Many neurodegenerative diseases including amyotrophic lateral sclerosis (ALS) are linked to the accumulation of specific protein aggregates in affected regions of the nervous system. SOD1, TDP-43, FUS and optineurin (OPTN) proteins were identified to form intraneuronal inclusions in ALS patients. In addition, mutations in OPTN are associated with both ALS and glaucoma. As the pathological role of OPTN in neuronal degeneration remains unresolved, we created a yeast model to study its potential for aggregation and toxicity. We observed that both wild type and disease-associated mutants of OPTN form toxic non-amyloid aggregates in yeast. Similar to reported cell culture and mouse models, the OPTN E50K mutant shows enhanced toxicity in yeast, implying a conserved gain-of-function mechanism. Furthermore, OPTN shows a unique aggregation pattern compared to other disease-related proteins in yeast. OPTN aggregates colocalize only partially with the insoluble protein deposit (IPOD) site markers, but coincide perfectly with the prion seed-reducing protein Btn2 and several other aggregation-prone proteins, suggesting that protein aggregates are not limited to a single IPOD site. Importantly, changes in the Btn2p level modify OPTN toxicity and aggregation. This study generates a mechanistic framework for investigating how OPTN may trigger pathological changes in ALS and other OPTN-linked neurodegenerative disorders.
Multiple human diseases are associated with failures of protein quality control, for which the formation of protein aggregates is a common pathological marker. Disorders including Alzheimer disease, Huntington disease and amyotrophic lateral sclerosis (ALS) are each linked to the accumulation of specific protein aggregates in affected tissues (Ross and Poirier, 2004). These aggregates, or their oligomeric precursors, disrupt intracellular mechanisms, impair protein homeostasis and ultimately are linked to cell death. For each disease, the death of various types of neurons is believed to be linked to the aggregation of specific proteins. In the case of ALS, there is selective loss of upper and lower motor neurons, which ultimately leads to patient death, usually from respiratory failure (Zinman and Cudkowicz, 2011).
Several subtypes of ALS can be categorized by the protein species identified within the cytoplasmic aggregates of diseased neurons. The protein SOD1 has long been implicated in familial ALS (Rosen et al., 1993), where its misfolding and aggregation is associated with a gain-of-function toxicity (Karch et al., 2009); as such, ALS cases have been defined as SOD1-positive or SOD1-negative. More recently, several additional proteins have been linked to familial and sporadic forms of SOD1-negative ALS; examples include TDP-43, FUS, ATXN2, TAF15 and UBQLN2 (Kabashi et al., 2008; Sreedharan et al., 2008; Kwiatkowski et al., 2009; Vance et al., 2009; Elden et al., 2010; Deng et al., 2011a). The TDP-43 and FUS proteins, which have received much attention recently, are structurally and functionally similar nuclear DNA/RNA-binding proteins that regulate different aspects of RNA metabolism (Lagier-Tourenne and Cleveland, 2009). Both proteins are aggregation-prone (Johnson et al., 2009; Sun et al., 2011) and form cytoplasmic aggregates in many cases of sporadic and familial ALS, as well as frontotemporal dementia, a disease that shares clinical and pathological features with ALS (reviewed in Lagier-Tourenne et al., 2010; Ince et al., 2011). Successful modelling of TDP-43 and FUS aggregation in different organisms (discussed further below) strongly suggests that these proteins form toxic aggregates independently of other proteins and thus are directly responsible for the formation of pathological neuronal inclusions.
Another protein recently proposed to play a role in the pathogenesis of both sporadic and familial ALS is optineurin (OPTN), which has been found to form characteristic skein-like inclusions in the cytoplasm and neurites of spinal anterior horn cells from ALS patients (Maruyama et al., 2010; Deng et al., 2011b). Several ALS-associated OPTN mutants were identified that result in the production of truncated and rapidly degraded protein variants (loss of function mutations) (Maruyama et al., 2010; Iida et al., 2012), consistent with reports that indicate a possible protective role of OPTN in the nervous system based on an inverse relationship between OPTN expression and the pattern of neuronal loss in Huntington disease (Okita et al., 2011). However, other mutations (for example OPTN E478G) are heterozygous and may result in a gain of toxic function, as is hypothesized for many SOD1 and TDP-43 mutants. Furthermore, different missense mutations in the OPTN gene are also linked to glaucoma, a degenerative eye disease, with OPTN E50K being the most prevalent and most associated with severe phenotypes (Rezaie et al., 2002; Aung et al., 2005).
Intriguingly, OPTN-positive inclusions have been identified for a number of pathological structures, including neurofibrillary tangles and dystrophic neurites in Alzheimer disease, Lewy bodies and Lewy neurites in Parkinson disease, neuronal nuclear inclusions in polyglutamine diseases and ballooned neurons in Creutzfeldt–Jakob disease (Osawa et al., 2011). However, the significance of such a wide distribution remains to be determined (Hortobagyi et al., 2011). OPTN is a ubiquitous cytosolic 67 kDa protein with several coiled-coil domains and a special C-terminal ubiquitin-binding domain. It is partially associated with the Golgi (Sahlender et al., 2005) and other cellular membranes and is involved in several intracellular membrane trafficking steps (Chibalina et al., 2010; Bond et al., 2011). It forms functional complexes with the motor protein myosin VI and the small GTPase Rab8, playing a role in Golgi ribbon formation and exocytosis (Hattula and Peranen, 2000; Sahlender et al., 2005), and also interacts with several other proteins (reviewed: Kachaner et al., 2012; Ying and Yue, 2012). Recently, OPTN has been suggested to function as an autophagy receptor, linking ubiquitinated cargoes with microtubule-associated protein light chain 3 (LC3)-decorated autophagosomal membranes (Wild et al., 2011).
In recent years the yeast Saccharomyces cerevisiae has emerged as a very productive eukaryotic system to study heterologous expression of human disease-related proteins (Winderickx et al., 2008; Khurana and Lindquist, 2010; Kryndushkin and Shewmaker, 2011). It often recapitulates observations made in higher organisms and offers the possibility for high-throughput genetic and small-molecule screening to reveal important information about conserved disease-related molecular pathways. Examples of successful modelling of protein misfolding disorders in yeast include the expression of polyglutamine (Q103), alpha-synuclein (α-SYN), TDP-43 and FUS, each of which results in cytotoxicity and the formation of multiple cytoplasmic aggregates (foci) per cell (reviewed in Gitler and Shorter, 2011; Kryndushkin and Shewmaker, 2011). Genetic screens utilizing these models identified non-overlapping specific sets of proteins that contribute to, or mitigate, the toxicity. As an example, multiple proteins involved in RNA processing were discovered using the FUS expression model (Ju et al., 2011; Sun et al., 2011), suggesting that FUS may contribute to disease pathogenesis via an RNA-mediated pathway. Importantly, many of the findings were confirmed in more complex organism models, demonstrating the outstanding predictive potential of the yeast system.
To manage the accumulation of misfolded proteins, organisms from bacteria to humans have intracellular mechanisms for directing aggregated proteins to specific locations, although the localization of deposition sites and the mechanisms of sorting may differ. The details of aggregate targeting and sorting are of particular interest because of the ubiquity of protein-misfolding diseases. Mammalian cells have an active microtubule-based transport mechanism to localize protein aggregates to a cytoplasmic perinuclear structure called the ‘aggresome’ (Johnston et al., 1998), sequestering potentially toxic misfolded proteins and facilitating their clearance by autophagy (Olzmann et al., 2008). In yeast, two similar cytoskeleton-dependent mechanisms were proposed recently. Wang et al. suggested a microtubule-dependent mechanism that targeted an overexpressed polyglutamine tract (huntingtin exon 1 with expanded polyglutamine and polyproline domains; Q103-P) to a single genuine aggresome site at the yeast centrosome (spindle pole body), marked with the spindle pole body marker Spc72 (Wang et al., 2009). Similarly, Kaganovich et al. characterized a perivacuolar aggresome-like site, coined the insoluble protein deposit (IPOD), where amyloid, prions and other terminally misfolded proteins are actively consolidated (Kaganovich et al., 2008). They also identified a second perinuclear quality control site, termed the JUNQ, where ubiquitinated proteins that retain relatively high mobility are processed (Kaganovich et al., 2008). Neither the JUNQ nor the IPOD was concluded to form at the yeast spindle pole body, but it is nonetheless clear that yeast use the cytoskeletal system to consolidate protein aggregates in a manner that is reminiscent of the mammalian aggresome (Ganusova et al., 2006; Kaganovich et al., 2008; Wang et al., 2009; Mathur et al., 2010).
Significant progress has been made towards understanding the cellular machinery that drives the sequestration of misfolded proteins to specific sites. The small heat shock protein Hsp42 localizes to the aggresome-like IPOD compartment and is required for aggregate formation (Specht et al., 2011). We previously described Btn2p, the yeast homologue of the mammalian Hook proteins (Kama et al., 2007), as a novel anti-prion protein in yeast with aggregate-managing potential (Kryndushkin et al., 2008). We hypothesized that Btn2p is involved in trafficking of protein aggregates to an aggresome-like site, lowering the amount of active prion seeds and preventing their efficient distribution to progeny cells, thus eliminating the prion from the population (Kryndushkin et al., 2008). Indeed, recent evidence suggests that Btn2p can cooperate with the chaperone Hsp42 and Sis1p to promote the sorting of misfolded proteins to both the IPOD and JUNQ (Malinovska et al., 2012). It was initially proposed that the microtubule-based transport is primarily involved, based on the sensitivity to the microtubule-depolymerizing drug benomyl (Kaganovich et al., 2008; Wang et al., 2009). However, an elegant study by Specht and co-workers challenged this finding and posited that the formation of the JUNQ and IPOD in yeast depends likely on intact actin filaments, not microtubules (Specht et al., 2011).
The OPTN-positive skein-like cytoplasmic inclusions found in ALS patients (Maruyama et al., 2010; Deng et al., 2011b) may be harmful by impairing cellular protein homeostasis, or they may trap vital cellular proteins leading to a loss-of-function effect. Given the productivity of yeast models in establishing the molecular basis of protein toxicity, such as for FUS-linked proteinopathy, we utilized a similar approach for OPTN to uncover its intrinsic potential to be a primary driver of protein aggregation and cytotoxicity. By heterologous expression of OPTN in yeast cells we establish that cellular toxicity is associated with cytoplasmic accumulation of non-amyloid OPTN species. Furthermore, we link the OPTN E50K mutation to greater cellular toxicity, recapitulating previous findings in cell culture and suggesting that the OPTN yeast model has potential use in high-throughput screening for genetic or small-molecule modifiers of toxicity. Lastly, we use the formation of OPTN (and FUS) inclusions as a model for how yeast target and process protein aggregates. We provide evidence for the helpful role of targeting misfolded protein, via the actin cytoskeleton, to a single aggregate in yeast to mitigate proteotoxicity and show that Btn2p aids in this process.
OPTN–GFP forms aggresomes in mammalian cells
During transient expression in various cell lines (including human retinal pigment epithelial cells), OPTN–GFP fusion protein demonstrated diffuse cytosolic distribution with perinuclear dots or granular structures (De Marco et al., 2006; Park et al., 2006; Shen et al., 2011). Overexpression of OPTN was reported to cause reduction in proteosomal activity, induction of autophagy and finally apoptotic cell death (Park et al., 2006; Shen et al., 2011). Curiously, stimulation of autophagy with rapamycin decreased the number and size of OPTN–GFP foci, whereas inhibition of autophagy showed the reverse effect, suggesting that the overexpressed OPTN was cleared via autophagy (Shen et al., 2011) and the observed perinuclear foci may represent aggregates of misfolded OPTN–GFP molecules.
To explore whether ubiquitin–proteasome system impairment can cause accumulation of OPTN aggregates in a similar manner to other aggregation-prone proteins (Johnston et al., 1998), we utilized transient expression of OPTN–GFP or GFP for 28 h in HEK293 cells that were exposed to the proteasome inhibitor lactacystin during the final 20 h. Untreated cells showed the formation of several visible foci close to the nucleus as reported previously (Fig. 1; De Marco et al., 2006; Park et al., 2006), and about 5% of cells displayed a single perinuclear fluorescent spot. In contrast, more than ∼70% of lactacystin-treated cells showed the formation of a large single aggregate (or a conglomerate of aggregates) at the microtubule organizing centre (MTOC), which was identified by staining with the MTOC marker γ-tubulin (Fig. 1, upper panels). Moreover, the perinuclear location of the aggregate depended on intact microtubules as multiple scattered small foci were observed after treatment of the cells with nocodazole preventing microtubule polymerization (data not shown). Vimentin staining showed an altered structure of the intermediate filament network in cells with prominent perinuclear OPTN aggregation, another characteristic feature for aggresome formation (Fig. 1, lower panels). Therefore, similar to other aggregation-prone proteins associated with neurodegenerative diseases (Wong et al., 2008), OPTN can form aggresomes in mammalian cells when present at high concentrations.
OPTN forms toxic aggregates in yeast cells, distinct from TDP-43 and FUS
To determine whether yeast cells may recapitulate OPTN aggregation with pathological consequences, we generated a yeast strain harbouring centromeric (low-copy) or high-copy plasmids for OPTN-inducible expression, similar to previous studies of TDP-43, FUS and polyglutamine (Q103) (reviewed in Gitler and Shorter, 2011; Kryndushkin and Shewmaker, 2011). We placed both OPTN and OPTN–GFP under control of the tightly regulated galactose-inducible GAL1 promoter to allow rapid, strong and synchronous induction of protein expression in all cells. The expression of either OPTN or OPTN–GFP was equally toxic to yeast cells (Fig. 2A), slightly less toxic compared to TDP-43 and FUS (Fig. 2B), and resulted in the formation of cytoplasmic foci (Fig. 2C; we will refer below to these foci as aggregates based primarily on their morphological properties as visualized by fluorescence microscopy). Intriguingly, the appearance of OPTN–GFP foci was dramatically different from those formed by TDP-43–GFP and FUS–GFP. In all cells the majority of OPTN–GFP accumulated in a single large spot often with few additional small foci, whereas both TDP-43–GFP and FUS–GFP formed multiple foci that were heterogeneous in size, scattered throughout the cytoplasm (Fig. 2C) and did not colocalize with OPTN (Fig. 2D and data not shown). As the GFP tag does not alter the toxicity of OPTN (Fig. 2A) and it simultaneously permits the monitoring of OPTN aggregation, we decided to use OPTN–GFP in further experiments.
Similar to TDP-43 and FUS, but in contrast to polyglutamine stretches, the toxicity and aggregation of OPTN–GFP was not enhanced or affected by any known yeast prions. The infectious amyloid form of the yeast prion protein Rnq1, designated [RNQ+] or [PIN+], is known to serve as a template for both the formation of yeast prions and the aggregation of other heterologous proteins, especially glutamine/asparagine-rich sequences (Derkatch et al., 2001; Meriin et al., 2002). Hsp104 is required for the propagation of all known yeast prions. The deletion of RNQ1 or HSP104 from a strain that carried the [RNQ+] prion resulted in no obvious change in toxicity or aggregation (Fig. S1). Therefore, the aggregation of OPTN–GFP is different from that of TDP-43 and FUS, but also occurs via a prion/amyloid-independent mechanism.
The yeast model reveals greater toxicity and aggregation propensity of the glaucoma-linked OPTN E50K mutant
Next, we examined the properties of two common disease-linked OPTN mutations: E50K [found in glaucoma patients (Rezaie et al., 2002)] and E478G [found in ALS patients (Maruyama et al., 2010)]. Both wild type and mutated OPTN–GFP proteins were expressed from centromeric (low-copy) plasmids and accumulated to similar levels after 5 h of protein expression induced with galactose (Fig. 3A). Low-copy centromeric plasmids were used in this experiment to avoid any variability that might result from using high-copy plasmids. We observed higher cytotoxicity for OPTN E50K–GFP compared to wild type protein (Fig. 3C), consistent with the behaviour of this mutant in the reported cell culture and mouse models (Park et al., 2006; 2010; Chi et al., 2010; Meng et al., 2012). Interestingly, we observed a higher average number of foci per cell for this mutant (Fig. 3B). In contrast, the OPTN E478G–GFP expression resulted in similar visual aggregation and the same toxicity as wild type OPTN (Fig. 3B and C). This suggests that the disease-related deficiencies for these two OPTN mutants are likely different, and that OPTN E50K may have a toxic gain of function that can be recapitulated in yeast cells.
The N-terminal region of OPTN containing a Rab8 binding site is necessary and sufficient for cytotoxicity
The OPTN protein has several coiled-coil domains, a leucine zipper [amino acids (aa) 143–164] and a LC3-interacting motif (aa 169 to 209) at the N-terminus, and a ubiquitin-binding region at the C-terminus (reviewed: Kachaner et al., 2012). Unlike proteins such as FUS, TDP-43 and many other disease-related proteins, OPTN contains neither prion-like domains nor large intrinsically disordered segments. We generated a series of OPTN truncations and expressed each of the truncated constructs as C-terminal GFP fusions to determine their subcellular localization and toxicity (Fig. 4A). All constructs had comparable expression levels (data not shown) and all but one, in which both coiled-coil regions were missing, formed fluorescent foci (Fig. 4B). However, only constructs containing the N-terminal domain were both toxic and aggregate-forming. As this domain harbours essential elements for RAB8 binding, these results suggest that it is the presence of a RAB8-binding site that renders OPTN aggregates cytotoxic (Fig. 4C). Importantly, the two closest Rab8 homologues in yeast, Sec4p and Ypt1p, are essential for cell viability.
To define whether the OPTN-linked toxicity is due to the functional deficiency of the yeast Rab8 homologues following their recruitment into OPTN aggregates, we expressed either SEC4 or YPT1 simultaneously with OPTN–GFP and compared the levels of toxicity. We found that Ypt1p overproduction slightly reduced the OPTN–GFP toxicity levels (Fig. 4D), consistent with such a scenario; however, factors other than Ypt1p depletion seem to concurrently contribute to the toxicity. Overproduction of Sec4p alone strongly reduced cell growth (data not shown) and thus could not be used in combination with OPTN to assess its effect on OPTN-mediated toxicity. Interestingly, the toxicity mediated by OPTN E50K–GFP was also affected by Ypt1p overproduction; simultaneous overexpression of OPTN E50K–GFP (centromeric vector) and Ypt1 (high-copy vector) showed only minor toxicity, whereas overexpression of OPTN E50K–GFP alone was highly toxic (Fig. 4D, bottom and Fig. 3C). It implies a strong interaction between OPTN E50K and the yeast Rab8 homologue, which is consistent with earlier findings (Nagabhushana et al., 2010; Park et al., 2010).
OPTN forms non-amyloid aggregates in yeast and in vitro
The formation of highly ordered filamentous protein aggregates, known as amyloid, is common to many neurodegenerative diseases. By recruiting monomeric protein molecules of the same kind into the growing fibres, amyloid provides a common pathological mechanism by which protein aggregates can propagate within and between cells. We examined whether OPTN aggregates have amyloid-like properties in yeast cells. Amyloid fibres are rich in beta-sheet structure and share unusual resistance to strong anionic detergents like sodium dodecyl sulphate (SDS) or sarkosyl, reflecting the strength of hydrogen bonding within the structure. The degree of this resistance is heavily dependent on the protein: while yeast prions, polyglutamine stretches and many other amyloid aggregates are resistant to 1% SDS treatment at room temperature, non-amyloid aggregates, formed by FUS or TDP-43, are SDS-sensitive, but may be partially resistant to a weaker anionic detergent, such as 1% sarkosyl. We examined whether OPTN–GFP aggregates share this resistance. Lysates from cells expressing Q25–GFP (the stretch of 25 glutamines does not aggregate in yeast and serves as a negative control), OPTN–GFP or FUS–GFP were treated with 1% sarkosyl for 10 min at room temperature and analysed by a filter retardation assay (Muchowski et al., 2000). While FUS–GFP formed sarkosyl-insoluble aggregates unable to pass through a 0.2 μm cellulose acetate membrane, OPTN–GFP aggregates passed freely through the membrane (Fig. S2A). Therefore, based on their biophysical resistance to mild detergent treatment, yeast OPTN–GFP aggregates are not amyloid-like, with a lower resistance to detergents than FUS and TDP-43, and no resistance to SDS (data not shown).
We also examined the capacity of bacterially expressed recombinant OPTN to form amyloid under neutral conditions. OPTN was purified under both native and denaturing conditions and exchanged into phosphate buffer (pH = 7.4) with reducing agents. Regardless of the purification method, OPTN remained soluble for at least 24 h at room temperature without agitation. Circular dichroism (CD) spectroscopic analysis revealed an initial high α-helical content for the soluble OPTN species [> 80% α-helical content, according to the Dichroweb (Lobley et al., 2002)], but during prolonged incubation (days) at higher concentrations (> 0.5 mg ml−1 protein) aggregation was evident by both turbidity (absorbance at 395 nm) and low-speed sedimentation (data not shown). For comparison, TDP-43 and FUS are both inherently prone to aggregation and rapidly precipitated from solution after a very short lag phase (Johnson et al., 2009 and data not shown). The OPTN aggregates were SDS-sensitive and did not yield the amyloid-specific increase in fluorescence when incubated with the dye thioflavin T (Fig. S2B). Moreover, OPTN aggregates differ from FUS aggregates morphologically as revealed by transmission electron microscopy: OPTN aggregates were amorphous, whereas FUS formed skein-like assembles (Fig. S2C).
OPTN partially localizes to the IPOD and reveals the complexity of peripheral protein deposition in yeast
Overexpression of conformationally unstable proteins in yeast cells often results in the formation of cytosolic aggregates (Khurana and Lindquist, 2010). So far, at least two distinct patterns of aggregation have been reported: the formation of one to two large protein inclusions [examples include accumulation of misfolded proteins at the IPOD and JUNQ (Kaganovich et al., 2008)] or the formation of multiple aggregates scattered randomly throughout the cytoplasm [as observed with FUS, TDP-43 (Fig. 2C) and Q103–GFP expression, or with expression of yeast prion proteins in cells harbouring the corresponding prion]. Interestingly, a particular pattern is not an absolute; some proteins can change their aggregation manner depending on other factors. For example, Q103–GFP alone forms multiple aggregates (Wang et al., 2009; Fig. 5C); however, it can be targeted to a single inclusion by coexpression with certain glutamine/asparagine-rich proteins (Wang et al., 2009; Fig. S3).
The aggregation pattern of OPTN–GFP in yeast is unique, representing a mixture between two previously reported patterns. At early time points of OPTN–GFP expression (1 h induction on the galactose medium) only a single focus can be detected by fluorescent microscopy, whereas after 5 h of induction (about three- to fourfold increase in the protein expression level compared to 1 h induction) additional (one to three) smaller foci become visible (Fig. 5A). We examined whether the main inclusion coincides with the IPOD and JUNQ sites, utilizing a destabilized Ubc9 variant that marks both the IPOD and JUNQ (Kaganovich et al., 2008). Both sites were distinguished by their different localizations, the JUNQ being perinuclear and the IPOD more peripheral. We were able to detect localization of OPTN–RFP at the IPOD, but not the JUNQ (Fig. 5B). As the Ubc9 marker behaved inconsistently in our hands and often generated abundant diffuse fluorescence that interfered with IPOD detection, we utilized other published IPOD markers to quantify the presence of OPTN at this site. The yeast prion proteins RNQ1–GFP, Ure2–GFP and the chaperone Hsp104–GFP were reported to be targeted to the IPOD (Kaganovich et al., 2008), and we confirmed that these proteins form a single aggregate in all cells during coexpression (Fig. S3). Next, these proteins were each coexpressed with OPTN–RFP and the frequencies of colocalization of the foci in a single cell were counted. We observed that in 84% (64 from 76), 20% (13 from 76) and 61% (78 from 128) of the cases, the main OPTN–RFP focus overlapped partially or completely with the RNQ1–GFP, Ure2–GFP and Hsp104–GFP foci respectively (Fig. 5C). In the other cases, OPTN foci were distinct from the JUNQ as well, as they were often not juxtanuclear (see Fig. 6B). These data suggest that peripheral protein deposit sites could not be restricted to a single location as was originally proposed (Kaganovich et al., 2008). Instead, misfolded proteins can be partitioned into a few separate peripheral aggregates.
Surprisingly, various combinations of overlap and non-overlap were observed for OPTN with selected amyloidogenic proteins. For example, OPTN does not colocalize with amyloid-forming Ure2p or Q103 during coexpression (Fig. 5C); however, it colocalizes significantly with amyloid-forming Rnq1 (84% colocalization) and colocalizes in all cases with Q103-P (100 cases observed; Fig. 5C). Importantly, the four mentioned amyloidogenic proteins colocalize in all cases in pair-wise comparisons during coexpression (Fig. S3). This apparent contradiction suggests that the observed colocalization events did not occur in a single defined physical compartment, and the extent of colocalization may reflect the specific affinity between overproduced proteins. On top of that, the non-amyloid proteins FUS and TDP-43 form a separate category, generating multiple peripheral aggregates in a single cell that do not colocalize with OPTN (Fig. 2D), the nucleus or various amyloidogenic proteins (Kryndushkin et al., 2011). Thus, the current classification of peripheral protein deposit sites in yeast requires further refinement.
Formation of OPTN inclusions is likely actin-dependent
Next, we examined whether the cytoskeleton plays an active role in OPTN foci formation in yeast cells. The expression of OPTN–GFP plasmid was induced for 5 h in the presence of either benomyl, an inhibitor of microtubule polymerization, or the actin-depolymerizing drug latrunculin A (LatA). Both treatments blocked formation of a single large OPTN–GFP aggregate, and instead multiple smaller aggregates appeared in the yeast cells (Fig. 6A). This phenomenon was recently described by Specht and co-workers for IPOD-linked protein aggregates (Specht et al., 2011). They concluded that the effect of benomyl was non-specific because a benomyl-resistant tubulin mutant likewise failed to form aggregates at the IPOD following benomyl treatment. Instead, in a yeast strain with a mutation that prevents LatA-mediated disassembly of the actin cytoskeleton (Ayscough et al., 1997), aggregates were effectively targeted to both the IPOD and JUNQ in the presence of LatA. Our results suggest that OPTN–GFP behaves similarly to IPOD substrates described by Specht and co-workers (Specht et al., 2011), consistent with its partial localization with these proteins. Indeed, in more than 50% of the cases, the main OPTN–GFP focus was not juxtanuclear (Fig. 6B) and did not colocalize with the spindle pole body (yeast MTOC) marker Spc72 (Fig. 6C). Therefore, we conclude that the formation of OPTN–GFP foci depends on a functional actin cytoskeleton, although additional microtubule dependence cannot be excluded.
Formation of aggregates at the IPOD relieves proteotoxic stress
Although multiple proteins can form large cytosolic aggregates during overproduction in yeast, only a subset shows cytotoxicity. In fact, even some disease-related misfolded proteins cause no growth disadvantage during aggregation. Examples of such benign behaviour include the yeast expression of Ure2N–GFP (the amyloid-forming prion domain of Ure2p), PrP–GFP (mouse prion protein; Fig. S4) and Q103-P–GFP (Wang et al., 2009), which all show the formation of a single large aggregate at the IPOD (Fig. S3). In contrast, overproduction of α-SYN, FUS, TDP-43, Q103 in yeast, as well as Sup35p and Rnq1p in the presence of corresponding prions, is toxic and is associated with formation of multiple foci (large protein aggregates) in a single cell (Khurana and Lindquist, 2010; Kryndushkin and Shewmaker, 2011). Interestingly, the higher numbers of foci often correlate positively with cytotoxicity (Johnson et al., 2009; McGlinchey et al., 2011). The mechanisms of toxicity may be protein-dependent, but can also involve sequestration of an essential protein into aggregates (Vishveshwara et al., 2009; Gong et al., 2012; Treusch and Lindquist, 2012). Similar to the mammalian aggresome, the formation of a single aggregate at the IPOD may sequester toxic protein species and serve as a protective mechanism. Indeed, coexpression of Q103 and Q103-P was sufficient for targeting Q103 into a single aggregate, preventing its cytotoxicity (Wang et al., 2009).
We explored factors that can drive an aggregation-prone protein to the IPOD. We utilized the previously published yeast expression of the ALS-linked protein FUS, which accumulates into multiple toxic foci distributed randomly in the cytoplasm with no IPOD colocalization (Kryndushkin et al., 2011) (Fig. 7A and B). Unexpectedly, we found the aggregation behaviour and toxicity level were greatly dependent on the orientation of the GFP tag (used to visualize FUS aggregates). While the C-terminal FUS–GFP fusion expression was similarly toxic as wild type FUS (Kryndushkin et al., 2011), the N-terminal GFP–FUS fusion expression was no longer toxic (Fig. 7A) and accumulated in one or two spots at the IPOD, as shown by colocalization with RNQ1–RFP (Fig. 7B and C). It also had a much slower aggregation rate (as judged by abundant diffuse cytoplasmic staining) compared to FUS–GFP. Paradoxically, a similar GFP–FUS fusion protein was previously reported to be toxic (Ju et al., 2011). We confirmed this finding using the original construct (designated GFP–FUS*) (Fig. 7A) that differs from our GFP–FUS by having a slightly different GFP version and a 30-aa linker between GFP and FUS (instead of the six-aa linker in our construct). Comparison of expression levels showed that GFP–FUS*, driven by the same promoter, accumulated to much higher levels than GFP–FUS during induction under identical conditions (Fig. 7D). The GFP–FUS* aggregates displayed often two or more foci per cell, representing an intermediate behaviour between that of GFP–FUS and FUS–GFP in respect to aggregation, toxicity and extent of IPOD targeting (Fig. 7A and B, and data not shown).
We found a similar phenomenon by placing GFP at the amino terminus of OPTN instead of the carboxy terminus. Similar to OPTN–GFP, GFP–OPTN accumulated in a single predominant aggregate (Fig. S5B), surrounded by diffuse fluorescence; additional smaller foci were rarely observed. Moreover, GFP–OPTN was not toxic (Fig. S5A) and a substantial proportion of cells contained no aggregates (Fig. S5C). The latter implies a lower aggregation rate for GFP–OPTN, which is consistent with the lower expression of this protein relative to OPTN–GFP (Fig. S5D). Therefore, protein expression levels and/or subtle changes in conformation are important factors that influence the effectiveness by which misfolded proteins can be sequestered to IPOD-like sites; efficient IPOD targeting may be sufficient for relieving cytotoxicity caused by protein accumulation.
Btn2p is involved in managing OPTN aggregates and partially suppresses OPTN cytotoxicity
We previously discovered that the Btn2 protein cures the yeast prion [URE3] when overproduced (curing indicates the elimination of the prion during cell proliferation) and colocalizes with the [URE3] prion aggregates during the curing process (Kryndushkin et al., 2008). Very recently, Btn2p was shown to be involved in sorting of misfolded non-amyloid proteins to both the IPOD and JUNQ via cooperation with Hsp42 and Sis1p (Malinovska et al., 2012). We examined whether Btn2p can participate in managing OPTN aggregates. Indeed, Btn2p and the main OPTN inclusion coincide in all cases during coexpression, suggesting that Btn2p marks the site where OPTN aggregates are deposited (Fig. 8A). Moreover, deletion of BTN2 generated a significant increase in the number of cells with multiple (more than three) foci for OPTN–GFP, when OPTN–GFP was expressed from the low-copy vector (Fig. 8B). We also detected fewer minor OPTN–GFP fluorescent foci with Btn2p coexpression (Fig. 8C), suggesting that Btn2p helps to sequester OPTN–GFP to a main IPOD-like inclusion. We could not detect direct binding between recombinant Btn2p and OPTN proteins in pull-down experiments under near physiological conditions (using fusions to glutathione S-transferase; Fig. S6), consistent with our previous result for Btn2p and Ure2p prion protein (Kryndushkin et al., 2008). Btn2p may interact with OPTN and other substrates indirectly, perhaps via chaperone cofactors Hsp42 and Sis1p (Malinovska et al., 2012).
Next, we examined whether Btn2p can control the toxicity associated with OPTN–GFP expression, testing both deletion and overexpression of BTN2. Coexpression of Btn2p and OPTN–GFP partially suppressed the cytotoxicity, whereas the deletion of BTN2 increased cytotoxicity (Fig. 8D). Note, that the OPTN–GFP expression levels were unchanged regardless of the Btn2p expression (Fig. S7A). Finally, we checked for additional proteins that colocalize with Btn2p and found that aggregates of Q103-P, PrP and Ure2N coincide with Btn2p, whereas Rnq1p shows only partial colocalization (in ∼50% of cases) (Fig. S7B). This indicates that the function of Btn2p is not restricted to non-amyloid aggregates as previously hypothesized (Malinovska et al., 2012).
OPTN, as well as SOD1, TDP-43 and FUS, is each known to accumulate independently as skein-like cytoplasmic inclusions in spinal cord neurons of ALS patients, and mutations in corresponding genes are associated with inherited forms of ALS. Moreover, OPTN inclusions are present in a number of other neurodegenerative disorders (Osawa et al., 2011), calling for a deeper understanding of OPTN biology. We hypothesized that abnormal accumulation of OPTN due to mutations [such as shown for OPTN E478G (Maruyama et al., 2010)] or under stress conditions (Shen et al., 2011) may result in its misfolding and/or aggregation leading to neuronal death in the cases of ALS and glaucoma. Here, we have established a new yeast model to investigate how altered expression of OPTN (or its mutants) can contribute to ALS or glaucoma pathology. We provide evidence that during overproduction the protein forms aggresomes in mammalian cells and non-amyloid aggregates at an IPOD-like site in yeast, becoming toxic with higher expression, thus suggesting a possible dominant role for OPTN in disease pathogenesis. Yeast expression of OPTN recapitulates core findings observed in the mammalian system, such as aggregation-associated cytotoxicity, aggresome-like accumulation and an increased toxicity of the glaucoma-linked OPTN E50K mutation. In addition, we show the specific nature of OPTN aggregates. First, they do not share the detergent resistance or [RNQ+] dependence of yeast prions and Huntington disease-related polyglutamine aggregates expressed in yeast. Second, OPTN has a different pattern of aggregation than other human disease-related proteins expressed in yeast. This suggests that yeast is a suitable system to investigate different mechanisms of protein aggregation and toxicity in eukaryotic cells. The described model can be used for high-throughput genomic and chemical screening to uncover conserved molecular pathways associated with OPTN cytotoxicity.
There is currently no agreement on how OPTN can trigger disease pathology in humans. Based on our data, several pathways likely concurrently contribute to OPTN-mediated toxicity in yeast, similar to what was reported for other yeast disease models (Su et al., 2010). One common mechanism that has been proposed involves sequestration of an essential protein into aggregates. We determined that the Rab8-binding region of OPTN, but not the ubiquitin-binding domain, is responsible for toxicity. Moreover, overexpression of the Rab8 yeast homologue, Ypt1p, partially restores cell growth, indicating possible recruitment and depletion of this essential protein. Interestingly, the greater toxicity of the OPTN E50K mutant was also reduced by Ypt1p overproduction, supporting an interaction between OPTN E50K and the yeast Rab8 homologue.
We utilized the OPTN model to gain new insights in how yeast cells control protein aggregation. The current paradigm stands for the accumulation of misfolded proteins in the specific intracellular JUNQ and IPOD sites (Kaganovich et al., 2008; Specht et al., 2011). Our data strongly suggest the existence of multiple IPOD-like sites formed by overlapping sets of substrates. Two unstable proteins that are coexpressed in a single cell can form separate or overlapping inclusions, likely depending on mutual affinity. Therefore, the existence of a physical space in a cell where all misfolded proteins are targeted and accumulated appears to be an oversimplification. This is also evident from earlier studies of yeast prion aggregates (Zhou et al., 2001). Nevertheless, certain classes of proteins do show remarkable colocalization within a group, such as FUS and TDP-43 (Kryndushkin et al., 2011; Sun et al., 2011), or amyloidogenic glutamine-rich sequences (Kaganovich et al., 2008; Fig. S3). In fact, Malinovska and co-workers came to a similar conclusion, clearly distinguishing that non-amyloid and amyloidogenic proteins should be sequestered into distinct locations (Malinovska et al., 2012). Based on our data, the real picture can be even more complicated and non-amyloid proteins may or may not be partitioned together with amyloidogenic proteins. Further studies will be required to classify such groups in detail.
Importantly, the inability of a protein to be contained within a single inclusion often leads to cytotoxicity. Our data indicate that such a protein can be forced to aggregate at a single site by manipulating its expression level (as shown for GFP–FUS), conformation, or by coexpression with a different protein that provides a targeting motif that facilitates inclusion formation [the latter scenario was previously shown for Q103 aggregation (Wang et al., 2009)]. The formation of a single inclusion instead of multiple aggregates restricts toxic species, reduces pressure on protein quality control machinery and may eventually restore normal cellular growth. It can be seen as ‘controlled aggregation’, a mechanism that appears to be conserved in all living cells. OPTN demonstrates only an intermediate toxicity compared to FUS, TDP-43 or polyglutamine in yeast; and its aggregates are often in a single focus. However, with greater OPTN expression, or with mutant OPTN E50K, the formation of more than a single focus correlates with the severity of toxicity. Conversely, reduction in OPTN concentration mitigates the toxicity. Perhaps, there is a certain threshold concentration for a misfolded protein where it can no longer be efficiently recruited to a single site and neutralized; proteins with efficient targeting would only cause cytotoxic effects upon accumulation above this threshold. Alternatively, ‘aggresome’ targeting does not ameliorate cytotoxicity in yeast in the presence of specific prions (Gong et al., 2012).
Cellular proteins involved in managing misfolded protein species can be anticipated to mitigate cytotoxicity. We examined Btn2p, a protein that marks yeast prion aggregates during the prion-curing process and prevents their effective transmission to daughter cells (Kryndushkin et al., 2008), for its ability to function on OPTN aggregates. We found that Btn2p colocalizes with the OPTN main inclusion in all observed cases and promotes assembly of the inclusion (Fig. 8). Importantly, Btn2p leads to partial suppression of OPTN-mediated cytotoxicity (Fig. 8D). This result is consistent with the previously observed effect of Btn2p on [URE3] prion propagation, where the deletion of Btn2p stabilized and strengthened the prion, whereas overexpression reduced prion seed number through accumulation of prion aggregates into a Btn2p-labelled inclusion (Kryndushkin et al., 2008). Taken together, Btn2p helps to sequester misfolded aggregated protein species into a large intracellular aggregate, playing an important role in cellular protection. Intriguingly, this aggregate or Btn2p itself colocalizes only partially with the IPOD markers, suggesting that Btn2p works preferentially with specific substrates, such as OPTN or Ure2p. We also found perfect colocalization of Btn2p with Q103-P and mouse prion protein PrP expressed in yeast, arguing that the Btn2p function is not restricted to non-amyloid aggregates as previously suggested (Malinovska et al., 2012), but extends to amyloidogenic proteins. Interestingly, a human homologue of Btn2p, Hook2 was shown to promote the formation of an aggresome with a model substrate and enhance the recruitment of aggresome components to the centrosome (Szebenyi et al., 2007). Based on this and our data, we predict beneficial effects resulting from Hook protein expression in mammalian proteotoxicity models, as well as in Batten disease models, as Hook proteins are linked to Batten disease (Luiro et al., 2004; Weimer et al., 2005).
Our work, in addition to many others, establishes a plausible scenario for OPTN-related pathological alterations that underlie neuronal degeneration in diseases like ALS or glaucoma. Cellular stresses, like bacterial infection or proteosomal inhibition, may cause upregulation of OPTN expression (Shen et al., 2011; Wild et al., 2011). The immune regulator tumour necrosis factor has also been shown to increase OPTN transcription (Li et al., 1998). Chronic stimulation of OPTN expression could result in pathological accumulation of the protein. Alternatively, dominant mutations (for example, OPTN E50K or E478G) may further destabilize OPTN structure and/or alter functional interactions (for example, with Rab8). Both ways may lead to the accumulation of cytotoxic OPTN species, deregulation of protein homeostasis and, eventually, neuronal cell death. Evaluation of the OPTN protein levels in diseased tissue could feasibly test this hypothesis.
Yeast strains, growth conditions, microscopy and plasmids
Plasmids used in this study are listed in Table 1. Plasmid construction is described in the Supplementary methods and primers are listed in Table S1. Strains BY4741 (MATa his3 leu2 met15 ura3, [PIN+], [psi-]) and W303 (MATa can1–100 his3–11,15 leu2–3,112 trp1–1 ura3–1 ade2–1, [PIN+], [psi-]) were used for protein expression. Strain 74-D694 (MATa his3–11,15 leu2–3,112 trp1–1 ura3–1 ade1-14, [PIN+], [PSI+]) and its derivative bearing a HSP104 deletion (kindly provided by Susan Liebman) were used in Fig. S1. Yeast cells were grown at 30°C in standard synthetic media (SD) containing 2% glucose, or 2% raffinose (SDRaf), or 2% galactose and 1% raffinose (SDGal) as a carbon source. Only required amino acids were added during cell growth. YPD/glycerol medium (20% glycerol was added to the standard YPD) was used for freezing and saving yeast strains. GuHCl treatment was performed for 60 generations on plates YPD supplemented with 4 mM GuHCl. For protein induction cells were grown overnight in SDRaf medium, adjusted to OD600 = 1 and then incubated in liquid SDGal medium for 5 h. For toxicity assays, yeast cells were grown overnight in liquid SDRaf. Cultures were then normalized by OD600, serially diluted in sterile 96-well plates, spotted onto solid SDGal medium and grown at 30°C for 3 days. For most toxicity experiments the strain BY4741 was used. Some toxicity assays were replicated using W303 with similar outcome.
Imaging of live yeast cells expressing the appropriate GFP- and RFP-tagged fusion proteins was performed on a Zeiss Pascal laser scanning confocal microscope equipped with a Plan-Apochromat 100×/1.4 oil objective (Carl Zeiss) and appropriate filters. The pinhole size was set at 0.5 μm. Image acquisition and analysis were carried out with LSM 5 software. Nucleoplasmic protein Pus1p was used as a nuclear localization marker. During the colocalization experiments, cells were grown overnight in SDRaf medium and then 2% galactose was added for the induction. After 5 h, cells were collected, washed and analysed by confocal microscopy. Two per cent poly-L-lysine and 2% concanavalin A (Sigma, St Louis, MO, USA) were used for coating glass slides (5 min at room temperature) to immobilized yeast cells.
Cell culture, transfection and immunofluorescence microscopy
HEK293 cells (Flp-In™-293, Invitrogen) were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS. For immunofluorescence microscopy, cells were grown on glass coverslips coated with poly-D-lysine and transiently transfected using FuGENE HD Transfection Reagent (Roche Applied Science, Indianapolis, IN, USA) according to the manufacturer's instructions. Forty-eight hours post transfection, cells were fixed either with methanol for 15 min at −20°C (for γ-tubulin staining) or with 4% paraformaldehyde (PFA) in 100 mM sodium cacodylate buffer containing 3 mM of each CaCl2, MgCl2 and KCl, pH 7.4, for 30 min at room temperature (for vimentin staining). Methanol-fixed cells were rehydrated in PBS and PFA-fixed cells were permeabilized in 0.1% Triton X-100/PBS for 5 min at room temperature before blocking with 1% gelatin from cold water fish skin (Sigma, St Louis, MO, USA) in PBS for 1 h. Cells were incubated with primary antibodies diluted in blocking solution overnight at 4°C, followed by washes and incubation with secondary antibodies for 30 min at room temperature. Rabbit anti-GFP (ab290, 1:2000; Abcam, Cambridge, UK) was combined with either mouse anti-γ-tubulin (GTU-88 ascites 1:200, Sigma T6557) or mouse anti-vimentin (LN-6, 2.5 μg ml−1; Sigma V2258), and these pairs of primary antibodies were visualized with DyLight 488 donkey anti-rabbit and DyLight 549 donkey anti-mouse secondary antibodies (Jackson ImmunoResearch). Coverslips were mounted on glass slides with ProLong Gold antifade reagent with DAPI (Molecular Probes) and allowed to cure overnight before imaging. Images were captured using a Zeiss Pascal laser scanning confocal microscope equipped with a Plan-Apochromat 63×/1.4 oil objective (Carl Zeiss) and processed using LSM 5 and Adobe Photoshop software.
Lysate preparation, immunoblotting and filter retardation assay
Yeast cultures were grown for 6 h in SDGal to induce expression of OPTN–GFP or other proteins. Cells were harvested and washed once with water. Cell pellets were resuspended in lysis buffer [50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 3 mM EDTA, 5% glycerol, 5 mM DTT, and Complete (Roche, Indianapolis, IN, USA) protease inhibitor mixture]. Cell disruption was performed in a bead beater with an equal volume of acid-washed glass beads for 3 min at 4°C. Cellular debris was removed by low-speed centrifugation (10 min at 1500 g). BCA reagent (Pierce, Rockford, IL, USA) was used to normalize protein concentrations. Yeast lysates were subjected to SDS/PAGE (any kD gels, Bio-Rad) and transferred to a PVDF membrane (Bio-Rad). Immunoblotting was performed according to standard protocols and described in Kryndushkin et al. (2011). Primary antibodies were used as follows: anti-GFP monoclonal antibody (Roche); anti-Hsc70 monoclonal antibody (LSBio); anti-Hsp104 polyclonal antibody (Stressgen); anti-OPTN polyclonal antibody (Sigma); all dilutions were 1:1000. Alkaline phosphatase-conjugated anti-mouse and anti-rabbit secondary antibodies were used at 1:5000.
Filter retardation assay using cellulose acetate membranes were performed as described previously (Muchowski et al., 2000). Briefly, lysates of cells expressing OPTN–GFP or other proteins for 5 h in SDGal were equalized, treated with 1% sarkosyl for 10 min at room temperature, vacuum blotted to cellulose acetate, washed with blotting buffer and stained with an anti-GFP antibody (Roche, Indianapolis, IN, USA).
Protein expression and purification
Btn2p was purified as previously described (Kryndushkin et al., 2008). OPTN and FUS were expressed from FS309 and FS198 respectively, which encode full-length proteins with an additional C-terminal hexahistidine tag. Freshly transformed BL21-CodonPlus (DE3) RIPL cells (Agilent) with expression plasmids were grown overnight in LB medium supplemented with 100 μg ml−1 ampicillin, 40 μg ml−1 streptomycin and 40 μg ml−1 chloramphenicol. Cells were collected, washed and grown in fresh LB medium from initial OD600 = 0.05 to OD600 = 0.7. At this stage, 0.1 mM isopropyl b-D-1-thiogalactopyranoside (IPTG) was added to induce protein expression. After 4 h induction, cells were collected, washed and saved by freezing at −80°C. Proteins were purified by affinity chromatography using Ni-NTA Agarose (Qiagen) columns and AKTAprime plus chromatography system (GE Healthcare) under both native and denaturative conditions according to the manufacturer's instructions. Proteins were eluted with the following buffers: 50 mM NaH2PO4, 300 mM NaCl, 250 mM imidazole, 10 mM beta-mercaptoethanol, pH = 8.0 (native purification) and 6 M GuHCl, 50 mM NaH2PO4, 300 mM NaCl, 200 mM imidazole, 10 mM beta-mercaptoethanol, pH = 8.0 (denaturative purification). After purification, proteins were concentrated to 10 mM using Amicon Ultra centrifugal filter units (10 kDa molecular weight cut-off; Millipore). Protein was then centrifuged for 30 min at 16 100 g to remove any aggregated material. After centrifugation, the protein concentration was determined by OD280 or Bradford assay (Bio-Rad) and the proteins were used immediately for aggregation reactions.
Thioflavin T (ThT) fluorescence assay
To induce aggregate formation, purified OPTN or Sup35NM (see Shewmaker et al., 2006 for purification procedure) was exchanged into PBS + 3 mM beta-mercaptoethanol (20 μM final concentration), incubated for 24 h at room temperature with mild rotation. One hundred microlitres of aliquots were mixed with a 100 μl ThT solution (10 μM final concentration; 50 mM Tris, pH = 8.5) in a 96-well plate and incubated for 5 min at room temperature. Fluorescent spectra were recorded using Synergy H1 Hybrid microplate reader (Bio-Tek) with settings at 442 nm (excitation) and 470–560 nm (emission). Freshly purified OPTN was used as a monomeric control.
Far-UV CD spectroscopy
OPTN protein, freshly purified under native conditions, was exchanged into 20 mM phosphate pH 7.4 and 3 mM beta-mercaptoethanol using PD-10 buffer exchange columns (GE Healthcare), and diluted to 0.1 mg ml−1 with the same buffer. Circular dichroic spectra were recorded on a JASCO J-815 spectropolarimeter with the bandwidth set at 1 nm. Peptide samples were analysed at 20°C in quartz cuvettes of 1 mm path length. Each spectrum represents the background-corrected average of five successive scans. The mean residue molar ellipticity, [θ], was calculated according to the equation, [θ] = (θ × MR)/(10 × l × c). θ is the measured ellipticity in millidegrees, MR is the mean residue mass (molecular weight of the peptide divided by the number of amino acid residues), l is the optical path length in cm and c is the protein concentration in mg ml−1. Deconvolution of far-UV CD spectra into secondary structural components (helix, strand, random coil) was performed using the K2D algorithm from the suite of programs available at the online server DICHROWEB (http://dichroweb.cryst.bbk.ac.uk) (Lobley et al., 2002).
Transmission electron microscopy
Recombinant FUS and OPTN protein samples were dispensed as 10 μl aliquots onto carbon-coated copper grids (Ted Pella). After about 2 min, the protein samples were blotted off with absorbent paper and 10 μl of water was quickly added to the grid surface. Next, the water was blotted off and 10 μl of 1.5% uranyl acetate stain was added to the grid surface. After about 2 min, the stain was blotted off and the grid was allowed to air-dry. Images were collected using a Philips CM100 operating at 80 kV.
We thank Herman Edskes (NIH, USA) for providing plasmids and for his thoughtful suggestions; Nathalie Gerassimov for her assistance with mammalian cell transfection and immunofluorescence; Ernest Maynard for helping with CD spectroscopy; Gregory Petsko (Brandeis University, USA), Kathryn Ayscough (University of Sheffield, UK), Beatrice Yue and Susan Liebman (University of Illinois, USA) and Michael Sherman (Boston University, USA) for sharing plasmids and strains. We also thank the USUHS BIC facilities for providing assistance with confocal microscopy. The research was supported by Uniformed Services University of the Health Sciences.
Conceived and designed the experiments: DK, FS. Performed the experiments: DK, TCP, FS. Contributed reagents/methods: GI. Analysed the data: DK, FS. Wrote the paper: DK, GI, FS.