4.1. Overproduced Hsp104 Acts Differently at the Two Levels of Yeast Prion Structure
Regrettably, many studies of yeast prions do not differentiate between prion fibrils, prion aggregates, and propagons. We observed that the Sup35 prion particles represent higher order aggregates composed of a number of relatively short fibrils made up of 10–50 Sup35 protomers, as well as some additional proteins
[44]. A propagon is a genetic entity defined by the assay developed by Cox et al.
[104] where the multiplication of prion particles is supposed to be blocked by GuHCl and these particles are allowed to segregate to individual cells, while an original cell divides for several generations and forms a small colony. This colony is then spread to single cells on a plate that lacks GuHCl and the propagons are then counted as the number of colonies retaining prions. Thus, propagons should be equivalent to prion higher order aggregates, provided that these aggregates neither merge nor split during their growth in the presence of GuHCl. GuHCl fully blocks the fragmentation of prion fibrils
[44], but the same was not strictly shown for aggregates. Propagons are clearly different from prion fibrils. Strong [
PSI+] fibrils include, on average, about 20 protomers
[44]. A yeast cell with strong [
PSI+] variants contain about 80,000 Sup35 molecules
[105], 4000 fibrils, and 200–1000 propagons
[65][106]. Thus, each propagon from a strong [
PSI+] contains, on average, 4 to 20 Sup35 fibrils. While fibrils have a strong amyloid structure that is insoluble in SDS, higher order aggregation is based on some weaker interactions that are sensitive to SDS
[44].
Our old experiments revealed paradoxical details regarding the effect of overproduced Hsp104 (a multicopy of
HSP104 on a glucose medium) on Sup35 prion aggregates. The size of prion aggregates (as revealed by centrifugation) substantially decreased, while the size of prion fibrils (as assayed by agarose gel electrophoresis) increased or remained constant (
Figure 5). These effects were observed for strong [
PSI+] variants and artificial prions with the Sup35 N domain from the yeast
Pichia methanolica [44][107], while weak [
PSI+] variants could not propagate in such conditions
[45]. The differential fate of aggregates and fibrils was not understood previously, but it can be explained now with the aid of later observations. We observed that the Sup35 N and M domains in their prion state contained both amyloid structured and unstructured regions
[45]. The latter are likely to mediate the aggregation of prion fibers either directly or through chaperones, since the aggregation is sensitive to Hsp104. The fibril-fragmenting activity of Hsp104 is likely to be restricted not by Hsp104 itself, but by the limited amounts of Sis1 and Ssa that are able to bind to a fibril due to the very close spacing of protomers in amyloids (
Figure 3). The proportion of Ssa1 and Ssa2 proteins to Sup35 observed in ex vivo [
PSI+] prions is about 1:2
[73], which is similar to the maximal value observed for α-syn amyloids
[95]. Thus, it appears that Sup35 prion fibrils are already saturated with Ssa, since they have the same 0.48 nm protomer spacing as α-syn amyloids. The Hsp104 fragmenting activity cannot be increased since it depends on the amount of prion-bound Ssa, which also cannot be increased. This could explain why the size of the Sup35 fibrils, in many cases, does not change in response to the strong overproduction of Hsp104. The interactions observed between fibrils in the higher-order aggregates do not have such restrictive architecture and, thus, can be responsive to an increase in Hsp104.
Figure 5. Overproduced Hsp104 disassembles Sup35 prion aggregates, but not fibrils. Hsp104 was either endogenous, or produced from a multicopy plasmid under control of the native promoter (~30-fold excess). (
A) Sup35 aggregates of yeast lysates fractionated by centrifugation. Sup35PS, [
PSI+PS1], and [
PSI+PS2] relate to Sup35 with prion domain from yeast
P. methanolica and M and C domains from
S. cerevisiae and its two prion variants. (
B) SDD-AGE of Sup35 prion fibrils. (
A,
B) Western blot staining for Sup35NM. Data reproduced from previous research
[107] (
A),
[44] (
B).
4.2. Elimination of Yeast Prions by Hsp104
Regarding the attempts to disassemble amyloids in animal models with the help of Hsp104 that was described in the last chapter, it is of interest to consider the Hsp104-mediated curing of prions in yeast. The curing data are paradoxical. Endogenous Hsp104 can cure a high proportion of [
PSI+] variants that emerge in the presence of the weakened Hsp104 mutant, T160M
[42][108], and mildly overproduced Hsp104 can cure some [
PSI+] variants
[58], but some other variants are not cured even at high Hsp104 overproduction
[45].
The data on prion curing by strongly overproduced Hsp104 are contradictory. Though initially it appeared that Hsp104 overproduction should proportionally increase its prion fragmenting and disaggregating activities
[29], this is apparently not so. It is often stated that [
PSI+] may be the only yeast prion that can be efficiently cured by excess Hsp104, since [
URE3] and [
PIN+] are resistant to such impacts
[109][110]. However, some level of [
URE3] curing has been shown in later work
[111]. Among [
PSI+] variants, weak ones are readily cured by excess Hsp104, while strong [
PSI+] variants are more resistant. This difference depends somewhat on the specifics of Hsp104 overproduction. When the
HSP104 gene was expressed from a multicopy 2-micron plasmid (this results in about 30-fold Hsp104 overproduction
[112]), in the testing of eight weak and seven strong [
PSI+] isolates, all the weak ones were fully cured, while all strong ones were fully resistant
[45]. Production under a low copy inducible
GAL1 promoter results in a similar 20–40 fold Hsp104 excess, but it cures a strong [
PSI+] variant almost completely in ten generations of growth
[106]. The difference in the curing of strong [
PSI+] variants appears to depend on the carbon source (galactose versus glucose)
[45]. We also observed (unpublished data) that the absence or the malfunction of mitochondria allows the complete curing of strong [
PSI+] variants by the multicopy of
HSP104 on glucose.
The easier curing of weak Sup35 prions appears paradoxical, since they should be less efficiently recognized and disassembled by chaperones compared to the strong ones, as suggested by their larger fibrils
[44] and higher mechanical strength
[56][90]. Greene et al. proposed that the curing of Sup35 prions mainly occurs through the Hsp104-mediated trimming of prion fibers from the ends. The extraction of Sup35 protomers from the prion ends does not generate new prion particles and it is likely to proceed relatively fast, since the ends of a fiber could be more accessible and the terminal protomers are easier to unfold as they are joined to just one other protomer, rather than two. These authors tried to confirm this idea by microscopic observation of the dissolution of the aggregates of GFP-labeled prion particles, after the start of Hsp104 overproduction, in several reviewed works
[113]. However, this approach does not appear adequate for the question. The visible prion–GFP foci represent large aggregates, including many prion fibrils
[44]. As we already noted, overproduced Hsp104 can disassemble these aggregates into smaller particles, down to single fibrils
[107] that are undistinguishable from monomers by light microscopy, without decreasing the size of these fibrils
[44]). The best way to observe trimming would be to monitor the size of prion fibrils by agarose electrophoresis, called SDD-AGE
[44], but this was not done.
Serio et al. proposed that Sup35 prion fibrils smaller than a certain threshold size may be unstable and would be dissolved, and that such a threshold is higher for weak prions. Hsp104 action not just at, but also near, prion fiber ends would act as a trimming process, generating no new prion particles; the higher the minimal size of weak Sup35 prions could explain their easier curing by Hsp104. However, while a mathematical model of this effect was provided, the experimental evidence in support of this idea
[114] does not appear sufficient. It is also unclear why weak Sup35 prions, which are mechanically sturdier than strong prions
[56] and less well-recognized by chaperones, require more protomers to stabilize their minimal particles.
Another more convincing group of data, although still partly contradictory, relates the [
PSI+] curing by overproduced Hsp104 to prion malpartition, rather than dissolution. The initial observation was that Hsp104 can bind directly to the Sup35 M domain region 129–148, which is important for curing, since Sup35 prions lacking this region are not cured by Hsp104 overproduction
[33].This Hsp104 binding is independent of Hsp40 and Hsp70 and, in relation to this, nonproductive
[66]. In line with this, in vivo Hsp104 shows two types of binding to Sup35 prions: one is labile, Hsp70-dependent, and exhibits a free exchange of Hsp104 with the pool of monomers; the other is a stable binding to the Sup35 M region that shows little exchange
[115]. The latter could be termed a direct and non-productive binding. Such a binding should sterically interfere or compete with the binding of different chaperones belonging to the productive fragmentation pathway, due to the relatively small space where such interactions occur (
Figure 3), and this could thus reduce prion fragmentation. In agreement with this, overproduced Hsp104 can substantially increase the size of strong Sup35 prion fibrils (
Figure 5B)
[44]. In contrast, Ness et al observed that Hsp104 overproduction under a low copy
GAL1 promoter did not impair fragmentation and did not alter Sup35 oligomer size. In their setup, the curing of strong [
PSI+] variants, which occurred at a rate of about 10% per generation, was shown to be due to prion retention in a proportion of mother cells. These authors proposed that Hsp104 mediated prion binding to some subcellular structures, thus causing their malpartition
[106]. Regrettably, such a study was not made for weak [
PSI+] variants where curing would be more pronounced.
[
PSI+] variants can also be cured with moderate efficiency by short-term heat shock. Such a shock increases the proportion of Hsp104 to Ssa proteins, and the prion curing also occurs through the asymmetric segregation of the prion
[116][117].
Notably, the weak [
PSI+] variant shows a larger proportion of stable Hsp104 binding to the Sup35 M domain
[115], which can explain the easier curing of weak [
PSI+] variants. In turn, the more efficient direct binding to weak Sup35 prions could be explained by our recent data. The Hsp104 target region of Sup35 is 129-148, which coincides well with amyloid Core 3 (124–153) that forms in the majority of strong [
PSI+] variants but is rarely found in weak [
PSI+] variants
[45]. It is reasonable to assume that Hsp104 can bind the target region in its unfolded state, but not while in an amyloid fold, so non-productive Hsp104 binding is less likely in strong [
PSI+] variants.
On the other hand, the productive binding of Hsp104 through Sis1 and Ssa should be more efficient in strong [
PSI+] variants. These chaperones bind to the unfolded regions of the Sup35 N domain, rich in tyrosine, which stimulates fragmentation best
[57][118]. Such regions are smaller in weak Sup35 prions due to the presence of Core 2 (~90–123) amyloid structure that is rare, or less pronounced, in strong [
PSI+] variants
[45]. Thus, the reduced productive and increased non-productive Hsp104 binding could define the easier curing of weak [
PSI+] variants by overproduced Hsp104.
Thus, despite some experimental discrepancies, the only confirmed mechanism of [PSI+] curing by overproduced Hsp104 is the inefficient partitioning of propagons in cell divisions, due to either reduced prion fragmentation, or Hsp104-mediated prion anchoring. Prion curing through its dissolution currently lacks sufficient evidence. Curing by normal Hsp104 levels of the prions generated in the presence of mutant Hsp104 is of great interest in this respect, but its mechanism has not been studied in detail.
4.3. The Therapeutic Potential of Protein Disaggregases
While the human Hsc70 system can efficiently disaggregate toxic oligomers and short amyloid fibrils, its activity against large, less toxic amyloid aggregates is severely impaired
[119]. Yeast Hsp104 is a much more powerful disaggregase than the Hsp70 (Ssa)-Hsp40 (Sis1) system alone
[94], but animals have no Hsp104 homolog
[91]. This raised hope that Hsp104 can disaggregate pathological amyloids if reintroduced to animal cells
[120][121]. Luckily, Hsp104 can efficiently collaborate with the animal disaggregation machinery and strongly improves the reactivation of heat-denatured luciferase
[94][122]. In vitro, Hsp104 can dissolve fibrils associated with human diseases: amyloid β, α-syn, prion protein, tau, amylin, and polyglutamine
[90][122][123]. In animal disease models, Hsp104 reduces polyglutamine toxicity in
Caenorhabditis elegans, fly, and rodent models
[124][125][126][127]. In a rat model of Parkinson’s disease, Hsp104 reduced the formation of phosphorylated α-syn inclusions and prevented nigrostriatal dopaminergic neurodegeneration
[122].
However, the activity of yeast Hsp104 against pathological aggregates can be insufficient even at high levels of Hsp104
[90]. This prompted attempts to enhance Hsp104 disaggregation activities, and this was, surprisingly, achieved through minor changes and even single missense mutations
[128]. Potentiated Hsp104 variants have been developed and are capable of suppressing toxicity associated with α-syn, TDP-43, and FUS in yeast
[129][130][131][132][133]. Many such potentiating mutations were found in the regulatory coiled-coil middle (M) domain of Hsp104 (residues 411–538), which mediates interactions of Hsp104 with Hsp70. Hsp104 potentiation often correlates with the destabilization of the M domain. However, some of these mutations show off-target toxicity. To overcome this problem, scanning mutagenesis of the M domain was performed
[134], as well as mutagenesis of Hsp104 NBD1 and NBD2
[129][133], which allowed the isolation of non-toxic potentiated Hsp104 mutants. The screening of a cross-kingdom collection of Hsp104 homologs in yeast proteotoxicity models revealed that prokaryotic ClpG reduces TDP-43, FUS, and α-syn toxicity, whereas prokaryotic ClpB is ineffective. The latter is not surprising, since Reidy et al. showed that bacterial ClpB did not properly interact with yeast chaperones and required its bacterial partner chaperones to function
[135]. Distinct eukaryotic Hsp104 homologs were uncovered that selectively antagonized α-syn condensation and toxicity in yeast and dopaminergic neurodegeneration in
C. elegans. Surprisingly, this therapeutic variation does not manifest as enhanced disaggregase activity, but rather as an increased passive inhibition of the aggregation of specific substrates
[136]. An Hsp104 variant that is efficient against TDP-43, α-syn, and polyglutamine, that lacked toxicity, was obtained from the thermophilic fungus
Calcarisporiella thermophila [137].
While animals lack the mitochondrial Hsp104 homolog, Hsp78, disaggregation in mitochondria can be performed by Skd3, another chaperone of the AAA+ family. Skd3 shows a homology with bacterial ClpB, but has only one NBD. Mutations in Skd3 that reduce its disaggregating activities are associated with 3-methylglutaconic aciduria, a severe mitochondrial disorder. Thus, Skd3 is a potent mitochondrial protein disaggregase which can be used for treating mitochondrial protein aggregation
[138].
AAA+ proteins that do not belong to the Hsp104 family can also be used to counteract toxic protein misfolding in animals. One such protein is archaeal PAN, an unfoldase homologous to the eukaryotic proteosomal 19S particle. PAN associates with the 20S catalytic particle and unfolds substrates before their degradation
[139]. A PAN variant was recently constructed with a C-terminal FLAG epitope tag (PANet), which impedes PAN interactions with the 20S proteasome, but does not affect unfolding. The expression of PANet in rod photoreceptors in a mouse model of retinopathy mitigates photoreceptor degeneration caused by protein misfolding without causing significant side effects
[140]. Thus, protein disaggregases of the AAA+ family have significant therapeutic potential.
It is important to keep in mind that the disaggregation of amyloids can have both positive and negative effects. Negative consequences might occur when large amyloids are broken into smaller pieces, but the latter are not efficiently destroyed. This resembles the situation with yeast prions, which are propagated by Hsp104. In animals, amyloids of smaller size are (1) often more toxic and (2) have a much higher potential for a prion-like spread between cells and tissues. Such problems were highlighted recently by Tittelmeier et al. who observed that reducing disaggregation in
C. elegans by knocking down Hsp110 caused a beneficial decrease of the amounts of toxic α-syn species and a reduction in the intercellular propagation of α-syn aggregates. A similar treatment decreased the amount of polyQ aggregates and their toxic effects in another
C. elegans model
[141]. This suggests that the optimal disaggregation activity would be one that destroys small aggregates, but cannot cope with large ones. Humans have three NEFs for Hsp70, which stimulate the entropic pulling effect to a different extent
[95]. Possibly, optimal activity might be achieved by adjusting levels of these NEFs.
The described data raises certain hopes and reveals some fundamental problems. The most optimistic aim would be the complete dissolution of amyloids, but this has never been shown, even in yeast. Still, some hope comes from observations in yeast where prions that appear in the presence of weakened Hsp104 are less resistant to Hsp104, while human amyloids that appear in the absence of Hsp104 yeast prions can be cured through retention in the mother cell, but such a scenario is not valuable for multicellular organisms, where most cells do not actively divide. However, they possess multiple mechanisms for the intercellular movement of amyloids (reviewed in previous research
[142]) and these are likely to be affected by factors, which influence prion retention in dividing yeast cells.
A central technical problem would be to deliver Hsp104 to every cell or extracellular location containing an amyloid, which does not seem currently feasible. Finally, a tool able to disassemble all amyloids could also disassemble functional amyloids, and, in particular, amyloids that are involved in long-term memory.
Nevertheless, disaggregases of the Hsp104 type can alleviate the symptoms of amyloidoses in some animal models. Although there might not be a single ideal disaggregase, different agents tailored for each type of amyloidosis might be possible. In any case, the most promising strategy would be to adjust disaggregases so that they disassembled more toxic, but less resistant, small amyloids, while leaving less toxic and more resistant larger aggregates intact.