Epigenetic biomarker to support classification into pluripotent and non-pluripotent cells

Abstract

Quality control of human induced pluripotent stem cells (iPSCs) can be performed by several methods. These methods are usually relatively labor-intensive, difficult to standardize, or they do not facilitate reliable quantification. Here, we describe a biomarker to distinguish between pluripotent and non-pluripotent cells based on DNA methylation (DNAm) levels at only three specific CpG sites. Two of these CpG sites were selected by their discriminatory power in 258 DNAm profiles - they were either methylated in pluripotent or non-pluripotent cells. The difference between these two β-values provides an Epi-Pluri-Score that was validated on independent DNAm-datasets (264 pluripotent and 1,951 non-pluripotent samples) with 99.9% specificity and 98.9% sensitivity. This score was complemented by a third CpG within the gene POU5F1 (OCT4), which better demarcates early differentiation events. We established pyrosequencing assays for the three relevant CpG sites and thereby correctly classified DNA of 12 pluripotent cell lines and 31 non-pluripotent cell lines. Furthermore, DNAm changes at these three CpGs were tracked in the course of differentiation of iPSCs towards mesenchymal stromal cells. The Epi-Pluri-Score does not give information on lineage-specific differentiation potential, but it provides a simple, reliable, and robust biomarker to support high-throughput classification into either pluripotent or non-pluripotent cells.

Figure 1. Derivation of the Epi-Pluri-Score.

(a) Schematic overview of the work flow. (b) Illustration of the two criteria used for identification of relevant CpG-sites is exemplarily depicted for a CpG site in ANKRD46: the margin between pluripotent and somatic samples (arrows; first criterion) and the fraction of correctly classified in vitro differentiated iPSCs (second criterion). (c) Margins between the ranges of pluripotent and somatic cells (negative values correspond to overlapping ranges) for all investigated CpG-sites. (d) Plot of the two criteria, which led to the selection of the CpG-site within ANKRD46. (e) DNAm levels of the two CpGs of the Epi-Pluri-Score (C14orf115 and ANKRD46) and various pluripotency-associated genes in 258 samples of the training-dataset. (f) Classification of samples of the training-dataset according to Epi-Pluri-Score and DNAm level at cg13083810 (POU5F1).

Figure 2. Validation of the Epi-Pluri-Score.

(a) DNAm levels at selected CpG sites in samples of the validation-dataset (264 pluripotent and 1,951 non-pluripotent samples; all analyzed with the Illumina HumanMethylation27 BeadChip). The best performing CpG site of the corresponding gene is depicted and CpGs of the Epi-Pluri-Score (C14orf115, ANKRD46) facilitate best classification. (b) Classification of the validation-dataset using the Epi-Pluri-Score and DNAm of cg13083810 (POU5F1). There are only 2 false positive and 3 false negative classifications. Annotation of a partially reprogrammed cell line (green) was later corrected.

Figure 3. Epi-Pluri-Score analysis by pyrosequencing of selected CpGs.

(a) Pyrosequencing assays were designed for the three relevant CpGs (indicated in red). (b) Using this method, various different cell types were analyzed (including ESCs, iPSCs, MSCs, HDFs, tumor cell lines, and other primary cells; Supplemental Table S3). All of the analyzed samples were correctly classified by the Epi-Pluri-Score. (c) DNAm levels and (d) gene expression levels (both based on microarray data) of the genes ANKRD46, C14orf115 and POU5F1 in iPSCs and MSCs indicate that hypomethylation in C14orf115 and POU5F1 may be associated with up-regulation of gene expression (* = P < 0.05; ** = P < 0.02; *** = P < 0.001).

Figure 4. Analysis of pluripotency markers during differentiation of iPSCs towards MSCs.

(a) Immunofluorescence analysis of TRA-1-60 (red) and OCT4 (green) in iPSCs and in the course of differentiation towards iPSC-derived MSCs (nuclear counterstaining with DAPI: blue; scale bar = 100 μm). (b) Time course of DNAm levels in cg23737055 (ANKRD46), cg22247240 (C14orf115), and cg13083810 (POU5F1). Three different iPSC clones are indicated by different symbols. (c) Epi-Pluri-Score classification of iPSC-derived MSCs in the course of differentiation. DNAm levels in cg13083810 (POU5F1) discriminate early differentiation changes better – it is therefore useful to complement the Epi-Pluri-Score. (d) Quantitative RT-PCR analysis of ANKRD46, C14orf115, and POU5F1 in the course of iPSC-differentiation towards MSCs. (n = 3; differential expression compared to iPSCs: * = P < 0.05; ** = P < 0.02; *** = P < 0.001).