Using artificial intelligence to find new treatments for cystinosis

Nephropathic cystinosis is a rare lysosomal storage disorder caused by loss of CTNS, leading to cystine accumulation in lysosomes. This disrupts lysosome–autophagy function and drives early proximal tubule (PT) dysfunction, clinically presenting as Fanconi syndrome with low-molecular-weight (LMW) proteinuria and progression to chronic kidney disease with multisystem involvement. Cysteamine lowers cystine and delays kidney failure, but PT dysfunction often persists and long-term use is limited by tolerability and dosing burden. There is therefore a clear need for therapies that improve PT homeostasis downstream of cystine storage.

To accelerate therapy development, I applied an AI-assisted platform integrating multi-omics and curated knowledge to prioritize approved drugs predicted to restore lysosomal competence and PT function in cystinosis. Among top candidates, I prioritized palbociclib (CDK4/6 inhibitor) based on ranking, mechanistic plausibility, and clinical tractability, while explicitly incorporating known safety liabilities (notably neutropenia) into an exposure-guided strategy.

Preliminary validation in CTNS-deficient primary mouse PT cells and zebrafish indicates robust target engagement at non-cytotoxic doses and improved PT epithelial and lysosomal readouts without significant changes in wholeorganism cystine, consistent with a downstream-of-storage mechanism of action.

Two gaps must be resolved to support a realistic repurposing decision: (i) how CDK4/6 inhibition improves lysosomal competence and PT function in cystinosis, and (ii) whether benefit can be achieved within a paediatric-compatible exposure window. I will address these gaps through two aims:

Aim 1 (Mechanism): Define how palbociclib rescues CTNS-deficient PT cell homeostasis and function by separating proliferation-linked effects from lysosome-intrinsic improvements. I will test TFEB/TFE3-dependent lysosomal biogenesis and TFEB/TFE3-independent lysosome reformation (or combinations) and distinguish CDK4 versus CDK6 contributions to the rescue phenotype.

Aim 2 (Translation): Establish an exposure-aligned efficacy and safety window in mammalian cystinosis models with targeted validation in patient-derived kidney cells. Using urinary CC16 as the primary endpoint (with LMW proteinuria as a clinically anchored PT readout), I will identify the minimal exposure that achieves PT target engagement with acceptable toxicity and test preventive and therapeutic efficacy in CtnsKO rats, including evaluation of additivity or interference with cysteamine.

By combining AI-guided nomination, mechanistic resolution, and exposure-aligned preclinical validation, this project will deliver (i) a mechanistic framework linking CDK4/6 inhibition to improved lysosomal competence and PT homeostasis and (ii) an integrated exposure–PD–efficacy–safety dataset enabling a clear paediatric go/no-go decision. More broadly, the present project establishes a scalable downstream-of-storage strategy in cystinosis and a reusable repurposing-to-validation workflow applicable to other lysosome-related kidney disorders.