Three independent organoid differentiation batches are profiled; from each, 15-25 organoids are pooled and dissociated to one cell suspension per batch (3 libraries total). A target of ~10,000 cells loaded per library (expecting ~6,000-8,000 recovered). Multiplexing via cell hashing is optional to pool batches and reduce cost while controlling batch effects. A primary/fetal intestine reference dataset is used for label transfer. Doublet and ambient-RNA controls are computational. Each library is sequenced to ~30,000-50,000 reads/cell.
BSL-1/BSL-2 for human iPSC-derived organoids in a Class II cabinet. Matrigel and Cell Recovery Solution are murine-derived biologicals — handle on ice, biohazard waste. DNase, TrypLE, and dead-cell-removal magnetic beads are low hazard but used RNase-free. 10x and library reagents include guanidinium/ethanol — irritants, used in a fume hood where required; segregate chemical waste. Decontaminate cell waste with 10% bleach. RNase-free practices (gloves changed often, dedicated reagents) protect both worker and sample integrity. Sharps and pipette tips to biohazard sharps containers.
Computational negative control: ambient-RNA correction (SoupX/CellBender) and empty-droplet filtering (EmptyDrops) control for background. Doublet detection (Scrublet/DoubletFinder) controls for multiplets. A primary or fetal human intestine scRNA-seq reference serves as the positive benchmark for cell-type identity (label transfer). Cell-hashing demultiplexing (if used) controls for batch/sample swaps and identifies cross-sample doublets. A no-cell (buffer-only) GEM run controls for reagent contamination. Viability staining gates input quality before loading.
Post-QC: median 2,000-5,000 genes/cell, <15% mitochondrial reads, doublet rate <8%. Clustering resolves LGR5/OLFM4+ stem, FABP2/APOA1+ enterocyte, MUC2/TFF3+ goblet, CHGA/NEUROD1+ enteroendocrine, and proliferative MKI67+ crypt clusters, plus a VIM/COL1A1+ mesenchymal population. Marker fidelity and proportions should approximate fetal intestine; absence of differentiated lineages or dominance of off-target (e.g., neuronal) clusters indicates poor differentiation.
To generate a high-quality single-cell transcriptomic atlas of human iPSC-derived intestinal (mid/hindgut) organoids by enzymatically dissociating organoids to >=85% viable single cells, capturing ~8,000-10,000 cells via 10x Genomics 3' chemistry, sequencing, and computationally identifying the major intestinal epithelial and mesenchymal cell types to assess maturation and inter-organoid heterogeneity.
Independent variables: differentiation batch, organoid age, (optional) hashing tag/sample identity. Dependent variables: per-cell QC metrics (genes/UMIs, % mitochondrial), recovered cell number, cluster identities and proportions, marker-gene expression, maturation score vs. reference. Controlled variables: dissociation enzyme and timing, target loading cell number (~10,000), sequencing depth (~30-50k reads/cell), reagent kit lot, Cell Ranger reference genome/version, ambient/doublet pipeline parameters.
Mature (day 30+) iPSC-derived intestinal organoids contain transcriptionally distinct epithelial subtypes — LGR5/OLFM4+ stem cells, FABP2/APOA1+ enterocytes, MUC2/TFF3+ goblet cells, CHGA/NEUROD1+ enteroendocrine cells, and a VIL1+ proliferative crypt-like compartment — alongside a VIM/COL1A1+ mesenchymal niche. scRNA-seq will resolve these clusters, and their proportions and marker fidelity will correlate with differentiation maturity versus fetal intestine references.
Raw reads processed with Cell Ranger (GRCh38) to a filtered cell-by-gene matrix. In Seurat/Scanpy: filter cells (200-7,000 genes, <15-20% mito), normalize (log/SCTransform), integrate batches (Harmony/RPCA), PCA -> UMAP -> Leiden clustering. Clusters annotated by canonical markers and label transfer from a reference. Cell-type proportions computed per batch; maturation scored by module scores. Ambient RNA (CellBender) and doublets (Scrublet) removed before clustering.
Differential expression per cluster by Wilcoxon rank-sum with Benjamini-Hochberg FDR (q < 0.05) in Seurat FindMarkers. Cell-type proportion differences across batches tested by a Dirichlet/beta-binomial model (scCODA) or by ANOVA on logit-transformed proportions (alpha = 0.05). n = 3 biological batches; batch is a random effect. Correlation of organoid pseudobulk to reference quantified by Spearman's rho. Multiple-cluster comparisons FDR-corrected. Power addressed by capturing >=6,000 cells/library to detect populations >=1% with >95% probability.