Direct reprogramming of human fibroblasts into insulin-producing cells using transcription factors

**Background:** Diabetes is characterized by a lack of functional pancreatic β cells, and cell replacement therapy is a promising treatment. Direct lineage reprogramming, converting one differentiated cell type into another without passing through a pluripotent stage, offers a potential source of β-like cells. While transcription factors Pdx1, Neurog3, and MafA (NPM) can convert pancreatic acinar cells into insulin-producing cells, fibroblasts were previously thought resistant to such conversion. This study investigates whether human fibroblasts can be directly reprogrammed into β-like cells using a combination of transcription factors.

**Methods:** Human foreskin fibroblasts (HFF1) were infected with a polycistronic adenovirus encoding Neurog3, Pdx1, MafA, and Cherry (Ad-NPM). After optimization, cells were cultured in RPMI-1640 with 6% FBS. To enhance β-cell specificity, Pax4 and Nkx2-2 were added sequentially (5TF protocol: Ad-NPM on day 1, Ad-Pax4 on day 4, Ad-Nkx2-2 on day 7). At day 10, cells were analyzed by qPCR, immunofluorescence, calcium imaging, and insulin secretion assays. Spheroids were generated by transferring cells to low-attachment plates on day 7. RNA sequencing was performed on 5TF spheroids and parental fibroblasts. For in vivo studies, 300 5TF spheroids were transplanted into the anterior chamber of the eye (ACE) of normoglycemic NSG mice, and 3500–5000 spheroids were transplanted into the omentum. Human insulin was measured in aqueous humor and plasma.

**Key Results:** Ad-NPM induced INS mRNA expression in fibroblasts, reaching 0.12% of human islet levels. Adding Pax4 and Nkx2-2 (5TF) increased INS expression and induced NKX6-1 and PAX6 while reducing GCG. At day 10, 67.9 ± 6.2% of cells were insulin-positive by immunofluorescence. 5TF cells showed reduced proliferation (44×10³ vs 238×10³ cells/well) but maintained viability. RNA-seq revealed 2806 differentially expressed genes (1186 upregulated, 1620 downregulated) between 5TF spheroids and fibroblasts, with enrichment for pancreas/β-cell and peptide hormone metabolism pathways and repression of β-cell disallowed genes (e.g., OAT, LDHA, SMAD3). In calcium imaging, 65% of 5TF cells responded to glucose or KCl. In 2D culture, insulin secretion was constitutive (stimulation index ~1.08), but in 3D spheroids, glucose-stimulated insulin secretion was observed (stimulation index 2.02 ± 0.18). In ACE transplants, human INS mRNA was detected at levels comparable to pre-transplant spheroids, and 43.5 ± 2.8% of HLA+ cells were insulin-positive at day 10. Human insulin was detected in aqueous humor of all 17 mice with 5TF grafts (76–1103 pmol/L), ~20-fold lower than human islet grafts. In omentum transplants, plasma human insulin increased from 3.6 ± 0.9 to 13.9 ± 3.7 pmol/L after glucose injection (p=0.014).

**Clinical Implications:** This study provides the first evidence that human fibroblasts can be directly reprogrammed into insulin-producing cells using transcription factors, challenging previous assumptions. The approach offers potential advantages: fibroblasts are readily accessible, can be autologously transplanted, and avoid tumorigenic risks associated with pluripotent stem cells. However, the generated cells are immature compared to primary β cells, with lower insulin expression, absence of mature insulin granules, and limited in vivo survival. Further optimization—such as extended culture, inclusion of soluble factors, or improved transplantation techniques—is needed to enhance functionality and longevity. Despite these limitations, this proof-of-concept establishes a foundation for developing alternative cell sources for diabetes cell therapy.