Organoids from cervical precancer flag NF1 as the gatekeeper: losing it yields tumors in 42 days, and an AI-designed peptide cuts lesions in mice
Synopsis
High-grade squamous intraepithelial lesions of the cervix usually take years to possibly become invasive cancer, and the molecular gate restraining that transition was unclear. Combining patient-derived precancerous organoids, single-cell transcriptomics and CRISPR-Cas9 selection in vitro and in vivo, the study pins NF1 as the top suppressor of malignant transition in HPV-positive precancerous organoids: NF1-deficient organoids formed tumors within 42 days under the skin of immunodeficient mice, while control organoids failed to engraft. AlphaFold3-guided modeling then produced Pep-NF1, an NF1 GAP-mimetic peptide that lowered cervical lesion-associated bioluminescence in an HPV16-driven mouse model alongside local immune changes.
Interpretation
Malignant transition of cervical precancer has a molecular constraint that functional screening can locate, and losing NF1 is enough for HPV-positive precancerous organoids to acquire malignant phenotypes within the experimental window. Recurrent genomic alterations in cervical cancer such as PIK3CA, PTEN and KRAS did not explain where precancerous epithelium becomes irreversibly malignant, and the absence of faithful precancer models made that step hard to test directly. In four HPV-positive high-grade squamous intraepithelial lesion organoid lines without pathogenic NF1 mutations by whole-genome sequencing, a CRISPR-Cas9 library targeting 222 candidate genes was carried through 90 days of in vitro passage and through tumor selection in NCG mice; NF1 showed higher beta-scores in tumor-derived samples than in matched long-term in vitro cultures across all four lines and ranked as the top positively enriched hit among all screened genes.
NF1 loss pushes precancerous epithelium to expand by raising both RAS-MAPK and PI3K-AKT signaling and by opening proliferation-associated chromatin regions. Single-cell trajectories had already shown RAS pathway activity rising with lesion stage, but that was correlation and did not say what restrains the pathway. In NF1-knockout versus matched control precancerous organoids, ATAC-seq with three samples per group yielded 8,906 high-confidence peaks and 7,137 differentially accessible regions, of which 7,082 had higher signal in the NF1-deficient group and the gained regions were enriched for AP-1-related motifs; RNA-seq of the same model showed upregulation of MKI67, EGFR, FGFR2, DUSP6 and TP63, and a GST-Raf-RBD pull-down showed NF1-GRD expression reduced GTP-bound RAS.
Pep-NF1, an NF1-GAP-mimetic peptide designed with AI structural modeling, showed antitumor activity in patient-derived tumor organoids and in an HPV16-driven mouse model. Full-length NF1 is large and structurally complex, so it cannot simply be replaced or restored, which motivates compressing its tumor-suppressive interface into a deliverable molecule. AlphaFold3 in multimer mode reconstructed the NF1-RAS-GTP complex and identified Pep-NF1 as the candidate with the most favorable predicted binding energy, and the modeled interface retained direct and water-mediated contacts over a 400-nanosecond molecular dynamics simulation; Pep-NF1 reduced viability of patient-derived cervical tumor organoids in a dose- and time-dependent manner at 2, 5 and 10 micromolar, with marked reduction within 48 hours at 5 micromolar or more; in HPV16/K14 transgenic mice dosed intraperitoneally at 20 mg/kg three times weekly from day 60 to about day 90, lesion-associated bioluminescence fell relative to PBS and irrelevant-peptide controls.
Perspective
The demonstrated scope is patient-derived cervical precancerous and cancerous organoids, cervical cancer cell lines, HPV16/K14 transgenic mice, and tissue microarray specimens from cervix, esophagus and lung; the organoid and mouse models capture epithelial-intrinsic processes and cannot reproduce the long-term evolutionary pressures or stromal complexity of human disease.
No recurrent pathogenic NF1 mutations appeared in the precancerous organoids, so functional attenuation may come from transcriptional or post-transcriptional regulation, which remains undetermined; whether the collapse of tumor organoids after NF1 restoration reflects cell-state reversion or preferential elimination of malignant cells cannot be distinguished without longitudinal single-cell profiling after reexpression; pharmacokinetic optimization of Pep-NF1 and validation in immune-competent and humanized models remain to be done.
