Substantial somatic genomic variation and selection for BCOR mutations in human induced pluripotent stem cells

Abstract

Human Induced Pluripotent Stem Cells (hiPSC) are an established patient-specific model system where opportunities are emerging for cell-based therapies. We contrast hiPSCs derived from different tissues, skin and blood, in the same individual. We show extensive single-nucleotide mutagenesis in all hiPSC lines, although fibroblast-derived hiPSCs (F-hiPSCs) are particularly heavily mutagenized by ultraviolet(UV)-related damage. We utilize genome sequencing data on 454 F-hiPSCs and 44 blood-derived hiPSCs (B-hiPSCs) to gain further insights. Across 324 whole genome sequenced (WGS) F-hiPSCs derived by the Human Induced Pluripotent Stem Cell Initiative (HipSci), UV-related damage is present in ~72% of cell lines, sometimes causing substantial mutagenesis (range 0.25-15 per Mb). Furthermore, we find remarkable genomic heterogeneity between independent F-hiPSC clones derived from the same reprogramming process in the same donor, due to oligoclonal populations within fibroblasts. Combining WGS and exome-sequencing data of 452 HipSci F-hiPSCs, we identify 272 predicted pathogenic mutations in cancer-related genes, of which 21 genes were hit recurrently three or more times, involving 77 (17%) lines. Notably, 151 of 272 mutations were present in starting fibroblast populations suggesting that more than half of putative driver events in F-hiPSCs were acquired in vivo. In contrast, B-hiPSCs reprogrammed from erythroblasts show lower levels of genome-wide mutations (range 0.28-1.4 per Mb), no UV damage, but a strikingly high prevalence of acquired BCOR mutations of ~57%, indicative of strong selection pressure. All hiPSCs had otherwise stable, diploid genomes on karyotypic pre-screening, highlighting how copy-number-based approaches do not have the required resolution to detect widespread nucleotide mutagenesis. This work strongly suggests that models for cell-based therapies require detailed nucleotide-resolution characterization prior to clinical application.

Effects of somatic mutations on cellular differentiation in iPSC models of neurodevelopment

“…In addition, we did not find any BCOR LoF variants among the parental fibroblasts of the iPSC lines, although they could still be present at very low frequency as subclones. Pathogenic mutations in the BCOR gene have been found to be recurrently mutated in blood-derived iPSC lines and positively selected for under iPSC culture conditions 11 . In this regard, we cannot determine whether the BCOR mutations driving the impaired DN differentiation originated in vivo or in vitr o, but they also expanded during iPSC reprogramming.…”

“…To test whether the overall burden of somatic mutations acquired by iPSC lines in vitro influenced their differentiation ability, we studied 384 cell lines (251 individual donors) from the HipSci project. Exomes were sequenced for both the parental fibroblast of donors and their derived iPSC lines 7 , allowing us to distinguish between variants present already in the donors from those acquired, or positively selected for, in the subsequent reprogramming process (hereon ' in vitro -acquired mutations') 11 ( Methods ) A median of 35 mutations (37 when including CNV) was observed per exome in the iPSCs, of which 14 were annotated as deleterious. In line with previous publications 16 , half (50.5%) of the mutations (SNVs and dinucleotides) were predicted to be missense or LoF.…”

“…A joint variant calling ( BCFtools/mpileup and BCFtools/call, version 1.4.25, human genome assembly GRCh37d5) between 384 pairs of parental fibroblasts and their corresponding iPSC lines was performed to identify 18,999 somatic mutations that were acquired or positively selected throughout iPSC reprogramming as described in 11 . This calling was performed for single-nucleotide variants (SNVs), dinucleotides, indels and copy number variants (CNVs) for both autosomal chromosomes and chromosome X.…”

“…We leveraged whole-exome sequencing data from 832 iPSC cell lines to evaluate the genetic determinants on several iPSC differentiations, where germline and somatic mutations were a priori indistinguishable. Yet, for 384 of the 832 lines, a joint variant calling between the available parental fibroblast and the corresponding iPSC (Rouhani et al 2021) was performed to detect the somatic mutations that were positively selected during reprogramming as a proxy to identify acquired somatic mutations in vitro . Effect of the mutational burden on iPSC differentiations.…”

“…WES data for the managed access samples is deposited in the European Genotype-Phenotype Archive (EGA, https://ega-archive.org ), with normal and specific disease cohort datasets available upon request and data access agreement. The variant call sets of acquired mutations in vitro and the code used to generate them are available at 11 .…”