Yeast [PSI+] prion aggregates are formed by small Sup35 polymers fragmented by Hsp104.

The yeast [PSI+] determinant is related to formation of large prion-like aggregates of the conformationally altered Sup35 protein. Here, we show that these aggregates are composed of small Sup35 prion polymers and associated proteins. In contrast to other protein complexes of yeast lysates, but similarly to amyloid fibers, these polymers are insoluble in SDS at room temperature. The polymers on average are about 30-fold smaller than the aggregates and comprise from 8 to 50 Sup35 monomers. The size of polymers is characteristic of a given [PSI+] variant and differs between the variants. Blocked expression of Hsp104 chaperone causes gradual increase in the size of prion polymers, while inactivation of Hsp104 by guanidine HCl completely stops their fragmentation, which shows indispensability of Hsp104 for this process.

Some proteins can change their fold from normal to a specific alternative form, called prion, which is able to catalyze this change (1). In man and animals such process causes prion diseases like Creutzfeldt-Jacob disease, bovine spongiform encephalopathy, and scrapie of sheep. A similar autocatalytic mechanism is shared by human amyloid diseases, which are noninfectious, in contrast to prion diseases (2). In yeast, there are several proteins, which can undergo prion-like structural conversion. The most studied of them are translation termination factor eRF3, also called Sup35, and Ure2 involved in regulation of nitrogen metabolism (3). The prion state of these proteins can propagate stably for many cellular generations. This may be observed by characteristic phenotypes, [PSI ϩ ] and [URE3], since these proteins in their prion form are aggregated and functionally inactive. In particular, the prion state of Sup35 results in low levels of soluble functional Sup35 and impaired translation termination, which is manifested as a nonsense-suppressor [PSI ϩ ] determinant. Similarly to mammalian prions, different variants or "strains" of [PSI ϩ ] were observed, apparently related to variations in prion structure. "Strong" [PSI ϩ ] variant exhibited strong suppression and high mitotic stability, while "weak" [PSI ϩ ] showed weak suppression and low stability. The efficiency of the suppression correlated inversely with the levels of soluble Sup35 (4,5).
The prion aggregates formed by prionogenic proteins in vivo represent large complexes, whose structure is not well characterized. In vitro, purified Sup35 and Ure2 formed fibers of uniform structure, which shared the key properties of amyloid fibers (6,7). This allowed proposing that the prion aggregates in vivo are composed of similar structures (called, hereafter, prion polymers). It is not clear whether the aggregates represent single prion polymers or their higher order complexes. The electron microscopy of [URE3] cells revealed networks of small fibers coinciding in location with the Ure2 molecules (8), which indicates in favor of the latter opportunity.
The [PSI ϩ ] propagation may be affected by the chaperones of Hsp70 and Hsp40 families, but the most significant role is played by Hsp104, which is the only chaperone strictly required for [PSI ϩ ] and other prions. The overexpression of Hsp104, however, caused frequent [PSI ϩ ] loss or antisuppressor effect when [PSI ϩ ] was not lost (9). Hsp104 was shown to break large aggregates of denatured protein into smaller pieces (10,11). We proposed that Hsp104 acts similarly on fiber-shaped prion polymers, thus fragmenting them into shorter polymers and increasing their number (12). This is essential for their inheritance and accelerates the prion conversion by multiplying the ends of prion polymers, where the conversion occurs. The overproduction of Hsp104 should cause excessive fragmentation, increased levels of soluble Sup35, and possibly [PSI ϩ ] loss. An alternative model proposed that Hsp104 is primarily required to facilitate the prion conversion in one or another way (13,14).
These two models may be distinguished, since they make different predictions for alteration of the size of prion particles upon inhibition of the Hsp104 function. By the former model, the size should increase due to blocked fragmentation, while by the latter it should stay constant or decrease due to block of polymerization. Recent studies provided some support for the "fragmentation" model. Decrease of the Hsp104 expression caused increase in the size of Sup35 prion aggregates, suggesting decreased disaggregation by Hsp104 (15). The activity of Hsp104 is inhibited by growing yeast cells in the presence of 3-5 mM guanidine HCl (GuHCl) 1 (16). Such treatment cures efficiently [PSI ϩ ] (17) and other known yeast prions. Study of the kinetics of [PSI ϩ ] loss in the presence of GuHCl allowed concluding that it blocks replication of prion "seeds" (18,19). Thus, Hsp104 inhibition correlates with the block of fragmentation (replication) of prion particles (seeds). However, in these experiments the relation of the studied prion entities to the Sup35 polymers considered by the above models was not characterized. The prion seeds were defined genetically, but their physical nature was not studied. In the work (15) the size of prion aggregates was estimated by fluorescence of Sup35-GFP fusions or immunofluorescent staining. However, this approach insufficiently characterizes the role of Hsp104, since (i) the fluorescence reveals only the largest aggregates, not all of them; (ii) it is unclear whether the observed fluorescent foci represent single prion particles or their higher order complexes; and (iii) the lack of quantitative estimates does not allow to conclude whether Hsp104 is critical for the fragmentation or just moderately modulates it.
To study the structure of prion aggregates and characterize the role of Hsp104, we developed here a novel approach for purification of prion particles and determination of their size. Using it, we showed that the Sup35 prion aggregates observed in vivo represent complexes of relatively small prion polymers with amyloid-like properties. The size of these polymers varied significantly between different [PSI ϩ ] variants. It depended greatly on the activity of Hsp104 in a manner that strongly supports the idea that Hsp104 mediates fragmentation of Sup35 prion polymers.  (20) were described earlier, and others were obtained in this work. The strong [PSI ϩ 1 ] variant was mainly studied unless otherwise indicated.Yeast were grown at 30°C in rich (YPD, 10 g yeast, 20 g peptone, 20 g glucose per 1 liter) or synthetic (SC, 6.7 g yeast nitrogen base, 20 g glucose supplemented with the required amino acids) media (22). 3 mM GuHCl or 5 g/ml doxycycline was added where indicated.

Strains and Genetic
Plasmids-To obtain a repressible HSP104 gene, plasmid pCM188 was used that contains tetracycline-repressible TET promoter (23). The ARS1-CEN4 block of pCM188 was deleted by EheI and BstXI sites, to produce pCM188i. The BsaAI-HindIII fragment containing the 5Ј-half of the HSP104 ORF was cloned into the HpaI and HindIII sites of pCM188i. For genomic integration into 5V-H19, this plasmid was cut by unique BglII site. The integration resulted in the complete HSP104 gene under the control of TET promoter and a non-functional 5Ј part of HSP104 and the strain T104 -5V-H19. The overexpression of Hsp104 was achieved by transforming yeast cells with the pFL44L-HSP104 plasmid (24).
Preparation and Analysis of Yeast Cell Lysates-Yeast cultures were grown in liquid media to an OD 600 of 1.5. The cells were harvested, washed in water, and lysed by glass beads in a buffer A: 25 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM dithiothreitol. To prevent proteolytic degradation, 10 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, and Complete TM protease inhibitor mixture (Roche Applied Science) were added. Cell debris was removed by centrifugation at 10,000 ϫ g for 2 min. To equalize the amount of protein in the electrophoretic samples, the protein concentration was determined according to Ref. 25. To determine the size distribution of Sup35, the cell lysates were fractionated by centrifugation through 15-40% sucrose gradient in the SW-41 rotor (Beckman) for 15 min at 35,000 rpm (170,000 ϫ g).
To check that glass beads lysis does not impair prion polymers, 5V-H19 [PSI ϩ ] cells were treated with zymolyase to remove cell walls. The obtained spheroplasts were lysed either by adding SDS to 1% or by vortexing with glass beads with standard intensity and duration. The experiment was also performed for the cells, in which the size of prion polymers was increased 4-fold by growth in the presence of 3 mM GuHCl for two generations. In both cases, the size of prion polymers did not depend on the lysis method (data not shown).
Electrophoresis-Protein electrophoresis in 10% polyacrylamide gels was performed according to Ref. 26, except that where indicated, boiling of the samples was avoided, and agarose was used for the concentrating gel. For separation of prion particles we used horizontal 1.8% agarose gels in the Tris-acetate-EDTA buffer (27) with 0.1% SDS. Samples were incubated in the sample buffer (0.5ϫ Tris-acetate-EDTA, 2% SDS, 5% glycerol, and 0.05% bromphenol blue) for 5 min at 37°C. After the electrophoresis, proteins were transferred from gels to Immobilon-P PVDF sheets (Millipore) by vacuum blotting overnight or electro-phoretically. Bound antibody was detected using the Amersham Biosciences ECL system.
Determination of the Nonsense Read-through-The UAA nonsense read-through levels were determined as a ratio of ␤-galactosidase activity in cells transformed with pUKC817 plasmid to that in cells with pUKC815 as described previously (28).
Calculations-The masses of Sup35-containing complexes in the sedimentation experiment (Fig. 3A) were calculated relatively to the material in fraction 2 by assuming that the mass is proportional to the sedimentation distance to the power of 3/2 (29). The average masses were calculated as ⌺(M n ϫ P n ), where M n is a relative mass of the fraction n, and P n is a proportion of Sup35 in this fraction relatively to the total amount of Sup35. The molecular mass of prion polymers was calculated from their mobility in agarose gels assuming a linear dependence of the logarithm of mass and the migration distance: lgM X ϭ aS X ϩ b, where M X is mass, S X is migration distance, and a and b are constants (29). The same equations are true for titin and nebulin, for which the masses M T ϭ 4200 kDa and M N ϭ 740 kDa, and migration S T and S N , are known. Solving the system of these three equations, we obtain: The relative increase in the molecular mass of Sup35 polymers in the experiments presented in Fig. 5 was calculated as a ratio of masses corresponding to the geometric centers of Sup35 spots in each lane.

[PSI ϩ ] Prion Aggregates Are Composed of Sup35 Prion
Polymers-To check the available models for the structure and maintenance of yeast prions, it was required to determine the size of yeast prion polymers. However, the methods, which would allow doing this with a reasonable precision, were unavailable. Furthermore, it was likely that the aggregates observed by the existing methods, e.g. centrifugation or GFP visualization (13,21), represent not polymers, but their higher order complexes possibly including some other molecules. This prompted us to develop an electrophoresis-based approach for analysis of prion polymers. First, it was required to find the conditions, which destroy associations of prion particles with other molecules and with each other, but do not disrupt these particles. Earlier, Paushkin et al. (30) used GuHCl at high concentrations to obtain Sup35 prion aggregates containing no detectable contaminants. However, we found later that while GuHCl solubilized the majority of cellular proteins, it also caused secondary precipitation of some proteins of yeast lysates (data not shown). We tested several other reagents at different concentrations for their ability to purify prion particles, including urea, high salt (KCl), deoxycholate, zwittergent 3-10, cetyltrimethyl ammonium bromide, N-lauroyl sarcosyl, and sodium dodecyl sulfate (SDS). For this, the aggregated fraction of 5V-H19 [PSI ϩ ] lysates was dissolved in buffers containing these reagents and centrifuged at high speed, and the proportion of Sup35 to other proteins in the pellets was studied (data not shown). The best purification of Sup35 was achieved with Nlauroyl sarcosyl and SDS applied at room temperature. Further tests were focused on the use of SDS, which had minor advantages over sarcosyl.
It was important to confirm that the presence of SDS does not impair the Sup35 prion particles. For this, the aggregate fraction of 5V-H19 [PSI ϩ ] lysate was dissolved in the Laemmli sample buffer containing 2% SDS, and aliquots were incubated at different temperatures ranging from 20 to 100°C. Without further boiling the samples were subjected to SDS-PAGE and analyzed for Sup35 by Western blotting (Fig. 1A). Sup35 was represented by two bands: a monomer band of about 80 kDa and polymers stuck at the top of the separating gel. In the samples incubated at 42°C or lower temperatures, low levels of monomers were observed, which are likely to represent ribosome-bound monomeric Sup35. Thus, Sup35 prion particles were resistant to SDS at these temperatures. They dissolved to a small extent at 50°C and almost completely at 70°C and higher temperatures. Earlier, a similar ability to withstand SDS at room temperature was demonstrated for Sup35 amyloid-like polymers obtained in vitro (31). This supports the structural similarity of Sup35 prion polymers and Sup35 amyloid fibers.
The described properties of SDS allow its use for electrophoretic analysis of the prion particles. However, since boiling of the samples was excluded, correct estimation of the mass of Sup35 prion particles required showing that no additional proteins remained associated with them in the presence of SDS. For this, we analyzed the proteins trapped at the top of the separating gel together with prion Sup35. A strip of the separating gel, containing Sup35 polymers, was cut off, boiled in SDS sample buffer, and the proteins separated using another SDS-PAGE gel as described in the legend to Fig. 1B. In this way, one can observe the proteins, which run as polymers before boiling, but as monomers after boiling. These should include Sup35 prion, other proteins with similar properties, and the proteins associated with Sup35 particles during the first electrophoresis. The only abundant protein band represented Sup35 (Fig. 1B), as confirmed by immunoblotting (not shown). Although in this experiment some minor proteins could escape detection by Coomassie Blue, it is possible to conclude that SDS dissolves most of the yeast non-prion protein complexes and removes most of the proteins, which could be associated with Sup35 prion polymers.
The large size of prion polymers, presumably of the megadalton range, precluded the use of polyacrylamide gels, even at lowest possible concentration of about 3%. So, in further exper-iments agarose gels were used. As a molecular mass standard, the preparation of rabbit myofibrils, which contains giant proteins titin and nebulin, was used (32). We assumed that the molecular masses of titin and nebulin are equal to 4200 and 740 kDa, which corresponds to the coding capacities of respective human genes (33,34). The homologous rabbit genes were not completely sequenced, but the partial amino acid sequence of rabbit titin was highly similar (99%) to its human counterpart (34).
A range of SDS concentrations for the treatment of samples was tried using 5V-H19 [PSI ϩ ] lysate to check that they do not affect the size of prion particles. The samples were analyzed using horizontal agarose gel as described under "Experimental Procedures" (data not shown). The mobility of Sup35 oligomers did not increase when the SDS concentration increased from 0.5 to 5% or duration of treatment increased from 3 to 10 min. This confirms that SDS does not impair Sup35 polymers; otherwise, we would observe a decrease in the polymer size with the increase of SDS concentration and duration of treatment. The polymer mobility was somewhat decreased at SDS concentrations lower than 0.5%. Most likely, these low levels of SDS were insufficient to destroy all non-prion interactions. In further work, buffers containing 2% SDS were used for incubation of lysate samples at 37°C. This removes the associated proteins from Sup35 prion polymers but leaves them intact and allows to analyze their size using agarose gel containing 0.1% SDS. This method was called semi-denaturing detergent-agarose gel electrophoresis (SDD-AGE).
A comparison of [PSI ϩ ] and [psi Ϫ ] lysates with this method distinguished them clearly (Fig. 2). In [PSI ϩ ] lysate, Sup35 was mostly in the polymer form, while in [psi Ϫ ] it was monomeric. In contrast, in the centrifugation analysis a significant portion of Sup35 was usually found in the aggregated fraction of both [PSI ϩ ] and [psi Ϫ ] lysates (Fig. 3A). The prion polymers were heterogeneous in size, ranging from 700 to 4000 kDa, which should correspond to 9 -50 Sup35 monomers.
Since we found that SDS removes associated proteins from Sup35 prion polymers, it was of interest how the size of these polymers compares with the size of Sup35-containing aggre- gates observed under the native conditions used in standard centrifugation methods. For this, we analyzed the lysate of 5V-H19 [PSI ϩ ] strain by centrifugation in the presence of 2% SDS and its absence. Comparison of the Sup35 sedimentation profiles showed that the presence of SDS greatly reduced the size of Sup35 prion complexes (Fig. 3A). On average, the mass of Sup35 prion aggregates was 35-fold greater than that of the polymers (see "Experimental Procedures"). So great a difference in sizes of prion polymers and aggregates may be due to two reasons, the association of Sup35 polymers in aggregates with some other proteins and macromolecular complexes, such as ribosome, and the presence of more than one Sup35 polymer per aggregate. However, the relative contribution of these two factors is difficult to determine. The Sup35 prion aggregates varied greatly in their size. The smallest of them were comparable with prion polymers, and the largest were about 100-fold bigger. However, the prion polymers, constituting these aggregates, differed much less in their size. The centrifugation fractions obtained without SDS were analyzed by SDD-AGE (Fig.  3B). The average size of Sup35 polymers in the slowest sedimenting fraction was about 900 kDa and in the pellet about 1500 kDa, which constitutes only 1.7-fold difference.
[PSI ϩ ] Variants Differ in the Size of Sup35 Polymers-Different isolates of [PSI ϩ ] may vary in the strength of the nonsense suppressor phenotype and in mitotic stability (35). Such variation was also observed for hybrid prion [PSI ϩ PS ] based on Sup35 prion domain from yeast P. methanolica (20). Usually, stronger suppression correlates with higher stability. It was proposed that the variation in [PSI ϩ ] properties reflects the difference in structure of prion particles (36 -38). This may result in variation of prion polymerization speed and the frequency of fragmentation of prion polymers and, therefore, in their different size (12). To check this, we compared the size of Sup35 polymers in different [PSI ϩ ] isolates. A significant variation was found, with the size being inversely related to the strength of [PSI ϩ ] (Fig. 4) GuHCl Inhibits Fragmentation of Sup35 Prion Polymers-GuHCl was shown to inactivate Hsp104, which should be a reason of its prion curing effect (16,19). To investigate the effect of GuHCl on the size of prion polymers, the 5V-H19 [PSI ϩ ] strain was grown in the YPD medium containing 3 mM GuHCl. A half-volume of the culture was taken for analysis once per generation and replaced with an equal amount of fresh medium with GuHCl to keep the culture density constant. After the third generation, the cells were transferred to YPD lacking GuHCl, and aliquots were taken once per hour. Cell lysates were obtained and analyzed by SDD-AGE (Fig. 5A). In the presence of GuHCl the average size of the Sup35 polymers increased about 2-fold per generation. Therefore, the fragmentation of Sup35 polymers was impaired, rather than their growth by accretion of new Sup35 molecules. Furthermore, the observed polymer growth rate represents a maximum, achievable on conditions that all Sup35 monomers incorporated into polymers and that the polymers did not experience fragmentation. Indeed, in such case the total number of Sup35 prion polymers remains constant, but at each generation the number of cells doubles, and therefore, the number of Sup35 polymers per cell should decrease 2-fold. Since the amount of Sup35 per cell remains constant and most of it is in the polymer form, the number of Sup35 molecules per polymer should double. This allows concluding that GuHCl entirely blocks the fragmentation of Sup35 prion polymers. The doubling of polymer size at the first generation suggests that the fragmentation was blocked immediately after addition of GuHCl. After removal of GuHCl from the medium, the size of Sup35 polymers was unchanged for about 1 h and then gradually returned to the starting values. The delay in decrease of the size of prion polymers could reflect the time required for washing out of GuHCl from the yeast cells.
Reduced Levels of Hsp104 Increase the Size of Sup35 Prion Polymers-To study the effects of the lack of Hsp104 directly, the chromosomal HSP104 gene was put under control of the tetracycline-regulatable TET promoter. This allowed to inhibit the synthesis of Hsp104 by addition of antibiotic to culture medium. However, it was not possible to achieve a rapid decrease of the Hsp104 levels, since this chaperone degrades very slowly, and its amount should decrease mainly due to dilution at cell divisions (15). In the absence of antibiotic, the TET promoter was induced and provided about 2-fold increased level of Hsp104 compared with the native HSP104 promoter. This difference did not alter significantly the size of Sup35 polymers and did not cause antisuppression or [PSI ϩ ] loss, often associated with excess Hsp104. The experiment was started with addition of the tetracycline analog, doxycycline, to 5 g/ml to repress the TET promoter. Culture aliquots were taken once per generation, and cell lysates were analyzed by SDD-AGE. The size of Sup35 polymers started to increase after two generations of HSP104 repression, reaching about 4-fold increase after six generations (Fig. 5B). This is a significantly slower increase than the one caused by GuHCl, which should be related to the presence of decreasing amounts of Hsp104 during the experiment. The basal level of Hsp104, determined after 2 days of growth in the presence of doxycycline, was about 5% of its wild type levels.
The Size of Sup35 Prion Polymers Does Not Decrease with the Increase of Hsp104 Levels-Overexpression of Hsp104 in [PSI ϩ ] strains increased the levels of soluble Sup35 and decreased the size of prion aggregates, as determined by centrifugation under native conditions (20,21). It was of interest to check the effects of excess Hsp104 by the SDD-AGE approach. The increased levels of monomeric Sup35 were indeed observed in the [PSI ϩ ] lysate with excess Hsp104, while no soluble Sup35PS was observed in [PSI ϩ PS ] cells, independently of the Hsp104 overexpression (Fig. 5C). Surprisingly, the size of prion polymers increased in the [PSI ϩ ] and was unchanged in the [PSI ϩ PS ] lysates, in contrast to a decrease in the size of aggregates.

Polymers and Aggregates: Two Levels in the Structure of
Yeast Prions-Yeast prions, the same as prions of higher eukaryotes, are presumed to be structurally similar to amyloid fibers (39). This assumption is related in part to observations that the yeast prionogenic proteins Sup35 and Ure2 tend to form amyloid fibers in vitro (6,7). In [PSI ϩ ] cells, yeast prions form large aggregates (21). However, no direct data were available on the structure and composition of these aggregates and their relation to amyloid fibers. Here, we showed that these aggregates represent complexes formed by relatively small Sup35 prion polymers and other associated proteins. The prion polymers were physically different from all other protein complexes; while the majority, if not all, of protein complexes of yeast lysates were dissolved in 2% SDS at 37°C, the prion polymers were resistant to this condition. This unique property of prion polymers is also typical of the Sup35 amyloid fibers formed in vitro (31). This strongly suggests that the prion polymers represent amyloid fibers. The ability to withstand SDS at room temperature was also observed for polymers of another prion protein, Rnq1, in [PIN ϩ ] lysates (data not shown), which suggests this to be a general property of yeast prions. It is also interesting that while non-prion protein complexes were dissolved by SDS, many individual proteins retain their function and, therefore, structure in its presence. For example, 1% SDS does not affect the activity of proteinase K (27), yeast cellular proteases, and the fluorescence of green fluorescent protein (data not shown). Thus, the interactions between monomers within the prion polymer are comparable by their strength to intraprotein interactions. By the amyloid model of Jarrett and Lansbury (40), a certain minimal number of monomers is required for a stable amyloid polymer. This number is related to the number of monomers per turn of helical structure of polymer and may define the kinetics of appearance of amyloid seeds de novo. The smallest polymers observed in our work contained about eight monomers. However, we do not exclude that smaller polymers could be stable, but in vivo they were preferentially eliminated by chaperones.
By centrifugation, the Sup35 prion polymers were on average about 35-fold smaller than the aggregates, which they form   Fig. 2. Immunostaining for Sup35 is shown. (Fig. 3). SDD-AGE revealed that the polymers comprised on average about 15-30 Sup35 molecules, depending on the prion variant. This value may be used to estimate the cellular number of prion polymers. It is possible to suggest that there are about 30,000 Sup35 molecules per yeast cell. Although this value was not directly determined for Sup35, it was obtained for Sup45 (41), a functional partner of Sup35. These proteins form a heterodimer functioning in the translation termination (28), their genes show similar codon bias indexes, and the SUP35 mRNA is slightly more abundant than the SUP45 one (41). This allows suggesting that Sup35 and Sup45 are expressed at similar levels. With an average prion polymer containing 20 Sup35 molecules, there should be about 1500 such polymers per cell. On the other hand, Eaglestone et al. (18) calculated that there are about 60 heritable prion seeds per cell. These seeds are likely to be equivalent to the Sup35 aggregates. The difference in numbers also suggests that the seeds are not equivalent to the polymers. Then there should be about 25 prion polymers on average per prion seed (aggregate), which agrees well with our centrifugation data (Fig. 3A).
It is interesting to note that the prion polymers and aggregates are relatively independent in defining the prion phenotypes. The prion conversion and the suppressor phenotype should depend on the number and properties of the polymers, while the stability of inheritance depends on the number of aggregates. These numbers usually correlate, but at least in one case, that of Hsp104 overexpression, the correlation could be broken: the number of polymers decreased, while the number of aggregates increased. Indeed, the numbers of these particles are inversely related to their size, and upon Hsp104 expression the size of Sup35 polymers increased, while the size of aggregates decreased.
The ability of prion polymers to catalyze further polymerization implies that a polymer represents a minimal particle able to act as a heritable seed. The aggregation of several polymers into a single particle reduces the number of seeds to a single seed, but this should not decrease the efficiency of polymerization. It should be noted that the term "prion aggregates" does not necessarily imply the presence of multiple polymers in them. This term was only used to denote the prion particles revealed by centrifugation in the absence of SDS. The smallest aggregates were almost of the same size as prion polymers and thus were likely to contain just one prion polymer.
The prion visualization using Sup35-GFP fusions usually reveals most of prion Sup35 being localized in several bright foci (13), which contrasts with the several tens of prion seeds calculated by Eaglestone et al. (18). Apparently, these foci represent the largest prion aggregates, which, although comprising most of cellular Sup35, constitute only a small portion of prion seeds. The majority of seeds should be too small to be detectable by fluorescent microscopy. Thus, despite the convenience of GFP test for prion detection, it may not be used for judging the quantitative aspects of prion aggregation, e.g. the number or size of the aggregates. Also, in some cases no foci were detected in the cells containing GFP fusions to prions based on Sup35 and Ure2 (42,43). This phenotype is compatible with prions being polymers but with a low degree of aggregation.
[PSI ϩ ] Variants Differ in Properties of Sup35 Prion Polymers-Earlier it was proposed that the variation of [PSI ϩ ] properties, known as "prion strains" or "prion variants" results from variation in the structure of prion particles (36 -38). In the frame of polymerization-fragmentation model, this means that the prions may differ in the polymerization speed (hereafter, the "speed" is used to characterize growth of a polymer; the "rate" describes overall prion protein polymerization and is equal to the product of the speed and the number of polymers in a cell) and in the efficiency of prion recognition and fragmentation performed by Hsp104 (12). These parameters define the size of Sup35 polymers, which, therefore, may differ between prion variants. Indeed, we observed a significant variation in the length of prion polymers in independent isolates of  (Fig. 4). The length of polymers correlated inversely with the suppressor strength of the [PSI ϩ ] variants, with a single exception of strong [PSI ϩ PS-1 ], which was distinguished by large polymers. Such correlation may be anticipated, as well as some exceptions from it. Smaller polymers mean their higher number per cell, higher polymerization rate, lower levels of monomeric Sup35, and higher nonsense suppression. In this logical sequence, however, one step is not strict; the polymerization rate is proportional not only to the number of prion polymers but also to the polymerization speed. The correlation observed among the studied [PSI ϩ ] variants suggests the lack of significant variation of this speed between them, but for some other [PSI ϩ ] variants, for example [PSI ϩ PS-1 ], this may not be true. A further reservation should be made for the cells carrying strong [PSI ϩ PS ] variants, which contained very low levels of soluble Sup35. These levels were lower than those in [PSI ϩ ] variants of similar suppressor efficiency and were below the 5-10% threshold minimally required for the cell viability (44). This forces us to suggest that polymerized Sup35 retains, although decreased, functional activity. It appears likely that in the prion aggregates the functional C-terminal domain of Sup35 is not structurally rearranged, but its linkage to a bulky aggregate sterically interferes with its interaction with the ribosome and/or other essential targets. This interference could be decreased for hybrid Sup35PS, since its middle domain, which could act as a spacer between the prion and functional domains, is larger than in S. cerevisiae Sup35 (162 amino acids versus 130). Therefore, in the [PSI ϩ PS ] variants with very low levels of monomeric Sup35, the efficiency of translational suppression may depend on the structure of prion particles, while in other variants the suppression should primarily depend on the levels of soluble Sup35.
The increased Sup35 polymer length in weak prion variants suggests fewer prion polymers per cell, which agrees well with low mitotic stability of these variants. Although one can presume that prion stability depends on the number of prion aggregates, rather than polymers, these numbers should correlate, since longer polymers seem to have higher propensity to aggregate (Fig. 3). Increased size of polymers should cause an even greater increase in the size of prion aggregates.
Hsp104 Is Required to Fragment Prion Polymers-Hsp104 plays a key role in propagation of yeast prions, since it is the only chaperone, which is indispensable for their propagation. To explain the requirement for Hsp104 in prion propagation, we proposed that Hsp104 is necessary to fragment prion polymers (12), which is required for their inheritance and for faster prion conversion. An alternative model proposed that Hsp104 is required for the prion conversion coupled with polymerization by helping to obtain certain unfolded intermediate form of prionogenic protein (13,14). It is possible to discriminate between these models, since they make different predictions for the case of lack of Hsp104 function; the size of prion polymers should increase by the former model and should not by the latter. Here, the Hsp104 function was decreased in two ways: by decreasing its levels and by GuHCl-mediated inhibition. In both cases a rapid increase in the size of Sup35 polymers was observed, which indicates a defect of fragmentation rather than of polymerization. These data agree with those of Ness et al. (45), who showed continued Sup35 polymerization and the lack of prion seeds replication in the presence of GuHCl. The only difference is that, as we noted above, the seeds are likely to be equivalent to prion aggregates rather than to polymers. These data support the first model and contradict to the second. Furthermore, the GuHCl experiment revealed a complete block of prion fragmentation. The size of Sup35 polymers grew 2-fold per cell generation, a rate that may only be achieved upon the complete block of fragmentation. It may be assumed that Hsp104 was solely responsible for the block, even though some other proteins involved in [PSI ϩ ] propagation may also be affected by GuHCl. Since the [PSI ϩ ] curing by GuHCl could be prevented by mutations in the HSP104 gene (46), other proteins are not critical for the fragmentation block. The levels of free monomeric Sup35 were not increased during the first generation in the presence of GuHCl (Fig. 5A), which also confirms that Sup35 prion polymerization was not inhibited. It was reported recently that the propagation of prions based on some Sup35 modifications altered at the N or M domains do not depend on Hsp104 (42,47). These data suggest a possibility of the prion fragmentation mechanism independent from Hsp104. Nevertheless, the strict dependence on Hsp104 of [PSI ϩ ] and other natural prions suggests that this mechanism does not play a noticeable role in these cases.
It should be noted that Hsp104 functions both at the level of prion polymers and aggregates. Earlier data showed that Hsp104 acts to dissolve the aggregates (20), (15). Here, we showed that it is essential for fragmenting the prion polymers. The processes of disaggregation and fragmentation are likely to differ significantly, despite that they may be driven by an essentially similar action of Hsp104. Prion polymers are regular, while aggregates are likely to be irregular, structures. The polymers grow via accretion of monomers; the aggregates, in addition, can grow by joining preexisting polymers or aggregates. Besides, the polymers are more solid, since the aggregates, but not polymers, are dissolved in the presence of SDS. Due to these differences, the polymers and aggregates may interact differently with various chaperones. For example, overexpression of Hsp104 makes aggregates smaller but may enlarge the polymers.
It is interesting to note that the speed of prion polymerization, at least in strong [PSI ϩ ] variants, is substantially restricted by the availability of soluble Sup35. This follows from the GuHCl experiment, during which the polymerization speed increased. Indeed, during the first two generations in the presence of GuHCl the cellular rate of Sup35 polymerization was constant and approximately equal to the rate of Sup35 synthesis, since only a relatively small amount of soluble Sup35 accumulated. After the two generations, the number of polymers per cell decreased 4-fold, and therefore, the speed of Sup35 polymerization increased 4-fold due to increasing levels of Sup35 monomers.
Overproduced  (20). The electrophoretic analysis of Sup35PS (Fig. 5C) from these [PSI ϩ PS ] cells revealed no soluble Sup35PS, which explains the absence of antisuppression. The lack of soluble Sup35PS was also observed when analyzing the same non-boiled samples using 10% acrylamide gel (data not shown). Apparently, the centrifugation did not distinguish monomeric Sup35PS and small aggregates, presumably single prion polymers, mainly present in the "soluble" fraction.
The excess of Hsp104 decreased the size of Sup35 aggregates in both [PSI ϩ ] and [PSI ϩ PS ] cells (20), in agreement with the assumption that Hsp104 breaks prion particles into smaller fragments (12,21). Surprisingly, the electrophoretic analysis showed that the size of Sup35 prion polymers increased in the presence of overproduced Hsp104, and the size of Sup35PS polymers was unchanged. Thus, excess Hsp104 might have the opposite effect on the size of Sup35 prion aggregates and of Sup35 polymers. The decrease in size of prion aggregates could be an indication of increased activity of Hsp104, which disrupted associations between Sup35 polymers in an aggregate and/or removed associated proteins. But the increase in size of Sup35 polymers was not predicted previously, and its explanation requires additional assumptions. It is probable that individual prion polymers vary in their susceptibility to Hsp104. The polymers could be less accessible to Hsp104 if they are associated with each other or with other macromolecular complexes, and such associations are more likely for larger polymers. Then smaller polymers would be preferentially eliminated, while a small number of resistant polymers would grow faster due to the increased pool of Sup35 monomers. A critical parameter of this model is the difference of Sup35 prion polymers in their susceptibility to Hsp104. If this difference is negligible, one would observe a decrease in the size of Sup35 polymers, as it was assumed earlier. At certain intermediate levels of difference, the size of prion polymers would be unchanged upon Hsp104 overexpression, the situation that was observed here for [PSI ϩ PS ] variants. The data obtained in this work suggest a dual role of Hsp104 in prion propagation: dismantling of prion aggregates into smaller pieces and fragmentation of prion polymers constituting these aggregates. This may result in three types of events important for the prion propagation and inheritance. New prion seeds may be generated by separating one or several polymers from an aggregate or by cleaving off a fragment of one polymer. New polymers may appear without new seed generation when a polymer is cleaved but remains attached to the aggregate.
While this work supported the amyloid-like nature of the yeast prions, the final proof of it could be the electron microscopic observation of purified polymers. We are currently attempting such purification, based on the methods and conditions described in this work. It would also be of interest to use the approaches described here to analyze other prion and amyloid proteins and especially the mammalian prion, PrP.