Amyotrophic lateral sclerosis (ALS) is the most common form of motor neuron disease and is known to be clinically, genetically and pathologically linked to frontotemporal dementia (FTD), the second most common form of presenile dementia after Alzheimer’s disease. Thus, both disorders relate to one another on a disease spectrum for which there is currently no cure nor disease-modifying treatment available. A considerable number of studies have provided evidence for a link between the dysregulation of RNA homeostasis and defective autophagy in ALS/FTD. Furthermore, recent literature from the autophagy and RNA research fields has provided intriguing connections between these key cellular mechanisms, with implications for our understanding of ALS/FTD pathogenesis.

Autophagy is an evolutionarily conserved lysosomal catabolic pathway that ensures nutrient recycling and removal of unwanted substrates. Consequently, autophagy plays a crucial homeostatic role for maintaining healthy and functional cells, enabling adaptation to changing cellular demands and protection from stress. Even though most studies have focused on autophagy-mediated protein catabolism, accumulating data supports autophagy-mediated RNA degradation as an important additional cellular RNA quality control mechanism. RNA, RNA-binding proteins (RBPs) and RNA-protein complexes (such as RNA granules) can all be degraded by autophagy, although the mechanisms of these processes and their biological consequences have not yet been fully determined and this is now an active field of research. Furthermore, new non-conventional functions of autophagy machinery proteins impacting key aspects of RNA homeostasis have been revealed.

The dysregulation of RNA and protein homeostasis is a major contributor to ALS/FTD pathogenesis. Notably, abnormal accumulation of insoluble protein inclusions in neurons and neuroglia in affected brain regions is a key hallmark of ALS/FTD. These inclusions are frequently characterized by the presence of RBPs involved in RNA processing, such as TAR-DNA binding protein (TDP-43), and proteins from the autophagy machinery, like the selective autophagy receptor, p62. Moreover, ALS/FTD genetics have also uncovered a substantial number of genes involved in autophagy and RNA metabolism. Thus, autophagy and RNA homeostasis are two key dysregulated molecular processes and there is an emerging perspective of an interplay between them in disease.

In this review, we will first provide an overview of important aspects of ALS/FTD and autophagy and then present novel findings regarding conventional and non-conventional functions of the autophagy machinery in RNA catabolism. We will then summarize key evidence for defective autophagy in ALS/FTD, dissecting the relevance of disease-associated proteins in the different stages of this process and defective transcriptional and post-transcriptional regulation. Finally, we will discuss the emerging impact of the interplay between autophagy and RNA homeostasis with particular reference to the complexities in C9orf72-ALS/FTD.

ALS and FTD are characterized by progressive degeneration of neurons in affected areas of the nervous system: motor neurons in the motor cortex, brainstem, and spinal cord in ALS and cortical neurons in the frontal and temporal lobes in FTD (Abramzon et al., 2020). Degeneration in these areas results in the clinically observed symptoms of loss of motor control in ALS and behavioural and language dysfunctions in FTD. Despite being clinically distinct, symptoms of both diseases are regularly observed in the same person or segregating within affected families (Abramzon et al., 2020). FTD and ALS are also linked by a common pathological signature, with the vast majority of familial and sporadic cases of ALS and a significant proportion of FTD displaying pathological accumulation of the RBP TDP-43 into cytoplasmic inclusions in neurons and also in glial cells (Neumann et al., 2007). Apart from the clinical and pathological overlap, genetics also unifies both disorders. Classic familial inheritance is observed in 5%–10% of ALS cases and in 30%–50% of FTD cases; whereas some genes like SOD1 or MAPT are predominantly linked to ALS or FTD, respectively, disease-causing variants in a number of genes have been identified in ALS/FTD families (see Table 1). Of particular note, in 2011, the discovery of a pathogenic G₄C₂ hexanucleotide repeat expansion in a non-coding region of C9orf72 gene causing both ALS and FTD, strengthened the genetic link between these neurodegenerative disorders (DeJesus-Hernandez et al., 2011; Renton et al., 2011). The C9orf72 mutation is the most common cause of familial ALS/FTD with around 40% and 25% of familial ALS or FTD carrying the C9orf72 repeat expansion, respectively, and accounting for around 6% of sporadic cases in both as well (Majounie et al., 2012).

The findings that sporadic and familial ALS/FTD are phenotypically and pathologically indistinguishable and that pathogenic variants in many ALS/FTD-linked genes are described in patients without a family history of disease highlight the value of studying the monogenic, heritable forms (Shepheard et al., 2021). Although ALS/FTD-associated genes do not fit easily into a single cellular pathway, they direct the focus to two key cellular processes: RNA metabolism and autophagy degradation, as many of the ALS/FTD-linked genes determined are implicated in these biological pathways. The first category of ALS/FTD-linked genes fall into those encoding for RBPs, like TARDBP (Sreedharan et al., 2008), FUS (Vance et al., 2009), hnRNPA1 (Kim et al., 2013), ANXA11 (Smith et al., 2017) and SMN (Blauw et al., 2012), implicated in RNA metabolism (see Table 1). RBPs bind to RNA at specific sequences or secondary structures to facilitate several steps of the RNA life cycle, both in the nucleus and cytoplasm. Under physiological conditions, the majority of ALS/FTD-associated RBPs are largely nuclear; however, under pathological conditions, such as when they are affected by ALS/FTD-associated mutations, they are often mislocalised to the cytoplasm and form large insoluble inclusions (Maziuk et al., 2017; Nussbacher et al., 2019).

Mislocalisation and aggregation of mutant RBPs impact their normal functions regulating RNA metabolism. This can result in a loss of function effect on their respective nuclear or cytoplasmic mRNA targets, but aggregation of ALS/FTD-related RBPs can also favour pathological interactions and sequestration of other RBPs or RNA (Nussbacher et al., 2019). An additional gain of function patho-mechanism may derive from the formation of aberrant RNA granules. This mechanism relies on the intrinsic aggregation-prone properties of RBPs; apart from RNA recognition motifs and nuclear import and export sequences, a feature of all RBPs is the presence of low complexity domains (LCDs). These domains allow transient multivalent interactions facilitating the formation of membraneless RNA-protein complexes, such as stress granules (SGs), which require the capacity to form and dissolve rapidly. Many ALS/FTD-causing mutations in RBPs occur in LCDs, increasing the tendency to form poorly dynamic and solid-like complexes which may form the seed for subsequent aggregation (reviewed in Mann and Donnelly, 2021; Solomon et al., 2021). Additionally, mutations in the stress granule components TIA1 and ataxin-2 (ATXN2) have also been associated with ALS (Polymenidou et al., 2011; Sephton et al., 2011; Mackenzie et al., 2017). As already noted, the most frequent RBP which aggregates in ALS/FTD is TDP-43 although mutations in TARDBP itself only underly 1%–4% of ALS/FTD cases, showing that alternative mechanisms can drive RBPs to aggregate (Sreedharan et al., 2008). Similarly, inclusions of the RBP FUS are found in brain tissue of some patients with ALS/FTD both with and without FUS mutations (Vance et al., 2009; van Langenhove et al., 2010).

A second category of ALS/FTD-linked genes fall in those encoding for proteins involved in the autophagy-lysosomal pathway (see Table 1). The products of these genes, such as those belonging to the autophagy machinery, like ubiquilin-2 (gene: UBQLN2) (Deng et al., 2011), optineurin (OPTN) (Maruyama et al., 2010), and p62 (SQSTM1) (Fecto et al., 2011; Rubino et al., 2012), are also identified in both neuronal and glial inclusions in post-mortem ALS/FTD patient tissue from both familial and sporadic cases, and sometimes colocalise with the above mentioned RBPs (Ramesh and Pandey, 2017). While this is also the case for two genes causing FTD (CHMP2B (Skibinski et al., 2005) and GRN), mutations in the MAPT gene predominantly lead to FTD with a different brain pathology, characterized by abnormal hyperphosphorylated tau filaments. Tauopathy is a feature of many diseases and, while autophagy also seems to be a key player in dealing with tau aggregates (Spillantini and Goedert, 2013), we will not discuss this within this review.

Similar to other ALS/FTD genes, the brain pathology of C9orf72 ALS/FTD patients shows typical TDP-43 and p62 inclusions in neurons and glial cells (Al-Sarraj et al., 2011; Murray et al., 2011; Cooper-Knock et al., 2012). However, C9orf72 patients additionally show unique RNA and protein pathologies stemming from the nature of the mutation. The large intronic repeat expansion mutation in C9orf72 is transcribed into sense and antisense repeat RNA which form aggregates termed RNA foci in neurons and glia (Mizielinska et al., 2013). Although in a non-coding genomic region, the sense and antisense repeat RNA can also be translated via an unconventional mechanism into dipeptide-repeat proteins (DPRs) which form protein aggregates in C9orf72-ALS/FTD patient brain and spinal cord (reviewed in Balendra and Isaacs, 2018). The mutation also causes haploinsufficiency of the encoded C9orf72 protein itself. Research is revealing multiples roles for the C9orf72 protein in autophagy, and although its loss does not seem to be the dominant mechanism for neurodegeneration, recent literature supports a model where a combination of C9orf72 haploinsufficiency and gain of function repeat RNA and DPR toxic mechanisms drive disease pathogenesis via a synergistic effect (reviewed in Shi et al., 2018).

Thus, ALS and FTD genetics and pathology have a strong association with dysfunction of RNA homeostasis and autophagy pathways and indicate them as key disease mechanisms that warrant further review.

Autophagy is particularly important in neurons and neuroglia where unique autophagy signalling and functions operate (Stavoe and Holzbaur, 2019; Belgrad et al., 2020; Sung and Jimenez-Sanchez, 2020). Neuronal autophagy is temporally and spatially fine-tuned to face their post-mitotic and long-lived nature and to accommodate their highly polarized structure. In neurons, autophagy plays a key role maintaining neuronal homeostasis as the major degradative pathway for the clearance of larger targets, such as aggregates and dysfunctional organelles. Furthermore, autophagy regulates specific neuronal functions, such as neurodevelopment, including axonal outgrowth and synapse formation, and neuronal activity and plasticity, which underlies the unique function of neurons to pass signals along neuronal networks and enables nervous system functionality. Unsurprisingly, defective neuronal autophagy has been widely linked to proteotoxic stress, aging and neurodegeneration (Stavoe and Holzbaur, 2019).

Neuroglia, comprising microglia, astrocytes, and oligodendrocyte lineage cells, are the other major brain cell types and in recent years have emerged as key players in brain development, physiology and metabolism. However, functions of autophagy in glial cells are less well known than in neurons (Strohm and Behrends, 2020). For example, in microglia, the resident mono-phagocytic cells in the CNS, autophagy participates in immune-related processes such as inflammasome activation (Houtman et al., 2019) and has been shown to be crucial for synaptic homeostasis in mice (Kim et al., 2017). Astrocytes, which help neuronal survival via the release of neurotrophic factors (Li et al., 2019), secrete bioactive molecules in extracellular vesicles in a process that involves divergent secretory organelles, including lysosomes and secretory autophagic vesicles (Verkhratsky et al., 2016), as further explained later. Lastly, in oligodendrocytes, autophagy is involved in myelination, which is one of the crucial functions of these cells (Belgrad et al., 2020). As briefly exemplified here, autophagy contributes to the biology of different cell types in the brain, and thus understanding how autophagy differs or is differentially regulated in neurons and neuroglia is an area of research that warrants the increasing attention it is receiving (Fleming et al., 2022).

However, across all cell types, the autophagy process shares general common features, reviewed here. Three distinct autophagy pathways can be described depending on a cargo’s delivery route to the lysosome: chaperone-mediated autophagy (CMA), microautophagy, and macroautophagy. CMA utilizes cytosolic chaperone proteins, typically the heat shock protein family A (Hsp70) member 8 (HSPA8/HSC70), to bind soluble proteins and direct them to the lysosome. CMA protein targets contain a KFERQ-like recognition motif and are delivered across the lysosomal membrane one-by-one by the lysosomal-associated membrane protein 2A (LAMP2A), which acts as a receptor to translocate them to the lysosomal lumen for degradation (Tekirdag and Cuervo, 2018). During microautophagy cytoplasmic entities destined for degradation are directly taken up by lysosomes or late endosomes (endosomal microautophagy). Although it is the least studied form of autophagy, microautophagy has now proven able to be selective (reviewed in Tekirdag and Cuervo, 2018).

Macroautophagy, herein referred to as autophagy, is the most prevalent and well-characterized. Since it is by far the most studied autophagic pathway in the context of RNA catabolism and ALS/FTD, we will focus on this process. Autophagy involves de novo formation of cytoplasmic double-membrane organelles termed autophagosomes. Autophagosomes sequester cellular components, undergo further maturation steps and finally fuse with lysosomes leading to the degradation of the autophagic cargo by hydrolases. The resultant membranes and nutrients are recycled back to the cytosol for reuse (Figure 1). Thus, each stage of the autophagy process, with its own set of regulatory factors, provide a potential point of dysfunction in disease. Here we provide more details on the autophagy steps implicated in ALS/FTD pathogenesis, providing a context to the disease-associated changes reviewed later.

The signalling mechanisms leading to autophagy activation under different types of cellular stressors have been extensively investigated (Yu L. et al., 2018). Two well-known kinases are implicated in the control of this tightly regulated the initiation process: the nutrient sensor, mTOR (mammalian target of rapamycin) complex 1 (mTORC1) and the energy-sensing AMP-activated protein kinase (AMPK), whose inhibition or activation, respectively, triggers autophagy. The ULK1 (Unc-51-like autophagy-activating kinase) complex, which comprises ULK1 itself and ATG13, FIP200 (FAK family kinase interacting protein of 200 kDa; also known as RB1CC1) and ATG101, integrates upstream mTORC1 and AMPK signalling to coordinate the induction of the first steps of the autophagy process (Saxton and Sabatini, 2017).

To initiate autophagosome nucleation, the active ULK1 complex, together with the only multiple transmembrane-spanning ATG protein, ATG9, are translocated to autophagosome formation sites (Mercer et al., 2018); these have been described at different cellular compartments including the plasma membrane, Golgi, mitochondria, lipid droplets, but in particular at the ER (reviewed by Gómez-Sánchez et al., 2021). This translocation is followed by the recruitment of the VPS34/class III PI3K complex (comprising VPS34/PIK3C3, VPS15, Beclin1, and ATG14). Through phosphorylation of members of this complex, the ULK1 promotes PI3KC3 activity, generating the phospholipid PI3P (phosphatidylinositol 3-phosphate) to enable vesicle elongation. PI3P incorporation promotes the recruitment of proteins with PI3P-binding domains, such as FYVE domain-containing proteins (e.g. ALFY, autophagy FYVE-linked protein), which helps in the maturation of the new structure, the phagophore.

Two ubiquitin-like protein cascades, which are highly conserved, are required for the next steps. Two ubiquitin-like proteins, ATG12 and LC3 are conjugated to ATG5 or the lipid phosphatidylethanolamine (PE), respectively. Similar to ubiquitin, covalent conjugation of these proteins is achieved through a cascade of activities that are catalysed by E1 activating enzymes, E2 conjugating enzymes, and E3 ligases (Mizushima, 2020). Briefly, the ATG4 protease cleaves LC3 by removing C-terminal amino acids and exposing a glycine residue. This glycine is used in the first cascade mediated by the E1-like ATG7, followed by ATG3, an E2-like enzyme. Next, in the second cascade, the ATG12/ATG5–ATG16L1 complex, (elongation complex), is formed, involving the conjugation of ATG12 to ATG5 by the E1-like ATG7 and the E2-like ATG10 and the association of ATG16L1 with ATG5. The elongation complex, considered to have E3-like activity, determines the lipidation site of LC3 a process required for the association of LC3 with the autophagosomal membrane (Fujita et al., 2008). During this, LC3 is covalently bound to PE in the membrane (Mizushima, 2020). As the phagophore expands, LC3-PE (also known as LC3-II) is incorporated into the membrane. Consequently, the level of LC3-II can be used to determine the number of autophagosomes in cells (Klionsky et al., 2021). Finally, the edges of the phagophore fuse (vesicle completion) to form the autophagosome, which engulf the cargo for degradation. The endosomal sorting complexes required for transport (ESCRT) machinery is required for this step (Takahashi et al., 2018; Flower et al., 2020). Membrane-bound LC3-PE controls critical steps during the autophagy process apart from the autophagosome closure, such as movement of the autophagosomes or its fusion with the lysosomes, and degradation of the autolysosome inner membrane (Mercer et al., 2018). LC3 is also a key molecule for the binding to adapter proteins in selective autophagy, as further reviewed later.

Autophagosomes undergo fusion with endolysosomal compartments, including early endosomes, multivesicular bodies and late endosomes to form amphisomes (maturation). Autophagosomes/amphisomes are transported along microtubules to the perinuclear region where they fuse with lysosomes to form autolysosomes. The steps of docking and fusion involve a large set of molecules, including cytoskeleton components and related motor proteins, Rab GTPases, SNAREs and other proteins (Kriegenburg et al., 2018). Most of the molecular machinery that is responsible for these events involve proteins that are shared with the endocytic pathway (Mercer et al., 2018). Finally, the enclosed cargo is degraded by lysosomal hydrolases, allowing for the release and recycling of the newly generated metabolites and lysosomal reformation. An acidic environment in the lysosomal lumen, generated by the vacuolar H⁺ ATPase (V-ATPase) at the lysosomal membrane, is essential for this degradation (Yu L. et al., 2018).

Autophagy is also controlled at the transcriptional regulation level (di Malta et al., 2019). The transcription factor TFEB (transcription factor EB) and related MiTF/TFE (microphthalmia-associated transcription factor/transcription factor E) family members recognize and bind to coordinated lysosomal expression and regulation (CLEAR) motifs in the promoter region of many lysosomal and autophagy genes, leading to an increase in their transcription and a global enhancement of autophagy flux (Sardiello et al., 2009; Palmieri et al., 2011; Settembre et al., 2012; Martina et al., 2014; Ploper et al., 2015). We here highlight the role of the master regulator TFEB. TFEB shuttles between the nucleus and the cytoplasm, which is mainly regulated by mTORC1 and Rag GTPases on the lysosomal membrane (Puertollano et al., 2018). In response to amino acid signals, Rag GTPases recruit TFEB to the lysosome where it is phosphorylated by mTORC1, inhibiting its nuclear translocation and activation (Puertollano et al., 2018). Therefore basally, TFEB remains in its inactive phosphorylated form in the cytosol. Upon nutrient deprivation, or other cellular stressors, such as lysosomal damage or oxidative stress (Jia et al., 2018; Wang H. et al., 2020), TFEB gets dephosphorylated (activated) and translocates into the nucleus. Here, it enhances the expression of important genes for the autophagy-lysosomal pathway such as lysosomal proteins, for example, LAMP1/2 and hydrolases, and key proteins of the autophagy machinery such as beclin1 or p62 (Puertollano et al., 2018).

New studies have provided evidence for an additional regulatory layer of the autophagy process at the post-transcriptional level, governed by RBPs which regulate the processing and translation of specific autophagy-related transcripts (Sakellariou et al., 2021). Recently, noncoding RNAs have also been added to this regulatory function, furthering the interplay of autophagy and RNA (Xu et al., 2017; Yang et al., 2017; Zhao et al., 2020). However, although post-transcriptional/co-translational regulation of autophagy has received increasing attention in recent years, it is still poorly understood.

Apart from the conventional RNA decay pathways (reviewed by Houseley and Tollervey, 2009; Tatosyan et al., 2020), autophagy-mediated RNA degradation complements these quality control mechanisms to remove obsolete/defective RNA molecules. Additionally, RBPs and RNA-protein complexes, such as RNA granules, are also autophagy substrates, indirectly affecting RNA homeostasis. RNA degradation by autophagy was initially suggested in early observations from the 80–90s (Sameshima et al., 1981).These pioneering studies demonstrated increased RNA degradation during starvation-induced autophagy. In agreement with a lysosomal RNA degradation pathway, several acid ribonucleases (RNases) have been identified within lysosomes (Schröder et al., 2010). Interestingly, loss-of-function mutations in the RNase RNASET2 leads to familial cystic leukoencephalopathy and displays lysosomal accumulation of ribosomal RNA in neurons and undigested substrates in microglia (Haud et al., 2011; Hamilton et al., 2020). Although the role of autophagy-dependent RNA catabolism has not been fully characterized, recent studies have described three different mechanisms - RNautophagy, ribophagy and granulophagy (Figure 2).

Uptake of mRNA molecules into lysosomes by the autophagy mechanism of RNautophagy does not require the formation of a double-membrane vesicle but instead resembles CMA (Schröder et al., 2010; Haud et al., 2011; Hamilton et al., 2020). During RNautophagy, the uptake occurs by direct binding of RNA to the lysosomal membrane proteins LAMP2C or SIDT2 (SID1 transmembrane family member 2), which act as nucleic acid receptors in an ATP-dependent manner (Fujiwara et al., 2013). The cytoplasmic region of LAMP2C/SIDT2 binds to RNA/DNA via its arginine-rich motif and possesses binding preference for stretches of poly-guanosines (Fujiwara et al., 2015; Hase et al., 2015; Hase et al., 2020). This binding may be regulated by other RNA-interacting proteins such as RBPs or DNA-binding proteins in the cytosol. Indeed, the ALS/FTD linked RBP hnRNPA1 was found to interact with LAMP2C in an RNA-dependent manner, in addition to a range of other RBPs (Fujiwara et al., 2013). In vitro studies have identified other lysosomal membrane proteins as RNA interactors, such as the mammalian CMA effector HSPA8 and the yeast lysosomal membrane V-ATPase subunit Vma1; however, the biological significance of these findings in terms of RNA-mediated autophagy degradation has not yet been determined (Castello et al., 2012). Indeed, further work is required to understand the substrates, sensors and regulation in RNautophagy. At present, the relative contributions of conventional RNA decay and this autophagy-mediated pathway in mRNA degradation are unclear.

Ribosomal RNA (rRNA) is the primary component of ribosomes and represents about 80% of total cellular RNA (Darnell, 1968; Warner, 1999; Weinberg et al., 2016). There are several pathways of eukaryotic ribosomal RNA decay (reviewed in Inada, 2020). However, ribosomes can also be targeted and removed by autophagy (An and Harper, 2020). Indeed, in electron microscopy studies, ribosomes are frequently detected inside autophagic structures in starved cells, leading to the general notion that ribosomes are non-selective cargo during bulk autophagy (Eskelinen et al., 2011). However, a growing body of evidence suggests that selective autophagic mechanisms directed to ribosomes (ribophagy) are employed in both yeast and mammalian cells (Beese et al., 2020). In mammalian cells, the protein NUFIP1 (nuclear fragile X mental retardation-interacting protein 1) has been identified as a ribophagy receptor for the 60S subunit, interacting with LC3B and ribosomes in response to nutrient stress (Wyant et al., 2018). However, a recent work has challenged this role, finding that NUFIP1 deletion has no effects on ribophagy flux (An et al., 2020). Indeed, a study from the same group has suggested that the overall contribution of ribophagy to ribosome homeostasis during nutrient stress is small relative to other pathways, such as the ubiquitin-proteasome system (An and Harper, 2018; An et al., 2020).

mRNAs exiting translation often accumulate in distinct cytoplasmic RNA-protein complexes, such as SGs and P-bodies. These are cytoplasmic membraneless organelles comprised of RNA and RBPs with key roles in the regulation of gene expression and cellular homeostasis (Alberti and Carra, 2018). SGs are assembled during a stress response, when global translation initiation is inhibited; thus, they are transient, reversible structures and their formation constitutes an adaptation mechanism for maintaining RNA metabolism homeostasis during unfavourable environmental conditions. Indeed, several studies have shown a pro-survival role for SGs (Reineke and Neilson, 2019; Samir et al., 2019). Interestingly, recent evidence suggests that translation of mRNAs can take place at the boundary of SGs, arguing against an exclusive role for SGs in inhibition of protein synthesis (Mateju et al., 2020). Physiologically, once the stress is resolved, SGs can be cleared by several mechanisms including disassembly by molecular chaperones (Ganassi et al., 2016; Mateju et al., 2017; Mediani et al., 2021), clearance by the ubiquitin–proteasome system (Turakhiya et al., 2018) and by selective autophagy – thus termed granulophagy (recently reviewed in Alberti and Carra, 2018). Under basal conditions only 5%–10% of SGs are degraded by granulophagy. However, autophagy-dependent SG degradation increases dramatically during stress and disease (Buchan et al., 2013; Mateju et al., 2017; Chitiprolu et al., 2018), which will be discussed later.

Autophagic clearance of SGs has been implicated by their recruitment of key autophagic proteins, such as p62 and LC3 (Buchan et al., 2013; Ganassi et al., 2016). A key study demonstrated that the molecular chaperone Cdc48 (human homologue: VCP), an ALS/FTD associated protein, is critical to autophagic degradation of SGs in yeast (Buchan et al., 2013). However, another study has shown that ULK1/2 localises to SGs where it phosphorylates VCP and increases its potential to drive SG disassembly in mammalian cells (Wang et al., 2019), which was not shared by other effectors of autophagy. Therefore, demonstrating both autophagy dependent and independent roles for classical autophagy proteins in SG clearance. Generally, selective autophagy involves the targeting of cargo by ubiquitination; this modification is recognized by the ubiquitin-associated domain of autophagy receptors (Conway et al., 2020), such as p62 and optineurin, which are also associated with ALS/FTD. However, one study has reported an alternative C9orf72 protein-mediated ubiquitin-independent granulophagy (Chitiprolu et al., 2018). Instead, symmetric arginine methylation of SG components by PRMT5 (protein arginine methyltransferase 5) was necessary for the selective targeting of SGs. This non-canonical signal was specifically recognized by p62 in a complex with C9orf72 via the Tudor protein SMN (survival motor neuron protein). Interestingly, impaired autophagy can also lead to anomalies in SG formation and morphology, which may be due to aberrant recruitment of defective ribosomal proteins (Seguin et al., 2014). These abnormal SGs contained noncanonical components, such as the 60S ribosomal subunit, which is normally absent. Furthermore, p62 and the autophagy receptor CALCOCO1 (also known as NDP52) have been implicated in autophagy-mediated degradation of retrotransposon RNA, localized to both RNA granules, P-bodies and SGs. This autophagy-mediated control in retrotransposon RNA degradation affects their insertion within the genome, which might be relevant for tempering somatic mosaicism (Guo et al., 2014). These selective mechanisms share the essential autophagy steps and machinery reviewed above, and so any impact on them may have a concomitant effect on RNA-selective autophagy degradation.

In addition, there are noncanonical roles for autophagy in RNA metabolism. Secretory autophagy is a newly defined autophagy function that bypasses the degradative process to allow the secretion of autophagic structures (Xu et al., 2018); of note, nucleic acids and RBPs have been found to be secreted by this pathway (Leidal et al., 2020). Interestingly, this process was also apparent in glial cells. Therefore, in addition to a known role for secretory autophagy in the release of aggregation-prone proteins (Lee J.-G. et al., 2016), the autophagy pathway may be implicated in non-cell autonomous control through excretion of nucleobases and RBPs (van Niel et al., 2018; Xu et al., 2018). LC3 has been reported to be responsible for the specific cargo loading of RBPs and small noncoding RNAs into vesicles to be secreted (Leidal et al., 2020). Interestingly, many RBPs are found to contain LC3-interacting (LIR) consensus motifs (Leidal et al., 2020). Furthermore, LC3 possesses a direct RNA-binding capacity through its own arginine-rich motif (Kraft et al., 2016), which has already been shown to play a role in the translational regulation of certain RNA (Zhou et al., 1997).

Another emerging role for autophagy in RNA metabolism is related to RNA transport, which is especially relevant for maintaining highly polarized cells such as neurons (Salogiannis and Reck-Peterson, 2017). The transport of RNA from neuronal soma to distal sites such as synapses allows local protein synthesis which is essential for the precise spatiotemporal control required for neuronal function. mRNAs associate with RBPs and are transported as RNA-protein complexes. Recent works have described a transport mechanism for RNA granules, known as “hitchhiking”, where they are indirectly transported along microtubules by docking onto endosomes, lysosomes or other membrane-bound organelles (Pohlmann et al., 2015; Gershoni-Emek et al., 2018; Cioni et al., 2019; Liao et al., 2019; Olgeiser et al., 2019). The ALS-related RBP Annexin A11 which is a component of SGs but also has phosphoinositide-binding capacity is reported to act as a molecular tether between RNA granules and lysosomes (Markmiller et al., 2018; Liao et al., 2019). Thus, due to the integration with the endocytic and lysosomal pathways, autophagy may also influence RNA transport.

Several stages of the autophagy pathway, in addition to selective autophagy mechanisms, are affected by ALS/FTD-linked genes (Figure 3). Most of these are ubiquitous expressed, including in neurons and neuroglia, and although the majority of studies have focussed on the role of these genes in neurons, some neuroglia specific roles have been reported. Here, we summarize the roles of these ALS/FTD genes relative to the specific stages of autophagy.

Regarding autophagy-mediated RNA homeostasis in ALS/FTD, the majority of the evidence comes from studies on SGs. As mentioned previously, the contribution of autophagy to the clearance of SGs under basal conditions is low but increases significantly upon cellular stress or in disease (Ganassi et al., 2016; Mateju et al., 2017; Mediani et al., 2021). Indeed, accumulating evidence suggests that SGs are targeted for selective autophagic degradation when granules become pathological (Sun et al., 2020). Although, SGs can be part of a pro-survival mechanism, chronic SGs prevent the normal functionality of a cell. If not efficiently cleared, persistent SGs may form the seed for aggregation of RBPs (Baradaran-Heravi et al., 2020), and chronic optogenetic induction of SGs has been shown to drive neurotoxicity (Zhang et al., 2019). Thus, under disease conditions the selective autophagic degradation of SGs may play a crucial role.

Mutations are found in several ALS/FTD genes encoding proteins involved in selective autophagy. The ALS-associated autophagy receptor p62 is recruited to SGs upon cellular stress (Chitiprolu et al., 2018; Turakhiya et al., 2018); TBK1 (TANK-binding kinase 1; (Cirulli et al., 2015; Freischmidt et al., 2015)) phosphorylates p62 (Pilli et al., 2012; Matsumoto et al., 2015) and another autophagy receptor - optineurin (Matsumoto et al., 2015), which enhances affinity to cargo and promotes selective autophagy of damaged mitochondria or intracellular pathogens (Richter et al., 2016). There have also been reports that TBK1 regulates upstream components of the autophagy signalling cascade such as AMPK (Zhao Y. G. et al., 2018) or the autophagosome-localized SNARE protein Syntaxin-17 (Kumar et al., 2019). Interestingly, SMCR8, a C9orf72 protein binding partner, is also a substrate of TBK1 (Sellier et al., 2016), which may therefore have implications in pathogenesis from the range of roles of the C9orf72-WDR41-SMCR8 complex in autophagy. As explained above, C9orf72, p62 and SMN are also involved in an alternative ubiquitin-independent granulophagy dependent on symmetric arginine methylation of SG components. Indeed, C9orf72-ALS patient cerebellar tissue exhibits an accumulation of symmetrically arginine-methylated proteins (Chitiprolu et al., 2018), suggesting deficient turnover of SGs, and deregulation of autophagy is observed in cell culture and animal models of SMN-associated neurodegeneration (Garcera et al., 2013; Custer and Androphy, 2014). Although in most cases, studies have not investigated a direct link between granulophagy and ALS/FTD associated autophagy genes, disease associated mutations in VCP (Gitcho et al., 2009; Johnson et al., 2010) have been shown to induce persistent SGs which also contain TDP-43 (Buchan et al., 2013; Gwon et al., 2021) and depletion or inhibition of VCP impairs SG formation (Seguin et al., 2014). Interestingly, the C9orf72 DPR poly(GA) has been shown to sequester VCP (Božič et al., 2021), which may similarly inhibit VCP’s function in selective autophagy and/or SG dynamics. VAPB patient muscle biopsies also contain VAPB aggregates that colocalise with TDP-43 and the SG protein TIAR1 (Tripathi et al., 2021), suggesting impaired autophagy-dependent SG clearance.

Perturbed RNA granule dynamics has drawn much attention from the ALS-FTD field in recent years; pathological granules transition from a rapidly reversible liquid state to a more solid gel-like state which renders them less dynamic and persistent (Wolozin and Ivanov, 2019; Solomon et al., 2021). Investigation of the C9orf72 arginine-rich DPR interactomes found a wide number of RBPs and were enriched in those involved in membraneless organelles including SGs components and major ALS/FTD RBP proteins, such as TDP-43 and hnRNPA1 (Lee K.-H et al., 2016; Boeynaems et al., 2017; Chew et al., 2019). Consequently, these arginine-rich DPRs—both poly(GR) and poly(PR) - have been shown to alter SG composition and dynamics by disrupting multivalent interactions and promoting the formation of pathological SGs (reviewed by Solomon et al., 2021). Similarly, mutants of several ALS/FTD-related RBPs, such as FUS and TDP-43, promote the formation of poorly dynamic SGs (Dewey et al., 2011; Baron et al., 2013; Murakami et al., 2015; Patel et al., 2015; Gopal et al., 2017; Ding et al., 2021). SGs can also accumulate defective ribosomal proteins (Seguin et al., 2014) and misfolded proteins, such as mutant SOD1 (Mateju et al., 2017), which can drive them into a pathological state (Alberti and Carra, 2018). In line with their role in the clearance of pathological SGs, persistent mutant FUS-positive SGs were found to co-localize with LC3-II (Ryu et al., 2014). Pharmacological activation of autophagy can also reduce FUS-and TDP-43-related cytotoxicity, which was associated with a reduction in the number of TDP-43- or FUS-positive SGs (Wang et al., 2012; Boyd et al., 2014; Ryu et al., 2014; Marrone et al., 2019). The ALS/FTD protein ubiquilin 2 can also be recruited to SGs where its interaction with the RBP FUS is proposed to cause FUS dissociation from SGs, a process which is impaired in disease mutants (Alexander et al., 2018; Dao et al., 2018). Optineurin can also modulate SG dynamics and clearance in autophagy dependent and independent ways; in the latter, reduced optineurin or disease mutants lead to an upregulation of TIA1 expression causing reduced SG clearance and accumulation of ubiquitinated TDP-43 (Kakihana et al., 2021). Furthermore, SGs also participate in autophagy regulation by recruiting signalling molecules, such as the components of the mTORC1 complex, raptor and mTOR; this association regulates TORC1 inactivation-reactivation during cell stress and recovery (Takahara and Maeda, 2012; Yan et al., 2012; Wippich et al., 2013). Sequestration of these signalling molecules or other RBPs into poorly dynamic SGs may compromise the autophagy–lysosomal pathway indirectly.

Finally, autophagic degradation of SGs and other RNA granules also relies on proper trafficking. As noted previously, ALS-associated variants in ANXA11 have been shown to disrupt RNA granule-lysosome docking, impeding their transport in neurons (Liao et al., 2019) and ANXA11 mutants can impede calcium homeostasis and stress granule disassembly (Nahm et al., 2020).

In summary, although thus far relatively few studies have connected autophagy and RNA homeostasis directly, there is clear evidence for a significant impact of autophagy on RNA homeostasis in FTD/ALS, particularly in relation to the clearance of pathological SGs.

Current knowledge reviewed here has established a multilayer understanding of autophagy regulation as well as novel non-canonical functions for the autophagy machinery. An emerging and less-well understood area of autophagy research involves its role in RNA metabolism. Here we discussed new evidence of specialized mechanisms for selective autophagy-mediated RNA degradation and other autophagy functions also affecting RNA homeostasis. We have also described how RBPs might not serve only as autophagic cargos but also exert regulatory functions on the autophagy process, adding additional complexity to the RNA-autophagy interplay.

This complex autophagy-RNA interplay is well evidenced in ALS/FTD, where dysregulated RNA homeostasis and autophagy defects are major interconnected disease mechanisms. We discussed here the role of ALS/FTD-associated proteins in the different stages of autophagy, disrupting general autophagy but also specific RNA-mediated autophagy, with SGs as major substrates in this process. The evidence presented here reinforce a unifying connection between many ALS/FTD genes and suggest a model in which pathogenic variants that dysregulate autophagic clearance and promote an accumulation of pathological RNA granules and drive cytoplasmic protein aggregation culminate in ALS/FTD.

The mechanisms associated to the C9orf72 mutation have been a focus of the ALS/FTD research field as major known cause of disease but also for its intriguing complex disease mechanisms. C9orf72 mutation studies nicely illustrate the relevance and complexity of the autophagy–RNA homeostasis interplay in health and disease. As reviewed above, as part of the C9orf72-WDR41-SMCR8 complex, the C9orf72 protein is involved in the initiation of autophagy and localised to the lysosome. Here, loss of C9orf72 results in inappropriate sorting of hydrolase precursors, impaired synaptic vesicle recycling, lysosomal accumulation, and alterations in glial secretomes. This latter finding has wide implications on the neuroimmune system and may contribute to neuroinflammation and non-cell autonomous neurotoxicity. It can also impact gene regulation of autophagy via its inactivation of mTORC and autophagosome maturation and trafficking via interaction with cofilin. Through these roles haploinsufficiency of the C9orf72 protein can affect RNA homeostasis by selective autophagy. However, the C9orf72 protein additionally impacts granulophagy by a direct role in a ubiquitin-independent pathway targeting arginine-methylated SG substrates for degradation.

In C9orf72-FTD/ALS, gain-of-function mechanisms emanating from the repeat are proposed to be the dominant cause of neurodegeneration (Mizielinska et al., 2014; Moens et al., 2019), and both repeat RNA and DPRs can affect the autophagy-RNA homeostasis interplay. Dysfunctional nucleocytoplasmic transport from both the repeat RNA and DPRs may impact the nuclear translocation of transcription factors that regulate autophagy genes (Kim and Taylor, 2017; Solomon et al., 2021); indeed, the master regulator TFEB is affected by expression of the C9orf72 repeat in Drosophila and cell models, an effect that appeared to be predominantly driven by repeat RNA with only a small effect from poly(GA) (Cunningham et al., 2020). The arginine rich DPRs bind to and disrupt a wide range of membraneless organelles (Boeynaems et al., 2017; Lee K.-H et al., 2016), including the nuclear pore through which all nucleocytoplasmic transport occurs, thus may contribute to the aforementioned disruption. However, as described above the arginine rich DPRs also bind SG components and perturb SG dynamics, which may result in the persistent pathological SGs that are targeted for degradation by autophagy. As noted, these effects are associated with the arginine rich DPRs which may mimic other arginine rich motifs such as the RGG motif which is common to RBPs involved in RNA homeostasis; similarly to the arginine rich DPRs, this motif permits interaction with both proteins and nucleic acids (Rajyaguru and Parker, 2012). This motif also undergoes regulation by arginine methylation—a signal already described to target cargo to SGs for autophagic degradation involving the C9orf72 protein (Chitiprolu et al., 2018). Indeed, the arginine rich DPRs also undergo arginine methylation, with different methylation types regulating its pathogenicity (Sakae et al., 2018; Gittings et al., 2020). Finally, the aggregates formed by the C9orf72 repeat RNA and DPRs can sequester proteins, for example the sequestration of VCP by poly(GA) (Božič et al., 2021), which promotes further dysfunction by these pathways. Gain-of-function DPR pathomechanisms can also be exacerbated by the coincident loss of C9orf72 protein levels in C9orf72-ALS/FTD which reduce their autophagic clearance. Indeed, complete knockout or reduction of C9orf72 exacerbates reduced survival in C9orf72 repeat expansion models in neuron-like cell lines, zebrafish and mice (Shao et al., 2019; Boivin et al., 2020; Zhu et al., 2020) and from DPR expression in control patient induced motor neurons (Shi et al., 2018). Conversely C9orf72 protein expression can ameliorate reduced lysosomal number, DPR accumulation and degeneration in C9orf72-ALS/FTD patient induced motor neurons (Shi et al., 2018).

Although studies have identified an effect of loss of C9orf72 protein function on clearance of DPRs, it has not yet been investigated whether it could also affect the degradation of C9orf72 repeat RNA, given the emerging evidence for the role of autophagy in RNA catabolism. Indeed, the mechanisms for the degradation of repeat RNA in repeat expansion disorders are still poorly understood. A recent study has revealed a role for the RNA exosome, which is generally involved in RNA quality control, in the degradation of the C9orf72 sense and antisense repeat transcripts (Kawabe et al., 2020). The RNA exosome exerts its functions in an autophagy/lysosomal-independent manner. Instead, the catalytic component EXOSC10 promotes the substrate loading into the tunnel of the barrel-like core of the RNA exosome complex for its degradation (Kilchert et al., 2016). However, additional mechanisms for C9orf72 expanded G₄C₂ repeat RNA are tempting to hypothesis when the RNautophagy receptors LAMP2C and SIDT2 have already shown to have binding preference for stretches of consecutive guanosines (Hase et al., 2015). Indeed, SIDT2 was found to interact with expanded CAG repeats in exon1 of the HTT transcript, which code for polyglutamine-expanded proteins, linked to Huntington’s disease. Importantly, overexpression of SIDT2 promoted degradation of HTT expanded transcripts, reducing the levels of polyglutamine-expanded huntingtin aggregates. Thus, C9orf72-derived expanded G₄C₂ repeat RNA could be directed to lysosomes for RNautophagy degradation in a similar manner. Although, in the case of the C9orf72 repeat RNA, hairpin and G-quadruplex structures will need to be considered (Fratta et al., 2012; Haeusler et al., 2014). It is also of note that the C9orf72 protein itself is a substrate for autophagy in a cell-type dependent manner (Leskelä et al., 2019).

In summary, both synergistic and feedback loops in autophagy and RNA homeostasis contribute to C9orf72-ALS/FTD pathogenesis, from gain and loss of function mechanisms. Loss of the C9orf72 protein may impact several stages of autophagy as well as granulophagy directly, which can exacerbate gain of function mechanisms by compromising DPR protein and potentially repeat RNA clearance. Then, the DPRs and repeat RNA can also impact autophagy and RNA homeostasis themselves, through effects on transcription factors, SG dynamics and potentially RNautophagy. Interestingly, a recent genome-wide data study has revealed that autophagy genes are significantly associated with ALS risk in C9orf72 expansion carriers, supporting a modifying role for autophagy from a genetic perspective as well (Saez-Atienzar et al., 2021). But it does not end there, as with the majority of ALS and a large proportion of FTD cases, the C9orf72 mutation is associated with pathology of the RBP TDP-43, and RBPs can also modulate autophagy. Indeed, the ALS/FTD-associated RBPs TDP-43, FUS and hnRNPA1 can all regulate the transcription of autophagy-related genes, with TDP-43 also able to modulate the transcription factor FOXO and FUS to both initiation and maturation of autophagosomes. Conversely, many ALS/FTD-linked genes can impact autophagy and RNA homeostasis via contribution to granulophagy, with evidence for the proteins encoded by SQSTM1 (p62) and VCP playing crucial roles.

Another point to consider is that multiple different cell types contribute to ALS/FTD pathologies (Ragagnin et al., 2019; Vahsen et al., 2021) and the autophagic-RNA homeostasis interplay may vary between these. Indeed, accumulation of insoluble protein, often RBP, inclusions are found in both neurons and neuroglia of ALS/FTD post-mortem tissue indicative of impaired protein clearance (Neumann et al., 2007). Of particular interest is also the role for secretory autophagy in the release of extracellular vesicles (di Malta et al., 2012; Verkhratsky et al., 2016) and excretion of nucleobases and RBPs (Shah et al., 2018; Varcianna et al., 2019; Leidal et al., 2020) from glial cells and their potential contribution to non-cell autonomous patho-mechanisms in ALS/FTD. However, as much of the research into the interplay of autophagy and RNA catabolism is very recent, further studies will be required to dissect contributions of different cell types to pathology via these pathways, which has implications for therapeutic targeting.

Indeed, an important point to consider is that these studies have opened a new avenue of research searching for therapeutic approaches for these fatal diseases. Autophagy has long been explored as a therapeutic target for ALS/FTD with contradictory results, depending on the ALS/FTD causing gene or disease-model. For instance, pharmacologically autophagy activation exacerbated neurotoxicity in Drosophila TARDBP (TBPH) knockout and SOD1^G93A mice (Zhang et al., 2011; Hsueh et al., 2016; Xia et al., 2016; Zhou et al., 2017). While a neuroprotective role for autophagy induction was reported in several FUS and TARDBP disease models (Caccamo et al., 2009; Wang et al., 2012; Burkhardt et al., 2013; Barmada et al., 2014; Boyd et al., 2014; Cheng et al., 2015) or even in the same SOD-1 mouse model (SOD1^G93A) (Fornai et al., 2008; Mancuso et al., 2014; Zhang T. et al., 2014). However thus far relatively little consideration has been given to the impact of targeting RBPs on autophagy, and we are not aware of any translational approaches for autophagy-dependent RNA catabolism in the ALS/FTD field as of yet.

In summary, we have reviewed the wide evidence for the contribution of ALS/FTD-related proteins and RNA to disrupt the different stages of autophagy. Through this dysfunction they have the potential to impact RNA catabolism in addition to their roles in direct RNA homeostasis deregulation. Of particular note is the interplay at the level of SG clearance (granulophagy), as alterations in SG dynamics and the contribution to disease pathology is of major interest to the field. Thus, future research may focus on more specific pharmacological interventions targeting pathological SGs for clearance via autophagy-dependent mechanisms. In addition, increasing our understanding of the molecular mechanisms of autophagy-dependent RNA homeostasis and the context in which these processes become dysfunctional in disease will be beneficial to design effective therapeutics particularly in ALS/FTD.