The coming-of-age of nucleocytoplasmic transport in motor neuron disease and neurodegeneration

Abstract

The nuclear pore is the gatekeeper of nucleocytoplasmic transport and signaling through which a vast flux of information is continuously exchanged between the nuclear and cytoplasmic compartments to maintain cellular homeostasis. A unifying and organizing principle has recently emerged that cements the notion that several forms of amyotrophic lateral sclerosis (ALS), and growing number of other neurodegenerative diseases, co-opt the dysregulation of nucleocytoplasmic transport and that this impairment is a pathogenic driver of neurodegeneration. The understanding of shared pathomechanisms that underpin neurodegenerative diseases with impairments in nucleocytoplasmic transport and how these interface with current concepts of nucleocytoplasmic transport is bound to illuminate this fundamental biological process in a yet more physiological context. Here, I summarize unresolved questions and evidence and extend basic and critical concepts and challenges of nucleocytoplasmic transport and its role in the pathogenesis of neurodegenerative diseases, such as ALS. These principles will help to appreciate the roles of nucleocytoplasmic transport in the pathogenesis of ALS and other neurodegenerative diseases, and generate a framework for new ideas of the susceptibility of motoneurons, and possibly other neurons, to degeneration by dysregulation of nucleocytoplasmic transport.

Introduction

Amyotrophic lateral sclerosis (a.k.a. Lou Gehrig’s disease) as first coined and described by Jean-Martin Charcot in the mid-nineteenth century for its neuropathology [1, 2] is a neurodegenerative disease of motoneurons that has typically bulbar or spinal onsets owing to the respective dysfunction and loss of upper corticospinal or lower spinal cord somatic motoneurons [3–5]. ALS is fatal and with a prognosis of life expectancy between 1.5 and 5 years after diagnosis [3, 4]. Notably, some motoneurons, such as ocular-motoneurons, are spared in ALS for reasons that are elusive [6–8]. ALS is a genetically heterogeneous disease with sporadic (sALS) and familial origins (fALS) [9, 10]. sALS in aggregate comprises ~ 90% of ALS, and the genetic and environmental causes of sALS remain largely unknown. fALS is caused by ~ 20 genes [9, 10], but genotype–phenotype correlations also support allelic heterogeneity between ALS, ALS syndromes and other forms of motoneuron diseases [4, 11, 12]. For example, mutations in the motor domain of the microtubule-based motor protein isoform of kinesin-1, KIF5A, cause hereditary spastic paraplegia and Charcot–Marie–Tooth disease type 2 [13–15]. In contrast, mutations affecting the cargo-binding tail domain of KIF5A cause neonatal intractable myoclonus or ALS [16–19]. Mutations in Cu/Zn superoxide dismutase (SOD1), chromosome 9 open reading frame 72 (C9ORF72), TAR DNA-binding protein (TARDBP), which encodes the TDP-43 protein, and fused in sarcoma (FUS) cause the most common forms of fALS [5, 9]. The dominant inheritance of fALS has supported the notion that ALS mutations promote gain-of-function and neurotoxicity of ALS gene products [10]. To date, there are no effective treatments to ALS [20].

Current challenges in ALS

In spite of significant advances in the identification of ALS genes and overt pathological hallmarks shared by several forms of ALS that can be heralded during the past decade, several fundamental questions about ALS and its pathogenesis remain unresolved. These outstanding issues have hampered therapeutic developments against ALS.

First, ALS genes have ubiquitous expression and diverse functions [9]. This genetic heterogeneity and pleiotropy complicate the understanding of the molecular pathogenesis of ALS. For example, several ALS genes are expressed in motoneurons and glial cells, such as astrocytes and microglia, where ALS genes play an important role in disease onset and progression [20–25]. Second, it remains obscure why motoneurons are preordained to dysfunction and degeneration by mutations in ALS genes that are ubiquitously expressed as well as by potential environmental insults. Likewise, it is obscure why fast and large motoneurons are most vulnerable to dysfunction and degeneration, while small and slow motoneurons are the most resistant to degeneration at the end-stage of disease [26, 27]. Third, pathological hallmarks shared by several forms of ALS have been uncovered [5], but the pathophysiological roles of many pathological markers are incomplete or obscure. For example, TDP-43 inclusions are present in 97% of ALS; however, TDP-43 pathology is not pathognomonic for ALS, because TDP-43 inclusions are also present in several other neurological diseases and even in some healthy brains of the elderly [5]. The concept that intracellular inclusions, which are hallmark to ALS and many other neurodegenerative diseases, are neurotoxic has been challenged by the notion that protein nucleation processes leading to protein aggregation subsumes a series of detoxifying responses by neurons against the detrimental effects of soluble misfolded protein species [28–32]. Further, several ALS mouse models of C9ORF72 recapitulate ALS pathologies without untoward pathophysiological effects, such as ALS motor behaviors [33–35], while other C9ORF72 models develop pathologies and motor deficits linked to ALS [36, 37]. Hence, some ALS pathologies appear to blur the lines of phenotypic demarcation of ALS. Finally, the genetic heterogeneity of ALS and its rapid progression complicates the development of surrogate and predictive pathophysiological measures of the onset and progression of ALS [38–40]. Isolating motor and non-motor endophenotypes that offer pathognomonic signs of sALS and fALS during the preclinical stage of the disease is critical to neurodegenerative diseases with rampant progression, such as ALS. Answers to these complex and challenging but fundamental questions will likely lead to unifying biological and mechanistic principles underpinning ALS and motoneuron biology, and possibly other neurodegenerative diseases.

An emerging and unifying principle in ALS

In spite of the aforementioned and unresolved questions surrounding ALS, a unifying pathobiological and organizing principle has recently emerged that cements the notion that several forms of ALS, and other neurodegenerative diseases, co-opt the dysregulation of nucleocytoplasmic transport [41–55]. Embedded in this principle is the notion that mutations in several ALS genes lead to aberrant nucleocytoplasmic partition of ALS-causing gene products and that this impairment subsumes the subcellular mislocalization and genesis of cytoplasmic inclusions of ALS gene products and accessory factors. Before expanding on some recent findings that lend support to this principle, these developments are encumbered with their own intricate constrains and complexities that forces consideration of basic and current principles underpinning nucleocytoplasmic transport. Some of these principles have been overlooked in past reviews and models attempting to explain impairments in nucleocytoplasmic transport caused by mutations in ALS genes. These principles will help to appreciate the roles and intricacies of nucleocytoplasmic transport in the pathogenesis of ALS and other neurodegenerative diseases, and in the susceptibility of motoneurons to degeneration by impairments in nucleocytoplasmic transport.

Principles of nucleocytoplasmic transport

Nuclear-cytoplasmic gradient of nucleotide-bound Ran GTPase

Ran GTPase is a small Ras-related nuclear protein, which is highly abundant and conserved between yeast and humans [56]. Although Ran GTPase alone harbors very low intrinsic rates of GTPase activity, Ran GTPase switches between GTP and GDP-bound conformational states in the nuclear and cytosolic compartments, respectively [57–60]. The asymmetric nucleocytoplasmic distribution of GTP and GDP-bound states of Ran GTPase is attained by the distinct subcellular localization of two critical regulators of Ran GTPase (Fig. 1). The nuclear localization of the chromatin-associated guanine nucleotide exchange factor (GEF), called the regulator of chromosome condensation 1 (RCC1), stimulates the GDP to GTP exchange of Ran GTPase by 100,000-fold in the nucleus [61–64]. By contrast, the cytoplasmic localization of the Ran GTPase-activating protein-1 (RanGAP1) stimulates the hydrolysis of Ran-GTP by 100,000-fold [64–67]. The resulting asymmetric gradient of Ran-GTP and Ran-GDP between the nucleus and cytoplasm is critical to impart unidirectional transport to Ran nucleotide-bound ensembles between the nuclear and cytoplasmic compartments [57–60, 68–70]. Ran-GDP is imported from the cytoplasm to the nucleus by the accessory factor, nuclear transport factor 2 (NTF2) [71, 72]. Nuclear trafficking occurs through nuclear pore, the gatekeeper of nucleocytoplasmic transport [73], in an energy-independent manner [74–77]. The nucleocytoplasmic shuttling of molecules < 40 kDa or less than 5–10 nm of diameter occurs by passive diffusion, whereas those of greater masses proceed by facilitated translocation in a manner which depends on nuclear transport receptors and interactions with nucleoporins, the components of the nuclear pore complex [78–85]. Importantly, passive and facilitated transports across the nuclear pore are not dynamically coupled [86–88]. It is estimated that a single nuclear pore has a mass flow of 100 MDa/s, translocation rates of 10³ s⁻¹ and a translocation speed through the central channel of the pore of ~ 0.5 µm s⁻¹ [82].

Fig. 1.

The Ran GTPase cycle. The nucleocytoplasmic gradient of the guanine-nucleotide-bound state of Ran GTPase is determined by the asymmetric localizations of the chromatin-associated guanine nucleotide exchange factor (GEF), regulator of chromosome condensation 1 (RCC1), in the nucleus, and of Ran GTPase-activating protein-1 (RanGAP1), in the cytoplasm. RCC1 stimulates the GDP to GTP exchange of Ran GTPase, whereas RanGAP promotes the hydrolysis of Ran-GTP. In vertebrates, RanGAP is also post-translationally modified by SUMOylation (SUMO-RanGAP) and recruited to Ranbp2, which localizes at cytoplasmic filaments of the nuclear pore and captures Ran-GTP-bound cargoes exiting the nuclear pore. The nuclear transport factor 2 (NTF2) binds to and transports Ran-GDP from the cytoplasm to the nucleus

Nuclear transport receptors (NTRs)

NTRs in the human comprise 20 structurally related members of the importin-β/karyopherin-β-related nuclear transport receptors. They consist of 8 nuclear export receptors (exportins) and 10 nuclear import receptors (importins) depending on the directionality of cargoes that they transport, but few exceptions to this rule exist for at least three receptors with bidirectional properties (e.g., exportin-4, exportin-7/Xpo7 and importin-13) [89–91]. Exportins have broad (e.g., exportin-1/Xpo1/CRM1) [92] or high substrate specificity toward cargoes (e.g., exportin-2/CAS/Xpo2) [93] exported from the nucleus, whereas importins mediate the nuclear import of cargoes from the cytoplasm (Fig. 2). These NTRs are characterized by a conserved Ran-GTP-binding motif at the N-terminal region, whereas their C-terminal regions are diverse and are implicated in substrate recognition [94, 95]. Several mammalian NTRs lack orthologues in lower organisms, such as yeast, and apparently, they evolved late in evolution for the recognition of selective and species-specific substrates. Importantly, importin-β-like receptors protect Ran-GTP from hydrolysis by RanGAP1 alone [96, 97]. This is an important feature of these receptors and it critically distinguishes them from two other structural and functional unrelated and high-affinity Ran-GTP-binding proteins, Ran-binding protein 1 (Ranbp1) and Ranbp2 (a.k.a Nup358; discussed next section). As a side note, it is important to note that some importin-β/karyopherin-β-related nuclear transport receptors were confusingly termed also Ranbp_n (e.g., Ranbp16/exportin-7) [98, 99], but these are structurally and functionally unrelated to Ranbp1 and Ranbp2. Importins display high affinity (low nM) towards Ran-GTP [100–102], whereas exportins in the absence of a nuclear export substrate have intermediate affinities for Ran-GTP in the micromolar range [93, 103–108]. Exportin and importin family members present different specificities toward the nuclear export and import of cargoes. For example, exportin-1 (a.k.a. CRM1/Xpo1) recognizes cargoes with the nuclear export sequence (NES) that comprise a leucine-rich export signal [92, 103, 104, 109–112]. Importin-β1 imports cargoes by binding these directly [113, 114] or indirectly via one of the seven importin-α isoforms of adaptors that recognizes cargoes with the classical nuclear localization sequence (NLS) [115–118], or by heterodimerization with importin 7 (e.g., histone H1) [119]. Importin-β2 (a.k.a., transportin-1) and importin 3 (a.k.a. transportin-2) recognize cargoes with proline–tyrosine (PY)-NLS [120, 121]. The structural context of recognition of signal sequences in cargoes is regulated at multiple levels, such as primary and secondary structural and functional contexts, formation of high-order ensembles and post-translational modifications [51, 90, 122–127]. These and other features may play important roles in the regulation of transport of cargoes. Finally, the nuclear import and export rates of model NES and NLS-substrates have similar flux rates (~ 56 molecules/s/nuclear pore, assuming ~ 2000 nuclear pores/nucleus), and translocation efficiencies (18 and 11 ms/event for nuclear export and import, respectively) [82, 88, 128]. These rates were close to the transport rates observed independently for other substrates, such as importin-β and NTF2-Ran-GDP complex (~ 5 ms) [129, 130]. Together, these results indicate that the gating properties of the nuclear pore (and likely the interactions between cargoes and NTRs) determine the dynamics of transport across the nuclear pore [131, 132].

Fig. 2.

Nuclear export and import of protein substrates. Nuclear export (left) of proteins is mediated by nuclear export receptors, exportins, such as exportin-1 (Exp1), which recognizes cargoes containing nuclear export sequences (NES). Cooperative association of exportin-1, Ran-GTP and NES-containing substrates promotes the nuclear export of nuclear cargo ensembles (1). Disassembly of nuclear export ensembles occurs when they encounter Ran-binding protein 2 (Ranbp2; a.k.a. Nup358) (2), which forms the cytoplasmic filaments of the nuclear pore, or Ranbp1 in the cytosol (3). Ranbp2 has a domain with zinc-finger motifs (ZnF), which binds exportin-1, and four Ran-GTP-binding domains (RBDs; only a single domain is represented). Ranbp1 has a single Ran-GTP-binding domain. The RBDs of Ranp2 and Ranbp1 destabilize Ran-GTP from its cargo ensemble and promotes the hydrolysis of Ran-GTP by RanGAP in the cytosol or by SUMOylated RanGAP (SUMO-RanGAP) at Ranbp2. Finally, free exportin-1 (Exp1) diffuses back to the nuclear compartment (4). Nuclear import (right) of proteins is mediated by the nuclear import receptors, importins, such as importin-β (Impβ), or its adaptors, such as importin-α (Impα), which recognizes cargoes containing nuclear localization sequences (NLS) (1). Nuclear import ensembles are transported to the nucleus (2), where they are disassembled upon encountering a high concentration of Ran-GTP, which binds to importin-β (Impβ) (3). Importin-β (Impβ)-bound to Ran-GTP is recycled back to the cytoplasm, where it is released from Ran-GTP by the RBDs of Ranbp2 (4) or Ranbp1 (5). RanGAP in the cytosol and SUMOylated RanGAP (SUMO-RanGAP) at Ranbp2 promote the hydrolysis of Ran-GTP

Ran GTPase effectors

Importin-β-like receptors exit the nuclear pore bound to Ran-GTP. The release of Ran-GTP from importin-β enables its binding to substrates destined for nuclear import, whereas the release of Ran-GTP from exportin-1 bound to NES-containing substrates unloads these cargoes from exportin-1. Two cytosolic proteins that were introduced earlier, Ranbp1 and Ranbp2, play central roles in displacing Ran-GTP from importin-β-like receptors after they exit the nuclear pore [97, 133, 134]. Ranbp1 is a single domain 26 kDa protein, which is highly conserved from yeast to man (but it is absent in D. melanogaster and C. elegans) [135–137]. In contrast, Ranbp2 is a large 358 kDa and multi-domain protein [138–141], which comprises the cytoplasmic filaments emanating from the nuclear pore [139, 140, 142]. Ranbp2 is absent in yeast and it is well conserved in mammals and some vertebrates, but not in other metazoans [143, 144]. Ranbp2 contains four highly conserved Ran-GTP-binding domains (RBDs) that are highly homologous to Ranbp1 and they are interspersed between other structurally and functionally unrelated domains. Among these domains is a domain with several zinc-finger motifs (ZnF) that binds specifically exportin-1 in a Zn²⁺-dependent manner [145]. The displacement of Ran-GTP from importin-β-like receptors by Ranbp1 and the RBDs of Ranbp2 in the cytosol enables Ran-GTP hydrolysis by RanGAP, which renders the disassembly reaction irreversible (Fig. 2) [97, 133, 134, 146, 147]. In vertebrates, RanGAP is SUMOylated and this modification causes the recruitment of SUMOylated Ran GAP to Ranbp2 at the cytoplasmic filaments of the nuclear pore (Fig. 2) [148–151]. Hence, Ranbp1 in the cytosol and the RBDs of Ranbp2 at the cytoplasmic face of the nuclear pore act as effectors of Ran GTPase by promoting the disassembly importin-β-like receptors from Ran-GTP. Finally, another protein, Ranbp3, has been described as a Ran-binding protein. The terminology used for Ranbp3 is misleading however, because Ranbp3 has poor homology (< 25%) to Ranbp1 and RBDs of Ranbp2, Ranbp3 has very weak binding activity for Ran-GTP (K_d ~ 10 µM) and Ranbp3 is best known to act as a non-essential cofactor of exportin-1 in the assembly of cargoes for nuclear export in the nucleus, where Ran-GTP concentration is very high (> 10 µM) [152–155].

Nuclear export of mRNA

In contrast to most protein cargo ensembles, tRNA and pre-microRNAs, whose nuclear export is dependent on Ran-GTP and importin-β/karyopherin-β-related nuclear transport receptors (e.g., exportin-t and exportin-5) [156, 157], the nuclear export of the bulk mRNA appears to be independent of the nucleocytoplasmic gradient of nucleotide-bound Ran GTPase (Fig. 3) [158]. Instead, the nuclear export of bulk mRNA in metazoans is orchestrated by the mRNA export receptor heterodimers, TAP (a.k.a., nuclear export factor 1—Nxf1) and p15 (a.k.a., NTF2-related export protein 1—Nxt1) [159–161]. These receptors are recruited to the messenger ribonucleoprotein (mRNP) particle, which is loaded with components of the transcription–export (TREX) complex, such as ALY (a.k.a. REF), and various heterogeneous nuclear ribonucleoproteins (hnRNPs) with various roles in RNA processing and export [162–168]. There are four major steps in the co-transcriptional maturation of mRNA in preparation for its export. These comprise: (i) the 5′ capping with 7′-methylguanosine and the CBP20 and CBP80 of the cap-binding complex (CBC) of the nascent pre-mRNA; these prevent the degradation of the pre-mRNA and promote its multi-step maturation and nuclear export [169–172]; (ii) the splicing of the pre-mRNA by proteins of the exon junction complex (EJC) that are recruited to the sites of exon fusions [163, 173, 174]; (iii) the 3′-end cleavage of the pre-mRNA downstream of its polyadenylation site [175]; (iv) the polyadenylation of the pre-mRNA [175, 176]. Notably, the capping and splicing steps of pre-mRNA processing promote the formation of nuclear export-competent mRNAs by recruiting the TREX complex. Among other factors, this complex includes the ATP-dependent DEAD-box RNA helicase and UAP56 (a.k.a. HEL), which recruits ALY to the mRNP [177]. Finally, pre-mRNAs and mRNAs appear to form “closed-loops” owing to the crosstalk and physical association between factors of the CBC/cap complex and those at the poly(A)⁺ site [178].

Fig. 3.

Nuclear export of messenger ribonucleoproteins (mRNPs). Two distinct nuclear export pathways of mRNA are represented on the left and right diagrams. The left diagram (1) shows that the nuclear export of mRNAs containing alternative mRNA nuclear export (ALREX)-promoting elements in the signal sequence-coding region (SSCR) at the 5′-end is dependent on the nuclear export receptors of mRNAs, TAP/p15, and poly(A)-binding proteins (PABP), but independent of transcription export (TREX) complex, and splicing and 5′cap structure of pre-mRNAs. The nuclear export factor(s) (Exp?) that mediate export of SSCR mRNAs are not known. The SSCR sequence interacts directly with a domain of Ranbp2 that contains zinc-finger motifs (ZnF) and this interaction potentiates the synthesis (or targeting) of secreted proteins from mRNAs co-translationally targeted to the endoplasmic reticulum (ER) by mechanisms not understood. It is possible that the nucleoporin, Gle1, at the cytoplasmic face of the nuclear pore and its partner, DEAD-box and RNA-dependent ATPase helicase, DDX19, promote also the nuclear export of SSCR mRNAs by catalyzing the remodeling (uncoating) of mRNPs. The right diagram (2) shows that the nuclear export of mRNAs is dependent on the cap-binding complex (CBC), which contains importin-α, TAP/p15, TREX and exon junction complexes (EJC) and PABP. Gle1 and DDX19 promote the nuclear export and remodeling of mRNPs as soon as the 5′-capped mRNP exits the nuclear pore. Release of the CBC and exchange with eIF4E is stimulated by the association of importin-α in the CBC to free importin-β after its regeneration by the Ran-GTP-binding domains (RBDs) of Ranbp2 (and SUMOylated RanGAP). The pioneer round of translation by ribosomes promotes the uncoating of mRNPs in preparation for steady-state translation. The top diagram (3) is like the mRNA export of diagram 2 with the exception that the remodeling the 5′-cap (CBC) takes place in the cytoplasm away from the filaments of the nuclear pore and it is catalyzed by the regeneration of importin-β by Ranbp1 (and RanGAP)

Elegant studies by Daneholt and coworkers and that were extended also by other groups have detailed the highly ordered escort process of Balbiani ring mRNP particles from the chromatin to the nuclear basket of the nuclear pore, its translocation through the central pore and extrusion at the cytoplasmic face of the nuclear pore [179–183]. This process involves the docking of the 5′ cap of the mRNP to the nuclear basket followed by its exit at the cytosolic face of the nuclear pore. The binding of nuclear export receptors of mRNA, TAP/p15 (a.k.a. Nxf1/Nxt1), to FG-repeats of some nucleoporins (Nups) of the nuclear pore appears to mediate the docking of mRNP to the nuclear pore and its translocation across the pore [166, 184–186]. Although two distinct hypotheses have been proposed to explain the transport of mRNPs from the transcription sites to the nuclear pores [187–191], the movement of mRNPs in the nuclear matrix and to the nuclear pore is governed by thermal/Brownian motion (passive diffusion) and thus this process depends on the size of the mRNP particles and viscosity of the nucleoplasm [192]. In contrast, the translocation through the nuclear pore is facilitated by the interaction of the nuclear export receptors with intrinsically disordered and flexible stretches of phenylalanine–glycine (FG)-repeats of ~ 10 nucleoporins that face the central channel of the pore and form a permeability barrier [131, 193, 194].

After exiting the nuclear pore, mRNPs encounter the nucleoporins, Nup214 (a.k.a., CAN), Nupl2/hcG1 (Nup42), Gle1 (Gle1B isoform), and the DEAD-box and RNA-dependent ATPase helicase, DDX19 (a.k.a. Dbp5 in yeast) at the cytosolic face of the nuclear pore [195–202]. Gle1 binds to hCG1 (a.k.a. Nup42) and DDX19 helicase and it stimulates DDX19 ATPase activity [201–203]. Nup214 enhances the Gle1-mediated ATPase activity of DDX19 and promotes its tethering to the cytoplasmic face of the nuclear pore [202]. The Gle1-mediated helicase activity of DDX19 stimulates the remodeling of mRNPs by removing proteins, such as TAP/p15 and the poly(A)⁺-binding protein Nab2, from the mRNP complex [197, 199, 202, 204–206]. This process also ensures that mRNAs cannot reenter the nuclear pore. Perturbations in the loading or removal of these receptors to or from mRNA decrease the efficiency of nuclear export and may even result in the retrograde transport of the mRNP from the pore [206]. Hence, the asymmetric localization of Gle1 and DDX19 at the cytosolic face of nuclear pore is critical to the directionality of the release of mRNPs in the cytosolic compartment (Fig. 3). Finally, the total translocation time of mRNAs, such as actin, across the nuclear pore (docking, transport, and release) is estimated to be ~ 180 ms [207], and only one-third of mRNPs are estimated to be successfully exported from the nucleus [208].

Another critical feature of the remodeling of mRNP as its 5′ cap exits the nuclear pore involves the replacement of the cap-binding complex (CBC) in mRNAs by eIF4E (and eIF4G and eIF4A) [172, 209, 210]. eIF4E is critical for the recruitment of the small ribosomal subunit and initiation of translation even before the export of the mRNP from the pore is completed [182]. In this regard, Gle1 and DDX19 may also play a role in translation and gene expression (this topic is discussed further below) [211, 212]. In addition to the classical roles of importin-β/α in nucleocytoplasmic transport, importin-β and importin-α exert important roles in cap-binding activity. Importin-α partakes in CBC in the nucleus and importin-β triggers the replacement of CBC-importin-α complex by eIF-4E at mRNA caps when they exit the nuclear pore [213–215]. Ranbp1 and/or Ranbp2, and possibly with the cooperation of RanGAP, triggers the liberation of importin-β from Ran-GTP after exiting the nuclear pore and thus allows importin-β to engage with the capped mRNP-CBC-importin-α complex in the cytosolic compartment. Hence, a non-canonical role of some importins is to regulate the capping and remodeling of mRNPs (Fig. 3). The remodeling of the 5′cap serves also two critical functions. First, CBC-capped mRNPs undergo a pioneer round of translation, which serves as a quality control mechanism to survey mRNAs destined for premature termination of translation and/or nonsense-mediated mRNA decay [216–219]. This step also promotes the remodeling of mRNPs by uncoating spliced mRNAs from other components, such as exon junction complex proteins [214, 220]. Second, the replacement of the CBC-capped mRNA by eIF-4E promotes the steady-state initiation of translation and protein synthesis. Finally, CBC is strongly regulated by growth factors and environmental stressors that by this mechanism regulate gene expression (e.g., suppression of translation) [215, 221–223].

Thematic variations in nucleocytoplasmic transport

In spite of the canonical roles of NTRs and TAP/p15 in nucleocytoplasmic transport that were described in the preceding sections, there are important and mounting variations to this theme. Due to space constraints, few examples are highlighted next that may provide rich mechanistic hints and insights to fundamental and interconnected biological and pathophysiological processes. I envisage that these non-canonical modalities of nucleocytoplasmic transport will play important roles in chartering the refinement, integration, and extension of organizing principles of nucleocytoplasmic transport in normal and disease states.

First, CRM1/exportin-1 also mediates the nuclear export of mRNAs and other RNAs that are released in the cytoplasm by hydrolysis of Ran-GTP. However, the nuclear export of RNA by exportin-1 depends on adaptors with NES signals, because exportin-1 does not bind RNA [196, 224–228]. Initially cited in this context is the nuclear export of the signal recognition particle (SRP) by exportin-1 in S. cerevisiae [229, 230]. The SRP is a cytosolic RNP comprised of a RNA molecule and core SRP proteins. According to the classical view, which has been recently refined [231], the SRP recognizes signal sequences at the amino-terminal end of proteins destined to the plasma membrane and secretion and thereby enabling their co-translational delivery and translocation to the endoplasmic reticulum. Hence, these early studies hint that exportin-1-mediated nuclear export regulates the biogenesis of proteins destined for secretion or to the plasma membrane. In vertebrates, however, it appears that exportin-5 mediates the nuclear export of SRP RNA [232, 233] and that mRNA export by exportin-1 is dependent at least on three different adaptors, such as RNA-binding protein human antigen (HuR), leucine-rich pentatricopeptide repeat protein (LRPPRC) and nuclear export factor 3 (Nxf3) [224–228].

Second, accumulating studies indicate that the large, multi-domain and peripheral nucleoporin, Ran-binding protein 2 (a.k.a. Nup358) serves as a scaffold and multidimensional platform to orchestrate the coupling of nuclear export and translation of mRNAs of secretory and plasma membrane proteins [234]. This effect enhances the translational potential of these mRNAs possibly by remodeling messenger ribonucleoproteins (mRNPs) when they exit the nuclear pore. Ranbp2 comprises the cytoplasmic filaments emanating from the nuclear pore [140, 142]. Exportin-1 docks to a domain of Ranbp2 containing several zinc-finger motifs (ZnF) [145] and this and the N-terminal leucine-rich domain of Ranbp2 are capable to interact directly with single-stranded RNA [234, 235]. The signal sequence-coding region (SSCR) of mRNAs coding secreted proteins contains RNA elements that stimulate an alternative mRNA nuclear export (ALREX) pathway and in particular of a subset of mRNAs characterized by adenine-less tracts in SSCR of ALREX-elements [234, 236]. This pathway appears dependent on TAP/p15, but independent of TREX, and splicing and 5′cap structure of pre-mRNAs [236]. The ALREX-promoting elements in SSCR potentiate the synthesis of reporter and native secreted proteins by mechanisms not understood yet [234, 236]. Interestingly, mutations of ALREX elements in SSCR lead to inhibition of nuclear export and formation of stress granules in the cytoplasm that include the mutated mRNAs and the translational initiation factor, eIF3B [234]. The ALREX-promoting elements interact selectively with the ZnF-containing domain of Ranbp2, while knockdown of Ranbp2 suppresses the synthesis of secreted and membrane-bound proteins, and possibly also of mitochondrial proteins [234]. Parenthetically, a recent study, which revisited the role of SRP in protein translation and targeting, found that loss of SRP leads to the mistargeting of mRNAs from the ER to mitochondria and thereby causing mitochondrial dysfunction [231].

Third, Ranbp2 was independently discovered as a multi-modular and cyclophilin-related protein with strong expression in cone photoreceptor neurons and as a candidate with biological functions similar to the cyclophilin encoded by ninaA of Drosophila that mediates the biogenesis of a subset of opsins [138]. Subsequent studies on the chaperone activity of the cyclophilin domain of Ranbp2 in cell culture and mice have shown that Ranbp2 controls the production of a subset of transmembrane G protein-coupled receptors, such as red/green opsin, in cone photoreceptor neurons, and without impairing the levels of the coding mRNAs [237–240]. Recent genetic studies in mice have also found that loss of Ranbp2 leads to pronounced accumulation of the levels of selective mRNAs that encode the secretory protein, chemokine ligand 14 (Cxcl14) in spinal motoneurons, and acetyl-CoA carboxylase 1 (Acc1) in the optic nerve (P. Ferreira, unpublished observations) [241]. The pronounced increase of Cxcl14 mRNA was accompanied by a sharp decrease of the levels of Cxcl14 protein, thus suggesting the uncoupling between export/targeting and translation of Cxcl14 and possibly of other mRNA species [241].

Fourth, exportin-1 mediates the nuclear export of eukaryotic translational initiation factor (elF4E)-dependent mRNAs with an elF4E-sensitive element (4E-SE) in their 3′UTR [242]. In this instance, the level of elF4E controls the remodeling of nuclear pore components at the cytosolic face, such as of Ranbp2, Nup214, Ranbp1, Gle1 and DDX19 [242]. As described earlier, these factors mediate the nuclear export and translational potential of mRNAs. The elF4E-mediated effects arise by changing indirectly the localizations of Ranbp2 and Nup214 (and the levels of Ranbp2) and by directly affecting the nuclear export of Ranbp1, Gle1 and DDX19 mRNAs containing 3′UTR 4E-SE and that are targets of elF4E [242]. Hence, a regulatory feedback loop appears to exist between factors regulating the translation initiation of selective mRNAs and their nuclear export.

Fifth, apparently conflicting experiments support and exclude the involvement of Ranbp2 in the nuclear export of bulk mRNA. In insect cells of fruit flies (D. melanogaster), which lacks Ranbp1 and express a Ranbp2 isoform with a much simplified molecular architecture than mammalian Ranbp2, knockdown of Ranbp2 promotes the nuclear accumulation of bulk poly(A)⁺ mRNA and loss of Nxf1 localization at the nuclear pore without changes in nuclear export mediated by exportin-1 [143]. Mouse embryonic fibroblasts lacking Ranbp2 also accumulate bulk poly(A)⁺ mRNA in the nucleus, but without nucleocytoplasmic mislocalization of Nxf1 [243]. By contrast, knockdown of Ranbp2 in human HeLa and U2OS cells does not affect the nucleocytoplasmic partitioning of bulk poly(A)⁺ mRNA [234, 244]. The reasons for these apparent discrepancies remain elusive, but they may be rooted on cell-type, compensatory and species-specific differences that underlie the orchestration of nucleocytoplasmic transport by Ranbp2 and its accessory partners. This notion is supported by genetic complementation studies in mice that found loss of Ran-GTP-binding activity of the Ran-binding domains, RBD2 and RBD3, of Ranbp2 is essential to the survival of selective cell types only, and thus, they are not biologically equivalent to other RBDs of Ranbp2 (e.g., RBD1 and RBD4) [241, 245]. Further, some accessory factors or substrates undergo species and cell type-specific recruitment to Ranbp2 (and Ranbp1) because Ranbp1, several domains of Ranbp2, and accessory factors thereof (e.g., SUMOylated RanGAP) are poorly conserved between species or some do not even exist in lower organisms (e.g., Ranbp2) [143]. These non-conservation attributes extend also to several nucleoporins that are absent or have poorly conserved primary structures in several species (e.g., Nup153) [194]. Finally, loss of Ranbp2 in spinal motoneurons (and retinal ganglion neurons; Ferreira, P. unpublished observations) cause perturbations in a restricted and largely non-overlapping set of transcripts [241]; thus supporting that Ranbp2 co-opts accessory factors for the cell type-dependent orchestration of nucleocytoplasmic transport of mRNAs perhaps by combinatorial mechanisms. Together, these distinctions and limitations will likely extend also to other nucleoporins [194]. Notably, these shared and unique attributes of nucleocytoplasmic transport have been overlooked in models of neurodegenerative diseases in which impairment of nucleocytoplasmic transport is emerging as a pathogenic driver of neurodegeneration.

Finally, alternate components of the cap-binding complex (CBC) exist and they play critical roles in mRNA export under stress stimuli. Another conserved and alternate, but less understood complex, TREX-2, which is mostly localized to nuclear pores, appears also to bridge the transcription and transfer of mRNAs to Nxf1 [246, 247]. A component of the TREX-2 complex is the germinal-center associated nuclear protein (GANP; a.k.a. Sac3 in yeast), which binds mRNA and interacts with RNA polymerase II and facilitates its recruitment to selective genes [248].

Collectively, these findings only scratch the surface of the intricacies and complexities of the molecular mechanisms underpinning nucleocytoplasmic transport and its regulation. As the functions, mechanisms, and regulations of many NTRs, nucleoporins and accessory factors are distilled in greater detail and in pathophysiological settings that are undergoing rapid flux, they will serve to validate, refine and extend the physiological significance of existing concepts across cell types and diseases in the human and model organisms. These advances are anticipated to eclipse in the near future the current knowledge of nucleocytoplasmic transport in cell functions and disease and harness their therapeutic potential.

Bridging nuclear export with anterograde transport in neurons

Neurons epitomize the polarization and compartmentalization of cellular architecture. The linear length of an axon of a lower motoneuron of the spinal cord can reach a meter in size in the human [249], whereas the total length of axons of striatal dopaminergic neurons is estimated to be ~ 8000 m in the rat [250, 251]. The exquisite sizes and polarized morphologies of many neural cell types are thought to preordain neurons to neurodegeneration by dysregulation of intracellular transport. In line with this point, it remains largely unresolved how cargoes exported from the nucleus are captured by the intracellular transport machinery, such as microtubule-based motors, for the “fast” and polarized delivery of cargoes to distal compartments, such as dendrites, axons or synapses. The coordination of the coupling of nuclear export and intracellular transport of cargoes poses another dimension of complexity to nucleocytoplasmic trafficking events, whose molecular underpinnings remain incomplete but are of unique significance to neural function and survival. In other words, foremost here are the spatiotemporal challenges intrinsic to neurons and that these need to overcome to prevent the entrapment of cargoes and accessory factors in somata after they exit the nuclear pore. Another point of critical importance that needs consideration is how the roles by moonlighting proteins partaking in nucleocytoplasmic and intracellular transport machineries are discriminated and segregated spatially and temporally. As expanded later in this review, mounting evidence indicates that dysregulation of these processes underpins the manifestation of pathological traits and most importantly, the pathomechanisms of several forms of neurodegenerative diseases, such as ALS.

The cytoplasmic nucleoporin, Ranbp2/Nup358, and the microtubule-based motor protein of the kinesin superfamily proteins (KIF), such as kinesin-1 (KIF5), have emerged as prime candidates to couple nuclear export with fast anterograde transport of cargoes exiting the nuclear pore. Ranbp2 associates directly via a non-conserved kinesin-binding domain (KBD) with the conventional kinesin-1 isoforms, KIF5B and KIF5C [252]. This domain of Ranbp2 jump-starts the motor activity of kinesin-1 [253, 254]. Further, the Ran-GTP-binding domains, RBD2 and RBD3, of Ranbp2 that flank its KBD, enhance the Ran-GTP-independent association of Ranbp2 with kinesin-1 and boost its motor activity [253, 254]. To date, Ranbp2 is still the only native and known cargo that can directly activate and boost motor activity of a kinesin in a minimal reconstitution system of purified components. In retinal ganglion neurons, whose nuclear envelopes are crowded with nuclear pores, Ranbp2 is found at the nuclear rims, where it colocalizes with Ran GTPase [255] (Fig. 4). Ranbp2 and Ran GTPase also colocalize to a subpopulation of discrete granules present along “highways” that radiate from the nuclear rim toward the axon hillock where they coalesce (Fig. 4) [255]. This implies that Ranbp2 can be released from the cytoplasmic filaments of nuclear pore. In this regard, genetic ablation of Ranbp2 has shown that the half-life of Ranbp2 at the nuclear pore is much shorter than the lifespans of many other nucleoporins (e.g., < 2.5 days vs months) [241, 256]. The compositions of the Ranbp2 and Ran-GTPase-containing granules in the cytoplasm are unknown, but independent studies indicate that they may consist of mRNPs transported by kinesin-1. For example, Kanai and coworkers found that in mouse brain and dendrites of hippocampal neurons kinesin-1 binds directly and transports mRNP granules of 1000S [257]. These granules are composed mRNAs, such as Arc and CamKIIα, and 42 proteins, such as RNA helicases, ALY, hnRNPs (e.g., hnRNPA/B, hnRNP-U), eIF factors, Fragile X mental retardation protein and related members (e.g., FMRP1, FXR1, and FXR2) and Staufen [257]. Other studies have also found that kinesin-1 mediates the transport of Staufen-2 (Stau-2), FRMP-containing RNPs and the ribosome receptor, p180 [258–260].

Fig. 4.

Complete co-localization of Ran-binding protein 2 (Ranbp2) and Ran GTPase at the nuclear rim and to a subset of granules (arrowheads) in threads (“highways”) radiating from the nuclear rim and toward the axon hillock of large retinal ganglion neuron (magnocellular neurons). Thin arrow on the left shows the co-extrusion or co-displacement of Ranbp2 and Ran GTPase from the nuclear rim. Thick arrow on the right points to an axon of an unmyelinated retinal ganglion neuron in which there is complete co-localization of Ranbp2 and Ran GTPase. Top (a) and middle panels (b) show the immunostaining of the retina ganglion cell layer of a cryosection of bovine retina immunostained for Ranbp2 (top) and Ran GTPase. The lower panel (c) is an overlay image of a Nomarski (differential interference contrast—DIC) image and the image a and b. Numerous granules organized along longitudinal bundles (“highways”) are visible and a subset of these co-immunolabel with Ranbp2 and Ran GTPase. These granules likely constitute RNPs. Cell labeled with a “asterisk” shows the nucleus of a small ganglion neuron with intense nuclear immunolabeling of Ran GTPase, but lacking Ranbp2 at the nuclear envelope/pores.

Image adapted from reference [255]

mRNPs illustrate a class of nuclear export cargoes whose long-distance transport and compartmentalized localization are critical to enable the spatial and temporal orchestration of localized protein synthesis. This function is critical for the maintenance of distal subcellular structures and functions of neurons and responses to multiple stimuli. In contrast to other cell types, the translational repression, transport, remodeling, and disassembly of mRNPs are competing imperatives that need to be precisely balanced in a spatial and temporal manner by neurons to ensure neural homeostasis. For example, elF4AIII, is a component of the exon junction complex (EJC), which is localized primarily to the nuclear compartment of HeLa and glial feeder cells [261–263], where mRNPs are remodeled and mRNAs are translated soon after exiting the nuclear pore. By contrast, elF4AIII is localized to somata and dendrites of hippocampal and cortical pyramidal neurons, where it associates presumably with 3′-UTR introns of translationally silent mRNAs in granules containing other RNPs, such as FMRP and Stau-1 [263]. Down-regulation of elF4AIII stimulates quantal synaptic transmission by increasing the levels of Glur1 receptor and Arc, which is required for long-term potentiation at synapses [263]. elF4AIII regulates Arc protein synthesis by degradation of Arc mRNA upon translation by a nonsense-mediated decay (NMD) mechanism and termed, translation-dependent mRNA decay [263]. Another critical implication of these findings is that EJC proteins in mRNPs provide a chronological window of the translational status of mRNAs in neurons [263]. In non-neural cells, EJC proteins are displaced during the pioneer round of translation when NMD also occurs and soon after the mRNP exits the nuclear pore. Hence, a localized translational control mechanism in neural compartments is emerging as a common feature in neural function and with consequential pathophysiological implications.

Finally, two discrete roles were cited previously for importin-β. A third emerging role is that importin-β acts as a sensor of axonal injury [264, 265]. In this regard, importin-β mRNA is transported to axons via a cis-acting regulatory sequence in its 3′-UTR [265]. This mechanism allows the segregation of nucleocytoplasmic transport functions of importin-β with its local translation in axons. Here, importin-β acts a sensor for retrograde signaling in dorsal root ganglion neurons of mice with sciatic nerve injury by mechanisms that are not fully eluciated yet. In light of the preceding discussion, it is tempting to hypothesize that importin-β’s role in axons is inextricably linked to the remodeling of mRNP in a manner reminiscent to that of pioneer round of translation, but that involves the regulation of local translation of mRNPs in axons [264]. Parenthetically, Ran GTPase, Ranbp1 and the nuclear shuttling factor, Stat3, may also partake in the local regulation of protein synthesis elicited by the importin-β-dependent retrograde injury signaling [266, 267]. Although axonal protein synthesis has been a contentious topic [268], these and other studies, such as local translation in axons of commissural neurons and retinal ganglion neurons, support the importance of protein synthesis in axons [269–272]. Importin-β is also localized prominently at the nuclear rim, lower fibers and synaptic pedicles of cone photoreceptor neurons, where it is abundantly expressed [255]. Ranbp2 also associates with importin-β and Stat3 [97, 142, 239]. Stat3 and Ranbp2 mediate the neuroprotection of photoreceptor neurons against physical and/or genetic injury in mice [273–275], and Ranbp2 regulates the trans-activation potential of Stat3 in gene expression [240]. Hence, it is likely that the importin-β-dependent injury-signaling pathway is shared by different neurons in stress signaling.

Dysregulation of nucleocytoplasmic transport: a pathogenic driver in ALS and neurodegeneration

While the guiding principles described in the preceding sections articulate general and alternate concepts of nucleocytoplasmic transport, we are now beginning to appreciate its intricacies and complexities to rationalize another dimension of complexity linked to neurodegenerative diseases, such as sALS and fALS. While this field is still replete with unresolved issues as cited at the beginning of this review, a theme is beginning to emerge that cements the notion that mutations in several ALS genes impair nucleocytoplasmic transport of their products and accessory partners. This impairment is emerging as a major pathogenic driver of neurodegeneration in ALS as well as in other neurodegenerative diseases. Next, I highlight some ALS targets with shared and unique attributes in defects of nucleocytoplasmic transport.

Cu/Zn superoxide dismutase 1 (SOD1)

The first study of nucleocytoplasmic dislocation of NTRs and accessory substrates linked to ALS was reported by Zhang et al. in motoneurons of the anterior horn in a mouse model of fALS with the G93A mutation in SOD1 [41]. In this study, Tg-SOD^G93A mice presented subcellular redistribution of importin-β and importin-α from the nuclear to the cytosolic compartments and immunoreactivity of Lewy body-like hyaline inclusions to these NTRs and their substrates, such as histone H1 [41]. The same laboratory subsequently reported that sALS patients lacked importin-β in nuclei of motoneurons, that there were irregular nuclear contours of the nuclear envelope immunolabeled with Nup62, even at presymptomatic stages of the disease, and that worsened at later stages of the disease [42]. Although SOD1 inclusions are found in ALS patients with SOD1 mutations, there is controversy surrounding the presence of misfolded SOD1 in the central and peripheral nervous system (e.g., spinal cord and cortex tissues) of sALS and non-SOD1 fALS patients using conformationally sensitive antibodies [276]. Regardless, a provocative study indicates that nuclear clearance of mutant SOD1 exerts neuroprotective effects against cytotoxicity caused by gain-of-function mutations in SOD1 [51]. This study showed that wild-type SOD1 was distributed between the nuclear and cytoplasmic compartments, whereas mutant forms of SOD1 were found primarily in the cytoplasm. Misfolded SOD1 and ALS-causing mutations in SOD1 expose a normally buried NES-like sequence that leads to the exportin-1-mediated nuclear export of misfolded SOD1. Hence, this study suggests that an exportin-1-mediated nuclear export mechanism promotes nuclear proteostasis against the accumulation of misfolded SOD1 in the nucleus. Alternatively, it is possible that saturation or equilibrium shifts of a nuclear export and saturable pathway caused by changes of exportin-1 association to misfolded SOD1, which is an abundant protein, promotes pathogenicity owing to the dysregulation of other cargoes of exportin-1 and/or to rearrangements or displacement of selective nucleoporins (e.g., Nup62).

Chromosome 9 open reading frame 72 (C9orf72)

The subcellular redistribution of importin-β were also confirmed in sALS and fALS patients with noncoding hexanucleotide (G₄C₂) repeat expansions of C9orf72 [46]. In particular, spinal motoneurons of these patients had loss of localizations of importin-β, Ran GTPase and a short protein isoform encoded by a splice variant of C9orf72, C9-S, at the nuclear rim [46]. Notably, isoform-specific antibodies showed that C9-S relocated from the nuclear to the plasma membrane of spinal neurons of ALS patients with and without mutations in C9orf72 [46]. It is possible that dipeptide repeat proteins (DPRs) expressed from the repeat-associated non-AUG (RAN) translation of bidirectionally transcribed expansion repeats in C9orf72 (C9Ran) cause also the sequestration of importin-β, Ran GTPase and other proteins, such as RanGAP. Expression of DPRs reduces the nuclear-cytoplasmic distribution of Ran GTPase in C9-ALS iPSC neurons [48]. Notably, TDP-43 was also redistributed from the nuclear to the cytosolic compartment in sALS and fALS with expansion of C9orf72 [5, 36, 46]. In mice, DPRs form intranuclear and cytoplasmic inclusions that immunostain for ubiquitin primarily in neurons [36]. In transgenic mice and C9-ALS iPSC-derived motoneurons, the C9Ran transcripts formed RNA foci and colocalized with hnRNPA1 and Pur-α [37, 277]. Finally, genetic screens in yeast and photoreceptor neurons of the compound eye of Drosophila melanogaster against DPR toxicity identified Aly/REF of the TREX complex, Gle1, and selective nucleoporins and substrates of the Ran GTPase cycle, such as importins and exportins, as strong genetic modifiers of the proteotoxicity of DPRs [47, 48, 278, 279].

Transactive response (Tar) DNA-binding protein 43 (TDP-43)

Mutations in TDP-43 cause ALS and frontotemporal dementia (FTD) [280, 281]. Although initial studies found that TDP-43 acts a splicing and transcriptional regulator [282–285], TDP-43 is a nucleocytoplasmic shuttling hnRNP, whose steady-state localization appears to be primarily nuclear [286, 287]. ALS mutations in TDP-43 shift its localization to the cytosol where it aggregates and becomes ubiquitinated [288–290]. Studies have shown that TDP-43 associates with introns of pre-mRNAs, and its own 3′-UTR and that of Ran GTPase, and thereby regulating the levels and splicing of thousands of pre-mRNAs [45, 291–294]. As a splicing regulator, TDP-43 was also found to modulate the splicing of multiple pre-mRNAs, such as hnRNPA1, that leads to the production of an hnRNPA1 variant, hnRNPA1B, with a longer prion-like domain and an increased propensity to aggregation [295]. The cytotoxicity of hnRNPA1B may be enhanced by TDP-43 mislocalization to the cytosol. TDP-43 is also transported as mRNPs via microtubules in axons of primary cortical neurons [296, 297]. ALS mutations in TDP-43 suppress the axonal trafficking of mRNP granules without apparently affecting mitochondrial transport and thus, supporting the notion that this inhibition promotes distal axonopathy as observed in ALS [297, 298]. Although the molecular basis of TDP-43 dysfunction in axonal transport of mRNPs is obscure, this may result from the dysregulation of microtubule-based motors associated to hnRNPs and/or accessory factors that bridge interactions between motors and one or more components of the mRNP granule.

The nuclear shuttling of TDP-43 is carried out by importin-β/karyopherin-β-related NTRs [299]. TDP-43 harbors a classical NLS at its N-terminal domain, but surprisingly none of the ALS mutations affect the NLS [43]. Knockdowns of importin-β and cellular apoptosis susceptibility (CAS) receptor (a.k.a. exportin-2), which acts as a nuclear export receptor for importin-α, result in the cytoplasmic localization and aggregation of TDP-43 [299]. FTD-ALS patients have also reduced levels of CAS in the brain, but not spinal cord [299]. In brains of sporadic FTD and C9ALS/FTD, importin-α2 (KPNA2) was depleted from the nucleus and accumulated in the cytoplasm and there was an overall decrease of its expression levels [54]. Importantly, this pathological manifestation was frequently observed without pTDP-43 inclusions in sporadic FTD and C9ALS/FTD or DPR inclusions in C9ALS/FTD patients, thus suggesting that soluble rather than aggregate forms of pTDP-43 and DPR exert neurotoxicity [54].

As it relates to the nuclear export of TDP-43, recent studies with primary cortical neurons indicate that redundant nuclear export pathways by multiple export receptors, such as exportin-1, exportin-7 and Nxf1, mediate the nuclear egress of TDP-43 and that selective inhibitors of nuclear export targeting exportin-1 partially rescue motor deficits in a rat model of TDP-43-induced paralysis [300]. This study appears to support the observation that NES mutations in TDP-43 ameliorate TDP-43 toxicity caused by its over-expression in fruit flies and primary rat cortical neurons [301, 302]. However, other studies have produced conflicting results. For example, a study found that the nuclear egress of TDP-43 and FUS (discussed see next section) in HeLa cells are independent of CRM1/exportin-1, exportin-5, the mRNA export factor ALY/REF, and of its RNA-binding function(s) [303]. In contrast, artificial enlargement of TDP-43 (and FUS) causes the impairment of their nuclear export, whereas overall inhibition of transcription stimulated TDP-43 nuclear egress [303]. Hence, passive diffusion appears to mediate the nuclear export of TDP-43. The exportin-1-independent nuclear export of TDP-43 and its nuclear egress by passive diffusion was also concordant with another independent study with HeLa and primary hippocampal neurons [304]. The reasons behind these conflicting reports are unclear, but they may partially arise from the cell-type dependent expression of nuclear export accessory factors.

Finally, recent studies found that ectopic expression of mutants of TDP-43 and a 25 kDa C-terminal fragment of TDP-43 in cultured cells sequestered or compromised the localization of nucleoporins and NTRs and promoted the nuclear retention of poly(A)⁺ mRNA presumably via prion-like domains shared by these proteins [305]. However, it is important to inject a cautionary note that TDP-43 pathology is not pathognomonic for ALS, because TDP-43 inclusions are also present in several other diseases, such as Alzheimer’s, Lewy body disease, Parkinsonism, chronic traumatic encephalopathy and even in healthy brains of the elderly [5].

Fused in sarcoma (FUS)

Mutations in FUS cause ALS and FTD [306, 307]. Unlike TDP-43, FUS harbors a non-classical NLS with a PY motif (PY-NLS) and about half of the known mutations in ALS and FTD affects this sequence [43, 306, 307]. Like TDP-43, FUS localization is primarily nuclear and ALS mutations promote FUS relocation to the cytoplasm, where FUS inclusions are also formed [43, 306–308]. Notably, there is a correlation between the degree of impairment of nuclear import by FUS mutations and the age of disease onset [43]. Nuclear import of FUS is mediated by importin-β2 (a.k.a. transportin) and impairment of this nuclear import pathway promotes the relocation of FUS to the cytoplasm, where it is sequestered into stress granules of cultured cells and of spinal neurons of patients with fALS and sFTD-FUS [43]. Like TDP-43, FUS binds thousands of mRNAs [309] and it appears to mediate mRNA transport to neural dendrites [310, 311].

Heterogeneous nuclear ribonucleoproteins (hnRNPs)

Mutations in hnRNPA1 and hnRNPA2B1 (this gene produces two isoforms, hnRNPA2 and hnRNPB1) cause a rare and dominant form of ALS and multisystem proteinopathy (MSP), which is a degenerative disorder affecting motoneurons, brain, muscle and bone [312]. The disease mutations fall within a prion-like-domain (PrLD) that is common to other hnRNPs, and they appear to increase the propensity to fibrillization of hnRNPA1 and hnRNPA2 [312]. Although hnRNPA1 and hnRNPA2B1 have primarily nuclear localization, the pathogenic D290V mutation in hnRNPA2 promotes its relocation and accumulation in cytoplasmic inclusions of the mouse muscle [312].

hnRNPA1 harbors a 38 residue non-canonical NES and NLS domain, termed M9, that is implicated in nuclear import and export and thus, M9 alone mediates the nuclear shuttling of hnRNPA1 [313–315]. hnRNPA2/B1 appear also to share the M9 signal sequence [314]. Importin-β2 mediates the nuclear import of hnRNPA1 [316]. hnRNPA1 and hnRNPA2B1 are implicated in multiple facets of RNA metabolism, such as splicing, trafficking and ribostasis of mRNAs [317–319]. For example, cis-acting hnRNPA2 response elements (A2RE) that bind hnRNPA2B1 are found in mRNAs (e.g., myelin basic protein and Arc) and they are involved in mRNA trafficking to neural dendrites, where hnRNPA2B1 is also found [320]. Further, hnRNPA2B1 expression and transport to synaptic dendrites are stimulated by BDNF [321].

Recently, transcriptome-wide screening of targets of hnRNPA2B1 in human iPSC-derived motoneurons and mouse spinal cord found a restricted number of transcripts with a shared UAG(G/A) signature in their 3′-UTR even though there was little overlap between the human and mouse mRNAs [322]. The alternative polyadenylation and splicing of some of the mRNAs was affected by hnRNPA2B1 depletion [322]. Further, the pathogenic D290V mutation appears to exert a toxic and perhaps gain-of-function by promoting exon exclusion; however, it produced divergent alternative splicing events between fibroblast and iPSC-derived motor neurons of affected and unaffected individuals as well as from those caused by hnRNPA2B1 depletion [322]. Regardless, these findings contrast to the widespread transcriptomal effects caused by TDP-43 or FUS depletion reported by other studies [291, 323]. However, the pathophysiological relevance of the heterogeneous transcriptomal changes controlled by hnRNPA2B1, TDP-43 or FUS are obscure. Another transcriptomal screening found that hnRNPA2B1 acts as nuclear “reader” of a consensus motif comprising N⁶-methyladenosine modification of mRNA and that impacts alternative splicing as well as primary miRNA processing [324]. Sumoylation of hnRNPA2B1 also controls its binding to Exomotifs in miRNAs and their loading to exosomes [325].

Gle1 and other nucleporins

As cited earlier, Gle1 is a ubiquitous nucleoporin at the cytosolic face of the nuclear pore that mediates mRNA export [200–202]. Autosomal recessive mutations in GLE1 caused by a 3-residue PFQ-insertion or compound heterozygous mutations, V617M and I684T, lead to the fetal motoneuron diseases, such as lethal congenital contracture syndrome 1 (LCCS1) or lethal arthrogryposis with anterior horn cell disease (LAAHD) [326]. This insertion in Gle1 appears to affect the self-association of Gle1 and its dysfunction at the nuclear pore [200]. Subsequent studies found that heterozygous mutations in GLE1 caused ALS (e.g., R697C) [327]. Notably, while the autosomal recessive mutations in Gle1 do not affect Gle1 localization at the nuclear pore, ALS-associated Gle1 mutations caused a reduction of wild-type Gle1 and loss of the mutant Gle1 at the nuclear pore that results from the inability of mutant Gle1 to bind the scaffold nucleoporin, Nupl2/hcG1, likely due to loss of Gle1stability [202, 327]. Hence, differential losses of Gle1 activities promote distinct pathological outcomes likely by mechanisms that differentially affect the remodeling of mRNPs as they exit the nuclear pore.

The Ran-binding protein 2 (a.k.a., Nup358) is a peripheral nucleoporin that forms the cytoplasmic filaments of the nuclear pore [142]. Although no human mutations linked to ALS have been found in RANBP2 yet, recent studies have shown that mice with loss of Ranbp2 in motoneurons develop rapidly progressive ALS-like motor traits, such as paralysis, dysphagia, and respiratory distress that culminate in the death of mice [241]. These traits are accompanied by axonopathy without TDP-43 pathology and by the disruption of nucleocytoplasmic transport, such as nucleocytoplasmic mislocalization of exportin-1, importin-β and Ran GTPase, and accessory substrates, such as HDAC4 [241]. hnRNPH3 was a new substrate affected by loss of Ranbp2 and its immunogenicity was lost in situ in motoneurons, but not in immunoblots of homogenates, thus indicating the misfolding or aggregation of hnRNPH3 by loss of Ranbp2 [241]. Further, spinal motoneurons presented dysregulation of chemokine signaling, which was characterized by the formation of intracellular deposits of components of this signaling pathway, such as its Cxcl14 ligand, Cxcr4 receptor and latent and activated Stat3 effector [241]. As described in the preceding sections conflicting results have implicated Ranbp2 in the nucleocytoplasmic export of bulk mRNA. However, differential transcriptomal analysis of sciatic nerve of mice with and without Ranbp2 in motoneurons showed that a limited number of mRNAs were affected by Ranbp2 loss [241]. Among these, there was pronounced accumulation of Cxcl14 mRNA both in the spinal cord and sciatic nerve but paradoxically this was accompanied by a decrease of Cxcl14 protein levels [241]; thus, it is possible that loss of Ranbp2 promotes the uncoupling of translation from Cxcl14 mRNA due to improper mRNP assembly or remodeling after exiting the pore. Finally, a novel intranuclear and long-lived Ranbp2 isoform, which is unique to spinal motoneurons, relocates to the cytosolic compartment where it localizes to the mitochondria after losses of Ranbp2 and of the short-lived Ranbp2 isoform at the nuclear pores [241]. In light of the findings that Ranbp2 controls the activation of kinesin-1, which is the primary motor for mitochondrial transport [254, 328–330], it will be important to define the role of this nuclear sequestered Ranbp2 isoform in mitochondrial transport and motor behavior. In this regard, a novel and small isoform of Ranbp2 that co-purifies and colocalizes with the mitochondria has been recently identified [331]. Finally, hnRNPA2B1 is a substrate for the cyclophilin domain (CY) of Ranbp2 and its cis–trans peptidyl–prolyl isomerase (PPIase)/chaperone activity [239]. Loss of the PPIase activity of Ranbp2 in mice promotes a post-transcriptional decline of hnRNPA2B1 levels [239], whereas novel small molecules against the PPIase pocket of CY of Ranbp2 promote a decline of hnRNPA2B1 in HeLa cells [240]. Hence, Ranbp2 emerges as a therapeutic target to control the pathogenicity of mutations in hnRNPA2B1 and the neuroprotection potential of other substrates, such as Stat3, in ALS [240, 332].

Recent studies also indicate that mutations in the nucleoporin, NUP88, cause lethal fetal akinesia deformation sequence (FADS), which involves congenital malformations related to impaired fetal movement [333]. Like Gle1, Nup88 is also involved in mRNA export. Hence, Nup88 appears to be another promising nucleoporin candidate gene for ALS.

ALS genes with unrelated nucleocytoplasmic transport functions

The roles of several other ALS genes, whose functions were once thought not to be involved in nucleocytoplasmic transport and/or RNA metabolism, may also present functions related to nucleocytoplasmic transport events. These include the vesicle-associated membrane protein-associated protein B and C (VAPB) [334, 335], optineurin [336, 337] and profilin 1 [338, 339]. ALS mutations in VAPB cause nuclear envelope defects possibly caused by the disruption of transport of nuclear envelope proteins from the ER-Golgi intermediate compartment [340]. Optineurin, which is implicated in ALS and glaucoma, translocates to the nucleus in response to an apoptotic stimulus and intranuclear inclusions of optineurin are found in patients with neuronal intranuclear inclusion disease (NIID) [336, 337, 341, 342]. Finally, ALS mutations in profilin 1, a small actin-binding protein which controls actin dynamics, influences stress granule dynamics [343]. Profilin 1 is exported from the nucleus by exportin-6, which appears to act as a constitutive suppressor of actin polymerization in the nucleus [344], where its accumulation may be pathogenic. This may represent a back sorting mechanism of extrusion of cytoplasmic proteins from the nucleus by exportins owing to leakage of otherwise cytosolic proteins into the nucleus (e.g., Ranbp1, RanGAP) [84, 345–347].

Concluding remarks

A unifying theme is emerging from provocative and accumulating evidence and hints in the literature that nucleocytoplasmic transport plays a central role in the pathogenesis of several forms of ALS. This affords a novel opportunity and fertile ground to illuminate and interface the underpinnings of a fundamental biological process with disease manifestations that compromise proteostasis, ribostasis, and neural functions in a yet more physiological context. Accumulating evidence also support that alterations in nucleocytoplasmic flux or partition of disease-causing substrates may constitute pathogenic drivers for other diseases other than ALS. For example, formation of cytoplasmic aggregates of mutant huntingtin, TDP-43 and artificial-β-sheets impairs the localization and sequesters several components of the nuclear import and export machinery and affects the integrity of the nuclear pore [49, 52–54]. Likewise, similar findings have been recently extended to tau, whose intracellular aggregation in neurofibrillary tangles is a hallmark of Alzheimer’s disease and tauopathies [55]. The findings indicate that nuclear pore and nucleocytoplasmic transport dysfunctions may promote neurotoxicity induced by tau. Finally, the nucleocytoplasmic export pathway has become a major therapeutic target for cancer and viral diseases and more recently, neurological diseases causing axonal damage, such as ALS [48, 348, 349]. Specifically, a growing armamentarium of reversible and irreversible and structure-based designed inhibitors directed toward exportin-1/CRM1 have been developed. These exportin-1/CRM1 inhibitors [a.k.a. selective inhibitors of nuclear export (SINE)], appear to hold therapeutic promise in preclinical models (e.g., KPT-276 and KPT-350) [48, 348], even though off-target effects, and thus toxicity, are a potential concern due to the broad-spectrum substrate specificity of exportin-1/CRM1 [92]. In spite of these advancements, the challenges ahead are considerable, given the spectrum of biological processes that intersect the biogenesis and pathogenic potential linked to the misprocessing of mRNAs and formation of protein aggregates and inclusions. Looking ahead, I can envision that mechanistic-guided concepts and approaches will illuminate in greater detail the mysteries on how moonlighting proteins of the nucleocytoplasmic transport machinery intersect with multitasking substrates of pathogenic potential to produce neural type-restricted and deleterious manifestations. The emerging new concepts will enable to harness regulatory mechanisms of nucleocytoplasmic transport into therapeutic approaches of ALS and other neurodegenerative diseases that are currently lacking.

Abbreviations

ALS

Amyotrophic lateral sclerosis

Ranbp2

Ran-binding protein 2

Ranbp1

Ran-binding protein 1

RBDs

Ran-GTP-binding domains

ZnF

Zinc-finger motif

Nup

Nucleoporin

CY

Cyclophilin

RCC1

Regulator of chromosome condensation 1

RanGAP1

Ran GTPase-activating protein-1

NTRs

Nuclear transport receptors

NES

Nuclear export sequence

NLS

Nuclear localization sequence

mRNP

Messenger ribonucleoprotein

TREX

Transcription–export complex,

hnRNPs

Heterogeneous nuclear ribonucleoproteins

CBC

Cap-binding complex

EJC

Exon-junction complex

eIF

Eukaryotic initiation factor

SRP

Signal recognition particle

SSCR

Signal sequence-coding region

ALREX

Alternative mRNA nuclear export

Cxcl14

Chemokine ligand 14

Acc1

Acetyl-CoA carboxylase 1

SOD1

Cu/Zn Superoxide dismutase 1,

C9ORF72

Chromosome 9 Open Reading Frame 72

TARDBP

Transactive response DNA-binding protein (TDP-43)

FUS

Fused in sarcoma

DPRs

Dipeptide repeat proteins

LCCS1

Lethal congenital contracture syndrome 1

LAAHD

Lethal arthrogryposis with anterior horn cell disease

Ran

Ras-related nuclear protein

Funding

This work was in part funded by National Institutes of Health Grants GM083165, GM083165-03S1 and EY019492 to P.A.F.

Compliance with ethical standards

Conflict of interest

The author declares no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. The author consents for the publication of this study.