Arsenite interferes with protein folding and triggers formation of protein aggregates in yeast

Certain metals and metalloids are known to adversely impact living organisms. For example, arsenic (As), cadmium (Cd) and chromium (Cr) are clearly toxic to cells and several compounds containing these agents are classified as human carcinogens (Beyersmann and Hartwig, 2008). Metal/metalloid toxicity is often attributed to inhibition of protein function through binding to free thiols or other functional groups in native proteins, or through displacement of essential metal ions in metal-dependent proteins (reviewed by Beyersmann and Hartwig, 2008; Sharma et al., 2011; Wysocki and Tamás, 2010). Recent in vitro studies indicated that certain metals and metalloids inhibit refolding of chemically denatured proteins (Ramadan et al., 2009; Sharma et al., 2008). Whether non-native or nascent proteins are targeted also in living cells is still unknown and the cellular basis of metal/metalloid toxicity is poorly understood.

Folding of most nascent proteins is co-translational and assisted by molecular chaperones (Frydman, 2001). However, the processes of protein folding and quality control are prone to errors and susceptible to environmental stress conditions resulting in aberrant protein conformers and aggregation, as in the case of several neurodegenerative diseases and age-related disorders (Buchberger et al., 2010; Goldberg, 2003). To prevent the accumulation of potentially toxic protein aggregates, cells utilize highly conserved defense mechanisms including molecular chaperones that facilitate disaggregation and reactivation of aggregated proteins, and the ubiquitin–proteasome pathway that contributes to aggregate removal through degradation (Buchberger et al., 2010; Goldberg, 2003; Liberek et al., 2008; Sharma et al., 2009).

Trivalent arsenic [arsenite, As(OH)₃ or As(III)] has been shown to interact with several native proteins in mammalian cells, such as actin, tubulin and thioredoxin reductase, and the classical view is that enzyme inhibition is due to binding (Aposhian and Aposhian, 2006; Menzel et al., 1999; Zhang et al., 2007). As(III) disrupts the actin and tubulin cytoskeleton in budding yeast Saccharomyces cerevisiae (Pan et al., 2010; Thorsen et al., 2009), but whether this is caused by direct binding to actin and tubulin is not known. S. cerevisiae responds to As(III) exposure by strongly enhancing expression of chaperone- and proteasome-encoding genes; upregulation of proteasomal components is mediated by the transcription factor Rpn4p and cells lacking this factor are As(III) sensitive (Haugen et al., 2004; Thorsen et al., 2007; Thorsen et al., 2009). Rpn4p is also required for Cd (Thorsen et al., 2009) and Cr tolerance (Holland et al., 2007). Together, these observations suggest that metal-treated cells may accumulate damaged proteins that need to be eliminated for optimal tolerance.

Although several lines of evidence indicate that protein homeostasis and quality control are affected by metals and metalloids, the current understanding of how these agents impact the proteome in living cells is limited. This study demonstrates that As(III) interferes with protein folding and triggers formation of protein aggregates in S. cerevisiae. We also show that As(III)-induced protein aggregates can act as seeds committing other, labile proteins to misfold and aggregate. This mechanism of toxic action may explain the suggested role of As(III) in the etiology and pathogenesis of certain protein folding disorders.

To investigate whether metals/metalloids cause protein aggregation in living yeast cells, we monitored the subcellular distribution of a green fluorescent protein (GFP)-tagged version of the chaperone Hsp104p that promotes protein disaggregation and refolding (Glover and Lindquist, 1998). Hsp104p–GFP was evenly distributed throughout the cytosol in unexposed cells, whereas As(III) (NaAsO₂) triggered Hsp104p–GFP redistribution to distinct foci (Fig. 1A), previously shown to represent sites of protein aggregation (Kawai et al., 1999; Lum et al., 2004). Cells exposed to 0.1 mM As(III) formed 1–3 foci/cell, while cells exposed to 0.5 mM As(III) formed larger and more foci/cell (Fig. 1A), indicating that As(III)-induced aggregate formation was concentration dependent. Protein aggregation was triggered by intracellular As(III), since Hsp104p–GFP foci were absent in cells overexpressing the plasma membrane-localized As(III) efflux permease Acr3p (Wysocki et al., 1997) (Fig. 1B). To confirm biochemical association of Hsp104p with As(III)-induced protein aggregates, we exposed cells to 0.1 mM or 0.5 mM As(III), isolated aggregated proteins by sedimentation, and probed the localization of Hsp104p by western blot analysis. For untreated cells, the majority of Hsp104p was in the soluble fraction, whilst As(III) exposure generated a concentration-dependent shift of Hsp104p from the soluble to the aggregate fraction (Fig. 1C). The total amount of Hsp104p increased in response to As(III), as reported earlier (Sanchez et al., 1992) (Fig. 1C).

We next asked whether other xenobiotic metals also trigger protein aggregation in vivo. For this, we exposed yeast to As(III), Cd (CdCl₂) or Cr (CrO₃) using concentrations that inhibited growth to similar degrees (Jin et al., 2008), and quantified protein aggregation by counting the fraction of cells with Hsp104p–GFP foci. Although all agents triggered protein aggregation, they did so to various degrees (Fig. 1D); As(III) was the most potent (∼75% cells had aggregates), followed by Cd (∼50%) and Cr (∼15%). We conclude that As(III) is a potent trigger of protein aggregation in vivo.

As(III) is known to induce formation of cytosolic structures in mammalian cells called processing bodies (P-bodies) and stress granules (SG) that consist of mRNAs and various RNA-binding proteins (Balagopal and Parker, 2009). We asked whether As(III) triggers P-body and/or SG formation also in budding yeast, and whether Hsp104p is part of those structures. For this, we simultaneously monitored the distribution of Hsp104p–GFP and red fluorescent protein (RFP)-tagged Dcp2p (P-body marker) or cyan fluorescent protein (CFP)-tagged Pab1p (SG marker) in As(III)-exposed cells. Similar to mammalian cells, P-bodies and SGs were formed in As(III)-treated yeast cells (Fig. 1E). However, the majority of the Hsp104p-containing foci were separate from Dcp2p- and Pab1p-containing foci. Even though we cannot exclude that Hsp104p is part of P-bodies and/or SGs, our data suggest that Hsp104p-containing aggregates represent largely distinct structures in As(III)-exposed S. cerevisiae cells.

We next monitored Hsp104p–GFP distribution over time. Hsp104p–GFP was localized to foci after 1 hour of As(III) exposure, but after 3 hours, the majority of wild-type cells had an even cytosolic Hsp104p–GFP distribution like in untreated cells (Fig. 2A). In contrast, Hsp104p–GFP foci were still present at the 3-hour time-point in rpn4Δ cells lacking the transcriptional regulator of proteasomal gene expression Rpn4p (Fig. 2A). Western blot analysis confirmed that Hsp104p was present in comparable amounts and associated with aggregates similarly in wild-type and rpn4Δ cells (Fig. 1C). Together, these data suggest that wild-type cells can clear the cytosol from protein aggregates, whilst rpn4Δ cells are impaired in aggregate clearance.

To explore the role of the proteasome in aggregate clearance, we measured the activity of the 26S proteasome in cell extracts prepared from untreated and As(III)-exposed cells. As(III)-treated wild-type cells had ∼7- to 8-fold greater 26S activity (Fig. 2B) and a small increase in the amount of proteasomal components (Fig. 2C) compared to untreated control cells. Addition of As(III) to cell extracts from unexposed control cells did not affect 26S activity (Fig. 2B), indicating that As(III) does not stimulate 26S activity in vitro. To test whether transcriptional activation of proteasomal genes is required for enhanced 26S activity, we also performed activity measurements using the rpn4Δ mutant. Although rpn4Δ cells were capable of increasing proteasomal activity, this mutant had clearly lower 26S activity (Fig. 2B) and reduced amounts of proteasomal components (Fig. 2C) compared to the wild-type, both in the absence and presence of As(III).

Deletion of the RPN4 gene sensitized cells to As(III) (Fig. 2D) as previously reported (Haugen et al., 2004; Thorsen et al., 2009). The proteasome-defective mutants cim5-1/rpt1-1 [defective in one of the ATPases of the 19S regulatory particle (Ghislain et al., 1993)] and pre1-1 pre4-1 [defective in the β-type subunits of the catalytic 20S core (Hilt et al., 1993)] were also clearly more As(III) sensitive than the corresponding wild-type (Fig. 2D). Taken together, aggregate clearance involves the proteasomal system and enhanced proteasomal activity is important for As(III) tolerance.

One way As(III) could promote protein aggregation is by causing errors during translation. To test this, we used a yeast mutant carrying an ade1-14 UGA stop codon and scored red pigmentation as a read-out for mRNA mistranslation (Liu and Liebman, 1996). The drug paromomycin, that is known to cause mistranslation, suppressed the red pigmentation associated with the ade1-14 allele (Fig. 3A). As(III) did not cause any color change of this mutant, and addition of both As(III) and paromomycin did not produce a stronger suppression of the red pigmentation than paromomycin alone (Fig. 3A). Cells grown in liquid medium were more sensitive to As(III) and paromomycin together than to either agent on its own (Fig. 3B). In contrast to As(III), Cr suppressed red pigmentation of the ade1-14 strain (Holland et al., 2007) and caused a strong synergistic toxicity with paromomycin (Fig. 3C), consistent with the notion that Cr induces mistranslation (Holland et al., 2007). We conclude that As(III) is not an efficient inducer of protein mistranslation.

To better understand how As(III) impacts the proteome, we exposed wild-type cells to 1.5 mM As(III) and isolated aggregated proteins by sedimentation. The isolated proteins were separated by SDS-PAGE and identified by mass spectrometry. A total of 143 aggregation-sensitive proteins were unambiguously identified in As(III)-exposed cells (supplementary material Table S1). To characterize these proteins, we searched for functional categories [according to FunCat, Munich Information Center for Protein Sequences (MIPS)] that were significantly enriched (false discovery rate (FDR) <5%) in the aggregated protein set compared to the S. cerevisiae genome (supplementary material Table S1; Fig. 1F). Highly overrepresented functions in the As(III)-aggregated protein set were related to protein synthesis (>3-fold enrichment; Fischer's exact test, P<10⁻⁹), metabolism (1.9-fold; P = 5×10⁻⁹), proteins with binding function or cofactor requirement (1.9-fold; P = 6×10⁻⁶), protein folding and stabilization (>4-fold; P<3×10⁻⁴), and unfolded protein response (>7-fold; P<10⁻⁹). The latter groups contained various chaperones including Hsp104p and its co-chaperones Ydj1p and Ssa1p, ribosome-associated chaperones (Ssb2p, Zuo1p, Ssz1p, Egd1p, Egd2p), components of the CCT (chaperonin-containing T) complex (Cct1p/Tcp1p, Cct3p, Cct4p, Cct8p), and the Hsp70 co-chaperones Sis1p and Sse1p (supplementary material Table S1). These chaperones are probably associated and cosedimented with their misfolded and aggregated protein substrates, suggesting that As(III) promotes protein misfolding in vivo.

To explore whether As(III) causes misfolding and aggregation of proteins during synthesis/folding or of already folded native proteins, we monitored Hsp104p–GFP distribution in cells treated simultaneously with As(III) and the protein synthesis inhibitor cycloheximide (CHX). Interestingly, no Hsp104p–GFP foci were formed when cells were exposed to As(III) in the presence of CHX, neither at the lower (0.1 mM) nor at the higher (0.5 mM) As(III) concentration (Fig. 4A). We next exposed cells to 42°C for one hour, a condition that can lead to thermal unfolding and aggregation of native proteins (Singer and Lindquist, 1998). Importantly, Hsp104p–GFP was redistributed to foci in response to high temperature both in the absence and presence of CHX (Fig. 4A), indicating that the mechanisms by which As(III) and heat cause protein aggregation are distinct. These data suggest that proteins in the process of synthesis/folding are likely to be the prime targets of As(III)-induced aggregation. To substantiate this finding, exponentially growing yeast cells were pulsed for 5 minutes with [³⁵S]methionine to label newly synthesized proteins, in the absence or presence of As(III) where As(III) was added 15 minutes prior to [³⁵S]methionine labeling. Next, translation was stopped by adding CHX and aggregated proteins were isolated by sedimentation. While equal amounts of radioactivity was incorporated into newly synthesized proteins in untreated and As(III)-exposed cells (total lysate), a higher proportion of the radioactivity was present in the aggregate fraction of As(III)-treated cells compared to untreated cells (Fig. 4B). Quantification of the ³⁵S-containing material showed a 2.5-fold increase in aggregation of newly synthesized proteins during As(III) treatment compared to unexposed cells (Fig. 4B).

The presence of ribosome-associated chaperones in the As(III)-aggregated protein set (supplementary material Table S1; see above) was another indication that nascent proteins are vulnerable to this metalloid. Co-translational protein folding in yeast is assisted by two functionally interconnected ribosome-associated chaperone systems called SSB-RAC (stress 70 B-ribosome-associated complex) and NAC (nascent polypeptide-associated complex) while the yeast Hsp110 chaperone Sse1p functions as a nucleotide exchange factor for both ribosome-associated and cytosolic Hsp70s (Frydman, 2001; Koplin et al., 2010). Loss of SSB (ssb1Δ ssb2Δ) or Sse1p (sse1Δ) function leads to protein aggregation and reduced cell viability under conditions of protein folding stress, these effects being aggravated by additional deletion of NAC (egd1Δ egd2Δ btt1Δ) (Koplin et al., 2010). Cells lacking either NAC or SSE (sse1Δ) were clearly more As(III)-sensitive than wild-type cells, whilst NAC SSE cells showed an additive sensitivity (Fig. 4C), consistent with the notion that As(III) causes protein folding stress. Unexpectedly, growth of wild-type and the mutants lacking SSB or NAC SSB was similarly affected by As(III) (Fig. 4C), even though these mutants accumulate aggregated proteins and experience protein folding stress (Koplin et al., 2010). Interestingly, SSB and NAC SSB deficient cells have strongly reduced translational activity (Koplin et al., 2010); hence, diminished translational activity may protect cells from As(III) toxicity.

To investigate whether As(III) has an inhibitory effect on the chaperone/protein folding systems, we first induced aggregate-/Hsp104p–GFP foci formation by heat (42°C), then lowered the temperature to 30°C and monitored aggregate clearance in the absence or presence of As(III). To prevent de novo formation of As(III)-induced aggregates when shifting the temperature back to 30°C, we added CHX as indicated (Fig. 4D). Three hours after the shift to 30°C, cells had cleared the cytosol from heat-induced aggregates/Hsp104p–GFP foci. In contrast, cells were defective in aggregate clearance in the presence of As(III) (Fig. 4D). Since As(III) does not inhibit but rather enhances proteasomal activity (Fig. 2), these data indicate that As(III) may interfere with chaperone activity in vivo.

To further explore the impact of As(III) on protein folding, we monitored folding of firefly luciferase using several in vitro assays. As(III) inactivated native luciferase in a slow time-dependent process (Fig. 5A), indicating that the folded protein is not particularly susceptible to As(III)-mediated inactivation. In contrast, As(III) efficiently inhibited the spontaneous refolding of chemically denatured luciferase (Fig. 5B). When the refolding assay was performed in the presence of the Escherichia coli DnaK/DnaJ/GrpE chaperone system and ATP, the rate and yield of refolding in the absence of As(III) increased about 4-fold (Fig. 5C). The inhibitory effect of As(III) was somewhat more pronounced in the presence of chaperones than for spontaneous refolding (Fig. 5E); nevertheless, the DnaK/DnaJ/GrpE chaperone system was still capable of increasing the yield of refolding in the presence of metalloid (Fig. 5C). Finally, As(III) interfered with chaperone-mediated disaggregation and refolding of stable aggregates of heat-denatured luciferase (Fig. 5D); in this case the inhibitory efficacy was slightly lower than in the case of chemically denatured luciferase (Fig. 5E). Collectively, these data demonstrate that As(III) is a potent inhibitor of protein folding in vitro, acting primarily on unfolded polypeptides.

Since accumulation of aggregated proteins is potentially toxic to cells (Goldberg, 2003), we asked whether treatments that prevent or exacerbate protein aggregation also affect As(III) tolerance. Indeed, treating yeast cells with CHX not only diminished protein aggregation (Fig. 4A) but also resulted in improved As(III) tolerance (Fig. 4E). The opposite correlation also exists; cells defective in aggregate clearance (rpn4Δ) are sensitized to As(III) (Fig. 2A,D). Hence, accumulation of protein aggregates generated during As(III) exposure contribute to the toxicity of this metalloid.

Protein aggregates can inhibit the native folding of other proteins, leading to inactivation and aggregation (Gidalevitz et al., 2006). We therefore addressed the possibility that stable As(III)-induced luciferase aggregates may inhibit native refolding of urea-unfolded luciferase in the absence of free As(III) ions. Following removal of free As(III) by gel filtration, we found that substoichiometric amounts of both As(III)-free aggregates (Fig. 6A) and As(III)-induced aggregates (Fig. 6B) inhibited native luciferase refolding. Remarkably, the IC₅₀ of As(III)-induced aggregates was four times lower than that of As(III)-free aggregates (Fig. 6C). Thus, luciferase refolding was affected by substoichiometric concentrations of aggregated molecules, a phenomenon evocative of seed-induced proteinaceous aggregation, as previously observed in the case of many disease-causing protein aggregates (Aguzzi and Rajendran, 2009; Ben-Zvi and Goloubinoff, 2002). We conclude that As(III)-aggregated species have a strong inhibitory effect on de novo folding of proteins that have not encountered any metalloid during refolding. Hence, As(III)-aggregated species can act as seeds committing other, labile proteins to misfold and aggregate.

Arsenic is a major environmental pollutant and chronic exposure is associated with neurotoxicity, cardiovascular abnormalities, nephrotoxicity, and with cancers of the skin, bladder and lung. This metalloid has also a long history of usage in medical treatment. Yet, the molecular details of its biological actions are not completely understood (Singh et al., 2011; Wysocki and Tamás, 2010). Here, we describe a novel mechanism of As(III) toxicity that involves accumulation of protein aggregates due to impaired protein folding in vivo.

We demonstrated that As(III) is a potent inducer of protein aggregation in S. cerevisiae; aggregate formation is concentration-dependent and triggered by ‘free’ intracellular As(III) (Fig. 1). Moreover, our data indicated that As(III)-aggregated proteins can act as seeds committing other proteins to misfold and aggregate (Fig. 6). We identified 143 proteins that aggregated during As(III) exposure and various molecular chaperones were found to co-sediment with As(III)-induced protein aggregates (Fig. 1F; supplementary material Table S1). These chaperones, including Hsp104p and its co-chaperones Ydj1p and Ssa1p, as well as the cytosolic chaperones Sis1p and Sse1p, are often found stalled on aggregates as a result of failed attempts to convert the aggregates into non-aggregated functional species (Shorter and Lindquist, 2008). Such a stalling effect adds to the well-established proteasome stalling, which is often observed in cases of aging or misfolding diseases (Hinault et al., 2011; Kern et al., 2010; Macario and Conway de Macario, 2002).

Interestingly, heat and As(III) appear to affect the proteome in distinct ways; whilst heat may cause thermal unfolding and aggregation of native proteins, As(III) generates protein aggregation primarily by interfering with folding of nascent polypeptides (Fig. 4). As(III) is a potent inducer of Hsp104p and other heat shock proteins; however, whilst cells lacking Hsp104p are 100- to 1000-fold more sensitive to heat than wild-type cells, Hsp104p provides only a 2- to 3-fold survival advantage during exposure to lethal concentrations of As(III) (Sanchez et al., 1992). There are several ways to interpret these observations. From an aggregation point of view, aggregates produced by heat and As(III) may be fundamentally dissimilar. From a toxicity point of view, the main lesion produced by As(III) may not be accessible to Hsp104p (different compartment) or may not be a protein. From a chaperone point of view, As(III) may directly inhibit Hsp104p or other chaperones required for the restoration of folding homeostasis accounting for its poor performance in As(III) tolerance and its inability to dissolve aggregates when As(III) is present.

Several observations indicate that As(III) triggers protein aggregation primarily by targeting unfolded/nascent proteins. Firstly, the presence of ribosome-associated chaperones in the As(III)-aggregated protein set suggested that nascent proteins are vulnerable to this metalloid. Indeed, cells defective in folding of nascent polypeptides were As(III)-sensitive probably due to enhanced protein folding stress (Fig. 4). Secondly, in vitro data demonstrated that native firefly luciferase was moderately affected by As(III), whilst spontaneous and chaperone-mediated luciferase refolding were strongly inhibited (Fig. 5). The notion that unfolded/nascent proteins are the main target for As(III)-induced aggregation in vivo was further supported by the translational shut-down experiment, in which CHX prevented As(III)-induced protein aggregation, and by the fact that As(III)-exposed cells showed a 2.5-fold increase in aggregation of ³⁵S-labeled material (Fig. 4A,B). Thirdly, in vitro data revealed that As(III) had a greater impact on the refolding protein than on the chaperone system; DnaK/DnaJ/GrpE increased the yield of luciferase refolding 3.3-fold in the presence of 600 µM As(III) and 4.3-fold in the absence of metalloid. Hence, 600 µM As(III) resulted in ∼24% lower chaperone activity while the spontaneous refolding luciferase had lost ∼70% of its efficiency (compare final yields in Fig. 5B,C). Although As(III) acts primarily on nascent proteins, the in vitro experiment above showed that this metalloid also interferes directly with chaperone activity. This interference was confirmed in vivo since As(III) inhibited clearance of heat-induced protein aggregates (Fig. 4D). Thus, folding inhibition contributes to As(III) toxicity in two ways; by aggregate formation and by chaperone inhibition.

This work also adds new knowledge on how As(III) affects the actin and tubulin cytoskeleton. The CCT chaperonin complex is required for proper folding of actin and tubulin, as well as many other substrates (Frydman, 2001). As(III) was previously shown to disrupt the actin and tubulin cytoskeleton in vivo (Pan et al., 2010; Thorsen et al., 2009) and to inhibit actin folding by CCT in vitro (Pan et al., 2010). However, it was not clear how As(III) inhibited CCT since neither substrate binding nor ATPase activity of CCT was affected by this metalloid (Pan et al., 2010). Here, we found four out of eight CCT components as well as the CCT substrate β-tubulin (Tub2p) in the set of As(III)-aggregated proteins. Currently, we do not know whether As(III) causes aggregation of de novo synthesized CCT components or whether nascent chains of actin and tubulin aggregate during folding, thereby pulling CCT into the insoluble fraction. In any case, aggregation of CCT components will probably affect the folding process, increase the cellular load of unfolded actin and tubulin, and delay cytoskeleton recovery during As(III) exposure.

Collectively, our in vivo and in vitro data show that folding of unfolded/nascent proteins is affected by As(III), resulting in accumulation of aggregated proteins. The fact that cells defective in chaperones and protein folding systems are As(III) sensitive underscores their importance for protection against metalloid toxicity.

We showed here that a failure in aggregate clearance correlates with As(III) sensitivity whilst diminished aggregation enhanced tolerance (Figs 2, 4). Accumulation of toxic aggregates puts a higher demand on the proteasomal system. Indeed, S. cerevisiae responded to As(III) by enhancing 26S proteasome activity and this increase in proteasomal activity was shown to be important for tolerance (Fig. 2). While the amount of proteasomal components was slightly elevated, the 26S activity rose about 7-fold. Moreover, rpn4Δ cells were capable of increasing 26S activity in response to As(III), despite being defective in transcriptional activation of proteasomal gene expression (Haugen et al., 2004). Apparently, proteasomal activity and/or assembly is regulated post-transcriptionally during As(III) stress. Mammalian cells respond to As(III) by inducing expression of the AIRAP protein which then associates with the proteasome. Although the exact role of AIRAP in vivo is not fully understood, AIRAP clearly improves proteasome stability and activity in vitro (Stanhill et al., 2006). Whether yeast has a protein with similar function as mammalian AIRAP, is currently unknown.

Our data also suggested that diminished translational activity can protect cells from As(III) toxicity (Fig. 4). Consistent with this view, recent genome-wide studies revealed that mutations in genes encoding ribosomal protein and biogenesis factors result in greater As(III) tolerance (Dilda et al., 2008; Pan et al., 2010). Given that cells treated with the translational inhibitor CHX accumulated fewer aggregates and displayed improved tolerance (Fig. 4A,E), it is tempting to speculate that cells slow down protein synthesis during As(III) exposure to avoid protein aggregation and toxicity. The fact that As(III)-exposed cells strongly downregulate expression of genes with functions related to protein synthesis with a concomitant upregulation of chaperone and proteasomal gene expression (Haugen et al., 2004; Thorsen et al., 2007) supports this notion.

We showed that several metals can trigger protein aggregation in vivo (Fig. 1). Although As and Cd as well as lead (Pb) and mercury (Hg) efficiently inhibit protein folding in vitro (Ramadan et al., 2009; Sharma et al., 2008; Sharma et al., 2011), Cd and Cr appeared less potent than As(III) in triggering protein aggregation in vivo. It is possible that Cd and Cr do not enter cells as efficiently as As(III). Alternatively, the observed difference in potency could be a consequence of distinct modes of biological action; for example, Cr causes protein mistranslation (Holland et al., 2007) whilst As(III) does not (Fig. 3); As(III) interferes with CCT activity in vitro whilst Cd does not (Pan et al., 2010). The fact that Cd is very efficient in inhibiting protein folding in vitro but less potent in triggering protein aggregation in vivo, could potentially be explained by a lack of CCT inhibition given that CCT participates in the folding of as much as 10–15% of all cytosolic proteins in mammalian cells (Thulasiraman et al., 1999).

This study demonstrated that As(III) is a potent inhibitor of protein folding, and a powerful trigger of protein aggregation in vivo. This novel mechanism of As(III) toxicity does not contradict previously identified targets and modes of action such as induction of oxidative stress, impairment of DNA repair and disruption of enzyme function (Aposhian and Aposhian, 2006; Beyersmann and Hartwig, 2008; Singh et al., 2011; Wysocki and Tamás, 2010). Chronic arsenic exposure is associated with a variety of neurodegenerative disorders caused by aberrant protein folding including Parkinson's and Alzheimer's diseases (Gong and O'Bryant, 2010; Singh et al., 2011), and our data suggest that the deleterious effects of As(III) may result both from short-term, direct interactions of relatively high concentrations of As(III) ions with folding proteins, alongside long-term indirect interactions of low concentrations of As(III)-aggregated seeds that in turn commit other, labile proteins to misfold and aggregate. In mammals, these effects may become amplified with age when the efficiency of chaperone- and protease-dependent proteostasis declines (Hinault et al., 2011). Hence, the mechanism of action described here could contribute to the effects of arsenic in such protein folding diseases or other human disorders.

S. cerevisiae strains and plasmids used in this study are listed in supplementary material Table S2. The HSP104–GFP rpn4Δ strain was generated by crossing BY4741 HSP104-GFP-HIS3-MX6 with BY4742 rpn4Δ::KanMX4 followed by sporulation, tetrad dissection and selection according to Kaiser et al. (Kaiser et al., 1994). Yeast strains were grown on rich YPD medium (1% yeast extract, 2% peptone, 2% glucose) or on minimal SC (synthetic complete) medium (0.67% yeast nitrogen base) supplemented with auxotrophic requirements and 2% glucose as a carbon source. Growth assays were carried out in liquid medium or on plates as previously described (Warringer and Blomberg, 2003; Wysocki et al., 2004). Sodium arsenite (NaAsO₂), chromium trioxide (CrO₃), cadmium chloride (CdCl₂), paromomycin sulfate, and cycloheximide (CHX) (all from Sigma-Aldrich) were added to the cultures at the indicated concentrations. Mistranslation was scored using a qualitative plate assay as previously described (Holland et al., 2007).

Yeast growing exponentially in SC medium were either untreated or exposed to 0.5 mM As(III). After 3 hours of exposure, cells were harvested by centrifugation, washed twice in phosphate-buffered saline (PBS: 137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 2 mM KH₂PO₄, pH 7.4) and resuspended in 700 µl cold extraction buffer (10% glycerol, 10 mM Tris-HCl pH 7.5, 1 mM EDTA, 0.5 mM DTT, 5 mM MgCl₂, pH 7.4). Cells were disrupted with glass beads and vortex (3×20 seconds with 2 minutes on ice in between) followed by centrifugation. The protein concentration of the resulting supernatant (whole-cell extract) was quantified according to (Bradford, 1976). To measure proteasomal activity, whole-cell extracts (20 µg of total protein) were incubated with 200 µM of the fluorogenic proteasome substrate suc-LLVY-AMC (Biomol International) for one hour at 30°C, and fluorescence intensity of the free AMC formed was monitored in a Synergy2 Multi-Mode Microplate Reader (BioTek). Proteasomal activity was also measured in the presence of 20 µM MG132 proteasome inhibitor (Biomol International), and 7-amino-4-methylcoumarin was used for obtaining the standard curve (Biomol International).

Yeast cells expressing GFP, CFP and RFP fusion proteins of interest were grown to mid-log phase in SC medium containing appropriate amino acid requirements. Cells were washed twice with water or PBS and the GFP, CFP or RFP signals were observed in living cells using a Leica DM RXA (Leica Microsystems) or Zeiss Axiovert 200M (Carl Zeiss MicroImaging) fluorescence microscope. The microscopes were equipped with 100× HCX PL Fluotar 1.30 (Leica) and Plan-Apochromat 1.40 (Zeiss) objectives and appropriate fluorescence light filter sets. Images were captured with digital camera [Hamatasu C4742-95 (Hamamatasu Photonics) or AxioCamMR3 (Zeiss)] and QFluoro (Leica) or AxioVision (Zeiss) software, and processed with Photoshop CS (Adobe Systems). Where indicated, cells were treated with As(III) or heat (42°C) in the absence or presence of 0.1 mg/ml cycloheximide. To compare protein aggregation by different metals, cells were exposed to 0.4 mM As(III), 0.4 mM Cr, or 5 µM Cd for 1 hour and Hsp104p–GFP localization monitored as above. To quantify protein aggregation, the fraction of cells with Hsp104p–GFP foci was determined by visual inspection of 350–900 cells.

Wild-type (BY4742) and rpn4Δ strains were grown to exponential phase (A₆₀₀∼0.6) in YPD medium and equivalent cell numbers (10 A₆₀₀ units) were used to analyze protein aggregation as described previously (Rand and Grant, 2006). Briefly, cells were disrupted in lysis buffer [50 mM potassium phosphate buffer, pH 7, 1 mM EDTA, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride and Complete Mini protease inhibitor cocktail (Roche)], and the membrane and aggregated proteins were isolated by centrifugation at 15.000 g for 20 minutes. Membrane proteins were removed by washing twice with 320 µl lysis buffer and 80 µl of 10% Igepal CA 630 (NP-40; Sigma-Aldrich), centrifuging at 15,000 g for 20 minutes each time, and the final aggregated protein extract was resuspended in 100 µl of lysis buffer. Soluble and aggregated protein extracts were analyzed by 12% reducing SDS-PAGE and western blotting using an anti-Hsp104p antibody (kindly provided by Mick Tuite, University of Kent). Bound antibody was visualized by chemiluminescence (Pierce) after incubation of the blot in donkey anti-rabbit immunoglobulin–horseradish-peroxidase conjugate (Amersham Pharmacia Biotech). To detect proteasomal components by western blot, primary antibodies targeting Rpt1p and Rpt3p (diluted 1:5000; Abcam) were used whilst an antibody targeting the kinase Hog1p (diluted 1:2000; Santa Cruz Biotechnology) was used as loading control. Secondary donkey anti-goat and donkey anti-rabbit antibodies (diluted 1:15,000; LI-COR Biosciences) and the Odyssey® Infrared Imaging System (LI-COR Biosciences) were used for visualization.

For protein identification, yeast cells were exposed to 1.5 mM As(III) for one hour, aggregated proteins were isolated as above, and the SDS-PAGE gels were stained using colloidal Coomassie blue (Sigma-Aldrich). Proteins were excised, trypsin digested, and identified using liquid chromatography and mass spectrometry (LC-MS) in the Biomolecular Analysis Facility (Faculty of Life Sciences, University of Manchester, UK). Proteins were identified using the Mascot mass fingerprinting program (www.matrixscience.com) to search the NCBInr and Swissprot databases. As a control, we also isolated and identified proteins that aggregated in the absence of As(III). Proteins found to aggregate both under control conditions and As(III) exposure were considered as generally aggregation prone and were not included in the analyses below.

Newly synthesized proteins were labeled with [³⁵S]methionine as previously described (Koplin et al., 2010). Cells were first starved for 1 hour in methionine-free SC medium, then labeled with 20 µCi/ml [³⁵S]methionine for 5 min, quickly chilled and treated with 300 µg/ml CHX to stop protein translation. Cells were disrupted with glass beads and vortex (3×1 minute) in lysis buffer (20 mM sodium phosphate, pH 6.8, 10 mM DTT, 1 mM EDTA, 0.1% Tween, 1 mM PMSF, protease inhibitor cocktail and 3 mg/ml zymolyase). Supernatants were adjusted to equal protein concentrations and aggregated proteins were pelleted at 16,000 g for 20 minutes at 4°C. Proteins were separated by SDS-PAGE (10%) and visualized by autoradiography (Molecular Imager FX, Bio-Rad).

Enriched functional categories were set with FDR <5%. Enrichments were calculated by taking the ratio of the relative protein content in the As(III)-aggregated set to the relative content in the genome for each category. P-values were calculated by Fisher's exact test.

Escherichia coli DnaK was expressed and purified as described previously (Feifel et al., 1996; Sharma et al., 2008), whilst DnaJ and GrpE (Schönfeld et al., 1995a; Schönfeld et al., 1995b) were kindly provided by Dr H. J. Schönfeld (F. Hoffmann-La Roche, Basel, Switzerland). Photinus pyralis luciferase was obtained from Sigma-Aldrich. Protein concentrations were determined and the proteins stored as described previously (Sharma et al., 2008).

To prepare chemically denatured luciferase, 1 mg lyophilized luciferase was dissolved in 1 ml of 50 mM Tris acetate, 50 mM potassium perchlorate, 15 mM magnesium acetate, pH 7.5, and precipitated by adding five volumes of acetone (−20°C, 30 min). After centrifugation (10 min, 10,000 g, 4°C) the pellet was re-dissolved in 1 ml of denaturing buffer [6 M guanidine HCl, 100 mM Tris acetate, 5 mM Tris(2-carboxyethyl)phosphine (TCEP, a non-thiol reducing agent), 50 mM potassium perchlorate, 15 mM magnesium acetate, pH 7.5].

To prepare stable aggregates of heat-denatured luciferase, 700 nM luciferase was incubated for 5 min at 45°C in 50 mM Tris acetate, 50 mM potassium perchlorate, 15 mM magnesium acetate, pH 7.5 (Ben-Zvi et al., 2004; De Los Rios et al., 2006). The residual activity after heat exposure was ∼2% of the initial. Luciferase activity was measured as described previously (Bischofberger et al., 2003) using a Victor Light 1420 Luminescence Counter (Perkin-Elmer).

To study the effect of pre-formed aggregates on luciferase refolding, chemically denatured luciferase was allowed to refold spontaneously at 25°C (final concentration of luciferase 1 µM) in refolding buffer, without or with 1 mM As(III) (NaAsO₂). About 20% and 5% of luciferase was refolded at 25°C in 2 hours, without and with 1 mM As(III), respectively. These samples containing 80% and 95% aggregated luciferase were separated from free As(III) using PD SpinTrap G-25 columns (GE Healthcare Life Sciences), and thereafter used to study their effect on the refolding of luciferase.

We thank Matthias Peter (ETH, Zurich), Dieter H. Wolf (University of Stuttgart), Elizabeth Craig (University of Wisconsin-Madison), Nicolas Talarek, Claudio de Virgilio (University of Fribourg) and Elke Deuerling (University of Konstanz) for providing strains and plasmids, Mick Tuite (University of Kent) for providing the anti-Hsp104p antibody, Hans-Joachim Schönfeld (F. Hoffmann-La Roche, Basel) for providing DnaJ and GrpE, and Thomas Nyström (University of Gothenburg) for critical reading of the manuscript.