Identifying Extrinsic versus Intrinsic Drivers of Variation in Cell Behavior in Human iPSC Lines from Healthy Donors

Abstract

Large cohorts of human induced pluripotent stem cells (iPSCs) from healthy donors are a potentially powerful tool for investigating the relationship between genetic variants and cellular behavior. Here, we integrate high content imaging of cell shape, proliferation, and other phenotypes with gene expression and DNA sequence datasets from over 100 human iPSC lines. By applying a dimensionality reduction approach, Probabilistic Estimation of Expression Residuals (PEER), we extracted factors that captured the effects of intrinsic (genetic concordance between different cell lines from the same donor) and extrinsic (cell responses to different fibronectin concentrations) conditions. We identify genes that correlate in expression with intrinsic and extrinsic PEER factors and associate outlier cell behavior with genes containing rare deleterious non-synonymous SNVs. Our study, thus, establishes a strategy for examining the genetic basis of inter-individual variability in cell behavior.

Keywords: SNV; cell adhesion; dimensionality reduction; fibronectin; genetic variation; high content imaging; iPSC; stem cell niche; stem cells.

Graphical abstract

Figure 1

Description of Phenotypic Dataset (A) Microscopic image showing cells 24 h after plating. Red: cell mask (cytoplasm); white: EdU incorporation (DNA synthesis, one EdU+ cell marked with asterisk); blue: DAPI (nuclei). Scale bar: 20 μm. (B) Schematic of phenotypic features measured in this study. (C) Correlation of different phenotypic measurements in all cells. (D) Distribution of main phenotypic features of all cell lines on three fibronectin concentrations (Fn1, red; Fn5, green; Fn25, blue). y axis: density measurements represent the cell number distributions. (E) Boxplots of mean cell area on three fibronectin concentrations in three biological replicates (batches). Each dot is one cell line. Asterisks (^∗∗∗∗p ≤ 0.0001) represent significance values from pairwise t tests performed between each condition. (F) Heatmap of mean cell area measurements for each cell line on three fibronectin concentrations in three independent experiments. Grey boxes correspond to replicates not performed.

Figure 2

Identification of Outlier Cell Lines for Individual Phenotypes The distribution of the Kolmogorov-Smirnov statistic (D) obtained by performing the Kolmogorov-Smirnov test of the distributions of each raw phenotypic feature for each individual cell line compared to all cell lines. 95^th percentile threshold is shown as a red line together with values of individual outlier lines (color coded). Lines listed in italic bold correspond to lines having outlier measurements in more than one batch.

Figure 3

Synthetic Phenotypic Features Capture Extrinsic and Intrinsic Contributions to Variance Plots showing the distribution of values for the 9 PEER Factors (F1–F9). Left and middle columns show distributions for three fibronectin concentrations (Fn1, red; Fn5, green; Fn25, blue). Asterisks represent significance values from pairwise t tests performed between each fibronectin condition (^∗∗∗∗p ≤ 0.0001; ^∗∗∗p ≤ 0.001; ^∗∗p ≤ 0.01; ^∗p ≤ 0.05; ns, not significant). Right-hand column shows the donor-concordance between two clonal lines of cells derived from the same donors. Values for one cell line in each pair are shown on the x axis and its “twin” on the y axis. Each dot corresponds to one cell line.

Figure 4

Using the “Extrinsic” and “Intrinsic” PEER Factors to Identify Genes That Correlate with Specific Cell Phenotypes (A) Heatmap showing the 3,879 genes correlated with either extrinsic PEER Factor 1, intrinsic Factor 9, or both. Color scale depicts correlation values. (B) GO analysis of genes correlating with PEER 1 (blue circles), PEER 9 (orange circles), or both factors (gray circles). All GO terms for the factors are shown. Circle size represents the frequency of the GO term in the underlying Gene Ontology Annotation (GOA) database; red color scale indicates p value. Each gene was mapped to the most specific terms applicable in each ontology. Highly similar GO terms are linked by edges, with edge width depicting the degree of similarity. Terms in black font were used to select the list of 175 genes in Table S3. (C) In 98 out of 175 genes, gene expression correlated significantly with cell area, tendency to form clumps (“clumpiness”), number of cells, and/or proliferation. The colors of the points correspond to the correlation values, while the shapes indicate correlation of a specific gene to the extrinsic (PEER 1; oval), intrinsic (PEER 9; rectangle) or both (triangles) factors. Grey dotted vertical lines separate genes correlating with one, two or four phenotypes (left to right). (D) Boxplots showing the expression values (vsn) of 32 out of the 38 genes (Table S5) with outlier gene expression in one or more outlier cell line. Color code (blue, orange, gray) as in (B).

Figure 5

Identification of Rare, Deleterious, and Destabilizing nsSNVs That Correlate with Outlier Cell Behavior (A) Analysis pipeline for selection of genes. The 3,879 genes associated with PEER 1 and 9 were screened for nsSNVs in over 700 cell lines from the HipSci resource and further filtered as shown. (B) Genes with at least one rare, deleterious, and destabilizing nsSNV in at least one cell line found to be an outlier for one or more phenotype. See Figure 2 for outlier KS analysis. Genes correlating with PEER Factor 1: blue; PEER Factor 9: orange; both: gray. The phenotypes of cell area, cell roundness, and nucleus roundness were significantly over-represented in outlier cell lines with one or more deleterious and destabilizing nsSNV (p ≤ 0.05). (C) Representative images of outlier cell line yuze_1 (top), control cell line A1ATD-iPSC patient 1 (center), and cell line not analyzed in the original screen ffdc_11 (bottom), on different fibronectin concentrations (Fn1, Fn5, and Fn25). (D) Protein structures of integrin α6 (top) and integrin β1 (bottom). nsSNVs detected in the two cell lines are shown with yellow spots indicated by red arrows.