The roles of intrinsic disorder-based liquid-liquid phase transitions in the “Dr. Jekyll–Mr. Hyde” behavior of proteins involved in amyotrophic lateral sclerosis and frontotemporal lobar degeneration

ABSTRACT

Pathological developments leading to amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) are associated with misbehavior of several key proteins, such as SOD1 (superoxide dismutase 1), TARDBP/TDP-43, FUS, C9orf72, and dipeptide repeat proteins generated as a result of the translation of the intronic hexanucleotide expansions in the C9orf72 gene, PFN1 (profilin 1), GLE1 (GLE1, RNA export mediator), PURA (purine rich element binding protein A), FLCN (folliculin), RBM45 (RNA binding motif protein 45), SS18L1/CREST, HNRNPA1 (heterogeneous nuclear ribonucleoprotein A1), HNRNPA2B1 (heterogeneous nuclear ribonucleoprotein A2/B1), ATXN2 (ataxin 2), MAPT (microtubule associated protein tau), and TIA1 (TIA1 cytotoxic granule associated RNA binding protein). Although these proteins are structurally and functionally different and have rather different pathological functions, they all possess some levels of intrinsic disorder and are either directly engaged in or are at least related to the physiological liquid-liquid phase transitions (LLPTs) leading to the formation of various proteinaceous membrane-less organelles (PMLOs), both normal and pathological. This review describes the normal and pathological functions of these ALS- and FTLD-related proteins, describes their major structural properties, glances at their intrinsic disorder status, and analyzes the involvement of these proteins in the formation of normal and pathological PMLOs, with the ultimate goal of better understanding the roles of LLPTs and intrinsic disorder in the “Dr. Jekyll–Mr. Hyde” behavior of those proteins.

KEYWORDS:

• amyotrophic lateral sclerosis
• frontotemporal lobar degeneration
• intrinsically disordered proteins
• liquid-liquid phase transition
• neurodegeneration
• proteinaceous membrane-less organelles

1. Introduction

Amyotrophic lateral sclerosis (ALS), also known as Lou Gehrig's disease, is a fatal form of the neurodegeration that represents a specific neuromuscular disease characterized by the degeneration of upper and lower motor neurons of the spinal cord leading to the progressive atrophy of abdominal, bulbar, limb, and thoracic muscles.^1,2 Respiratory failure is the most common cause of death of the afflicted patients, typically within 3–5 y after onset of the ALS symptoms.³ Despite being much less abundant than other neurodegenerative maladies, such as Alzheimer disease (AD) or Parkinson disease (PD), ALS is considered as the most common neurodegenerative disorder that affects motor neurons in adults, and is particularly infamous for its devastating physiological effects and rapid lethality. As with many other neurodegenerative diseases, most ALS cases are sporadic, and only about 8–10% of ALS cases are inherited.^1-3 As it is a neurodegenerative disorder with the progressive degeneration of motor neurons in the motor cortex of the brain, the brainstem motor nuclei, and the anterior horns of the spinal cord, progression of ALS is characterized by the spatio-temporal spread of pathology, where early stages are manifested by local symptoms, such as subtle cramping or weakness in the bulbar or limb muscles, that rapidly culminate in the global paralysis of almost all skeletal muscles and invariably lead to death.⁴ The characteristic symptom of ALS is progressive amyotrophy, or muscle wasting, caused by muscle denervation; i.e., the degeneration of spinal motor neurons results in atrophy of their target muscles.⁴ Since degeneration of different sets of motor neurons would affect different regions of the body, ALS is clinically characterized by its considerable heterogeneity, and multiple subsets of the disease can be distinguished. For example, based on the clinical manifestations at symptom onset, patients with sporadic ALS can be grouped into limb- and bulbar-onset subtypes, where limb onset cases are associated with the limb motor symptoms, such as weakness or atrophy or fasciculation, and where bulbar onset cases are characterized by the presence of only bulbar signs, such as dysarthria, swallowing difficulty, and tongue fasciculation.^5,6 Importantly, patients with limb- and bulbar-onset ALS subtypes are characterized by the significant difference in brain atrophy patterns and in clinical progression, which is more rapid in bulbar-onset than in limb-onset ALS.^5,7 In addition to the classical (Charcot) ALS, this spectrum of disorders includes several other syndromes, such as ALS with multisystem involvement (e.g., ALS-dementia); flail arm or Vulpian-Bernhardt syndrome; flail leg syndrome, also known as pseudopolyneuritic form; progressive bulbar palsy, PBP; primary lateral sclerosis, PLS; and progressive muscular atrophy, PMA.⁶ Among the pathological hallmarks of ALS are shrinking and degeneration of spinal motor neurons, and accumulation of rounded or thread-like deposits of aggregated proteins (collectively known as inclusions) in the cytoplasm of affected motor neurons.⁴ These inclusions are often ubiquitinated, and, in most cases of ALS, contain ubiquitinated TARDBP/TDP-43 (TAR DNA binding protein) as the major component.⁸

Because many ALS patients show cognitive impairment in addition to muscle atrophy, and develop frontotemporal lobar degeneration (FTLD, which is second to AD as a cause of dementia in patients under the age of 65, see ref. 9),¹⁰ and because many FTLD patients show ALS symptoms,¹¹ it is thought now that ALS and FTLD constitute 2 parts of a clinical spectrum of a disease.¹² It was also pointed out that the overlapping pathogenesis of ALS and FTLD indicated that in these diseases, motor neuron degeneration and cognitive deficits might originate due to a common molecular basis.² In fact, both of these diseases are proteinopathies, because they are associated with the presence of deposits or inclusions containing various misfolded proteins, such as ATXN2 (ataxin 2); C9orf72; dipeptide repeat proteins, DPRs; FUS (FUS RNA binding protein); HNRNPA1 (heterogeneous nuclear ribonucleoprotein A1); HNRNPA2B1 (heterogeneous nuclear ribonucleoprotein A2/B1); NEK1 (NimA related kinase 1); OPTN (optineurin); SOD1 (superoxide dismutase 1); TARDBP (TAR DNA binding protein); UBQLN2 (ubiquilin 2); VCP (valosin containing protein); or members of the ubiquitin-proteasome system (UPS). Therefore, pathologically, ALS and FTLD can be grouped into multiple subtypes based on the proteins involved in the formation of corresponding inclusions. These corresponding subtypes are ALS–ATXN2, ALS–C9orf72 or ALS–DPR, ALS–FUS, ALS–HNRNPA1, ALS–HNRNPA2B1, ALS–NEK1, ALS–OPTN, ALS–SOD1, ALS–TARDBP, ALS–UBQLN2, and ALS–VCP, or FTLD–DPR, FTLD–FUS, FTLD–MAPT, FTLD–TARDBP, and FTLD–UPS.²

Although all aforementioned subtypes are sporadic forms of ALS and FTLD, proteinaceous inclusions of many familial forms of these diseases contain defective and mutated forms of proteins typically seen in sporadic diseases. The corresponding examples of familial ALS cases include pathologies related to the mutated or otherwise altered forms of ATXN2 (polyQ expansion), C9orf72 (hexanucleotide repeat expansions of the intronic region of the C9orf72 gene between the noncoding exons 1a and 1b), FUS (mutations), OPTN (exon deletion, mutations), SOD1 (mutations), TARDBP (mutations), UBQLN2 (mutations), and VCP (mutations). In familial forms of FTLD, the genetic alterations include mutations in GRN/PGRN (granulin precursor) and MAPT, hexanucleotide repeat expansions near the C9orf72 gene,¹³ and mutations in several rare FTLD-related genes (or FTLD/ALS-related genes), such as CHMP2B (charged multivesicular body protein 2B); CHCHD10 (coiled-coil-helix-coiled-coil-helix domain containing 10); FUS; HNRNPA1; HNRNPA2B1; OPTN; SQSTM1 (sequestosome 1); TBK1 (TANK binding kinase 1); TARDBP; and VCP.¹⁴ It is important to note that sporadic ALS and FTLD often arise from spontaneous mutations that were not inherited.^6,15 For example, mutations in the TARDBP gene have been linked to both familial and sporadic ALS.^16,17 Furthermore, increased susceptibility to sporadic ALS is thought to be associated with mutations in APOE (apolipoprotein E),¹⁸ or in the KSP repeat region of the NEFH (neurofilament heavy) gene,^19,20 as well as with the decreased expression of the glutamate transporter SLC1A2/EAAT2 (solute carrier family 1 member 2)^21,22 or alterations in the VEGF (vascular endothelial growth factor) gene.²³ Therefore, there is a reason to think that all ALS subtypes may have a genetic root cause, whether or not familial inherited.

There are many mechanisms by which different proteins implicated in ALS or FTLD can be related to the pathogenesis of these diseases. These could be grouped into 2 major categories: loss of physiological function (the protein does not do what it is supposed to do), and gain of pathological function (the protein does what it should not do). Various factors, such as mutations, posttranslational modifications (PTMs), processing, changes in the environmental conditions (pH, temperature, ionic strength, exposure to membrane or metal ions, or some other small molecules, etc.), toxic insults, and failure of the protective system can lead to misrecognition, mistrafficking, mislocalization, misfolding, and aggregation, eventually resulting in the loss of normal functionality or the gain of pathological functions. In addition to all these factors, many of which were targets of countless studies whose outputs are covered in numerous dedicated reviews, a new threat is coming into the light of modern research, namely, the ability of some of the proteins to be engaged in the physiological liquid-liquid phase transitions (LLPTs) leading to the formation of various proteinaceous membrane-less organelles (PMLOs, such as different cytoplasmic RNA granules, including stress granules [SGs]). The goal of this review is to look into this mechanism using several illustrative examples of ALS- and FTLD-related proteins. To this end, we will first introduce some of the ALS- and FTLD-related proteins that were shown to be engaged in the LLPTs or related to the PMLOs and describe normal and pathological functions of those proteins. Then, we will consider the roles of these proteins, such as SOD1, TARDBP, FUS, C9orf72, and dipeptide repeat proteins generated as a result of the translation of the intronic hexanucleotide expansions in the C9orf72 gene, PFN1, GLE1, PURA, FLCN, RBM45, SS18L1/CREST, HNRNPA1, HNRNPA2B1, ATXN2, MAPT, and TIA1 in the formation of PMLOs. We will also analyze a link between the normal and pathological LLPTs, and illuminate the role of liquid-liquid phase transitions and intrinsic disorder in the “Dr. Jekyll–Mr. Hyde” behavior of these proteins.

2. Intrinsic disorder in ALS- and FTLD-related proteins

2.1. A brief overview of the intrinsic disorder phenomenon: What are the intrinsically disordered proteins and what are they for?

The fact that numerous biologically active protein regions and even entire proteins do not possess stable tertiary and/or secondary structures is rapidly becoming the new reality in protein science. These naturally flexible proteins are known now as intrinsically disordered proteins (IDPs),²⁴ a term selected by the scientific community from several other names used in the early literature dealing with this topic (e.g., natively denatured,²⁵ natively unfolded,²⁶ intrinsically unstructured,²⁷ and natively disordered proteins,²⁸ among many other terms). In addition to proteins that are intrinsically disordered as a whole, a much larger cohort of proteins is described as hybrid proteins that contain both ordered domains and IDP regions (IDPRs).²⁹ These proteins/regions exist as highly dynamic structural ensembles possessing multiple rapidly interconverting conformations.^30,31 The IDPRs are characterized by remarkable spatiotemporal heterogeneity, and their structures can range from expanded, random coil-like ensembles, to pre-molten globule-like ensembles, which are the highly disordered entities containing some localized regions of transient secondary structure, to molten globule-like species with well-developed secondary structure and mobile side-chains.^32–34

There are multiple levels at which IDPs and IDPRs differ from structured globular proteins and domains. Such differences can be found in amino acid composition, charge, flexibility hydrophobicity, and sequence complexity. For example, IDPs/IDPRs are significantly depleted in so-called order-promoting residues (I, L, V, W, F, Y, C, and N), and contain more disorder-promoting residues, such as A, R, G, Q, S, P, E, and K.^24,35–37 Based on these and other remarkable differences between the amino acid sequences of ordered proteins/domains on the one side and IDPs/IDPRs on the other side, numerous disorder predictors were developed. A short list of commonly used predictors include members of the PONDR family,^35,38 CH-plot,³⁹ NORSp,⁴⁰ GlobPlot,^41,42 FoldIndex,⁴³ IUPred,⁴⁴ DisoPred,^45–47 DisCon,⁴⁸ MFDp2,⁴⁹ and many others. It was pointed out that, because different computational tools use different sequence attributes and different definitions of disorder, additional important information related to the peculiarities of predicted disorder can be retrieved via comparing and combining the results of several predictors on an individual protein of interest or on a protein dataset.^50-57 Furthermore, several methods were developed for the prediction of disordered protein binding regions including alpha-MoRF-Pred,⁵⁸ ANCHOR,⁵⁹ MoRFpred,⁶⁰ PepBindPred,⁶¹ MFSPSSMpred,⁶² DISOPRED3,⁶³ MoRFCHiBi,⁶⁴ fMoRFpred,⁶⁵ and DisoRDPbind,^66,67 indicating that functions of IDPRs and IDPs are predictable from protein sequences. Finally, because sites of the enzyme-catalyzed posttranslational modifications, such as phosphorylation,⁶⁸ acetylation, methylation, and ubiquitination⁶⁹ are commonly located within the IDPRs, several computational tools utilizing this information have been developed, such as DisPhos (disorder-enhanced phosphorylation predictor), which can efficiently find IDPR-located phosphorylation sites with 76% accuracy for serine, 81% for threonine, and 83% for tyrosine.⁶⁸ More recently, another tool has been developed which is a unified sequence-based predictor of 23 types of posttranslational modification (PTM) sites.⁶⁹

The high abundance of intrinsic disorder in nature was illustrated in several computational studies of different proteomes where various disorder predictors were utilized. These same studies showed that the overall level of disorder of proteins increases with organism complexity, and that over half of eukaryotic proteins are predicted to have long IDPRs.^47,50,70,71

The conformational adaptability and structural plasticity of IDPs/IDPRs, their ability to quickly react in response to the changes in their environment and to fold under a variety of conditions,^{24,27,39,72–81} combined with their unique capabilities to fold differently while interacting with different binding partners and to be able to engaged in promiscuous binding,^78,82 define a wide set of functions exerted by IDPs/IDPRs in different biological systems, and explain and determine the broad participation of IDPs/IDPRs in various biological processes,^34,83,84 where they can be involved in various signaling processes,^85,86 regulation of numerous pathways,^87–94 cell protection,⁹⁵ protein protection,^96,97 cellular homeostasis,^98,99 and controlled cell death.^100–104

These same structural features also define the ability of IDPs to modulate and control functions of their binding partners and to promote and coordinate the assembly of various complexes.¹⁰⁵ IDPs can be engaged in a wide spectrum of interactions, ranging from those leading to the formation of highly stable assemblies to weak but specific signaling interactions.³² It is also important to remember that although binding of IDPs/IDPRs to their partners is often accompanied by the disorder-to-order transitions, many IDPs and IDPRs preserve significant amounts of disorder in their bound states,³² leading to the formation of so-called disordered, dynamic, or fuzzy complexes, being bound to ordered proteins,^106–111 other disordered proteins,^112–114 membranes,^115,116 or nucleic acids. Overall, IDPs/IDPRs can form static, semi-static, and dynamic or fuzzy complexes.³³ Static and semi-static binding modes range from the interaction-induced folding of a whole molecule to binding-triggered gaining of local structure on the surface of a binding partner, and from penetrating deep inside the binding partner to wrapping around the binding partner.¹¹⁷ IDPs can participate in one-to-many and many-to-one interactions, where one IDPR binds to multiple partners potentially gaining very different structures in the bound state, or where multiple unrelated IDPs/IDPRs bind to one partner.³²

Furthermore, there are several means by which biological activities of IDPs/IDPRs are tightly controlled and regulated in a precise and timely manner. These means include alternative splicing generating multiple protein isoforms with highly diverse regulatory elements^118–120 and various PTMs, such as acetylation, glycosylation, phosphorylation, etc.^{34,69,121,122} In fact, the multifunctionality of IDPs/IDPRs relies (at least in part) on the presence of multiple relatively short functional elements that are spread within their amino acid sequences³² and can be dramatically reshuffled by alternative splicing (AS),^119,120,123 which can generate a set of protein isoforms with a highly diverse set of regulatory elements.¹¹⁸

2.2. Proteoforms: Intrinsic disorder and one sequence–many functions paradigm

The functional multifariousness of IDPs/IDPRs represents an important illustration for the paramount conclusion that the proteome size, not the genome size, defines the complexity of a biological system.¹²⁴ This is because the number of functionally divergent proteins is typically higher than the number of genes. For example, in a human cell, the number of protein-coding genes approaches 21,000,¹²⁵ while the number of functionally different proteins is substantially higher (in the range of a few hundred thousand, if not millions).^126–130 There are several mechanisms that contribute to this functional diversification of proteins and can affect a given gene product at the DNA, mRNA, or protein levels. Among such diversifiers are single or multiple point mutations, indels (i.e., insertions or deletions), and single nucleotide polymorphisms causing allelic variations contributing at the DNA level; AS, mRNA editing, the use of alternative promoters, alternative initiation of translation, and other pre-translational mechanisms affecting mRNA; and different PTMs that can change physico-chemical properties of a polypeptide chain. Therefore, instead of the “one gene–one protein–one function” view, the actual gene-protein relationship is much more complex, being better described by the “one gene–many proteins–many functions” model (see Fig. 1).

Figure 1. Classic and modern protein structure-function relationships. Schematic representation of the classic “one gene–one protein–one function” paradigm (top part, blue) and its modification by alternative splicing and PTMs when affected genes encode ordered proteins (middle part, pink) or intrinsically disordered and hybrid proteins containing ordered and intrinsically disordered domains (bottom part, red). Reproduced with permission from ref. 132.

In other words, there are various means that ensure creation of a wide spectrum of rather different protein molecules from a single gene. This multilevel diversification of gene products is at the root of the proteoform concept,¹³¹ where proteoforms “designate all of the different molecular forms in which the protein product of a single gene can be found, including changes due to genetic variations, alternatively spliced mRNA, and PTMs”.¹³¹ Subsequently, it was pointed out that, in addition to the means for increasing the chemical variability of a polypeptide chain, the protein structural and functional diversity can be further extended via protein intrinsic disorder and function itself.^132–134 Here, any given protein can be considered as a basic (or intrinsic, or conformational) proteoform because it does exist as a dynamic conformational ensemble, members of which have different structures and can potentially have different functions. There are also inducible proteoforms generated by various allelic-, mRNA-, and PTM-based diversifications of the canonical protein sequence. Finally, because structural ensembles of both basic and induced proteoforms can be affected by protein function, interaction with specific partners, or placement inside the natural cellular environment (which is extremely crowded, characterized by the presence of high concentrations of various biological macromolecules,^135-137 has limited available volume,¹³⁸ and considerably restricted amounts of free water^{135,139–143}), protein functionality can produce functioning proteoforms. Obviously, any member of the inducible (or modified) proteoform (i.e., any mutated, modified, or alternatively spliced form) or functioning proteoform is a conformational proteoform itself because it also represents a structural ensemble.^132–134 Curiously, alternative splicing generating multiple protein isoforms^118–120 and various enzymatically catalyzed PTMs, such as acetylation, glycosylation, phosphorylation, etc.,^{34,69,121,122} are commonly located within the regions with intrinsic disorder, providing additional link between conformational/basic and inducible/modified proteoforms.

All these considerations, taken together, define the ability of a protein to have a multitude of structurally and functionally different states, which are constituents of the “protein structure-function continuum”, where a given protein exists as a dynamic conformational ensemble containing multiple proteoforms (conformational/basic, inducible/modified, and functioning) characterized by diverse structural and functional features.^132–134

2.3. Protein intrinsic disorder and human diseases

Since IDPs/IDPRs are common universal regulators and controllers, they are tightly controlled and regulated¹⁴⁴ via multiple mechanisms, such as interaction with chaperones^145–149 or nanny proteins,¹⁵⁰ binding to specific partners of proteinaceous and non-proteinaceous nature,^151–157 various PTMs,^87,158–161 and regulated degradation.^162–167 As the consequences of misbehavior of an important controller (reflected in the inability of such a protein to adopt and keep a functional state) or any distortion in the tightly controlled processes could be disastrous, it is not surprising that the development of particular pathological conditions is frequently associated with the dysfunction of IDPs manifested in protein misfolding and aggregation leading to the loss of normal function and gain of toxic function.^168–170 Fig. 2 represents some of the factors related to this phenomenon and shows that the aberrant regulation at the genetic level or various posttranslational (nongenetic) mechanisms can cause pathogenic transformation in an IDP/IDPR.¹⁷¹ The complexity of this picture, where numerous factors contribute to conformational diseases, is further increased by the ability of all these factors to not only work independently, but to also take action additively or synergistically.^171,172 Therefore, many IDPs and hybrid proteins contacting ordered domains and long IDPRs are related to the development of various diseases.^169,173,174 For example, in addition to the usual suspect, TP53/p53 (tumor protein p53),¹⁷⁵ some other well-known cancer-related proteins with experimentally confirmed IDPRs include APC (adenomatous polyposis coli),^176,177 TP53BPs/ASPPs (tumor protein p53 binding protein),^178,179 AXIN1 (axin 1),^176,180,181 BH3-only proteins,¹⁸² BRCA1 (BRCA1, DNA repair associated),¹⁸³ CTAs/cancer-testis antigens,¹⁸⁴ CREBBP/CBP (CREB binding protein) and its paralog, EP300/p300 (E1A binding protein p300),¹⁵⁶ EWSR1 (EWS RNA binding protein 1),¹⁸⁵ proteins of human papillomavirus (HPV),¹⁸⁶ PTEN (phosphatase and tensin homolog),¹⁸⁷ CDKN1A/p21 (cyclin-dependent kinase inhibitor 1A) and CDKN1B/p27Kip1 (cyclin-dependent kinase inhibitor 1B),⁹² SIRTs/NAD-dependent protein deacetylase sirtuins,^188,189 and many others. Many of the neurodegeneration-related proteins are also IDPs. Illustrative examples of highly disordered proteins engaged in various neurodegenerative diseases include a protein-chameleon, SNCA/α-synuclein,^190–195 associated with PD, dementia with Lewy bodies, AD, Down syndrome, and several other synucleinopathies, ataxin in spinocerebellar ataxia, prions in Creutzfeldt-Jakob disease and transmissible spongiform encephalopathies, amyloid β in AD, and MAPT in AD and many other tauopathies.^168,196 Among other human diseases for which involvement of IDPs was demonstrated are AIDS (HIV Rev protein),¹⁹⁷ cardiovascular disease (e.g., F2 [coagulation factor II, thrombin]),¹⁹⁸ cystic fibrosis (CFTR [cystic fibrosis transmembrane conductance regulator]),¹⁹⁹ and type II diabetes (IAPP/amylin [islet amyloid polypeptide]).¹⁷³

Figure 2. Factors causing pathogenic transformation in an IDP/IDPR. Schematic representation of different ways by which aberrant regulation at the genetic level or various posttranslational (non-genetic) mechanisms can cause pathogenic transformation in an IDP/IDPR. Reproduced with permission from ref. 171.

2.4. Intrinsic disorder in proteins related to ALS and FTLD

Fig. 3 demonstrates that all ALS- and FTLD-related proteins considered in this review contain significant levels of intrinsic disorder. The analogous results for the dipeptide repeat proteins poly(GA), poly(GR), poly(GP), poly(PA), and poly(PR) synthesized as a result of the repeat-associated non-ATG (RAN) translation of the sense and antisense repeat RNAs derived from the intronic region of the C9orf72 gene are not shown in this figure. This is because of the well-known fact that the repeat-containing proteins are often intrinsically disordered, with the more perfect repeats being less structured.²⁰⁰ Therefore, it is expected that these proteins should be mostly disordered, and in agreement with this expectation, all these proteins were predicted to be highly disordered.²⁰¹

Figure 3. Intrinsic disorder in proteins presented in this study. The intrinsic disorder predisposition of these proteins was evaluated by PONDR® VSL2. (A) SOD1; (B) TARDBP; (C) FUS; (D) C9orf72; (E) PFN1; (F) GLE1; (G) PURA; (H) FLCN; (I) RBM45; (J) SS18L1/CREST; (K) HNRNPA1; (L) HNRNPA2B1; (M) ATXN2; (N) TIA1; and (O) MAPT. In the corresponding plots, regions with disorder scores above the 0.5 threshold (shown as a thin black line) are predicted to be intrinsically disordered.

Results shown in Fig. 3 are in agreement with the results reported in previous studies. In fact, prevalence of intrinsic disorder in proteins associated with various neurodegenerative diseases was the subject of several dedicated studies.^{168,170,171,196,201–203} Also, some proteins associated with biological liquid-liquid phase transitions (LLPTs) and formation of various proteinaceous membrane-less organelles (PMLOs), including stress granules (SGs), are disordered.^204–206 As a result, in those studies, the intrinsic disorder status and biological roles of IDPRs were already systemized for several proteins considered in this review (SOD1, TARDBP, FUS, C9orf72 and DPRs, PFN1, TIA1, MAPT, HNRNPA1, HNRNPA2B1, and ATXN2).

Because interested readers can find details of those analyses in the corresponding publications,^{168,196,205,206} there is no need to reproduce them here. Therefore, this section represents some data on the intrinsic disorder status of the remaining ALS-FTLD-related proteins that also play important roles in SG formation (GLE1, PURA, FLCN, RBM45, and SS18L1). It is important to keep in mind that the in-depth analysis of the intrinsic disorder predispositions of all these proteins and the detailed consideration of the potential roles of their IDPRs is outside the scope of this article.

Nucleoporin GLE1 (GLE1, RNA export mediator; UniProt ID: Q53GS7) is the nucleoporin responsible for the export of the poly(A)-tail containing mRNAs from the nucleus into the cytoplasm^207–210 and is also known as a multifunctional regulator of gene expression.²¹¹ The N-terminal region (residues 1–30) of human GLE1 is required for interaction with the nuclear pore complex via the nucleoporin NUP155 (nucloporin 155),²¹² whereas the coiled-coil domain containing 2 coiled-coil regions is required for GLE1 self-oligomerization and localization to the nuclear pore complex.²¹⁰ Fig. 3F shows that the N-terminal half of this protein containing these functionally important regions is predicted to be mostly disordered.

A high level of disorder in this protein is also confirmed by the MobiDB database (http://mobidb.bio.unipd.it/)^213,214 which generates consensus disorder scores for a query protein by aggregating the output from 10 predictors. Such consensus MobiDB disorder content of human GLE1 is 25.50%. The conclusion on the presence of high levels of disorder in human GLE1 protein is further supported by Fig. 4A, which shows the results of the analysis of this protein using the D²P² database (http://d2p2.pro/)²¹⁵ that provides disorder evaluations by several computational tools and also shows some important disorder-related functional information, such as location of various curated PTMs and ANCHOR-predicted disorder-based protein-protein interaction sites,^59,216 known as molecular recognition features, MoRFs (see 58, 217–219). Fig. 4A shows that human GLE1 is predicted to have 11 MoRFs, all located within the N-terminal half of the protein, and a multitude of PTMs, such as phosphorylation, ubiquitination, and acetylation. All this clearly indicates that abundant disorder serves several important functions for this protein.

Figure 4. Multifactorial analysis of intrinsic disorder in GLE1 (UniProt ID: Q53GS7) (A) and PURA (UniProt ID: Q00577) (B). Intrinsic disorder propensity and some important disorder-related functional information generated for the corresponding proteins by the D²P² database are shown. Here, the green-and-white bar in the middle of the plot shows the predicted disorder agreement among 9 predictors, with green parts corresponding to disordered regions by consensus. The yellow bar shows the location of the predicted disorder-based binding sites (molecular recognition features, MoRFs), whereas colored circles at the bottom of the plot show the location of various PTMs.

Transcriptional Activator PURA/Pur-α (UniProt ID: Q00577; MobiDB consensus disorder content is 41.93%) is a highly conserved multifunctional single-stranded DNA- and RNA-binding protein²²⁰ that serves as a transcriptional activator. Fig. 3G and 4B show that PURA is predicted to have intrinsically disordered N- and C-terminal tails and also has a disordered region in the middle of the sequence. It is predicted to have 6 MoRFs and several phosphorylation, ubiquitination, and acetylation sites spread throughout the sequence of this protein (Fig. 4B).

FLCN (folliculin; UniProt ID: Q8NFG4; MobiDB consensus disorder content is 23.32%) is a tumor suppressor, that is predicted to contain 2 long IDPs and a disordered C-tail (Fig. 3H and 5A). The centrally located disordered region contains the coiled-coil domain (residues 287–310). FLCN is predicted to have 2 MoRFs located within the N-terminal IDPR, as well as several sites of phosphorylation and ubiquitination (Fig. 5A).

Figure 5. Multifactorial analysis of intrinsic disorder in FLCN (UniProt ID: Q8NFG4) (A), SS18L1/CREST (UniProt ID: O75177) (B), and RBM45 (UniProt ID: Q8IUH3) (C). Plots (A) and (B) were generated using the D²P² database. Here, keys are similar to those described in legends to Fig. 4. Because the D²P² database does not have related information for human RBM45, plot (C) represents disorder profiles generated by PONDR® VLXT (black line), PONDR® FIT (red line), PONDR® VL3 (green line), PONDR® VSL2 (yellow line), IUPred_short (blue line) and IUPred_long (pink line). The cyan dashed line shows the mean disorder propensity calculated by averaging the disorder profiles of the individual predictors. The light pink shadow around the PONDR® FIT shows the error distribution. In these analyses, the predicted intrinsic disorder scores above 0.5 are considered to correspond to the disordered residues/regions, whereas regions with disorder scores between 0.2 and 0.5 are considered flexible.

SS18L1/CREST (UniProt ID: O75177; MobiDB consensus disorder content is 86.16%) is a 396-residue-long, highly disordered transcription activator that has a disordered N-tail, 1 ordered domain that contains 2 MoRFs, and a long, highly disordered region covering the central and C-terminal parts of this protein (Fig. 3J and 5B). This disordered region serves as an excessive binding platform, because 11 of 13 MoRFs predicted in SS18L1 are located there. Overall, Fig. 5B shows that 55.6% of human SS18L1 residues are expected to be engaged in disorder-based protein-protein interactions.

RBM45 (RNA binding motif protein 45; UniProt ID: Q8IUH3; MobiDB consensus disorder content is 18.07%) is an RNA-binding protein with binding specificity for poly(C). Although there is no information about this protein in D²P², a multiparametric analysis of this protein by PONDR® VLXT, PONDR® FIT, PONDR® VL3, PONDR® VSL2, IUPred_short, and IUPred_long shown in Fig. 5C suggests that this protein contains several disordered and flexible regions; that is, regions with disorder scores exceeding 0.5 and ranging between 0.2 and 0.5, respectively. Also, there are 3 MoRFs in human RBM45.

3. Stress granules, other proteinaceous membrane-less organelles, and biological liquid-liquid phase transitions

Colocalization of biological macromolecules (proteins and nucleic acids) at high concentrations within a small cellular micro-domain leads to the formation of various proteinaceous membrane-less organelles (PMLOs), including stress granules (SGs). These PMLOs are highly dynamic assemblages, which, due to the presence of various RNAs and proteins, are generally known as ribonucleoprotein (RNP) granules, RNP bodies, or RNP droplets.²²¹ Although such PMLOs can typically be found in the cytoplasm and the nucleoplasm of eukaryotic cells, they are also present in mitochondria and, in plants, in chloroplasts (Fig. 6). As a matter of fact, the known set of PMLOs exceeds 25 species. The incomplete list of cytoplasmic PMLOs includes centrosomes,²²² germline P-granules (germ cell granules or nuage),^223,224 neuronal RNA granules,²²⁵ processing bodies or P-bodies,²²⁶ and stress granules.²²⁷ Most commonly known nuclear PMLOs are Cajal bodies,²²⁸ chromatin,²²⁹ cleavage bodies,²³⁰ histone locus bodies,²³¹ nuclear gems (Gemini of Cajal bodies),^232,233 nuclear pores,²³⁴ nuclear speckles or interchromatin granule clusters,²³⁵ nuclear stress bodies,^236,237 nucleoli,²³⁸ Oct1/PTF/ transcription (OPT) domains,²³⁹ paraspeckles,²⁴⁰ perinucleolar compartment,²⁴¹ PcG bodies (polycomb bodies, subnuclear organelles containing polycomb group proteins),²⁴² promyelocytic leukemia nuclear bodies or nuclear dots,²⁴³ and the Sam68 nuclear body.²⁴¹ Mitochondria and chloroplasts have RNA granules²⁴⁴ and chloroplast SGs, respectively.²⁴⁵ This realm of cytoplasmic, nuclear, mitochondrial and chloroplast PMLOs is depicted in Fig. 6. One should keep in mind that although in this review SGs are the most frequently discussed PMLOs, pathological cellular transformations may also engage other organelles (e.g., P-bodies, neuronal RNA granules, CREBBP-containing nuclear bodies, and paraspeckles among others, see below).

Figure 6. Diversity of PMLOs found in eukaryotic cells. Schematic representation of the multitude of cytoplasmic, nuclear, mitochondrial and chloroplast PMLOs.

The important feature of PMLOs that clearly differentiate them from other cellular organelles is the fact that, contrary to the “traditional” membrane-encapsulated organelles, the structural integrity of PMLOs is not supported by encapsulation in the membrane. Instead, their biogenesis is entirely controlled and mediated via a multitude of protein-protein, protein-RNA, and/or protein-DNA interactions.²⁴⁶ Furthermore, this lack of membrane encapsulation defines the liquid-like behavior, such as dripping, relaxation to spherical structures upon fusion, and wetting, commonly described for various PMLOs.^{223,227,247,248} Also, the lack of encapsulation defines the highly dynamic nature of PMLOs and ensures the direct contact of their components with the surrounding cytoplasm or nucleoplasm.^249,250 Based on all these considerations, as well as on the fact that PMLOs are just slightly denser than the rest of the cytoplasm or nucleoplasm^251,252 and possess high level of internal dynamics, these organelles are defined as liquid-droplet phases of the nucleoplasm/cytoplasm.^{223,227,247,248,253,254} Curiously, although PMLOs are considered as a different liquid state of the cytoplasm or nucleoplasm, their major biophysical properties are rather similar to those of the rest of the intracellular fluid.²²¹ PMLOs have unique morphologies, specific distribution patterns, and, at the molecular level, their structures are determined by a specific set of resident proteins.

It is assumed that the formation of SGs and other PMLOs represents a result of the liquid-liquid phase transition (or fluid-fluid demixing).^{221,223,247,255–258} Biophysically, besides the obvious requirement of the presence of high enough concentrations of the related constituents compartmentalized within a limited enough volume, the efficiency of such LLPTs depends on the flexibility and multivalency of the participating components and on the ability of these constituents to be engaged in multiple highly controllable low-affinity interactions.^{253,259–270} It was emphasized that highly dynamic structures, a low complexity of amino acid sequences that can be considered as a string of repetitive amino acids or groups of amino acids, and the ability to be engaged in multiple low-affinity interactions with other proteins and nucleic acids of IDPs and IDPRs, make them ideal drivers of such LLPTs, and, in agreement with this hypothesis, many PMLOs abundantly contain IDPs/IDPRs.^{205,206,271–273}

Since SGs are the PMLOs which are most frequently considered in association with ALS, these RNP droplets are discussed below in greater detail. Readers interested in other PMLOs are can peruse recent reviews.^{206,221,273,274} As with all other PMLOs, SGs are complex assemblages that include proteinaceous and RNA constituents and are typically formed as a specific cellular response to stressful conditions or to a variety of environmental and cellular cues. SGs include components of the translation-competent 48S complex (eukaryotic initiation factors EIF3 complex, EIF4G1/EIF4F, EIF4B, the small ribosomal subunits, PABPC1/PABP1 (poly[A] binding protein cytoplasmic 1), and stalled mRNA transcripts),²⁷⁵ many other RNA binding proteins (RBPs; primary SG formers) that are involved and regulate mRNA stabilization, processing, translation, and transport,²⁷⁶ and, being overproduced, are able to induce SG formation in the absence of stress (such as PABPC1, TIA1, TIA1L/TIAR [TIA1 cytotoxic granule associated RNA binding protein like 1], G3BP1 [G3BP stress granule assembly factor 1], CPEB [cytoplasmic polyadenylation element binding protein], FXR1 FMR1 autosomal homolog 1], FMR1/FMRP [fragile X mental retardation 1], and CASC3/MLN51 [cancer susceptibility 3]),^276–283 as well as a variety of other proteins that are recruited to the SGs via protein-protein interactions with primary SG formers.²⁷⁵ Furthermore, these cytoplasmic foci contain various mRNA stalled at the pre-initiation stage and can therefore be considered as cytoplasmic messenger ribonucleoprotein particles that possess defined cytoprotective functions.²⁸³ Among the biological functions ascribed to SGs are modification of the local patterns of translation within the cell, sequestration of various signaling molecules that regulate cell viability, and sequestration and silencing of mRNAs encoding house-keeping proteins.^275,284 Furthermore, under stress conditions SGs participate in shifting RNA translation towards cytoprotective proteins, such as chaperones and heat shock proteins.²⁸⁵ The list of regulatory proteins sequestered by SGs includes several regulators of apoptosis, such as RACK1 (receptor for activated protein C kinase 1), TRAF2 (TNF receptor-associated factor 2),^281,286,287 and MTORC1 (mechanistic target of rapamycin complex 1) that serves as a regulator of cell growth and metabolism.^288,289

Of special importance is the fact that, unlike amyloid or amyloid-like fibrillation, the process of SG formation is completely reversible and is tightly controlled.²⁹⁰ The formation of SGs, being an illustrative example of biological LLPT, is a very complex process. In brief, SG nucleation is initiated by several RNA-binding proteins, with the most commonly examined being TIA1, TIAL1, ZFP36/TTP (ZFP36 ring finger protein), and G3BP,²⁷⁵ whereas SG maturation involves incorporation of many pro-apoptotic proteins (e.g., TRAF2, ROCK1 (Rho associated coiled-coil containing protein kinase 1), and RACK1 and other regulators of apoptosis (such as RPS6KA3/RSK2 [ribosomal protein S6 kinase A3] and FASTK [Fas activated serine/threonine kinase]), leading to the inhibition of the apoptotic response.²⁹⁰ Among apoptosis-unrelated signaling and regulating proteins incorporated into the SGs are MAPK8/JNK1 (mitogen-activated protein kinase 8), MAP2K7/MKK7 (mitogen-activated protein kinase kinase 7), RHOA (ras homolog family member A), AKAP9/AKAP350 (A-kinase anchor protein 9), WDR62 (WD repeat domain 62), and HDAC6 (histone deacetylase 6).²⁹⁰

An interesting recent development is the recognition of the importance of SGs for the pathology of several neurodegenerative diseases,^{283,290–292} including ALS and FTLD.^283,293-299 Relation of SGs to neurodegeneration comes from the following observations:

(i)	SG marker proteins, such as PABPC1 and G3BP, are found in cytosolic inclusion bodies of patients with AD, ALS, and FTLD, indicating that distorted SG dynamics might contribute to pathogenesis of these and other neurodegenerative diseases300;
(ii)	in many neurodegenerative diseases, there is an obvious colocalization of SGs with insoluble protein aggregates290;
(iii)	SGs frequently contain RNA-binding proteins related to the pathogenesis of various neurodegenerative diseases, such as TARDBP and FUS, related to the pathology of ALS and FTLD, SMN (survival motor neuron protein) related to the spinal muscular atrophy pathology, spinocerebellar ataxia 2-related ATXN2, OPTN related to primary open angle glaucoma and amyotrophic lateral sclerosis 12, and ANG (angiogenin) that is involved in amyotrophic lateral sclerosis 9 pathology;
(iv)	Mutations in the synaptic functional regulator FMR1, which is a protein that plays a central role in neuronal development and synaptic plasticity through the regulation of alternative mRNA splicing, mRNA stability, mRNA dendritic transport, and postsynaptic local protein synthesis of a subset of mRNAs301–305 and is associated with fragile X syndrome) or HNRNPA1 (an RNA-binding protein related to ALS-FTLD pathogenesis) cause severe impairment of SG dynamics300;
(v)	also, MAPT,290 PRNP/ PrP (prion protein),306 HTT (huntingtin),307 and some other Q/N-rich proteins308 are associated with SGs and modulate SG formation.

The roles of various proteins related to ALS and/or FTLD and involved in SG formation are considered below.

4. Normal and pathological functions of the ALS- and FTLD-related proteins and their roles in biogenesis of stress granules and other PMLOs

4.1. SOD1/Cu-Zn superoxide dismutase 1

4.1.1. Structural properties of SOD1

The SOD1 gene is located at position 22.11 within the long (q) arm of chromosome 21 (genetic location 21q22.11), and its coding region consists of 5 exons interrupted by 4 introns.³⁰⁹ Struturally, human SOD1 (UniProt ID: P00441) is a homodimer whose protomers are held together by hydrophobic interactions.^310,311 The structure of this protein is characterized by the highly conserved topology found in all the intracellular eukaryotic SODs, where each protomer has 153 amino acids arranged in a Greek key β-barrel structure composed of 8 antiparallel β-strands.³¹² Homodimeric SOD1 is characterized by a very high conformational stability, with the melting temperature (T_m) exceeding 90°C.³¹³ This high stability is determined by 3 factors associated with the posttranslational maturation of this protein, such as the coordination of 1 Zn²⁺ and 1 Cu²⁺ ion by each protomer of the homodimer, the formation of the intrachain disulfide bond Cys58-Cys147, and dimerization. In fact, disulfide-reduced apo-SOD1 (i.e, a protein with metal ions removed) has a T_m of ∼42°C,^314,315 which makes it highly succeptible for spontaneous partial unfolding followed by misfolding and aggregation under physiological conditions. Amyloid-like fibrils, which are a pathological hallmark in SOD1-related patients with familial ALS,³¹⁶ are formed only by the disulfide-reduced apo-SOD1.³¹⁷ This clearly indicates that intracellular deregulation of metal binding and/or disulfide formation might be among the key factors triggering the pathological misfolding of SOD1.³¹⁸ In agreement with this hypothesis, many ALS-related mutations of SOD1 affect the post-translational control of SOD1 maturation,^319,320 leading to an increase in the intracellular populations of the apo-,³²¹ and/or disulfide-reduced forms of this protein.³²²

Reduction of the disulfide bond causes increased disorder in the apo-SOD1 generating a conformation significantly different from the structure of the wild-type SOD1 (WT-SOD1) protomer.³¹⁸ These observations are in strong agreement with a systematic computational analysis of the folding and structural dynamics of monomeric and dimeric forms of SOD1 in their holo- and apo-forms, both in the disulfide-intact and disulfide-reduced states by the all-atom discrete molecular dynamics (DMD) simulations, which revealed that although the disulfide bond formation has less pronounced stabilizing effects than the metal binding, the disulfide bond nevertheless has a significant contribution to protein dynamics.³²³ It was also shown that the reduction of this disulfide bond is necessary for initiation of fibril formation.³²⁴

4.1.2. Normal and pathological functions of SOD1

SOD1 is an antioxidant metalloenzyme that protects cells from oxidative stress (OS) by catalytic neutralization of the toxicity of superoxide radicals via a reaction that converts superoxide (O₂⁻) to O₂ ions and H₂O₂.^325–327

SOD1 is one of the genes that are most commonly involved in the ALS pathogenesis. SOD1 mutations have been associated with a great number of familial ALS (fALS) cases, and more than 170 mutations of SOD1 were found to cause ∼20% of the cases of fALS and ∼5% of the cases of sporadic ALS (sALS).^328–331 However, it is worth keeping in mind that the numbers reflecting the incidence of SOD1 mutations in ALS are changing over time, and it is currently accepted that ∼10% and ∼2% of fALS and sALS cases, respectively, are associated with mutations in the SOD1 gene.³³² Although many of the SOD1 mutations lead to a loss of function, at least some of the ALS-related mutations are thought to cause the enzyme to gain damaging properties,³³³ such as ability to damage mitochondria, proteasomes, and protein chaperones.³³⁴ Conformationally, many SOD1 mutants have similar properties, possessing solvent-exposed hydrophobic regions that increase the propensity of this protein to aggregate.³³⁵

Interestingly, the “Dr. Jekyll–Mr. Hyde” behavior of SOD1 is not only triggered by mutations, because misfolded WT-SOD1 has also been implicated in sporadic ALS.³³⁵ Curiously, although the discovery of the dominant mutations in the SOD1 gene made almost 25 years ago³³⁶ marked the beginning of the molecular era of the analysis of ALS, no consensus on the main toxicity of mutant SOD1 has emerged as of yet.⁴ Currently, there is a wide spectrum of potential toxic mechanisms proposed to explain how mutations in SOD1 might mediate the degeneration and death of motor neurons,⁴ and such a lack of consensus serves as an indication of the overall complexity of this problem.

SOD1 protein maturation involves the appropriate coordination of Cu²⁺ (by His47, His49, His64, and His121) and Zn²⁺ ions (by His64, His72, His81, and Asp84) and the formation of a kinetically stable disulfide bond between Cys57 and Cys146.³³⁷ A very unusual feature of SOD1 is its striking capability to maintain an intrasubunit disulfide bond even under the reducing conditions of the cytosol. In fact, in the vast majority of proteins the sulfhydryl groups of cysteine residues have a pKa >8.0. As a result, in the absence of oxidative stress and in the reducing environment of the cytoplasm at physiological pH, these sulfhydryl groups remain protonated, and cytoplasmic proteins, as a rule, do not contain disulfide bonds.^338,339 The CCS (copper chaperone for superoxide dismutase) is required for the formation of this disulfide bond in SOD1, because the spontaneous disulfide formation in SOD1 by O₂ is a slow process, which is greatly accelerated by the Cu²⁺-bound form of CCS.³³⁷ It was pointed out that both metal occupancy and disulfide status are crucial for the formation of a stable functional state of SOD1,³⁴⁰ because without these factors the protein cannot gain its stable native conformation and cannot dimerize. Obviously, the formation of these functional homodimers with stable native conformation and intact disulfide bonds can also be altered by the pathological SOD1 mutations,³³⁵ or by oxidation, demetallation, and other altered PTMs.^341–344 It was also pointed out that some aberrant PTMs induced by oxidative damage (e.g., appearance of oxidized carbonyl groups in the protein leading to its over-oxidation) force the WT-SOD1 to acquire binding and toxic properties of mutant forms of this protein.^341–344 This gives some support to the idea that SOD1 “is subject to misfolding following the loss of normal PTM and/or the induction of aberrant modification”.³³⁵

There are several ways by which mutated and misfolded SOD1 and misfolded WT-SOD1 might act as a neurotoxin. First, pathogenic SOD1 may exert its neurotoxic effects via glutamate excitotoxicity, which causes an abnormally high release of calcium ions in the post-synaptic cleft that triggers subsequent neurotoxic events including mitochondrial dysfunction. Pathogenic SOD1 is theorized to cause this excitotoxicity due to the loss of SLC1A2/EAAT2/GLT1 (reducing reuptake of the neurotransmitter) and excessive glutamate efflux. Second, astrocytes expressing a SOD1 mutant lack the ability to effectively regulate their α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, whose subunit composition is in constant flux, which can cause different permeability. These mutant astrocytes lack the GRIA2/GluR2 subunit of these receptors that increases their permeability to Ca²⁺ ions. This sensitization, along with increased calcium production/decreased reuptake, act in tandem to cause excitotoxicity.³³⁵ Third, OS by reactive oxygen species (ROS) is increased in cells with mutant SOD1. However, this stress is not due to loss of function by SOD1 but rather due to upregulation of the ROS. These ROS are formed via many different mechanisms including NAPDH oxidase and mitochondrial respiration.³³⁵ Fourth, mutant SOD1 puts stress on endoplasmic reticulum (ER) proteins such as DERL1 (derlin 1), and the resulting ER stress has been linked to both familial and sporadic ALS. In particular, DERL1 is crucial for the elimination of misfolded proteins from the endoplasmic reticulum. Interactions between mutant SOD1 and DERL1 cause the inhibition of this process, resulting in ER stress and ultimately apoptosis via ERN1/IRE1 (endoplasmic reticulum to nucleus signaling 1).³³⁵

The mitochondria is another target for mutant SOD1, which has been shown to accumulate exclusively on the cytosolic side of mitochondria in the spinal cord even before ALS symptoms are observed.³³⁵ Binding of mutant SOD1 to several target proteins, including the integral membrane proteins BCL2 (BCL2, apoptosis regulator) and VDAC1 (voltage dependent anion channel 1), causes irregularities in mitochondrial homeostasis. Over-oxidized WT-SOD1 also forms a toxic complex with BCL2 in the lymphoblasts in some patients with sporadic ALS.³³⁵

Vesicle transport is extremely important for long motor neurons to function normally, and distortion in this transport can potentially contribute to selective motor neuron death. Therefore, another pathological mechanism of mutant SOD1 has been linked to abnormal axonal synaptic vesicle transport in motor neurons during the early stages of ALS.³⁴⁵ In fact, at an early asymptomatic ALS stage preceding loss of spinal motor neurons and peripheral axons, fast axonal transport is impaired in transgenic mice carrying G93A SOD1^G93A or SOD1^G86R mutations^345,346 mutations equivalent to those found in subsets of human fALS. In line with the idea that the oxidized WT- SOD1 shares an aberrant conformation with the ALS-related mutants of this protein, a detailed analysis revealed that recombinant, oxidized WT- SOD1 and the wild-type protein immunopurified from sALS tissues were equally able to inhibit the kinesin-based fast axonal transport, acting similarly to the fALS-linked SOD1 mutants,³⁴³ suggesting an overlapping pathogenic mechanism.³³⁵

Finally, recent evidence suggests that SOD1 aggregation also exerts toxicity via prion-like propagation. This idea is supported by the fact that aggregating SOD1 is engaged in both cell-to-cell transmission and seeded aggregation. Both wild-type and mutant SOD1 form aggregates that are released from the injured cells as part of an exosome.³⁴⁷ Also, mutant SOD1 can be released via a main secretory pathway mediated by CHGA (chromogranin A) and CHGB that serve as components of neurosecretory vessels.³⁴⁸ Likely, this pathway can also be related to the release of the oxidized WT-SOD1 which was shown to interact with CHGB.^341,349 In addition, both wild-type and mutant SOD1 aggregates seed aggregation of WT- SOD1 in vivo, supporting the theory of prion-like propagation of SOD1-related pathology.³³⁵

4.1.3. SOD1 and biogenesis of stress granules and other PMLOs

It has been established that both the WT-SOD1 and mutant SOD1 forms (mtSOD1 containing A4T, G41S, G85A, or G93A substitutions), as well as TARDBP, and YWHAB/14-3-3 protein α/β (tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein beta) can be involved in interaction with NEFL (neurofilament light) mRNA and differentially modulate the stability of this mRNA, which is destabilized by mutant SOD1,³⁵⁰ but stabilized by TARDBP³⁵¹ and YWHAB.³⁵² Furthermore, interaction of WT-SOD1 or mtSOD1 with YWHAB results in efficient disruption of the interaction of YWHAB with NEFL mRNA,³⁵³ suggesting that SOD1 might have dual effects on the mRNA stability, via direct binding and by eliminating stabilizing YWHAB-NEFL mRNA interaction. Importantly, using RNA-IP-PCR in ALS spinal motor neurons, it has been also demonstrated that NEFL mRNA is sequestered in SGs and P-bodies, suggesting that NEFL mRNA processing is dramatically altered in ALS.³⁵³ Although the colocalization with the NEFL mRNA-containing stress granules and P-bodies was unequivocally established for TARDBP, the ability of YWHAB, WT-SOD1 and mtSOD1 to interact with NEFL mRNA suggests that all these proteins can be related to alteration of the processing of this mRNA leading to its compartmentalization within both SGs and P-bodies.³⁵³ It is important to note that the aforementioned NEFL, is not the only mRNA sequestered by SOD1 and TARDBP aggregates and SGs. As a matter of fact, it is now well-established that TARDBP and SGs in general bind up thousands of various RNAs, making NEFL mRNA, at best, a singular model RNA for the types of cellular mechanisms these promiscuous RNA-binding proteins and RNA granules might be involved with.^354–358

Direct involvement of SOD1 in the regulation of SG dynamics was recently established by showing the ability of mtSOD1 (L144F) to interact, in an RNA-dependent manner, with G3BP1 that plays a critical role in SG dynamics.³⁵⁹ In fact, the aforementioned mtSOD1-G3BP1 interaction is rather specific (involving F380 and F382 residues of the RNA-binding RNA-recognition motif [RRM] domain of G3BP1), because almost no interaction is seen between the mtSOD1 and other RNA-binding proteins implicated in ALS.³⁵⁹ It was also shown that the SG formation induced in N2A cells by hyperosmolar shock and arsenite treatment is delayed by the mtSOD1, indicating that one of the culprits of pathogenic SOD1 mutations in ALS could be related to the alterations of the RNA metabolism and the SG dynamics.³⁵⁹

4.2. TARDBP/TDP-43 (TAR DNA binding protein)

4.2.1. Structural properties of TARDBP

TARDBP/TDP-43 is an RNA- and DNA-binding protein that is 414 amino acids long with a molecular mass of 44 kDa. TARDBP belongs to the family of the heterogeneous ribonucleoproteins (hnRNPs), members of which are RNA-binding proteins that bind to pre-mRNA and are involved in RNA splicing. The TARDBP gene is located on chromosome 1p36.2 and contains 6 transcribed exons.³⁶⁰ Human TARDBP protein (UniProt ID: Q13148) has an N-terminal domain (NTD, residues 1–103), 2 RNA recognition motifs, RRM1 (residues 104–200) and RRM2 (residues 191–262), a nuclear export signal (NES), a nuclear localization signal (NLS), and a glycine-rich C-terminal domain (residues 274–413), which is known as a prion-like domain and is predicted to be highly disordered (Fig. 3B). This RRM1-RRM2-C-tail domain organization is similar to that of hnRNP family proteins such as HNRNPA1 and HNRNPA2B1.³⁶¹ Although TARDBP is mainly a nuclear protein, it moves between the inside and outside of the nucleus, with this shuttling regulated by its NES and NLS motifs. The C-terminal domain is used by TARDBP for binding to single-stranded DNA, RNA, and proteins, and most of the ALS- and FTLD-related mutations found in TARDBP are located within this prion-like domain encoded by exon 6 of the TARDBP gene.³⁶² This prion-like domain has also been associated with the misfolding and self-aggregation of this protein.³⁶³ According to comprehensive computational analysis, the C-terminal region of human TARDBP contains 4 disorder-based protein-protein interaction sites²⁰¹ and, therefore, clearly serves as a major “docking port” for its binding partners.

Human TARDBP is present in 2 proteoforms generated by alternative splicing, with isoform #2 being different from the canonical form of the protein by having a changed N-terminal tail, where the ¹MSEYIRVTEDENDEPIEI¹⁸ sequence is substituted to MPQMLAGEIWCMLSTIQK, and where the 19–134 region is missing.

4.2.2. Normal and pathological roles of TARDBP

TARDBP has many functions, most of which are related to RNA handling due to its ability to bind RNA through its RNA recognition motifs.³⁶⁴ Some of these functions include regulation of transcription, pre-mRNA splicing, as well as processing, stability, and transport of mRNA.^365,366 In addition, TARDBP can participate in protein-protein interactions with itself and several other proteins. Examples of its binding partners include HNRNPA1, HNRNPA2B1 and FUS.³⁶⁷ These proteins are required for TARDBP to perform some of its biological functions.³⁶⁷ An important feature of TARDBP is its intrinsic ability to form aggregates, and this aggregation process is mediated by the C-terminal fragments of TARDBP.³⁶⁸ TARDBP is a component of ubiquitin-positive MAPT-negative insoluble protein aggregates in both neurons and glial cells in sporadic ALS.³⁶⁹ In addition, it was shown that the cleavage of TARDBP by CAPN (calpain) makes this protein more prone for aggregation than the full-length protein, and that both phosphorylated and nonphosphorylated TARDBP fragments resulting from this proteolytic cleavage remain intracellularly long enough to self-aggregate.³⁷⁰ Furthermore, because the phosphorylated TARDBP becomes resistant to the CAPN cleavage it is likely that phosphorylation (and some other PTMs) may play a regulatory role in the TARDBP pathology of ALS and FTLD.³⁷⁰ The pathological phosphorylation of TARDBP is controlled by the phosphatase PPP3/calcineurin that can bind to and catalyze the removal of pathological C-terminal phosphorylations of TARDBP.³⁷¹

In line with the hypothesis on the important roles of the TARDBP proteolytic cleavage for pathogenesis of neurodegeneration is the fact that various TARDBP fragments generated by the apoptotic CASP3 (caspase 3) and CASP7, as well as by CASP6 and CASP8 after activation by CASP3, accumulate in motor neurons and serve as a post mortem hallmark of different neurodegenerative diseases.³⁷² In fact, activation of the CASPs under some stress conditions might result in the fast appearance of the C-terminal fragments (CTFs) of TARDBP with the molecular mass of 35 kDa that accumulate in the cytoplasm, and the formation of this truncated form of TARDBP can be efficiently inhibited by the A90V mutation found in the FTLD/ALS patient with a family history of dementia,³⁷³ whereas another mutant, D169G, is cleaved by CASP3 more efficiently than the wild-type TARDBP.³⁷⁴ Also, the 35-kDa CTFs of TARDBP form ubiquitin-negative cytoplasmic inclusion bodies, which are mildly toxic, whereas the 25-kDa TARDBP CTFs are able to efficiently form toxic ubiquitin-positive inclusion bodies and sequester cellular RNA.³⁷⁵ These observations are in line with the hypothesis that the CASP-mediated cleavage pattern of TARDBP determines the rate of clearance and cytotoxicity of this protein.³⁷⁶ They also indicate that TARDBP possesses dual vulnerability to proteolytic attack by both CAPNs and CASPs.³⁷⁷

4.2.3. TARDBP, SGs, and other PMLOs

Undoubtedly, among the ALS-FTLD-related RNA-binding proteins, the involvement of TARDBP in SG formation represents one of the best studied cases of the regulated protein aggregation phenomenon. One of the first reports indicating the existence of a potential link between TARDBP, SGs (which are liquid-like ribonucleoprotein complexes where protein synthesis is temporarily arrested), and ALS was published in 2009, when, based on axotomized motor neuron analysis, it was hypothesized that the upregulated TARDBP expression combined with the prominent localization of this protein to the cytosolic SGs might represent a physiological response to the motor neuron injury caused by degenerative processes, such as ALS.³⁷⁸ In the same year, careful analysis of the response of a motoneuronal cell line to oxidative stress and environmental insults revealed that TARDBP is capable of efficient SG assembly, with the RRM1 (residues 104–200) and the 216–315 region of the TARDBP C-tail being crucial for the recruitment of this protein into SGs.³⁷⁹ Also, analysis of the ALS spinal motor neurons revealed that TARDBP plays a fundamental role in compartmentalization of NEFL mRNA (processing of which is fundamentally altered in ALS) within SGs and P-bodies.³⁵⁹ The SG recruitment of TARDBP was further evidenced by the large scale analysis of the TARDBP interacting proteins, many of which are known SG components.³⁶⁴ Furthermore, this analysis revealed that the TARDBP interacting proteins can be functionally clustered into the cytoplasmic/translation- and nuclear/splicing-related PPI networks, suggesting that TARDBP can function in both the nucleus and the cytoplasm, playing multiple roles in RNA metabolism.³⁶⁴

Many subsequent studies provided direct support to the idea that TARDBP, SGs, and ALS-FTLD are intimately linked. For example, an analysis of the cultured cells under the stress conditions revealed the existence of the TARDBP-SG association promoted by both the direct binding of TARDBP to specific SG proteins, such as TIA1, and via the direct interactions of TARDBP with RNA.³⁸⁰ Furthermore, it was pointed out that the “regulated protein aggregation” leading to the SG formation might be associated with the ALS pathology, because the increased formation of TARDBP-containing SGs is associated with accumulation of TARDBP-based detergent insoluble complexes both in cell cultures and pathological brain tissues.³⁸⁰ Also, treatment with the translational inhibitors that suppress or reverse SG formation inhibits the formation of inclusions mediated by the wild-type or mutant TARDBP,³⁸⁰ suggesting that there is an interplay between TARDBP aggregation, SG formation, and the ALS-FTLD-associated TARDBP mutations, an understanding of which is vital to the clarification of the roles of TARDBP in neurodegeneration.²⁹⁴ It was indicated that pathological TARDBP aggregates can be formed via at least 2 different mechanisms, either following the “independent model”, where the formation of such aggregates is independent of SG formation, or via the “precursor model,” where SG formation might seed the formation of TARDBP aggregates.²⁹⁴ Therefore, TARDBP undergoes transitions between soluble, droplet, and aggregate phases, and these transitions might have important implications for the pathological aggregation and disease development.³⁸¹

There are also several studies emphasizing that the TARDBP-based formation of SGs, due to being regulated by various means, represents an illustrative example of “regulated protein aggregation”. For example, TARDBP involvement in SG formation in SH-SY5Y neuronal-like cells exposed to chronic oxidative stress induced by overnight treatment with the mitochondrial inhibitor paraquat is affected by inhibition of MAPK8/JNK1, MAPK3/ERK1 or MAPK14/p38.³⁸² In addition to oxidative stress, formation of TARDBP-positive SGs can be promoted by ER stress, which is induced in both familial and sporadic forms of ALS, and the extent of which is dependent on the overexpression of ALS-linked mutant TARDBP.³⁸³ Intracellular levels of 4-hydroxynonenal (HNE, a marker of oxidative stress), are elevated in sporadic ALS patients. Treatment of cells with HNE induces noticeable aggregation and phosphorylation of TARDBP, affects cellular localization of this protein, and decreases its accumulation in SGs, suggesting that the aberrations in the SG dynamics induced by the enhanced HNE levels might represent an ALS risk factor.³⁸⁴

The paraquat-induced formation of the TARDBP-positive SGs in SH-SY5Y cells is successfully suppressed by inhibiting the MAPK/ERK-induced phosphorylation of TARDBP by bioavailable bis(thiosemicarbazonato)copper(II) complexes.³⁸⁵ Subsequent screening of various kinase inhibitors using an in vitro model of the formation of TARDBP-positve SGs revealed that the TARDBP accumulation is successfully decreased by the inhibitors of CDKs (cyclin-dependent kinases) and GSK3 (glycogen synthase kinase 3), and these same inhibitors are able to reverse the pre-formed TARDBP-positive SGs.³⁸⁶ A more detailed analysis revealed that the SG integrity is dependent on CDK2 phosphorylation.³⁸⁷ Here, only phosphorylated CDK2 is able to colocalize with TARDBP accumulated within SGs and to phosphorylate HNRNPK (heterogeneous nuclear ribonucleoprotein K) that interacts with TARDBP, whereas nonphosphorylated CDK2 fails to phosphorylate HNRNPK, preventing its incorporation into SGs and thereby preventing accumulation of TARDBP in SGs.³⁸⁷

The fact that the ALS-linked mutations of TARDBP (A315T and Q343R) are able to increase the tendency of this protein not only to aggregate, but also to form SGs provides another important link between ALS pathology and aberrant SG dynamics.³⁸⁸ Fig. 7 provides an illustration for this idea by presenting effects of pathological TARDBP mutants, such as the fALS-associated mutation A315T, sALS-related mutations G294A and N390S, the fALS- and sALS-associated mutation G348C, and truncated TARDBP (residues 1–324) on SG formation and characteristics in HEK293T cells.³⁸⁹ Fig. 7 shows that although the wild-type and all mutant proteins are expressed at similar levels (Fig. 7B) and are predominantly localized to the nucleus in the unstressed HEK293T cells (see “Control” line in Fig. 7A), a multitude of SGs is formed in the cytoplasm of stressed HEK293T cells by each pathological mutant after one h of osmotic stress induced by 0.4 M sorbitol, and these “pathological” SGs are more numerous and noticeably larger than the SGs formed by the wild-type TARDBP under the same conditions (see Fig. 7C for the quantification of the size of SGs formed by various TARDBP species).³⁸⁹ However, not all ALS-related mutations in TARDBP are made equal, and a recent study showed that the A382T mutation causes a significant reduction in the number of SGs per cell, reduces the percentage of cells that form SGs, and leads to a significant decrease of viability, suggesting that the A382T mutation-induced loss of TARDBP function in SG nucleation reduces the ability of cells to respond to stress.³⁹⁰

Figure 7. Effect of the ALS-related mutations within the C-terminal region of human TARDBP on the generation and morphology of SGs in response to osmotic stress. (A) Localization of the exogenous wild-type TARDBP, as well as transiently expressed pathological TARDBP mutants and TARDBP lacking the C-terminal tail in HEK293T cells. Exogenous TARDBP was stained with anti-MYC antibody; nuclei, with ToPro-3. Bar: 10 μm. (B) Levels of expression of proteins shown in panel (A). (C) Quantification of SG size (pixels²/granule) following 1 h of sorbitol stress. Shown are mean granule sizes ± SEM for wild-type (open bar) and mutant (filled bars) TARDBP (*, P < 0.05). Reproduced with permission from ref. 389.

Although many of the aforementioned findings seem to link TARDBP, SGs, and ALS pathogenesis, since alterations in SG dynamics have been suggested to play a key role in the TARDBP protein aggregation process,^{283,292,294,299,391,392} this hypothesis remains hotly contested. This is because SGs are typically observed in studies that are usually conducted under gross overexpression of TARDBP or other ALS proteins, or in the presence of the nonpathologically relevant mutations to those proteins, or following harsh treatments with agents inducing cell stress, such as sodium arsenite or H₂O₂; SGs are not too often observed in ALS patients. To date, the primary criticism of the SG hypothesis is that it generated few “leads” that have withstood the test of time, proven highly reproducible in other laboratories, or proven more generally true in ALS patients. Despite these concerns, the important roles of TARDBP in regulation of SG biogenesis and dynamics are not under question in model studies when cells are stressed and undergo development of SG (and other PMLOs). Obviously, future studies are needed to shed more light on this important problem.

4.3. FUS/fused in sarcoma/translocated in liposarcoma (FUS RNA binding protein)

4.3.1. Structural properties of FUS

The second most common suspect linking aberrant SGs to ALS-FTLD is FUS (FUS RNA binding protein). In fact, FUS is an important player in ALS, mutations of which are responsible for a noticeable subset of both fALS and sALS,^393,394 where they account for ∼4% and 1% of total cases, respectively.^360,395 Furthermore, FUS mutations account for ∼35% of fALS in patients younger than 40 years old.³⁹⁶ The FUS gene is positioned within chromosome 16 (its cytogenic location is 16p11.2, which is the short [p] arm of chromosome 16 at position 11.2) and has 15 exons. It encodes a 526 residue-long RBP, which belongs to the FET family of RPBs that includes FUS, EWSR1, and TAF15 (TATA-box binding protein associated factor 15).^397,398 Human FUS (UniProt ID: P35637) contains several distinct functional domains including a RNA-recognition motif and a highly-conserved C-terminal NLS.³⁹⁵ This C-terminal domain is a hot-spot for the identified ALS-related mutations.³⁹⁹

4.3.2. Biological and pathological roles of FUS

The FUS protein is expressed both in the nucleus and cytoplasm, and can shuttle between these 2 locations. Although the nucleus of neurons is a site of the predominant localization of FUS in the norm, characteristic FUS-immunoreactive inclusions are found in the cytoplasm of the ALS–FUS and FTLD–FUS patients, suggesting that the mislocalization of this protein to the cytoplasm might contribute to the neurodegeneration via the gain-of-toxicity mechanism.^399,400 This protein is involved in several biological functions, which include DNA repair,^401,402 regulation and control of transcription of target genes,⁴⁰³ and RNA processing via regulation of pre-mRNA splicing.³⁵⁴ Recent studies revealed that target gene expression can be significantly altered by FUS mutations due to the ability of mutant forms to bind to the target gene mRNA.⁴⁰⁴ It was also indicated that the ALS-associated R521C mutation in human FUS is able to interfere with the normal function of the wild-type protein by interacting with it and by interfering with normal FUS binding to HDAC1 (histone deacetylase 1).⁴⁰⁵ This mutant FUS also forms more stable complexes with the mRNA of BDNF (brain derived neurotrophic factor) than the wild-type protein and altered alternative splicing of this mRNA.⁴⁰⁵ Furthermore, an interesting feed-forward regulatory loop has been established, where some mutant forms of FUS protein upregulate microRNAs MIR141 and MIR200A that, in their turn, affect FUS protein synthesis.⁴⁰⁶ Although both TARDBP and FUS share the ability to form abnormal cytoplasmic aggregates in tissues of ALS and FTLD patients via interaction with mRNAs, these proteins bind different sets of cytoplasmic mRNAs that, however, converge on common cellular pathways, and differently affect the post-transcriptional fate of their partner mRNAs in the cytoplasm.⁴⁰⁷ Similar to TARDBP, FUS can interact with RNA and with several members of the hnRNP family of RNA-binding proteins.⁴⁰⁸ Although aggregation of FUS and TARDBP is observed in the cytoplasm of ALS patients, the related mechanisms of pathogenesis are assumed to be different.³⁷⁰ This hypothesis is based on the observations that TARDBP aggregates are observed primarily in adult patients, whereas FUS aggregates are found in juvenile patients.³⁷⁰

4.3.3. FUS and PMLOs

In ALS-FTLD, the cytoplasm-located SGs preferentially contain mutant FUS but not the wild-type protein,^409,410 suggesting that the pathogenic FUS mutations are related to the altered dynamics of SG assembly and reflect the presence of important differences between normal and disease physiology.^409–411 Furthermore, OS causes mutant FUS recruitment into SGs, leading to the sequestering of the wild-type FUS, disruption of RNA processing, and initiation of cell death.⁴¹² The involvement of fALS-associated mutations in FUS cytoplasmic mislocalization is attributed to the fact that the majority of these mutations affect the NLS positioned within the C terminus of FUS, thereby impairing nuclear import of this protein.^400,413,414 Similar effects were described for the protein with a deleted C-terminal 32 amino acid residues containing an effective NLS.⁴¹⁵

All these observations emphasize the crucial roles of the ALS mutations or deletion of the NLS of FUS in the distortion of RNA metabolism via impairment of FUS nuclear localization and induction of the cytoplasmic inclusions and SGs.⁴¹⁵ The biological and pathological importance of the observed effects of the C-terminally truncated FUS was recently reemphasized by finding an ALS-related mutation in the FUS gene consisting of a 2-base pair deletion, c.1509_1510delAG, resulting in a truncated protein, p.G504Wfs*12, lacking the NLS.⁴¹⁶ Expression of this truncated FUS leads to the severe cytoplasmic mislocalization of mutant FUS and strong colocalization of this protein with SGs.⁴¹⁶

ALS-related mutations in FUS have a complex effect on SG dynamics. In fact, although SG assembly is delayed in cells expressing mutant FUS, once formed, the mutant-FUS-containing SGs are noticeably different from the SGs lacking FUS, being more dynamic, larger, and more abundant, and disassembling more rapidly once stress is removed.⁴¹⁷ ALS-linked C-terminal mutations have profound effects on cellular localization and functions of FUS protein, which is diffusely mislocalized in the cytoplasm in the form of SGs and spontaneously formed large aggregates. In addition to mutant FUS, these inclusions contain wild-type FUS and RNA-binding proteins HNRNPA1, HNRNPA2B1, and SMN1.⁴¹⁸ These large, spontaneously formed FUS mutant-derived aggregates in the cytoplasm sequester a variety of RNA binding proteins and mRNAs suggesting toxic gain of function leading to the disruption of the various aspects of RNA equilibrium and biogenesis.⁴¹⁸

In addition to ALS-related mutations, the methylation of FUS arginines by PRMT1 (protein arginine methyltransferase 1) plays an important role in regulation of the nuclear-cytoplasmic shuttling of wild-type FUS and its amyotrophic lateral sclerosis 6-associated mutants participating in the mutation-induced toxic gain of function.^419,420 Also, oxidative stress promotes the accumulation of mutant FUS in cytoplasmic SGs leading to the sequestration of the wild-type protein.⁴²¹

Curiously, the abilities of the endogenous wild-type FUS to translocate from the nucleus to the cytoplasm and to be engaged in SG formation are related to the physiological cellular response to the hyperosmolar stress induced by sorbitol, with the cell viability in response to sorbitol being mostly lost in cells with reduced expression of FUS.⁴²² It was pointed out that the recruitment of FUS in SGs might represent a unique cellular defense mechanism against irreversible aggregation of FUS protein mislocalized to the cytoplasm.⁴²³

4.4. Protein C9orf72 and dipeptide repeat proteins Poly(GA), Poly(GR), Poly(GP), Poly(PA), and Poly(PR) generated due to the hexanucleotide repeat expansion of the intronic noncoding region of the C9orf72

4.4.1. Structural properties of C9orf72 and DPRs

The most common genetic change underlying both FTLD and ALS is related to the alterations of the C9orf72 (chromosome 9 open reading frame 72) gene. The cytogenic location of C9orf72 is 9p21.2, which is the short (p) arm of chromosome 9 at position 21.2. The C9orf72 gene consists of 12 exons, and multiple transcript variants encoding different isoforms can be generated by alternative splicing. The major pathological alteration of C9orf72 is the hexanucleotide repeat expansion (GGGGCC/CCCCGG) in the noncoding region located in an intronic region of the C9orf72 gene between its noncoding exons 1a and 1b.^424,425 This hexanucleotide repeat expansion can vary from 10 to thousands of repeats and is associated with 30 to 60% of fALS cases, 8% of sALS cases, 25% of FTLD cases, and up to 88% of cases in familial ALS-FTLD.^{332,424–430} Said hexanucleotide repeat expansion is now considered as the most common cause of familial and sporadic TARDBP-positive FTLD and ALS (C9orf72-FTLD/ALS).^431,432

Although the hexanucleotide repeat expansion affects the noncoding region of the C9orf72 gene, the expanded repeats can be bi-directionally transcribed, generating numerous sense and antisense repeat RNAs.^433–436 Furthermore, the resulting sense and antisense repeat RNAs can be translated to generate a series of the dipeptide repeat proteins (DPRs or C9-related polypeptides synthesized via RAN-translation). When sense GGGGCC repeat RNAs are used as mRNAs in the RAN-translation, 3 different proteins, poly(GA), poly(GR), and poly(GP), are synthesized,^437,438 whereas RAN-translation of the antisense CCCCGG repeat RNAs generates poly(PA), poly(PR), and poly(GP) proteins.^434,436,439 Expanded repeat-containing RNAs are engaged in the formation of RNA foci that sequester RNA-binding proteins, whereas RAN-generated DPRs are found as aggregates throughout the CNS of C9orf72-ALS-FTLD patients.^{434,436–440}

Alternative splicing of human C9orf72 generates 2 proteoforms (UniProt ID: Q96LT7), a canonical or long isoform consisting of 481 residues, and a short isoform that has 222 residues due to lack of the C-terminal half of the sequence (residues 223–481).

As was already mentioned, due to their highly redundant sequences, all DPRs, including poly(GA), poly(GR), and poly(GP), poly(PA), and poly(PR),^433-436 synthesized by RAN-translation using the sense and antisense repeat RNAs generated as a result of the bidirectional transcription of the hexanucleotide repeat expansions of the C9orf72 gene are expected to be mostly disordered, and in agreement with this expectations, all these DPRs are also predicted to be highly disordered.²⁰¹

4.4.2. C9orf72 in the norm and pathology

Although the biological functions of the C9orf72 protein remained unknown for a long time, recent analysis revealed that this protein can be involved in interactions with CFL1 (cofilin 1) and other actin binding proteins, thereby regulating the axonal actin dynamics and also serving as a modulator of small GTPases.⁴⁴¹ Also, C9orf72 takes part in the regulation of endosomal trafficking.⁴⁴² Furthermore, in neuronal cell lines, primary cortical neurons and human spinal cord motor neurons, C9orf72 colocalizes with RAB/RAS-related proteins implicated in autophagy and endocytic transport—RAB1, RAB5, RAB7 and RAB11—indicating that C9orf72 might be related to autophagy regulation.⁴⁴²

The repeat-positive patients are characterized by the reduced expression levels of both AS isoforms of human C9orf72.^443–445 However, the exact position of the C9orf72 loss-of-function mechanism in the C9-FTLD/ALS pathogenesis is not clear as of yet. In fact, although several studies ruled out a loss of function of the C9orf72 protein as a contributor to ALS pathology⁴⁴⁶ (because, e.g., cellular survival in vitro⁴⁴⁷ or in vivo⁴⁴⁸ of both cortical and motor neurons is not affected by the reduction of the C9orf72 levels), many laboratories continue to study the roles of C9orf72 in ALS pathology. Also, C9orf72-depleted motor neurons are characterized by reduced axonal actin dynamics, which is associated with the altered capacity of neurons to maintain axons and axon terminals in the cellular pathology underlying ALS and FTLD.⁴⁴¹ The reduction in the axonal actin dynamics is associated with a pathway involving phosphorylation of the actin binding protein CFL1 through the C9orf72-dependent modulation of the activity of the small GTPases ARF6 (ADP ribosylation factor 6) and RACK1 (receptor for activated C kinase 1) leading to the enhanced activity of LIMK1 LIM domain kinase 1) and LIMK2.⁴⁴¹ Furthermore, loss of C9orf72 in mouse brain neuroblastoma cells alters functions of a complex containing WDR41 (WD repeat domain 41) and SMCR8/Smith-Magenis syndrome chromosome region, candidate 8) that is involved in modulation of the GTPase functions of the members of RAS oncogene family, RAB8A and RAB39B, resulting in impaired autophagy in neurons.⁴⁴⁹

It was pointed out that the mutations in C9orf72 not only cause the appearance of malformed RNA molecules, but also promote mislocalization of TARDBP to the cytoplasm.⁴⁵⁰ Conversely, in various models, DPRs alter cellular functions and induce toxicity in different ways.⁴⁴⁰ For example, the arginine-rich DPRs—poly(GR) and poly (PR)—display the highest toxicity in primary neurons and fly models, where they trigger nucleolar stress, nuclear transport defects, protein mislocalization, and RNA processing alterations.^447,451,452 In contrast, expression of poly(GA) proteins in primary neurons results in aggregation of the transport factor UNC119, proteasome impairment, and impairment of RAD23B/HR23B and nucleocytoplasmic transport proteins.^453–455 Finally, very little or almost no toxicity is reported for poly(GP) and poly(PA).^434,447

4.4.3. C9orf72, DPRs, and SGs

It has been established that the expanded repeats located within the noncoding region of the C9orf72 gene can be bidirectionally transcribed, and both sense and antisense repeat RNAs are engaged in the formation of RNA foci in the CNS of ALS and FTLD patients.^433–436 Besides RNA foci containing hexanucleotide sense and antisense repeat RNAs, the afflicted patients are characterized by the presence of diverse proteinaceous inclusions throughout the central nervous system (CNS).⁴⁵⁶ These inclusions are composed of the DPRs (or C9-related polypeptides synthesized via RAN-translation) generated from the GGGGCC repeat RNA via repeat-associated non-ATG (RAN) translation. Therefore, the hexanucleotide repeat expansion of the C9orf72 gene produces a double-hit impact on the cells via sequestration of some of the RNA-binding proteins into RNA foci (e.g., as is observed when a GGGGCC₃₁ repeat expression in NSC34 and HeLa cells causes sequestration of PURA and its binding partner, FMR1/FMRP, in the intracytosolic SGs)⁴⁵⁷ and via formation of various intracellular inclusions (including SGs) containing RAN-generated DPRs.⁴⁵⁸ Systematic analysis of the interactomes of all DPRs revealed that arginine-containing poly(GR) and poly(PR) are engaged in binding to RBPs and proteins with low complexity sequence domains, which are common components of several PMLOs, such as SGs, nucleoli, and the nuclear pore complex.⁴⁵⁸ Furthermore, poly(GR) and poly(PR) alter dynamics and functions of multiple PMLOs.⁴⁵⁸ For example, localization of these 2 DPRs to the nucleolus induces suppression of ribosomal RNA synthesis, impairs SG formation, and causes translocation of the key nucleolar component NPM1 (nucleophosmin 1), leading to nucleolar stress and cell death.⁴⁵⁹ Finally, the C9orf72 protein itself is able to interact with several ALS-related proteins, such as UBQLN2, HNRNPA2B1, HNRNPA1, and ACTB (actin beta).⁴⁴² Furthermore, in cells overexpressing C9orf72, in addition to the formation of nuclear C9orf72 aggregates, prominent HNRNPA1- and HNRNPA2B1-positive SGs emerge, and the proportion of these HNRNPA1- and HNRNPA2B1-positive SGs noticeably increases when the cells overexpressing C9orf72 are treated with proteasome inhibitors.⁴⁴²

4.5. PFN1 (profilin 1)

4.5.1. Structural properties of PFN1

Human PFN1 (profilin 1; UniProt ID: P07737) is a small, globular, actin-binding protein, mutations in which constitute a relatively rare cause of fALS,⁴⁶⁰ mostly among patients with European ancestry.⁴⁶¹ The PFN1 gene is characterized by the cytogenetic location 17p13.2, which is the short (p) arm of chromosome 17 at position 13.2. It has 6 exons and can generate at least 6 alternatively spliced variants.

4.5.2. Normal and pathological functions of PFN1

Human PFN1 is a 140 residue-long ubiquitous eukaryotic protein that serves as a major regulator of filamentous actin growth via binding actin-ADP monomers, promoting the conversion of actin-ADP to actin-ATP, transporting the actin-ATP monomers, and interactions within FMNs/formins located at the growing end of actin filaments.^462,463 PFN1 also interacts with phosphoinositides, thereby serving as a link between the phosphatidylinositol cycle and actin polymerization.^464,465 Besides serving as an essential component of the cytoskeletal rearrangement signaling pathway, PFN1 can interact with a large cohort of proteins with poly-l-proline stretches.^466,467 Curiously, only mutant forms of PFN1 are engaged in the formation of insoluble aggregates, disrupt cytoskeletal structure, inhibit filamentous actin formation, and elevate UBC (ubiquitin C) and SQSTM1/p62 (sequestosome 1) levels in motor neurons, whereas the wild-type protein does not have such activities.⁴⁶⁸

It was shown recently that the native conformation of PFN1 in vitro is severely destabilized by the ALS-linked mutations, and the turnover of the PFN1 mutant forms is dramatically accelerated in cells.⁴⁶⁹ This mutation-induced structural destabilization of PFN1 defines the increased propensity of ALS-linked variants to aggregate and also their loss-of-function phenotypes.⁴⁶⁹ In SH-SY5Y neuroblastoma cells, transient expression of ALS-linked PFN1 mutants (and not of the wild-type protein) causes the formation of cytoplasmic SQSTM1- and ubiquitin-positive aggregates that sequester endogenous TARDBP, induce accumulation of TARDBP, and promote conversion of normal TARDBP into its abnormal (detergent-insoluble and phosphorylated) form, suggesting the presence of the gain-of-toxic function scenario, where the ALS-related mutations equip PFN1 with a novel toxic function that causes TARDBP aggregation.⁴⁷⁰

4.5.3. PFN1 and stress granules

Although mutations in the PFN1 gene are associated with fALS and provide a potential link between the actin cytoskeleton abnormalities and ALS pathogenesis, these ALS-linked PFN1 mutations are rare, and the actual role of mutations in this protein in neurodegeneration is poorly understood. Recently, some insights into the potential pathological mechanisms of the PFN1 mutations were gained using the budding yeast Saccharomyces cerevisiae that has a PFN1 ortholog (in fact, PFN1 is highly conserved through evolution) as a model and utilizing the power of yeast genetics in the form of an unbiased, genome-wide synthetic lethal screen with yeast cells lacking profilin (pfy1∆).⁴⁷¹ This analysis revealed that in the pfy1∆ background, deletion of some genes related to SGs and P-bodies is synthetically lethal, suggesting that PFN1 can be related to the SG and P-body dynamics.⁴⁷¹ This hypothesis is supported by the analysis of mouse primary cortical neurons and human cell lines that show association of PFN1 and the related protein PFN2 with the SGs, and revealed that SG dynamics is indeed altered by the ALS-linked mutations in PFN1.⁴⁷¹

4.6. Nucleoporin GLE1 (GLE1, RNA export mediator)

4.6.1. Structural properties of GLE1

The gene encoding the human nucleoporin GLE1 protein is located at position 34.11 of the long (q) arm of chromosome 9 (cytogenetic location: 9q34.11) and contains 22 exons, generating at least 11 transcript variants encoding different alternatively spliced isoforms. GLE1 serves as a shuttling mRNA export factor that is needed for the export of the poly(A) tail-containing mRNAs from the nucleus into the cytoplasm,^207-210 and also may play a role in the terminal step of mRNA transport through the nuclear pore complex (NPC) via interaction with nucleoporins NUPL2/CG1 (nucleoporin like 2) and NUP155.⁴⁷² Being an essential multifunctional modulator of DEAD-box RNA helicases in yeast, Gle1 is involved in translation initiation through regulation of a DEAD-box protein, the initiation factor Ded1,⁴⁷³ and plays a role in termination of translation via interaction with another DEAD-box protein, Dbp5.⁴⁷⁴ There are 2 alternative splicing-generated isoforms of GLE1 in human, with the canonical form (or GLE1B) consisting of 698 residues, and with the GLE1A form missing the C-terminal tail (residues 660–698) and harboring the change IEAI → YQAC at residues 656–659 constituting a new C terminus. Although GLE1B and GLE1A share 655 identical residues, these isoforms are characterized by different intracellular distribution. In fact, although the canonical GLE1B isoform is localized predominantly to the NPC, the short GLE1A isoform is spread diffusely throughout the nucleus and cytoplasm.²⁰⁸ Localization of GLE1B to the NPC is ensured by the interaction of its N-terminal tail (residues 1–29) with NUP155,²¹² as well as via binding of the C-terminal tail (residues 656–698) to NUPL2.⁴⁷² In addition to its NUP155- and NUPL2-binding motif located at the N- and C-terminal tails, respectively, GLE1 (UniProt ID: Q53GS7) possesses a region related to the mediation of protein shuttling between the nucleus and the cytoplasm (residues 444–483). This protein also has 2 coiled-coil regions (residues 151–277 and 306–356).

4.6.2. Functions and dysfunction of GLE1

Two rare deleterious ALS-associated mutations in GLE1 (a splice site 1965-2A>C and a nonsense 209C>A mutation) deplete GLE1 at the nuclear pore, thereby abrogating function of this protein in nuclear mRNA export.⁴⁷⁵ The splice site mutation affects intron 14 of GLE1, where it destroys a splice acceptor site, resulting in the expression of a mutant protein, in which the C-terminal 44 residue-long region is replaced with a novel 88 amino acid C-terminal domain.⁴⁷⁵ A nonsense c.209C>A mutation introduces a premature stop codon in exon 2 of GLE1 thereby encoding a truncated GLE1-S70X protein of only 69 residues.⁴⁷⁵ The resulting c.209C>A mRNA is highly unstable and the truncated GLE1-S70X protein is not found.⁴⁷⁵ In addition to these 2 deleterious mutations, some ALS patients possess a missense variation c.2089C>T encoding a mutant protein harboring the R697C substitution.⁴⁷⁵ Functional implementations of this mutation have not been assessed as of yet.

Functional analysis revealed that the splice site mutant possesses functions ascribed to both GLE1A and GLE1B isoforms, suggesting that it is not subject to the normal mechanisms of GLE1 regulation.⁴⁷⁶ Besides causing ALS, mutations in human GLE1 are associated with autosomal recessive conditions that lead to death of fetuses before the 32^nd week of gestation or soon after birth, lethal congenital contracture syndrome 1 (LCCS1 caused by the T144 → TPFQ and R569H mutations) and lethal arthrogryposis with anterior horn cell disease (LAAHD caused by the V617M and I684T mutations).^477–479

4.6.3. GLE1 in PMLOs

One of the fALS forms associated with mutations in the GLE1 gene result in the production of a protein that is defective in SG regulation, and has some translation-related cytoplasmic functions.⁴⁸⁰ As was already pointed out, GLE1 protein is present in cells as at least 2 isoforms generated by alternative splicing. These 2 forms are differently expressed (with the canonical form being more common), have different cellular localizations (with GLE1B being primarily localized to the nuclear envelope rim and with GLE1A being mostly found in the cytoplasm), and have different functions (with GLE1B interacting with NPC and playing a role in efficient and expedient mRNA export from the nucleus,⁴⁸¹ and with GLE1A being involved in essential cytoplasmic functions related to translational modulation and regulation of proper assembly and disassembly of SGs⁴⁸⁰).^{208,210,212,472} The effect of the ALS-related mutation on the GLE1 sequence can be considered from the GLE1A and GLE1B angles. From the GLE1B perspective, this mutation causes the replacement of the 44-amino acid NUPL2 binding site with a novel 88 amino acid C-terminal domain, whereas from the GLE1A perspective, the GLE1-c.1965-2A>C mutation in the GLE1 gene results in the replacement of a short C-tail with a long C-terminal domain. Because this new inserted domain cannot bind to the nucleoporin NUPL2,⁴⁷⁵ it was expected that GLE1-IVS14-2A>C would be functionally similar to GLE1A.⁴⁷⁶ In agreement with this hypothesis, GLE1-IVS14-2A>C is absent from the nuclear envelope rim, causes an increase in the cytoplasmic functions of GLE1, is recruited to SGs, and is able to regulate assembly and disassembly of SG2, resulting in the formation of larger SGs.⁴⁷⁶ Furthermore, both the GLE1A and the GLE1-IVS14-2A>C form cytoplasmic “fibrous” aggregates that do not colocalize with the components of SGs but recruit TARDBP and the molecular chaperone HSP90.⁴⁷⁶ However, in contrast to normal GLE1A that is not involved in mRNA export from the nucleus, GLE1-IVS14-2A>C is able to support some level of mRNA export in cells with endogenous GLE1 depletion.⁴⁷⁶ Therefore, the ALS-related GLE1-IVS14-2A>C isoform is capable of acting as both GLE1A and GLE1B, and may contribute to ALS pathogenesis via distortion of the normal mechanisms of GLE1 regulation.⁴⁷⁶

4.7. Transcriptional activator PURA/Pur-α (purine rich element binding protein A)

4.7.1. Structural properties of PURA

The gene encoding the transcriptional activator PURA/PUR1 (purine rich element binding protein A) has the cytogenetic location 5q31.3 (i.e., at position 31.3 of the long [q] arm of chromosome 5) and contains 5 exons. Originally, transactivator PURA was shown to bind to the purine-rich single strand of the PUR element located at a putative origin of replication site in the MYC gene.⁴⁸² Subsequent studies revealed that PURA is involved in (mostly negative) regulation of several genes, such as SPN/leukosialin/CD43 (sialophorin),⁴⁸³ FAS (Fas cell surface death receptor),⁴⁸⁴ APBB1/FE65 (amyloid beta precursor protein binding family B member 1),⁴⁸⁵ MBP (myelin basic protein),⁴⁸⁶ CHRNB4 (cholinergic receptor nicotinic beta 4 subunit),^487,488 and ACTA2 (actin, alpha 2, smooth muscle, aorta).⁴⁸⁹

Human PURA (UniProt ID: Q00577) is a 322-residue long protein that possesses 2 regions with compositional biases, a Gly-rich domain (residues 11–53) and a Gln- and Glu-rich region (residues 293–322) that might serve as a part of the transcriptional activation domain, and a central DNA-binding domain containing 3 class I repeats and 2 class II repeats.⁴⁹⁰

4.7.2. Biological and pathological roles of PURA

PURA is a highly conserved, ubiquitous and multifunctional protein engaged in a sequence-specific interaction with single-stranded DNA and RNA.²²⁰ Among different functions ascribed to this protein are targeting mRNA to neuronal dendrites,^491–493 DNA replication, DNA repair, and gene transcription.^490,494,495 Being a single-stranded nucleic acid-binding protein, PURA possesses DNA helix-destabilizing helix unwinding activities.⁴⁹⁶ The preferred DNA sequence recognized by PURA is composed of the GGN repeats, and binding of PURA to this recognition sequence is accompanied by the formation of multimeric complexes and is controlled and amended by interaction with other transcription factors.^490,497,498 In addition to interaction with single-stranded DNA and RNA, PURA can bind to a cohort of cellular regulatory proteins, such as the CCNA1 (cyclin A1)-CDK2 complex,^499,500 CCNT1 (cyclin T1),⁵⁰¹ CDK9,⁵⁰¹ E2F1 (E2F transcription factor 1),⁵⁰² RB1 (RB transcriptional corepressor 1),⁵⁰³ SP1 (Sp1 transcription factor),⁴⁸⁶ and YBX1/YB1 (Y-box binding protein 1).⁵⁰⁴ Furthermore, this protein is considered to be a major player in the regulation of the cell cycle and can be involved in oncogenic transformation,⁴⁹⁰ and its intracellular level varies during the cell cycle, peaking during mitosis and decreasing at the onset of S phase.⁴⁹⁹

It was pointed out that PURA might be involved in the pathogenesis of the FUS mutation-associated form of ALS through the control of mRNA translation, indicating that altered protein synthesis may be implicated in the disease.⁵⁰⁵

4.7.3. Roles of PURA in SG biogenesis

Engagement of PURA in ALS pathogenesis is related to the sequestration of this important RNA/DNA-binding protein in the inclusions containing the expanded GGGGCC repeat RNAs generated by the expansion of the hexanucleotide repeats in the first intron of C9orf72.^506-508 Also, in the presence of the GGGGCC₃₁ repeats, PURA efficiently accumulates in the intracytosolic granules positively stained with SG markers.⁴⁵⁷

Furthermore, PURA is able to specifically bind to the FUS C-terminal region, physically interact with the mutated FUS in an RNA-dependent manner, and colocalize with the mutated FUS to the SGs found in the motoneuronal cells derived from ALS patients.⁵⁰⁵ In ALS patient cells carrying disease-causing FUS mutations and challenged with stress and in mammalian neuronal cells, PURA efficiently colocalizes with mutant FUS in constitutive SGs.⁵⁰⁹ Because SG formation in mammalian cells is significantly reduced by the shRNA-mediated knockdown of endogenous PURA, it has been concluded that this protein is essential for SG assembly.⁵⁰⁹

4.8. FLCN (folliculin)

4.8.1. Structural properties of FLCN

Human FLCN/BHD with cytogenetic location 17p11.2, which is the short (p) arm of chromosome 17 at position 11.2, contains 32 exons, and encodes a 579 residue-long protein (folliculin). FLCN serves as a tumor suppressor, mutations in which are associated with BHD syndrome, a rare inherited genodermatosis characterized by kidney tumors, lung cysts, and benign tumors of the hair follicle (hair follicle hamartomas).^510-513 Mutations in FLCN are also associated with primary spontaneous pneumothorax^514-516 and familial renal cell carcinoma.⁵¹⁷ FLCN is highly conserved across species, with human protein showing 92% identity to the mouse FLCN, being 22–36% identical (44–56% similar) to the Drosophila melanogaster protein and 27–28% identical (44–52% similar) to the Caenorhabditis elegans ortholog.⁵¹¹

Human FLCN (UniProt ID: Q8NFG4) has 1 coiled-coil region (residues 287–310) and might be present in 3 isoforms generated by alternative splicing, where isoforms #2 and #3 differ from the canonical form by missing long regions C-terminally located regions (residues 343–579 and 198–579) and possessing the DLEEESESWD…SGCGSWQPRK→GEAGVLL PGP…LREAHPPISV and CPGREGPIFF…FLLGKVRGII→SLVATEPVSV…ELREESCWTC changes at residues 291–342 and 134–197, respectively. Curiously, most cancer-related mutations in this protein are associated with the C-terminal truncations in FLCN suggesting that this domain might play a role in tumor suppression.⁵¹⁸ A crystal structure solved for the C-terminal domain of human FLCN (residues 341–566) shows that this domain is characterized by an αβ structure, where a core β-sheet and helices packed on the one side are followed by an all-helical region, and which is similar to the differentially expressed in normal cells and neoplasia (DENN) domain of DENND1B/connecdenn 2 (DENN domain containing protein 1B; PDB ID: 3V42, see ref. 518).

4.8.2. Functionality and dysfunctionality of FLCN

FLCN is related to energy and/or nutrient sensing and the MTOR (mechanistic target of rapamycin) pathway via the FLCN-interacting proteins FNIP1 and FNIP2 that also bind to AMP-activated protein kinase (AMPK), which is involved in energy sensing and negative regulation of MTOR activity.^519,520

FLCN is related to ALS via its direct interactions with TARDBP (where the residues 202–299 of FLCN interact with the RRM domains of TARDBP) leading to the modulation of TARDBP cytoplasmic translocation and aggregation.⁵²¹ FLCN is needed for TARDBP nuclear export, but not nuclear import.⁵²¹ The overexpression of FLCN is associated with enhanced TARDBP cytoplasmic mislocalization, whereas reduced FLCN levels lead to the preferential deposition of TARDBP to the nucleus.⁵²¹

4.8.3. FLCN and PMLOs

The FLCN-driven cytoplasmic mislocalization of TARDBP leads to SG formation and accumulation of TARDBP aggregates that are recognized by the UPS and autophagy-lysosome pathway.⁵²¹ Curiously, under stress conditions, FLCN is needed for the successful incorporation of TARDBP into the SGs, because in the arsenite-treated cells containing endogenous FLCN, TARDBP is preferentially found in the cytoplasm where it colocalizes with SGs, whereas after arsenite treatment of the FLCN-depleted cells, TARDBP dissociates from the SGs and shuttles back into the nucleus.⁵²¹

4.9. RBM45 (RNA binding motif protein 45)

4.9.1. Structural properties of RBM45

The gene encoding human RBM45/DRB1 (RNA binding motif protein 45) has the cytogenic location 2q31.2. The RBM45 gene has 8 exons and can generate 3 alternatively spliced isoforms.

The amino acid sequence of human RBM45 (UniProt ID: Q8IUH3) consists of 476 residues and possesses 3 RRMs (residues 26–106, 121–195, and 392–464) as well as a C-terminal nuclear localization sequence (within the C-tail residues 454–472).⁵²² There is also a homo-oligomer assembly (HOA) domain located within the linker region between RRM2 and RRM3 that promotes homo-oligomerization of RBM45 and is highly conserved across species.⁵²² In addition to the full-length canonical form, alternative splicing generates 2 proteoforms, where intron retention in isoform 2 generates a sequence missing the 225–226 dipeptide, missing a C-terminal domain (residues 331–476) and having a D330E substitution. Isoform 3 is minimally different for the canonical form, being characterized by the missing 225–226 dipeptide.

4.9.2. Biology and pathology of RBM45

RBM45 protein belongs to the family of neural RRM-type RNA-binding proteins that play a number of important roles in neural development.⁵²³ It was also reported that human RBM45 preferentially interacts with the poly(C) RNA.⁵²³ Although this protein is predominantly found in the nucleus, it contains a nuclear localization signal and can shuttle between the nucleus and cytoplasm.⁵²⁴ Recent immunoprecipitation and mass spectrometry analysis revealed that 132 proteins can specifically interact with RBM45, and that this excessive protein-protein interaction (PPI) network includes many RBPs engaged with RBM45 mostly via RNA-dependent interactions in the nucleus.⁵²⁵ Among the RBM45-interacting proteins discovered in this study were several ALS-linked RBPs, such as FUS, HNRNPA1, and TARDBP,⁵²⁵ and among the biological processes and pathways ascribed to the RBM45-interacting proteins are RNA processing, RNA splicing, cytoplasmic RNA translation, and EIF2 (eukaryotic translation initiation factor 2) and EIF4 pathways.⁵²⁵

Recently, it was reported that levels of RBM45 are statistically elevated in the cerebrospinal fluid of patients with sporadic and familial ALS, and cytoplasmic inclusion bodies containing RBM45, ubiquitin, and TARDBP are found in 91% of ALS, 100% of FTLD–TARDBP and 75% of Alzheimer disease cases.⁵²⁶ Immunochemical analysis also revealed that RBM45 forms a punctate pattern within nuclei of neurons and glia in the brain and spinal cord that lack either TARDBP or ubiquitin. Patients with the C9orf72 hexanucleotide repeat expansion display the most extensive RBM45 pathology,⁵²⁶ likely due to the aforementioned binding preference for poly(C) RNA. However, the generality of this conclusion is questionable, because only 3 C9orf72 patients were analyzed in this study. Subsequent studies also showed that RBM45 is able to bind and colocalize with the C-terminal fragment of TARDBP.⁵²⁷ This protein also modulates the antioxidant response in ALS through interactions with KEAP1 (kelch like ECH associated protein 1), which is an inhibitor of the antioxidant response transcription factor NFE2L2 (nuclear factor, erythroid 2 like 2).⁵²⁸ It was also shown that one of the pathological mechanisms linking RBM45 with ALS can be related to the mislocalization of this protein and perturbation of its nuclear-cytoplasmic trafficking that could be responsible for the induction of the formation and accumulation of toxic cytoplasmic aggregates.⁵²⁴

4.9.3. RBM45 and SGs

The cytoplasmic inclusions containing RNA-binding protein RBM45, being present in neurons and glia in a majority of ALS and FTLD patients, incorporate TARDBP and are marked with ubiquitin.⁵²⁶ It was pointed out that, while similar to TARDBP and FUS, RBM45 possesses RRMs, and does not have a characteristic glycine-rich prion-like domain (PrLD),⁵²⁶ but contains a bipartite nuclear-localization sequence located at the C-terminal region.⁵²² Besides RNA binding, RBM45 is able to interact with the C-terminal domain of TARDBP⁵²⁷ both in the nucleus and cytoplasm,⁵²² can bind to FUS in an RNA-dependent manner,⁵²² and is able to form homo-oligomers via the evolutionarily conserved motif located within the linker region between RRM2 and RRM3.⁵²²

Although RBM45, TARDBP and FUS are predominantly localized to the nucleus in healthy cells, these proteins form cytoplasmic inclusions in cells of ALS-FTLD patients, with RBM45 homo-oligomerization being needed for the efficient association with other ALS-linked proteins.⁵²² Also, when cells are exposed to oxidative stress, all 3 proteins shuttle from the nucleus to the cytoplasm, where they are engaged in SG formation.^{283,299,528,529} Altogether, currently available data suggest that RBM45 self-oligomerization in the cytoplasm plays a role in the incorporation of this protein into SGs, where it promotes RBM45 interaction with TARDBP.⁵²²

4.10. SS18L1/CREST (SS18L1, nBAF chromatin remodeling complex subunit)

4.10.1. Structural properties of SS18L1

The SS18L1/CREST/KIAA0693 (SS18L1, nBAF chromatin remodeling complex subunit) gene is located on chromosome 20q13.33 and contains 11 exons, encoding a protein required for calcium-dependent dendritic growth and branching in cortical neurons.⁵³⁰

Human SS18L1 (UniProt ID: O75177) is a 396-residue-long protein that has an N-terminal autoinhibitory-autoregulatory domain (residues 1–148) needed for SMARCA4/BRG1 (SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4) binding,⁵³⁰ followed by a methionine-rich domain (residues 149–232), a multifunctional domain (MFD, residues 246–317) related to the regulation of the subcellular localization and transactivation of SS18L1 and involved in the protein self-dimerization,⁵³¹ and a C-terminal transactivation domain (residues 325–396) related to the control of nuclear localization of SS18L1 (residues 334–396) containing a binding region crucial for interaction with CREBBP (residues 387–396).⁵³⁰ Furthermore, there are 4 potential SH2-binding motifs in human SS18L1, residues 50–53, 353–356, 371–3789, and 391–394. The C-terminal half of SS18L1 (residues 190–396) is characterized by low sequence complexity (it possesses compositional bias, due to being enriched in Gln residues).⁵³⁰ It was also pointed out that this C-terminal part satisfies the criteria for a PrLD,^530,532 serving as a primary cause of SS18L1 aggregation, a statement supported by the observation that protein aggregation is completely abolished by the deletion of this domain.⁵³³

Human SS18L1is present as at least 5 isoforms generated by alternative splicing, where, in comparison with the full-length canonical form, isoforms 2, 3, and 4 are characterized by missing N-terminal regions 1–23, 1–86, and 1–131, respectively, and isoform 5 possesses the MSVAFA SARP…SKGKTAECTQ → MQSLSTEARY…ICPRSPPARR substitution at the 1–49 region.

4.10.2. Normal and pathological functions of SS18L1

SS18L1 is another protein whose mutations are associated with ALS.^532,534 SS18L1 is a nuclear calcium-regulated protein involved in transcriptional activation via its interaction with CREBBP⁵³⁵ mediated by the last 9 residues of SS18L1, and in chromatin remodeling via binding to chromatin remodeling proteins ARID1A/BAF250 (AT-rich interaction domain 1A) and SMARCA4/BRG1.⁵³⁰ In fact, SS18L1 serves as a constituent of the chromatin remodeling complex including SMARCD3/BAF (SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily d, member 3) playing an important regulatory role at various stages of neural development by modulating transcription of specific sets of genes.⁵³⁶ Because SS18L1-deficient mice display defective dendritic branching, early lethality, and motor disturbances, it has been hypothesized that the physiological function of this protein is related to the control of the normal development of neuronal dendritic trees.⁵³⁰

Association of SS18L1 with ALS pathology is based on finding mutated forms of this protein (e.g., Q388stop and Ile123Met) in sporadic ALS,⁵³² and Ala264Thr and p.Gln222_Ser224del mutations in patients with the familial form of ALS.⁵³⁴ Besides being involved in ALS, SS18L1 is associated with synovial sarcoma, which is an aggressive soft tissue tumor caused by the t(X;18)(p11;q11) translocation, where the SS18 gene on chromosome 18 is fused to one of the SSX genes on the X chromosome.^537-540 Among such fusions are SS18-SSX1, SS18-SSX2, and SS18-SSX4 that account for more than 95% of the synovial sarcomas^537-540 and a SS18L1-SSX1 fusion generated by the in-frame fusion of nucleotide 1216 (exon 10) of SS18L1 with nucleotide 422 (exon 6) of SSX1.⁵⁴¹ In the resulting chimeric SS18L1-SSX1 protein, the last 8 residues of SS18L1 are replaced by 78 C-terminal residues of the SSX1 protein, which is a transcriptional repressor domain.⁵⁴¹

4.10.3. SS18L1 and biogenesis of stress granules and other PMLOs

In the nucleus, SS18L1 is localized to the nuclear bodies of unknown nature that are positive for the histone acetyltransferase CREBBP/CBP, and the C-terminal 164 residues are needed for this nuclear body localization of SS18L1.⁵⁴² Furthermore, SS18L1 belongs to the set of paraspeckle proteins.⁵⁴³ A recent study showed that SS18L1 is involved in interaction with FUS and possesses a strong tendency for aggregation.⁵³³ In its aggregated form, SS18L1 affects morphology and integrity of paraspeckles via trapping other paraspeckle proteins,⁵³³ thereby supporting a hypothesis that compromised paraspeckle formation can serve as a pathogenic factor in ALS-FTLD.^544,545 Finally, similar to many other ALS-FTLD-related proteins, SS18L1 is found in SGs.⁵³³ Therefore, SS18L1 possesses an intricate ability to be recruited to at least 3 different types of PMLOs, such as CREBBP-containing nuclear bodies, paraspeckles, and SGs.

SS18L1 is one of the ALS-related proteins that contain a PrLD. Curiously, a recent comprehensive review indicated that almost 70 human RBPs possesses PrLDs, which are “low-complexity domains that possess a similar amino acid composition to prion domains in yeast, which enable several proteins, including yeast Sup35 and Rnq1, to form infectious conformers, termed prions.”⁵⁴⁶ Human RBPs routinely utilize PrLDs in their numerous functions. Furthermore, the presence of PrLDs defines the ability of RBPs to undergo LLPTs leading to the formation of various PMLOs.⁵⁴⁶

4.11. HNRNPA1 (heterogeneous nuclear ribonucleoprotein A1) and HNRNPA2B1

4.11.1. Structural properties of HNRNPA1 and HNRNPA2B1

Human genes HNRNPA1 (with cytogenetic location 12q13.13, which is the long [q] arm of chromosome 12 at position 13.13) and HNRNPA2B1 (located on chromosome 7p15.2, which is the short [p] arm of chromosome 7 at position 15.2) encode members of a family of ubiquitously expressed hnRNPs. The HNRNPA1 and HNRNPA2B1 genes have 13 and 15 exons, respectively, and are able to generate multiple spliced isoforms (at least 9 and 13, respectively).

Both HNRNPA1 and HNRNPA2B1 have similar domain organization. Human HNRNPA1 (UniProt ID: P09651) is 372-residue-long protein possessing 2 globular domains (A and B, residues 4–94 and 95–108) portions of which form RRMs (residues 14–97 and 105–184). Region 218–240 represents the RNA-binding RGG-box, whereas residues 320–357 constitute the nuclear targeting sequence (M9). The C-terminal part of HNRNPA1 (residues 195–372) is a Gly-rich PrLD. There are 2 alternatively spliced isoforms of this protein, where the HNRNPA1-B isoform is considered as a canonical full-length form of the protein, and the HNRNPA1-A isoform (which is 20 times more abundant than HNRNPA1-B) is considered different from the canonical form due to missing residues 252–303.^547-549 Analogously, human HNRNPA2B1 (UniProt ID: P22626) is a 353-residue-long protein that has 2 RRMs (residues 21–104 and 112–191), a low complexity, Gly-rich C-terminally located PrLD (residues 193–253), a nuclear targeting sequence (residues 308–347), and a nuclear localization signal (residues 9–15). Structural information is available for the N-terminal regions of both proteins (residues 2–196 of human HNRNPA1 containing 2 RNA recognition motifs, PDB ID: 2LYV⁵⁵⁰; and residues 1–103 corresponding to the first RRM of the human HNRNPA2B1, PDB ID: 1X4B).

Both HNRNPA1 and HNRNPA2B1 proteins are present as 2 proteoforms generated by alternative splicing, with the HNRNPA2B1 proteoforms being different by a 12 amino acid insertion close to the N terminus,^547-549 but both containing C-terminally located PrLDs that encompass residues 197–353 in HNRNPA2B1 and residues 185–341 in hnRNPA2.^363,551 The HNRNPA1 proteoforms are different due to missing amino acids 252–303 in the shorter HNRNPA1-A isoform, but both harbor PrLDs located at their C-termini that span residues 186–320 in HNRNPA1-A and residues 186–372 in HNRNPA1-B.^363,551

4.11.2. Biological and pathological functions of HNRNPA1 and HNRNPA2B1

HNRNPA1 and HNRNPA2B1 are 2 ubiquitous RNA-binding proteins with a PrLD that are mutated in ALS.⁵⁵² HNRNPA1 is an RNP involved in several RNA metabolic processes, such as pre-mRNA splicing and transport of cellular RNAs, which it accomplishes by shuttling between the nucleus and the cytoplasm.^361,553-555 Curiously, different RNA-binding profiles were assigned to the cytoplasmic and nuclear HNRNPA1, with the cytoplasmic protein possessing a high affinity for the AU-rich elements,^556,557 and with the nuclear protein having a high affinity for a polypyrimidine stretch bordered by AG at the 3′ ends of introns.^558,559

HNRNPA2B1 is among the most abundant RBPs in eukaryotic cells and is highly expressed in brain and muscle. The major function of this RBP is to form the core of the RNP complex that associates with the nascent pre-mRNAs, packaging them into hnRNP particles and thereby playing important roles in various processes ranging from transcription, to pre-mRNA processing, to RNA nuclear export and subcellular location, to mRNA translation, and to modulating the stability of mature mRNAs.⁵⁶⁰ Human HNRNPA2B1 is engaged in the formation of a multitude of HNRNP particles via interaction with several (at least 20) other HNRNPs and heterogeneous nuclear RNA.⁵⁶⁰ The name to this protein is given due to the existence of 2 functional isoforms, HNRNPA2 and HNRNPB1, generated by alternative splicing, with the HNRNPB1 serving as a full-length canonical form and with the HNRNPA2 isoform missing residues 3–14 close to the N terminus.

Mutations in PrLD of human HNRNPA2B1 (Asp302Val) and HNRNPA1 (Asp314Val or Asp314Asn) are associated with ALS and multisystem proteinopathy (also known as inclusion body myopathy with early-onset Paget disease and FTLD). To gain some knowledge on the molecular mechanisms connecting mutations in HNRNPA1 and HNRNPA2B1 with fALS, a structure-based threading algorithm ZipperDB that is designed to predict the presence in a query protein of steric zippers, short (6-amino acid-long) segments able to form 2 self-complementary β-strands,⁵⁶¹ was recently used.⁵⁵² This analysis revealed that the potent steric zipper motifs were introduced by various ALS-associated mutations into the PrLDs of HNRNPA1 and HNRNPA2B1.⁵⁵² A subsequent study provided experimental support to this hypothesis and revealed that although the wild type versions of HNRNPA1 and HNRNPA2B1 have a noticeable intrinsic tendency to form self-seeding fibrils, this fibrillation propensity is dramatically enhanced by the disease-associated mutations that accelerate the formation of self-seeding fibrils that cross-seed polymerization of wild-type proteins.⁵⁵¹

4.11.3. HNRNPA1, HNRNPA2B1, and PMLOs

In addition to enhancing the formation of cytoplasmic inclusions, the ALS-related mutations foster the excess incorporation of HNRNPA2B1 and HNRNPA1 into SGs.⁵⁵¹ In fact, the constitutive SGs contain higher levels of mutant forms of the HNRNPA2B1 and HNRNPA1 proteins than the corresponding wild-type proteins. Furthermore, arsenite treatment of model cells results in more rapid incorporation of mutant proteins into SGs than the wild-type proteins, indicating that disease-related mutations alter the recruitment of HNRNPA2B1 and HNRNPA1 not only into the cytoplasmic inclusions but also into SGs.⁵⁵¹

4.12. ATXN2 (ataxin 2)

4.12.1. Structural properties of ATXN2

Human ATXN2 is a 1,313-residue-long protein involved in the trafficking of EGFR (epidermal growth factor receptor) via binding to SH3GL2 (SH3 domain containing GRB2 like 2, endophilin A1) and SH3GL3 (SH3 domain containing GRB2 like 3, endophilin A3), which are the brain-expressed members of the SH3GL/endophilin A family involved in synaptic vesicle endocytosis.⁵⁶² ATXN2 is a 1,313 residue-long protein most known for its involvement in the pathogenesis of spinocerebellar ataxia type 2 (SCA2) caused by the expansion of the polyglutamine (polyQ) region of this protein encoded by the expanded CAG trinucleotide repeat within the open reading frame of the ATXN2 gene located at position 24.12 on the long (q) arm of chromosome 12 (cytogenetic location: 12q24.12) and encoding multiple transcript variants due to alternative splicing and the presence of 26 exons. In fact, although CAG repeats are known to vary in length in normal alleles of ATXN2, the most common allele carries 22 CAG repeats and the maximal ATXN2 repeat size is 31, whereas expansion of the ATXN2 CAG repeat region to carry 34–59 repeats is associated with SCA2.^563-568

In addition to the canonical full-length form, human ATXN2 (UniProt ID: Q99700) has at least four alternatively spliced isoforms. Isoform 2 is missing the C-terminal tail (residues 996–1313) and harbors the PLYPIPMTPMPVNQAK → YQICPNSGKTSIIRVP substitution in the preceding region (residues 980–995). Isoform 3 is the shortest and the most diversified form of ATXN2, missing N-terminal (residues 1–981) and C-terminal tails (residues 1258–1313), missing the 1106–1123 region, and possessing YPIPMTPMPVNQAKTYR→MYYAVEILFNRQSAFFS and AHVQSGMVP→VIPALAN FL substitutions at residues 982–998 and 1249–1257, respectively, in addition to harboring the Ile1124Val substitution. In isoform 4, the C-tail (residues 1244–1313) is missing, whereas isoform 5 is characterized by missing regions 1–265 and 1106–1124 and possesses the ACPKLPYNKETSPSF YFAI → V substitution. The canonical form has several regions with compositional bias, such as three Pro-rich regions (residues 47–158, 551–734, and 929–1085), a poly-Pro region (residues 55–64), a poly-Gln region (residues 166–187), and a poly-Ser region (residues 213–223). It was indicated that several of these isoforms are present in mice and humans, might have different functions, and are expressed in different tissues, such as brain, spinal cord, cerebellum, heart, and placenta.⁵⁶⁹

4.12.2. Normal and pathological roles of ATXN2

This cytoplasmic RNA-binding protein is considered as an ALS susceptibility factor, because intermediate-length polyglutamine expansions (Q27-Q33) in ATXN2 are associated with increased ALS risk.^570-572 Curiously, although SCA2 is associated with the polyQ expansions exceeding 34 repeats all encoded by pure CAG repeat expansions in the ATXN2 gene, the corresponding expansions associated with ALS are all interrupted, containing 1 to 3 CAA codons (also encoding glutamine) within the CAG repeats.⁵⁷³ Despite the fact that CAG expansion in genes encoding several proteins such as ATXN1, ATXN2, ATXN3, ATXN7, TBP (TATA-box-binding protein), ATN1 (atrophin 1), and HTT (huntingtin) is associated with a set of polyQ diseases, expansions in polyQ disease genes other than ATXN2 are not associated with ALS, suggesting that the repeat-induced alterations in the biological functions of ATXN2 and not the mere presence of an expanded polyQ repeat represent the ALS risk factor.⁵⁷⁴ Because no ATXN2-associated polyglutamine expansion was found in patients with sporadic and familial FTLD, it has been concluded that this polymorphism is ALS specific.⁵⁷⁵ The pathological effect of the ATXN2 polyQ expansion in ALS is associated with the enhancement of the stress-induced activation of CASP3 and increase TARDBP pathological modifications, such as C-terminal cleavage and phosphorylation.⁵⁷⁶ Furthermore, the presence of the ATXN2 intermediate-length expansion affects the morphology of the TARDBP inclusions, because the motor neurons of ALS patients harboring ATXN2 polyQ expansions contain primarily skein-like or filamentous TARDBP deposits and do not contain large round Lewy body-like inclusions abundantly found in ALS patients without ATXN2 polyQ expansions.⁵⁷⁷

4.12.3. ATXN2 in the biogenesis of PMLOs

In addition to TARDBP and FUS, ALS-FTLD-related mutations, which are known contributing factors to SG formation due to the mislocalization of carrier proteins to the cytoplasm, ATXN2 is also known as a constituent protein of SGs in mammalian cells, suggesting the existence of the common pathological cascade formed by TARDBP, FUS, and ATXN2 and related to the aberrations in SG dynamics.⁵⁷⁸ The importance of ATXN2 in the controlling dynamics of PMLOs is supported by the observations that this protein can interact with the DEAD/H-box RNA helicase DDX6 (DEAD-box protein 6), which is an important component of P-bodies and SGs, and that the alterations in the ATXN2 levels might interfere with the assembly of SGs and P-bodies.⁵⁷⁹

Based on the analysis of cellular ALS models, it was also suggested that ATXN2 with an ALS-linked intermediate-length repeat (ATXN2 Q31) might serve as a potent modifier of the FUS-based pathology, promoting FUS translocation to the cytoplasm, inducing ER stress, and promoting Golgi fragmentation.⁵⁸⁰ Furthermore, because ATXN2 colocalizes with FUS in motor neurons derived from sALS and FUS-linked fALS patients, colocalizes with FUS in the ER-Golgi compartments in neuronal cell lines and coprecipitates with FUS in ALS spinal cord lysates, it was hypothesized that the interplay between the mutated FUS and ATXN2 with intermediate-length polyQ expansions provides potential molecular and cellular mechanisms linked to ALS pathology.⁵⁸⁰ Similarly, a connection between the TARDBP-based pathology and ATXN2 with intermediate-length polyQ expansions was noted based on the comparative analysis of the TARDBP-positive inclusions in the ALS cases with and without ATXN2 possessing intermediate-length polyQ expansions.⁵⁷⁷ Here, fewer motor neurons in the ALS cases without extended ATXN2 contain abundant large round (likely SGs) and skein-like TARDBP inclusions, whereas only skein-like TARDBP inclusions are predominantly found in almost ALS cases harboring ATXN2 polyQ expansions.⁵⁷⁷ It was also found that an increased ATXN2 expression might cause TARDBP and FUS mislocalization, leading to RNA dysregulation.^581,582

4.13. TIA1 (TIA1 cytotoxic granule associated RNA binding protein)

4.13.1. Structural properties of TIA1

In humans, the TIA1 gene is located on chromosome 2p13.3 (at position 13.3 of the short [p] arm of chromosome 2). It has 15 exons, generates at least 11 transcript variants and the gene product is an RNA-binding protein involved in alternative pre-RNA splicing and regulation of mRNA translation.⁵⁸³ TIA1 is another RNP containing poly-glycine-rich PrLD that promotes reversible physiological aggregation of TIA1 into SGs,⁵⁸⁴ but, under stress conditions, can initiate formation of pathological RNA granules.^585,586 TIA1 is characterized by a well-developed interactome, shown to co-immunoprecipitate with 163 brain proteins.⁵⁸⁶ Therefore, based on the presence of a domain with profound intrinsic aggregation propensity and the overall high binding promiscuity of TIA1 it is not surprising to find that abnormal TIA1-containing deposits are found in the brain tissue of patients with AD, ALS, Creutzfeld-Jakob disease, frontotemporal dementia with Parkinsonism (FTDP-17), FTLD–TDP43, Huntington disease, and spinal muscular atrophy.^290,584,585

Human TIA1 (UniProt ID: P31483) is a 386-residue-long protein that contains 3 RRMs (residues 7–83, 106–184 and 214–286), which serve to bind RNA and a Gly-rich domain that mediates protein aggregation.²⁹⁰ In addition to the canonical full-length proteoform, 2 isoforms are produced by alternative splicing. Here, isoform 2 has a SSTVVSTQRSQD → N substitution at the 93–104 region, whereas isoform 3 is missing the C-terminal tail (residues 215–386) and has a SNTKQLSYDEVVNQSSPSNC → CRCIGEEKEMW NFGEKYARF substitution in the preceding region (residues 195–214).

4.13.2. Biology and pathology of TIA1

TIA1 can serve as an apoptosis-promoting factor that is able to modulate alternative splicing of dedicated transcripts.⁵⁸⁷ Careful analysis of the consequences of the Tia1 deletion in mouse spinal cord and cerebellum revealed several strong dysregulations, including noticeable effects on the cell cycle and apoptosis regulators, as well as on lipid storage and membrane trafficking factors.⁵⁸⁸

4.13.3. TIA1 and SGs

TIA1 is thought to serve as one of the core RBPs responsible for the primary step (nucleation) of RNA granule formation followed by recruitment of secondary RBPs to form mature RNA granules, such as SGs. SGs act as key players in environmental stress- and viral infection-induced translational suppression due to their role in deciding the fate of sequestered mRNA by sorting it for re-initiation, storage, or degradation.⁵⁸⁶ One of the facts supporting the involvement of TIA1 in SG biogenesis is formation of SGs in cells infected with poliovirus.⁵⁸⁹ Such SGs, formed at early stages of infection, contain TIA1, translation initiation factors, RNA binding proteins, and mRNA.⁵⁸⁹ TIA1 shuttling from the nucleus in response to environmental stress, and subsequent aggregation of a C-terminal proteolytic TIA1 fragment containing glutamine-rich PrLD play a crucial role in the formation of cytoplasmic SGs,^{269,277,590,591} suggesting that the controlled prion-like aggregation of TIA1 might be related to the regulation of SG formation.^290,592 In fact, it was pointed out that TIA1 is one of a few key RBPs, such as TIAL1/TIAR, ZFP36, G3BP1, G3BP2, and FMR1, which play a crucial role in the initial stress-induced nucleation of SGs.^269,290,592 The central role of TIA1 and TIAL1 in SG biogenesis and related stress-induced inhibition of translation initiation is supported by the fact the stress-induced phosphorylation of translation initiation factor EIF2A is followed by the TIA1- and TIAL1-dependent recruitment of most cytoplasmic mRNAs to SGs.²⁷⁷ In this way, TIA1 and TIAL1 act as translational silencers needed for controlling the duration of stress-induced translational arrest.²⁷⁷

4.14. MAPT (microtubule associated protein tau)

4.14.1. Structural properties of MAPT

The MAPT gene located on chromosome 17q21.31 (at position 21.31 of the long [q] arm of chromosome 17) has 16 exons and the gene product is a cytosolic axonal protein that binds and stabilizes microtubules,^593,594 thereby serving as an important structural component of the microtubule-based transport system of the neurons.

Human MAPT (PMID: P10636) exists as a family of proteoforms migrating in SDS gel electrophoresis at close bands of 55–62 kDa. Alternative splicing of MAPT mRNA is (at least in part) responsible for this heterogeneity. In fact, although the MAPT gene has 16 exons, only 8 of them (1, 4, 5, 7, 9, 11, 12, and 13) are found in all isoforms of this protein and are therefore considered as constitutive, whereas the remaining exons (2, 3, 4A, 6, 8, 10, and 14) are subject to alternative splicing.⁵⁹⁵ Because exons 1 and 14 are parts of the promoter and the 3′ untranslated region of the MAPT mRNA, respectively, they are not found in any isoform of MAPT. None of the brain-specific MAPT isoforms have exons 4A, 6, and 8, which are found only in the MAPT mRNAs of the peripheral tissue.^596,597 The variability of the brain-specific MAPT isoforms is caused by the alternative splicing of exons 2, 3, and 10 (see ref. 597) that generates 6 mRNA isoforms^598,599 containing 2+3+10+, 2+3+10−, 2+3−10+, 2+3−10−, 2−3−10+, and 2−3−10− combinations of these exons and encoding 6 protein isoforms. Because exons 2 and 3 each encode 29-residue-long sequences located in the N-terminal part of MAPT (designated N inserts), and because exon 10 encodes 1 of the 4 31-residue-long microtubule-binding repeats located within the C-terminal part of MAPT and encoded by exons 9, 10, 11, and 12 (designated as R regions), alternative splicing generates MAPT variants that differ from each other by having either 0, 1, or 2 N-terminal inserts, and 3 or 4 C-terminally located R-regions (designated, respectively, as 2N4R, 2N3R, 1N4R, 1N3R, 0N4R, and 0N3R). The longest isoform in the CNS (441 amino acids total) has 4 repeats (R1, R2, R3, and R4) and 2 inserts, whereas the shortest isoform (352 amino acids total) has 3 repeats (R1, R3, and R4) and no insert.^600,601 Although the normal cortex is characterized by equal expression levels of the 3- and 4-repeat forms, the ratio of isoforms is changed in tauopathies.^602,603 Furthermore, in the peripheral nervous system and almost in all CNS neurons that extend processes into the peripheral nervous system, the big MAPT (with a molecular mass of ∼110 kDa) is expressed that contains an additional 254-residue insert in the amino-terminal half encoded by the additional large exon (4a) included in the corresponding mRNA.^604-606

Phosphorylation and other PTMs serve as an additional source of MAPT microheterogeneity.⁶⁰⁷ GSK3B, CDK5, and MAPK kinases are involved in phosphorylation of many residues of MAPT during brain development⁶⁰⁸ and multiple sites of MAPT are phosphorylated by several kinases in vitro (for a review, see ref. 609). The primary targets of the in vitro phosphorylation are residues within the microtubule-interacting region (repeat-containing domain) and its flanking regions. Many of these sites are also phosphorylated in paired helical filaments (which are the major form of MAPT deposition in AD).^610,611 In fact, 10 major phosphorylation sites have been identified in MAPT isolated from paired helical filaments from patients with AD.^610,611

4.14.2. MAPT function and dysfunction

Among functions ascribed to MAPT in vitro are binding to microtubules, regulation of their assembly, and control of their dynamic instability.^612-616 In situ, the axons have high levels of MAPT,⁶¹⁷ and brain cells in general are estimated to have high proportion of this protein, which might account for 0.025–0.25% of total soluble brain protein.^618,619 Although at normal conditions, the levels of the microtubule-unbound MAPT are kept low, the increase of soluble MAPT caused by the impairment of the protein-degradation system⁶²⁰ or aberrant phosphorylation⁶²¹ might result in aggregation of this protein.⁶²²

Aggregation of MAPT is associated with a broad spectrum of neurodegenerative diseases collectively known as tauopathies (e.g., AD, FTLD, ALS, progressive supranuclear palsy, cortico basal degeneration, Pick disease, etc.).⁶²³ FTLD is a group of complex diseases characterized by high heterogeneity of their pathological features. In fact, pathologically, FTLD variants are diversified and classified based on the major specific protein found in their inclusions, with FTLD–MAPT (which is an FTLD variant where MAPT serves as the major specific protein of the inclusions) being one of the most common pathological subtypes of this disease.⁶²⁴

Among the mechanisms connecting MAPT to ALS is cleavage of this protein and generation of the neurotoxic MAPT_45–230 fragment that is highly abundant in lumbar and cervical spinal cord specimens obtained from ALS subjects, being present in the ALS upper motor neurons located in the precentral gyrus, and forming aggregates in the spinal cord of ALS patients.⁶²⁵

4.14.3. MAPT stress granules, and other PMLOs

ALS and FTLD are tauopathies, which are a large group of neurodegenerative diseases associated with aggregation and dysfunction of MAPT. Curiously, the TIA1 protein, which is typically considered as a SG marker and participates in SG nucleation, was recently shown to be localized to the AD-related neurofibrillary tangles composed of the hyperphosphorylated and aggregated MAPT, with TIA1 being able to directly interact with MAPT.⁵⁸⁵ Although normal cytosolic MAPT only weakly interacts with TIA1, the abnormal hyperphosphorylation promotes this interaction.⁵⁸⁵ A dedicated study was recently conducted to check if, in its turn, MAPT can be associated with TIA1-containing SGs.⁶²⁶ This study revealed that when MAPT-expressing cells are treated with arsenite, a prominent SG response is observed and some formed SGs contain MAPT.⁶²⁶ Furthermore, when extracellular MAPT is internalized in the recipient cells (in line with the cell-to-cell propagation of the MAPT-related pathology model⁶²⁷), a prominent formation of SGs is observed, and MAPT is abundantly present in these cytoplasmic nonvesicular cellular bodies.⁶²⁶ Because in these studies, a pseudohyperphosphorylated MAPT-E14 variant carrying 14 phosphomimetic (serine/threonine to glutamate) mutations is more efficient in induction of SG formation and is able to more strongly interact with TIA1, it has been hypothesized that the hyperphosphorylation-induced oligomerization may be an important factor promoting SG recruitment of the internalized MAPT species.⁶²⁶ Based on these observations it was suggested that SGs and TIA1 might be important players in the cell-to-cell transmission of MAPT pathology.⁶²⁶

5. ALS- and FTLD-related proteins, PMLOs, and autophagy

In order to appropriately balance sources of energy at critical times in development and in response to nutrient stress, a cell can utilize a self-degradative process known as autophagy (self-eating) that is responsible for the disposal of various intracellular species.^103,628-634 Macroautophagy (hereafter autophagy) provides a way for elimination of aberrant protein aggregates, superfluous or damaged organelles, and even entire bacteria.⁶³⁵ Structures doomed for autophagy-based clearance are first sequestered by phagophores, the precursors to autophagosomes, which are large, double-membrane bound vesicles, and then delivered into the interior of the lysosome or vacuole, where they are digested by resident hydrolases.^635,636 When the autophagosome fuses with the lysosome, the autolysosome is formed, and the content of the autophagosome is catabolized.^637,638 It is important to remember that the formation of a new autophagosome does not cover all autophagy processes.²⁹⁸ For example, in a chaperone-mediated autophagy, individual protein substrates are directly targeted to the lysosome by chaperones and then translocated across the lysosomal membrane, whereas in microautophagy, invagination of the lysosomal membrane directly engulfs the cytoplasmic content.⁶³⁹ Because autophagy can sequester and degrade large protein complexes and entire organelles, it represents an important alternative to the ubiquitin-proteasome system.^298,638,640

In ALS, aberrant SGs cannot disassemble and therefore accumulate and are consequently degraded by autophagy.⁶⁴¹ There are several observations linking together ALS- and FTLD-related proteins, stress granules (and potentially other PMLOs), and autophagy. Furthermore, it is thought that there might be a special chaperone-mediated SG surveillance system, granulostasis, which might regulate composition and dynamics of SGs, thereby playing a role in ALS pathogenesis.^641,642 The first lines of evidence on the connection between autophagy and ALS have been derived from animal models.²⁹⁸ For example, an autophagy pathway was shown to be enhanced in a transgenic mouse model of SOD1-ALS,⁶⁴³ and also was observed in spinal cord tissues from SOD1-ALS patients.⁶⁴⁴ In line with the granulostasis hypothesis,⁶⁴² recruitment of chaperones prevents formation of aberrant SGs in cells expressing ALS-linked variants of SOD1, and promotes disassembly of such SGs when the stress subsides.⁶⁴⁵

TARDBP can also be involved in autophagy regulation, mostly by controlling biogenesis of autophagosomes and lysosomes.^646-649 For example, autophagosome-lysosome fusion is blocked by TARDBP loss of function.⁶⁴⁸ Also, TARDBP targets a key component of the MTORC1 complex, RPTOR/raptor (regulatory associated protein of MTOR complex 1), thereby regulating the MTORC1-TFEB (transcription factor EB) signaling pathway.⁶⁴⁹ Stability of RPTOR mRNA is decreased when TARDBP function is lost, leading to the enhancement of autophagosomal and lysosomal biogenesis in an MTORC1-dependent manner. Furthermore, loss of TARDBP is also associated with impaired autophagosome-lysosome fusion in an MTORC1-independent manner.⁶⁴⁹

Another study in this area was a work by Farg et al., who established that besides regulating endosomal trafficking, C9orf72, via its interaction with RAB proteins related to autophagy and endocytic transport, such as RAB1, RAB5, RAB7 and RAB11, can be involved in regulation of autophagy in neurons.⁴⁴² This idea is further supported by the colocalization of C9orf72 with UBQLN2 and autophagosome marker MAP1LC3B/LC3B (microtubule associated protein 1 light chain 3 beta)-positive vesicles, and comigration of this protein with lysosome-stained vesicles in neuronal cells.⁴⁴² Finally, a possible dysregulation of trafficking in ALS patients bearing the C9orf72 repeat expansion was postulated based on the immunohistochemical analysis of motor neurons of ALS patients that revealed an increased colocalization between C9orf72 and RAB7 and RAB11 compared with controls.⁴⁴² Recently, it was shown that the loss of C9orf72 might cause deregulation of autophagy via reduction of the activity of the negative regulator of autophagy MTOR, and the increase of the levels and nuclear translocation of TFEB, indicating that C9orf72 might serve as a negative regulator of autophagy, playing an important role in coupling the cellular metabolic state with autophagy regulation.⁶⁵⁰

An analysis of the role of autophagy in regulation of FUS-positive SGs was conducted in the same year by Ruy et al.⁶⁵¹ Here, it was shown that the FUS-positive SGs formed under oxidative stress in neurons bearing the ALS-linked FUS^R521C mutation are colocalized to the zinc finger protein TRIM27 (tripartite motif containing 27) RFP-LC3-positive autophagosomes and are preferentially accumulated in autophagy-deficient neurons,⁶⁵¹ whereas rapamycin-enhanced autophagy is able to reduce the accumulation of FUS-positive SGs in an autophagy-dependent manner.^300,651 It was also indicated that autophagy activation not only reduces mutant FUS-positive SG accumulation, but also inhibits neurite fragmentation mediated by mutant FUS, and reduces cell death in response to oxidative stress in disease-associated cultured neurons.³⁰⁰ Therefore, the enhancement of the autophagy activity leading to the reduction in the number of SGs containing mutant FUS can be neuroprotective,^651,652 whereas distortion of the autophagic process can contribute to the ALS-FTLD pathogenesis.⁶⁵³

6. Conjointness of the LLPT-based pathogenicity: Collective responsibility of proteins involved in PMLO formation

Because any PMLO (physiological or aberrant) represents a proteinaceous assemblage containing multiple proteins, LLPT-based pathologies represent a collective responsibility of proteins involved in PMLO formation. In essence, such kind of pathogenicity driven by aberrant PMLOs represents an example of a misdeed conducted by good guys assembled with bad company in the wrong place at the wrong time. Several examples below show how ALS-related proteins might cooperate, while affecting dynamics of PMLOs.

Although in the norm, TARDBP and FUS are preferentially localized in the nucleus, being transported there via the nuclear import receptors in a RAN (RAN, member RAS oncogene family) GTPase-dependent manner,⁴¹³ they also play a role in regulation of the RNA life cycle, being involved in the normal biogenesis of the intracytoplasmic RNA-containing PMLOs (e.g., SGs) in addition to regulation of the transcription, processing, transport, and stability of RNA.³⁹² In ALS and FTLD, both of these proteins mislocalize to the cytoplasm, where they form cytoplasmic inclusions, affect SG dynamics, and impair the RNA quality control system.²⁹³ For example, the rate of TARDBP incorporation into the SGs is accelerated by the ALS-related mutations in this protein, and the mutated TARDBP forms larger SGs than the wild-type protein.³⁸⁹ Similarly, FUS with the ALS-associated mutations efficiently localizes to SGs causing their assembly into larger structures.⁵⁰⁵ However, TARDBP and FUS do not act alone, being engaged in interaction with each other, and this interaction is enhanced by the ALS-related mutations in TARDBP.⁶⁵⁴ Furthermore, ATXN2 with intermediate-length polyglutamine expansions that serves as an ALS susceptibility factor^570-572 and is an SG constituent RNA-binding protein, is associated with both FUS-⁵⁸⁰ and TARDBP-based pathologies.⁵⁷⁷ For example, longer polyglutamine expansions in ATXN2 enhance interaction of this protein with TARDBP and increase the efficiency of the TARDBP cytoplasmic mislocalization.⁵⁷⁰ These observations suggest that the ALS-FTLD pathology can be at least in part driven by the mislocalization of these 3 RNA-binding proteins to the cytoplasm that results in triggering the aberrant LLPTs leading to the formation of abnormal cytoplasmic PMLOs and promoting impairment of the RNA quality control system.^293,578

TARDBP, FUS, and ATXN2 that interact with each other, potentially forming the core of the ALS-FTLD pathological cascade⁵⁷⁸ are not exceptions to the rule, and other proteins associated with ALS-FTLD pathogenicity may also contribute to the aforementioned PMLO-related collective responsibility. For example, among the known C9orf72 targets are some proteins related to ALS, such as actin, HNRNPA1, HNRNPA2B1, and UBQLN2, and overexpression of this protein is accompanied by the formation of both nuclear C9orf72 aggregates and prominent cytoplasmic SGs containing HNRNPA1 and HNRNPA2B1.⁴⁴² It was also indicated that HNRNPA1 and HNRNPA2B1 can interact with TARDBP via their PrLDs.^364,367,552 Mutations of these proteins trigger their mislocalization from the nucleus to the cytoplasm, promote their aggregation and formation of sarcoplasmic inclusions, and cause multisystem proteinopathy and ALS.⁵⁵² Curiously, pathological mutations in HNRNPA1 and HNRNPA2B1 also stimulate mislocalization of wild-type TARDBP from the nucleus to the cytoplasm and result in formation of TARDBP-positive cytoplasmic inclusions. Typically, the inclusions formed by TARDBP, HNRNPA2B1, and HNRNPA1 are physically separated and only occasionally overlap.⁵⁵¹ Similarly, pathological mutations in VCP promote formation of cytoplasmic inclusions by the wild type versions of TARDBP, HNRNPA2B1, and HNRNPA1,⁵⁵¹ suggesting that the same phenotypic changes leading to TARDBP pathology and disease can be promoted by perturbations of the VCP, HNRNPA2B1, and HNRNPA1 pathways.⁵⁵² This important phenomenon is illustrated by Fig. 8 that represents the results of the analysis of the effect of mutant forms of VCP (A232E and R155H) on the formation of SGs and TARDBP recruitment to those SGs.⁶⁵⁵ It is seen that the constitutive SGs containing TARDBP are formed only as a result of the overexpression of the mutated VCP, whereas no such effects were found when the wild-type VCP was overexpressed in HeLa cells.⁶⁵⁵

Figure 8. Effect of mutated forms of VCP (A232E and R155H) on the formation of constitutive SGs and the recruitment of TARDBP to those SGs. HeLa cells were transfected with plasmid expressing wild-type or mutant VCP-GFP forms and stained for the SG marker EIF3B and TARDBP. Overexpression of mutant but not wild-type VCP resulted in the formation of TARDBP-containing SGs. Scale bar: 10 µm. Reproduced with permission from ref. 655.

PURA can interact with the C-terminal region of the mutated FUS in an RNA-dependent manner, and these 2 proteins are colocalized in the SGs present in ALS patients.⁵⁰⁵ Furthermore, accumulation of PURA within the intra-cytosolic SGs is enhanced in the presence of the (GGGGCC)₃₁ repeats⁴⁵⁷ due to the ability of this protein to interact with the GGGGCC expansions of the C9orf72 gene.⁵⁰⁸ Interaction of FLCN with TARDBP can modulate shuttling of TARDBP between the nucleus and the cytoplasm,⁵²¹ thereby playing a role in regulation of SG dynamics. RBM45 shows some sequence similarity to TARDBP and FUS and forms both nuclear and cytoplasmic inclusions.⁵²² Furthermore, the homo-oligomeric form of this protein found in the cytoplasm can interact with both TARDBP and FUS, and accumulation of RBM45 in the cytoplasm leads to the formation of the TARDBP-positive SGs.⁵²² SS18L1 can both self-aggregate and interact with FUS.⁵³³ Curiously, this protein serves as an illustration of the hypothesis that the ALS-FTLD pathology can be caused by misbehaving PMLOs other than SGs (e.g., compromised paraspeckle formation^544,545). In fact, the dynamics, morphology, and integrity of paraspeckles can all be affected by aggregated SS18L1 due to the ability of this species to entrap other paraspeckle proteins.⁵³³

7. Concluding remarks: Linking PMLOs with neurodegeneration via protein intrinsic disorder

Observations presented in the previous sections clearly indicate that many proteins considered in this article and related to ALS and FTLD are either directly or indirectly involved in the LLPTs and formation of SGs and other PMLOs. Furthermore, being purified and analyzed in vitro, some of these proteins show high aggregation potential and the ability to undergo liquid-liquid demixing, leading to the formation of phase-separated droplets⁶⁵⁶ or polymer hydrogels.⁶⁵⁷ It is assumed that these membrane-less cellular compartments represent liquid droplets that phase separately from the cytoplasm and stably coexist with their environments.⁶⁵⁸ Although typically LLPTs leading to the formation and disintegration of PMLOs represent normal physiological processes related to the organization of macromolecules into compartments needed for the proper functioning of a cell, abnormal liquid-liquid phase transitions and aberrant PMLOs can be associated with age-related neurodegenerative diseases,⁶⁵⁸ via impairment of the cytoplasmic RNA quality control system²⁹³ or some other related mechanisms.

It was also pointed out that LLPTs are often driven by IDPs containing low-complexity regions or domains.^{205,206,226,269,271,657,659–666} In agreement with these earlier observations Fig. 3 shows that the majority of the proteins considered in this article that are related to ALS and FTLD contain long IDPRs. Although C9orf72 is the most ordered protein in this set, the pathological involvement of this protein to neurodegeneration is related to the hexanucleotide repeat expansions of the intronic region of the C9orf72 gene leading to the bidirectional transcription of the expanded repeats associated with the generation of various sense and antisense repeat RNAs^433-436 that serve as mRNAs in RAN-translation to generate a series of DPRs.^434,436-439 All these DPRs including poly(GA), poly(GR), poly(GP), poly(PA), and poly(PR) proteins are highly disordered. Furthermore, many of the ALS- and FTLD-related proteins are able to interact with nucleic acids and many protein partners, and many of them are involved in SG formation, and are associated with the assembly of SG-like inclusions in ALS and FTLD. For example, in addition to being involved in the formation of SGs, TARDBP and FUS are associated with several other PMLOs, such as DNA damage repair sites, paraspeckles, other RNA granules, and transport granules.^{299,402,664,667,668}

In other words, intrinsic disorder might represent a feature unifying very different proteins in their relation to the pathogenesis of neurodegeneration via aberrant LLPTs and abnormal PMLOs. Therefore, it is likely that the “Dr. Jekyll–Mr. Hyde” behavior of proteins related to ALS and FTLD pathogenesis can at least in part be due to the ability of these proteins to be engaged in the intrinsic disorder-based LLPTs.

Abbreviations

APBB1/FE65	=	amyloid beta precursor protein binding family B member 1
ACTA2	=	actin, alpha 2, smooth muscle, aorta
ACTB	=	actin beta
AD	=	Alzheimer disease
AKAP9/AKAP350	=	A-kinase anchor protein 9
ALS	=	amyotrophic lateral sclerosis
AMPA	=	α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid
AMPK	=	AMP-activated protein kinase
ANG	=	angiogenin
APC	=	adenomatous polyposis coli
APOE	=	apolipoprotein E
ARF	=	ADP ribosylation factor
ARID1A/BAF250	=	AT-rich interaction domain 1A
AS	=	alternative splicing
ATN1	=	atrophin 1
ATXN	=	ataxin
ATXN1	=	ataxin 1
ATXN2	=	ataxin 2
ATXN3	=	ataxin 3
ATXN7	=	ataxin 7
AXIN1	=	axin 1
BCL2	=	BCL2, apoptosis regulator
BDNF	=	brain derived neurotrophic factor
BHD	=	Birt-Hogg-Dubé syndrome
BRCA1	=	BRCA1, DNA repair associated
C9orf72	=	chromosome 9 open reading frame 72
CASC3/MLN51	=	cancer susceptibility 3
CCNA1	=	cyclin A1
CCNT1	=	cyclin T1
CCS	=	copper chaperone for superoxide dismutase
CDK	=	cyclin dependent kinase
CDKN1A/p21	=	cyclin dependent kinase inhibitor 1A
CDKN1B/p27Kip1	=	cyclin dependent kinase inhibitor 1B
CHCHD10	=	coiled-coil-helix-coiled-coil-helix domain containing 10
CHMP2B	=	charged multivesicular body protein 2B
CHRNB4	=	cholinergic receptor nicotinic beta 4 subunit
CFTR	=	cystic fibrosis transmembrane conductance regulator
CHGA	=	chromogranin A
CHGB	=	chromogranin B
CNS	=	central nervous system
CPEB	=	cytoplasmic polyadenylation element binding protein
CREB	=	cAMP response element binding
CREBBP/CBP	=	CREB binding protein
CTA	=	cancer/testis antigen
CTF	=	C-terminal fragment
DDX6	=	DEAD-box protein 6
DDX19B/DBP5	=	DEAD-box helicase 19B
DEAD box	=	protein with Walker B motif with the amino acid sequence D-E-A-D
Ded1	=	ATP-dependent RNA helicase ded1
DENND1B/connecdenn 2	=	DENN domain containing 1B
DERL1	=	derlin 1
DMD	=	discrete molecular dynamics
DPR	=	dipeptide repeat proteins
E2F1	=	E2 transcription factor 1
EIF2	=	eukaryotic translation initiation factor 2
EIF2A	=	eukaryotic translation initiation factor 2A
EIF3	=	eukaryotic translation initiation factor 3
EIF4B	=	eukaryotic translation initiation factor 4B
EIF4G1/EIF4F	=	eukaryotic translation initiation factor 4 gamma 1
EP300	=	E1A binding protein p300
ER	=	endoplasmic reticulum
ERN1/IRE1	=	endoplasmic reticulum to nucleus signaling 1
EWSR1	=	EWS RNA binding protein 1
fALS	=	familial form of ALS
F2	=	coagulation factor II, thrombin
FAS	=	Fas cell surface death receptor
FASTK	=	Fas activated serine/threonine kinase
FET	=	protein family including FUS, EWSR1, and TAF15
FLCN	=	folliculin
FMR1/FMRP	=	fragile X mental retardation 1
FNIP	=	folluculin interacting protein
FTLD	=	frontotemporal lobar degeneration
FUS	=	FUS RNA binding protein
FXR1	=	FMR1 autosomal homolog 1
G3BP1	=	G3BP stress granule assembly factor 1
G3BP2	=	G3BP stress granule assembly factor 2
GLE1	=	GLE1, RNA export mediator
GLFG	=	G-L-F-G repeat region of nucleoporin
GRN/PGRN	=	granulin precursor
GSK3	=	glycogen synthase kinase 3
HDAC	=	histone deacetylase
HNE	=	4-hydroxynonenal
HOA	=	homo-oligomer assembly domain
HNRNP	=	heterogeneous nuclear ribonucleoprotein
HNRNPA1	=	heterogeneous nuclear ribonucleoprotein A1
HNRNPA2B1	=	heterogeneous nuclear ribonucleoprotein A2/B1
HPV	=	human papillomavirus
HTT	=	huntingtin
IAPP/amylin	=	islet amyloid polypeptide
IDP	=	intrinsically disordered protein
IDPR	=	intrinsically disordered protein region
KEAP1	=	kelch like ECH associated protein 1
LCCS1	=	lethal congenital contracture syndrome 1
LAAHD	=	lethal arthrogryposis with anterior horn cell disease
LIMK	=	LIM domain kinase
LLPT	=	liquid-liquid phase transition
MAP1LC3B	=	microtubule associated protein 1 light chain 3 beta
MAP2K7/MKK7	=	mitogen-activated protein kinase kinase 7
MAPK3/ERK1	=	mitogen-activated protein kinase 3
MAPK8/JNK1	=	mitogen-activated protein kinase 8
MAPT	=	microtubule associated protein tau
MBP	=	myelin basic protein
MFD	=	multifunctional domain
MoRF	=	molecular recognition feature
mtSOD1	=	mutant SOD1
MTORC1	=	mechanistic target of rapamycin complex 1
NEFH	=	neurofilament heavy
NEFL/NFL	=	neurofilament light
NEK1	=	NimA related kinase 1
NES	=	nuclear export signal
NimA	=	never in mitosis A
NLS	=	nuclear localization signal
NPC	=	nuclear pore complex
NPM1	=	nucleophosmin 1
NFE2L2	=	nuclear factor, erythroid 2 like 2
NTD	=	N-terminal domain
NUP155	=	nucleoporin 155
NUPL2	=	nucleoporin like 2
OPTN	=	optineurin
OS	=	oxidative stress
PABPC1/PABP1	=	poly(A) binding protein cytoplasmic 1
P-body	=	processing body
PBP	=	progressive bulbar palsy
PD	=	Parkinson disease
PFN1	=	profilin 1
pfy1∆	=	yeast cells lacking profilin
PLS	=	primary lateral sclerosis
PMA	=	progressive muscular atrophy
PMLO	=	proteinaceous membrane-less organelle
PPI	=	protein-protein interaction
PrLD	=	prion-like domain
PRNP/PrP	=	prion protein
PRMT1	=	protein arginine methyltransferase 1
PTEN	=	phosphatase and tensin homolog
PTM	=	posttranslational modification
PURA	=	purine rich element binding protein A
PV	=	poliovirus
RAB	=	RAB, member RAS oncogene family
RACK1	=	receptor for activated C kinase 1
RAD23B/HR23B	=	RAD23 homolog B, nucleotide excision repair protein
RAN	=	RAN, member RAS oncogene family
RAN-translation	=	repeat-associated non-ATG translation
RB1	=	RB transcriptional corepressor 1
RBM45	=	RNA binding motif protein 45
RBP	=	RNA binding protein
RHOA	=	ras homolog family member A
RNP	=	ribonucleoprotein
Rnq1	=	[PIN+] prion protein RNQ1
ROCK1	=	Rho associated coiled-coil containing protein kinase 1
ROS	=	reactive oxygen species
RPTOR/Raptor	=	regulatory associated protein of MTOR complex 1
RRM	=	RNA-recognition motif
RPS6KA3/RSK2	=	ribosomal protein S6 kinase A3
sALS	=	sporadic form of ALS
SCA2	=	spinocerebellar ataxia 2
SG	=	stress granule
SIRT	=	sirtuin
SLC1A2/EAAT2	=	solute carrier family 1 member 2
SMARCA4/BRG1	=	SWI/SNF related, matrix associated, actin dependent regulator of chromatin, subfamily a, member 4
SMCR8	=	Smith-Magenis syndrome chromosome region, candidate 8
SMN	=	survival motor neuron protein
SOD1	=	superoxide dismutase 1
SP1	=	Sp1 transcription factor
SPN/CD43	=	sialophorin
SQSTM1/p62	=	sequestosome 1
SS18L1/CREST	=	SS18L1, nBAF chromatin remodeling complex subunit
Sup35	=	eukaryotic peptide chain release factor GTP-binding subunit
TAF15	=	TATA-box binding protein associated factor 15
TANK	=	TRAF family member associated NFKB activator
TARDBP/TDP-43	=	TAR DNA binding protein
TBK1	=	TANK binding kinase 1
TBP	=	TATA-box binding protein
TFEB	=	transcription factor EB
TIA1	=	TIA1 cytotoxic granule associated RNA binding protein
TIAL1/TIAR	=	TIA1 cytotoxic granule associated RNA binding protein like 1
TP53	=	tumor protein p53
TP53BP/ASPP	=	tumor protein p53 binding protein
TRAF2	=	TNF receptor associated factor 2
TRIM27	=	tripartite motif containing 27
UBC	=	ubiquitin C
UBQLN2	=	ubiquilin 2
UNC119/RG4	=	unc-119 lipid binding chaperone
UPS	=	ubiquitin-proteasome system
VCP	=	valosin containing protein
VDAC1	=	voltage dependent anion channel 1
VEGF	=	vascular endothelial growth factor
WDR41	=	WD repeat domain 41
WDR62	=	WD repeat domain 62
WT-SOD1	=	wild-type SOD1
YBX1/YB1	=	Y-box binding protein 1
YWHAB	=	tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein beta
ZFP36/TTP	=	ZFP36 ring finger protein

Conflicts of interest

The author declares no conflict of interest.

Acknowledgments

I am thankful to Alexey Uversky and Prof. Daniel Klionsky for careful reading and editing the manuscript of this article.

Funding

This work was supported in part by a grant from the ALS Association to V.N.U. (grant number IIP-265).