An ancient polymer lost by cancer cells, and the organoid assay that catches it

What the work claims

This is a primary mechanistic preprint. Its central claim is that the inorganic polymer polyphosphate (polyP), which accumulates to high levels at the plasma membrane of many primary cells, is a structural component of filopodia: actin-rich protrusions that mediate cell-cell adhesion, contact inhibition and chemotaxis. The authors argue polyP scaffolds IRSp53, the inverse BAR domain protein whose oligomerization bends membrane to initiate filopodia, promoting its clustering and phase separation. Raising polyP (by expressing the bacterial polyP kinase EcPPK1) stabilizes filopodia and increases adhesion; depleting polyP (with a degron-controlled yeast exopolyphosphatase, ScPPX) accelerates filopodial disassembly and promotes migration¹.

The translational claim: metastatic fibroblasts and breast cancer organoids carry markedly reduced and intracellularly reorganized polyP compared with non-transformed counterparts, and restoring it via lipid-nanoparticle (LNP) delivery suppresses invasive phenotypes and reverses pro-metastatic gene expression signatures, implicating polyP loss as a new molecular hallmark of metastatic cells¹.

How it works

The mechanistic chain is unusually complete for a preprint. In NIH3T3 fibroblasts, EcPPK1 expression raised polyP and increased the number and length of MyoX-positive filopodia with no change in steady-state ATP, ruling out an energy-metabolism confound; the effect reproduced in human U2OS cells and in COS-7 cells that normally form lamellipodia. Live imaging showed EcPPK1 filopodia barely change length over 5 minutes, and the filopodial F-actin dissociation rate slowed at least threefold, from about 3.5 minutes to more than 11 minutes. The converse manipulation, Shield-stabilized ScPPX, cut cellular polyP within 24 hours and reduced average filopodia number by roughly 40 percent while shortening persistence times. Membrane-associated polyP co-localized with IRSp53 and CDC42, recruited purified IRSp53 into clusters on synthetic giant unilamellar vesicles, induced IRSp53 phase separation in vitro, and proximity-ligation and cryo-electron tomography placed polyP along the filopodial axis with a denser membrane-to-actin connector network when polyP was high¹.

For the cancer arm, the team turned to organoids from two genetically engineered mouse models of aggressive breast cancer: MMTV-PyMT, whose expression profile clusters with human luminal B disease, and C3(1)-Tag, a basal-like triple-negative model. Organoids embedded in fibroblast growth factor-supplemented collagen I invade robustly over 5 to 7 days. Tumor organoids from either model showed strongly reduced polyP, most pronounced at the plasma membrane, versus healthy mammary organoids (n = 3 mice per experiment). A 24-hour pretreatment with polyP-300-loaded LNPs at two loading doses significantly raised endogenous polyP, and after 7 days of FGF-stimulated invasion the treated MMTV-PyMT organoids were more than 50 percent less invasive by inverse circularity, with near-identical results in C3(1)-Tag organoids. RNA sequencing at day 0 and day 7 (n = 4 per treatment and time point) showed coherent reversal: oxidative phosphorylation genes rose, while epithelial-to-mesenchymal transition, apical-junction disorganization and migration programs fell¹.

A plausible reason tumors lose polyP: the polyphosphatases NUDT3 and H-Prune1 are strongly upregulated in aggressive breast cancers by DepMap analysis, and both enzymes' cancer roles have previously been attributed to other mechanisms. The authors propose polyP loss itself, not just the enzymes, as the causal event¹.

Where a skeptic should push

The most load-bearing assumption is that the membrane polyP-IRSp53 interaction is the causal driver of the invasion phenotype, rather than a correlate. The evidence is correlative at the structural level: no structure of the polyP-IRSp53 interface exists, the authors have not anchored how polyP sits at the membrane, and inositol phosphates, which seed polyP synthesis and touch dozens of pathways, are an acknowledged confound the paper cannot exclude¹.

Second, the therapeutic leap. There is no in vivo metastasis experiment here at all: no mouse treated with polyP-LNPs and challenged for lung colonization. The organoid invasion assay, however elegant, is a proxy. The molecule itself is a problem: polyP-300 is a distribution of chain lengths, not a defined molecular entity, which is a manufacturing and regulatory headache, and two recent colon-cancer studies found the opposite effect, with exogenous polyP increasing tumor proliferation, a discrepancy the authors attribute to delivery route but have not resolved. Finally, note the declared conflict: the senior authors hold a pending patent application on LNP-mediated polyP delivery. That does not invalidate the data, but it prices the enthusiasm¹.

The authors' own limitations statement concedes the decisive gap: before polyP depletion can be called a unifying feature of metastatic cancer, someone must survey a broad panel of patient-derived organoids, measuring polyP against matched non-tumor tissue and testing whether LNP supplementation suppresses invasion. That survey does not exist in this paper.

polyP as a target organoid invasion assays can score

Strip away the therapy framing and the durable contribution is an assay. The 3D invasion readout used here, tumor organoids embedded in FGF-supplemented collagen I scored by inverse circularity after 5 to 7 days, is cheap, morphological, and directly reports the adhesion-versus-invasion decision the paper mechanizes. Anti-invasive compounds are poorly served by standard viability screens, which rank killing power, not restraint of movement; this format is exactly the kind of mechanistically anchored, image-based endpoint that organoid drug-discovery platforms can run at scale, and the polyP-LNP rescue arm shows how a positive-control rescue should be built into such screens¹.

The generalization warning is the flip side. Both tumor models are mouse; both agree, which is reassuring but is two points in genotype-and-stroma space, not a property of breast cancer. Human patient-derived organoids differ in matrix dependence, stromal content and invasion strategy, so the numerical efficacy, and possibly the direction of the effect, must be re-derived in human panels before this becomes a screen anyone should trust for candidate selection. The same caution applies to LNP co-development: LNP tropism for metastatic deposits is unproven, and the field has burned itself before on mouse-matrix results that evaporated in human tissue. The honest near-term play is to use polyP biology as a mechanistic probe and the invasion organoid as the readout, while treating LNP-polyP as what it is today: a patented, preclinical, heterogeneous-polymer delivery concept with zero in vivo efficacy data¹.

The bottom line

Established: polyP stabilizes filopodia through IRSp53 scaffolding, with bidirectional genetic control across several cell systems; metastatic mouse mammary tumor organoids deplete membrane polyP; and LNP-mediated polyP restoration cuts 3D invasion by more than half while reversing pro-metastatic transcription. Hypothesis: polyP loss is a general driver of metastasis and its restoration is therapeutic. What would confirm it: a patient-derived organoid survey measuring polyP against matched normal tissue, structural definition of the polyP-IRSp53 interface, and an in vivo metastasis experiment with LNP delivery. What would break it: human PDO panels that fail to show polyP depletion or fail to respond to supplementation, or in vivo studies showing no effect on metastatic burden.