The involvement of stress granules in aging and aging‐associated diseases

Abstract Stress granules (SGs) are nonmembrane assemblies formed in cells in response to stress conditions. SGs mainly contain untranslated mRNA and a variety of proteins. RNAs and scaffold proteins with intrinsically disordered regions or RNA‐binding domains are essential for the assembly of SGs, and multivalent macromolecular interactions among these components are thought to be the driving forces for SG assembly. The SG assembly process includes regulation through post‐translational modification and involvement of the cytoskeletal system. During aging, many intracellular bioprocesses become disrupted by factors such as cellular environmental changes, mitochondrial dysfunction, and decline in the protein quality control system. Such changes could lead to the formation of aberrant SGs, as well as alterations in their maintenance, disassembly, and clearance. These aberrant SGs might in turn promote aging and aging‐associated diseases. In this paper, we first review the latest progress on the molecular mechanisms underlying SG assembly and SG functioning under stress conditions. Then, we provide a detailed discussion of the relevance of SGs to aging and aging‐associated diseases.

conditions, and are thought to influence cellular signaling pathways, and mRNA function, localization, and turn over (Buchan, 2014;Buchan & Parker, 2009;Kedersha, Ivanov, & Anderson, 2013). They present across eukaryotes, including mammalian, plant, and fungal (yeast) cells. In mammalian cells, SGs form in response to heat stress, arsenite exposure, UV irradiation, and viral infection (Kedersha et al., 2013). In plant cells, SGs are induced by hypoxia, high-salt stress, and oxidative stress, as well as methyl jasmonate, potassium cyanide, and myxothiazol exposure (Gutierrez-Beltran, Moschou, Smertenko, & Bozhkov, 2015;Nover, Scharf, & Neumann, 1983;Pomeranz et al., 2010;Sorenson & Bailey-Serres, 2014;Weber, Nover, & Fauth, 2008;Yan, Yan, Wang, Yan, & Han, 2014). SGs form from pools of untranslated mRNA and contain various translation initiation factors, as well as a variety of RNA-binding proteins and many non-RNA-binding proteins (Guzikowski, Chen, & Zid, 2019). SGs are dynamic, complex, and variable assemblies, with composition and structure that can vary dramatically under different types of stresses, such as heat shock, oxidative stress, osmotic stress, nutrient starvation, and UV irradiation. This diversity of SGs supposedly relies on diverse interactions between proteins and RNAs within the SGs and reflects the ability of cells to respond quickly to various environmental stresses (Buchan & Parker, 2009;Protter & Parker, 2016). Recent evidence has revealed that mammalian SGs show liquid-like behavior while yeast SGs have characteristics similar to those of solid material (Kroschwald et al., 2015). Results from super-resolution fluorescence microscopy and fluorescence recovery after photobleaching (FRAP) experiments reveal that SGs have two distinct layers with different components, functions, and dynamics: a stable inner core structure surrounded by a less dense shell layer. The components in the core structure are believed to be less dynamic, while the components in the shell layer are more dynamic. Two models were proposed for SG assembly. The first is that SG assembly is initiated by formation of a stable core containing a diverse proteome, followed by rapid growth of this core. Subsequently, these initial small granules merge to form larger mature stress granules, through liquid-liquid phase separation (LLPS). Alternatively, LLPS of translationally repressed ribonucleoproteins occurs first, and then, high concentrations of proteins in phase-separated droplets promote the formation of the core Wheeler, Matheny, Jain, Abrisch, & Parker, 2016).

| S TRE SS G R AN ULE FORMATI ON AND REG UL ATION
Unlike cellular compartments surrounded by lipid bilayer membranes, which physically separate the interior and exterior of the compartments, SGs lack a physical barrier to separate their components from the surrounding medium. For many years, it remained elusive how molecules could be clustered into a discrete area without a surrounding barrier, and how such clustered molecules could modulate their structures and internal biochemical activities. Increasing evidence indicates that LLPS driven by multivalent weak macromolecular interactions (protein-protein, protein-RNA, and RNA-RNA interactions) is an important organizing principle underlying the self-assembly of membraneless compartments (Gomes & Shorter, 2019;Hyman, Weber, & Julicher, 2014). LLPS is a process in which a well-mixed liquid solution separates into two distinct liquid phases.
One phase is enriched in some certain components, and the other is depleted of these components (Alberti & Dormann, 2019). In recent years, an increasing body of evidence has indicated that LLPS plays an important role in cellular organization and the assembly of membraneless organelles, which are comprised of highly concentrated proteins and RNAs (Alberti & Dormann, 2019). In contrast to membrane-bound organelles, which utilize active transport of molecules across the surrounding membranes to maintain their compositions, the liquid state of the two LLPS phases ensures that the components within these two phases can easily rearrange and can quickly exchange materials with each other. This allows liquid non-membranebound compartments such as SGs to remain separate from the liquid cytoplasm but able to rapidly respond to environmental stresses. SGs can accomplish this rapid response through entry of macromolecules into the SGs or release of certain components from the SGs (Alberti & Dormann, 2019; Hyman et al., 2014). However, evidence indicates that this process is sensitive to cellular environmental changes such as temperature, pH, and even concentration of molecules (Alberti & Hyman, 2016). High concentrations of macromolecules must reach a critical threshold to start LLPS. However, if a certain concentration is exceeded, macromolecules within the SGs, such as proteins, will tend to form aggregates. This can lead the LLPS-derived liquid assembly to take on gel-like or solid-like properties, which may not  (IDRs). RNA and DNA molecules that harbor multiple interaction regions for binding to other proteins or nucleic acids can also undergo phase separation independently or synergistically with proteins (Jain & Vale, 2017;Molliex et al., 2015). Different kinds of molecular interactions between the diverse regions discussed above can promote LLPS. We have summarized these interactions that promote LLPS and SG formation in Table 1. In the following sections, we mainly focus on the molecular mechanisms underlying SG assembly and discuss the essential factors responsible for the regulation of SG assembly.

| Roles of proteins with intrinsically disordered regions in SG formation
Multiple high-content studies have generated detailed lists of SG components in both yeast and mammalian cells (Youn et al., 2019).
Here, we summarize the results of these works in Table 2 to provide an overview of all the SG components identified together with the corresponding large-scale experimental methods applied. Only a fraction of these are thought to be necessary for the formation and maintenance of SGs under stress conditions. Proteins that are essential and sufficient to drive the formation of SGs are called scaffold proteins. In the absence of these scaffold proteins, the compartments do not form or are unstable (Ditlev, Case, & Rosen, 2018). For example, certain translation repressors, such as caprin-1 and TIA-1, RNA-binding proteins, like G3BP, and enzymes with ATPase activities, have been shown to function as scaffold proteins (Buchan & Parker, 2009;Gilks et al., 2004;Jain et al., 2016;Kedersha et al., 2016;Tourriere et al., 2003). These proteins can interact with RNAs and recruit other client proteins to the SGs to form a promiscuous interaction network. They are essential for the formation of SGs.
Multivalent interactions among protein intrinsically disordered regions or intrinsically disordered proteins (IDPs), which contain low-complexity domains (LCDs) or prion-like domains (PrLDs), are major contributors to protein phase separation and SG formation.
These low-complexity sequences have low amino acid diversity (Gomes & Shorter, 2019) and can be identified by their similarity in amino acid composition to the known human and yeast prions (Gilks et al., 2004). Many RNA-binding proteins found in SGs are characteristically composed of these sequences, which have been shown to be important for SG formation and targeting. Examples of proteins with prion-like domains that are responsible for granule assembly include T-cell intracellular antigen-1 (TIA-1) and fused in sarcoma (FUS). The prion-like domain of TIA-1 is essential for its recruitment to SGs and is also necessary for promotion of SG formation. Further, the function of the TIA-1 PrLD can functionally substitute with the PrLD of the yeast Sup35 protein (Gilks et al., 2004). The RNA-binding protein FUS localizes to SGs through its N-terminal IDR, and mutations in this region prevent SG accumulation (Kato et al., 2012).
Recent studies have pointed out the importance of promiscuous interactions caused by these prion-related, disordered regions.
Through such interactions, the proteins are able to act cooperatively in promoting LLPS . For example, short linear motifs (SLiMs) within IDRs can interact with the surface of other well-folded protein domains (Jonas & Izaurralde, 2013). Also, interactions between IDRs might form cross-strand beta-zippers to stabilize granules (Protter & Parker, 2016). Recently, a new type of beta-sheets has been identified, namely kinked cross-beta-sheets or low-complexity aromatic-rich kinked segments (LARKS). Unlike classic steric zippers of amyloid fibrils, LARKS are characterized by relatively weak stability. These transient cross-beta-contacts might also contribute to LLPS (Hughes et al., 2018). RGG/RG motifs are segments that occur in low-complexity disordered regions and have a high affinity for RNA.
Thus, multivalent interactions between RGG motifs and RNAs are thought to underlie phase separation in the formation of SGs (Chong, Vernon, & Forman-Kay, 2018;Gomes & Shorter, 2019). Recent evidence has also indicated that intracellular interactions between tyrosine residues from PrLDs and arginine residues in RGG motifs could drive phase separation (Bogaert et al., 2018;Qamar et al., 2018;Wang, Choi, et al., 2018;Yoshizawa et al., 2018). Besides these interactions, IDRs could also work together with RNA recognition motifs (RRMs) to modulate phase behavior synergistically. For example, FUS phase separation is governed by interactions between tyrosine residues in a PrLD and arginine residues in an RNA-binding domain (Wang, Choi, et al., 2018). These domains could form diverse weak interactions such TA B L E 1 Molecular interactions that promote liquid-liquid phase separation and SG formation  (2013) Kinked cross-β-sheets between IDRs Interact weakly through polar atoms and aromatic side chains FUS, hnRNPA1, nup98 Hughes et al. (2018) Interactions between RGG/RG motifs and PrLDs Cation-π interactions FUS Bogaert et al. (2018), Qamar et al. (2018), Wang, Choi, et al. (2018) Interactions between RNAbinding domain and IDR Tyrosine-arginine interactions (different with generic cation-π interactions) FET family proteins Wang, Choi, et al. (2018) (2018) as π-π interactions, cation-π interactions, charge-charge interactions, and intermolecular cross-β-contacts, which are all thought to be important drivers of phase separation (Gomes & Shorter, 2019).

| RNAs are involved in stress granule assembly
Stress granule formation requires a pool of RNA molecules that can bind to the RNA-binding domains of numerous proteins to form multivalent RNA-protein interactions and contribute to the formation of higher-order assemblies. In addition, RNA can influence other aspects of SG assembly. First, SG formation begins with inhibition of translation initiation and polysome disassembly; adding of drugs that trap mRNAs in polysomes will reduce SG assembly (Buchan, Muhlrad, & Parker, 2008;Kedersha et al., 2000;Kedersha, Gupta, Li, Miller, & Anderson, 1999  assembly in a repeat length-and G-quadruplex structure-dependent manner, indicating essential roles for RNA and RNA structure in SG assembly (Fay, Anderson, & Ivanov, 2017). Similarly, Langdon and colleagues observed that the secondary structure of mRNA determines the molecular composition of membraneless compartments (Langdon et al., 2018

| Protein post-translational modification and the cytoskeleton system regulate stress granule formation
Protein post-translational modification (PTM) is another mechanism by which cells control the localization of SG components and assembly/disassembly of SGs (Buchan & Parker, 2009 Khong et al. (2017) poly(ADP-ribosylation) (see Table 3 for a full list of currently known  (Tsai, Ho, & Wei, 2008).
Previous work also showed that phosphorylation of Ras-GTPaseactivating protein SH3 domain-binding protein 1 (G3BP1) on Ser 149, a key player in SG assembly, impairs G3BP1 self-association and inhibits arsenite-induced stress granule assembly (Tourriere et al., 2003). However, a recent study showed that G3BP1-S149 is indeed phosphorylated, but S149 phosphorylation does not change upon stress treatment and therefore is not a simple switch that regulates SGs. The phenotype of the original S149E mutant emerged because of an accidental S99P mutation. Despite this, Panas et al.
pointed out that phosphorylation of S149 could not be excluded from influencing other processes related to SG or signaling, which is regulated by interactions between G3BP1 and other proteins (Panas et al., 2019 that covalently attaches to protein substrates in a manner similar to ubiquitination, known as neddylation. Neddylation of serine/arginine (SR)-rich splicing factor 3 (SRSF3), an SG regulator, is necessary for SG formation (Jayabalan et al., 2016). Reversible protein acetylation is another important protein PTM modulated by histone acetylases and histone deacetylases. Histone deacetylase 6 (HDAC6), a cytoplasmic deacetylase, exhibits deacetylase activity that is important for regulation of SG formation, indicating a new role for acetylation in the stress response (Kwon, Zhang, & Matthias, 2007).
Based on RNA-mediated interference screening, some genes that modification were found to be important for SG assembly. Further, the substrates of O-GlcNAc modification might be major components of SGs (Ohn, Kedersha, Hickman, Tisdale, & Anderson, 2008).
SUMO is a small ubiquitin-like molecule that is covalently attached to target proteins to regulate protein-protein interactions, protein localization, and protein function (Hannoun, Greenhough, Jaffray, Hay, & Hay, 2010). SUMOylation has also been reported to play a role in recruitment of proteins to SGs, since SG assembly is impaired when SUMOylation of eukaryotic initiation factor eIF4A2 is disa-  by PARPs, regulates many physiological processes in the nucleus (Schreiber, Dantzer, Ame, & Murcia, 2006). Recent studies have found that PARylation level can serve as a major regulator in the assembly and disassembly of SGs. Specific PARPs and pADPr glycohydrolases (PARGs) are localized to cytoplasmic SGs . Overexpression of PARPs can induce the formation of SG without stress, and overexpression of PARGs can prevent SG formation. Also, a high level of PARylation in the cell delays disassembly of SGs (Catara et al., 2017;Leung, 2014;Leung et al., 2011).
An additional study found that the formation of stress granules is a strategy adopted by cancer cells to resist chemo/radiotherapy. Besides the above-mentioned PTMs, the cytoskeletal system is another factor involved in the regulation of SG formation (Kwon et al., 2007;Loschi, Leishman, Berardone, & Boccaccio, 2009). SGs are dynamic structures that exchange materials with other granules and/or with their surroundings. The cytoskeleton system is thought to be a scaffold for SG assembly and dynamic maintenance, as well as movement, fusion, and fission ( Figure 1a). For example, microtubules and motor proteins are critical for SG fusion and disassembly, as evidenced by the finding that drugs that disrupt microtubules and inhibit motor proteins reduce the appearance of SGs (Kwon et al., 2007). In addition, the motor proteins kinesin and dynein are localized in SGs and regulate SG assembly and disassembly (Loschi et al., 2009).

| FUN C TI ON S OF S TRE SS G R AN ULE S
Cells are frequently exposed to fluctuating, potentially adverse environmental conditions. Thus, formation of SGs allows cells to adapt to diverse environmental changes and provides protection for key cellular components. This is reflected in some SGdefective mutants, which are more sensitive to stress. These include HDAC6-deficient mouse embryo fibroblasts, p90 ribo- The second role of SGs is signal transduction in response to cellular stress. Some specific factors that are sequestered in SGs will alter the regulation of particular signaling pathways, in which these factors participate (Kedersha et al., 2013). For example, target of rapamycin complex 1 (TORC1) signaling can be modulated by SGs upon heat stress: TORC1 signaling is blunted by sequestering TORC1 in SGs, and signaling is reactivated upon SG disassembly (Takahara & Maeda, 2012). Additionally, astrin-induced localization of mTORC1 components in SGs suppresses oxidative stress-induced apoptosis (Thedieck et al., 2013). Moreover, SG formation can negatively regulate the stress-activated p38 and JNK MAPK (SAPK) pathways that trigger the apoptotic response. This is accomplished through sequestration of the receptor for activated C kinase (RACK1) in the SG, thus protecting the cell from death (Arimoto, Fukuda, Imajoh-Ohmi, Saito, & Takekawa, 2008). Furthermore, ribosomal subunit eIF4G can interact with and recruit TNF-α receptor-associated factor 2 (TRAF2) to SGs when a cell is confronted with stress, thus inhibiting tumor necrosis factor signaling by lowering TRAF2 biological activity in the cytoplasm (Kim, Back, Kim, Ryu, & Jang, 2005). Plasminogen activator inhibitor-1 (PAI-1), an activator of senescence, is also localized to SGs in senescent cells. This indicates that SGs can counteract senescence by recruiting a key factor to the SGs and disrupting its senescence function (Omer et al., 2018).
The third function of SG is a protective one: Formation of SGs allows cells to "bounce back" and thrive once the stress subsides.
SGs do this by temporarily storing and protecting mRNAs and proteins from autophagy and degradation by proteasomes throughout the duration of the stress, allowing a rapid restart of translation and other signaling pathways upon release from the stress (Guzikowski et al., 2019). Yeast poly(A)-binding protein (Pab1), a marker of SGs, can act as a stress sensor and undergo a phase transition to a hydrogel to protect stressed cells and then to help cells better recover upon stress relief (Riback et al., 2017). Results have also shown that the yeast pyruvate kinase protein Cdc19 can form reversible aggregates that co-localize with SGs during stress, which is a protective mechanism from stress-induced degradation.
Upon cessation of stress and resumption of growth, it allows quick re-entry of Cdc19 into the cell cycle without re-expression (Saad et al., 2017).

| Aging and aging-related diseases
Aging is a ubiquitous, progressive, and usually irreversible biological process accompanied by a decline in physiological and reproductive functions. This process normally begins shortly after the formation of the fertilized egg and spans the whole lifespan of the F I G U R E 1 Effects of aging on SG assembly, dynamics, and clearance. Formation of SGs begins with nucleation of various RNA-binding proteins and RNAs. The SGs then grow into larger assemblies via additional proteinprotein and protein-RNA interactions. These complexes coalesce into higherorder SGs in a cytoskeleton systemdependent manner (a). Aging-associated mitochondrial dysfunction and inactive metabolism might lead to limited control of this process and aberrant SGs (d). In addition, aging-associated disease-causing proteins, misfolded proteins caused by protein homeostasis decline (b), and other chronic stress (c) during aging lead to impaired SG dynamics and persistent SGs. Aberrant SGs can be cleared by autophagy under normal conditions, but with age, disturbed PQC can have a negative effect on the removal of aberrant SGs (e) body (Veitia, Govindaraju, Bottani, & Birchler, 2017). It has been less than 40 years since aging research entered the era of modern molecular biology with isolation of the first long-lived nematode mutant (Klass, 1983 Aging is a predominant risk factor for many disorders, including cancer, neurodegenerative diseases, and cardiovascular diseases. Aging itself is not a disease, but aging and aging-related diseases sometimes share the same basic molecular and cellular mechanisms (Franceschi et al., 2018). For example, as aging progresses, the ability of cells to maintain protein homeostasis declines, leading to the formation of widespread intracellular protein aggregates (Taylor & Dillin, 2011). In addition, neurodegenerative diseases such as

| Aberrant stress granules: a driver of aging and aging-related diseases?
Age-associated proteostasis disruption is characterized by the ap- More importantly, the classic SG proteins PAB-1 and TIAR-2 can form aggregates in aged C. elegans, and a high level of aggregation of these SG components is associated with smaller animal size, reduced fitness, and shorter lifespan . This suggests that SG protein aggregation might accelerate aging and reduce lifespan.
LCDs of RNA-binding proteins (RBPs) play an important role in SG formation because these motifs allow multiple interactions; along with their flexibility in folding, this allows the formation of a dense protein-protein and protein-RNA interaction network (Buchan, 2014;Guzikowski et al., 2019). However, their conformational flexibility also leads these proteins to be aggregation-prone  Neumann et al., 2006Neumann et al., , 2009).

In addition to RBP mutants and RBP-including aggregates within
SGs, which can disturb SG dynamics, some proteins can form aggregates independently from SG formation. They can aggregate in association with lipid membranes such as mitochondrial membranes and endoplasmic reticulum, which can lead to dysfunction of these membrane-bound organelles (Hebda & Miranker, 2009;Stefani, 2010;Zhou et al., 2014). Also, some cytosolic amyloid-like protein aggregates such as TDP-43 have been shown to disturb nuclear Abnormal accumulation of misfolded proteins that form insoluble fibrillar aggregates is thought to be one of the major causes of many neurodegenerative diseases (Furukawa & Nukina, 2013).
Although fibrils are toxic to some degree, numerous studies suggest that protein oligomers are more strongly correlated with disease severity than fibrils in neurodegenerative diseases ( formation of fibrils, Sup35 would lose its function as translation terminator and confer cell a survival advantage under oxidative stress (True, Berlin, & Lindquist, 2004). In animals, amyloid formation has been proposed to modulate signal transduction and clear misfolded proteins (Furukawa & Nukina, 2013;Li et al., 2012). Thus, the complexity of the potential consequences of protein aggregation on aging and diseases, as well as direct causal link between aggregation and neurotoxicity, requires further investigation.

| How do aging and aging-related diseases affect stress granules?
Emerging evidence suggests that LLPS-driven SG assembly is associated with cancer, virus infections, and age-related neurodegenerative disorders . A recent study found that SG components were mislocalized from the nucleus to the cytoplasm in aged animals when they were exposed to mild stressful conditions, but no mislocalization occurred in young animals under the same conditions . It is well shown that during aging, cellular surveillance systems are disturbed, resulting in the aberrant cytoplasmic localization of nuclear RBP. We will therefore discuss how stress granules might be influenced by aging and aging-related diseases in detail below.

| The relationship between cellular senescence state and SG formation
Cellular senescence is an irreversible cell cycle arrest state. The number of cells entering this state increases with aging, and it has been widely assumed that cellular senescence contributes to aging (Lopez-Otin et al., 2013). Cellular senescence has been found to impair the proper formation of both canonical and noncanonical SGs in kidney cells (Moujaber et al., 2017). Canonical SG formation is reduced by the depletion of transcription factor Sp1, which regulates the abundance of the SG-nucleating proteins G3BP1 and TIA-1/ TIAR. In addition, senescence can cause translation initiation factor eIF2α hyperphosphorylation and the loss of CreP, which also correlates with the aging-related hyperphosphorylation of eIF2α (Moujaber et al., 2017). These results indicate that when cells enter the senescence state, two essential SG proteins, eIF2α and the transcription factor Sp1, can work as direct aging-related targets, causing significant deficiencies in SG production. Similarly, a recent study found that constitutive exposure to stress could induce the formation of SGs in proliferating cells, but not in fully senescent human fibroblasts (Omer et al., 2018). It has also been reported that the formation of SGs during early stages of senescence is sufficient to prevent senescence. This process involves the recruitment of PAI-1 to the SGs. Recruitment of PAI-1 to SGs interferes with secretion of PAI-1 and, consequently, cytoplasmic retention of cyclin D1, which further promotes the phosphorylation of pRB, leading to the preven-  (Lian & Gallouzi, 2009). This increase correlates with a rapid decrease in the expression levels of the senescence-associated gene, p21 (Lian & Gallouzi, 2009), which is not found in senescent cells under chronic and persistent stress (Omer et al., 2018). This suggests that how senescence process and the behavior of senescent cells respond to stress might depend on the method used to induce stress and whether or not the stress is physiologically relevant. At the same time, Lian et al. found that, accompanied by the formation of increased number of SGs under acute stress, the disassembly of SGs is also delayed when cells are fully senescent upon acute stress removal. As a consequence of this, translation and synthesis of some vital proteins in fully senescent cells would occur at a slower rate than in proliferative cells after stress removal, affecting normal cellular processes (Lian & Gallouzi, 2009). Thus, we can see that although senescent cells respond differently to chronic stress and acute insult, their ability to regulate SGs and adaptability to stress are impaired in general. This could in part explain the well-known phenomenon that senescent cells recover from stress more slowly than nonsenescent cells; it could also help to explain their high correlation with age-related diseases (Gallouzi, 2009;Honda & Matsuo, 1987;Rosenfeldt et al., 2004;Zarzhevsky, Menashe, Carmeli, Stein, & Reznick, 2001).

| Aging-induced cellular environmental and metabolic changes affect stress granule dynamics
The process of aging is linked with the decline of multiple cellular events, including loss of protein homeostasis, decreased vacuolar acidity, and increased reactive oxygen species (ROS) damage (Denoth Lippuner, Julou, & Barral, 2014;Lopez-Otin et al., 2013).  (Molliex et al., 2015;Patel et al., 2015). These concentration-induced changes in phase behavior could also trigger changes in the binding affinities of IDPs with their binding partners, leading to recruitment of distinct molecules into RNP granules. This in turn could lead to different RNP granule composition, physical properties, and structural organization (Alberti & Hyman, 2016).
Several lines of evidence indicate that aging is associated with the loss of ability to control gene expression and maintain protein homeostasis (Figure 1b). In the nematode C. elegans, extensive proteome remodeling and imbalances have been found to occur during aging, with a large amount of proteins increasing or decreasing in abundance, accompanied by widespread protein aggregation (Walther et al., 2015). Changes in protein concentration and impairment in the solubility of RNP granule-forming proteins during aging would serve to impair SG formation or recruit different clients to RNP granules. Thus, the progressive failure of protein homeostasis that occurs during aging should impact SG assembly or disturb SG dynamic maintenance, which would lead to increased aggregation load.
In addition, during aging, cells are exposed to diverse types of stress, such as constant oxidative stress and declining acidity (Denoth Lippuner et al., 2014). Oxidative stress can induce the oxidation of proteins, lipids, and DNA, resulting in irreversible structural and functional damage (Gandhi & Abramov, 2012). Oxidative stress is also a known inducer of SG formation and a key factor in neurodegenerative disorders (Federico et al., 2012;Gandhi & Abramov, 2012;Patten, Germain, Kelly, & Slack, 2010). Acute oxidative stress promotes SG formation, and SGs disassemble when the stress is removed. However, persistent oxidative stress could trigger aggregation and oligomerization of some pathological RBPs such as TDP-43, FUS, and tau (the most toxic species in neurodegenerative tauopathies), which could subsequently be recruited to SGs, serving to stabilize them (Chen & Liu, 2017;. Thus, it can be speculated that these aging-associated chronic stress conditions could accelerate the formation of pathological aggregates of RBPs, causing additional functional defects and enhancing the nucleation of pathological SGs (Figure 1c), especially when the protein quality control system is overwhelmed.
It is worth noting that most of the conditions used to study SGs might have no or very little physiological relevance to aging and aging-related diseases. For example, SGs can be induced by different oxidative stress reagents such as sodium arsenite and peroxide, but the constitutions of these SGs are different (Chen & Liu, 2017).
This means that SGs induced by these reagents represent specific responses of cells to environmental changes, but these conditions do not necessarily mimic the intracellular state of aging cells. Also, when cells encounter acute stress, SGs can be induced and then removed after the insults subside, which is considered to be a protective measure of the cells. In contrast, during aging, persistent stress might induce continuous formation of SGs , interfering with translation and the normal functions of important proteins trapped in the SGs. Thus, considering the differences between chronic stress and acute stress, more studies of SGs, including both their components and dynamics, should be performed under chronic stress conditions that mimic aging or disease-related intracellular environmental changes.

| The influence of RBP mutations associated with aging-related Neurodegenerative diseases on SG dynamics
RBP mutations that increase SG formation or limit SG clearance are known causative factors in some neurodegenerative diseases (Li, King, Shorter, & Gitler, 2013;Ramaswami, Taylor, & Parker, 2013).

| Declining protein quality control system in aged cells impacts SG composition and dynamics
The protein quality control (PQC) system is an integrated network of molecular chaperones and two main degradative systems, which selectively degrade misfolded proteins and dysfunctional organelles, namely the ubiquitin-proteasome system (UPS) and autophagy, respectively (Amm, Sommer, & Wolf, 2014;Boya, Reggiori, & Codogno, 2013). When cells age, however, the PQC system and protein homeostasis are disrupted (Josefson, Andersson, & Nystrom, 2017).
Chaperone expression levels and activity have been found to decrease in the brains of mice as they age, and this aging-dependent decline contributes to the susceptibility of aged neuronal cells to misfolded proteins (Yang, Huang, Huang, Gaertig, Li, & Li, 2014). Further, analysis of chaperone expression in humans found that ATP-dependent chaperone machines are repressed both in the aging human brain and in aging-associated diseases (Figure 1e), indicating the importance of chaperones in aging and in prevention of the pathogenesis of neurodegenerative disease (Brehme et al., 2014). In addition, a large body of evidence supports an overall age-dependent decrease in the amount of proteasomes or UPS subunits, as well as proteasomal activity, during the aging process (Figure 1e; Baraibar & Friguet, 2012;Bulteau, Szweda, & Friguet, 2002;Kastle & Grune, 2011;Saez & Vilchez, 2014).
Autophagy is a conserved PQC system involving the removal of damaged proteins and organelles in a lysosomal-mediated pathway.
Severe stress-and aging-related misfolded proteins could specifically accumulate and aggregate within SGs, which could alter SG composition, impair SG dynamics, and, finally, lead to aberrant conversion from a liquid-like to a solid-like state (Figure 1b). For example, under severe stress conditions like robust heat stress, stress granule components can interact with misfolded proteins via their PLDs, promoting the seeding of SGs (Kroschwald et al., 2015). Under mild stress conditions or normal growth conditions, the cellular chaperone machinery and degradation systems are sufficient to manage the surveillance of such aberrant interactions between RBPs and other aggregation-prone proteins.
During aging, however, PQC systems decline, resulting in compromised PQC systems that can be overrun, which might affect SG dynamics.
In support of this, researchers have found that inhibition of Hsp70 function in yeast and mammals leads to increased SG formation and delayed SG disassembly through an increase in the number of SGs containing misfolded proteins (Cherkasov et al., 2013;Mateju et al., 2017;Walters, Muhlrad, Garcia, & Parker, 2015). Ganassi and collaborators identified the HSPB8-BAG3-HSP70 chaperone complex as a key regulator of SG composition and dynamics. Once this chaperone-mediated SG surveillance mechanism is disturbed, misfolded proteins and defective ribosomal products (DRiPs) accumulate in SGs, triggering an aberrant liquid-to-solid conversion with defective SG disassembly (Ganassi et al., 2016). The 26S proteasome can be recruited to SGs to promote their clearance (Turakhiya et al., 2018), and inhibition of the UPS induces SG formation, although this is not because of the failure of degradation of SG assembly factors (Mazroui, Marco, Kaufman, & Gallouzi, 2007).
In young cells, aberrant SGs can be cleared through another backup degradative system-autophagy-in which aberrant SGs are targeted to the vacuole. Inhibition of autophagy could affect SG clearance in mammalian cells, in which autophagic clearance of SG may be an important factor in reducing the pathology from various diseases (Buchan et al., 2013). However, in aged cells, impaired autophagy may be associated with enhanced SG formation and/or disturbed SG clearance, resulting in aberrant SG persistence after removal of the stressor (Figure 1e). This may be a cause of aging-related neurodegenerative diseases. This is consistent with the association of reduced autophagy with accelerated aging, and the correlation between autophagy/ubiquitination failure and a wide range of disorders, including cancer and neurodegeneration (Saez & Vilchez, 2014;Yang & Klionsky, 2010). It may be due to that incompetence of the PQC systems results in the inability to eliminate aberrant SGs and protein aggregations in a timely manner (Madeo, Tavernarakis, & Kroemer, 2010;Rubinsztein, Marino, & Kroemer, 2011). In 2014, Seguin and colleagues found that autophagy, lysosomal activity, and VCP activity are not only involved in SG clearance, but also participate in proper SG assembly. Inhibition of their activities causes defective SG formation and alterations in SG morphology and composition (Seguin et al., 2014). This shows that autophagy can affect SG assembly and disassembly in a complicated process that requires further investigation.
Apart from this, less is known about how neuronal autophagy is regulated. In contrast to autophagy in non-neuronal cells, which can be induced by starvation and other cellular stressors, autophagy in neurons is not significantly induced by such stressors (Maday, Wallace, & Holzbaur, 2012;Wong & Holzbaur, 2014).
Also, although some progress has been made in revealing the relationship between autophagy and aging in non-neuronal cells (Chang et al., 2017;Hansen, Rubinsztein, & Walker, 2018), little is known about whether these essential mechanisms work similarly in neurons. What we know is that disrupting autophagy in neurons has led to neurodegeneration in some animal models (Hara et al., 2006;Komatsu et al., 2006;Zhao et al., 2013). Recently, Stavoe and colleagues found that aging could affect autophagosome biogenesis and induce morphological changes of autophagic structures in neurons through regulation of the dynamics and phosphorylation state of WIPI2 (Stavoe et al., 2019;Stavoe & Holzbaur, 2020). In addition to initial autophagosome biogenesis, Nixon found that lysosomal integrity in later stages of autophagy also decreased with age in neurons (Nixon, 2017). Despite this, the regulation of other steps in neuronal autophagy, such as autophagosome closure, retrograde transport of autophagosomes, and cargo degradation, during aging is yet to be fully understood. Also not fully understood is how aging influences SG assembly and disassembly through a deficiency of neuronal autophagy.

| P OTENTIAL REL ATI ON S HIPS B E T WEEN OTHER MEMB R ANELE SS ORG ANELLE S AND AG ING OR AG ING -REL ATED D IS E A S E S
In addition to SGs, there are many other membraneless compartments that may be related to aging. These compartments differ in their composition, localization, size, and function. Examples of these non-membrane-bound assemblies include P bodies, P granules, neuronal transport granules, nucleoli, and Cajal bodies. Unlike SGs, which are enriched in translation initiation factors, P bodies are cytoplasmic RNA granules comprised primarily of the factors involved in mRNA degradation, translational repression, and RNAmediated gene silencing (Luo, Na, & Slavoff, 2018). Despite their differences, SGs and PBs also share many RNA-binding proteins and RNA components, and are in close proximity and even overlap with each other (Buchan et al., 2008;Souquere et al., 2009). Given the existence of RNA decay factors in P bodies, they have been proposed to be involved in mRNA degradation and turnover. Currently, whether P bodies function in mRNA storage and/or mRNA decay remains actively debated (Luo et al., 2018;Standart & Weil, 2018).
Combined with the finding that defects in RNA metabolism can trigger aging (Mazzoni & Falcone, 2011), to some extent these findings imply potential links between P bodies and aging. In addition, studies have indicated that P bodies are targeted for autophagy during the stationary phase, which suggests that aging-related PQC declines might also affect P-body dynamics (Buchan et al., 2013). P granules are membraneless organelles composed of proteins and RNAs and found in the germline cytoplasm (Seydoux, 2018).
They are thought to be sites of small RNA biogenesis and post-transcriptional regulation, and are essential for the differentiation of germ cells into functional gametes during postembryonic development (Seydoux, 2018;Strome & Updike, 2015 (Brangwynne et al., 2009(Brangwynne et al., , 2011Kroschwald et al., 2015;Seydoux, 2018 will be an important question to be answered. Moreover, it is known that aggregation-prone proteins can be recruited to SGs and that this could result in aberrant or persistent SGs during cellular stress and after the stress subsides. These aberrant SGs might induce a series of effects that can be attributed to reduced stress resistance with age. Such aberrant SGs may also act as seeds to facilitate the formation of irreversible mature protein aggregates in aged cells, further accelerating the decline of the cellular functions of these proteins. Thus, it seems that maintaining a proper SG dynamic might be a potential strategy to delay aging and increase lifespan. Two key questions that remain to be answered are as follows: (a) What kind of proteins are prone to form aggregates during aging? And (b) is aggregation triggered by interactions between aggregation-prone proteins and SG components? Answers to these questions will have important implications for our understanding of the machineries underlying the relationship between SGs and aging. Key components identified from such studies might generate valuable pharmaceutical targets for the treatment of aging-related diseases.

CO N FLI C T S O F I NTE R E S T
The authors have no conflicts of interest to disclose.

AUTH O R CO NTR I B UTI O N S
XC and XJ wrote the manuscript with input from BL. XC, XJ, and BL contributed to the design of the manuscript. BL supervised the overall direction, planning, and writing of the manuscript.