TDP-43 is an evolutionarily conserved DNA/RNA binding protein comprising four primary domains: an N-terminal domain (NTD, aa: 1–103) with a nuclear localization signal (NLS, aa: 82–98), two RNA-recognition motifs (RRMs, aa: 104–200, 191–262), and an intrinsically disordered C-terminal low-complexity domain (LCD, aa: 274–413) (see Figure 1). Insights from key structural features and disease causative mutations in TDP-43 highlight important mechanisms that drive TDP-43 pathogenesis. The vast majority of ALS/FTD causing mutations reside within its LCD and were shown to increase the propensity of TDP-43 to aggregate (Johnson et al., 2009; Buratti, 2015). In cases of familial ALS the A315 residue is mutated to E or T (A315E and A315T), which can lead to the assembly of tightly packing steric zippers that form stable fibrils (Nelson et al., 2005). The A315E mutation and phosphorylation of A315T also introduce electrostatic interactions into the amyloidogenic core regions of TDP-43, stabilizing larger fibril conformations that can seed irreversible aggregates (Cao et al., 2019). These TDP-43 aggregates could cause translation inhibition of sequestered mRNAs or affect translation by association with factors that regulate protein synthesis (Ramaswami et al., 2013). Interestingly, oligomerization of TDP-43 through the N-terminal domain via salt bridges between D22 and R55, E3 and R52, as well as E17 and R52 has been shown to antagonize TDP-43 proteinopathy by spatially separating LCDs that may otherwise aggregate (Afroz et al., 2017). Furthermore, the tandem RRMs of TDP-43 are of central importance to the specificity and activity of the wild-type protein, contributing to most of its RNA binding activity. The two RRMs have distinct nucleotide specificities, but mainly recognize (UG/TG)n repeat RNA via the organization of key aromatic and charged residues (Lukavsky et al., 2013). Recent findings suggest that the RNA binding activity of TDP-43 may be of particular interest in disease, as association with long sequences of UG rich RNA has been shown to dissolve optogenetically induced TDP-43 aggregates in vitro and reduce neurotoxicity (Mann et al., 2019; Grese et al., 2021). Another interesting feature of the RRM domains is the presence of a salt bridge, R151/D247, that is not required for RNA binding per se, but regulates affinity and specificity (Flores et al., 2019). Disrupting the R151/D247 salt bridge affects RNA binding and destabilizes the protein, which mitigates toxicity.

Recently, a short TDP-43 (sTDP-43) isoform upregulated by neuronal hyperactivity has been identified (Weskamp et al., 2020). It results from alternative splicing that removes most of the C terminus unstructured domain and introduces a functional nuclear export signal (NES) that causes it to localize to the cytoplasm where it can drive the aggregation of endogenous, full length TDP-43 (Weskamp et al., 2020). TDP-43 has also been shown to undergo liquid-liquid phase separation (LLPS) via its LCD, which may facilitate aggregation (Babinchak et al., 2019; Conicella et al., 2020; Pakravan et al., 2021). A recent study characterized a novel mouse model expressing LLPS-deficient TDP-43 (TDPΔCR) caused by deletion of residues 321–340 that represent the conserved region (CR) required for LLPS from within the LCD (Gao et al., 2021). Structure function studies defined residues 99–105 as the TDP-43 translation factor binding motif due to its ability to bind several translation factors and its requirement for enhanced protein synthesis observed in the TDPΔCR mouse model (Gao et al., 2021). Altogether, these findings highlight the importance of RNA binding, RNA sequestration as well as TDP-43 protein aggregation in disease pathology, and outline potential avenues for TDP-43 involvement in translational control.

Under physiological conditions, TDP-43 resides predominantly in the nucleus where it modulates transcription, splicing and miRNA biogenesis (Casafont et al., 2009). Due to its scarce presence outside the nucleus, the extent to which full length TDP-43 performs physiological operations in the cytoplasm remains unclear, considering that the predicted NES does not function as such (Ederle et al., 2018). In numerous neurodegenerative diseases, however, TDP-43 is depleted from the nucleus and mislocalizes to the cytoplasm forming pathological inclusions (Neumann et al., 2006; McAleese et al., 2017). Substantiated by these observations, the mechanism of TDP-43 pathogenesis is postulated as a loss of nuclear function and/or cytoplasmic gain-of-function (Vanden Broeck et al., 2014; Ederle and Dormann, 2017). To test the nuclear depletion hypothesis, early studies using gene knockout methods revealed that loss of TDP-43 is lethal, both in embryonic development and postnatally when inactivated using Cre-inducible gene excision (Kraemer et al., 2010; Sephton et al., 2010; Wu et al., 2012). Other studies focused their attention on recapitulating the cytoplasmic accumulation of TDP-43, demonstrating that overexpression of either the full length wild type protein (Johnson et al., 2008; Li et al., 2010; Liachko et al., 2010; Tsai et al., 2010; Wils et al., 2010; Igaz et al., 2011) or a disease-associated variant (Kabashi et al., 2009; Wegorzewska et al., 2009; Liachko et al., 2010; Zhou et al., 2010; Estes et al., 2011; Huang et al., 2012), is sufficient to cause ALS phenotypes in various animal models. Using other mutant forms of TDP-43 to manipulate its localization and function provided additional support for cytoplasmic toxicity as a likely disease mechanism. For instance, variants of TDP-43 harboring RRM mutations that affect RNA binding do not result in cytoplasmic mislocalization or induce toxicity (Elden et al., 2010; Voigt et al., 2010; Ihara et al., 2013; Coyne et al., 2015). Conversely, deletion of the NLS causes TDP-43 (ΔNLS-TDP-43) to accumulate in the cytoplasm inducing toxicity in primary cortical neurons of rats and neurodegeneration in mice, with rare cytoplasmic inclusions in motor neurons (Barmada et al., 2010; Igaz et al., 2011; Walker et al., 2015; Spiller et al., 2016). Collectively, these animal models support both loss and gain of function mechanisms and highlight the relationship between cytoplasmic accumulation of TDP-43 and neurodegeneration.

As an RNA binding protein, TDP-43 has a broad influence on gene expression, however, its impact on translation in particular is not well understood. Aside from direct alterations to the translation machinery, disturbances in granule dynamics and mRNA localization may contribute considerably to changes in the translatome. As new transcripts are generated in canonical cap-dependent translation (reviewed in Jackson et al., 2010), the growing pre-mRNA strands are available to associate with RNA binding proteins and form highly dynamic ribonucleoprotein (RNP) complexes (Moore and Proudfoot, 2009). The composition of an RNP complex changes with respect to its location and function in RNA metabolism, providing a mechanism for controlling the translational fate of the mRNAs it associates with (Singh et al., 2015). In the cytoplasm, translationally repressed RNP complexes can remodel into higher order structures such as stress granules, neuronal RNA transport granules, and P-bodies (Bowden and Dormann, 2016). Stress granules are a well-studied subset of RNP complexes that can rapidly assemble in response to stress or when translation initiation is inhibited, resulting in sequestration of mRNAs and translation initiation factors (reviewed in Anderson and Kedersha, 2009; Buchan and Parker, 2009). The abundance of highly disordered, prion-like domains in several RNA binding proteins predisposes stress granules to pathogenicity (King et al., 2012; Ramaswami et al., 2013). As proposed in the ribostasis hypothesis, stochastic assembly of RNA binding proteins within stress granules into self-propagating amyloid structures creates aberrant RNP aggregates that disrupt RNA homeostasis eventually leading to cell death (Ramaswami et al., 2013). Disease associated TDP-43 variants (A315T, M337V, Q343R, R361S) have been shown to disrupt stress granule dynamics (McDonald et al., 2011; Liu-Yesucevitz et al., 2014; Gordon et al., 2019). Although recent reports do not support the involvement of stress granules in TDP-43 proteinopathy, or suggest that TDP-43 association with stress granules may in fact be protective (McGurk et al., 2018; Gasset-Rosa et al., 2019; Mann et al., 2019), loss of function studies for the stress granule assembly factor G3BP1 revealed that stress granules facilitate but are not required for TDP-43 aggregation (Fernandes et al., 2020).

Translation can also occur independent of the 5′ cap, driven by the presence of Internal Ribosome Entry Sites (IRESs), initially discovered within viral RNAs and subsequently identified within mRNAs with certain structural features (Pelletier and Sonenberg, 1988). Cap independent translation mechanisms remain less understood compared to canonical translation and rely on RNA binding proteins known as IRES Trans-Activating Factors (ITAFs), which also play a variety of other roles in the cell (Hellen and Sarnow, 2001; Komar and Hatzoglou, 2011). Cap independent translation is prevalent during cellular stress and may play a role in neurological disorders such as ALS/FTD where stress granule dynamics appears to be altered (reviewed in Pelletier and Sonenberg, 1988; Hellen and Sarnow, 2001; Spriggs et al., 2008; Komar and Hatzoglou, 2011; Yang and Wang, 2019).

The ribostasis hypothesis posits that cytoplasmic accumulation of TDP-43 leads to mRNA sequestration and translation inhibition. In support of this hypothesis, various reports indicate that TDP-43 may act as a global repressor of translation. Indeed, TDP-43 depletion in HEK cells by siRNA causes a global increase in translational yield (Freibaum et al., 2010; Fiesel et al., 2012). Corroborating these findings, recent studies show that overexpression of cytoplasmically restricted TDP-43 (ΔNLS-TDP-43) in HEK or SHSY5Y cells causes a global reduction in protein synthesis, as demonstrated by Surface Sensing of Translation (SUnSET) assays that quantify puromycilation of nascent proteins (Russo et al., 2017; Charif et al., 2020). Further substantiating these findings in vivo, overexpression of cytoplasmic TDP-43 in the forebrains of transgenic mice (ΔNLS-TDP-43) results in decreased global translation as evidenced by both SUnSET and polysome profiling (Charif et al., 2020). Interestingly, puromycin incorporation experiments in TDPΔCR mice and primary neurons revealed an increase in global protein synthesis. Since ribosomal protein levels were unaltered, the authors concluded that ribosome assembly was increased (Gao et al., 2021).

TDP-43-induced changes in translation may occur through TDP-43 association with components of the protein synthesis machinery (Freibaum et al., 2010). For instance, induction of short-term oxidative stress in HeLa cells causes TDP-43 to shift from non-ribosomal fractions to monosomal fractions where it associates with stalled ribosomes via mRNA binding (Higashi et al., 2013). Following longer recovery time after arsenite induced stress, TDP-43 associates temporarily with polyribosomes indicating its ability to act both as a negative and as a positive regulator of translation (Higashi et al., 2013). Subsequent reports that TDP-43 co-migrates with non-ribosomal fractions and polyribosomes in Drosophila models of TDP-43 proteinopathy lend support to the notion that TDP-43 also associates with the translation machinery in vivo and may play a pivotal role in the dysregulation of translation in disease (Coyne et al., 2015). Consistent with these findings, polysome profiling in a motor neuron-like cell line or primary cortical neurons transfected with human TDP-43 (hTDP-43) shows that in the absence of stress, TDP-43 fractionates with polyribosomes and shifts to lighter RNP fractions upon polysome disruption with EDTA treatment (Neelagandan et al., 2019). Further analyses using ribosome profiling confirmed that TDP-43 associates with polyribosomes in actively translating cells (Neelagandan et al., 2019). Furthermore, ribosome footprinting in primary cortical neurons identified specific mRNA translational targets of TDP-43 (Neelagandan et al., 2019; see more in “Specific Candidate Targets,” below and in Table 1).

These studies suggest that TDP-43 may play a direct role in translational control, perhaps by direct association with ribosomes. In SHSY5Y neuroblastoma cells, TDP-43 associates with ribosomes via RACK1, a WD40 scaffold protein that binds the 40S ribosomal subunit near the mRNA exit channel and functions as a docking site for several translation machinery proteins (Gallo and Manfrini, 2015; Russo et al., 2017). Immunostaining of mouse hippocampal neurons indicates that TDP-43 puncta strongly colocalize with RACK1 containing granules that correspond to ribosomes/polyribosomes within neurites. Furthermore, RACK1 colocalizes with TDP-43 in pathological inclusions in ALS patient spinal cords, further substantiating the notion that TDP-43’s association with ribosomes may promote TDP-43 aggregation and neurodegeneration (Russo et al., 2017). Several reports highlight TDP-43 dependent alterations in translation that may be indirect, resulting from impaired cellular homeostasis. When used in mouse or Drosophila models of TDP-43 proteinopathy, Translating Ribosomes Affinity Purification (TRAP) revealed complex changes in several targets and pathways within the in vivo motor neuron translatome, including RNA metabolism and cytoplasmic translation itself (MacNair et al., 2016; Lehmkuhl et al., 2021). Additionally, TDP-43 overexpression in Drosophila has been shown to cause the phosphorylation of eukaryotic initiation factor-2alpha (eIF2alpha), an indicator of stress granule formation and consequently, translational repression (Kim et al., 2014). Taken together, these findings support a role for TDP-43 in translation, mediated at least in part through direct association with ribosomes via RACK1 protein.

Although TDP-43 appears to play a critical role in modulating RNP granule dynamics and mRNA localization, identifying specific translational targets of TDP-43 remains paramount in the development of novel therapeutics. Early reports identified Rac1 and futsch/Map1b, key regulators of neural plasticity, as specific mRNA targets of TDP-43 (Majumder et al., 2012; Coyne et al., 2014). Rac1 is a known positive regulator of AMPAR-mediated spinogenesis and is shown to be upregulated upon depletion of TDP-43 in mouse hippocampal neurons (Majumder et al., 2012). Consistent with these findings, TDP-43 has been shown to cooperate with FMRP to co-repress the translation initiation of Rac1, GluR1, and Map1b in dendrites (Majumder et al., 2016). In the context of TDP-43 proteinopathy modeled by TDP-43 overexpression, neuronal activation-induced dissolution of post-synaptic RNPs containing Map1b, GluR1, and CamkII is impaired and accompanied by reduced translation as evidenced by ribopuromycylation assays (Wong et al., 2021). In Drosophila, overexpression of TDP-43 (wild type or the disease associated variant G298S) in motor neurons represses the translation of futsch, as evidenced by a reduction in Futsch protein expression at the NMJ and a shift in futsch mRNA from actively translating polysome fractions to non-translating RNPs (Coyne et al., 2014). Dysregulation of futsch mRNA reduces microtubule stability at the neuromuscular junction, which is mitigated by restoring futsch levels using overexpression in the context of TDP-43 proteinopathy (Coyne et al., 2014). Another mRNA target with functional consequences on synaptic physiology is hsc70-4 mRNA (Coyne et al., 2017). More specifically, TDP-43 impairs the translation of hsc70-4 mRNA due to its sequestration by mutant TDP-43^G298S, as evidenced by soluble vs. insoluble and polysome fractionations (Coyne et al., 2017). Upon restoration of hsc70-4 levels via overexpression, synaptic vesicle endocytosis measured using FM1-43 dye uptake is rescued, suggesting that dysregulation of hsc70-4 translation mediates TDP-43 toxicity at the neuromuscular synapse (Coyne et al., 2017). To identify specific mRNAs transported by TDP-43, in vitro systematic evolution of ligands by exponential enrichment (SELEX) demonstrated that RNAs bound by TDP-43 contained G-quadruplexes and were transported to distal neurites along with TDP-43 for local translation, consistent with G-quadruplexes being novel structural motifs within RNAs associated with TDP-43 (Ishiguro et al., 2016).

Additional translational targets of TDP-43 have recently been reported. In aged mice expressing mutant TDP-43^A315T, TRAP and microarray analyses revealed seven genes that are differentially expressed. Of these, four candidates (Ddx58, Ccl4, Prickle4, and Mthfsd) were confirmed by immunofluorescence in mouse spinal cord motor neurons. Immunohistochemistry experiments in patient derived spinal cords confirmed that DDX58 and MTHFSD, both of which are RNA binding proteins, are differentially expressed in ALS compared to controls (MacNair et al., 2016). Another study characterized changes to the motor neuron translatome in TDP-43^A315T transgenic mice, specifically at the onset of motor symptoms (Marques et al., 2020). By comparing transgenic mice with wild type littermates and asymptomatic mice hemizygous for the wild type hTDP-43 transgene (Chat bacTRAP; hTDP-43^WT), translational changes specifically associated with disease could be distinguished. Candidate mRNAs Syngr4 and Plekhb1 were up- and downregulated, respectively, at the transition from asymptomatic to early symptomatic motor dysfunction, as identified by TRAP and RNA sequencing approaches in spinal cord motor neurons (Marques et al., 2020). Validation experiments confirmed that SYNGR4 and PLEKHB1 protein levels were dysregulated as predicted. Interestingly, these targets were altered in two different mouse models of TDP-43 proteinopathy (TDP-43^A315T and TDP-43^Q331K), suggesting that they may play critical role in the progression of mutant TDP-43 driven ALS (Marques et al., 2020).

Moreover, overexpression of TDP-43 (wild type and A315T) in motor neuron-like cells and in primary cortical neurons enhanced the translation of Camta1 and Mig12 mRNAs through 5′UTR binding. Conversely, translation of Dennd4a was enhanced specifically by mutant TDP-43^A315T through the 3′UTR (Neelagandan et al., 2019). Importantly, independent studies in cultured ALS motor neurons and an in vivo murine model of Parkinson’s disease identified CAMTA1 and DENND4A as “master regulators” of transcriptional programs in neurodegenerative disease (Brichta et al., 2015; Ikiz et al., 2015).

A more recent study using RNA immunoprecipitation assays, TRAP, and bioinformatics analyses, reported the glypican Dally like protein (Dlp) as a novel TDP-43 candidate target, based on the enrichment of dlp mRNA in TDP-43 complexes and depletion from ribosomes in the context of TDP-43 proteinopathy (Lehmkuhl et al., 2021). This study further shows that surprisingly, while Dlp expression in synaptic terminals at the neuromuscular junction is significantly reduced, Dlp protein accumulates in puncta within the ventral cord neuropil suggesting that a combination of translation and transport defects are at play, and that these effects may be compartment specific. Interestingly, TDP-43 knock-down by RNAi (TBPH^RNAi) was sufficient to deplete Dlp from the NMJ, but not induce significant ventral cord neuropil puncta, highlighting that TDP-43 nuclear depletion and cytoplasmic accumulation have distinct contributions to disease pathomechanism. Restoring Dlp by overexpression, in motor neurons specifically, restores Dlp expression at the synaptic terminal and mitigates TDP-43-dependent locomotor deficits. Importantly, these findings align with ALS patient data indicating that GPC6 protein, a human homolog of Dlp, accumulates in puncta within the spinal cord, mimicking the Dlp accumulations in the Drosophila ventral cord neuropil (Lehmkuhl et al., 2021). Additional targets identified in Drosophila using TRAP include metabolic pathways (e.g., pentose phosphate, nuclear encoded components electron transport chain components, oxidative stress). Of note, overexpressing glucose 6 phosphate dehydrogenase (G6PD), which is predicted to be downtranslated in the context of TDP-43^G298S, mitigates locomotor deficits in Drosophila models and highlights a role for metabolic rewiring in degenerating motor neurons (Lehmkuhl et al., 2021).

Interestingly, several recent studies have identified mRNA targets of TDP-43 that encode distinct populations of proteins, such as ribosomal proteins (RPs), mitochondrial proteins, and translation factors (Nagano et al., 2020; Altman et al., 2021; Gao et al., 2021; Lehmkuhl et al., 2021). For instance, cytoplasmic TDP-43 caused a specific decrease in nuclear-encoded mitochondrial proteins, particularly ATP5A1, Cox4i1, and Ndufa4 (Altman et al., 2021). Notably, mRNA levels of these three proteins were unchanged and even moderately increased in TDP-43ΔNLS samples, suggesting TDP-43 accumulation in axons directly impairs local translation of nuclear encoded mitochondrial proteins (Altman et al., 2021). TDP-43 depletion has also been shown to reduce the number and functionality of mitochondria in MN axons, and subsequent treatment with nicotinamide (NAM), a precursor for nicotinamide adenine dinucleotide (NAD +), rescues axon growth defects and capacity for protein synthesis (Briese et al., 2020). In agreement with these findings, depletion of TDP-43 in cortical neurons causes a specific downregulation of ribosomal protein mRNAs in neurites that coincides with axonal extension defects, which are rescued upon overexpression of RP mRNAs Rp141, Rp126, or Rps7 (Nagano et al., 2020). In addition to translational changes in mitochondrial and ribosomal protein mRNA, several translation factors, namely PABPC4, PABPC1, RPS6, EEF1A1, and RPL7, have been shown to co-precipitate with wild type TDP-43 and are enriched in TDPΔCR mutant mice (Gao et al., 2021). Taken together, these findings highlight a role for TDP-43 in regulating various cellular pathways including synaptic metabolism, cellular energetics, and cytoplasmic translation (Marques et al., 2020; Nagano et al., 2020; Altman et al., 2021; Lehmkuhl et al., 2021).