Figure 2: Pan-Ligandome Perturb-Seq Screen in KOLF2.1J hiPSCs via OASIS.

(A) Schematic of Pan-ligandome Perturb-Seq in KOLF2.1J hiPSCs. Cells were transduced with our lentiviral library on Day 0 in mTeSR media. On Day 4, one plate was harvested, and single-cell RNA sequencing (scRNA-Seq) was performed, representing the Day 4 mTeSR condition. Concomitantly, a second plate was swapped from mTeSR to E6 media, allowed to propagate until Day 6, and subsequently harvested for scRNA-Seq representing the Day 6 E6 condition. Altogether, ~97,000 single cells expressing a single ligand were identified post-QC. Panels B-G correspond to Day 4 mTeSR screen; panels H-Q correspond to Day 6 E6 screen. (B) Screen QC statistics for Day 4 mTeSR screen. Left: plot of UMIs per cell, representing 73.2K cells with a median of 8.5K UMIs per cell. Right: number of cells captured per ligand, with 81% of screened ligands having >25 associated cells. (C) Uniform Manifold Approximation and Projection (UMAP) visualization of transcriptional profiles of cells in Day 4 mTeSR screen. Gray cells represent non-NTC cells, while colored cells represent NTC cells, with the color scale demonstrating local density. (D) UMAP visualization colored by Leiden cluster. Cluster 12 is significantly enriched for ligands pertaining to interferon signaling, as identified by permutation testing (see Methods, adj. p < 0.05). (E) Volcano plot of differential gene expression between Leiden cluster 12 and all other clusters. Colored dots represent FDR < 0.05, |L2FC| > 2. Enriched GO:BP terms for upregulated genes are shown. (F) Energy distance test (E-test) results for all ligands vs NTC, with significant (adj. p < 0.05) ligands labeled. The E-test measures the extent of perturbation-induced transcriptional remodeling. (G) Clustermap of ligand-ligand transcriptional correlation for ligands identified in clusters by HDBSCAN. Insets highlight clusters related to interferon signaling and tumor necrosis factor (TNF) response. (H) Screen QC statistics for Day 6 E6 screen. Left: plot of UMIs per cell, representing 23.8K cells with a median of 20K UMIs per cell. Right: number of cells captured per ligand, with 51% of screened ligands having >25 associated cells. (I) Left: UMAP visualization of transcriptional profiles of cells in Day 6 E6 screen. Gray cells represent non-NTC cells, while colored cells represent NTC cells, with the color scale demonstrating local density. Right: marker gene expression overlays showing pluripotency (POU5F1), muscle progenitor (PAX7, TBX1), and neural (ZIC1) markers. (J) UMAP visualization colored by Leiden cluster. Cluster 1 is significantly enriched for ligands pertaining to FGF signaling, as identified by permutation testing (see Methods, adj. p < 0.05). (K) Overlay of FGF16, 4, 6, 3, 18 positions on UMAP visualization. (L) Volcano plot of differential gene expression between Leiden cluster 1 and all other clusters. Colored dots represent FDR < 0.05, |L2FC| > 2. Enriched GO:BP terms for upregulated genes are shown. (M) Energy distance test (E-test) results for all ligands vs NTC, with significant (adj. p < 0.05) ligands labeled. (N) Clustermap of ligand-ligand transcriptional correlation for ligands identified in clusters by HDBSCAN. Insets highlight clusters related to interferon and FGF signaling. (O) Mechanistic diagram of FGF-mediated signaling, indicating downstream upregulation of ETV4/5, DUSP6, and SPRY4. Modified from Ornitz and Itoh(25). (P) Log 2 fold changes (vs NTC) of downstream transcription factors in the FGF signaling pathway identified from log normalized counts from single-cell dataset. Significance is determined by differential expression via DESeq2 on unnormalized counts from single-cell dataset (see Methods, ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05).