TDP-43

Amyotrophic lateral sclerosis (ALS) is a progressive adult-onset neurodegenerative disorder of upper corticospinal motor neurons (Betz cells in the primary motor cortex) and lower somatic α-motor neurons in the brainstem and spinal cord ventral horn, sharing a clinicopathological spectrum with frontotemporal lobar degeneration (FTLD). Although over 30 genes harbor causal or high-risk variants for familial (~10–15%) and sporadic (~85–90%) ALS—led in prevalence by the C9orf72 (GGGGCC)_n hexanucleotide repeat expansion, Cu/Zn Superoxide Dismutase 1 (SOD1 on Chr21q22.11), Transactive Response DNA-Binding Protein 43 kDa (TDP-43, encoded by TARDBP on Chr1p36.22), and FUS—neuropathological stratification separates the disease into two distinct molecular entities:

1. The TDP-43 Proteinopathy Majority (>97% of all ALS cases and ~45% of FTLD cases): Across virtually all sporadic ALS, C9orf72-ALS/FTD, GRN and TBK1 cases, and TARDBP-mutant ALS, endogenous TDP-43 is depleted from the cell nucleus and sequestered into hyperphosphorylated (pSer409/410), ubiquitinated cytoplasmic skein-like or dense round inclusions.
2. The TDP-43-Negative Exceptions (~2–3% of total ALS, comprising SOD1-ALS and FUS-ALS): Motor neurons from patients carrying pathogenic mutations in SOD1 (~15–20% of familial ALS) or FUS (~4% of familial ALS) accumulate cytoplasmic misfolded SOD1 or FUS aggregates while retaining normal nuclear TDP-43 localization.

Recent transcriptomic and cryo-EM discoveries have demonstrated that loss of nuclear TDP-43 splicing repression—rather than cytoplasmic aggregate toxicity alone—acts as a primary driver of motor neuron death by unmasking unannotated intronic cryptic exons and premature polyadenylation sites inside essential neuronal transcripts, most notably Stathmin-2 (STMN2 on Chr8q21.13) and Unc-13 Homolog A (UNC13A on Chr19p13.11).

Tandem RRM Domains and the C-Terminal Prion-Like Low-Complexity Domain

TARDBP (12.9 kb, 6 exons) encodes TDP-43, a ubiquitously expressed 414-amino-acid (44.7-kDa) heterogeneous nuclear ribonucleoprotein (hnRNP) that shuttles between the nucleus and cytoplasm but resides predominantly (>90%) inside the nucleus under physiological conditions. Its modular domain architecture comprises four functional regions:

1. The N-Terminal Domain (NTD, residues 1–76) and Bipartite Nuclear Localization Signal (NLS, residues 82–98): The globular ubiquitin-like NTD mediates reversible head-to-tail physiological homopolymerization (via Leu71, Val72, and Trp68), which enhances pre-mRNA splicing efficiency while sterically separating the aggregation-prone C-terminal tails of adjacent monomers. Following the NTD, a bipartite basic NLS binds Importin-α/β (KPNA/KPNB1) to maintain nuclear import.
2. Tandem RNA-Recognition Motifs (RRM1, residues 106–176, and RRM2, residues 191–262): Folded into canonical βαββαβ globular domains, RRM1 and RRM2 bind single-stranded RNA and DNA with sub-nanomolar to low-nanomolar affinity (K_d ≈ 2–10 nM) and strict sequence specificity for long alternating uridine-guanosine ((UG)_n, n ≥ 6) dinucleotide tracts, which occur across >6,000 intronic sites in the human transcriptome (~30% of all human pre-mRNAs).
3. The Intrinsically Disordered C-Terminal Glycine-Rich Prion-Like Domain (LCD, residues 274–414): Enriched in uncharged polar glutamine, asparagine, serine, and glycine residues alongside spaced aromatic phenylalanines and tryptophans (plus a transient α-helical segment spanning residues 320–340), this low-complexity domain drives reversible liquid–liquid phase separation (LLPS) into nuclear ribonucleoprotein granules (gems, paraspeckles) and cytoplasmic stress granules.

Over 95% of the >70 autosomal dominant ALS-causing missense mutations in TARDBP—including Ala315Thr (A315T), Gln331Lys (Q331K), Met337Val (M337V), Gly298Ser (G298S), and Ala382Thr (A382T)—cluster within Exon 6 encoding the C-terminal prion-like LCD. These mutations lower the thermodynamic barrier for a liquid-to-solid phase transition, converting dynamic LLPS condensates into irreversible cross-β amyloid-like fibrils (whose cryo-EM core in ALS/FTLD brain spans residues 282–360 in a double-spiral-shaped fold). Furthermore, TDP-43 tightly autoregulates its own mRNA levels via a negative feedback loop: when nuclear TDP-43 concentration rises, it binds (UG)_n repeats inside the 3′-UTR (TDPBR region) of its own TARDBP pre-mRNA, promoting alternative splicing of Intron 6 and Intron 7 that targets the transcript for nonsense-mediated decay (NMD). Once cytoplasmic aggregates trap newly synthesized TDP-43 in the cytosol, nuclear TDP-43 depletion relieves 3′-UTR autorepression, driving a self-reinforcing surge of TARDBP mRNA translation that feeds the cytoplasmic sink.

STMN2 Exon 2a Truncation and Axonal Regeneration Failure

Under physiological conditions, nuclear TDP-43 binds (UG)_n repeat tracts inside introns to act as a splicing repressor, sterically preventing the spliceosome (U2AF2/U2 snRNP) from recognizing weak, unannotated "cryptic" 3′ and 5′ splice sites. When TDP-43 is cleared from the nucleus in ALS motor neurons, dozens of intronic cryptic exons are spliced into mature mRNAs, either introducing premature termination codons that trigger nonsense-mediated decay (NMD) or generating truncated, non-functional proteins.

In 2019, independent RNA-sequencing studies identified Stathmin-2 (STMN2, encoding the neuronal growth-associated protein SCG10) as the single most abundant and severely downregulated transcript in TDP-43-depleted human motor neurons and post-mortem ALS spinal cords:

1. Normal Splicing in the Presence of Nuclear TDP-43: TDP-43 binds a 24-nt tract containing three tandem (GU)_n motifs located inside Intron 1 of STMN2, immediately downstream of a cryptic 3′ splice acceptor site (Exon 2a). This binding blocks Exon 2a recognition, ensuring that Exon 1 splices properly to canonical Exon 2 (producing full-length 179-amino-acid STMN2, which is palmitoylated at Cys22/Cys24 and transported to axonal growth cones to regulate microtubule dynamic instability after axonal injury).
2. Cryptic Exon 2a Splicing and Polyadenylation upon TDP-43 Loss: When nuclear TDP-43 falls by >50%, the spliceosome splices Exon 1 directly into cryptic Exon 2a (227 bp), which contains an in-frame stop codon after only 16 amino acids and an internal canonical AAUAAA polyadenylation signal. This premature cleavage and polyadenylation truncates the STMN2 transcript (producing a tiny 1.5-kDa non-functional peptide), depleting full-length STMN2 protein by 70–95% and paralyzing the ability of motor axons to regenerate neuromuscular junctions (Agrin–MuSK synapses). Crucially, steric-blocking 2′-O-methoxyethyl (2′-MOE) antisense oligonucleotides (such as QRL-201) that hybridize to the Exon 2a cryptic splice/polyadenylation site completely restore full-length STMN2 expression and axonal outgrowth in TDP-43-deficient human motor neurons.

UNC13A Cryptic Exon Insertion and GWAS Risk SNP Epistasis

A second critical target of TDP-43 splicing repression solved one of the longest-standing enigmas in ALS human genetics. For over a decade, genome-wide association studies (GWAS) identified two common intronic single-nucleotide polymorphisms (SNPs) inside Intron 20 of UNC13A—rs12608932 (A>C) and rs12973192 (C>G)—as top genetic risk factors for sporadic ALS (OR ≈ 1.25–1.35) and modifiers of shorter survival (reducing median survival by 6–12 months), yet neither SNP altered protein sequence or baseline gene expression in healthy brain tissue.

In 2022, two landmark studies revealed that TDP-43 normally binds a UG-rich sequence inside UNC13A Intron 20 (between Exon 20 and Exon 21) to repress a 128-bp (or 178-bp) cryptic exon. When nuclear TDP-43 is depleted in ALS or FTLD neurons, this cryptic exon is spliced into the mature UNC13A mRNA, introducing a premature stop codon that triggers nonsense-mediated decay (NMD) and depletes UNC13A (Munc13-1), an essential presynaptic active-zone protein required for SNARE-mediated synaptic vesicle priming and neurotransmitter release. Crucially, the risk SNP rs12973192(G) sits directly inside the cryptic exon (and rs12608932(C) lies 34 bp downstream of the cryptic donor site), where the C→G substitution disrupts a core UG-repeat binding motif, reducing TDP-43 binding affinity by ~2.5- to 4-fold. Consequently, in healthy neurons with abundant nuclear TDP-43, the risk allele remains silent; however, as soon as nuclear TDP-43 begins to decline in early ALS, UNC13A pre-mRNAs carrying the rs12973192(G) risk allele lose TDP-43 repression first, accelerating cryptic exon inclusion, synaptic failure, and patient mortality.

Dimer Destabilization, Metal Loss, and Tofersen RNase H1 Knockdown

In contrast to TDP-43-positive ALS, approximately 15–20% of familial ALS (and ~1–2% of sporadic ALS) is caused by >200 predominantly autosomal dominant missense mutations distributed across all five exons of SOD1 (Chr21q22.11), which encodes the ubiquitous 154-amino-acid (32-kDa homodimeric) cytosolic/intermembrane-space Cu/Zn Superoxide Dismutase. Each SOD1 monomer folds into an 8-stranded Greek-key β-barrel that coordinates one catalytic copper ion (Cu²⁺/Cu⁺ ligated by His46, His48, His63, and His120) and one structural zinc ion (Zn²⁺ ligated by His63, His71, His80, and Asp83), stabilized by an intramolecular disulfide bond between Cys57 and Cys146.

Rather than causing disease via loss of superoxide dismutase enzymatic activity (many pathogenic mutants such as G93A and D90A retain nearly 100% wild-type enzymatic activity, and Sod1 knockout mice do not develop spontaneous motor neuron disease), SOD1 mutations—including the rapidly fatal North American founder variant Ala4Val (A4V, median survival 1.1–1.4 years), Gly93Ala (G93A), His46Arg (H46R), and Asp90Ala (D90A)—confer a toxic gain of function by thermodynamically destabilizing the native homodimer interface (ΔΔG ≈ −2 to −5 kcal/mol), impairing Cu/Zn metalation (apo-SOD1), and reducing the Cys57–Cys146 disulfide bond. Monomeric, demetalated SOD1 exposes hydrophobic β-strands and free cysteines (Cys6, Cys111) that nucleate misfolded oligomers and prion-like cross-β fibrils that damage outer mitochondrial membranes (binding BCL-2 and VDAC1) and impair axonal transport, while nuclear TDP-43 remains intact (and STMN2/UNC13A cryptic exons are absent).

Because SOD1-ALS toxicity is driven strictly by mutant protein abundance, reducing SOD1 mRNA synthesis directly targets the root cause. Tofersen (BIIB067 / Qalsody) is a 20-nucleotide gapmer antisense oligonucleotide (ASO) administered via lumbar intrathecal injection that hybridizes to the 3′-UTR of both wild-type and mutant human SOD1 mRNA, recruiting nuclear and cytoplasmic RNase H1 to cleave the transcript. In the Phase III VALOR trial and open-label extension, tofersen lowered cerebrospinal fluid SOD1 protein by ~35% and reduced plasma neurofilament light chain (NfL)—a quantitative biomarker of active axonal degeneration—by 55–60%, stabilizing or improving muscle strength in patients treated early and leading to accelerated FDA approval in April 2023.

How can TDP-43 cryptic exon splicing be detected in living ALS patients as a fluid biomarker?

While cryptic exons in STMN2 and UNC13A lead to rapid mRNA truncation or nonsense-mediated decay (NMD), several other TDP-43-repressed cryptic exons—such as those inside HDGFL2 (hepatoma-derived growth factor-like 2) and ACTL6B—are spliced in-frame without triggering NMD. Consequently, ribosomes translate the cryptic exon into a novel cryptic neo-peptide that never exists in healthy individuals. Ultrasensitive immunoassays (MSD/Simoa) using neo-epitope antibodies against cryptic HDGFL2 detect TDP-43 loss of function in cerebrospinal fluid and blood even in presymptomatic C9orf72 carriers prior to symptom onset.

Why does intermediate polyglutamine expansion in Ataxin-2 (ATXN2) increase ALS risk through TDP-43?

Normal ATXN2 alleles (Chr12q24.12) contain 22 or 23 CAA-interrupted CAG glutamines, whereas expansions ≥34 CAGs cause Spinocerebellar Ataxia Type 2 (SCA2). Intermediate-length expansions of 27–33 glutamines occur in ~4.7% of sporadic ALS cases (OR ≈ 3.0–5.0). Ataxin-2 is an RNA-binding protein (LSm and PAM2 domains) that localizes to stress granules and interacts directly with TDP-43 in an RNA-dependent manner; intermediate polyQ tracts enhance stress-granule condensation, promoting TDP-43 pSer409/410 phosphorylation and cytoplasmic mislocalization.

Why are SOD1-ALS and FUS-ALS cases negative for TDP-43 pathology and STMN2 cryptic exons?

Both SOD1 and FUS mutations trigger motor neuron degeneration through pathways that sit either downstream of or parallel to nuclear TDP-43. Misfolded SOD1 damages mitochondrial bioenergetics and ER-Golgi trafficking directly, while FUS mutations (P525L, R521C in the C-terminal PY-NLS) selectively disrupt Transportin-1 (TNPO1)-mediated nuclear import of FUS (without impairing Importin-α/β nuclear import of TDP-43), causing a distinct RNA-processing defects and DNA repair failure while leaving nuclear TDP-43 intact.

1. Neumann, M., Sampathu, D. M., Kwong, L. K., et al. (2006). "Ubiquitinated TDP-43 in frontotemporal lobar degeneration and amyotrophic lateral sclerosis." Science, 314(5796), 130–133. https://doi.org/10.1126/science.1134108
2. Melamed, Z., López-Erauskin, J., Baughn, M. W., et al. (2019). "Premature polyadenylation-mediated loss of stathmin-2 is a hallmark of TDP-43-dependent neurodegeneration." Nature Neuroscience, 22(2), 180–190. https://doi.org/10.1038/s41593-018-0293-z
3. Ma, X. R., Prudencio, M., Koike, Y., et al. (2022). "TDP-43 represses cryptic exon inclusion in the FTD–ALS gene UNC13A." Nature, 603(7899), 124–130. https://doi.org/10.1038/s41586-022-04424-7
4. Miller, T. M., Cudkowicz, M. E., Genge, A., et al. (2022). "Trial of antisense oligonucleotide tofersen for SOD1 ALS." New England Journal of Medicine, 387(12), 1099–1110. https://doi.org/10.1056/NEJMoa2204705