOD1 subunits to a level that allows the formation of normal quaternary structure. Further studies of the C57S variant indicate that it is capable of binding Cu in the normal Cu-site, but for WT human SOD1 there is evidence that the disulfide must be reduced in order for Cu to be loaded into the Cu-site (Furukawa et al. 2004; Sea et al. 2015). Although the binding of Cu certainly contributes to the stability of SOD1, the binding of Zn could be viewed as an earlier and more critical event (Bruns and Kopito 2007). The acquisition of post-translational modifications and assembly into homodimers imparts tremendous stability to WT SOD1 such that it resists proteolytic digestion at concentrations up to 1 mg/ml for 30 min at 37°C (Ratovitski et al. 1999). Studies of WT SOD1 maturation in vitro have demonstrated that nascent SOD1 rapidly forms a monomeric intermediate that is resistant to low levels of proteinase (10 µg/ml) but susceptible to higher levels (1 mg/ml), with resistance to the higher levels of proteinase being acquired post-translationally (t1/2 ~19 minutes) (Bruns and Kopito 2007). The binding of Zn is required to achieve the more protease resistant conformation, and the kinetics of intra-subunit disulfide bond formation, and dimerization, follow closely the kinetics of acquiring high protease K resistance (Bruns and Kopito 2007). All of these data present a picture of SOD1 maturation that suggests that one could expect that any given molecule of nascent WT SOD1 should mature into a fully metallated dimeric enzyme in a period of ~60 minutes.

The effect of fALS mutations in SOD1 on the kinetics of maturation varies to some extent, but in general the effect is to slow the rate of maturation (Bruns and Kopito 2007). ALS-mutations in residues involved in the binding of Cu (H46R and H48Q) slow the rate of dimerization and intramolecular disulfide bond oxidation, but do not prevent these post-translational events (Bruns and Kopito 2007). ALS-mutations in the cysteine residues that produce the intramolecular disulfide bond (C57S and C146R) prevent this modification from occurring, and predictably slow maturation. Interestingly, 50% of the C146R-SOD1 can acquire high resistance to proteinase (Bruns and Kopito 2007). Collectively, these data indicate the post-translational modification of SOD1 influences the rate of maturation, but inherent attributes of the SOD1 sequence largely dictate whether the protein will ultimately fold to acquire native conformation.

The relationship between the maturation of mutant SOD1 and the appearance of misfolded aggregates of these proteins is of interest because understanding this relationship may have implications for pharmacologic intervention to prevent aggregation. Given the effects of ALS mutations on the maturation of SOD1, it is possible that the misfolded forms of the protein may represent species of SOD1 that stray from the normal maturation pathway very early in its biogenesis. However, such a scenario does not fit well with the natural history of the disease in which symptoms and SOD1 aggregates appear months (in mice) or decades (in humans) after birth. From a thermodynamic perspective, it is somewhat easier to imagine that nascent SOD1 might be most prone to fold off-pathway and eventually aggregate. However, another scenario could postulate that SOD1 proteins that have aged and acquired damage might be the source of misfolded protein. In a previous study of SOD1 half-life in vivo, Borchelt and colleagues determined that SOD1, including mutant SOD1, is axonally transported in the slow component in the motor neurons of mice (Borchelt et al. 1998). Given the long interval required for transport in the slow component (months), motor neurons would be expected to have a tremendous burden of “older” mutant SOD1. Thus, there are two equally compelling sources of mutant SOD1 in motor neurons in vivo that could produce misfolded aggregates.

The aggregation of SOD1 in living cells can be monitored by tagging the protein with fluorescent fluorophores, such as yellow fluorescent protein (Prudencio et al. 2010). A study of SOD1 kinetics of aggregation by live cell imaging reported that the G93A variant of human SOD1 converts from a diffuse distribution (soluble), to a punctate distribution (insoluble inclusion) about 120 minutes after cells are treated with a proteasome inhibitor to induce aggregation (Kim et al. 2014). Relatively large inclusions form rapidly, over 40 minutes, suggesting that the pool of pre-existing SOD1 is captured into the aggregates. Fluorescence resonance imaging by this same group suggested that the source of the misfolded SOD1 was likely monomeric. In the present study, we have taken advantage of the Dendra2 fluorophore technology in which excitation can be used to switch the output color from green to red, thereby clearly marking pre-existing protein in a live cell (Gurskaya et al. 2006). We have used this technique, in cells that transiently express the A4V variant of SOD1 and spontaneously produce intracellular inclusions. We determine that newly made SOD1 is a significant driving force in generating aggregates, but older protein can also be drawn into inclusions once the process has begun. Our findings may have implications for therapies that target SOD1 expression as a therapeutic approach. Older, long-lasting pools of mutant A4V-SOD1 may represent a persistent reservoir of protein that can sustain the production of toxic misfolded SOD1 within cells well after production of new protein has been diminished by gene-silencing therapies.

Methods

SOD1 expression constructs

SOD1 expression plasmids, using the pEF.Bos vector (Mizushima and Nagata 1990) for WT, A4V, G85R, D101G, D101N, and G37R human SOD1 cDNA have been previously described (Ayers et al. 2014; Borchelt et al. 1994; Prudencio et al. 2009). To visualize the aggregation of mutant SOD1, we created a fusion construct of WT and A4V human SOD1 cDNA with the Dendra2 fluorophore (Clontech Laboratories, Mountain View, CA, USA). This construct is similar to SOD1 fusions to yellow fluorescent protein that we have previously described (Prudencio et al. 2010; Qualls et al. 2013a; Qualls et al. 2013b; Roberts et al. 2012). The fusion construct of WT-SOD1:Dendra2 and A4V-SOD1:Dendra2 was generated by amplifying the Dendra2 cDNA using oligonucleotides that modified the 5’ sequence of Dendra2 to remove the start codon and align the coding frame to be in-frame with SOD1.The construct was produced by excising the YFP fluorophore of pEF.Bos vectors encoding WT-SOD1:YFP and A4V-SOD1:YFP (Prudencio et al. 2010) and inserting the Dendra2 cDNA amplicon, using the Infusion cloning kit (Clontech Laboratories). The final construct was sequenced in its entirety to confirm the presence of the A4V mutation, and that the Dendra2 was fused in-frame with no unintended mutations. All recombinant DNA work was approved by the University of Florida Division of Environmental Health & Safety. No live animals were used in this study.

Time Lapse Imaging and Analysis

CHO cells were plated in a 35 mm glass-bottom dish. The cells were washed with DPBS and subsequently transfected with SOD1 A4V-Dendra2 DNA (2 µg) and Lipofectamine 2000 (5 µl) (Invitrogen, Carlsbad, CA, USA), for 24 hours at 37°C with 5% CO₂. The transfection was followed by the addition of complete media (DMEM with 10% horse serum and 2 mM L-glutamine) to the cells. The cells were transferred to a live-scope imager (Nikon Ti-E Inverted Live Cell Imaging System, Tokyo, Japan) at 16 hours post-transfection. After the first time point (T0) image, the cells were flashed for 30 seconds with blue light (405nm) at 50% maximum fluorescence intensity to photo convert the green fluorophore to red. The normalized excitation before and after photoactivation switched from wavelengths of 490 to 553nm and 507 to 573nm for green and red protein, respectively. Images were taken at T0 pre and post-flash over a span of 23 hours at an interval of 15 minutes using the NIS-Elements Image Capture and Analysis Software.

Tissue Culture Transfection

For biochemical studies of SOD1 disulfide bond formation, protease sensitivity, and metal content, the SOD1 constructs (4ug) were transfected into HEK293FT cells using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA), for 24 or 48 hours, as indicated in the figure legends, at 37°C with 5% CO₂. After 3–5 hours, complete media (DMEM with 10% horse serum and 2 mM L-glutamine) was added to the cells. The cells were then harvested as described below for the various analyses.

Detergent extraction and trypsin digestion of SOD1 for disulfide bond analysis

The methodology used for determining the detergent solubility of mutant SOD1 and for determining sensitivity to proteolytic digestion have been previously described (Ayers et al. 2014). Briefly, after a 24-hour transfection, cells were scraped from the dish in DPBS and washed 3 times in DPBS. After the final wash, cells were lysed by sonication three times for 10 seconds each in 1X TEN buffer containing 1% NP-40, 200 mM iodoacetamide, and 1:100 v/v protease inhibitor cocktail (in experiments involving trypsin digestion, the protease inhibitor cocktail was omitted from all buffers). The lysate was then centrifuged at >100,000 × g for 5 min in an AirFuge to produce supernatant 1 (S1) and pellet (P1) fractions. The S1 fractions were put on ice, and the P1 fractions were washed with the same extraction buffer, sonicated three times for 10 seconds each, and centrifuged at >100,000 × g for 5 min. The supernatant was discarded and the remaining pellet (P2) was resuspended by pulsed sonication in 1X TEN buffer containing 0.5% NP-40, 100 mM iodoacetamide, and protease inhibitors. The protein concentrations of the S1 and P2 fractions were then determined by BCA assay (Pierce Biotechnology, Rockford, IL, USA).

To assess sensitivity to proteolysis, 5 µg of protein in the S1 fraction and 20 µg of protein in the P2 fraction were incubated with trypsin at final concentrations of 0, 10, and 100 µg/ml for 30 minutes at 37°C. Digestions were stopped by the addition of Laemmli sample buffer and immediately boiling the sample at 100°C.

Immunoblot analyses

The methodology for immunoblot analyses has been previously described (Ayers et al. 2014). Briefly, 5 µg of protein from the S1 fractions and 20 µg of protein from the P2 fractions were analyzed by SDS-PAGE in 18% Tris-Glycine gels (Invitrogen, Carlsbad, CA, USA). For denaturing/reducing gels, the samples were boiled for 5 minutes in Laemmli sample buffer with β-mercaptoethanol (βME). For non-reducing gels, the βME was omitted from the sample buffer. For non-reducing SDS-PAGE, an in-gel reduction was accomplished by incubating gels in transfer buffer with 2% βME for 10 minutes prior to transferring to nitrocellulose membrane. The immunoblots were probed with two different rabbit polyclonal antibodies, one designated as m/hSOD antibody (Borchelt et al. 1994) or one designated as hSOD antibody (Bruijn et al. 1998). Primary antibodies were revealed with goat anti-rabbit secondary antibody at 1:5000 (KPL, Gaithersburg, MD, USA) and chemi-luminescence reagents (Thermo Scientific Inc., Rockford, IL, USA), using a Fujifilm imaging system (FUJIFILM Life Science, Stamford, CT, USA). The aggregation propensity was assessed by comparing the ratio of immunoreactive SOD1 bands in the P2 versus S1 fractions as previously described (Prudencio et al. 2009).

SOD1 metal-binding characterization

The level of Cu and Zn bound to soluble SOD1 isolated from cultured cells, which had been transfected for 24 hours, was determined using methods previously described (Ayers et al. 2014; Lelie et al. 2011). The method involved a combination of size exclusion chromatography coupled directly to ICP-MS where Cu, Zn, manganese, and iron concentrations were then measured in real-time allowing for accurate Cu and Zn concentration from the SOD1 peak. SOD1 metallation was determined by dividing SOD1 metal concentration by the SOD1 protein concentration.

Statistical analyses

All statistical analyses were conducted using in GraphPad PRISM 5.01 Software (La Jolla, CA), with the tests used indicated in figure legends. In the live-imaging studies, aggregation propensity was assessed by statistical analysis of the ratio of red to green SOD1-Dendra2 protein in inclusions at each time point.

Results

To visualize the aggregation of A4V-SOD1 in living cells, we constructed fusion proteins of human A4V-SOD1 with the fluorophore Dendra2. We have previously demonstrated that ALS mutant SOD1 tagged with YFP (SOD1:YFP), including the A4V variant, forms cytoplasmic inclusions when expressed in cultured cells, whereas WT-SOD1:YFP does not (Prudencio et al. 2010; Prudencio and Borchelt 2011; Qualls et al. 2013a; Qualls et al. 2013b; Roberts et al. 2012). Similarly WT SOD1 tagged with green fluorescent protein does not readily form inclusions when expressed in cultured cells and is fully active (Stevens et al. 2010). In most of our prior studies of mutant SOD1 aggregation, we have used the human HEK293 cell model because it allows for the high levels of expression that are required to produce aggregation in the time frame of a typical cell culture experiment. In the present study, however, we found the HEK293 cell model to be problematic for live cell imaging due to cell rounding, which produces shifts in the focal plane and obscures visualization of the cytoplasmic inclusions. In identifying a replacement, we observed that CHO cells were superior in live cell imaging morphology. In previous work, we have demonstrated that fusions of mutant SOD1:YFP proteins readily produce inclusion structures in CHO cells, whereas WT-SOD1:YFP do not (Roberts et al. 2012). To confirm that the Dendra2 fluorophore was not inducing SOD1 aggregation, we generated a WT-SOD1 fused to Dendra2 and expressed it in CHO cells for 48 hours. No inclusions were observed in the cells expressing WT-SOD1:Dendra2, whereas abundant inclusions were produced in cells transiently transfected with the SOD1 mutant A4V fused to Dendra2 (A4V-SOD1:Dendra2) (Fig 1).

Fig. 1.

Effects of Dendra2 fusion on SOD1 inclusion formation. Representative images of CHO cells at 48 hours following transient transfection with WT-SOD1:Dendra2 or A4V-SOD1:Dendra2.

The basic paradigm used here was to transfect cells, then wait 16 hours before exposing to a blue light flash to photo-convert the emission of the fluorophore from green to red. By this method, any A4V-SOD1:Dendra2 protein made over the initial 16 hours would emit in the red spectrum and any newly-made protein would emit in the green spectrum. The flashed cells were then incubated for an additional 22 hours in a live imaging microscope. In analyzing the resulting video images, we focused our attention on two types of cells. One focus was on cells that contained diffusely distributed A4V-SOD1:Dendra2, and the other focus was on cells that had already developed inclusions by 16 hours post-transfection.

In a subset of cells that demonstrated only diffusely distributed A4V-SOD1-Dendra2 (green) at pre-flash, inclusions could be detected forming at about 12 hours post-flash (Fig. 2, Fig. S1). In these cells, we observed a fairly similar intensity of fluorescence in both the green (newly-made) and red (older protein) channels. As the inclusions grew in size over an additional 6 hours (18 hours post-flash), there seemed to be a relatively equivalent increase in both the green and red emitting forms of A4V-SOD1:Dendra2 in the inclusions (Fig. 2, Table 1, Fig. S2). We observed that all of the cells that gained red fluorescent inclusions over time also gained newly-made, green inclusions (Table 1). In cells in which inclusions were evident prior to the flash, we observed a relatively rapid capture of newly-made A4V-SOD1:Dendra2 (green) protein by the pre-existing inclusions (red) (Fig. 3, Table 1, Fig. S2). This phenomenon appeared in almost all cells that had pre-existing inclusions at 0 hours (Table 1). Collectively, these observations suggest that newly-made A4V-SOD1:Dendra2 protein is highly vulnerable to capture in growing aggregates, but much older forms of this protein can also be captured. In this model, it seems that both the newly-made and older, presumably more mature, pools of A4V-SOD1:Dendra2 are similarly vulnerable to capture into growing inclusions.

Fig. 2.

Images captured from time-lapse live imaging of CHO cells expressing A4V-SOD1:Dendra2 forming inclusions after 16 hours post-transfection. CHO cells were transiently transfected with expression plasmids for A4V-SOD1:Dendra2 and incubated for 16 hours before placement in an environmental chamber attached to a Nikon live-imaging microscope. Post-flash images are shown at time 0, 6, 12, and 18 hours.

Table 1.

Quantification of cells transiently transfected with A4V-SOD1:Dendra2

	Total red fluorescent cells at 0 hr	No. of cells with no inclusions at 0 hr and no inclusions at 18 hr	No. of cells with no inclusions at 0 hr that gained red inclusions by 18 hr / no. of those cells that also gained green inclusions 1	No. of cells with red inclusions at 0 hr / no. of those cells that also gained green inclusions by 18 hr 2
Trial 1	286	94	34/34	53/52
Trial 2	133	45	12/12	32/28
Trial 3	97	24	7/7	29/25

¹Column represents cells depicted in Figure 1A

²Column represents cells depicted in Figure 1B

Fig. 3.

Images captured from time-lapse live imaging of CHO cells expressing A4V-SOD1:Dendra2 forming inclusion before 16 hours post-transfection. CHO cells were transiently transfected with expression plasmids for A4V-SOD1:Dendra2 and incubated for 16 hours before placement in an environmental chamber attached to a Nikon live-imaging microscope. Post-flash images are shown at time 0, 6, 12, and 18 hours.

To more fully understand the dynamics of WT and mutant A4V-SOD1 maturation in cultured cell models, we examined the oxidation state, protease sensitivity, and metalation of expressed protein. For this component of the study, we turned back to the HEK293FT cell model that has been used extensively to study the aggregation of untagged mutant SOD1 (Karch and Borchelt 2010; Prudencio et al. 2009; Wang et al. 2003). Prior studies by Zetterstrom and colleagues demonstrated that SOD1 proteins lacking the normal intramolecular disulfide bond can be discriminated from proteins with the correct linkage by non-reducing, denaturing, SDS-PAGE (Zetterstrom et al. 2007). SOD1 with a normal intramolecular disulfide (oxidized – O) migrates slightly faster in these gels than the reduced protein due to differences in the shape of the molecules. As previously reported (Ayers et al. 2014), when over-expressed, WT SOD1 is slow to acquire the normal intramolecular disulfide bond, but eventually slightly more than 50% of the protein is correctly oxidized by 48 hrs post-transfection (Fig. 4; Supplemental Fig. S3). By contrast, >80% of the A4V variant protein never acquires an intramolecular disulfide bond.

Fig. 4.

Variations in the formation of intramolecular disulfide bonds by soluble WT and A4V SOD1 extracted from HEK293FT cell. Following transfection in HEK293FT cells for either 24 or 48 hours, as indicated, with WT SOD1, mutant SOD1 constructs A4V, G85R, D101G, or D101N, or left untransfected (UT), cells were detergent extracted as described in “Experimental Procedures”. 5µg of the detergent soluble fractions were analyzed by SDS-PAGE without reducing agent followed by an in-gel reduction in 2% β-mercaptoethanol. The ratio of oxidized (O) to reduced (R) SOD1 was quantified for each construct at both 24 and 48 hours [mean ratio ± S.E. (error bars) of three replicate experiments]. Analysis of variance (ANOVA) and Tukey’s test were used to determine the statistical significance for mutant SOD1 oxidation at either 24 or 48 hours when compared to WT at the same timepoint: *** P ≤ 0.001.

We next examined maturation of WT and A4V SOD1 into protease-resistant conformations. In previous work, we have demonstrated that natively folded WT SOD1 is extremely resistant to proteolytic degradation (Ratovitski et al. 1999). To examine the protease sensitivity of mutant SOD1, we have used a strategy in which soluble and insoluble forms of the protein are analyzed separately by first fractionating cell lysates by detergent extraction and centrifugation (Wang et al. 2003). At 48 hours post-transfection, both oxidized and reduced forms of WT SOD1 are detected in NP40 soluble fractions [Fig. 5A; Fig. S3; also see (Ayers et al. 2014)]. Both of these forms of WT protein are resistant to proteolytic digestion (Fig. 5A and B; Fig. S3), with the reduced form being somewhat more sensitive (Fig. 5B). The vast majority of NP-40 soluble A4V SOD1 in these cells is reduced, and this reduced form of the protein is highly sensitive to protease digestion (Fig. 5A and C; Fig. S3). At the exposure used in these immunoblots, the endogenous SOD1 in these cells cannot be seen (see Fig. S3) and thus the faint band of reactivity that migrates as oxidized protein is A4V mutant SOD1 that is largely resistant to protease attack (Fig. 5A and C). Notably, there is also a minor fraction of reduced A4V that is resistant to digestion (Fig. 5A). Overall, however, in buffers containing NP40, the A4V variant of SOD1 is much less resistant to proteolytic digestion than the WT variant.

Fig. 5.

Sensitivity of the oxidized and reduced forms of NP-40 soluble SOD1 to proteolytic digestion. HEK293FT cells were transfected with the WT and A4V SOD1 constructs for 48 hours and detergent extracted as described in “Experimental Procedures”. A, 5µg of the NP-40 detergent soluble fractions were treated with various concentrations of trypsin and analyzed by SDS-PAGE without reducing agent followed by an in-gel reduction in 2% β-mercaptoethanol. The migration of the 20kDa band of the ladder is indicated on the left of the panel. B-G, the reduced (black bars) and oxidized (white bars) band intensities for each SOD1 construct were quantified in relation to the non-trypsin treated fraction [mean ratio ± S.E. (error bars) of at least three replicate experiments]. ANOVA and Tukey’s test was used to determine the statistical significance of the trypsin treated reduced or oxidized band intensities when compared to the non-treated samples: ** P ≤ 0.01, *** P ≤ 0.001.

Other ALS variants of SOD1 exhibit similar properties when over-expressed (Fig. S3). NP40 soluble forms of the G85R-SOD1 are likely reduced (difficult to definitively demonstrate by electrophoretic migration), and are highly sensitive to proteolytic digestion. As previously described (Ayers et al. 2014), most of the NP40 soluble D101N and D101G variants of SOD1 that are detected in over-expressing cells migrate as reduced protein that is protease sensitive (Fig. S3). The G37R variant contrasts somewhat, as a greater percentage of the reduced form of this variant is resistant to protease digestion (Fig. S3). Still, as a whole, when highly over-expressed, ALS variants of SOD1 are less able to mature into protease-resistant conformations.

Mutant forms of SOD1, including the A4V variant, misfold and aggregate to become insoluble in NP40 (Wang et al. 2003). By 48 hours post-transfection, very little of WT SOD1 adopts the NP40-insoluble conformation (Wang et al. 2003)(Suppl. Fig. S4A). Similar to the soluble A4V-SOD1, the insoluble forms of this protein are also sensitive to trypsin digestion (Fig. S4). Detergent-insoluble A4V SOD1 is partially resistant to trypsin at 10 µg/ml, but completely degraded at 100 µg/ml. Interestingly, aggregated forms of the mutants examined here (A4V, G37R, G85R, D101N, D101G) show remarkably similar profiles of protease sensitivity (Fig. S4), suggesting the aggregated forms of these mutants may share structural features.

An important point to emphasize in the foregoing studies is that the trypsin digestions were carried out in buffers containing NP40. When we liberate soluble WT and A4V SOD1 by treatment of cells with saponin (Prudencio and Borchelt 2011) and digest the proteins in the buffers containing saponin, then the sensitivity of reduced A4V SOD1 to digestion is less dramatic, although still distinguishable from WT (Fig. S5). Similarly, reduced forms of the G37R variant show less sensitivity to digestion (Fig. S5). Presumably, the difference in degree of sensitivity to digestion in the two experiments is due to effects of NP40 on the accessibility of trypsin cleavage sites in the mutant protein, with the detergent acting to increase accessibility. Notably, reduced forms of other mutants, such as G85R, D101G, and D101N [also see (Ayers et al. 2014)] display high sensitivity to protease digestion in buffers containing saponin. These data provide an indication of the influence of detergents on the protease sensitivity of mutant SOD1.

To determine metal content of the proteins in these cells, we isolated the over-expressed SOD1 by HPLC fractionation and then analyzed the SOD1 containing fractions by ICP-MS (Ayers et al. 2014). The metal content of A4V-SOD1 isolated from these cells at 24 h post-transfection was similar to that of WT-SOD1 [data for untransfected cells and cells expressing WT SOD1 reproduced from (Ayers et al. 2014)]. The level of bound Cu was less than 0.5 equivalents per dimer, with both WT and A4V SOD1 containing about 2 equivalents of Zn per dimer (Fig. 6; Fig. S6). In this over-expression model, the metal content in the A4V variant was similar to that of G37R variant (Fig. S6). By contrast, the G85R, D101G, and D101N variants were all distinguished by low incorporation of both Cu and Zn [Fig. S6; see (Ayers et al. 2014)]. Comparing these data to what is observed for endogenous SOD1 in these same cells, it is clear that over-expressed WT or mutant SOD1 did not acquire Cu efficiently, with the incorporation of Zn being more variable between mutants.

Fig. 6.

Metal content of soluble SOD1 isolated from HEK293FT cells expressing WT and mutant SOD1. The copper (Cu) and zinc (Zn) metallation state of soluble human SOD1 isolated from either untransfected HEK293FT cells or cells transfected for 24 hours with the SOD1 constructs listed were measured by HPLC-ICP-MS as described in “Experimental Methods”. The amount of Cu (black bars) and Zn (white bars) per dimer was calculated by dividing the metal concentration for each construct by its protein concentration [mean ratio ± S.E. (error bars) of at least three replicate experiments]. ANOVA and Tukey’s test was used to determine the statistical significance of the Cu and Zn levels for each of the mutants in comparison to the WT levels: ** P ≤ 0.01, *** P ≤ 0.001.

Discussion

In cells over-expressing mutant SOD1, we have modeled the spontaneous aggregation of A4V mutant protein, using both visual and biochemical means of assessing aggregation. In our study of SOD1 fused to the Dendra2 fluorescent reporter protein, our primary goal was to determine whether newly synthesized A4V SOD1 was disproportionately prone to aggregation and inclusion formation. As SOD1 matures, it acquires several post-translational modifications that would stabilize structure, including the binding of Cu, Zn, and an intramolecular disulfide bond. Intuitively, it would seem far easier for mutant protein that has failed to mature to be recruited into aggregates than to have a mature protein essentially unfold. Although we find that newly made A4V SOD1 is highly prone to aggregation, we also find that older, pre-existing, mutant SOD1 can be recruited to growing aggregates relatively efficiently. Presumably, older protein, which in this study includes proteins that are 1 to 2 hours old, would be more mature than newly made protein (Bruns and Kopito 2007). Thus, on face value, our data from the SOD1-Dendra2 study would suggest that more mature A4V-SOD1 can be templated to misfold in cells with growing aggregates. Importantly, in all cells producing A4V-SOD1-Dendra2 inclusions, the newly-made protein was invariably recruited to such structures, indicating that once aggregation has initiated, newly-made mutant SOD1 may be highly vulnerable to accretion into growing inclusions. An important feature of the cell models used here to examine the spontaneous aggregation of mutant SOD1 is that the mutant proteins must be highly over expressed. Our biochemical analysis of over-expressed A4V SOD1 indicates that a large portion of the protein in these cells does not acquire two of the structure-stabilizing post-translational modifications (Cu binding and intramolecular disulfide bond formation). Over-expressed A4V SOD1 is generally less well folded, even after 48 hours post-transfection, as indicated by sensitivity to trypsin digestion in non-ionic detergents. It is important to note that when WT SOD1 is over-expressed in the same manner, it too fails to bind Cu, with a large fraction ~50% failing to form the normal disulfide bond. Yet, when expressed to equivalent levels as mutant SOD1, the WT protein does not spontaneously misfold and aggregate. Thus, a lack of post-translational maturation due to over-expression is not necessarily a defining feature that causes aggregation.

Still, our data suggest that one of the reasons that it may be possible to model mutant SOD1 aggregation in over-expressed cells is that a large portion of the over-expressed protein fails to acquire structure stabilizing modifications. All mutants we have examined here, in this paradigm, largely fail to acquire Cu and oxidize the disulfide. In buffers containing NP40, the conformation of the reduced form of expressed mutant SOD1 leaves these proteins more sensitive, to varying degrees, to protease digestion than WT SOD1. Interestingly, although soluble forms of the mutants we have examined show differing sensitivities to protease digestion, the aggregated forms of these mutants show remarkably similar profiles of protease sensitivity (see Fig. S4), suggesting the aggregated forms of these mutants may share structural features. The collective conclusion from these studies is that in cell culture models in which spontaneous aggregation of mutant SOD1 is observed, the majority of expressed protein has not properly matured. Importantly, we and others have shown that the mutant SOD1 that accumulates as detergent insoluble aggregates in SOD1-ALS mouse models appears to be derived from immature precursors, lacking the normal disulfide bond (Karch and Borchelt 2008; Karch et al. 2009; Zetterstrom et al. 2007). Studies in vitro have similarly shown that intramolecular disulfide bond formation is a key event in controlling the fibrillation of SOD1 (Chattopadhyay et al. 2015). The binding of Zn to immature SOD1 has also been shown in influence its aggregation in vitro, producing amorphous aggregates rather than fibrillary structures (Leal et al. 2015). Notably, our analysis of mutant A4V SOD1 metalation in our cell model indicated that most of the protein possessed 2 Zn per dimer (see Fig. 6). We conclude that if immature forms of mutant SOD1 (particularly disulfide-reduced forms) exist at any level under physiological conditions, then such forms of the protein would be very prone to aggregate. This conclusion implies that small molecules that improve the maturation of mutant SOD1 might reduce the burden of toxic protein and slow the progression of SOD1-linked ALS. One example of such a molecule might be CuATSM, which improves Cu delivery in the G93A-SOD1 mouse model of over-expression and extends life span (Williams et al. 2016). Our data also predict that inhibition of mutant SOD1 production in patients by gene-silencing therapies should substantially reduce the burden of misfolded protein aggregates by reducing the pool of immature SOD1 proteins that are highly prone to aggregation.

Abbreviations used

ALS

amyotrophic lateral sclerosis

CHO cells

Chinese Hamster Ovary cells

fALS

familial ALS

HEK293 cells

Human embryonic kidney #293 cells

PBS

phosphate buffered saline

SDS-PAGE

sodium dodecyl sulfate-polyacrylamide gel electrophoresis

SOD1

Cu/Zn superoxide dismutase 1

YFP

yellow fluorescent protein