Small Heat Shock Proteins, Big Impact on Protein Aggregation in Neurodegenerative Disease

1Maintenance of cellular protein homeostasis (proteostasis) is crucial for cell function and survival. Neurons are particularly sensitive to dysregulated proteostasis as evidenced by the accumulation and aggregation of amyloidogenic proteins, which are a hallmark of neurodegenerative disease. Proteostasis is regulated by networks of interacting proteins that include translation machinery, secretory pathways, proteases, the ubiquitin-proteasome system (UPS), autophagic machinery, and molecular chaperones. In addition to their vital role in folding nascent proteins, molecular chaperones recognize and triage proteins if misfolding events occur. Chaperones can be classified as holdases (that bind and hold partially folded protein intermediates to prevent their aggregation), foldases (that assist the proper folding of proteins in an ATP-dependent manner), and unfoldases (that convert misfolded proteins into transiently unfolded intermediates to provide an opportunity for spontaneous proper refolding). ATP-dependent chaperones, like the 70 kDa heat shock protein (Hsp70) and the 90 kDa heat shock protein (Hsp90), facilitate refolding, degradation, or sequestration of these misfolded proteins. In mammalian systems, chaperones often work together in multisubunit heterocomplexes, comprised of an ATP-dependent chaperone as well as co-chaperone and accessory proteins. These discrete heterocomplexes allow for specific client selection, the dynamic modulation of these clients by regulating ATPase activity, or, if a protein is unable to refold, delivery to additional proteostasis machinery for degradation or sequestration. Here, we review evidence for modulation of neurodegenerative disease-relevant protein aggregation by a family of molecular chaperones, known as small heat shock proteins (sHsps).

2sHsps are a special class of molecular chaperones that lack an ATPase domain. They are characterized as small proteins, between 12 and 43 kDa, containing a core α-crystallin domain (ACD) flanked by variable N-terminal and C-terminal domains. There are 10 α-crystallin domain-containing mammalian sHsps, designated HspB1 (Hsp27), HspB2 (myotonic dystrophy kinase-binding protein, MKBP), HspB3 (Hsp17), HspB4 (αA-crystallin), HspB5 (αB-crystallin), HspB6 (Hsp20), HspB7 (cardiovascular Hsp, cvHsp), HspB8 (Hsp22), HspB9 (cancer/testis antigen 51, CT51) and HspB10 (outer dense fiber protein 1, ODFP1) (see Table 1 ). In general, sHsps interact with intermediately folded proteins through surface exposed hydrophobic residues in order to stabilize the protein and prevent further misfolding and/or aggregation. Therefore, sHsps act early in the chaperone processing of misfolded proteins, often prior to refolding attempts by ATP-dependent chaperone complexes. The dynamic nature of sHsp structure appears to be an important determinant for client protein binding and chaperone function. sHsp dynamics occur at five levels, including through 1) flexible domains flanking the ACD, 2) polydisperse self-multimerization, 3) multimerization with other sHSPs, 4) subunit exchange, and 5) regulation by the cellular environment including post-translational modifications, which are discussed in more detail below.

3The disordered nature of the N- and C-termini flanking the ACD allows for tertiary flexibility that likely supports dynamic interactions with client proteins as well as other sHsp subunits. Observations of intrinsic disorder propensity based on primary structure are presented in Figure 1. A universal intrinsic disorder propensity within the C-terminal domains appears to be a common trait among human sHsp family members; while disorder propensity among N-terminal domains are more variable. HspB4 and HspB5 (the α-crystallins) distinguish themselves as having the most ordered N-terminal domains. The N-terminal domains of HspB1, HspB2, HspB3, HspB6, and HspB10 have discrete segments of disorder; while HspB7, HspB8, and HspB9 have a more widespread disorder propensity in the variable N-terminal domain ( Figure 1 ). It is notable that, for many of the sHsps, high disorder propensity also extends into the conserved ACDs, which are generally regarded to have a well-structured immunoglobulin-like fold of β-sheets, and may reflect differences in ACD flexibility among sHsps.

4sHsps utilize flexible hydrophobic surfaces to interact with exposed hydrophobic surfaces of client proteins, which are made available during misfolding or denaturation. Both high affinity and dynamic low affinity interactions of sHsps with client proteins have been described. Generally, high affinity interactions are associated with amorphous misfolding clients while sHsp effects on amyloid fibril formation are often due to weak or transient client interactions. Expression of sHsps, with their own exposed hydrophobic surfaces, could potentially pose a cellular proteostasis problem of its own. One way for cells to address this problem is to restrict expression of sHsps until stress events (like heat shock or oxidative stress) necessitate chaperone activity. Another mechanism is to sequester exposed hydrophobic regions into large, dormant, multimeric structures (sHsp oligomers); keeping sHsp chaperones poised to handle early misfolding events prior to transcription and translation of other stress-inducible chaperones. Indeed large, inactive, oligomeric sHsps disperse into smaller oligomers as they become active. This transition can be precipitated by stress events, post-translational modifications, and by direct competition of client proteins with sHsp oligomeric interactions.

5Dynamic sHsp homodimers and polydisperse, homooligomeric structures regulate chaperone function as well as client protein specificity and binding. Many factors regulate the size of these structures. sHsp expression (or concentration) itself alters the size of oligomers in vitro. HspB1 and HspB5 each form higher order homooligomers of greater than 20 subunits in vitro, while other family members generally form smaller oligomers or dimers. Moreover, sHsp homooligomers with a discrete number of subunits may actually be composed of a variety of dynamic structures, since polymorphic arrangement of the dimer and oligomer subunits have been reported.

6Chaperone activity is also regulated by the formation of dynamic sHsp heterodimers and heterooligomers that undergo continuous subunit exchange. Native α-crystallin in the eye lens is made up of polydisperse HspB4/HspB5 heterooligomeric complexes. The crystal structure of a tetrameric HspB2/HspB3 heterocomplex, which is found in muscle cells, was recently reported. While a complete picture of the physiologically relevant heterooligomeric sHsp combinations has not been elucidated, in vitro evidence for a variety of heterooligomeric sHsp groupings is emerging, including: HspB1/HspB5, HspB1/HspB6 and HspB2/HspB6 heterooligomeric complexes as well as HspB8 interactions with HspB1, HspB2, HspB5, HspB6, and HspB7. The relative expression levels of discrete sHsp family members will influence the complement of oligomers resulting from subunit exchange, which may guide an interminable number of oligomeric combinations.

7Changes in the cellular environment and post-translational modifications also regulate chaperone activity by affecting the structural and multimerization dynamics of sHsps. Environmental influencers include alterations in temperature and pH. The most commonly studied post-translational modification of sHsps is phosphorylation within the N-terminal domain, which is generally considered to decrease oligomer size, but opposite effects have been reported. Phosphomimetics of HspB5 have been shown to form smaller but more polydisperse oligomers.

8Dynamic sHsp oligomeric structures would be expected to reach equilibrium in a constant environment, but the current literature paints a picture of something more complicated. Since sHsp dynamics are sensitive to many factors, there may be no set equilibrium in the context of an ever-changing cellular environment. This dynamic nature of sHsps can make pinning down a specific structure difficult; and any identified structure of a single oligomeric conformation will likely represent only a snapshot of the evolving assembly. Continuously morphing sHsp oligomers may facilitate the broad recognition of clients that this family of molecular chaperones regulates.

9Of the 10 mammalian sHsp proteins, HspB1, HspB5, and HspB8 are expressed in the brain, as well as, HspB6, and HspB7, albeit at lower levels. HspB2 and HspB3 mRNA expression have also been detected at relatively low levels in adult brain ( Table 1 ). HspB4 expression appears to be restricted to the eye lens, while HspB9 expression is restricted to testes and heart, and HspB10 expression is restricted to testes, eye, and muscle; these sHsps show no or negligible gene and protein expression in the brain under normal conditions. In addition to the difference in tissue specificity, the regulation of expression is also varied among the HspB family members. Heat shock stress upregulates the expression of HspB1, HspB5, and HspB8. However, in vitro HspB8 responds to heat shock in only a subset of cell types, suggesting that stress-induced expression may not be a ubiquitous phenomenon across all cell types in vivo. Brain expression of sHSP family members may be induced by other stressors. For example, HspB1, HspB5, HspB6, and HspB8 were upregulated in response to oxidative stress and hyperosmotic stress in rat hippocampal neurons ( Table 1 ). Surprisingly, HspB4, which is not found in normal brain tissues, has been reported to have detectable mRNA and protein expression in cultured primary rat astrocytes, even in the absence of stress induction. It is unclear whether this is reflective of HspB4 expression in a subpopulation of normal brain cells or if expression was induced by culture conditions; either way it is clear that sHsps not normally found in brain tissue have the potential to become expressed in response to stressors, which may include pathological situations. HspB1 and HspB5 are more predominantly expressed in glial cells under basal conditions, rather than in neurons. Multiple reports demonstrate HspB1 and HspB5 upregulation in neurodegenerative disease. For example, in the AD brain, along with other tauopathies including PSP, CBD, and FTDP-17, HspB1 and HspB5 are upregulated in reactive glial cells and can be found to colocalized with glial tau inclusions. HspB5 upregulation in the astrocytes of ALS patient spinal cord and HspB1 upregulation in a transgenic mouse model of ALS have been reported.

10HspB1 and HspB5 share many characteristics that differentiate them from other sHsps (see Figure 2 ). They both readily form higher order homooligomers of 20+ subunits, while other family members generally form smaller oligomers of less than 10 subunits or dimers. Comparison of the number of human sHsp-interacting proteins also reveals that HspB1 has the largest interactome followed by HspB5 and HspB2 (BioGRID 3.5.170) ( Figure 2 ). HspB1 and HspB5 also exhibit chaperone activity across a wider spectrum of model aggregating proteins, compared to other sHsps ( Figure 2 ). HspB1 and HspB5 subunits comprise ubiquitous large, homooligomeric and, perhaps, heterooligomeric complexes, while other family members appear to function as smaller (under 10 subunits) sHsp oligomers. Alternatively, sHsps that form smaller homooligomers may become incorporated into large sHsp heterooligomers through subunit exchange that correlates with their expression level and sHsp interactome; imbuing new interaction preferences for clients as well as additional proteostasis machinery. Known physical interactions between sHsp family members are summarized in Figure 3 (BioGRID 3.5.170).

11sHsps interact with numerous client proteins, especially under stress conditions. Therefore, it can be difficult to determine whether cellular effects of sHsps result from direct interactions with a given client protein or indirectly through modulation of other proteins and/or signaling pathways. Indeed, cytoprotective roles for sHsps have been described, including inhibition of apoptotic signaling pathways, inhibition of neuroinflammation, and buffering against damage from oxidative stress. Therefore, sHsps will undoubtedly affect cellular proteostasis at multiple levels.

12Direct sHsp interactions with misfolded proteins can prevent further irreversible misfolding events and facilitate client triage by other chaperones that more actively influence client fate through specific proteostasis pathways. Generally, HspB1, HspB4, and HspB5, which are all heat shock responsive, are commonly reported to play a role in substrate refolding, while HspB6, HspB7, and HspB8 have been shown to preferentially promote the degradation of clients through autophagy. However, this simple categorization of HspBs does not hold true in many instances, including many examples discussed in this review. HspB8, the best characterized sHsp in regard to proteostasis mechanisms, forms a complex with the co-chaperone BCL2-Associated Athanogene 3 (Bag3), Hsp70, and C-terminus of Hsc70-interacting protein (CHIP), which appears to drive aggresome formation and promote autophagic degradation of client proteins. HspB8 overexpression has been reported to lead to enhanced proteasome expression and activity as well as influence its subcellular localization, at least in cardiac tissues. The authors also demonstrated that HspB8 was found in complex with proteasomes together with HspB1 and other chaperones. This may have a general, indirect effect on the degradation of aberrant proteins. HspB1, HspB5, and HspB6 have also been shown to interact with Bag3. In neurodegenerative disease, Bag3 and HspB8 are upregulated in astrocytes, which may help clear protein aggregates. HspB6 has been reported to regulate autophagy through direct interaction and stabilization of Beclin 1, a key regulator of autophagosome formation. Hsp90 inhibition and proteasome inhibition both activate the heat shock response (;, 90), which upregulates HspB1, HspB4, and HspB5. Proteasome inhibition has also been shown to upregulate HspB8. Heat shock response has also been shown to promote HspB1 phosphorylation, regulate its oligomerization state, and enhances UPS activity. Upregulated HspB1 in response to heat shock has been reported to promote proteasomal degradation of certain client proteins. Bag1 competition for Bag3 occupancy in Hsp70/CHIP complexes drives client degradation through the UPS. Aging and oxidative stress correlates with a switch in expression from Bag1 to Bag3 that may shift the client proteins from UPS-mediated degradation to an autophagic pathway. Expression and activity of sHsps may influence shuttling misfolded proteins between proteasomal and autophagic degradation pathways, allowing for the alternative system to be utilized when the primary degradation pathway for a given client is impaired or insufficient.

13Chaperone activity of sHsps was first demonstrated by suppression of stress-induced formation of amorphous aggregates. The first observation that sHsps also inhibit amyloid fibril formation under physiological conditions in vitro was reported using apolipoprotein C-II. Since then, a role for sHsp modulation of amyloidogenic protein aggregation has been reported for a variety of client proteins relevant to neurodegeneration. Here, we focus on selected amyloidogenic proteins in neurodegeneration with evidence for direct sHsp interactions and/or chaperone activity ( Figure 3 ). However, it should be noted that for cellular and in vivo applications, direct chaperone activity may be only part of the sHsp-associated mechanisms modifying aggregation or toxicity. Moreover, in disease, while many chaperones still support a healthy environment for the cell, certain chaperone complexes appear to exacerbate the problem of protein misfolding by stabilizing or in some cases promoting the stabilization and accumulation of toxic, prefibrillar oligomers.

14Amyloid β (Aβ) peptides are cleavage products of the amyloid precursor protein (APP) that are prone to aggregate into toxic oligomers, amorphous aggregates, and fibrils; extracellular deposition of Aβ-rich plaques is considered a hallmark of AD. Multiple sHsps have been reported to affect Aβ, including HspB1, HspB2, HspB5, HspB6, and HspB8. HspB1, HspB2, and HspB6 associate with senile plaques in AD brain tissue. In vitro associations between Aβ and HspB1, HspB5, HspB6, and HspB8 have been described. While a connection between HspB1 and Aβ1-42 has been reported, little is known about the molecular details of this interaction. What is known is that, in cellular systems, extracellular Aβ1-42 induces the expression of HspB1; inversely, exogenous exposure of extracellular HspB1 dose-dependently ameliorated Aβ1-42 cellular toxicity. Recombinant protein studies have suggested that the role of HspB1 is to affect Aβ quaternary structure, sequestering toxic oligomers into less toxic, larger aggregates; while other studies have demonstrated a reduction in Aβ aggregation. A more detailed interaction between HspB5 and Aβ has been described. HspB5 interacts with the hydrophobic core of Aβ1-40 and competes for Aβ–Aβ interactions HspB5 association with Aβ1-40 prevented amyloid fibril-associated cellular toxicity. HspB5 also binds to Aβ1-42 with micromolar affinity and inhibits Aβ fibril elongation and the nucleation of Aβ seeds. This work demonstrated that the N-terminal domain of HspB5 was not required for Aβ binding nor inhibition of Aβ fibrillation, whereas the β4–β8 groove of the ACD appears to be crucial for this interaction. HspB6 interacts with Aβ and the physical interaction was mapped to residues adjacent to the KLVFF oligomerization domain within Aβ. This interaction was strengthened by HspB6 N-terminal domain phosphorylation (;. Similar to HspB1, exogenous, extracellular HspB6 also reduced Aβ1-42 cellular toxicity. In addition, it was demonstrated that phosphorylation of endogenous HspB6 correlated with an attenuation of Aβ1-42 toxicity. A direct interaction of HspB8 with Aβ1-40 and Aβ1-42 was demonstrated using surface plasmon resonance (SPR), with a significantly greater affinity to Aβ1-40 with the E22N Dutch mutation that correlates with cerebral amyloid angiopathy. In vivo studies using a cross between HspB2/HspB5 knock-out and mutant APP mice revealed locomotor and sensory deficits, suggesting a negative synergy when these genotypes are combined. However, evaluation of changes to amyloid deposition was not possible, since the knock-out mice did not survive to an age wherein amyloid deposits can be detected. More work is needed to fully understand the effects of discrete sHsps or sHsp combinations on Aβ deposition and toxicity.

15The microtubule-associated protein tau (tau) is an intracellular intrinsically disordered protein that binds to and stabilizes axonal microtubules. Tau is the major constituent of neurofibrillary tangles (NFTs), protein aggregates that form in neurons and contribute to neurodegeneration seen in more than 20 tauopathic diseases, including AD. Only two sHsps have been characterized for their interactions with tau, HspB1 and HspB5. Interestingly, both HspB1 and HspB5 are upregulated and associated with pathological NFTs in tauopathies. HspB1 has been shown to bind directly to tau and reduce aggregation. HspB1 targets early tau aggregates and delays aggregation, but does not inhibit the fibril elongation process once started. As recently revealed by nuclear magnetic resonance (NMR), HspB1 recognizes two aggregation-prone regions within the microtubule binding repeat regions of tau via both the N-terminal domain and the conserved β4–β8 cleft in the ACD. Interestingly, this is the same surface used for sHsp dimerization as well as interaction with certain other clients and co-chaperone proteins. The fact that HspB1 binds to several different proteins in the same region, and that this binding is relatively weak, suggests that competitive protein–protein interactions may regulate its chaperone activity toward tau. This supports a hypothesis for large, inactive sHsp oligomers that convert to smaller, active oligomers through competitive tau interactions. Phosphorylation of the N-terminal domain, or the introduction of specific pseudophosphorylation mutations, reduces the size of HspB1 oligomeric structures and increases chaperone activity. HspB1 and a pseudophosphorylated mutant both bound tau and reduced or delayed tau aggregation in vitro, albeit at different levels. Although there are no reports of a direct interaction of tau and HspB8, Bag3 overexpression in neurons reduces tau aggregation; which may implicate a role for HspB8 and autophagy. Additionally, tau predominantly aggregates in excitatory neurons, with lower Bag3 and HspB8 expression, relative to inhibitory neurons. HspB1 overexpression was evaluated in brain using a transgenic mouse model of tau aggregation via intracranial injection of an adeno-associated viral (AAV) vector. In this model, wild-type HspB1 reduced neuronal tau levels and prevented long-term potentiation (LTP) dysfunction, a cellular correlate of learning and memory. The pseudophosphorylated mutant had opposite effects; increasing neuronal tau levels and failing to prevent the LTP deficit. These results suggest that while reduction of sHsp oligomer size increases chaperone activity in vitro, sHsp oligomeric complex dynamics are crucial for proper chaperone activity in vivo. Further studies are needed to test other sHsps for their interactions with tau and for their effects in tauopathic mouse models.

16α-Synuclein (α-syn) is an intrinsically disordered neuronal protein associated with the regulation of synaptic vesicles. It is also the primary structural component in Lewy Bodies, large protein aggregates that develop in neurodegenerative synucleinopathies, such as Parkinson’s Disease (PD) or dementia with Lewy Bodies (DLB). Point mutations (A30P, E46K, H50Q, and A53T) or whole gene triplication of the SNCA gene, encoding α-syn in humans, are associated with the familial forms of early onset PD. sHsps interact with multiple species of α-syn, from monomeric partially folded intermediates to fully formed fibrils, modulating both aggregation and cellular toxicity. HspB5 was the first sHsp reported to bind α-syn with low affinity and has been known for some time to inhibit amyloid fibril formation. It has also been reported that α-Syn can interact with at least 5 different sHsps as determined by SPR, with varying affinities; in order of highest affinity for α-syn: HspB5 (Kd = ∼60 nM) > HspB1 and HspB6 (∼400 nM) > HspB2 and HspB8 (∼4 µM). Each of these five sHsps delayed or inhibited α-syn fibril formation in vitro. Remarkably, HspB8 had the most robust inhibition of α-syn fibrillization in vitro, but had one of the lowest affinities for α-syn. HspB1 and HspB5 reduced α-syn aggregation, which was attributed to weak transient interactions, similar to the relationship of HspB1 with tau. This suggests that transient, dynamic sHsp interactions may best ameliorate fibril formation. While HspB5 and HspB8 reduced fibril formation for wild-type α-syn as well as for three familial PD-linked mutants, other sHsps showed more varied effects on discrete α-syn mutants. Surprisingly, HspB2 and HspB6 actually increased the formation of α-syn A53T fibrils. A recent study found that reduction of the disulfide-bond at the HspB1 β4–β8 dimer interface increased its chaperone activity toward multiple client proteins, including α-syn; perhaps implicating this interface as an important surface for interactions with α-syn. This study also demonstrates that the ACD region appears rigid in the dimer, but becomes dynamic and partially unfolded as a monomer; this is consistent with the predicted disorder propensity of the HspB1 ACD shown in Figure 1. HspB5 expression rescued a rough eye phenotype in a transgenic α-syn fly model. However, HspB5 overexpression was shown to promote mutant α-syn accumulation in a mouse model of PD through suppression of astrocytic autophagy, while knock-down promoted α-syn clearance. Additional in vivo studies are needed to fully understand the relationship between sHsps and α-syn variants.

17Genetic expansion of polyglutamine (polyQ) tracts in certain proteins promotes aggregation into fibrillar structures and the death of neurons in several diseases, including spinocerebellar ataxias (via ataxin proteins), spinal and bulbar muscular atrophy (via the androgen receptor protein), and Huntington’s disease (via the huntingtin protein). While sHsps have been implicated in an amelioration of aggregate formation in a number of cellular polyQ models, direct interaction between sHsps and polyQ domains has not been reported. This is not surprising as the polar polyQ tract would not directly provide hydrophobic surfaces for sHsp–client interactions. Effects of sHsps on polyQ toxicity may be attributed to direct interactions or secondary effects through an accentuation of autophagy via the interaction of HspB8 with the co-chaperone Bag3 or a reduction in oxidative stress. HspB5 has been reported to bind to the Josephin domain flanking the polyQ expansion of the ataxin-3 protein, slowing polyQ-driven aggregation. A protein engineering strategy similar to the action of HspB5, targeting polyQ-flanking regions with engineered single-chain antibody fragments, also successfully inhibited mutant huntingtin aggregation. Vos et al. performed an elegant comparison of all 10 sHsps on the aggregation of polyQ expanded huntingtin and ataxin-3, which demonstrated that HspB7 most potently reduced polyQ aggregation in a proteasome-independent manner. The well-studied and generally promiscuous sHsps, HspB1 and HspB5, were surprisingly not very effective at reducing polyQ aggregation. It is not yet known whether any of the polyQ huntingtin-active sHsps target a flanking region or act solely through secondary effects. In another study, HspB1 overexpression in a N-terminal huntingtin-fragment transgenic mouse model (R6/2) was unable to alter disease progression or huntingtin aggregation. However, astrocyte-specific HspB5 overexpression in a full-length huntingtin transgenic model (BACHD) showed neuroprotective effects and slowed disease progression. Moreover, HspB7, ameliorated an eye degeneration phenotype in a fly model expressing polyQ expanded ataxin-3 and has yet to be evaluated in mammalian polyQ disease models. More work is needed to understand the direct and indirect actions of sHsps on polyQ expanded aggregates in vivo.

18Superoxide dismutase-1 (SOD-1) is a free radical scavenging metalloenzyme that can contain aggregation-prone mutations linked to familial amyotrophic lateral sclerosis (ALS), a neurodegenerative motoneuron disease. HspB1, HspB5, and HspB8 can all affect the accumulation of this protein. HspB1 and HspB5 can directly interact with SOD-1 in vitro and suppress the aggregation of both wild-type and the familial ALS-linked G93A mutant SOD-1. As with polyQ containing proteins, HspB8 has been shown to facilitate the autophagic removal of misfolded or aggregated SOD-1. In this context, HspB8 interacts with a multimeric complex that also includes Bag3, Hsp70, and CHIP, however there is not yet any direct evidence for direct HspB8 interactions or “holdase” chaperone activity for SOD-1. Similar observations of HspB8-linked enhancement of autophagic clearance have been made for another ALS-linked protein, transactive response DNA-binding protein 43 kDa (TDP-43), without evidence for a direct interaction. However, there is evidence via high throughput co-fractionation data for an interaction of TDP-43 with HspB1. Two other studies evaluated overexpression of HspB1 in a transgenic SOD-1G93A model of ALS; one showed no effect on disease progression, while the other showed a modest protection at early stages of disease, which were not sustained and resulted in no improvement at later stages. Likewise, HspB5 knock-down in a transgenic SOD-1 mutant mouse model did not increase SOD-1 aggregation or disease progression. However, differential expression of HspB5 ALS transgenic mouse models with varied background correlated with slower disease progression in these animals. The effects of other sHsp family members on ALS-related aggregates have not been reported.

19In addition to intracellular sHsp dynamics, some sHsps have been reported outside of the cell. HspB1, HspB5, and HspB6 have been found in the extracellular environment, in the absence of a traditional secretory signal sequence, and there is evidence for both exosomal and endolysosomal secretion mechanisms. Elevated HspB1 protein levels in serum have been detected in multiple sclerosis and in both plasma and cerebrospinal fluid during ischemia. The presence of HspB5 autoantibodies in the serum of AD and PD patients provides additional evidence for the presence of extracellular sHsps in neurodegenerative disease. sHsp secretion appears to be regulated by post-translational modifications. A phosphorylation deficient HspB1 mutant colocalized with the lysosome-associated membrane protein-1 (LAMP-1) and was preferentially secreted; while a pseudophosphorylated mutant demonstrated decreased secretion and decreased LAMP-1 colocalization. Evidence for functional roles of these secreted sHsps are emerging. In a simple case of direct interaction, sHsps would need to be located outside of cells in order to associate with and reduce toxicity of extracellular misfolded proteins, like Aβ; which has been described for HspB1, HspB5, and HspB8 in extracellular, classic, senile plaques in AD brain. However, extracellular sHsps have also been implicated in cell-to-cell communication, signaling, inflammation, and immunity, which suggests multiple indirect effects on neuronal health. For example, extracellular HspB5 functions as a stroke neuroprotectant by modulating the immune system.

20Meanwhile, aggregation prone proteins that are traditionally considered to be intracellular, like tau, α-syn, TDP-43, and huntingtin, are also found to be secreted into the extracellular milieu, which may be a mechanism that enables pathological spreading from cell to cell. There is growing evidence that this cell-to-cell transfer contributes to the anatomical pattern of pathology progression to distinct brain regions that typify specific neurodegenerative diseases. Both sHsps and neurodegenerative amyloids are reported to be secreted from cells through similar mechanisms of exosomal release or endolysosomal secretion; other mechanisms, including direct transport and tunneling nanotubes, may also contribute. Physical interactions between sHsps and amyloidogenic proteins may also suggest that this class of molecular chaperones could play a regulatory role in their non-canonical secretion or perhaps buffer against the spread of toxic aggregates to additional cells.

21There is growing evidence that formation of membrane-less organelles can be influenced by some aggregation-prone proteins, including proteins linked to neurodegenerative disease. Membrane-less organelles are viscous, liquid, subcellular compartments containing proteins and often RNA that develop as localized pockets of condensed cellular material formed by a process called liquid–liquid phase-separation (LLPS). Alterations in pH, ion concentrations, temperature, oxidative stress, molecular crowding, and post-translational modifications can trigger molecular supersaturation causing LLPS. Proteins with intrinsically disordered regions can also drive LLPS, due to conformational flexibility that allows for a multitude of transient multivalent interactions. AD, ALS, and frontotemporal dementia-related proteins, such as tau and TDP-43 can form LLPS droplets. There is evidence that aberrant membrane-less organelles can become more rigid over time, and this process can be accompanied (or preceded) by the misfolding and pathological aggregation of proteins residing in these organelles.

22sHsps share a number of properties with proteins that undergo LLPS, including regions of high intrinsic disorder ( Figure 1 ), capability to transiently interact with many substrates, and self-oligomerization. The same aggregation-prone motifs in the microtuble-binding regions of tau that are implicated in driving LLPS are also targeted by HspB1. HspB2 itself has been shown to undergo LLPS, through interactions of the disordered C-terminal domain, which results in the formation of liquid-like compartments in the nucleus that cause impaired nuclear homeostasis and transcription. Remarkably, this droplet formation is abrogated by co-expression of HspB3, which forms heterooligomeric complexes with HspB2. Emerging evidence for sHsp residency in membrane-less organelles may suggest additional roles for sHsps in LLPS, including regulation of membrane-less organelle biogenesis and clearance. For example, HspB1, HspB5, and HspB7 have been shown to associate with nuclear speckles. Furthermore, HspB1 and HspB8 are recruited to cytoplasmic stress granules, which are membrane-less organelles. While HspB8 has a role in autophagy, the Hsp70-HspB8-Bag3 complex also plays a role in both the autophagy-independent disassembly of stress granules as well as the extraction of misfolded proteins, known as defective ribosomal products (DRiPs). Additionally, down-regulation of HspB8-Bag3 leads to conversion of stress granules to an aberrant state. HspB1 may also affect stress granule dynamics, but perhaps only after stress granules that have already progressed to an aberrant state.

23Membrane-less organelles may provide a microenvironment for the seeding and/or propagation of neurodegeneration-related amyloids. The presence of sHsps in membrane-less organelles may be through a direct interaction with misfolded proteins or due to the intrinsic ability of discrete sHsps to phase separate. Either way, the commingling of neurodegenerative-associated proteins and sHsps in membrane-less organelles may represent a nexus for regulating aggregation. Therefore, modulation of chaperone protein expression or activity may be a potential therapeutic target to treat various neurodegenerative disease pathologies.

24Understanding the interactions and chaperone functions of sHsps with amyloidogenic protein aggregates may better inform small molecule and engineered biologic therapeutic strategies for neurodegenerative proteinopathies.

25Potential therapeutic strategies that aim to modulate endogenous sHsp expression or phosphorylation generally suffer from a lack of specificity for the sHsp family, let alone for discrete sHsps. Heat stress-responsive sHsps can be activated by drugs that generate a challenge to proteostasis, which includes proteasome inhibitors (e.g. Bortezomib), Hsp90 inhibitors (e.g. 17-AAG), and oxidative stress inducers (e.g. terrecyclic acid). However, these treatments also induce expression of other molecular chaperone families (e.g. Hsp70 and Hsp40) and are not specific for sHsp activation. Efforts to identify Hsp co-inducers, substances that potentiate stress responses without inducing a primary stress response on their own, may offer improved selectivity. Arimoclomol is a co-inducer currently in clinical trials for ALS (NCT00706147, NCT03491462), which may be further enhanced by co-administration with a heat shock response inducer, like Celastrol. As phosphorylation of sHsps can modulate their oligomeric status and chaperone activity, identifying compounds that activate specific kinases or phosphatases that modulate sHsps phosphorylation may also be an attractive therapeutic strategy. In fact, inhibitors of phosphodiesterase 9 were found to induce phosphorylation of HspB6 and reduce Aβ cellular toxicity; one of these compounds is currently being tested in an AD clinical trial (NCT00930059).

26Small molecules that interact with sHsps may be a promising strategy for therapeutics, but the nature of this family of chaperones makes drugability difficult. There are no known small molecule ligands to use as a scaffold to start from. The dynamic and polydisperse nature of these proteins taunt the idea of engineering a high affinity binding drug; indeed, these promiscuous proteins likely have many client binding sites with a variety of conformations. The dimer interface of the ACD, regions within the N-terminal domain, and the IPV motif of the C-terminal domain are important for dimer, higher order oligomer, and client protein interactions. sHsp-client protein interactions compete for oligomeric sHsp–sHsp interactions and activate chaperone activity; these sites have also been proposed as potential drug discovery targets. Oxysterols, like Lanesterol, are an example of potential sHsp-targeting small molecules reported to reverse HspB4 or HspB5 aberrant aggregation, which has been reported as a genetic cause of cataract formation in the eye lens. These compounds appear to reduce the formation of amyloid-like fibers through an interaction with the ACD-dimer interface; however, the high concentrations required for this effect may suggest a lack of specificity for this interaction. sHsp peptide fragments or client binding domains may be promising scaffolds to initiate drug discovery efforts. Small molecules or peptides, with optimally tuned affinities triggering cycling between large oligomers and smaller species, may be more effective “activators” of sHsp chaperone activity than traditional static high affinity binders; as cycling itself has been determined to be crucial for proper chaperone activity in vivo with respect to tau aggregation.

27Although we generally restricted this review to human sHsps that interact with amyloidogenic proteins involved in neurodegenerative disease, the diversity of sHsps from different organisms, from bacteria to humans, provides a rich set of ACD-containing proteins to explore for aggregation prevention activity. For example, a sHsp from a parasite was shown to be a potent inhibitor of Aβ fibrillation and reduced associated toxicity in a neuroblastoma cell model. Specific mutant or engineered sHsp variants, with altered oligomeric structure or client interactions, may prove to have increased chaperone activity towards amyloidogenic proteins. Additionally, effective holdase chaperones may be engineered as fusion proteins with domains from other chaperones (foldase or unfoldase) or co-chaperones. Small peptides derived from human HspB4 and HspB5 sequences, termed mini-chaperones, display chaperone-like activity. One of these constructs reduced cellular toxicity of Aβ. An important hurdle for eventual clinical translation of engineered molecular chaperones will be delivery to the site of aggregation. Preclinical evaluation of exogenous chaperone expression can be evaluated in mouse models through gene therapy delivery, as used to evaluate tau pathology by our group; but clinical translation will rely on the continued development of clinical gene therapy or protein transduction domain strategies.

28JW, AD and VU wrote sections of the manuscript. JW, VU, and LB created the figures, JW and LB edited the manuscript. All authors contributed to the manuscript, read, and approved the submitted version.

29This work was supported by the National Institute On Aging of the National Institutes of Health under Award Number RF1AG055088. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

30The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.