SOD1^A4V aggregation alters ubiquitin homeostasis in a cell model of ALS

Amyotrophic lateral sclerosis (ALS, also known as motor neuron disease, MND) is a progressive neurodegenerative disease leading to paralysis of voluntary muscles due to the death of motor neurons in the brain and spinal cord. The prognosis of ALS is poor, usually leading to death within 2 to 5 years of first symptoms. A fraction of patients also develop clinical or subclinical frontotemporal dementia (FTD) (Turner et al., 2013). In most cases of ALS, the cause remains unknown (sporadic ALS; sALS). However, ∼10% of cases are inherited (familial ALS; fALS). A large proportion (20%) of fALS cases can be attributed to mutations in the gene encoding superoxide dismutase 1 (SOD1) (Chen et al., 2013). SOD1 was the first gene discovered to cause fALS and is also the most widely studied. There are now over 20 genes known to cause ALS (Chen et al., 2013), including a growing list of genes associated with dysregulation of ubiquitin (Ub) signalling. Genetic mutations in VCP, SQSTM1, UBQLN2 and OPTN have all been associated with ALS, and are all part of the protein degradation machinery of the cell. In addition, recently discovered mutations in TBK1 (Cirulli et al., 2015) and CCNF (Williams et al., 2016) add to this growing list of degradation machinery associated with ALS. The precise role of each of these genes in the homeostasis of Ub is unknown, but Ub sequestration into insoluble inclusions is common to all forms of ALS (Ciryam et al., 2017).

Abnormal accumulation of proteins into insoluble aggregates is a hallmark of many neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease, Huntington's disease and ALS (Yerbury et al., 2016). In the context of ALS, there is growing evidence that a correlation exists between protein aggregate load and neuronal loss in the ALS spinal cord (Giordana et al., 2010; Brettschneider et al., 2014; Ticozzi et al., 2010; Leigh et al., 1991; Strong et al., 2005). Our previous work showed a correlation between in vitro aggregation propensity and rate of disease progression (McAlary et al., 2016), suggesting that protein aggregates are intimately linked with motor neuron cell death. Recent work also indicates that protein misfolding and aggregation may be responsible for disease progression through a prion-like propagation throughout the nervous system (Strong et al., 2005; Zeineddine et al., 2015; Münch et al., 2011; Sundaramoorthy et al., 2013; Grad et al., 2014). It is unlikely that misfolding alone is responsible for the disease, and post-translational modifications also seem to play an important role (McAlary et al., 2013). One crucial post-translational modification is ubiquitylation, which is necessary for protein degradation. Degradation defects that lead to inclusion formation are associated with a tendency for cells to be dysfunctional and undergo apoptosis (Atkin et al., 2014; Tsvetkov et al., 2013; Weisberg et al., 2012).

Inclusions associated with neurodegeneration consist of a variety of proteins including proteins specific to the disease [e.g. Aβ and tau in Alzheimer's disease (Chiti and Dobson, 2006)], proteins associated with cellular quality control machinery [e.g. molecular chaperones (Sherman and Goldberg, 2001; Yerbury and Kumita, 2010) and the proteasome (Huang and Figueiredo-Pereira, 2010)] and other unrelated aggregation-prone proteins (Ciryam et al., 2013, 2015). Based on analysis of human tissue, it has been shown that a large number of proteins are supersaturated in wild-type and ALS-associated mutant cells, with cellular concentrations under wild-type conditions that exceed their predicted solubility (Ciryam et al., 2013, 2015). These supersaturated proteins are associated with the biochemical pathways underpinning a variety of neurodegenerative diseases. Most recently, we have shown that proteins co-aggregating with SOD1, TDP-43 (also known as TARDBP) and FUS inclusions are supersaturated (Ciryam et al., 2017), consistent with a collapse of motor neuron protein homeostasis in ALS. Others have found that the proteins that co-aggregate with c9orf72 dipeptide repeats in cell models are also supersaturated (Boeynaems et al., 2017). The composition of inclusions found in ALS varies considerably depending on whether the disease is sporadic or familial, and the genetics of the familial forms.

Ub is a pervasive feature of inclusions in ALS, regardless of underlying genetic aetiology. Ub is a versatile signalling molecule responsible for controlling an array of cellular pathways including transcription, translation, vehicle transport and apoptosis (Hershko and Ciechanover, 1998). Ub labels substrate proteins via a highly ordered multi-step enzymatic cascade with specific differences in the length and topology of poly-ubiquitin chains determining a range of signalling outcomes, including proteolytic degradation via the proteasome (Ciechanover and Brundin, 2003; Pickart, 2001). Inside cells, Ub exists in a dynamic equilibrium between free Ub and Ub conjugates, and its conjugation to proteins is controlled by the opposing actions of Ub ligases and deubiquitylating enzymes (DUBs) (Dantuma et al., 2006; Groothuis et al., 2006). Recently, it has been proposed that the sequestration of Ub into insoluble aggregates may deplete the free Ub pool required by many essential cellular processes (Groothuis et al., 2006).

Although mounting genetic and functional evidence suggests an important role for the UPS in the development of ALS pathology, the distribution and availability of Ub in ALS models has not yet been described. In the work reported here, we sought to characterise the Ub landscape in a cell-based SOD1 model of ALS (Fig. S1). By following the distribution of fluorescently labelled Ub in live cells expressing mutant SOD1 (SOD1^A4V), we show that Ub homeostasis is disrupted in cells containing aggregates of SOD1^A4V. The aggregation of SOD1^A4V leads to an accumulation of the proteasome reporter tdTomato^CL1, indicative of UPS dysfunction. This dysfunction was further supported by the redistribution of the Ub pool and decrease in free Ub levels observed in cells with SOD1^A4V aggregates. Moreover, ubiquitome analysis confirmed that misfolded SOD1 was associated with Ub redistribution and subsequent alterations to mitochondrial morphology. This report highlights that disruption to Ub homeostasis is associated with aggregation of misfolded proteins and may play an important role in the pathogenesis of ALS.

We previously showed that all cellular SOD1 aggregates contained Ub at the time points tested (Farrawell et al., 2015). Here, we used real-time imaging using a Ub fusion protein to follow inclusion formation in a single cell from its genesis until it had taken up a large proportion of the cytoplasm. Previous work has demonstrated that GFP–Ub fusions behave identically to endogenous Ub and provide a robust fluorescent indicator of Ub distribution (Dantuma et al., 2006). We created a mCherry–Ub fusion protein that behaved in a similar manner to the GFP fusion protein (Fig. S2). Ub was observed in foci at the earliest detectable stages of SOD1^A4V aggregation (Fig. 1A, see insert). Ub was continuously added to inclusions throughout their formation, as was SOD1^A4V (Fig. 1A; Fig. S3). Approximately 51% of cellular ubiquitin was detected in SOD1^A4V aggregates (Fig. 1B).

To ensure that the Ub accumulation in SOD1^A4V inclusions was not an artefact of Ub overexpression, we next probed for endogenous Ub using antibodies. SOD1^A4V expression causes inclusions in ∼15% of NSC-34 cells (Fig. S4), and in those cells that contain inclusions we observed a high degree of overlap between SOD1^A4V aggregates and Ub when using both antibodies specific to K48 and K63 chains (Fig. 1C). We also found that SOD1^G93A aggregates contained Ub K48-linked chains in SOD1^G93A mouse spinal motor neurons (Fig. 1D).

We previously showed that a UPS reporter containing a CL1 degron peptide that is subject to rapid degradation accumulates to significantly higher amounts in ALS patient fibroblasts compared to controls, suggesting an overwhelmed UPS (Yang et al., 2015). Work from others suggests that SOD1 aggregates are toxic because they interfere with the quality control function of the juxtanuclear quality control compartment (JUNQ) (Weisberg et al., 2012). To examine the relationship between SOD1 aggregation and UPS dysfunction, we used transiently transfected NSC-34 cells that contained SOD1^A4V–GFP and tdTomato^CL1. We initially used flow cytometry to examine the effect of proteasome inhibition on reporter accumulation. While there was a significant difference in reporter signal between cells expressing wild-type SOD1 (SOD1^WT) and those expressing SOD1^A4V, both cell lines showed a similar dose-dependent increase in reporter signal with MG132 treatment (Fig. 2A). This suggests that, while SOD1^A4V expression causes significant UPS disruption, there is no specific vulnerability to proteasome inhibition of cells expressing mutant SOD1^A4V compared to SOD1^WT-expressing cells.

However, this analysis examined the entire population of SOD1^WT- or SOD1^A4V-expressing cells and previous work using a similar CL1-containing fluorescent reporter suggests that the aggregation of Huntingtin and CFTR cause UPS dysfunction (Bence et al., 2001). We hypothesised that inclusions formed by mutant SOD1 might influence UPS activity in a similar fashion, a phenomenon that may not be observed when analysing the entire population of cells, given that only 15% of NSC-34 cells expressing SOD1^A4V produce inclusions (Fig. S4). In order to identify cells with SOD1^A4V inclusions in our flow cytometry data, we tested the relationship between aggregation and total cellular fluorescence (Fig. 2B). Cells were transiently transfected with SOD1^A4V–GFP and imaged 48 h post transfection using confocal microscopy. The total cellular fluorescence for individual cells was plotted for both cells with and without inclusions. The fluorescence of both populations was normally distributed (Fig. 2B). The mean fluorescence of cells containing aggregates was significantly greater than that of cells without aggregates (Fig. 3C), presumably due to continued accumulation of SOD1 into inclusions. However, there was significant overlap between the two populations, so we could not entirely distinguish populations. We decided to analyse the most highly fluorescent cells (top 10%) across the entire population. In this subset of cells, ∼80% contained SOD1^A4V inclusions (Fig. 2D). We reasoned that analysing the top 10% fluorescent cells would sufficiently enrich for cells containing inclusions. Using flow cytometry, we observed a 4-fold increase in reporter fluorescence (P<0.001) – indicating UPS dysfunction – in the top 10% of SOD1^A4V–GFP-expressing cells compared to control cells (expressing SOD1^WT) in the same fluorescence range. That is, UPS dysfunction in SOD1^A4V cells is largely restricted to cells containing aggregates. By comparison, only a small increase (1.2-fold) in reporter fluorescence was observed in the entire population of SOD1^A4V–GFP-expressing cells compared with the entire SOD1^WT–GFP-expressing population (P<0.01) (Fig. 2E). This suggests that SOD1 aggregation is associated with compromised UPS function.

Ub exists in a dynamic equilibrium in the cell, partitioning into four major pools: (1) immobile in the nucleus, (2) immobile in the cytoplasm, (3) soluble polyUb chains, and (4) a small fraction as free monomeric Ub (Dantuma et al., 2006). Mobile Ub is a combination of free monomeric Ub, free Ub chains and Ub attached to diffusible proteins (Fig. S1). Immobile Ub is primarily bound to histones in the nucleus, and bound to organelles and cytoskeleton in the cytoplasm (Fig. S1). To determine whether the observed UPS dysfunction in cells with mutant SOD1 aggregates was associated with altered Ub homeostasis, we examined the distribution of Ub into different cellular pools following expression of wild-type or mutant SOD1^A4V by fluorescence recovery after photobleaching (FRAP) (Fig. 3A–D). The diffusion rate back into the bleached area provides information on the mobility of the Ub species present in solution and the amount of immobile fluorophore in the compartment of interest (i.e. bound to membrane vesicles or cytoskeleton in the cytosol, and bound to histones in the nucleus). We selected regions of interest (ROIs) in both the nucleus and cytoplasm (Fig. 3B) in cells co-expressing mCherry–Ub and GFP, SOD1^WT–GFP or SOD1^A4V–GFP. In the case of SOD1^A4V–GFP- expressing cells, we performed the analysis on two subpopulations, cells with and without aggregates. As previously reported (Dantuma et al., 2006), we find that Ub FRAP (indicating Ub mobility) is higher in the cytoplasm than in the nucleus, likely due to a large proportion of nuclear Ub being attached to histones. Patterns of recovery appeared to be similar in all treatments, with the exception of a lower cytoplasmic recovery in the cells containing SOD1^A4V–GFP aggregates (Fig. 3A). After calculating the mean half-life (T_1/2) of recovery in each case, we find no significant differences between any of the cell populations (Fig. 3C). This suggests that, within the mobile population of Ub, the presence of SOD1^A4V–GFP aggregates does not significantly alter the kinetics of Ub diffusion. Furthermore, the amount of mobile Ub available to the cell did not appear to be altered by the presence of aggregates, as no significant differences in the mobility of Ub were observed between cells containing SOD1^A4V aggregates and the controls (Fig. 3D).

To test whether cells containing SOD1^A4V–GFP aggregates may diminish free monomeric Ub, we next examined the relative amounts of monomeric Ub in cells expressing GFP, SOD1^WT–GFP or SOD1^A4V–GFP for 48 h, by western blotting (Fig. 4A). We did not observe any significant differences using this method, likely due to the fact that this analysis represents a total measure of all cells in the culture, including non-transfected cells (Fig. S5). We therefore turned to fluorescence recovery after nucleus photobleaching (FRANP) analysis to monitor free Ub in single cells. Monomeric mCherry–Ub will diffuse through the nuclear pore (<60 kDa), while mCherry–Ub incorporated into Ub chains will not (Dantuma et al., 2006). Hence, relative free monomeric Ub can be measured by monitoring diffusion through the nuclear pore following fluorescence bleaching solely in the nucleus or cytosol (Dantuma et al., 2006). We transfected cells with GFP, SOD1^WT–GFP or SOD1^A4V–GFP (Fig. 4B) for 48 h, then bleached the entire nucleus and quantified the amount of relative monomeric Ub diffusing through the nuclear pore back into the nucleus (Fig. 4C). We observed a significant drop in free monomeric Ub in cells containing SOD1^A4V–GFP aggregates, but no difference in the level of monomeric Ub between cells expressing SOD1^WT and SOD1^A4V in the absence of inclusions (Fig. 4D).

Free Ub exists in complex equilibrium with multiple conjugated forms, and ALS mutations may induce redistribution of Ub through altered activity in various cellular pathways. To investigate changes in the Ub-modified proteome (the ‘ubiquitome’) of cells expressing SOD1^WT compared with SOD1^A4V–GFP, we performed proteomics following enrichment of cell lysates for ubiquitylated proteins (Fig. 5A). We identified 316 ubiquitylated proteins common to cells expressing either SOD1^WT or SOD1^A4V; 55 proteins were uniquely present in the ubiquitome of cells expressing SOD1^A4V and 11 unique proteins were identified in the ubiquitome of cells expressing SOD1^WT (Fig. 5B; Table 1). Network analysis of known protein–protein interactions and ontology in the set of ubiquitylated proteins unique to cells expressing mutant SOD1^A4V using the String database (Szklarczyk et al., 2015) identified a number of enriched pathways, including metabolic pathways, the ubiquitin-proteasome system (UPS), and ribosome and mRNA processing and transport (Fig. 5C). Expression of a chronic misfolded protein has previously been observed to increase proteome misfolding and we have shown that ALS aggregates are composed of supersaturated proteins – that is, proteins which have expression levels higher than one might predict given their solubility (Ciryam et al., 2017).

We compared the supersaturation scores for both the unfolded and native states of ubiquitylated proteins found uniquely in SOD1^A4V–GFP-expressing cells to those of the whole proteome. The median supersaturation score of these 55 ubiquitylated proteins is 10× higher than that for the whole proteome in the unfolded state (σ_u) and 25× higher than that for the whole proteome in the native state (σ_f) (Fig. 5D). These data are consistent with proteome instability and resulting redeployment of Ub, driving altered Ub distribution and subsequent impairment of Ub homeostasis upon expression of a chronically misfolded protein.

Previous work has shown that mutations in SOD1 lead to mitochondrial dysfunction (Mattiazzi et al., 2002; Vande Velde et al., 2011; Song et al., 2013; Joshi et al., 2018). Our ubiquitome analysis (above) identified enrichment of mitochondrial/metabolic proteins in the ubiquitome of SOD1^A4V-expressing cells (Table 1), suggesting potential mitochondrial defects in these cells. Hence, we labelled mitochondria with Mitotracker and examined morphology using confocal microscopy (Fig. 6A–C). We found that SOD1^A4V-expressing cells contained a significantly higher number of mitochondria per cell compared to controls (Fig. 6B). In addition, these mitochondria are significantly more circular than those in control cells (Fig. 6C). We quantified Mitotracker fluorescence in cells by using flow cytometry, as above focusing on the top 10% most fluorescent cells (based on SOD1–GFP expression) to enrich for cells with aggregates (i.e. when expressing SOD1^A4V). This subset of SOD1^A4V-expressing cells has higher Mitotracker fluorescence compared to SOD1^WT-expressing cells exhibiting the same level of fluorescence – consistent with accumulation of mitochondria seen upon SOD1^A4V expression (Fig. 6D). In the same subset of aggregate-containing cells, we observed increased fluorescence of the functional reporter CMXRos, suggesting alterations in mitochondrial membrane potential (Fig. 6E).

A unifying feature of neurodegenerative diseases such as ALS is the presence of Ub within insoluble protein aggregates (Leigh et al., 1991; Lowe et al., 1988; Mori et al., 1987). Beyond labelling substrates for degradation via the proteasome, Ub is an important regulator of cellular processes such as transcription, translation, endocytosis and DNA repair. The sequestration of Ub into inclusions may therefore reduce the availability of free Ub essential for these processes, compromising cellular function and survival. To gain insight into the regulation of Ub homeostasis in ALS, we followed the dynamic distribution of Ub at a single-cell level in a well-established SOD1 cell model of ALS. Our results confirm that the expression of mutant SOD1^A4V leads to UPS dysfunction and corresponding disruption of Ub homeostasis, suggesting these processes plays a key role in the development of ALS pathology.

In cell models, Ub regulates the concentrations of TDP-43 and SOD1 (Scotter et al., 2014; Miyazaki et al., 2004), and proteasome inhibition can trigger their abnormal accumulation. While our previous work has established that SOD1 aggregates contain Ub (Farrawell et al., 2015), results here confirm that SOD1 co-aggregates with Ub, and that Ub is present at the earliest stage of aggregation. Furthermore, SOD1 aggregates were found to contain both K48- and K63-linked polyubiquitin chains, which signal degradation via the proteasome and autophagy pathways, respectively (Kwon and Ciechanover, 2017). Interestingly, both K48- and K63-linked ubiquitylation has been associated with the formation of inclusions (Tan et al., 2008), with previous studies showing that K63 polyubiquitylation directs misfolded SOD1 to the ubiquitin-aggresome route when the UPS is inhibited (Wang et al., 2012). These data are consistent with the need for tightly regulated Ub homeostasis in neurons, and suggest that any perturbation in this homeostasis could cause Ub depletion and subsequent toxicity. It is likely that protein aggregation causes Ub depletion, which could then drive further aggregation in a positive-feedback loop (Fig. 7).

In the nervous system, the UPS contributes to the regulation of many aspects of synaptic function, such as neuronal growth and development, neuronal excitability, neurotransmission, long-term potentiation (LTP) and synapse formation and elimination (Mabb and Ehlers, 2010; Kawabe and Brose, 2011). UPS dysfunction is therefore central to neuronal health and neurodegenerative disease (Yerbury et al., 2016). Impairment of the UPS has been implicated strongly in the pathogenesis of ALS (Scotter et al., 2014; Tashiro et al., 2012; Cheroni et al., 2009). Here, we reveal that the aggregation of SOD1 compromises UPS function, decreasing cellular capacity to degrade the proteasome reporter tdTomato^CL1. Our findings are consistent with the previous demonstration of UPS dysfunction in ALS, with accumulation of the fluorescent UPS reporter Ub^G76V–GFP observed in the spinal cord and cranial motor neurons of SOD1^G93A mice (Cheroni et al., 2009). Moreover, a motor neuron-specific knockdown of the proteasome subunit Rpt3 in the absence of ALS genetic background results in an ALS phenotype in mice – including locomotor dysfunction, progressive motor neuron loss and mislocalisation of ALS markers TDP-43, FUS, ubiquilin 2 and optineurin (Tashiro et al., 2012). These data support a model where a reduction in UPS capacity is sufficient to drive ALS pathology in mice. Not even overexpression of human mutant TDP-43 gives such an accurate reproduction of a human ALS-like phenotype in mice. In ALS patient spinal motor neurons, 85% (38/40) of cytoplasmic TDP-43 foci or inclusions were positive for Ub. In fact, almost all cellular Ub in these neurons was sequestered within large skein-like inclusions (Farrawell et al., 2015). Interestingly, Ub has been found to accumulate in inclusions without the aggregation of TDP-43 in sALS (Giordana et al., 2010), suggesting that aggregation of proteins, such as TDP-43, FUS and SOD1 may not be necessary for Ub depletion-induced toxicity. In human skin fibroblasts, it has been shown that the UPS reporter GFP^CL1 accumulates significantly more in ALS patient fibroblasts compared to controls, suggesting a compromised UPS (Yang et al., 2015). This raises the question of whether there are alternative means to control protein concentration when protein aggregates disrupt degradation mechanisms. It was recently proposed that widespread transcriptional repression may serve this purpose in Alzheimer's disease (Ciryam et al., 2016).

The work presented here suggests that the aggregation of misfolded SOD1 alters Ub homeostasis and subsequently depletes the free Ub pool in cells. Cellular Ub exists in a complex equilibrium between free and conjugated Ub (Dantuma et al., 2006). Neurons are vulnerable to a deficiency in free Ub, which, if prolonged, can lead to cell death (Tan et al., 2000, 2001). Many factors can influence Ub homeostasis. For example, proteasome inhibition depletes free Ub to as low as 5% of basal levels in less than 2 h (Mimnaugh et al., 1997; Patnaik et al., 2000). Inhibition of translation also depletes free Ub through reduced production, while toxicity can be rescued by overexpression of Ub (Hanna et al., 2003). Accumulation of ubiquitylated proteins in inclusions is an important potential mechanism for depletion of free Ub, and, in this context, free Ub levels can be partially restored by overexpression or removal of Ub from the aggregated protein through ubiquitin-specific proteases. It was recently shown that while the ataxia-associated mutation in Usp14 causes a reduction in free Ub and neuromuscular junction dysfunction in mice, overexpression of Ub restored free Ub levels in motor neurons and improved the neuromuscular junction (NMJ) structure (Vaden et al., 2015). Additionally, free Ub could be increased in cells containing huntingtin aggregates by overexpression of the de-ubiquitylation enzyme USP14 (Hyrskyluoto et al., 2014). Not only did this protect cells from aggregate-induced toxicity, it also reduced ER stress, which is thought to precede inclusion formation in ALS models (Atkin et al., 2014). Furthermore, perturbations in ubiquitin homeostasis caused by dysregulation of ubiquitin-like modifier activating enzyme 1 (UBA1) induces neuromuscular pathology in animal models of spinal muscular atrophy (Wishart et al., 2014; Powis et al., 2016). Systemic restoration of UBA1 has been shown to rescue this pathology (Powis et al., 2016). These data further support the concept that modulation of cellular Ub pools is an important factor in the pathogenesis of neurodegenerative disease.

The pathways responsible for modulating Ub homeostasis in ALS are not yet well understood. Our ubiquitomics analysis of cells expressing SOD1 reveals that the unique ubiquitome of cells expressing SOD1^A4V is enriched for proteins prone to aggregation (i.e. that are supersaturated). This is consistent with findings that proteins associated with neurodegeneration are supersaturated (Ciryam et al., 2013, 2015, 2017). Furthermore, a substantial proportion of the ubiquitylated proteins identified are from pathways known to be dysfunctional in ALS, including RNA processing, and metabolic and mitochondrial pathways. This is not surprising given that both RNA-binding proteins and oxidative phosphorylation and metabolic pathways are also prone to aggregation in neurodegeneration (Ciryam et al., 2015). These results suggest that mutations in SOD1 may lead to mitochondrial dysfunction by disrupting Ub homeostasis. Interestingly, we find that cells expressing mutant SOD1^A4V had altered mitochondrial morphology and function. Disturbances to mitochondrial morphology and function have been previously reported in human motor neurons carrying the SOD1^A4V mutation (Kiskinis et al., 2014) and NSC-34 cells expressing mutant SOD1^G93A (Joshi et al., 2018). Recent studies have revealed that mitochondrial impairment occurs soon after proteasome inhibition (Maharjan et al., 2014). In fact, continued proteasome dysfunction in mouse brain cortical neurons inhibited the degradation of ubiquitylated mitochondrial proteins and led to the accumulation of dysfunctional mitochondria (Ugun-Klusek et al., 2017). Misfolded proteins such as mutant SOD1 have also been shown to interact with mitochondrial proteins and translocate into the mitochondrial intermembrane space and mitochondrial matrix, where they accumulate and induce mitochondrial dysfunction (Vijayvergiya et al., 2005; Jaarsma et al., 2001; Ruan et al., 2017; Igoudjil et al., 2011; Fischer et al., 2011). However, this cytotoxicity can be attenuated through the ubiquitylation of misfolded but non-aggregated SOD1, which promotes its degradation via the UPS (Yonashiro et al., 2009). Collectively, these studies demonstrate an integral functional relationship between impairment of the UPS and mitochondrial dysfunction, and that modulation of cellular Ub pools may rescue mitochondrial dysfunction caused by the accumulation of SOD1.

In conclusion, we observe that the aggregation of mutant SOD1^A4V leads to altered UPS activity and redistribution of Ub, causing disrupted Ub homeostasis. Given that Ub controls many essential cellular pathways that are also dysfunctional in ALS – including transcription, translation, vesicle transport, mitochondrial function and apoptosis – these findings suggest that Ub homeostasis is a central feature of ALS pathogenesis. Further understanding of the contribution of Ub homeostasis to ALS pathology may be imperative to understanding the molecular pathways underpinning neurodegenerative disease more broadly.

pEGFP-N1 vectors containing human SOD1^WT and SOD1^A4V were generated as described previously (Turner et al., 2005). The tGFP construct (pCMV6-AC-GFP) was obtained from OriGene (USA). GFP–Ub and mRFP–Ub (Addgene plasmids 11928 and 11935, deposited by Nico Dantuma; Dantuma et al., 2006; Bergink et al., 2006) were acquired from Addgene. The mCherry–Ub construct was created by replacing the mRFP sequence in mRFP–Ub with mCherry fluorescent protein. The tdTomato^CL1 construct was obtained by cloning the CL1 sequence (ACKNWFSSLSHFVIHL) into pcDNA3.1(+)tdTomato.

Neuroblastoma×spinal cord hybrid NSC-34 cells (Cashman et al., 1992) were maintained in Dulbecco's Modified Eagle's Medium/Ham's Nutrient Mixture F12 (DMEM/F12) supplemented with 10% fetal bovine serum (FBS, Bovogen Biologicals, Australia). Cells were maintained at 37°C in a humidified incubator with 5% atmospheric CO₂. For confocal microscopy, cells were grown on 13 mm round coverslips in 24-well plates or on eight-well µ-slides (Ibidi, Germany). Cells were grown in six-well plates for cell lysate and cell sorting experiments. Cells were transfected using Lipofectamine 3000 (Invitrogen, USA) according to manufacturer's instructions with 0.5 μg DNA per well for a 24-well plate, 0.2 μg DNA per well for eight-well µ-slides and 2.5 μg DNA per well for six-well plates. For co-transfections the amount of DNA was divided equally between constructs. Animal studies were performed with the approval (AE11/29 and AE12/09) of the Animal Ethics Committee of the University of Wollongong (Wollongong, Australia).

To quantify UPS activity, NSC-34 cells were co-transfected with the fluorescent proteasome reporter tdTomato^CL1 and treated overnight (∼18 h) with 0–30 µM of the proteasome inhibitor MG132. The level of tdTomato^CL1 fluorescence was determined via flow cytometry on a Becton Dickinson flow cytometer 48 h post transfection as in Yang et al. (2015).

Imaging of NSC-34 cells co-transfected with SOD1^A4V–GFP and mRFP–Ub was performed 24 h post transfection on a Leica TCS SP5 confocal microscope. Cells were imaged every 15 min over 17 h in a CO₂ chamber maintained at 37°C with 5% CO₂. For fluorescence recovery after photobleaching (FRAP) and fluorescence recovery after nucleus photobleaching (FRANP) experiments, analysis was performed on transfected NSC-34 cells at 48 h post transfection using the LAS AF FRAP Application Wizard on the Leica TCS SP5 confocal microscope. Images were acquired using the 63× objective with two line averages and a scan speed of 700 Hz. Five pre-bleach images were acquired over 7.5 s with the 561 nm laser set at 20% power. The region of interest (ROI) was then bleached using the ‘zoom in ROI’ method over five frames of 1.5 s at 100% laser power. For FRANP analysis, the entire nucleus was bleached. Fluorescence recovery or loss was monitored for 120 s with the laser power set back at 20%.

To quantify SOD1 inclusion formation, images were acquired as described previously (McAlary et al., 2016). Briefly, z-stack images of NSC-34 cells transfected with SOD1–GFP constructs were acquired at 48 h post transfection using the 63× objective, 512×512 pixels, one line and frame average, and a scan speed of 400 Hz. z-stacks were processed into a single image using LAS-AF Lite software (Leica) and at least 100 transfected cells were analysed for the presence of inclusions.

GFP fluorescence in NSC-34 cells expressing soluble and insoluble (aggregated) SOD1^A4V–GFP was quantified from confocal images taken 48 h post transfection using ImageJ 1.48v (Schneider et al., 2012). A minimum of 100 transfected cells were analysed per experiment. Frequency distribution analysis was subsequently performed in GraphPad Prism version 5.00 for Windows (GraphPad software, USA).

The viability of cells expressing SOD1 was monitored over 68 h in an Incucyte automated fluorescent microscope (Essen BioScience, USA) as described in McAlary et al. (2016). Cells were dissociated at 24 h post transfection and re-plated in 96-well plates at a confluency of 20% in Phenol-Red-free DMEM/F12 containing 10% FBS. Images were acquired every 2 h and analysed using a processing definition trained to select GFP-positive cells. The number of GFP-positive cells was normalised to time zero before SOD1^A4V numbers were normalised to SOD1^WT values.

NSC-34 cells grown on coverslips were transfected with GFP-tagged SOD1^A4V and fixed for 20 min at room temperature (RT) with 4% paraformaldehyde (PFA) (Merck Millipore, USA) in phosphate-buffered saline (PBS) 48 h post transfection. Cells were permeabilised in 1% Triton X-100 (TX-100) in PBS for 30 min on ice before blocking for 1 h at RT with 5% FBS, 1% bovine serum albumin (BSA) 0.3% TX-100 in PBS. Cells were incubated with rabbit primary antibodies against K48 or K63 Ub chain linkages (05-1307/05-1308, Merck Millipore; 1:500 dilution) overnight at 4°C followed by Alexa Fluor 647-conjugated anti-rabbit-IgG secondary antibody (ab150079, Abcam, UK; 1:500 dilution) for 5 h at RT. All antibodies were diluted in 1% BSA, 0.1% TX-100 in PBS and cells were washed with PBS between each incubation step.

Spinal cord sections from the SOD1^G93A mouse were also stained for SOD1 and K48 or K63 Ub chain linkages. A pap pen (Daido Sangyo, Tokyo, Japan) was used to separate tissue sections mounted onto the same slide. Sections were then fixed with 4% PFA for 15 min at RT before permeabilisation at RT for 10 min with 0.1% TX-100 in PBS containing 2% (v/v) normal horse serum (NHS). Sections were then blocked for 20 min at RT with 20% NHS and 2% BSA in PBS followed by staining overnight at 4°C in a humidified chamber with sheep anti-SOD1 (ab8866, Abcam; 1:200 dilution), mouse neuron-specific βIII tubulin antibody (ab78078, Abcam; 1:200 dilution) and rabbit anti-ubiquitin K48 or K63 (Merck Millipore; 1:200 dilution). The following day, sections were incubated with 4 µg/ml Alexa Fluor-conjugated secondary antibodies reactive to sheep, mouse and rabbit IgG (A21100, Invitrogen; ab150114, Abcam; A181448, Invitrogen) for 1 h at RT. Following staining, coverslips were mounted onto slides using ProLong Diamond Antifade Mountant (Molecular Probes). All imaging experiments were performed on a Leica TCS SP5 confocal microscope.

NSC-34 cells grown in six-well plates and transfected with SOD1–GFP or mRFP–Ub were harvested 48 h post transfection with trypsin-EDTA (Gibco). Cells were washed with PBS before being resuspended in RIPA buffer [50 mM Tris-HCl pH 7.4, 1% (w/v) sodium deoxycholate, 150 mM NaCl, 1 mM EDTA, 1% TX-100, 0.1% SDS, 10 mM NEM, 1 mM sodium orthovanadate, Halt^TM Protease Inhibitor Cocktail (Thermo Scientific)]. Protein concentration was determined with a BCA assay.

Cell lysates with a total protein concentration of 100 µg were reduced with β-mercaptoethanol and heated for 10 min at 70°C before being analysed on a Mini-PROTEAN TGX Stain-Free Gel (Bio-Rad) for 2 h at 100 V. Proteins were transferred onto a nitrocellulose membrane (Pall Corporation) using the standard protocol on the Bio-Rad Trans-Blot Turbo Transfer System. The membrane was blocked in 5% skim milk powder in Tris-buffered saline with 0.2% (v/v) Tween-20 (TBST) for 1 h at RT before probing for Ub with rabbit anti-Ub antibody (ab137025, Abcam; 1:1500 dilution) overnight at 4°C. The following day the membrane was washed three times with TBST over 30 min before incubating with horseradish peroxidise (HRP)-conjugated anti-rabbit-IgG antibody (1706515, Bio-Rad; 1:2000 dilution) for 1 h at RT. The membrane was visualised with chemiluminescent substrate (Thermo Scientific) on the Amersham Imager 6600RGB.

NSC-34 cells transfected with SOD1^WT-GFP or SOD1^A4V–GFP were harvested 24 h post transfection with trypsin-EDTA (Gibco) and washed with PBS before being resuspended in Phenol Red-free medium containing 10% FBS and filtered through a 35 µm nylon mesh strainer (Greiner Bio-One, Austria). Enrichment of GFP-positive cells was performed on a FACS Aria III (BD Biosciences, USA) and a BD FACs Aria II (BD Biosciences). Cells were sorted based on GFP fluorescence (excitation 488 nm, emission 530/30 nm) and collected into 5 ml FACS tubes containing 1 ml FBS. Following enrichment, cells were lysed for ubiquitomics analysis.

Proteomics to identify the ubiquitin-modified proteome (the ‘ubiquitome’) in NSC-34 cells was performed as described in Nagarajan et al. (2017). Cells were lysed in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.5% NP-40, 1 mM DTT, 1× EDTA-free protease inhibitor cocktail, 10 mM N-ethylmaleimide, 1 mM sodium orthovanadate). Protein concentration was determined using a BCA assay, and 300 µg of total protein was used per sample to immunopurify mono- and poly-ubiquitylated proteins using 30 µl of VIVAbind Ub affinity matrix (Ubiquitin Kit, VIVA Bioscience). Samples were incubated for 2 h at 4°C with end-to-end mixing then matrix was collected by centrifugation (1 min; 4°C; 1000 g) and washed six times in 500 µl of detergent-free wash buffer [50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM DTT, 1x EDTA-free protease inhibitor cocktail, 10 mM N-ethylmaleimide, 1 mM sodium orthovanadate]. After washing, beads were digested for 30 min at 27°C, then reduced with 1 mM DTT and left to digest overnight at RT with sequencing-grade trypsin (5 µg/ml, Promega), as described previously (Turriziani et al., 2014). Samples were alkylated with 5 mg/ml iodoacetamide, and protease digestion terminated with trifluoroacetic acid. Trypsinised eluents were collected after a brief centrifugation then purified and desalted using self-packed tips with six layers of C18 Empore disks (Pacific Laboratory Products), then dried in a SpeedVac. Samples were then resuspended in 12 µl 5% formic acid, 2% acetonitrile and stored at −80°C.

For nano-liquid chromatography tandem mass spectrometry (nLC-MS/MS) analysis, 5 µl of each of the peptide samples were loaded and separated along a C18 column (400 mm, 75 µm ID, 3 µm silica beads) and introduced by nanoelectrospray into an LTQ Orbitrap Velos Pro coupled to an Easy-nLC HPLC (Thermo Fisher). MS/MS data were collected for the top ten most abundant ions per scan over a 140-min time gradient. The order of data collection was randomised to interchange between biological conditions with BSA run between each sample to minimise temporal bias. MS/MS raw files were analysed using the Andromeda search engine integrated into MaxQuant (v1.2.7.4) (Cox and Mann, 2008) against the Uniprot Mouse database. A false discovery rate of 1% was tolerated for protein, peptide and sites, and one missed cleavage was allowed. MaxQuant output data were filtered to remove contaminants, reverse hits, proteins only identified by site and proteins with <2 unique peptides (Croucher et al., 2016). An individual protein was defined as present under a particular condition if it was detected in a minimum of two replicates. Protein–protein interactions and KEGG pathway analysis among the resulting protein list was analysed using STRING (v10) (Szklarczyk et al., 2015) (with a confidence score of 0.700) and Cytoscape (v3.1.1) (Smoot et al., 2011).

NSC-34 cells transiently transfected with SOD1^WT-GFP or SOD1^A4V–GFP were incubated with 100 nM Mitotracker Deep Red FM (Molecular Probes) or Mitotracker CMXRos (Molecular Probes) for 30 min at 37°C prior to analysis by confocal microscopy or flow cytometry at 48 h post transfection. Mitochondrial morphology was examined with a mitochondrial morphology macro (Dagda et al., 2009) on ImageJ 1.48v. Mitotracker Red and Mitotracker CMXRos accumulation in cells expressing high levels of SOD1–GFP was also assessed by flow cytometry (Becton Dickinson).

The authors would like to gratefully acknowledge Clare Watson who helped with editing and proof reading the manuscript.