Single cell transcriptomic analysis of human pluripotent stem cell chondrogenesis

The therapeutic application of human induced pluripotent stem cells (hiPSCs) for cartilage regeneration is largely hindered by the low yield of chondrocytes accompanied by unpredictable and heterogeneous off-target differentiation of cells during chondrogenesis. Here, we combine bulk RNA sequencing, single cell RNA sequencing, and bioinformatic analyses, including weighted gene co-expression analysis (WGCNA), to investigate the gene regulatory networks regulating hiPSC differentiation under chondrogenic conditions. We identify specific WNTs and MITF as hub genes governing the generation of off-target differentiation into neural cells and melanocytes during hiPSC chondrogenesis. With heterocellular signaling models, we further show that WNT signaling produced by off-target cells is responsible for inducing chondrocyte hypertrophy. By targeting WNTs and MITF, we eliminate these cell lineages, significantly enhancing the yield and homogeneity of hiPSC-derived chondrocytes. Collectively, our findings identify the trajectories and molecular mechanisms governing cell fate decision in hiPSC chondrogenesis, as well as dynamic transcriptome profiles orchestrating chondrocyte proliferation and differentiation.

For single cell RNA sequencing (scRNA-seq), sequencing of mixed species was performed prior to the experimental samples to ensure the quality of sequencing with low cell multiplet rate (2.7%) (Fig. S3A). Furthermore, individual samples are collected by by time points (d1, d3, d7, d14, d28, d42) post-chondrogenic induction due to prohibitive cost of the sequencing per sample. However, two batches of d28 samples were collected from independent experiments for scRNA-seq. Canonical correlation analysis (CCA) was used to align cells from the 2 batches16 (Fig. S3B). The cells in the same cluster from different batches exhibited high correlation in their gene expression (Spearman's rank coefficient rs > 0.87 for all clusters) (Fig. S3C). Furthermore, genes that were highly conserved in one particular cluster (using cluster 0 as an example) showed similar expression patterns in the clusters from distinct batches, suggesting that our differentiation is highly reproducible.  (7):1094-1105), n = 3-4 pellets per group is sufficient to detect statistical significance at the 95% confidence level.
For quality control and to determine the heterogenous composition of cell populations, gene barcode matrices were input into the Seurat R package (version 2.4). Low-quality cells, defined as cells expressing less than 200 genes, more than 7,000 genes, and greater than 5% mitochondrial gene content, were removed. Note that these cutoff criteria need to be adjusted for individual studies, as gene expression levels can be cell type-dependent. Genes that were detected in less than three cells were also removed. Next, to reduce the variance introduced by "unwanted" sources, we regressed out variation in gene expression driven by cell cycle stages and mitochondrial gene expression with the vars.to.regress argument in the function ScaleData in Seurat. This exclusion criteria were ore-established based on the instruction provided on the Seurat website. All the scRNA-seq data analyses were performed after data filtering and cleaning (after i.e., quality control steps).
For single cell RNA sequencing (scRNA-seq), two batches of d28 samples were collected from independent experiments for scRNA-seq to evaluate the data reproducibility. Sequencing experiments were generally performed in biological duplicates. The high correlations between two batches of our d28 samples suggests that our experiments is highly reproducible. Furthermore, multiple independent cell lines were used for the current study to ensure the validity of our protocol and experiments. The chondrogenesis of all the independent cell lines will performed independently and the results are highly reproducible.
The collected/harvested samples was randomized and allocated into experiments.
The collected/harvested samples was randomized and allocated into experiments. The Researchers were blinded to group allocation. However, the bioinformatic analyses of the scRNA-seq samples were not randomized as it is required to know the time points of samples in order to construct differentiation trajectory. All the antibodies used were validated by the manufactures and the validating data was provided on their websites. Furthermore, we also used both positive and negative staining controls to further validate each antibody in our experiments.

nature research | reporting summary
Three distinct human induced pluripotent stem cells (hiPSC) lines were used in the current study: STAN, ATCC, and BJFF. STAN line was purchased from WiCell (#STAN061i-164-1), ATCC line was acquired from ATCC (#ATCCACS-1019), and BJFF was obtained from the Genome Engineering and iPSC Core at Washington University in Saint Louis. All three lines were reprogrammed by Sendai virus from human foreskin fibroblasts.
All cell lines used are evaluated to have pluripotency using terotomera formation and confirmed to be karyotypically normal.
All cell lines used are mycoplasma free.
No commonly misidentified cell lines were used in the study. male NOD mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, #005557, Jackson laboratory) at age of 18-20 weeks-old were used for human xenograft implantation in the in the dorsal region (subcutaneous) or in osteochondral defects in the knee joints of mice. Mice were housed under a 12hr light/12hr dark cycle with ambient temperature and humidity.
No wild animals were used in the study.
No field collected samples were used in the study.
All animal procedures were approved by Institutional Animal Care and Use Committee (IACUC) at Washington University in Saint Louis.