Cryptic splicing as a graded readout of TDP-43 dysfunction: mechanism, biomarkers, and a proposal for threshold-resolved staging across ALS, frontotemporal dementia, and LATE

Nuclear depletion of the RNA-binding protein TDP-43 is the shared molecular lesion of a disease spectrum far broader than the one in which it was discovered. It defines the neuropathology of more than ninety per cent of amyotrophic lateral sclerosis (ALS) and roughly half of frontotemporal lobar degeneration (FTLD), and it is present in some form in close to forty per cent of the brains of people who die past the age of eighty, where it is termed limbic-predominant age-related TDP-43 encephalopathy (LATE). Among the consequences of losing nuclear TDP-43, one has proved uniquely tractable: the de-repression of cryptic splice sites, normally silent sequences that TDP-43 occludes and that, on its departure, are spliced into mature messenger RNA. Since the founding observation in 2015, cryptic splicing has moved from a curiosity to the most direct, quantitative and specific molecular signature of TDP-43 loss of function available in human tissue, and to the target of at least four antisense oligonucleotide programmes now in or entering clinical trials. This review synthesises that literature: the cell biology of TDP-43 nuclear clearance; the catalogue of cryptic events and the evidence that they are drivers rather than passengers; their distribution across ALS, FTLD-TDP, LATE and Alzheimer's disease; and the therapeutic and biomarker programmes they have generated. It then argues that the field's central unresolved question is quantitative rather than qualitative. Independent titration experiments show that different cryptic events appear at different degrees of TDP-43 depletion — STMN2 at approximately a third, UNC13A at more than half, UNC13B at more than ninety per cent — implying that the pattern of cryptic events in a cell encodes the depth of TDP-43 loss, not merely its presence. Three confounds currently prevent that inference: nonsense-mediated decay masks frameshifting events and may generate apparent thresholds artefactually; published titrations quantified TDP-43 messenger RNA rather than protein; and bulk measurements cannot distinguish a graded response within cells from a binary switch across a distribution of cells. A three-aim research programme is proposed to resolve these. Aim 1 separates splicing thresholds from detection thresholds using titrated TDP-43 depletion under nonsense-mediated decay inhibition with protein-level and nascent-transcript readouts. Aim 2 resolves graded versus binary behaviour at single-neuron resolution using targeted long-read and amplicon-based single-nucleus sequencing of sorted human neuronal nuclei. Aim 3 translates a validated event hierarchy into a multiplexed, threshold-resolved biofluid panel and tests it against autopsy-confirmed pathology and as a pharmacodynamic readout. If cryptic splicing is graded, it supplies what the TDP-43 field has never had and what amyloid and tau imaging gave Alzheimer's disease: a staging variable.