From Evasion to Collapse: The Kinetic Cascade of TDP-43 and the Failure of Proteostasis

TDP-43 is a highly conserved protein made up of 414 amino acids. Its full length is organized into four major domains: (1) an N-terminal domain (NTD), that spans approximately 1–80 residues; (2) two RNA Recognition Motifs (RRMs, at residues 105–169 and 194–257, respectively), which are a characteristic feature of heterogeneous nuclear ribonucleoprotein (hnRNP) proteins responsible for RNA recognition and protein interactions; and (3) a highly disordered glycine-rich C-terminal domain (CTD) or prion-like domain (residues 277–414) (Figure 1) [21,22]. Other studies refer to the CTD as the glycine-rich region (GRR) or the low-complexity domain (LCD). LCDs in RNA-binding proteins (RBPs) are typically enriched in a small subset of amino acids (e.g., G, Q/N, S, Y, F, R), which reduces sequence diversity and results in “low sequence complexity.” The absence of side-chains in glycine confers flexibility, making these domains key regulators of protein phase behavior, especially in scenarios involving regulation and stress adaptation [23].

The CTD is unstructured and contains most of the documented ALS-associated mutations, thus it is considered a significant driving force of protein assembly [24,25]. Studies have revealed that the liquid droplet environment, electrostatic repulsion, the presence of CTD segments that act as steric zippers, and low-complexity aromatic-rich kinked segments (LARKS) that can be converted to irreversible interactions by fALS variants are all located in the CTD and play a common role in increasing the rate of TDP-43 aggregation [26,27]. As for the other TDP-43 domains, molecular modeling revealed different segments from the NTD (²⁴GTVLLSTV³¹), RRM1 (¹²⁸GEVLMVQV¹³⁵), and RRM2 (²⁴⁷DLIIKGIS²⁵⁴) as aggregation-prone and were confirmed to have amyloid-fibril propensities through various biophysical techniques and molecular dynamics simulations [28]. The simulations revealed that the RRM2 peptides transitioned from random coil to α-sheet-rich structures, forming parallel and anti-parallel α-sheets, even in the absence of aromatic residues, underscoring RRM2 as a critical structural region for TDP-43 aggregation [28].

Complementing these earlier findings on aggregation-prone regions, Kumar et al. applied proteolytic cleavage to recombinant TDP-43 filaments, revealing a hidden core sequence within the CTD (residues 279–360), which was found to be necessary for the propagation of TDP-43 proteinopathy [22]. Notably, this core sequence resembled that described by Arseni et al. when they isolated TDP-43 pathological filaments from the brains of ALS/FTLD-TDP (type A) patients, a subtype of FTLD (the other subtype is FTLD-tau) [29]. This core sequence (residues 282–360) had a double-spiral shape fold, whereas the former (in vitro) was buried and flanked by the NTD and RRMs (Figure 1) [22,29]. The findings suggest that the cellular environment and flanking sequences are important contributors to the final filament orientation. These differences likely originate from non-proteinaceous cofactors found in filament cavities and disease-specific PTMs (e.g., citrullination and monomethylation of R293), which have been shown to promote specific amyloid folds and stabilize structural variations in the human brain [22,29]. The emergence of these differences in fold structure, despite primary sequence similarities, underscores the importance of assessing full-length proteins in their native contexts.

Although the CTD of TDP-43 is intrinsically disordered and predisposed to aggregation, Afroz et al. (2017) [33] showed that the NTD also interacts to form physiological oligomers under stress. The formed TDP-43 oligomers result in the spatial separation of the CTD ends, minimizing the chances of contact between these regions and thereby reducing aberrant aggregation [33]. This suggests that TDP-43 has evolved an intrinsic self-regulatory mechanism to counteract its own tendency to aggregate. These findings were supported in a subsequent paper by Jiang et al. (2017), which showed that the NTD is primarily responsible for the formation of homodimers in solution, and further tetramerization occurs through intermolecular disulfide bonds [34]. More importantly, they showed that mutations destabilizing the NTD dimer significantly enhanced the formation of cytoplasmic inclusions, indicating that NTD dimerization is essential in protecting against aggregation [34].

Another study by Wang et al. (2018) [35] further demonstrated the critical role of NTD in enhancing TDP-43 polymeric assembly and liquid-liquid phase separation (LLPS), where proteins and RNA condense into dynamic, droplet-like compartments. Since the interaction of NTD is much weaker and less spontaneous than CTD, the authors hypothesized that TDP-43 phase separation is favored through the synergistic cooperation of NTD and CTD interactions [35]. Solubility shifts in TDP-43 constructs made of NTD alone, CTD alone, and full-length TDP-43 were compared using turbidity assays and microscopic observation of droplet formations. The LLPS of full-length TDP-43 was found to be more pronounced than that of CTD alone, suggesting the contributory role of NTD interaction sites [35]. As for NTD alone, the construct does not readily undergo LLPS but instead forms weak, reversible head-to-tail polymers in a linear arrangement. Most importantly, introducing a phosphomimetic substitution (S48E) in the NTD disrupted head-to-tail assembly, thereby weakening and raising the concentration threshold for LLPS, despite the CTD’s inherent ability to phase separate, further validating the synergistic involvement of the NTD in phase separation [35]. These findings, together with more recent work showing that RNA binding and macromolecular assembly regulate TDP-43 LLPS behavior in vivo, support the idea that phase separation is a tightly controlled, sequence-encoded property, rather than an inevitable prelude to aggregation [35,36,37]. Recent studies also showed that intact RNA binding can buffer TDP-43 against aberrant LLPS and aggregation, whereas RNA-binding-deficient or acetylation-mimicking mutants undergo distinct, aggregation-prone phase behavior [16,38].

While the cooperation between the NTD and the CTD underscores how TDP-43 regulates the functional assembly of oligomers with pathological aggregation, neuropathological findings reveal that this balance is not maintained equally across disease contexts, as evidenced by the absence of TDP-43 inclusions in certain forms of ALS. Specifically, Tan et al. found TDP-43 inclusions in the brains and spinal cords of sporadic ALS (sALS) and non-mutant superoxide dismutase 1 (SOD1) familial ALS (fALS), but not in two mutant SOD1 cases [39]. Subsequent clinicopathological studies have shown that ALS resulting from FUS mutations likewise exhibits FUS-positive, TDP-43-negative inclusions [40,41]. Their findings were corroborated by Mackenzie et al., who found TDP-43 inclusions in sALS and non-SOD1 fALS neurons, but not in SOD1 fALS tissue [36,42]. A question was then put forward as to what was making TDP-43 inclusions common in sALS and not as easily observable in fALS tissues [36]. This prompted another group to explore if TDP-43, on its own, was prone to aggregation or sequestered by other aggregated components, thus simply making it a “marker of disease” [43].

In their study, Johnson et al. demonstrated that recombinant TDP-43, without interacting cofactors, rapidly formed aggregates after a lag phase of approximately 5–10 min, as indicated by increased solution turbidity and a corresponding increase in pellet size following centrifugation [43]. Identical conditions were applied to control proteins (e.g., BSA, soybean trypsin inhibitor, creatine kinase, and GFP), but no aggregation was observed, indicating that TDP-43 is inherently prone to aggregation [43]. Using a yeast mutant model, the same group showed that reported ALS mutations (e.g., Q331K, M337V, Q343R) accelerated aggregation in vitro and increased the formation of aggregates in vivo.

We acknowledge that many of the studies discussed in this review rely on cellular, C. elegans, mouse, and zebrafish models. Interspecies differences in TDP-43 sequence, splicing targets, lifespan, and proteostatic capacity are likely to shift the threshold at which a second hit occurs, but do not fundamentally change the underlying cascade logic. Moreover, human neurons, particularly long-lived projection neurons, exhibit distinct vulnerabilities that could further accentuate specific stages of the cascade, such as clearance failure.

In summary, these findings highlight that TDP-43 is inherently prone to aggregation, with its sequence and domains already predisposing it to self-assembly. Consequently, even in its native state, the protein exhibits oligomerization hotspots that could ultimately manifest as pathological inclusions. This underscores why the protein is heavily dependent on stringent and continuous quality control. Two distinct degradatory and clearance pathways help clear aberrant, misfolded proteins and the aggregates they form. However, several factors can disrupt their principal components, which will be discussed in the succeeding sections.

The degradation of aberrant TDP-43 depends on two central quality control systems: (1) the Ubiquitin-Proteasome System (UPS), which degrades soluble, ubiquitinated monomers, and (2) the Autophagy-Lysosome Pathway (ALP), which sequesters larger aggregates in double-membrane vesicles for lysosomal fusion and protease-mediated degradation [44]. For larger, aggregated proteins, the ALP takes over, assisted by adaptor proteins such as sequestosome 1 (p62) and NBR1, which bind to polyubiquitin chains and simultaneously interact with LC3 [44]. Polyubiquitin tags are not limited to UPS, but are also recognized by the ALP via p62 and NBR1 [45,46,47].

While ubiquitin tagging can guide proteins toward either proteasomal or autophagic degradation, the exact functional partitioning between the two systems remained unclear for TDP-43. Scotter et al. (2014) explored this gap, ultimately demonstrating that the UPS processes soluble TDP-43, whereas the ALP handles aggregates to facilitate their clearance [48]. More importantly, they found that cellular macroaggregates of TDP-43 are reversible when both systems are functional, and further postulated that a “second hit” in either of these two systems drives the pathological cascade of TDP-43 aggregation in ALS and FTD [48]. In their study, stable cell lines were generated in which expression of different HA-tagged TDP-43 variants could be induced with doxycycline. The expressed constructs were: (1) wild-type (HA-TDP-43-WT), (2) ΔNLS-truncated mutant (HA-TDP-43-ΔNLS), and (3) C-terminal fragment (aa 181–414) (HA-TDP-43-CTF). Both the localization and expression of the constructs were then observed following treatment with autophagy inhibitors (e.g., 3-methyladenine (3MA), bafilomycin (Baf)) or UPS inhibitors (e.g., MG132 and epomoxin) [48].

The results revealed a distinct separation of TDP-43 processing between the two degradation systems. Autophagy was shown to be the leading player in clearing TDP-43 CTFs, whereas the UPS primarily targeted soluble WT/ΔNLS TDP-43 [48]. Inhibiting UPS resulted in the pronounced accumulation and aggregation of soluble TDP-43 species. The researchers used a doxycycline (DOX)-inducible system in SH-SY5Y and HEK293 cells to express HA-tagged TDP-43 [48]. Applying MG-132, a reversible UPS inhibitor, resulted in the formation of large, cytoplasmic, and ubiquitin-positive aggregates [48]. This was followed by a 48 h washout period to remove the MG-132, and end-point imaging revealed that the cells were devoid of TDP-43 and p62 aggregates, even when TDP-43 expression continued via DOX induction [48]. The cells were then immediately transferred to fresh medium containing 3MA (which blocks the initial stages of autophagy) and bafilomycin (Baf; which prevents the fusion of autophagosomes with lysosomes) to block autophagy upon restoration of UPS function [48]. The authors then observed that macroaggregates still disassembled into smaller, discrete particles (microaggregates), demonstrating that the initial fragmentation of larger aggregates is an autophagy-independent step. Nevertheless, in the presence of autophagy inhibitors, these microaggregates persisted in the cells. Conversely, the activation of autophagy with trehalose or lithium chloride (LiCl) enhanced CTF degradation in separate experiments, validating its principal role in clearing TDP-43 aggregates [48]. When these findings are viewed holistically, they build the following picture: (1) degradation of soluble TDP-43 species is mainly via the UPS, (2) once protein oligomerization is initiated, these species will have become too large to enter the proteasome pore, requiring the intervention of another pathway (i.e., autophagy) [48]. Notably, the study demonstrates that only when both systems are functional can the cells completely reverse TDP-43 macroaggregate formation, further substantiating the cooperative interplay between the UPS and autophagy pathways. An important limitation worth noting here is the biological cost of clearance, where 69.3% of aggregate-laden cells died during the 15 h clearance period [48]. Reversibility was only observed in the minority of cells that managed to survive the initial aggregate burden.

These findings are consistent with previously published work, which shows that different degradation pathways process distinct TDP-43 species. Specifically, Wang et al. (2010) showed that the truncated form of TDP-43 (TDP-25) was more reliant on the autophagy pathway [49]. Furthermore, Urushitani et al. (2010) revealed the synergistic effect of the proteasome and autophagosome in clearing polyubiquitinated TDP-43, corroborating the dual-pathway requirement [50]. Additionally, proteasome inhibition has been shown to drive the cytoplasmic accumulation and aggregation of neuronal TDP-43, supporting the central role of the UPS in degrading soluble TDP-43 species [51]. Conversely, the role of autophagy in clearing truncated or aggregation-prone species was demonstrated through the rapamycin-mediated activation of the pathway, resulting in the rescue of mislocalized TDP-43 and reduction in CTF accumulation in the cytosol [52]. The tight, cooperative interplay between the UPS and ALP systems is critical for TDP-43 proteostasis, providing the mechanistic rationale for how a ‘second hit’ to either system triggers the pathological cascade observed in ALS and FTLD.

The domains of TDP-43 are indeed critical in influencing its folding behavior, particularly the highly disordered CTD. Other domains, such as the NTD, were demonstrated to have the capability of counteracting this intrinsic disorder through specific structural arrangements that provide spatial separation of the CTD, thereby attenuating uncontrolled folding and aggregation. However, certain conditions, such as the presence of mutations, can enhance the stability of its folding, potentially shifting the balance toward more aggregation-prone conformations.

Particularly, mutations (G298S, Q331K, M337V) in the CTD introduce structural modifications in its primary sequence, which potentially lead to impairment in lysosomal or proteasomal degradation, as well as favoring irreversible aggregation over normal LLPS, thereby decreasing TDP-43’s susceptibility to degradation [53,54,55,56,57]. An early study by Ling et al. (2010) [58] explored the effect of these CTD mutations on the half-life of TDP-43. Here, tetracycline-inducible wild-type and mutant genes were expressed using a cytomegalovirus (CMV) promoter in isogenic cell lines. By randomly cycling the cells through brief incubation with [³⁵S]methionine/cysteine, newly synthesized proteins were radiolabeled and subsequently monitored over time to measure their half-lives [58]. The results showed that mutant proteins were degraded two to four times more slowly than wt-TDP-43 [58]. Moreover, immunoprecipitation results indicated stronger interaction between endogenous wt-FUS/TLS and TDP-43 mutants, suggesting that mutation-induced CTD stabilization not only affects folding but also enhances cross-interaction with FUS/TLS [58].

The structural disruptions resulting from these mutations (e.g., A321G, A321V, Q331K, M337V) were analyzed in detail using NMR spectroscopy, simulation, and microscopy, revealing a CTD subregion that transiently folds into a helix, thereby mediating TDP-43 phase separation [54]. The mutations disrupt this phase separation by inhibiting the interaction and destabilizing the α-helical structure. Notably, A321G and Q331K eliminated the phase separation of the CTD, while M337V decreased solution turbidity by approximately 50% [54]. Removing the region containing the α-helical structure and the mutations (Δ321–343), and the helix-disrupting mutation (A326P) resulted in the loss of phase separation, suggesting the participation of this helical region is essential in LLPS [54]. At low concentrations, the truncated CTD as well as the mutant CTD remained soluble but easily formed insoluble aggregates when their concentrations were increased, demonstrating that this helical region is necessary for stabilizing the liquid-like state of TDP-43 [54]. Zeng et al. (2024) similarly showed the disruptive effect of M337V in the same region with their own simulations, corroborating the above results [56]. Taken together, these findings suggest that residues 321–343 form a critical region that promotes and maintains the reversibility of TDP-43 LLPS through its intermolecular helix-helix interactions with the same region of other TDP-43 molecules. However, mutations within this segment disrupt these interactions, destabilizing LLPS and thereby shifting the equilibrium toward irreversible aggregation.

So far, we have established that mutations in the CTD are a disruptive force that favors the formation of irreversible aggregates. However, some mutations paradoxically render TDP-43 more vulnerable to protease-mediated cleavage. Investigations into the role of calpains in promoting the cleavage of mutant TDP-43 proteins were prompted by the hypothesis that the downregulation of adenosine deaminase acting on RNA 2 (ADAR2) enzymes enhances the generation of TDP-43 pathology [30]. ADAR2 enzymes convert the adenosine found in the Q/R site of GluA2 (the protein subunit of α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid or AMPA receptors) pre-mRNA to inosine, facilitating the impermeability of AMPA receptors to calcium ions (Ca²⁺) [59,60]. However, the motor neurons of patients with sporadic ALS have reduced levels of ADAR2, resulting in leaky AMPA receptors due to the decreased processing of GluA2 pre-mRNA [61,62]. The reduced ADAR2 leads to increased uptake of Ca²⁺, which can be neurotoxic. The increased Ca²⁺ uptake likewise prompted Yamashita et al. to investigate the proportional effect of Ca²⁺-dependent proteases on TDP-43 pathology, and they found an increased generation of C-terminal fragments (CTFs), with A315T and M337V TDP-43 mutants being cleaved faster than wild-type TDP-43 (Figure 1) [30]. The group further demonstrated that knocking out AR2 in mice led to TDP-43 mislocalization and cleavage, resulting from calpain activation triggered by Ca²⁺ AMPA receptors. In contrast, AR2res (AR2-restored) mice, which expressed Ca²⁺-impermeable AMPA receptors even in the absence of AR2, showed no TDP-43 pathology or calpain activation, confirming that Ca²⁺ influx is the trigger of TDP-43 pathology [30].

A separate study by Chiang et al. (2016) demonstrated that the D169G mutation exhibits higher thermal stability compared to wild-type TDP-43 and is cleaved by caspase-3 more efficiently, resulting in the production of more CTFs (TDP-35) in neuroblastoma cells [31]. In vitro, caspase-3 cleaves the D169G mutant approximately two-fold more efficiently than wt-TDP-43 (40.3 ± 3.8% vs. 20 ± 3.4% cleaved N RRM12 after 2 h) [31]. The D169G mutation is located in the RRM1 domain and induces a local conformational change in a α-turn, resulting in increased hydrophobic interactions within the RRM1 core (Figure 1) [31]. The resulting more rigid conformation possibly exposes more caspase-3 cleavage sites to the solvent, thereby enhancing cleavage efficiency. Based on these findings, the effects of TDP-43 cleavage appear paradoxical at first glance; cleavage is generally expected to be a biologically favorable process that facilitates the clearance of TDP-43 deposits from cells. However, this is not always the case; the cleavage site and the properties of the resulting fragments are critically important, as specific cleavage fragments can, in fact, exacerbate aggregation. Specifically, the generation of CTFs has been shown to promote aggregation due to their increased tendency to self-associate and nucleate TDP-43 oligomers [63,64,65,66,67,68]. In this case, cleavage is not inherently beneficial; however, the type and properties of the resulting fragments dictate the likelihood of their clearance or persistence within the cytoplasm.

As established in the section above, the UPS is the primary pathway for processing soluble TDP-43 monomers. The UPS and the autophagy-lysosome pathways have distinct roles, and the latter cannot compensate for the failure of the former to process TDP-43 soluble species. This functional specialization means that the persistence of pathological TDP-43 in the cytoplasm of affected neurons is a critical indication that the UPS is somehow overwhelmed or circumvented in disease states. Proteasomal evasion, therefore, is not just a consequence of aggregate size exceeding the 20S core aperture but involves specific molecular factors that impede efficient substrate recognition and digestion. The failure of the UPS is linked primarily to two factors that direct the toxic shift towards TDP-43 oligomerization and aggregation: (1) intrinsic substrate characteristics that disrupt ubiquitination, and (2) extrinsic factors that compromise proteasome function.

Building on what was introduced earlier, protein processing by the UPS can occasionally be inaccurate and exacerbated by the stressors mentioned above. Hence, misfolded TDP-43 oligomers that slip past the UPS could coalesce into larger aggregates that the ALP processes. Autophagy is the cell’s main bulk degradation pathway, which sequesters large, insoluble aggregates, damaged organelles, and aberrant proteins into specialized vesicles called autophagosomes, which then fuse with lysosomes, organelles that contain hydrolytic enzymes for degradation (Figure 2) [97,98,99]. Given its high processing capacity, the ALP should, in theory, be capable of processing and clearing the protein aggregates that escape the UPS. However, the unopposed formation of TDP-43 aggregates is associated with the dysfunction of the ALP at several critical stages, resulting from a failed recognition, impaired influx, and compromised downstream clearance, which will be discussed next.

TDP-43 normally resides and performs its regulatory functions within the nucleus. Still, a small fraction of its concentration is regularly shuttled to the cytoplasm to perform functions in RNA metabolism and regulation of translation [11]. In the nucleus, the RRM domains of TDP-43 bind with high affinity, stabilizing the protein and maintaining it in a soluble state [37,122,123]. One study demonstrated that the removal of the RRM domains using site-directed mutagenesis significantly reduced TDP-43 levels within the nucleoplasm, underscoring the valuable role of RRM domains in TDP-43 nuclear localization [123]. This cooperative binding produces a higher-order assembly on mRNA, facilitating the reduction in intermolecular interactions between the aggregate-prone NTD and CTD, which could result in aggregation [122].

However, once TDP-43 exits the nucleus, multiple factors further reduce its RNA-binding capacity. PTMs, such as the acetylation of lysines (e.g., K136 and K145), disrupt RNA binding and enhance phase separation through the CTD, leading to the formation of insoluble aggregates [10,124]. Acetylation mimicking mutants (K145Q) exhibited insoluble, hyperphosphorylated, and ubiquitinated aggregates [21]. Another mutation (K181E) found in the RRM-linking region also reduced RNA-binding affinity and enhanced aggregate formation [125]. These findings confirm that the solubility and conformational integrity of TDP-43 are significantly dependent on sustained RNA engagement.

Once RNA engagement is weakened, the resulting consequence significantly contributes to the increased formation of TDP-43 aggregates [72,126]. Additional studies have described the formation of anisotropic intranuclear liquid droplets exhibiting HSP70 chaperones, which translates to a state of pathological assembly [127]. Furthermore, RNA-binding-deficient TDP-43 was shown to undergo demixing in stress granules (SGs), where the TDP-43-rich phase has significantly reduced RNA levels compared to the G3BP1-rich phase [128]. G3BP1 (Ras GTPase-activating protein-binding protein 1) functions as a scaffolding protein that initiates stress granule assembly. Together with other RBPs and RNAs, it contributes to the formation of a dynamic, liquid-like phase that constitutes the structural matrix of normal SGs. TDP-43 is also included in this phase, but prolonged oxidative stress and the buildup of misfolded TDP-43 result in its “demixing” from the G3BP1 matrix, creating its own subphase within the granule [128]. Thus, the observed loss of RNA in the TDP-43 phase corroborates that maintaining TDP-43 solubility highly depends on its RNA-binding capability.

For additional context and to avoid confusion with regular liquid-liquid phase separation, intra-condensate demixing is an additional phase separation event within an existing biomolecular condensate, in which a previously uniform droplet spontaneously segregates into compositionally distinct subphases, such as a TDP-43-enriched core within a stress granule [128]. This additional step is thermodynamically favorable because it allows for maximizing the newly acquired interactions of the TDP-43 molecules while minimizing unfavorable contacts with other stress-granule components, lowering the overall free energy of the condensate [128]. This also increases the local abundance of reactive oligomeric species that can react with functional proteins and contribute to cellular toxicity. More details on the toxic species of TDP-43 are discussed in the following section.

Additionally, Yan et al. (2025) [128] concluded that although the formation of SGs increased TDP-43 concentration above a critical threshold, this alone was insufficient for aggregation. Conversely, when prolonged oxidative stress resulting in cysteine oxidation and structural destabilization within the RRM domain (and increased hydrophobic interactions in the CTD) was the only variable treated, but TDP-43 levels failed to reach the critical threshold in SGs, no aggregation was observed [128]. Therefore, both variables (the crucial concentration of TDP-43 in SGs and prolonged oxidation) are essential for the phase separation of TDP-43 to occur within the granules and form irreversible aggregates with pathological hallmarks.

Mechanistically, RNA binding increases the stability of the RRM domains and inhibits the conformational plasticity of the protein. Molecular dynamics simulations demonstrated that the absence of RNA promotes the tandem RRMs to exhibit enhanced conformational plasticity, particularly RRM1, which displays an increased likelihood of adopting partially unfolded conformations [129]. The interaction with RNA introduces constraints to TDP-43, increasing intradomain stability and inhibiting the structural transitions that lead to uncontrolled aggregation [129].

Moreover, single-molecule tracking revealed a significant reduction in TDP-43 mobility with increasing stress duration, where the effective diffusion coefficients dropped from a baseline of 3 μm²/s to 0.75 μm²/s after two hours of arsenite-induced stress [130]. This was observed for both free cytoplasmic TDP-43 and TDP-43 sequestered into SGs, suggesting that oligomeric hotspots, which could potentially lead to irreversible aggregates, may result from cytoplasmic TDP-43, not just those already sequestered into SGs.

Finally, the cytoplasm fundamentally lacks the quality control mechanisms that maintain TDP-43 homeostasis and solubility in the nucleus, fostering an environment that predisposes TDP-43 to misfolding and aggregation. Nuclear TDP-43 is delicately retained by forming large RNA-mediated macromolecular complexes that block its cytoplasmic diffusion in a size-dependent manner [37]. These nuclear compartments are also stabilized by RNA binding, oligomerization, and phase separation, where TDP-43 is organized into compartments that maintain its solubility and functionality [131].

Studies have also demonstrated that passive diffusion is the primary force driving TDP-43 cytoplasmic egress, where the protein continuously leaks into the cytoplasm but is actively shuttled back to the nucleus through nucleoporins and import factors [5]. Investigation of the interactome of TDP-43 insoluble aggregates revealed their enrichment in the components of the nuclear pore complex and nucleocytoplasmic machinery, indicating the direct sequestration of these essential karyopherins, leading to the impairment of the nuclear pore complex [38,132]. The result is a feed-forward mechanism in which the increased levels of aberrant cytoplasmic TDP-43 progressively sequester more components, disrupting nuclear import and leading to further accumulation of cytoplasmic TDP-43.

Another study revealed that the monomerization of TDP-43 is key in driving the nuclear export of the protein. Monomeric TDP-43 was demonstrated to preferentially interact with nuclear RNA export factor 1 (Nxf1), compared to its dimeric form [133]. Additionally, the monomerization was shown to be triggered by cellular stresses or dysfunctions (i.e., spliceosomal impairment), compared to physiological conditions, where TDP-43 dimerizes or forms higher-order oligomers to facilitate its nuclear retention and solubility [133].

Taken together, reduced affinity to RNA due to structural mutations or PTMs, reaching critical concentration thresholds within SGs (demixing), prolonged oxidative stress, the absence of quality control mechanisms in the cytoplasm, and the resulting vicious cycle induced by the sequestration of critical importins by the increased egress of aberrant TDP-43 are all significant drivers that produce a self-reinforcing cascade that drive irreversible aggregation.

The propensity of TDP-43 to naturally form aggregates has been described, with its early phase characterized as an oligomeric species that is stable, spherical, and amyloid-like, featuring exposed hydrophobic surfaces [43,82]. These oligomers were reported to have reduced nucleic-acid binding and direct neurotoxicity in neuronal models [82]. Regions in TDP-43 that facilitate its self-aggregation were identified, and candidate peptides targeting these regions were designed to reduce TDP-43 aggregation. Despite identifying two constructs that successfully reduced aggregation in a dose-dependent manner, the reduction in aggregation did not prevent cell death [134]. However, it must be noted that the authors considered only observable aggregates; thus, it is possible that cell death could mainly stem from the toxicity of soluble oligomers, which have much more exposed hydrophobic regions that can interact with functional cellular proteins relative to stable aggregates [135]. Moreover, liquid-like RNP granules in axons become more viscous and less dynamic with ALS-linked mutations, disrupting granule transport and inducing the toxic gain-of-function effect [136]. Finally, cytoplasmic liquid droplets that formed independently of classical SGs were reported to gel up or solidify under stress, which is enough to disrupt nuclear import and deplete nuclear TDP-43, resulting in cell death [38,130]. Given these findings, primary toxic TDP-43 species appear to arise from oligomeric and gel-like assemblies rather than end-stage fibrils, thereby combining direct proteotoxicity with the disruption of RNA metabolism and nucleocytoplasmic transport.

CTFs are also species that play a significant role in TDP-43 toxicity, as they have been characterized as able to seed full-length TDP-43. On their own, they are capable of inducing cell death through a toxic gain-of-function without sequestering their full-length counterpart [65], as well as disrupting proper proteasomal assembly [84]. Together, these findings support a conceptual framework in which soluble oligomers, viscous RNP granules, and CTFs act as the principal toxic species. At the same time, large aggregates may be later consequences that correlate with, but do not strictly determine, neuronal death. Simply put, pathogenicity may be best explained as a combination of misfolded assembly-mediated gain of function and depletion of functional TDP-43 within the nucleus.

Now that we have established the key molecular players involved in the mislocalization of TDP-43, we will discuss more factors that further accelerate the formation of TDP-43 aggregates once the initial stages have been fulfilled. These factors are critical to understand, as they substantially increase the rate of TDP-43 entanglement. Several steps within this process may also serve as therapeutic targets to slow or interrupt the pathological cascade.

The apparent paradox, where TDP-43 aggregates are predominantly cytosolic despite their steady-state nuclear concentration being much higher, can be clarified by considering the different buffering capacities of these compartments. In the nucleus, high-affinity RNA binding and the formation of large ribonucleoprotein assemblies, nuclear bodies, and anisotropic intranuclear droplets (anisomes) maintain TDP-43 in dynamic, liquid-like states, and keep its concentration of free, aggregate-competent monomers low [38,127]. By contrast, once monomerization and RNA disengagement enable TDP-43 entry into the cytoplasm, the protein reaches an environment with weaker RNA buffering, fewer dedicated chaperone elements, and more chronic stressors, driving phase separation in stress granules and demixed cytoplasmic droplets that could rapidly solidify into irreversible aggregates [154].

In this view, intranuclear anisomes represent a high-density but still regulated ‘buffer’ for RNA-free TDP-43, whereas the cytoplasm functions like a sink where the quality-control capacity is lower, nucleocytoplasmic transport is progressively impaired, and aggregates become trapped as hyperphosphorylated, ubiquitinated inclusions [38].

No new data were created or analyzed in this study. Data sharing is not applicable to this article.