|Home | About | Journals | Submit | Contact Us | Français|
Self-perpetuating protein aggregates transmit prion diseases in mammals and heritable traits in yeast. De novo prion formation can be induced by transient overproduction of the corresponding prion-forming protein or its prion domain. Here, we demonstrate that the yeast prion protein Sup35 interacts with various proteins of the actin cortical cytoskeleton that are involved in endocytosis. Sup35-derived aggregates, generated in the process of prion induction, are associated with the components of the endocytic/vacuolar pathway. Mutational alterations of the cortical actin cytoskeleton decrease aggregation of overproduced Sup35 and de novo prion induction and increase prion-related toxicity in yeast. Deletion of the gene coding for the actin assembly protein Sla2 is lethal in cells containing the prion isoforms of both Sup35 and Rnq1 proteins simultaneously. Our data are consistent with a model in which cytoskeletal structures provide a scaffold for generation of large aggregates, resembling mammalian aggresomes. These aggregates promote prion formation. Moreover, it appears that the actin cytoskeleton also plays a certain role in counteracting the toxicity of the overproduced potentially aggregating proteins.
Prions are protein isoforms that cause transmissible neurodegenerative diseases in mammals (for review, see reference 50) and control heritable traits in fungi (for review, see references 10 to 12). Most known prions are self-perpetuating amyloid-like ordered fibrous protein aggregates which propagate the prion state by immobilizing the soluble protein molecules of the same amino acid sequence. Saccharomyces cerevisiae prion [PSI+] is an aggregate of the translation termination factor Sup35. The prion domain of Sup35 is rich in glutamine (Q) and asparagine (N) residues, resembling poly-Q proteins, such as huntingtin, which is involved in Huntington's disease (for review, see reference 53). While recent data shed light on the major steps of propagation of the preexisting [PSI+] aggregates in yeast cells (for review, see references 12 and 47), the mechanism of initial prion formation from nonprion protein remains a mystery. It has been shown that de novo formation of the [PSI+] prion is induced by transient overproduction of the Sup35 protein or its prion domain (14, 19). This process is usually efficient only in cells containing other QN-rich protein aggregates, such as [PIN+], a prion form of Rnq1 (20, 22). Likewise, preexisting QN-rich prions promote aggregation and aggregation-related toxicity of heterologous poly-Q proteins expressed in yeast cells (33, 38). Possibly, preexisting QN-rich aggregates either provide initial nucleation centers for aggregation of other QN-rich proteins or sequester unknown antiaggregation factors.
Assembly of amyloid fibers resembles the assembly of cytoskeletal structures such as actin filaments. The QN-rich domain of Sup35 was shown to interact with the actin assembly protein Sla1 in the two-hybrid assay (4). Deletion of SLA1 decreases de novo induction of [PSI+] by excess Sup35. Sla1 is a component of the multiprotein machinery that regulates the assembly and disassembly of actin filaments involved in the formation of endocytic vesicles (for review, see reference 24). Another component of this machinery, Sla2, is a yeast homolog of the mammalian huntingtin-interacting protein Hip1 (29). Aggregates of the poly-Q fragment of huntingtin, generated in yeast cells, also interact with the vesicle assembly proteins and block endocytosis, possibly by sequestering components of the vesicle assembly machinery (34). Long-term disruption of the actin cytoskeleton by latrunculin A results in a high frequency of [PSI+] loss in the surviving cells (5). These data point to possible links between prions and the actin cytoskeleton in yeast.
Here, we have systematically studied interactions between the yeast prion protein Sup35 and components of the actin cortical cytoskeleton. Our results uncover a crucial role of the yeast actin cytoskeleton in the initial generation of prion aggregates by overproduced Sup35.
Two-hybrid experiments employed the [psi−] strain PJ69-4A, containing the PGAL-ADE2 and PGAL-HIS3 reporter constructs (27). Strains GT907-6A (MATa ade1-14 his3 ura3 leu2 SLA1-GFP HIS3 [PSI+ PIN+]) and GT908-6A (MATa ade1-14 his3 ura3 leu2 SLA2-GFP HIS3 [PSI+ PIN+]) were constructed by mating GT81-1D (see below) to, respectively, YBL007C or YNL243W green fluorescent protein (GFP)-fusion strains from the Invitrogen collection, followed by sporulating and dissecting the resulting diploids. All other yeast strains originated from the homozygous diploid GT81 (MATα/MATa ade1-14/ade1-14 his3/his3 leu2/leu2 lys2/lys2 trp1/trp1 ura3/ura3 [PSI+ PIN+]) and its isogenic haploid derivatives GT81-1C (MATa) and GT81-1D (MATα) (16). Gene deletions were generated by direct PCR-mediated transplacement (32). In each case, the whole open reading frame (ORF) was deleted and replaced with either the Saccharomyces cerevisiae HIS3 gene (in the case of sla1Δ [see reference 4]) or the Schizosaccharomyces pombe homolog of the S. cerevisiae HIS3 gene (in all other cases, see reference 32). Transplacement of the wild-type ACT1 allele with the act1-R177A allele was performed via the integration/excision procedure, using the plasmid pKWF46-R177A (see below), and verified by sequencing of the PCR-amplified ACT1 fragment. The [psi− PIN+] derivatives of the [PSI+ PIN+] strains were obtained by curing yeast cells of [PSI+] in result of transient overproduction of Hsp104 (17). The [psi− pin−] strains were generated from the [PSI+ PIN+] strains via curing yeast strains of both [PSI+] and [PIN+] by either GuHCl treatment or transient expression of the dominant negative Hsp104 derivative, Hsp104-KT (17, 49). In all cases, the presence or absence of [PIN+] was monitored in the functional assay, based on [PSI+] induction by excess Sup35 (as described in reference 17) and, in some cases, also verified by the differential centrifugation assay for aggregated Rnq1 protein (as described in reference 33).
The URA3 plasmid pmCUPNMsGFP, bearing the chimeric SUP35NM-GFP construct under the control of the copper-inducible PCUP1 promoter, and control plasmid pmCUPsGFP with the GFP construct were kindly provided by S. Lindquist (43). The URA3-based plasmids pRS316GAL-SUP35N and CEN-GAL-SUP35 and HIS3-based plasmid pLA1-SUP35N bearing, respectively, SUP35N, SUP35, and SUP35N constructs under control of the galactose-inducible PGAL promoter, as well as the respective empty control plasmids pRS316GAL and pLA1, were described earlier (reference 49 and references therein). The HIS3 plasmid pLA1-Sup35ΔS contained the NM region of SUP35 and about half of the C region, cut at the SalI recognition site; this construct, designated SUP35ΔS, was expressed from the PGAL promoter. The URA3-based plasmid pRS316GAL-SUP35NM-GFP was constructed by K. Gokhale in Y. Chernoff's lab via isolation of the BamHI-SacI fragment of pmCUPNMsGFP, containing the SUP35NM-GFP fusion gene, and insertion of it into pRS316GAL next to the PGAL promoter. The URA3 plasmid CEN-GAL-SUP35-RFP, containing the SUP35ΔS fragment fused to the red fluorescent protein (RFP)-coding gene and placed under the PGAL promoter, was constructed by K. Gokhale in Y. Chernoff's lab. The RFP-coding gene of Discosoma sp. was from the plasmid pDsRed1-N1, purchased from BD Biosciences (formerly Clontech). The pEMBL-yex-based series of multicopy plasmids, bearing the URA3 gene, partially defective LEU2-d allele, and SUP35 gene or its SUP35N region under the endogenous PSUP35 promoter, was described earlier (reference 19 and references therein). The TRP1 centromeric plasmid pFL39-CUP-SUP35C, constructed in this study, bears the SUP35C region under the PCUP1 promoter. The plasmid pmCUP1-SUP35-HA, expressing the hemagglutinin (HA)-tagged Sup35 protein from the PCUP1 promoter and described earlier (1), was used for the immunoprecipitation experiments. Basic two-hybrid vectors and plasmids bearing the complete SUP35 gene or SUP35N fragment fused to either the N terminus of the GAL4 DNA-binding domain (GAL4DBD) or to the C terminus of the GAL4 activation domain (GAL4ACT), as well as the plasmid bearing SUP35N fused to the C terminus of GAL4DBD and plasmid bearing the C-terminal fragment of SLA1 fused to the C terminus of GAL4ACT were described earlier (reference 4 and references therein). The plasmids bearing the complete ORFs of SLA2, END3, ARP2, or ARP3, fused each to GAL4ACT, were from the yeast two-hybrid tester strains of S. Fields' collection (48). All these plasmids were isolated from yeast, and both the presence and size of the inserts were confirmed by digestion with restriction endonucleases. The strain from the same collection, supposedly containing the plasmid with the complete SLA1 ORF, in fact contained a rearranged plasmid. Apparently, the fusion of GAL4ACT with the complete SLA1 gene cannot be maintained in the overproducing construct. Plasmid bearing the complete ACT1 ORF fused to the N terminus of GAL4ACT and the integrative plasmid pKWF46-R177A (7), bearing the act1-R177A (act1-155) mutant allele with the URA3 gene inserted into a flanking sequence, were kindly provided by D. Drubin. Episomal plasmid pWA1-SLA2-HA, coding for the HA-tagged Sla2 protein, was kindly provided by L. Hicke.
Yeast cultures were grown at 30°C, except for the experiments that involved the temperature-sensitive sla2Δ and arp3Δ derivatives, grown at 25°C. Standard yeast media and standard procedures for yeast cultivation, phenotypic and genetic analyses, transformation, sporulation, and dissection were used (44). In all cases where the carbon source is not specifically indicated, 2% glucose (Glu) was used. The solid synthetic medium containing 2% galactose (Gal) or liquid synthetic medium with 2% galactose and 2% raffinose (Gal+Raf) instead of glucose was used to induce the PGAL promoter. CuSO4 was added to synthetic medium at a concentration of 100 to 150 μM (solid medium) or 150 μM (liquid medium) in order to induce the PCUP1 promoter. Liquid cultures were grown with at least a 1/5 liquid:flask volumetric ratio in a shaking incubator at 200 to 250 rpm.
Individual transformants, bearing either the constructs that overproduce Sup35 and its derivatives or respective control plasmids were velveteen replica plated onto Gal medium selective for the plasmid(s) in order to induce the PGAL promoter or onto Glu medium with 150 μM CuSO4 in order to induce the PCUP1 promoter. After 3 to 4 days of incubation, plates were velveteen replica plated onto −Ade medium with glucose and without copper to score for [PSI+] induction. [PSI+] was detected by its ability to suppress the ade1-14 (UGA) mutant allele (see reference 17). At least eight independent transformants were tested for each strain-plasmid combination; the majority or all of them did show one and the same result in each case.
Yeast cultures, bearing plasmids with the SUP35-derived constructs under the PCUP1 promoter, were pregrown in synthetic liquid medium selective for the plasmid(s) and inoculated into medium of the same composition containing 150 μM CuSO4 at the starting concentration of 106 cells/ml. Aliquots were taken after certain periods of time and plated onto synthetic medium selective for the plasmid(s) and without addition of CuSO4. Grown colonies were velveteen replica plated onto yeast extract-peptone-dextrose (YPD) and −Ade media. The presence of [PSI+] was detected by both color on YPD (light pink rather than red) and growth on −Ade, which are indicative of ade1-14 (UGA) suppression (17). Mosaic colonies, containing both [PSI+] and [psi−] sectors, usually constituted a small fraction and were counted as half [PSI+] and half [psi−]. In both qualitative and quantitative experiments, a sample of the Ade+ colonies was tested for curability by GuHCl in order to confirm epigenetic inheritance of [PSI+] (17).
Standard procedures were used for isolation of DNA, restriction digestion, ligation, and bacterial transformation (42). Enzymes were purchased from New England Biolabs and Invitrogen. The thermal cycler was from Ericomp, Inc. Western analysis, chemiluminescent labeling, and hybridization were performed according to GE Healthcare protocols. Protein isolation and differential centrifugation were performed as described previously (17). Affinity chromatography analysis of the yeast extracts for proteins binding Sup35NM-His and coimmunoprecipitation analysis of yeast proteins were performed in accordance with our previously described protocols (1). Densitometry was performed on a Gel Logic 200 imaging system using the Kodak 1D program. Generation of antibodies specific to Sup35NM (49) and Ade2 (1) was described previously. Antibodies specific to HA and actin were purchased from Sigma. Antibodies specific to Rnq1 and Sla2 were gifts of S. Lindquist and D. Drubin, respectively.
GFP and RFP detection in the live exponentially growing yeast cells (17, 43) and staining of yeast vacuolar membranes with the lipophilic dye FM4-64 (34) were performed as described previously. For actin staining, cells were fixed by adding formaldehyde directly to the culture, to a final concentration of 3.7%, followed by incubation for 15 min at 30°C and staining with 0.165 mM rhodamine-phalloidin (Molecular Probes, Eugene, OR) in phosphate-buffered saline for 1 h on ice, followed by three washes in phosphate-buffered saline (30). Preparation for imaging was accomplished by placing an aliquot of fixed and stained cells onto a glass slide and sealing the coverslip to the slide with clear nail polish. Images were taken on a BX41 microscope (Olympus) with the Endow GFP bandpass emission (green) or tetramethyl rhodamine isocyanate (rhodamine/Dil; red) filters. The LSM510 confocal laser scanning microscope (Carl Zeiss Inc., New York, N.Y.) was used for more detailed analysis of aggregates and colocalization experiments. Image analysis was conducted using the Zeiss LSM image browser (Carl Zeiss, Jena, Germany) as described previously (5, 49). An argon laser with an excitation wavelength of 488 nm and pinhole size of 4.66 airy units was used for GFP detection, and a helium-neon laser with an excitation wavelength of 543 nm and pinhole size of 4.13 airy units was used for RFP, actin, and vacuolar membrane staining detection.
In the experiments that assessed effects of aggregates on cell viability and [PSI+] formation, live cells expressing GFP from the PCUP1-promoted constructs were sorted after 48 h of induction in the presence of CuSO4. A Precision Micromanipulator from Micro Video Instruments, Inc., attached to the Olympus BX41 microscope was used for cell sorting. The resulting cells were incubated on YPD medium to form colonies. Those not capable of producing colonies after 6 to 7 days of incubation were checked under the microscope. Typically, it was observed that for nonviable cells, the growth arrest occurred at the first or second cell division after sorting. Colonies that grew were checked for the presence or absence of [PSI+] by color and growth assays as described above and in reference 17. Those which turned [PSI+] positive were streaked on YPD, and about 100 subcolonies of each colony, produced by this procedure, were rechecked for the presence of [PSI+]. This enabled us to distinguish between the homogenous [PSI+] colonies and “mosaic” colonies containing both [PSI+] and [psi−] cells.
The C-terminal region of the actin assembly protein Sla1 was previously shown (4) and is now confirmed (Fig. (Fig.1A)1A) to interact with the N-terminal prion-forming domain of Sup35 (Sup35N) in a two-hybrid assay. Sla1 is a component of the actin assembly machinery that is involved in endocytosis. This machinery includes a number of other proteins, among them Sla2, End3, Arp2, and Arp3 (for review, see reference 24). We have now shown that complete Sla2, End3, Arp2, and Arp3 proteins also exhibit two-hybrid interactions with both complete Sup35 (data not shown) and Sup35N (Fig. (Fig.1A).1A). Interactions of Sla2 with the fragment of Sup35 including the prion domain were also confirmed biochemically, as endogenous Sla2 protein was pulled down from yeast extracts by resin containing the immobilized His-tagged N and M (middle) region of Sup35 (Sup35NM-His) (Fig. (Fig.1B1B).
As several actin assembly proteins interact with Sup35, we checked whether actin interacts with Sup35. No two-hybrid interaction between actin and Sup35 was detected (Fig. (Fig.1A).1A). However, actin was pulled down together with the HA-tagged Sup35 protein from yeast extracts (Fig. (Fig.1C).1C). No actin was detected in the immunoprecipitates of the other yeast protein, Ade2, indicating that actin is not binding nonspecifically to just any yeast protein in this kind of assay. The difference between the results of the two-hybrid assay and pull down assay could be explained if Sup35 either interacted specifically with filamentous actin or pulled actin down as part of a larger cytoplasmic complex containing actin assembly proteins, such as Sla2, that directly interact with Sup35. The artificial Gal4-actin fusion proteins are most likely unable to assemble properly into actin filaments and large actin-containing complexes. Besides, the two-hybrid interaction occurs in the nucleus, where some components of the cytoplasmic actin complexes are probably missing.
The Sup35 fragments, including the Sup35NM region, fused to fluorescent proteins such as green fluorescent protein (Sup35NM-GFP) or red fluorescent protein (Sup35-RFP) provide a useful tool for the detection of prion aggregation in yeast and for monitoring the process of prion formation. In cells containing no prions, Sup35NM-GFP (17, 43) and Sup35-RFP (this study) are usually diffusely distributed throughout the cytoplasm, while in [PSI+] cells, they generate cytologically detectable aggregates. Overproduction of Sup35NM-GFP in [psi−] yeast strains containing [PIN+], a prion isoform of the QN-rich protein Rnq1, results in both accumulation of huge cytologically detectable aggregates of Sup35NM-GFP and de novo formation of [PSI+] in a fraction of cells. The aggregation behavior of Sup35NM-GFP resembles that of the complete Sup35 protein under similar conditions, as proven by direct visualization of untagged Sup35 protein via secondary immunofluorescence (49, 52). In our hands, accumulation of aggregatesand [PSI+] formation were most efficient when Sup35NM-GFP was expressed from the copper-inducible PCUP1 promoter induced by 100 to 150 μM CuSO4, compared to the galactose-inducible PGAL promoter. As observed previously (52) and confirmed by us (Fig. (Fig.2A),2A), the Sup35NM-GFP aggregates induced in the [psi− PIN+] cultures can take different shapes, such as “worm-like” or “ring-like” structures and “clump-like” and “dot-like” structures. More detailed analysis of the “worms” and “rings” by confocal microscopy, which allows slicing and three-dimensional imaging of the sample, revealed that “ring-like” structures can be located either at the cell periphery (peripheral “worms” and “rings”) or inside the cell (internal “rings”) (Fig. (Fig.2A2A).
Similar to Sup35NM-GFP, overexpression of the Sup35-RFP fusion construct from the galactose-inducible PGAL promoter led to formation of both clump-dot and worm-ring aggregates (Fig. (Fig.2B)2B) in up to 21% of cells of the [psi− PIN+] culture and induced formation of [PSI+] in the [psi− PIN+] culture at a level comparable to that of PGAL-induced Sup35NM-GFP (data not shown).
More than 70% of cells with cytologically detectable aggregates of Sup35NM-GFP, isolated from the [psi− PIN+] culture by micromanipulation, were unable to form colonies, indicating that aggregates are toxic to the yeast cells (see additional detail below). Of about 30% of cells with aggregates that remained viable, 70% gave rise to the complete [PSI+] or mosaic [PSI+]/[psi−] colonies, while only about 10% of viable cells with diffuse fluorescence did so (Fig. (Fig.2C).2C). One should note that after PCUP1 transcription is turned down, the levels of the Sup35NM-GFP protein are likely to remain high for a while, until the excess Sup35NM-GFP is degraded. Therefore, it was predictable that a subset of cells showing no visible aggregates at the moment of micromanipulation could generate aggregates after shifting to the noninducing medium. It is possible that cells with diffused fluorescence producing [PSI+] or mosaic [PSI+]/[psi−] colonies actually belonged to this exceptional fraction.
Both the correlation of [PSI+] induction with the cytologically detectable Sup35NM-GFP aggregation and the decreased viability of cells containing ring-like structures agree with previous observations by Zhou et al. (52). However, in our hands cells with clumps or dots were also frequently inviable. This difference with the previously published data is possibly due to either a difference in the strain genotypic backgrounds or relatively small number of cells with clumps and dots analyzed individually in the Zhou et al. work (52), or both.
Interestingly, when Sup35NM-GFP or Sup35-RFP was overproduced in the cells containing the preexisting [PSI+] prion, only clumps or dots were usually found. This confirmed previous observations (52) and suggests that worms and rings represent intermediates involved in the early stages of [PSI+] formation rather than mature prion aggregates.
As Sup35N interacts with some components of the actin cytoskeleton involved in endocytosis (see above), we have checked whether Sup35NM-GFP aggregates colocalize with the cytologically detectable structures related to the endocytic pathway. In yeast, endocytic vesicles are presumed to be formed at or in close proximity to cortical actin patches and transported to the vacuole along actin cables (for review, see reference 24). While no colocalization between Sup35NM-GFP aggregates and actin patches was detected in the [PSI+] cultures, some clumps and the vast majority of the peripheral worms and rings, generated in the [psi− PIN+] cultures, overlapped with at least a fraction of actin patches (Fig. (Fig.2C2C and Table Table1).1). Overlap of the peripheral rings with actin patches could in principle be a coincidence due to the fact that rings occupy a large percentage of the cell cortex, but such an explanation is unlikely in the case of clumps. Previously, Zhou et al. (52) were unable to prove colocalization between actin patches and Sup35NM-GFP rings. However, these authors did not employ confocal microscopy and were therefore unable to distinguish between peripheral and internal structures. Such a distinction was crucial for our colocalization experiments. Interestingly, essentially all internal Sup35NM-GFP rings were assembled around the vacuole (Fig. (Fig.2E2E and Table Table11).
Sla1 and Sla2 proteins were previously shown to exhibit a punctate localization in the yeast cells, and some but not all of Sla1 and Sla2 dots coincide with actin patches (2, 3). Sla1 and Sla2 proteins tagged with GFP exhibited a punctate pattern in our hands (Fig. (Fig.2F)2F) that was similar to one published previously for native proteins with small epitope tags. This confirms that distribution of the GFP-tagged derivatives reflects that of the native proteins and does not represent an artifact of the GFP fusion. When Sup35-RFP was overexpressed in the [PSI+] or [psi− PIN+] cultures bearing either Sla1-GFP or Sla2-GFP, an increased concentration of Sla1-GFP or Sla2-GFP dots was observed in the areas corresponding to the Sup35-RFP aggregates (Fig. 2G and H and Table Table1),1), indicating that a significant fraction of the Sla1/2-containing structures is associated with aggregated Sup35. Taken together, fluorescence microscopy data clearly demonstrate colocalization of the Sup35-derived aggregates with at least some components of the actin cortical cytoskeleton involved in the endocytic pathway.
Next, we checked whether mutations in the genes coding for components of the cortical actin cytoskeleton influence [PSI+]. As the commercially available yeast deletion strains lack markers for monitoring [PSI+] formation and maintenance, we have generated deletions of the genes ARP2, ARP3, END3, SLA1, and SLA2 by direct PCR-based transplacement in the strain GT81-1C, which contains such a marker (see Materials and Methods). The GT81-1C derivative with the point mutation act1-R177A in the actin (ACT1) gene has also been generated by the integration/excision procedure. It turned out that arp2Δ is lethal in the genotypic background of our strains independently of the presence or absence of the [PSI+] and [PIN+] prions, while arp3Δ leads to gross defects of growth, mating, and cell morphology and is therefore essentially impossible to work with (data not shown). sla2Δ was lethal in the [PSI+ PIN+] background (see below), while sla1Δ, end3Δ, and act1-R177A were viable both with and without prions, so that [PSI+ PIN+], [psi− PIN+], and [psi− pin−] strains with these alterations were generated. sla1Δ, end3Δ, and act1-R177A are all known to decrease endocytosis to various extents and to cause detectable defects in accumulation and distribution of cortical actin patches (7, 8, 24). We have confirmed that the act1-R177A mutant essentially lacked cortical actin patches detectable by the rhodamine-phalloidin staining. Although this could be in part due to a defect of the actin-rhodamine binding associated with this alteration, a defect of the cortical actin cytoskeleton in this mutant has also been confirmed previously by other means (7). The end3Δ cells contained a decreased number of cortical actin patches, in comparison to the wild-type cells (data not shown). This confirms previous observations (8). All strains carried the reporter UGA allele ade1-14, so that UGA readthrough, caused by [PSI+], could be detected by either a lighter color on complete (YPD) medium or growth on medium lacking adenine (−Ade) (for review, see reference 17). end3Δ and especially act1-R177A (but not sla1Δ) decreased growth of all strains on the complete medium and of the [PSI+] strains on the −Ade medium. It is likely, although not proven, that decreased growth of these strains on −Ade was a consequence of the general growth defect rather than the specific effect on the [PSI+]-mediated nonsense suppression. Growth on −Ade was restored by mating of these mutant strains to the wild-type [psi−] strain, indicating that maintenance of the preexisting [PSI+] prion was not affected (data not shown). Among 500 to 700 colonies analyzed per each [PSI+] haploid culture, either none (wild type, sla1Δ, and act1-R177A) or only 0.4% (end3Δ) have become [psi−], confirming that mitotic stability of the preexisting [PSI+] prion was not significantly affected by cytoskeletal alterations.
Deletion of another cytoskeletal gene, SLA2, turned out to be impossible to generate directly in the haploid [PSI+] strain. When a diploid [PSI+ PIN+] strain heterozygous for the sla2Δ::HIS3 disruption was generated, sporulated, and dissected, no sla2Δ::HIS3 spores were recovered among 31 tetrads (Table (Table2).2). The URA3 plasmid with the SLA2 gene rescued the sla2Δ::HIS3 spores, but they were unable to grow on the 5-fluoroorotic acid medium that counterselects against the URA3 plasmid, showing that sla2Δ is lethal in the absence of the wild-type SLA2 allele.
However, when the heterozygous SLA2/sla2Δ::HIS3 diploid was cured of [PSI+] before sporulation (see Materials and Methods for technical details), about half of the haploid sla2Δ::HIS3 spores formed colonies (Table (Table2).2). Remarkably, the [psi− PIN+] sla2Δ haploids grew slower (exponential doubling time, 3.2 h) than the [psi− pin−] sla2Δ haploids (2.3 h) or both [psi− PIN+] and [psi− pin−] SLA2+ haploids (1.8 h). Therefore, growth defects of the sla2Δ deletion are exacerbated in the presence of [PIN+] and lead to lethality in the [PSI+ PIN+] background.
Next, we checked whether de novo [PSI+] induction by transient overproduction of the Sup35N-containing constructs is affected by cytoskeletal alterations. Experiments were performed in both galactose-inducible (Fig. (Fig.3A)3A) and copper-inducible (Fig. 3B and C) experimental systems. sla1Δ caused a moderate decrease in de novo [PSI+] induction, confirming our previous observation in a different genotypic background (4), while end3Δ and act1-R177A strongly inhibited de novo [PSI+] induction. This effect was observed for both constructs overexpressing Sup35N (Fig. (Fig.3A)3A) or Sup35NM-GFP (Fig. 3B and C) and constructs overexpressing complete intact Sup35 (Fig. (Fig.3A).3A). As predicted from the incompatibility between sla2Δ and [PSI+] (see above), no [PSI+] induction was detected in sla2Δ. Reduced [PSI+] induction in the cytoskeletal mutants was not due to a decrease in the levels of the overproduced inducer protein, as no significant differences in levels of Sup35NM-GFP protein induced from the PCUP1 promoter were detected between the wild-type strain and any mutant (Fig. (Fig.3D3D).
As some cytoskeletal alterations, such as act1-R177A, reduced growth of the [PSI+] strains on −Ade (see above), one could suggest that weak [PSI+] isolates induced in these cultures could remain undetectable, thus mimicking a defect in [PSI+] induction. To eliminate the possibility of such an artifact, we induced the PGAL-SUP35 construct on galactose medium in the haploid act1-R177A culture, velveteen replica plated yeast cells onto glucose medium in order to turn off the PGAL promoter, and then mated them by replica plating onto a lawn of the wild-type [psi−] strain of the opposite mating type. Resulting diploids, generated by mating, were velveteen replica plated onto −Ade medium to check for [PSI+]. In this design, Sup35 overproduction and [PSI+] induction occurred in haploids, but [PSI+] detection was performed in diploids, where the recessive act1-R177A mutation, being in the heterozygous state, could not influence [PSI+] detection. As the frequency of [PSI+] derivatives detected in such an assay was decreased in the diploids obtained from act1-R177A, in comparison to identical diploids obtained from the wild-type strain (Fig. (Fig.3E),3E), and mating capability of the yeast cells was not influenced by act1-R177A (Fig. (Fig.3F),3F), we concluded that act1-R177A is defective in induction of the viable [PSI+] derivatives rather than in detection of the newly induced [PSI+] derivatives.
As de novo [PSI+] induction by overproduced Sup35NM-GFP coincides with accumulation of the cytologically detectable Sup35NM-GFP aggregates (see above [Fig. [Fig.2C]),2C]), we have checked whether aggregate accumulation is affected by cytoskeletal alterations. Indeed, cytoskeletal mutations significantly decreased both the percentage of cells with visible Sup35NM-GFP aggregates (Fig. (Fig.4A)4A) and the proportion of the Sup35NM-GFP protein pelletable at low centrifugation speed, in comparison to the wild-type strain (Fig. (Fig.4B4B).
It is worth noting that sla2Δ exhibited the most severe defect inaggregate accumulation among all the cytoskeletal mutants tested. As [PSI+] formation is normally much more frequently detected in the colonies originating from cells with visible aggregates (see above [Fig. [Fig.2C]),2C]), these data suggest that the lack of de novo [PSI+] induction in the sla2Δ culture is not solely due to synthetic lethality between sla2Δ and [PSI+] but also is due to the decreased ability of sla2Δ to produce aggregates.
Remarkably, viable cells with detectable aggregates, isolated from mutant cultures by micromanipulation, produced [PSI+] colonies as frequently as viable cells with detectable aggregates isolated from the wild-type culture (Fig. (Fig.4C).4C). The only exception was the sla2Δ culture, which essentially lacked cells with aggregates. Therefore, decreased [PSI+] induction in the sla1Δ, end3Δ, and act1-R177A strains was due to a defect in generation of cells with detectable aggregates, rather than due to a defect at any step following aggregate formation.
Overproduction of complete Sup35 or its Sup35N-containing fragments inhibits growth of [PSI+] yeast strains (13, 18, 19) and, at very high levels of overproduction, even growth of [psi− PIN+] cells (20). Mutations sla1Δ, end3Δ, and act1-R177A increased toxicity of overproduced Sup35 or Sup35NM-GFP in a [PSI+ PIN+] strain (Fig. 5A and B). These mutations also slightly but reproducibly increased toxicity of overproduced Sup35NM-GFP in the [psi− PIN+] background (Fig. (Fig.5C).5C). A more dramatic increase of excess Sup35 toxicity by cytoskeletal mutants in the [psi− PIN+] background was detected in the presence of the multicopy SUP35-containing plasmid bearing the LEU2-d marker (Fig. (Fig.5D),5D), which allows plasmid copy number to be amplified up to 100 copies per cell by selection on −Leu medium (for description of this experimental assay, see references 19 and 20 and references therein). In the absence of excess Sup35, sla1Δ did not significantly affect growth rates, and quantitative measurements (Fig. 5B and C) confirmed that increased toxicity of excess Sup35NM-GFP in the end3Δ and act1-R177A mutants could not be explained simply by the general decrease of growth rates caused by these mutations.
Prion-dependent toxicity of overproduced Sup35 or its Sup35N-containing derivatives was previously explained by sequestration of the soluble Sup35 protein by aggregates, leading to a shortage of Sup35 at the terminating ribosomes. Indeed, simultaneous expression of the Sup35C region, which lacks the prion domain and cannot be incorporated into aggregates, ameliorates the [PSI+]-dependent toxicity of excess Sup35 in the wild-type strain, as confirmed by us (Fig. (Fig.5A).5A). However, Sup35C was not capable of completely ameliorating toxicity of overproduced Sup35 in the cytoskeletal mutants. This shows that the cytoskeleton-dependent component of excess Sup35 toxicity cannot be explained solely by sequestration of Sup35 into the prion aggregates.
The toxic effect of overproduced Sup35 could be a combination of both inhibition of growth and cell death. To specifically detect cell death, we isolated individual cells from [psi− PIN+] cultures overproducing Sup35NM-GFP by micromanipulation. As mentioned above, only about 30% of cells with detectable Sup35NM-GFP aggregates, isolated from a wild-type culture, were able to form colonies; the rest of them either did not divide or underwent only one division, indicative of immediate or almost immediate growth arrest. About the same fraction of cells with aggregates remained viable in the cytoskeletal mutants, with the exception of sla2Δ, where cells with aggregates were essentially not found (Fig. (Fig.5E).5E). In contrast, cells with diffuse fluorescence essentially always formed viable colonies in the wild-type strain, but about half of them were unable to do so in the mutant strains (Fig. (Fig.5E).5E). This lethality was dependent on Sup35NM, as cells with diffuse fluorescence from the mutant cultures overproducing GFP alone were perfectly viable (not shown). Therefore, mutations in the components of the cortical actin machinery increase toxicity of overproduced Sup35 derivatives specifically in cells not forming visible aggregates.
Assembly of endocytic vesicles and their movement towards the endosomal/vacuolar system involves the processes of actin polymerization and depolymerization, reminiscent of the assembly and disassembly of amyloid fibers (for review, see reference 24). These processes are modulated by the proteins of the vesicle assembly machinery, including the Pan1/Sla1/End3 complex and proteins Sla2, Arp2, and Arp3. Our data show that at least under conditions when Sup35 or its derivatives bearing a QN-rich prion domain are overproduced, interactions between them and some proteins of the cytoskeletal machinery involved in endocytosis can be detected (Fig. (Fig.11 and and2).2). Depletion of at least some components of the vesicle assembly machinery or mutational alteration of actin severely affects aggregation and prionization of overproduced Sup35 in yeast cells (Fig. (Fig.33 and and44).
One of the Sup35N-interacting cytoskeletal proteins, Sla1, contains a region of oligopeptide repeats (26) which resembles the “propagation element” of Sup35N (12, 39). These repeats are located within the C-terminal region of Sla1, which has been shown to interact with Sup35N (4). Our previous data indicated that the internal deletion Sup35Δ22-69, removing a portion of the repeat region, or point mutation PNM2 within the repeat region decreased or abolished two-hybrid interaction with Sla1. These alterations also decrease the ability of Sup35 to generate the prion state (19, 23, 31) but don't eliminate it completely (9, 21, 31). Likewise, sla1Δ decreases but does not completely eliminate [PSI+] induction (4). Our new data explain these observations by demonstrating that Sup35 interacts with the actin machinery through multiple components (Fig. (Fig.11 and and2).2). It is possible that these interactions also involve various regions of Sup35N. For example, Sla2 contains a Q-rich region (35, 51), resembling the QN-rich stretches of the yeast prion domains and potentially capable of interacting with them in a polar zipper-like fashion (40). The QN-rich stretch of Sup35N constitutes an aggregation element which is distinct from the propagation element (12, 39). It is worth noting that some components of the yeast translational machinery, for example eEF-1α, are known to directly bind actin (37). Therefore, association of Sup35 with the translational machinery may further strengthen its interaction with the actin apparatus.
Huge cytologically detectable aggregates of overproduced Sup35NM-GFP, generated in the [psi− PIN+] cells, are associated with various components of the endocytic/vacuolar pathway (Fig. (Fig.2).2). We favor the model proposing that these aggregates are initially assembled in the actin patches (major sites of the endocytic vesicle assembly) and then move inside the cell together with components of the endocytic machinery, eventually resulting in formation of internal rings surrounding the vacuole (Fig. (Fig.6).6). In such a scenario, actin-rich complexes involved in vesicle formation provide initial sites for the assembly of protein aggregates. Aggregate formation would therefore manifest itself as an abortive result of action of the same cellular motors that are normally involved in vesicle assembly and movement during endocytosis.
Remarkably, defects in aggregate formation in the cytoskeletal mutants are accompanied by increased toxicity of overproduced Sup35 (Fig. (Fig.5).5). This coincides with a high lethality among overproducer cells without visible aggregates that is observed only in the mutant cultures (Fig. (Fig.5E).5E). It is therefore likely that the spread of large quantities of the overproduced and possibly misfolded protein throughout the cell cytoplasm is toxic, while the ability of the cells to assemble overproduced proteins into large aggregates with help from the actin cortical cytoskeletal machinery plays, at least to a certain extent, a protective role.
Interestingly, even the sla2Δ cells still contained some amount of overproduced Sup35NM-GFP in the pelletable fraction (Fig. (Fig.4B),4B), indicative of the presence of some aggregates, despite the lack of large cytologically detectable aggregated structures. This agrees with the observation that most of the overproduced Sup35 (41) or Sup35NM-GFP (Fig. (Fig.4B)4B) protein in the [psi− PIN+] culture is present in the insoluble (aggregated) state, despite the fact that only a fraction of the cells contains large cytologically detectable aggregates. It is an attractive possibility that small agglomerates of misfolded proteins, which are capable of sequestering their interacting partners from the cytoplasm or inhibiting the protein folding and/or proteolytic machineries, could be more toxic than large structuralized aggregates confined to specific compartments and relatively isolated from the rest of the cytoplasm.
Although accumulation of large Sup35 aggregates is undoubtedly related to the process of prion formation, these aggregates are not equivalent to mature prions, as some cells with aggregates still give rise to [psi−] colonies (Fig. (Fig.2C).2C). It is probable that large aggregates represent intermediates in the prion generation pathway. Breakage of these aggregates into the smaller ordered oligomeric structures, apparently with the help of chaperones (for review, see reference 12) may initiate the prion propagation cycle (Fig. (Fig.6).6). Our preliminary observations show that Hsps are indeed associated with at least some Sup35NM-GFP aggregates (R. Wegrzyn, L. Ozolins, and Y. Chernoff, unpublished data).
Recent data suggest that in addition to aggregate breakage, the yeast chaperone Hsp104 may be involved in the initial conformational transitions leading to aggregation of Sup35 (45) and polyglutamines (25). This would agree with our previous observation that two-hybrid interactions between Sup35N and Sla1C are not detected in the strain lacking Hsp104 (4). It is therefore possible that Hsp104 is involved in generation of certain local folding alterations that promote both aggregation and interaction with the cytoskeletal components. However, it is also possible that the effect of Hsp104 on the Sup35N-Sla1C interaction is indirect, as Hsp104 is required for the maintenance of not only [PSI+] but also the other QN-rich yeast prions which can, in turn, promote Sup35 aggregation (for review, see reference 11). Thus, understanding the nature of the Hsp104 effect requires further investigation.
Cytoskeleton-assisted generation of large aggregates of prion-forming proteins in yeast cells somewhat resembles aggresome formation in mammalian cells (28). In contrast to mammals, yeast “aggresomes” exhibit peripheral or perivacuolar rather than perinuclear location. It is possible that the biological meaning of both phenomena is also the same. Overproduction likely results in accumulation of misfolded protein molecules due to an insufficient supply of chaperones that are necessary for normal folding. Proteins with high QN content, not degradable by the proteosomal system either due to structural patterns or simply because the proteosomal pathway is overwhelmed by high amounts of misfolded polypeptides, could be assembled into huge aggregates and transported to close proximity of the lysosome/vacuole for the purpose of their subsequent degradation, maybe via formation of the autophagosomes, eventually fused to the vacuole. However, aggregation of a protein with prion-forming potential also facilitates formation of the self-perpetuating prion polymers, which can escape degradation (Fig. (Fig.66).
It remains to be seen to what extent prion induction by protein overproduction resembles rare (10−6 per cell division ) spontaneous generation of [PSI+] cells in nonoverproducing yeast cultures. Cell-to-cell variations in protein levels are difficult to measure with existing experimental techniques. It is possible that spontaneous prion generation results from the rare appearance of cells accumulating increased amounts of Sup35 due to variations in efficiency of either gene expression or protein degradation, or both. In this case, our data may reflect involvement of the vesicle assembly networks and aggresome-like structures in spontaneous prion generation under natural conditions.
Aggregates seen in the mature [PSI+] cells are no longer associated with actin patches (Fig. (Fig.2C2C and Table Table1),1), but they still bind some other cytoskeletal components, such as Sla1 and Sla2 (Fig. 2G and H). Long-term disruption of the actin apparatus with latrunculin A results in significant loss of [PSI+] (5), suggesting that some components of the actin cytoskeleton may continue to play a role in propagation of the mature prions. Moreover, cytoskeletal mutants increase the toxicity of excess Sup35 in [PSI+] cultures (Fig. (Fig.5A).5A). The sla2Δ mutant, which exhibits the most severe defect in aggregate accumulation in [psi− PIN+] cells (Fig. (Fig.4A),4A), is lethal in the [PSI+ PIN+] background. The cytoskeleton-dependent component of toxicity is probably not related to sequestration of functional Sup35 by prion aggregates, as toxicity is not ameliorated by expression of Sup35C (Fig. (Fig.5A).5A). It is possible that the interaction of prion aggregates with cytoskeletal proteins helps cells to localize misfolded prion proteins and prevent them from causing significant damage.
While we have not detected any significant effect of cytoskeletal mutations on [PSI+] propagation under normal conditions, we have previously observed that sla1Δ increases [PSI+] loss in the presence of some chemical agents, such as dimethyl sulfoxide, or in the presence of excess Hsp104 (4). This suggests that the cytoskeletal apparatus could be involved not only in de novo prion formation but also in prion recovery from some prion-curing treatments. Interestingly, ring-like Sup35 aggregates, which are never seen in cells that propagate mature [PSI+] prion under normal conditions, can be observed in the presence of excess Hsp104 (reference 52; confirmed by our unpublished observations). An attractive possibility is that [PSI+] recovery from the disaggregating activity of excess Hsp104 occurs via the pathway that is controlled by the cytoskeletal apparatus and involves intermediate aggregated structures similar to those observed during de novo induction of [PSI+]. This possibility is currently under investigation.
It is not yet known whether proteins of the cortical actin cytoskeleton influence other prions in the same way as they influence [PSI+]. At the least, depletion of cytoskeletal proteins involved in endocytosis increases [PIN+]-dependent toxicity of the huntingtin-derived poly-Q constructs in yeast (34). Moreover, poly-Q aggregation in [PIN+] cells blocks endocytosis, suggesting a poisoning of the endocytic/vacuolar pathway as one of the mechanisms contributing to poly-Q toxicity. Other components of the cytoskeleton, specifically microtubules, were also shown to be required for the formation of poly-Q aggregates in yeast cells, and disruption of microtubules increased poly-Q toxicity (36). This parallels our observation that aggregate formation may play a protective role in certain situations. Hip1, one of the mammalian homologs of Sla2, interacts with huntingtin in mammalian cells (29). This points to the possible involvement of cortical cytoskeletal networks in poly-Q aggregation and toxicity in mammalian cells as well as in yeast. It has also been suggested that association with certain membrane structures (6) or recycling through the endocytic pathway (46) may play a role in the conversion of mammalian PrP into a prion state. It is therefore possible that the actin cortical cytoskeleton and other membrane-associated cytoskeletal networks play a universal role in promoting aggregation and/or prionization of various types of misfolded proteins.
We thank B. Chen, M. Gleason, K. Gokhale, and P. Winslett for help in some experiments, A. Meriin for discussion of results, and D. Drubin, S. Fields, L. Hicke, and S. Lindquist for plasmids, strains, and antibodies.
This work was supported by NIH grant R01GM58763 to Y.O.C. L.N.O. was a recipient of the Summer Internship from HHMI and of the Undergraduate Research Scholarship from the Emory-Georgia Tech Center for Engineering of the Living Tissue. S.B. was a recipient of the Presidential Undergraduate Research Assistantship from Georgia Institute of Technology.