TPC2 as a therapeutic target: exploration

Target exploration, 2026-09-26. Follow-on from the chlorpromazine ALS hypothesis investigation, which concluded "the interesting target here is TPC2, not chlorpromazine."

Curated artifacts: controversy_tpc2_direction_of_modulation in kb/modules/lysosomal_substrate_accumulation.yaml, and three added evidence items plus a revised rationale on hyp_als_phenothiazine_autophagy_repurposing in kb/disorders/Amyotrophic_Lateral_Sclerosis.yaml.

The headline, which revises last week's conclusion

"TPC2 is the interesting target" was half right, and the half it got wrong matters.

TPC2's ion selectivity is not fixed — it is set by which ligand opens the channel. A NAADP-mimetic agonist gives non-selective cation currents and strong Ca²⁺ signals; a PI(3,5)P₂-mimetic gives Na⁺-selective currents and weak ones; and the two are "coupled to opposing changes in lysosomal pH and exocytosis" (PMID:32167471). So "a TPC2 agonist" names at least two different drugs.

Once you split the arms, the evidence sorts cleanly, and chlorpromazine lands on the wrong side:

Arm	Compounds	In vivo efficacy
NAADP-mimetic, Ca²⁺-permeable agonism	NAADP, TPC2-A1-N, chlorpromazine	none anywhere
PI(3,5)P₂-mimetic, Na⁺-biased agonism	TPC2-A1-P; the LRRK2 "biased agonist"	MLIV mouse (rotarod, P62, astrogliosis); LRRK2 Drosophila behaviour
Antagonism	trans-Ned-19, tetrandrine, SG-094	rat tMCAO infarct reduction; LRRK2 correction

The chlorpromazine ALS paper benchmarks its evoked current against "TPC2 endogenous agonist NAADP and its mimetic agent TPC2-A1-N", blocked by trans-Ned-19 (PMID:40796055). That is the NAADP arm. Meanwhile the landmark storage-rescue paper — the one that made TPC2 look like a drug target — used TPC2-A1-P throughout: 115 mentions of A1-P, zero of A1-N, cells through iPSC neurons through the MLIV mouse. And in LRRK2 Parkinson's models, "reducing Ca²⁺ permeability with a novel biased TPC2 agonist corrected deviant Ca²⁺ entry and behavioral defects" (PMID:40279672).

So the ALS hypothesis runs through the one mode of the one channel for which no animal has yet been shown to benefit. That is a sharper objection than any in last week's report, and it is also more actionable: it names a different compound to try rather than just a reason to stop.

The direction of benefit is genuinely disputed

Three in vivo results, three directions, same channel:

• Activate. Small-molecule TPC2 activation "results in an amelioration of cellular phenotypes associated with LSDs such as cholesterol or lipofuscin accumulation" across mucolipidosis type IV, Niemann-Pick C1 and Batten patient fibroblasts and iPSC-derived neurons, with rotarod rescue, reduced P62 aggregates and reduced cerebellar astrogliosis in the MLIV mouse (PMID:35929194).
• Inhibit. "The pharmacological inhibition of TPC2 lysosomal channel by Ned-19 protects from focal ischemia by hampering a hyperfunctional autophagy" — reduced infarct volume and neurological deficit after rat tMCAO (PMID:36708960).
• TPC2 is the disease. Expressing human TPC2 in Drosophila, which lack TPCs, "phenocopied LRRK2 G2019S in perturbing dopaminergic-dependent vision and movement in vivo." The authors' own summary: "Thus, both inhibition and select activation of TPC2 are beneficial" (PMID:40279672).

These are reconcilable, and the reconciling principle is not disease-specificity. The stroke paper states it: "the role of autophagy in Stroke is controversial since excessive or prolonged autophagy activation exacerbates ischemic brain injury." If autophagic flux has an optimum, the therapeutic direction depends on which side of it the cell starts — and neither a module node nor a target name records that.

This is the same shape as the pre-existing gap_tfeb_direction_of_dysregulation_in_storage on the storage module, which asks whether TFEB should be raised or normalized. TPC2 sits one step upstream and adds a dimension TFEB does not have: a biased agonist can move it in two pharmacologically distinct directions, not just up or down. The new discussion says so and cross-references the TFEB gap rather than restating it.

What is genuinely attractive about the target

Human loss of function is tolerated. gnomAD v4 for TPCN2 (queried 2026-09-26): pLI ≈ 6.1 × 10⁻²⁶, LOEUF 0.98, 92 observed pLoF variants against 111 expected (o/e = 0.83). Monarch reports zero gene-to-phenotype and zero gene-to-disease associations for HGNC:20820. There is no Mendelian TPCN2 disease. For an antagonist programme that is close to an ideal safety prior — people who lack the channel are walking around undiagnosed.

The antagonist arm is structurally enabled. SG-094, "a synthetic analog of the Chinese alkaloid medicine tetrandrine with increased potency and reduced toxicity", has a solved mechanism: it arrests the IIS4 voltage sensor in a downward state, resembling gating modifiers of canonical voltage-gated channels (PMID:38815576). Tetrandrine itself came out of the Ebola-entry work. This is real medicinal chemistry, not a tool compound.

It is a convergence point across several dismech modules. TPC2 activation rescues diseases conforming to lysosomal_substrate_accumulation; its autophagy role sits on disabled_macroautophagy; and melanosomes are lysosome-related organelles, so the pigmentation phenotype touches lysosome_related_organelle_biogenesis. That is why the curated artifact is a discussion on an existing module rather than a new module — TPC2 dysfunction is not itself an established disease mechanism, it is an intervention point on mechanisms already curated.

What is genuinely unattractive

Both arms have an on-target liability, and they are different liabilities.

Agonism promotes exactly what the antagonist programmes are built to block. TPC2 loss attenuates cancer cell migration and metastasis, impairs choroidal neovascularization and iPSC-derived endothelial tube formation (PMID:42584788), and reduces infectivity for cholera toxin and Ebola virus (PMID:33465068). A chronic systemic TPC2 agonist is a pro-angiogenic, pro-migratory, pro-viral-entry drug by construction.

Antagonism has metabolic and — critically for ALS — neurodevelopmental costs. TPC2⁻/⁻ mice "show defects in cholesterol degradation, leading to hypercholesterinemia; TPC2 absence also results in mature-onset obesity, and a role in glucagon secretion and diabetes has been proposed" (PMID:33465068). And in motor neurons specifically, TPC2 is required: trans-Ned-19 and TPC2 knockout "attenuated CaP Ca²⁺ signaling and inhibited axon extension" in zebrafish caudal primary motor neurons (PMID:32546534), and TPC2 inhibition "decreased the normal ipsilateral correlation and contralateral anti-correlation, indicating a disruption in normal spinal circuitry maturation" (PMID:29577882).

That last pair is the sharpest thing in this exploration. In a motor-neuron disease, the better-developed arm of TPC2 pharmacology acts against the arm of TPC2 biology that motor neurons demonstrably need — using the very same compound. trans-Ned-19 is neuroprotective in stroke, corrective in LRRK2 models, and blocks motor axon extension. Whichever way you push TPC2 in ALS, you pay.

Most people do not carry reference TPC2. "Surprisingly, one variation, L564P, was found to be the predominant TPC2 isoform on a global scale", and L564P is a prerequisite for the blond-hair-associated M484L gain-of-function effect (PMID:33465068). Any construct, screen or potency measurement built on the reference sequence is measuring a minority allele. This is a target-validation hazard that would be easy to miss.

The best agonist is a poor drug. TPC2-A1-P was "rapidly eliminated, being undetectable by 240 min", and 20 mg/kg was chosen "to avoid off-target activity while providing a therapeutic dose for > 20 min." The MLIV mouse rescue was achieved with roughly twenty minutes of daily exposure. That is an encouraging efficacy signal and a discouraging pharmacokinetic one, and it means sustained TPC2 activation has not actually been tested for safety or sufficiency.

The field says so itself. "Despite its potential as a drug target, TPC2 is still in the early stages of therapeutic development", with the major challenge being "achieving high target specificity without inducing unintended effects on other endolysosomal channels" (PMID:39978661). TRPML1/ML-SA1 is the obvious comparator and was run head-to-head in the storage paper.

Loose ends worth knowing about

The gating controversy may have just been resolved. A 2026 preprint reports that NAADP "elicits two-pore channel currents by lifting Lsm12-mediated inhibition of PI(3,5)P₂ activation" (PMID:42039649) — i.e. one activation pathway with a removable brake rather than two independent ligands. If that holds, the two-arm model above is a description of two points on one mechanism rather than two mechanisms, which would change how a biased agonist should be designed. It is a bioRxiv preprint and not peer-reviewed; it is deliberately cited here and not in the KB.

A new regulatory layer. Mucin-type O-linked glycosylation at luminal Ser612/Ser613 acts as a "structural gating brake" — "genetic, enzymatic, or pharmacological glycan removal enhances basal and ligand-evoked TPC2 activity", driving lysosomal tubulation and cancer cell migration reversible by tetrandrine (PMID:42270676). A post-translational set point means TPC2 activity is not inferable from expression, which is worth remembering when reading the one existing dismech mention of TPC2 as a protein-level readout (Dominant_Deafness-Onychodystrophy_Syndrome, where TPC1/TPC2 are reported upregulated in a lysosomal-acidification defect).

Where this leaves the ALS hypothesis

Revised, not abandoned, and the revision is a different first experiment.

Last week's window study asked whether chlorpromazine's TPC2/TFEB engagement is reachable below the motor-toxicity threshold. It should now also read out the PI(3,5)P₂-biased mode, because if motor-neuron TDP-43 clearance tracks that mode instead, the useful compound is TPC2-A1-P or a successor — not an antipsychotic, and none of the dopaminergic window problem applies. That is a strictly better experiment and roughly the same cost.

And the obvious alternative — block TPC2, since that arm is better developed and human LoF is tolerated — is specifically contraindicated in a motor-neuron disease by the zebrafish axon-extension result. Both of those are now recorded on the ALS discussion.

Sources

Twelve TPC2 references cached under references_cache/; all snippets quoted in the KB are exact-quote verified (243/243 in the ALS entry, 23/23 in the module).

PMID	Role
32167471	Ion selectivity is set by the activating ligand; opposing effects on lysosomal pH and exocytosis
35929194	TPC2-A1-P rescues MLIV/NPC1/Batten; MLIV mouse in vivo; PK and ML-SA1 comparison
36708960	trans-Ned-19 inhibition is neuroprotective in rat tMCAO
40279672	TPC2 is pathogenic in LRRK2-PD; both inhibition and biased activation beneficial
40796055	Chlorpromazine's TPC2 current benchmarked against NAADP / TPC2-A1-N
39978661	Drug-development review: early stage, selectivity is the main challenge
38815576	SG-094 / tetrandrine antagonist structural mechanism
33465068	L564P is the predominant global isoform; TPC2⁻/⁻ mouse metabolic phenotypes
42270676	O-glycan gating brake; lysosomal tubulation; migration reversible by tetrandrine
42584788	TPC2 loss impairs choroidal angiogenesis and endothelial tube formation
32546534	TPC2 required for motor axon extension; trans-Ned-19 blocks it
29577882	TPC2 required for spinal circuit synchronization in primary motor neurons
42039649	Preprint, not peer-reviewed. NAADP acts by lifting Lsm12 inhibition of PI(3,5)P₂ activation

Identity and constraint checked 2026-09-26: HGNC:20820 TPCN2, "two pore segment channel 2", 11q13.3, UniProt Q8NHX9, OMIM 612163 (gene). gnomAD and Monarch figures as quoted above.