Chromosome X dosage modulates development of aneuploidy in genetically diverse mouse embryonic stem cells Authors: Alexander Stanton, Selcan Aydin, Daniel A Skelly, Tracy McGarr, Dylan Stavish, Kim Leonhard, Seth Taapken, Tenneille Ludwig, Erik McIntire, Matthew Pankratz, Anne Czechanski, Ted Choi, Steven P Gygi, Ivana Barbaric, Thomas Vierbuchen, Steven C Munger, Laura G Reinholdt, Martin F Pera Published: 2026 ABSTRACT SummaryThe genetic integrity of pluripotent stem cells (PSC) is critical to their applications in research and therapy, but it is compromised by frequent development of structural chromosome variants associated with malignancy. Many PSC lines exhibit remarkable genetic stability, but little is known about the basis of the known variation in genomic integrity among different PSC isolates. Here, we identify aneuploidies using RNA-seq and proteomics data from a panel of mouse embryonic stem cell (mESC) lines derived from 170 Diversity Outbred mice. We found 62 lines with detectable aneuploid subpopulations and a subset of originally XX lines that lost one chromosome X (XO). Strikingly, a much lower proportion of XX lines were aneuploid, compared to XY or XO lines. Two single-cell RNA-seq datasets demonstrated that aneuploid XY DO mESC also show lower chromosome X gene expression, and a prospective study confirmed that XY mESC accumulate higher aneuploid proportions in culture than isogenic XX lines. We identify potential mechanisms for this protective effect of X chromosome dosage, including our findings that the lines with two active X chromosomes have a higher proportion of 2-cell-like cells and higher expression of X-linked tumor suppressor genes, both features associated with the maintenance of genomic integrity, and that they compete more robustly in co-culture with chromosomally abnormal cells than XY cells.Keywords: mouse embryonic stem cell, female, genetic integrity, aneuploidy, X-chromosome, 2C state, cell competition, e-karyotyping, transcriptome, proteome Introduction One of the most significant obstacles to the laboratory or clinical use of pluripotent stem cells (PSCs) is their tendency to develop recurrent genetic abnormalities during prolonged culture. Structural or numerical chromosome abnormalities are the most commonly observed variants. The aneuploidies that occur can be segmental or whole chromosome copy number variations (CNVs). In human primed PSC, it is common to see duplications of chromosomes 1, 7, 12, 14, 17, 20, X, as well as smaller regions of these and some other chromosomes (Draper et al., 2004; Mitalipova et al., 2005; Catalina et al., 2008; Lefort et al., 2008; Spits et al., 2008; Mayshar et al., 2010; Amps et al., 2011; Halliwell et al., 2020) whereas in naive mouse embryonic stem cells (mESCs) it is common to see duplications of chromosomes 1, 6, 8, and 11 as well as loss of one sex chromosome (X in female lines and Y in male lines) (Sugawara et al., 2006; Rebuzzini et al., 2008; Kim et al., 2013; Gaztelumendi and Nogués 2014; Codner et al., 2016; Guo et al., 2018). Despite their disparate nature, these mutations can all result in similar changes to the cells harboring them. First, they can decrease differentiation potential, compromising their use in research or therapy, or in the mouse, their ability to contribute to the developing embryo in the generation of transgenic animal models (Werbowetski-Ogilvie et al., 2009; Ben-David et al., 2014; Zhang et al., 2016; Andrews et al., 2022). Second, they can confer a proliferative or survival advantage, leading to their expansion in undifferentiated cultures and posing a risk of cancer development if used for cell replacement therapy (Zhang et al., 2016; Andrews et al., 2022). While around a third of human PSC samples show these structural abnormalities, many cell lines can be maintained without issue through many generations, though it is unclear why this is the case (Amps et al., 2011). Given these problems, it is important to be able to understand what underlying variation contributes to the generation and maintenance of these aneuploidies in culture to improve our ability to propagate these valuable resources. The hypothesis that the genetic background of PSC lines might influence their genetic stability has not been investigated. A main advantage to using mESC as a model for studying PSC is the availability of stem cells from multiple mouse inbred strains and hybrids to facilitate genetic mapping (Czechanski et al., 2014; Skelly et al., 2020). However, most studies of mESC—not just in the context of aneuploidy—have had very little representative genetic diversity. One report noted variation in the frequency of specific chromosome duplications among cell lines from highly related genetic backgrounds (Gaztelumendi and Nogués 2014). Additionally, because most mESC used for research are derived from XY mice, there is little understanding of how chromosomal sex affects genetic stability in the pluripotent state. XX cells in the preimplantation epiblast, which mESC are derived from, have reactivated their second X chromosome and therefore have higher X-linked gene dosage compared to XY mESC (Mak et al., 2004; Okamoto et al., 2004). The most well documented phenotypic consequence of this dosage difference is in pluripotency signaling pathways. In serum-grown cell cultures, XX mESC exhibit a more naïve-like pluripotent state than XY mESC due to decreased MAPK/ERK signaling that presumably stems from dosage of signaling factors encoded on the X chromosome (Schulz et al., 2014). Genetically diverse, allelically balanced multiparent mouse populations like the Diversity Outbred (DO) mouse population have been developed to perform high resolution genetic mapping projects (Svenson et al., 2012) and more recently mESC have been derived for these populations from both XX and XY mice and have been used to study genetic factors that affect maintenance of pluripotency and cell differentiation potential (Ortmann et al., 2020; Skelly et al., 2020). Key features of the DO panels include a high level of widely distributed genetic diversity, high heterozygosity, and accumulated recombination events that facilitate high resolution genetic mapping (Swanzey et al., 2021). These features are not duplicated in existing large panels of hiPSC lines that are accrued randomly and have strong biases in their representation of genetic ancestry (Ghosh et al., 2022). While recent advances in developing chemically defined culture media have generated methods by which we can culture ESC from numerous strains, unfortunately these approaches have had adverse effects on aneuploidy frequency. Originally, mESC were cultured on a mitotically inactivated murine embryonic fibroblast (MEF) feeder cell layer which provided the mESC with the necessary paracrine signaling to proliferate and maintain pluripotency (Evans and Kaufman 1981). Eventually, efforts to define the culture conditions for mESC arrived at the combination of leukemia inhibitory factor (LIF) and inhibitors of the MAPK/ERK and GSK3-signaling pathways using the small molecules PD0325901 and CHIR99021, respectively, termed 2i (Silva et al., 2008). Another, less widespread method named alternative 2i (a2i) substitutes the Src-inhibitor CGP77675 for the MAPK/ERK-signaling inhibitor (Shimizu et al., 2012). While these culture conditions achieve the intended goal of a defined mESC culture media that maintains naive pluripotency and allowed the derivation of mESC from non-permissive strains, they incidentally have detrimental effects on the genomic stability of the cells cultured in these conditions compared to conditions that include MEF feeder cells (Choi et al., 2017; Guo et al., 2018). This highlights the need to further understand the causes of aneuploidy in mESC to develop additional media additives that can improve the maintenance of proper karyotype. To investigate what genetic variation affects the development of aneuploidy in mESC cultures, we have used multi-omic datasets from a panel of DO mESC. These mESC were cultured in serum + LIF and only one inhibitor, the GSK3 inhibitor (for methodology, see Skelly et al., 2020). Culturing in the presence of only this one inhibitor facilitates better understanding of the inherent differences between chromosomal sexes where most of the differences result from the MAPK/ERK-signaling pathway. Using bulk RNA-sequencing and mass spectrometry proteomics data from 170 genotyped mESC lines, we identified samples with accumulations of aneuploid cells by detecting significantly different changes in both transcript and protein expression due to changes in chromosome dosage. Analysis of these detected aneuploidies showed that cell lines with dual X chromosome dosage are less likely to be aneuploid, a finding confirmed in a prospective validation study scRNA-seq study that also showed lower X-linked gene expression in aneuploid cells. Relative to XY cell lines, XX lines also have higher expression of 2C-like (2CL) state genes and higher expression of tumor suppressor genes with functions in DNA damage response, and compete more robustly with chromosomally abnormal cells. Results Virtual karyotyping of DO mESC identifies recurring duplications in four autosomes To investigate how underlying variability in the genetics and chromosomal sex of PSC influences the development and accumulation of aneuploid cells, we analyzed paired-end bulk RNA-seq and multiplexed mass spectrometry proteomics data from a panel of 170 unique DO mESC lines as well as 10X single-cell RNA-seq (scRNA-seq) from an additional 12 unique DO mESC lines derived and cultured in serum + LIF/GSK3 inhibitor (CHIR99021) all collected in the Predictive Biology Inc laboratory. None of the cell lines represented in the single-cell dataset are also represented in the bulk RNA-seq and proteomics dataset. While the bulk dataset has a nearly even number of chromosomal sexes (88 XX and 82 XY cell lines), all 12 lines in the single-cell dataset are from XY embryos. After filtering out gene models, Riken genes, and genes not expressed in all mESC lines, the bulk data consists of 12,579 detected transcripts and 5,141 detected proteins. Our single-cell data consisted of 18,954 detected transcripts and 5,992 cells from between 427 and 557 per cell line. These bulk datasets do not have associated karyotyping data, so virtual karyotyping was used to discover which chromosomes have recurrent CNVs accumulating within the cell line cultures. Virtual karyotyping is the process of assessing CNV in a sample by assessing the resultant increase/decrease in downstream gene expression due to copy number changes. Our methods are similar to those implemented by other groups (Mayshar et al., 2010; Ben-David and Benvenisty 2011; Ben-David et al., 2012, 2013; Weissbein et al., 2016) but specialized to suit the structure of our population and dataset as well as the lack of known euploid reference control samples. Given the balanced genetic population structure of the DO model, we expect the summary expression of each chromosome (median of gene expression for autosomes and sum of gene expression for chromosome X) to fit a single Gaussian distribution and that a population with multiple samples harboring accumulations of CNV for that chromosome would disrupt the fit of this distribution. It is evident when looking at population-normalized gene expression across the genome that certain samples have chromosome expression levels similar to the median for the population (leftmost images of Figure S1A) while other samples have noticeable increases in expression across entire individual chromosome(s) (rightmost images of Figure S1A). Assaying each chromosome’s normality within the population by a Shapiro-Wilk test identified recurring duplications of chromosomes 1, 6, 8, and 11 by their right-skewed expression distribution for both transcript and protein (Figure S1B). Following up on the non-normally distributed chromosomes, we identified samples whose chromosome median gene expression was significantly outside of the expected distribution. After bootstrapping, we identified a total of 62/170 samples that had significant accumulations of cells with chromosome gains and none that had chromosome losses. For the purposes of brevity, the cell lines identified with this method will be referred to simply as aneuploid lines—with the implicit understanding that these are mosaics of cells with both normal and abnormal karyotype. Of the aneuploid cell lines, 10 lines (5.6%) had chromosome 1 duplications, 8 lines (4.4%) had chromosome 6 duplications, 35 lines (19.4%) had chromosome 8 duplications, and 24 lines (13.3%) had chromosome 11 duplications (Figures 1A and 1B). Lines with a single trisomy made up 75.81% of aneuploid samples (47/62), 78.72% of which were chromosome 8 or 11 aneuploidies (Figure 1C). Likewise, all aneuploid lines with multiple duplications had duplication of at least chromosome 8 or 11 in addition to another chromosome (Figure 1C). Identification of DO mESC lines with accumulated aneuploid cell populations using RNA-seq and proteomics (A–C) Comparison of median transcript and protein expression for all genes on specific chromosomes. (A) Commonly euploid chromosomes. (B) Commonly duplicated chromosomes. Purple triangles represent aneuploid mESC lines. (C) Number of cell lines with specific karyotypes in bulk DO datasets. (D) Comparison of median transcript and protein expression for XX, XY, and XX/XO mosaic cell lines. (E) XIST expression for XX, XY, and XX/XO mosaic cell lines from RNA-seq. Loss of chromosome X occurs in a subset of originally XX cell lines Another commonly observed chromosomal CNV in mESC is the loss of the X chromosome in XX cell lines (Gaztelumendi and Nogués 2014; Zhang et al., 2016). Because both X chromosomes are active during naive pluripotency, we employed the same methodology that was used to determine autosomal aneuploidy to determine copy number changes in the X chromosome for the XX lines in our dataset. The expression of X-linked genes in originally XX mESC showed a left-skewed distribution, indicating that chromosome X loss occurred in multiple DO mESC lines (Figure S1B). Bootstrapping results determined that 22/88 (25%) XX cell lines had significantly lower X chromosome expression, matching that of the XY mESC lines (Figure 1D). To rule out the possibility of improper X chromosome inactivation (XCI), we compared the expression of Xist between high and low X expressing XX lines compared to XY male lines. The lack of substantial increase in Xist expression from the low chromosome X gene expressing originally XX mESC lines rules out XCI, whereas the expression levels observed between these groups is more consistent with low levels of expression proportional to the number of X chromosomes harbored by the line (Figure 1E). Estimating the extent of mosaicism in aneuploid cell lines Since virtual karyotyping of bulk sequencing data alone cannot determine the percentage of aneuploid cells in a given line, we used the scRNA-seq dataset to estimate the range of aneuploid mosaics we were observing. Using the R/CONICSmat package, we were able to identify individual cells with trisomies for chromosomes 1, 6, 8, and 11 and directly compare expression of those chromosomes between euploid and aneuploid cells (Figures 2A and 2B). While non-duplicated chromosomes did not experience global changes in gene expression between ploidy conditions, samples with one specific trisomy had approximately a 1.4-fold increase in expression from that duplicated chromosome relative to samples with normal karyotype (Figure 2B). Knowing what chromosome-wide transcript expression change is expected if 100% of the mosaic is aneuploid for a given chromosome provides us with a way to evaluate the changes seen in the bulk RNA-seq data. Comparing each aneuploid sample to the median expression of the euploid samples, we estimate that the aneuploidies annotated in the bulk data represent between 3.44% and 9.06% cells with chromosome 1 trisomy, 3.90%–14.39% cells with chromosome 6 trisomy, 2.50%–11.63% cells with chromosome 8 trisomy, and 2.99%–17.27% cells with chromosome 11 trisomy (Figure 2C). Using DO mESC single-cell RNA-seq aneuploidy calls to estimate aneuploid content of cell lines in bulk DO data (A) Number of cells with specific karyotypes in single-cell data. (B) Gene expression changes from euploid chromosomes in cells with abnormal karyotype compared to cells with normal karyotype (top) as well as gene expression changes from aneuploid chromosomes in cells with a specific trisomy compared to cells with normal karyotype. (C) Estimation of aneuploid cell percentage of bulk data cell lines using the changes determined from single-cell data and the median chromosome RNA-seq values from the bulk data. Cells were studied at passages 7–9. Mouse duplications replicate genes and pathways represented in human autosomal duplications In order to understand the human relevance of the observed mouse aneuploidies, the synteny between mouse and human PSC (hPSC) duplications was assessed using The Jackson Laboratory Synteny Browser (syntenybrowser.jax.org). Genes from mouse chromosomes 1, 6, 8, and 11 were mapped to their corresponding syntenic genes in the human genome, highlighting genes from human chromosomes 1, 7, 12, 14, 17, 20, and X as well as the more minimal segmental duplications 1q32, 7p, 12p13.3, 17q (Halliwell et al., 2020). While mouse chromosomes 1, 6, and 11 shared broadly similar gene content to multiple human duplications, mouse chromosome 8 was surprisingly lacking in synteny with any human duplications (Figures S2A and S2B). Whereas the other mESC duplications likely influence the cells in similar ways as hPSC duplications due to their shared genes, it was curious to find that the most common mouse duplication in our data had no counterpart in human aneuploid PSC. For this reason, we wondered if the trisomy 8 mESC might alter the same pathways as human chromosomal duplications through a different subset of genes. To identify the consequences of trisomy 8, differential expression analysis was performed on the scRNA-seq dataset because it represents purer euploid and aneuploid populations than the bulk data. All cells harboring only trisomy 8 were compared to cells with no detected aneuploidies. As expected, the comparison showed differential expression of pathways related to gene expression/processing as expected when accommodating a large-scale increase in genetic material (Figure S2C). The trisomy 8 cells also differentially expressed genes related to the DNA replication, consistent with the observation that aneuploidies can drive a proliferative advantage relative to euploid cells (Figure S2C). Most of the downregulated pathway terms related to immune response (Figure S2C). Despite a lack of synteny with regions of the genome that are overrepresented in aneuploid human PSC, mouse trisomy 8 seems to result in alterations to expression of genes related to cell proliferation and DNA metabolism and repair, similar to the phenotype of hPSC with recurrent aneuploidies (Lamm et al., 2016) (and references therein). It would have been most informative to compare the effect of X chromosome dosage on recurrent genetic abnormalities in mESC with such an effect in naive human PSC, but the data on the latter were insufficient to support any definite conclusions (Nissim Benvenisty, personal communication). Loci on chromosome X associated with overall autosomal ploidy We began interrogating variation that affected autosomal aneuploidy in the DO samples by mapping genotypes associated with aneuploidy status. Genetic mapping was performed using the R/qtl2 package, using aneuploid and euploid status as a binary trait while correcting the covariates to reflect the recently identified XO lines previously labeled as XX. No significant loci were discovered for any individual trisomy (Figure 3A), however, a significant locus at chromosome X was discovered for general abnormal karyotype (p < 0.01; Figure 3B; Chr X: 143,530,457–157,497,091 in mm10). The founder effects under this interval indicate WSB/PhJ alleles associated with higher likelihood of aneuploidy and PWK/PhJ alleles associated with lower likelihood of aneuploidy (Figure 3C). Unfortunately, genes under this interval did not immediately stand out as good candidates. Therefore, we believe it is more likely that genetic variation in this region affects aneuploidy by affecting broader changes, such as structural variants might cause. Genomic mapping of DO mESCs using aneuploidy as a binary trait reveals influence of chromosome X variation (A and B) Assessing the relationship between genomic loci with duplication of chromosome 8 or 11 (A) and any autosomal duplication (B). (C) Effects of alleles from DO founder mouse strains on aneuploidy at the significant locus found on chromosome X. Cell lines with two active X chromosomes have lower aneuploid proportions Following from the observation that genetic variation on chromosome X was associated with abnormal karyotype, we assessed the proportion of aneuploid mESC lines within each chromosomal sex. There was a significantly lower proportion of aneuploid XX lines compared to XY or XO lines (11/66 XX lines, 41/82 XY lines, 10/22 XO lines; Chi-squared p value = 0.0001; Figure 4A). This pattern was not only observed with overall karyotype, but also for samples with trisomy 8 and 11 (4 and 3/66 XX lines, 26 and 16/82 XY lines, 5 and 5/21 XO lines; Chi-squared p value = 0.0006 and 0.0158, respectively; Figure 4A). Although too few samples were annotated as trisomy 1 or 6 to detect significance, the same trend between XX and XY or XO samples can be seen for these abnormalities as well (Figure 4A). Except for trisomy 11, the proportion of aneuploid XO lines fell in between the XX and XY samples, which suggested that X chromosome dosage was underlying the observed stratification since the mosaic of XO samples includes some proportion of XX cells as well. Inverse relationship between chromosome X gene dosage and autosomal duplications (A) Proportion of cell lines from the bulk DO datasets classified as aneuploid by chromosome X dosage: two (XX), mosaic of one and two (XX/XO), and one (XY). (B) Correlation between each chromosomal sex’s sum of normalized chromosome X transcripts or proteins and the proportion of aneuploid cell lines using DO bulk data. (C) From scRNA-seq of DO mESC lines, the correlations between X-linked gene expression changes in cells with a specific trisomy (relative to euploid cells). Red points are genes with shared directional changes—increasing in both trisomies or decreasing in both trisomies. (D) Average gene expression changes in cells with a specific trisomy and the proportion of genes with positive changes in each chromosome from the single-cell DO data. The duplicated chromosome is highlighted in purple and chromosome X is highlighted in red. (E) Percentage of XX and XY hiPSC and hESC cell lines submitted to WiCell with identified autosomal duplications by G-banded karyotyping. To see if there was a relationship between expression of the X chromosome and likelihood of being annotated as aneuploid, we assessed the correlation between global chromosome X gene expression and proportion of aneuploid lines within each chromosomal sex. For global chromosome X expression, the same sum of transcript or protein expression from Figures S1 and Figure 2 were used. There was a distinct negative correlation between expression of X-linked genes and the proportion of aneuploid lines in a sample (Pearson correlation = −0.95, p = 2.2 × 10−16 for transcripts and Pearson correlation = −0.94, p = 2.2 × 10−16 for protein; Figure 4B). This observation served as additional evidence that the aneuploid-associated loci on chromosome X reflect X-linked gene expression more broadly rather than a single causal gene under the interval. Using the aneuploidy annotations in the scRNA-seq dataset of all XY mESC lines, we were able to observe a similar pattern at the individual cell level. We performed differential expression analysis comparing cells with a single trisomy to completely euploid cells and analyzed the proportion of genes from each chromosome that were upregulated, downregulated, or experienced no changes. As expected, there was a large proportion of upregulated genes from the duplicated chromosome for each individual trisomy (Figure 4C). The remainder of the autosomes shared a similar ratio of upregulated and downregulated genes across trisomies, however, in all besides trisomy 1 there is a higher proportion of downregulated chromosome X genes (Figure 4C). The log2 fold-change in expression of X-linked genes was also significantly correlated between trisomy 6, 8, and 11 (0.315 between 6 and 8, 0.166 between 6 and 11, 0.199 between 8 and 11; all p < 0.001; Figure 4D). These results demonstrate that individual aneuploid cells, even within all XY cell lines, have lower X-linked gene expression. Because the scRNA-seq data are at a single time point, we cannot determine whether individual cells with lower chromosome X expression are more likely to develop aneuploidies (as observed in the bulk data) or if this is evidence that aneuploidy induces decreased expression of genes on chromosome X. We were interested to see if human PSC shared the observation from mouse models that chromosome X dosage modified the likelihood of aneuploidy. Collaborators at The WiCell Institute and Sheffield University analyzed the human ESC and iPSC lines submitted to The WiCell Institute for karyotyping. Although the differences were less pronounced than in the mESC, there was a general trend of increased abnormal karyotype in the XY cell lines compared to the XX lines in iPSCs, but not in ESCs (p = 1.59 × 10−12 in iPSCs; Figure 4E; the ESC sample was slightly underpowered to detect the 22% vs. 26% difference seen in iPSC at p = 0.05 and 90% CI). Although the effect size is smaller in human relative to mouse, we do not expect identical effects due to species differences in culture conditions and chromosome X dynamics during cell culture. Primed XX hPSC show X-inactivation (unlike naive mouse ESC with two active X chromosomes), but erosion of X-inactivation and concomitant increases in X chromosome transcript and protein levels are common findings in both human ESC and iPSC (Bar et al., 2019; Brenes et al., 2021; Topa et al., 2024). Erosion of X inactivation and protective effects of X-linked tumor suppressors that escape X-inactivation (see “discussion”) could account for some degree of protection from aneuploidy in female hPSC. These data from thousands of directly karyotyped cell lines by an external source are strong evidence that this is not a mouse-specific relationship, although the additional nuances of human PSC culture may mitigate this effect under some conditions, and the human data are potentially confounded by a lack of information on the passage level of the cells analyzed. XX mESC cultures accumulate less aneuploidy over time in culture than isogenic XY mESCs To validate our retrospective observation that fewer aneuploid XX mESC lines were identified than XY mESC lines in a different cellular system, we performed scRNA-seq on isogenic C3H/HeSn-Paf/J XX, XY, and XO mESC lines at the beginning and end of 10 passages in serum/LIF culture. The XO cell lines in this study were derived directly from XO embryos with no other genetic abnormality (the product of mating wild-type females with C3H/HeSn-Paf/J males). These latter mice have a small inversion that spans the pseudoautosomal region on their X chromosome (Burgoyne and Evans 2000). This leads to a higher-than-normal rate of non-disjunction of the XY pair of chromosomes during meiosis I. When C3H/HeSn-Paf/J males are bred to a non-carrier female, the result is a high rate of XO embryos (Lane and Davisson 1990), from which the cell lines used in our study were derived (and which can give rise to viable offspring). For each chromosomal sex, three independently derived cell lines were aged for the study. After cells were demultiplexed and filtered, our early isogenic scRNA-seq dataset consisted of 5,200 XX cells, 5,375 XY cells, and 3,564 XO cells, and our late scRNA-seq dataset consisted of 2,832 XX cells, 5,511 XY cells, and 16,600 XO cells (Figures S3 and S4). These early and aged datasets represented expression of 25,573 and 25,452 genes, respectively. The large increase in XO cells indicates that some of the XX and XY cells lost their second sex chromosome over time, although we cannot tell for certain because the demultiplexing was performed based on transcriptomic signatures of each chromosomal sex and not by external methods. First, we annotated the autosomal aneuploidies present at both time points to compare with the DO mESC datasets (Figure 5A). Our early time point began at 86.6% entirely euploid cells, which decreased to 19.7% by the end of 10 passages. The population of aneuploidies in the early time point was predominantly chromosomes 8 and 11 trisomies, matching the distribution observed in the DO mESC bulk and single-cell data. However, the aged aneuploid cells were mostly chromosomes 1 and 8 trisomies—some of which harbored additional chromosome duplications. This observation suggests that while duplication of chromosome 11 is more common in earlier cultures, chromosome 1 duplication may provide advantages that prevail over long periods of time. Another interesting observation was the identification of two duplications not seen in the DO datasets: duplication of chromosomes 14 and 19. These duplications were observed almost exclusively in conjunction with other previously noted aneuploidies (mostly chromosome 1) and have not been identified as commonly as others in the literature. Therefore, we suspect that they are secondary aneuploidies that provide an advantage in culture when in the presence of other acquired advantages from more common duplications. In vitro aging of isogenic cell lines with different chromosomal sexes (A) Proportions of specific chromosomal duplications in early and late (aged) passage mESCs identified from single-cell RNA-seq. Aneuploidies are ordered on the Y axis by proportion in early-stage cultures. (B and C) Before and after proportions of (B) total aneuploidy and (C) specific chromosomal duplications in each chromosomal sex. (D) Change in proportion of total aneuploidy and specific chromosomal duplications in each chromosomal sex. Importantly, analysis of the aneuploid annotations in these aged isogenic mESC also demonstrated a larger increase in aneuploid cell population in the XY and XO lines compared to the XX lines. While XX lines already had fewer aneuploid cells at the early collection, they also had a smaller change in aneuploid proportion over the course of 10 passages than the XY and XO lines had (Figures 5B–5D). When broken down by specific trisomy observed, the differences due to chromosomal sex were less distinct (Figure 5C). However, this is not unexpected if we assume that selection for specific chromosomal duplications occurs only after missegregation. Therefore, different cultures may accumulate different aneuploidies based on which chromosomes are randomly allocated to the wrong daughter cell. An alternate hypothesis could be that specific aneuploidies are differentially beneficial to cells with different suites of sex chromosomes. Overall, the data show that the X chromosome dosage of a cell line influences the buildup of aneuploid populations in culture, validating the observation from diverse mESC. Dusp9 expression correlates with mESC sexual dimorphisms and ploidy status While it may not be likely that a single causal gene is responsible for the differential aneuploidy proportions between single and dual X chromosome samples, we looked for genes on chromosome X whose expression is correlated with ploidy status to better understand the molecular underpinnings of this effect. Because XX lines are seldom used in mESC studies, the literature on their sexual dimorphisms is sparse. However, one literature review noted a difference in the pluripotency signaling pathways between XX and XY mESC under certain culture conditions (Schulz 2017). The review identified six X-linked genes (Eras, Nono, Dusp9, Tfe3, Nr0b1/Dax1, and Zfx) which are associated with pluripotency and able to at least partially replicate the differences in molecular markers of pluripotency (increased Nanog, Esrrb, and Tcl1 expression, Erk phosphorylation, and Akt activity; decreased Mek, Erk, and Gsk3 expression, and DNA methylation; prevention of differentiation)—of which Dusp9 overexpression was most similar to increased X dosage. Because protein expression is more highly attenuated than transcript expression, we calculated the transcript-protein correlation of X-linked genes in our bulk RNA-seq and proteomics data to better understand which genes are most affected by increases in X dosage at the functional level—paying attention to the genes identified in Schulz (2017). Dusp9 had the highest transcript-protein correlation (0.8207) of all six (Schulz 2017) genes, the next highest of which was Eras (0.6361; Figures 6A and 6B). Within each sex chromosome category (XX, XY, and XO), there is lower expression of Dusp9 transcript and protein in aneuploid lines compared to euploid lines (Figures 6B and 6C). Therefore, Dusp9 which is in large part responsible for the pluripotency differences conferred by increased X dosage (Schulz 2017) is more lowly expressed in more aneuploidy-prone single X lines and is also more lowly expressed in those lines that have accumulated aneuploid cells. Increased proportion of 2C-like state cells in mESCs with two X chromosomes (A) Dose sensitivity measured by the correlation between transcript and protein for all X-linked genes represented in the RNA-seq and proteomics in the DO bulk data. Genes that may account for the pluripotency state differences between XX and XY mESCs are labeled (Schulz et al., 2014). (B) Comparison of Dusp9 transcript and protein abundance in individual DO mESC lines colored by chromosomal sex. (C) Abundance of Dusp9 transcript and protein in individual DO mESC lines, comparing euploid and aneuploid lines within each chromosomal sex category. (D) Abundance of transcripts upregulated in the 2C-like state as identified in Macfarlan et al. (2012); Amano et al. (2013) within each chromosomal sex in bulk RNA-seq data. (E) Flow cytometry for the 2C-like state marker MERVL gag protein in XX, XO, and XY C3H/HeSn-Paf/J mESC lines. Signal from each cell (left) and percent of cells with increased MERVL-gag expression from each line (right). Cell lines with two active X chromosomes have higher expression of genes upregulated in 2CL state Differential Dusp9 expression led us to the consideration of a rare, metastable state recently discovered in mESC called the 2CL state. This is a state that mESC periodically transition in and out of during culture, and while generally less than 1% of a culture will be in the 2CL state at any given time depending on conditions and cell line, within about nine passages all of the cells will have gone through that transition (Zalzman et al., 2010). The 2CL state is named for the resemblance of its transcriptional profile to the 2-cell stage of embryonic development, particularly the expression of Zscan4 and endogenous retroviruses like murine endogenous retrovirus L (MERVL) (Macfarlan et al., 2012; Amano et al., 2013). Zscan4 functions in Tert-independent telomere elongation (Zalzman et al., 2010) and according to Amano et al. (2015) may either be involved in supernumerary chromosome removal or may bias survival in favor of normal karyotypes. The Dusp9 expression patterns that we observed were of interest because feeder-based mESC culture conditions not only favor genomic stability, but also promote an increased presence of 2CL state cells and show strong upregulation of Dusp9 compared to feeder-free culture (Guo et al., 2018). Since Dusp9 is expressed higher in dual X mESC lines, euploid mESC lines, as well as feeder-based mESC cultures and all of those are associated with improved genomic integrity, it may be that a common thread is the 2CL state. In order to better understand the relationship between X chromosome dosage and the 2CL state, we compared expression of 2CL state genes between sex chromosome categories. Lists of transcripts upregulated in 2CL state cells were obtained from Macfarlan et al. (2012); Amano et al. (2013). Within our DO mESC bulk RNA-seq data, we found expression of 344/525 (65.52%) of the Macfarlan genes and 45/99 (45.45%) of the Amano genes. The sum of the 2CL state gene expression Z scores were significantly higher in XX mESC lines compared to the XY or XO mESC lines in both gene sets used, suggesting either a stronger induction of these genes or a higher proportion of 2CL cells in the XX mESC (Figure 6D). To further investigate the relationship between the 2CL state, X dosage, and genetic stability, isogenic C3H/HeSn-Paf/J XX, XY, and XO mESC lines, used in the prospective study of genetic stability and X chromosome status used in the prospective study of genetic stability in Figure 5, were assayed by flow cytometry using an antibody directed against MERVL gag protein to quantify the number of 2CL cells. While the 2CL population was small in each cell line as expected, the number of 2CL cells was consistently higher in XX mESC compared to the XO or XY mESC, consistent with the observations from bulk RNA-seq data (Figure 6E). Biological replicates were obtained using independent isolates of these cell lines and a different anti-MERVL gag protein antibody (Figure S6). These results combined with the chromosomal sex bias in aneuploidy development suggest that an increased dosage of X-linked genes in XX mESC increases the proportion of 2CL cells which may contribute to the improved genomic stability of these cell lines. Differential expression of X-Linked tumor suppressor genes Because the human PSC subpopulation comparable to the 2C state, the 8-cell-like (8CL) state, has not been reported in the primed state culture conditions that most of the WiCell data comes from, we sought an alternative explanation for the shared effect of chromosome X dosage between species. We hypothesized that X-linked tumor suppressor genes (TSG), that control responses to DNA damage, apoptotic stimuli and cell proliferation, may be differentially expressed in XX cell lines, providing more protection against emergence of aneuploid cells (Motoyama and Naka 2004). We collected a list of 24 TSG on chromosome X from the Tumor Suppressor Gene Database (bioinfo.uth.edu/TSGene) and analyzed their dosage-sensitivity in our bulk datasets. Dusp9 also came up in our list of TSG, but two other genes had notable dosage sensitivity: Fhl1 and Kdm6a (Figure S5A). Of these, only Fhl1 was significantly associated with ploidy status as well as chromosome X dosage (Figures S5B and S5C). Additionally, because we noted that aneuploid XY mESCs are associated with lower chromosome X gene expression, we wanted to assess changes to these X-linked TSG specifically. When observing log2-fold changes in all aneuploid cells, there were clear decreases in expression of Eda2r, Fhl1, Flna, Rbmx, Dusp9, Kdm6a, Zfp185, Rbbp7, Dmd, Rps6ka6, and Rpl10, and three of the dosage sensitive genes we identified were among the most strongly downregulated (Figure S5D). However, when examining specific aneuploidies due to their unequal representation in the scRNA-seq dataset, it becomes apparent that these TSG exhibit variable expression between trisomies (Figure S5E). These data suggest that while some TSG are expressed higher in cells with two chromosome X copies, expression of any particular single X-linked TSG is not broadly responsible for the emergence of aneuploidy. Moreover, many TSG in this list show cell-type specificity, so it is difficult to directly implicate them generally in expansion of subpopulations of aneuploid mESC. XX cells compete more robustly with aneuploid cells in co-culture experiments compared to XY cells The prevalence of recurrent aneuploidies in stem cell culture depends on two factors: the inherent genetic stability or instability of the cell, and the degree to which particular aneuploidies confer a selective advantage under the culture conditions employed. An enhanced proportion of 2CL cells and differential expression of tumor suppressor genes could account for a lower incidence of the generation of aneuploidies in XX cells. In addition however, since a number of genes on the X chromosome drive pluripotency maintenance, and X dosage effects act to stabilize the naive pluripotent state in female cells (Schulz 2017), female cells may be better fit to compete with aneuploid cells. Therefore, we tested the ability of C57BL6/J XX and XY cells to compete in co-culture with a cell line bearing trisomy 11 and constitutively expressing mCherry, under normal culture conditions and in the presence of reduced levels of LIF (which might result in further selective advantage for aneuploid cells). Results of three independent experiments with three female and two male lines are shown in Figure 7. A 10% inoculum of mCherry+ Trisomy 11 cells expanded in proportion considerably over two passages in co-culture with male cells grown in control medium or medium with reduced LIF concentrations in 5/6 trials, while no expansion was observed in 6/6 co-cultures with female cells in control medium or in reduced LIF concentrations. In co-cultures with one male cell line, the aneuploid population stabilized after expansion, but with a second male line the proportion of aneuploid cells declined in the third passage. In co-cultures with female lines, the population of mCherry+ cells did not increase when LIF concentrations were lowered over the range studied, and declined at the third passage in most experiments. Competition between diploid male or female cell lines and a cell line bearing the mCherry gene under a constitutive reporter and trisomy 11 Trisomy 11 mCherry+ cells (10%) were combined with two male (8 and 4) and two female (12 and 11) diploid cells (90%) at the initiation of co-cultures. Cells were passaged three times under stem cell maintenance conditions (mouse embryo fibroblast feeders, 10% fetal calf serum, and 107 units/ml LIF, and 2i) or in the same conditions with reduced levels of LIF (5E6, 1E6, or 2E5 units/mL in experiments 1, 2, and 3, respectively). At each passage samples were taken for flow cytometric determination of the proportion of mCherry+ cells. Two to three technical replicates (wells) for each culture were obtained at every time point. Two-way ANOVA indicated differences between male and female lines were significant across time points in all experiments (p < 0.001). Unpaired t test indicated significant (p < 0.01) differences between male and female lines at all time points with normal or reduced LIF supplementation, except for 8M and 9F at D3 (experiment 1) and 4M and 11F at D1 (experiment 3) with a false discovery rate of <1%. Discussion Here, we have presented a methodology to virtually karyotype a diverse group of cell lines without euploid controls by using both transcriptomic and proteomic data. We were able to show patterns of aneuploidy that were similar to that found by an established R package for karyotyping single-cell datasets (Figures 1, 2, and 5). These patterns showed that the DO mESCs did not sustain any losses of autosomes, only losing chromosome X in a subset of the XX cell lines (Figures 1C and 1D). Of the duplications observed, chromosomes 1, 6, 8, and 11 were recurrent in both our bulk and single-cell datasets, although there was a distinctly higher likelihood of chromosomes 8 and 11 being duplicated (Figures 1B, 1E, and 2A). While we have tried to make these results generalizable by using a panel of cell lines with a highly diverse genetic background and even split of chromosomal sexes, there is still the possibility that trends in CNV may also reflect the culture conditions and variations therein may favor increased/decreased dosage of different gene sets. However, it seems that on the whole, the duplications commonly seen in mouse and human generally contain similar same genes or are predicted to modulate similar pathways (Figure S2), with the notable exception of chromosome X which is lost in mouse PSC cultures and gained in human PSC cultures (Mitalipova et al., 2005). Our results reveal a distinct connection between the expression of chromosome X genes and the maintenance of normal karyotype in mESC. There are clear examples of the direct association of lower chromosome X dosage with increased likelihood for aneuploidy (Figures 4A, 4B, and 5). Moreover, aneuploid cells with the same number of X chromosomes as their euploid counterparts in the single-cell dataset show statistically lower expression of chromosome X genes (Figures 4C and 4D). There is evidence in models outside the PSC field to suggest that there may be a bidirectional relationship between X-linked genes and genomic stability in multiple species. Individuals with Turner syndrome (XO) have an increased risk for solid tumors compared to XX individuals, while those with Klinefelter syndrome (XXY) have reduced solid tumor risk compared to XY individuals (Ji et al., 2016). Likewise, there is a noted male bias in the development of many types of cancer (Cook et al., 2009; Edgren et al., 2012; Chaligné et al., 2015; Siegel et al., 2020). In a follow-up on this observation, another lab discovered a reduced ratio of X to autosomal gene expression in a number of tumor samples (Zhang et al., 2018). There have also been experiments in Drosophila demonstrating that induced autosomal CNVs result in reduced expression of chromosome X, similar to what we observed in the single-cell data (Clemente-Ruiz et al., 2016). The mechanisms that underpin each directionality of this relationship might still be distinct. Certain genes on chromosome X may promote improved genomic stability, however, the expression of X-linked genes—including those promoting genomic stability—may be decreased as a consequence of autosomal duplications. These observations mostly come from cancer models, and to the best of our knowledge, this is the first observation of chromosomal sex bias in the genomic integrity of ESC. Our observations on X chromosome dosage and aneuploidy in mESC were reflected in a large cohort of hiPSC as well, though the effect size was smaller (Figure 4E). A differing impact of X dosage in the two species is not a wholly surprising result given the different and complex dynamics of the X chromosome during early development, and the variations across in vitro culture models between and within each species. While naive female mESC consistently have two active X chromosomes, most widely used hPSC are more akin to the primed pluripotent state where X chromosome inactivation has taken place (Geens and Chuva De Sousa Lopes 2017; Sahakyan et al., 2017). If uniform X inactivation was found in all hPSC lines, then we would anticipate a uniformly lower protective effect of XX genotype than in mouse. However, human ESC and iPSCs often show erosion of X inactivation and increases in levels of X-linked transcripts and proteins (Bar et al., 2019; Brenes et al., 2021; Topa et al., 2024). Thus, across a large cohort of XX hPSCs, we would expect variable dosage effects on X-linked genes due to variations in the degree of X inactivation erosion and the regions of the X affected. Therefore, we would anticipate a variable degree of protective effect against aneuploidy in female hPSC, owing to variations in X chromosome activity, as we observed. X-linked tumor suppressor genes which escape X inactivation may also help prevent chromosomal CNV as has been suggested in the cancer field (Dunford et al., 2017). Our data revealed the novel discovery that cell lines with two active X chromosomes have an increased proportion of cells in the rare 2CL state at a given time (Figures 6D and 6E). There have been studies showing that fostering the 2CL state benefits the genomic stability of mESC, possibly through enhanced telomere maintenance (Guo et al., 2018). Another study hinted at a restorative role for the 2C state, noting the emergence of a population of euploid cells in a culture of cells derived from trisomic donors through overexpression of the 2C gene Zscan4 (Amano et al., 2013). More recently, Zscan4 was shown to interact with Parp1 to promote DNA repair (Tsai et al., 2023), and to bind to microsatellite repeats prone to instability to protect against DNA damage (Srinivasan et al., 2020). Elimination of 2C-like cells using a suicide gene driven by the MERV-L LTR increased DNA damage in mESC culture, downregulated DNA repair and resulted in increases in SNV, CNV, and chromosomal rearrangements. Together this evidence argues strongly that transit through the 2C-like state provides genetic stability to mESC (Du et al., 2024). Therefore, the observation that XX mESC lines have higher 2CL proportion and less likelihood to develop aneuploid cells than XY or XO lines is a strong indication that the two are related. Proof of this hypothesis will require some means of enhancing transit of ES cells through the 2C-like state over a long-term culture period, something that is not attainable with current methodologies. During the conception and most of the course of this project, the presence of a totipotent state in PSC culture was considered a mouse-specific phenomenon. As of writing, several papers have been published on the existence of 8-cell-like cells (8CLC) within human naive PSC cultures (Mazid et al., 2022; Moya-Jódar et al., 2022; Taubenschmid-Stowers and Reik 2023; Yoshihara and Kere 2023). These 8CLC have similar transcriptomes to the 2CLC of mouse cultures, expressing totipotency genes such as Zscan family genes and endogenous retroviruses (Mazid et al., 2022; Moya-Jódar et al., 2022). These subpopulations appear in small quantities—similar to the 2CLC in mouse—when using culture media designed to promote naive PSC characteristics, such as PXGL, 5iLA, 5iLAF, t2iLGö, and RSeT, but interestingly they were not seen in expanded potential stem cells (Mazid et al., 2022; Moya-Jódar et al., 2022). It is not known whether naive PSCs bearing two X chromosomes would be more stable genetically, or if they would show a higher proportion of 8C-like cells (Figure 6E). Human XX naive PSC show variability in the activation of the X chromosome with varying patterns of erosion (Sarel-Gallily and Benvenisty 2022). Existing datasets on naive hPSC are not large enough to enable genetic analysis of chromosomal stability (Nissim Benvenisty, personal communication). Thus, the relationship between X dosage, the 8C-like state, and genetic stability in naive PSC remains uncertain. Since no evidence for the 8CL state in human primed PSC exists, it is possible that this shared mechanism was the increased dosage of TSG expressed from chromosome X, as there is evidence that some TSG escape XCI (Dunford et al., 2017). We do not notice a shared pattern of which TSG are correlated with euploid or aneuploid cells. While Dusp9 and Fhl1 stand out from the bulk data, the scRNA-seq data show a more trisomy-specific suite of TSG with higher expression in euploid single cells (Figure S5). These results are not indicative of a lack of TSG involvement, but rather that no one specific TSG is responsible for the aneuploidy mitigation from chromosome X dosage. When co-cultured with male diploid mESC, cells bearing trisomy 11 expanded in proportions during serial passaging whereas they failed to do so in co-culture with female diploid cells. This suggests that the known effects of X-linked gene dosage on the maintenance of naive pluripotency in mESC (Schulz et al., 2014; Schulz 2017; Song et al., 2019) might diminish the competitive advantage of cells bearing recurrent aneuploidies in the presence of female versus male diploid cells. The competitive advantage of trisomy 11 cells in co-culture with male ESC was more apparent in suboptimal conditions (decreased levels of exogenous LIF). Together, the increased proportion of 2CL cells and increased expression of tumor suppressors may decrease the overall incidence of aneuploidy in female mESC cultures, while the increased stability of the pluripotent state in XX cells may hold back the competitive advantage of cells with recurrent aneuploidies, particularly under suboptimal culture conditions. Our results have implications for the prevention of recurrent genetic abnormalities in hPSC. The finding of synteny between affected genomic regions in mESC and hPSC suggests common underlying selective mechanisms and validates the mESC as a model for human cells. The lack of evidence for specific autosomal loci with a strong influence on development of aneuploidy suggests that variations in cell culture environment may be critical to the variable occurrence of aneuploidy across cell lines and cell labs. This is in accord with recent observations from Barbaric and co-workers (Stavish et al., 2024). Our results also confirm that culture conditions (reduced levels of LIF) can modulate the selection for aneuploid cells. Lastly, it has long been known in human that males are more susceptible to cancer development than females. Our work suggests that the mESC system might be a powerful tool to investigate the underlying mechanisms of this sex-dependent effect, particularly as aneuploidy is strongly associated with progression of malignancy in mouse PSCs (Xiao et al., 2024). Limitations of the study Comparison of results in mESC with hPSC have inherent limitations. There are insufficient data on human naive cells to assess the effects of the X chromosome complement on aneuploidy. In human primed cells, owing to variations in the erosion of X-inactivation and concomitant stochastic X-reactivation across cell lines, it is likely that there would be variations in the effect of X chromosome dosage on genetic stability or stem cell maintenance in the panel of cell lines examined. Further study of erosion of X-inactivation and aneuploidy in large panels of cell lines could resolve this question. Molecular mechanisms underlying X chromosome promotion of genetic stability through an increase in the proportion of cells in the 2C-like state, increased dosage of tumor suppressor genes, or enhanced stem cell maintenance and competitiveness likely involved multiple genetic loci and will require further investigation. Resource availability Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Martin F. Pera (martin.pera@jax.org). Materials availability This study generated a mESC trisomy 11 cell line bearing a mCherry reporter that is available from the lead contact. Data and code availability •Diversity Outbred single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287621. Diversity Outbred single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287621. •C3H/HeSn-Paf/J scRNA-seq single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287717. C3H/HeSn-Paf/J scRNA-seq single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287717. •Analysis code and additional files to run analyses have been made available through https://github.com/StantonAR/mESC_Aneuploidy. Analysis code and additional files to run analyses have been made available through https://github.com/StantonAR/mESC_Aneuploidy. •Additional information required to analyze the data in this paper is available from the lead contact upon request. Additional information required to analyze the data in this paper is available from the lead contact upon request. Acknowledgments We would like to thank Gregory Carter and Christopher Baker of The Jackson Lab and Donna Slonim of Tufts University School of Medicine for consistent feedback throughout the investigation. We would like to thank Peter Andrews of Sheffield University for insights about differentially expressed gene sets in PSCs. Funding sources: The Jackson Laboratory. Dr. Rick Maser of the JAX Genetic Resource Sciences Group developed the 36H mCherry reporter line. Dr. Ben Lowe (JAX Genetic Resource Sciences Group) and Ms. Carly Rodriquez (Pera Lab) performed dd-PCR karyotyping. The following JAX Scientific Service Laboratories contributed to this study: Single Cell Laboratory, Flow Cytometry, Genome Technologies. Author contributions Diversity outbred and C57BL6/J mESC line derivation, culture, and collection, M.P., L.G.R., T.C., A.C., and T.V.; C3H/HeSn-Paf/J mESC line derivation, A.C. and L.G.R.; mass spectrometry proteomics, S.P.G.; RNA-seq and proteomics dataset preprocessing, D.A.S. and S.A.; DO single-cell RNA-seq dataset, T.C.; computational analyses, A.S. with input from D.A.S. and S.C.M.; human PSC karyotyping, K.L., S.T., E.M., and T.L.; human PSC karyotype analysis, D.S. with statistical analysis by A.S.; C3H/HeSn-Paf/J mESC experiments, A.S.; competition experiments, T.M.; manuscript writing and editing, A.S. and M.F.P.; study conception and design by M.F.P. and A.S. All authors read and commented on the manuscript. Declaration of interests T.C. has an equity interest in Predictive Biology Inc. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesRabbit polyclonal anti-MERV-L gagEpiGentekCat# A-2801-050Lots 2106036 and 2503049Mouse anti-POU5F1 monoclonal antibodyBD BiosciencesCat # 611202; RRID: AB_398736Donkey anti-rabbit IgG Alexa 488InvitrogenCat# A21206donkey anti-mouse IgG H + L Alexa 647InvitrogenCat# A32787Chemicals, peptides, and recombinant proteinsDMEM High GlucoseInvitrogenCat#11960069Embryonic Stem Cell FBSThermo FisherCat#10439024Mouse LIFMilliporeCat#ESG1107CHIR99021Stem Cell TechnologiesCat#72052Ghost Dye Red 780Tonbo BiosciencesCat#13-0865-T100ParaformaldehydeElectron Microscopy SciencesCat#15710-SPerm/Wash BufferBD BiosciencesCat# BD 611202AccutaseSTEMCELL TechnologiesCat#07920Trypsin-EDTAGibcoCat#25300-054Cat#25200-072Critical commercial assaysChromium Single Cell 3′ Reagent Kits v3.110X GenomicsCat#PN-1000268PN-1000120Deposited dataDO mESC proteomicsAydin et al., 2023; ProteomeXchange (http://www.proteomexchange.org)PXD033001DO mESC RNA-seqSkelly et al., (2020); ArrayExpress (https://www.ebi.ac.uk/arrayexpress/)E-MTAB-7728DO scRNA-seqThis studyhttps://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287621C3H/HeSn-Paf/J scRNA-seqThis studyhttps://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287717Experimental models: Cell linesDiversity Outbred mESCsPredictive Biology Inc.N/AC3H/HeSn-Paf/J mESCsLaboratory of Laura Reinholdt, JAXN/AC57Bl6J mESCLaboratory of Laura Reinholdt, JAXN/AC57Bl6/J mESC 40XY +11 with mCherry reporter in Cftr locusLaboratory of Martin Pera, JAXN/AMouse embryonic fibroblasts (MEFs) derived from C57Bl/6JLaboratory of Martin Pera, JAXN/AExperimental models: Organisms/strainsMouse: J:DOThe Jackson LaboratoryJAX: 009376Mouse: C3H/HeSn-Paf/JThe Jackson LaboratoryJAX: 001529OligonucleotidesLMO and LMO anchorLaboratory of Chris BakerLaboratory of Chris McGinnisPrimers for digital droplet PCR karyotypingJAX Genome Research SciencesN/ASoftware and algorithmsAnnotationHubBioconductorhttps://bioconductor.org/packages/release/bioc/html/AnnotationHub.htmlbeanplotRdocumenationhttps://www.rdocumentation.org/packages/beanplot/versions/1.3.1/topics/beanplotBioCircosbioCircos.jshttp://bioinfo.ibp.ac.cn/biocircos/biomaRtBioconductorhttps://bioconductor.org/packages/release/bioc/html/biomaRt.htmlbootN/AN/AclusterProfilerBioconductorhttps://bioconductor.org/packages/release/bioc/html/clusterProfiler.htmlCONICSmatDiaz lab, githubhttps://github.com/Neurosurgery-Brain-Tumor-Center-DiazLab/CONICS/blob/master/CONICSmat/R/InferBreakpoints.Rcowplotgithubhttps://github.com/wilkelab/cowplotdata.tableCRAN.Rhttps://r-datatable.com/DescToolsCRAN.Rhttps://cran.r-project.org/web/packages/DescTools/index.htmlDESeq2Bioconductorhttps://bioconductor.org/packages/release/bioc/html/DESeq2.htmldevtoolsCRAN.Rhttps://cran.r-project.org/web/packages/devtools/index.htmlDOSEBioconductorhttps://www.bioconductor.org/packages/release/bioc/html/DOSE.htmldplyrCRAN.Rhttps://cran.r-project.org/web/packages/dplyr/index.htmlensembldbBioconductorhttps://bioconductor.org/packages/release/bioc/html/ensembldb.htmlGGallyCRAN.Rhttps://cran.r-project.org/web/packages/GGally/index.htmlggcytogithubhttps://github.com/RGLab/ggcytoggplot2CRAN.Rhttps://cran.r-project.org/web/packages/ggplot2/index.htmlggpubrCRAN.Rhttps://cran.r-project.org/web/packages/ggpubr/index.htmlggridgesCRAN.Rhttps://cran.r-project.org/web/packages/ggridges/index.htmlggsciCRAN.Rhttps://cran.r-project.org/web/packages/ggsci/index.htmlggExtraCRAN.Rhttps://cran.r-project.org/web/packages/ggExtra/index.htmlggbeeswarmCRAN.Rhttps://cran.r-project.org/web/packages/ggbeeswarm/index.htmlgplotsCRAN.Rhttps://cran.r-project.org/web/packages/gplots/index.htmlggpointdensityCRAN.Rhttps://cran.r-project.org/web/packages/ggpointdensity/index.htmlmiscToolsCRAN.Rhttps://cran.r-project.org/web/packages/miscTools/index.htmlmixtoolsCRAN.Rhttps://cran.r-project.org/web/packages/mixtools/index.htmlorg.Mm.e.g.,.dbBioconductorhttps://bioconductor.org/packages/release/data/annotation/html/org.Mm.eg.db.htmlpheatmapCRAN.Rhttps://cran.r-project.org/web/packages/pheatmap/index.htmlplotlyCRAN.Rhttps://cran.r-project.org/web/packages/plotly/index.htmlplotrixCRAN.Rhttps://cran.r-project.org/web/packages/plotrix/index.htmlqtl2CRAN.Rhttps://cran.r-project.org/web/packages/qtl2/index.htmlreadrCRAN.Rhttps://cran.r-project.org/web/packages/readr/index.htmlreadxlCRAN.Rhttps://cran.r-project.org/web/packages/readxl/index.htmlreshape2CRAN.Rhttps://cran.r-project.org/web/packages/reshape2/index.htmlRtsneCRAN.Rhttps://cran.r-project.org/web/packages/Rtsne/index.htmlscrangithubhttps://github.com/elswob/SCRANSeuratSatija Lab githubhttps://github.com/satijalab/seuratstatsN/AN/AsquashCRAN.Rhttps://cran.r-project.org/web/packages/squash/index.htmlsurvcompCRAN.Rhttps://cran.r-project.org/web/packages/survcomp/index.htmltidyverseCRAN.Rhttps://cran.r-project.org/web/packages/tidyverse/index.htmlungevizWilke Lab githubhttps://github.com/wilkelab/ungevizUpSetRCRAN.Rhttps://cran.r-project.org/web/packages/UpSetR/index.htmlviridisCRAN.Rhttps://cran.r-project.org/web/packages/viridis/index.html Experimental model and study participant details Diversity outbred mESC aneuploidy The DO mESC lines used in this study were the same lines published in (Skelly et al., 2020; Aydin et al., 2023) and derived, cultured, and harvested as described in those methods sections. Briefly, DO mice were purchased from The Jackson Laboratory and bred at Predictive Biology Inc. The ICM from the resulting embryos were prepared as described in (Czechanski et al., 2014) to derive mESC. Cultures were tested for the presence of mycoplasma at the beginning and termination of the experiment, which ranged from passage 6–16 and was uniform for all cell lines in this study. For all ESC lines, pluripotency was assessed by germline chimera competence and cell identity by DNA sequencing. Prospective validation study mESC used to validate the incidence of aneuploidies and 2CLC differences between cell lines of varying chromosomal sex were derived from a cross between C3H/HeSn and Paj/J mice (JR #001529), designed to yield XO and XXY in addition to XX and XY embryos due to non-disjunction of the sex chromosomes. Three independently derived cell lines from each XX, XO, and XY embryos were obtained from the lab of Laura G. Reinholdt and maintained independently. Competition experiments A male C57BL/6J cell line bearing trisomy 11 and a Bxb1 landing pad in the Cftr locus was transduced with an mCherry reporter gene using Bxb1 integrase mediated recombination. Clone 36H6 was selected for high levels of mCherry expression using flow cytometry and fluorescence microscopy. Diploid male and female C57BL/6J cell lines were obtained from the laboratory of Laura Reinholdt. Population doubling times of 36H6 and all diploid cell lines grown in isolation were similar. Cell lines were analyzed by G-banded karyotype after derivation, and their ploidy was checked immediately prior to experimentation (3–4 passages after derivation) using dd-PCR with probes for Chromosomes 1, 6, 7, 8, 10, 11, 19, X and Y. Absence of mycoplasma infection was confirmed at this stage. Pluripotency was assessed by germ line transmission (diploid cells only). Method details Diversity outbred mESC aneuploidy Following derivation in serum + 2i/LIF media, mESC were transferred to a serum + 1i/LIF media (Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 15% mouse embryonic stem cell grade fetal bovine serum, 100 U/mL Penicillin-Streptomycin, 2mM GlutaMAX, 0.1mM non-essential amino acids, 1mM sodium pyruvate, 0.1mM 2-mercaptoethanol, approximately 2000U/ml LIF, and 3uM CHIR99021), and slowly tapered off feeder-based culture through expansion at a 1:4 split ratio every 3–4 days until collection for bulk or scRNA-seq and mass spectrometry proteomics. RNA was harvested from DO mESC at passage 7–9. Prospective validation study Cells were cultured on plates coated in gelatin and irradiated MEF feeders (50 Gy delivered from a Cesium-137 source) at 1E5 cells/cm2 and maintained on ESM (DMEM, 15% mouse embryonic stem cell grade fetal bovine serum from unselected lots, 1% Penicillin/Streptomycin, 1% GlutaMAX-1, 1% MEM NEAA, 100mM Sodium Pyruvate, 55 mM BME, 107 units LIF) by passage every 3 days at a 1:3 to 1:4 split ratio. Maintenance of XX karyotypes in diploid female lines in these experiments and in the competition experiments described in the next section was likely facilitated by routine passage at subconfluence (50–70%) and modest split ratio (1:3 to 1:4) and the combination of feeder cell support, serum and 2i conditions. Competition experiments Diploid male or female C57BL/6J cell lines were harvested and plated out in co-culture with Trisomy 11 36H6 cells at a 90:10 percent ratio (total 1E5 cells) on 6 well plates coated in gelatin and irradiated MEF feeders and maintained on ESM (15% mouse embryonic stem cell grade fetal bovine serum, 1% Penicillin/Streptomycin, 1% GlutaMAX-1, 1% MEM NEAA, 100mM Sodium Pyruvate, 55 mM BME, 107 units LIF) by passage every 3 days at a 1:3 to 1:4 split ratio. The cultures were harvested using Accutase for flow cytometry analysis, and were serially passaged every 3 days for a total of three passages in three separate experiments. 2–3 technical replicates (wells) for each culture were obtained at every time point. In some cultures, the added LIF concentration was reduced to 5E6, 1E6 or 2E5 units/ml. Diversity outbred mESC bulk RNA-seq RNA-seq data was collected, assayed, and processed as described in (Skelly et al., 2020; Aydin et al., 2023) with the following exceptions. Whereas (Skelly et al., 2020)uses the single-end version of this dataset, we use the paired-end data used in (Aydin et al., 2023). Both previous papers used reads that had been rank Z transformed, but here we omitted this step as this transformation removes the inherent structure necessary to identify CNVs from non-normally distributed expression values. Before analyses, genes not expressed in all cell lines, gene models, and Riken genes were excluded. Diversity outbred mESC proteomics Mass spectrometry proteomics data was collected and assayed as described in (Aydin et al., 2023) and processed as in (Chick et al., 2016) with the following exceptions. Like the bulk RNA-seq data, these data have not been rank Z transformed in order to preserve the structure necessary to identify CNVs from non-normally distributed expression values. Before analyses, genes not expressed in all cell lines, gene models, and Riken genes were excluded. Diversity outbred mESC single-cell RNA-seq DO mESC 10X Genomics scRNA-seq gene x cell matrix was provided by Ted Choi of predictive Biology. Data processing was performed mostly as described in the standard R/Seurat package tutorial. Only cells with less than 10% of total reads from mitochondrial genes, greater than 2,000 detected genes, and that had unique cell line identification were kept for further analysis. Cell cycle scoring was performed using mouse cell cycle genes from https://raw.githubusercontent.com/hbc/tinyatlas/master/cell_cycle/Mus_musculus.csv, but no significant effects were identified that would necessitate regression. Isogenic mESC single-cell RNA-seq For sequencing, cells were dissociated as they were for flow cytometry (see below for described methods). Once lifted, cells were allowed to settle in a 10cm culture dish to remove MEF feeders due to the differential attachment rate of mESC and feeders. The remaining cells were then passed through a 40 micron filter to remove clumping cells. Then, 500,000 cells from each cell line were labeled with lipid modified oligos (LMO) as described in (McGinnis et al., 2019). Briefly, cells were washed and resuspended in ice-cold PBS before incubating with 1:1 molar ratio of barcode oligo and LMO anchor (supplied by Chris Baker from the lab of Chris McGinnis) for 5 min at room temperature. Then, cells were incubated with LMO co-anchor (also from Chris Baker) on ice for 5 min. LMOs were blocked with 1% BSA and 1% FBS in ice-cold PBS, washed in the same solution, and sent for 10X library prep and sequencing in The Jackson Laboratory single-cell biology core facility. Cell viability was assessed on a Luna FX7 automated cell counter (Logos Biosystems), and up to 40,000 cells (4,000 cells per LMO) were loaded onto one lane of a 10X Chromium Controller. Single cell capture, barcoding and library preparation were performed using the 10X Chromium platform (Zheng et al., 2017) version 3.1 NEXTGEM chemistry and according to the manufacturer’s protocol (#CG00315) with modifications for generating the LMO libraries. cDNA and libraries were checked for quality on Agilent 4200 Tapestation, quantified by KAPA qPCR, and libraries were pooled using a ratio of 95% gene expression library and 5% LMO library before sequencing; each gene expression-LMO library pair was sequenced at 91% of an Illumina NovaSeq 6000 S4 flow cell lane, targeting 20,000 barcoded cells with an average sequencing depth of 100,000 read pairs per cell. Illumina base call files for all libraries were converted to FASTQ using bcl2fastq v2.20.0.422 (Illumina) and FASTQ files associated with the gene expression libraries were aligned to the GRCm38.93 reference assembly with GENCODE vM23 annotations (10x Genomics mm10 reference 2020-A) using the version 6.1.2 Cell Ranger count pipeline (10x Genomics). FASTQ files representing the LMO libraries were processed into cell x LMO digital count matrices using CITE-Seq-Count (version 1.4.5; https://zenodo.org/badge/latestdoi/99617772). Flow Cytometry A single-cell suspension was collected from each line by dissociation with 1X Accutase. Cells were stained for viability using Ghost Dye Red 780 on ice for 20 min. Immediately after, samples were fixed with 4% PFA on ice for 15 min, quenched in subsequent washes with 0.125 mM glycine pH 3.0. Fixed cells were permeabilized in 1X Perm/Wash buffer on ice for 15 min. Samples were incubated in 1:6000 MERVL-gag primary antibody at 4C for 1 h. Samples were incubated in donkey anti-rabbit IgG Alexa 488 secondary antibody on ice for 20 min, then data was collected on the BD FAC Symphony A5. Quantification and statistical analyses Computational analyses virtual karyotyping of bulk transcriptomics and proteomics Virtual karyotyping of diversity outbred mESC bulk data was performed on batch corrected expression data that had not been rank Z transformed. Gene models and Riken genes were removed from the dataset before analysis. Chromosome summary statistics for each sample were generated by calculating the median of expression values for all genes on autosomes and sum of expression values for Chromosome X. Chromosomes with CNVs were identified by Shapiro-Wilk normality test, where chromosomes determined non-normal by p ≤ 0.01 for RNA and protein were determined to represent reoccurring CNV. Specific samples harboring these abnormalities were identified by bootstrapping gene expression values to calculate p-values for RNA and protein chromosomal summary statistics in each sample, then combined using the Z-transformed method. Samples with a conservative combined p ≤ 0.0001 were annotated as aneuploid. Virtual karyotyping of diversity outbred single-cell RNA-seq Identification of aneuploid cells in single-cell RNA-seq data was performed using the R/CONICSmat package (Müller et al., 2018). Each cell line was analyzed separately from the other 11 to keep expression comparisons limited to isogenic backgrounds. Cells with p ≤ 0.1 were annotated as aneuploid for Chromosomes 1, 6, 8, and 11. Estimation of Aneuploid Cell content in bulk datasets The fold-change in gene expression across duplicated chromosomes in aneuploid cells were quantified using the ‘detectBreakpoints’ function in the R/CONICSmat package. Cells with only one trisomic chromosomes were compared to cells with a completely euploid karyotype. A summary statistic was calculated by taking the median of all genes across the chromosome. The aneuploid content of bulk RNA-seq data can then be determined by the formula: Where ‘Proportion aneup.’ is the proportion of cells in the sample that have the chromosome in question duplicated, ‘Line’ is the summary statistic for the cell line whose aneuploid proportion is being calculated, ‘Median eup.’ is the median of the summary statistic for lines annotated as euploid, and 0.4 is the proportional increase in expression expected if all cells in the sample were aneuploid for the chromosome in question, determined empirically. Synteny between mouse and human PSC duplications Identification of syntenic genes between mouse and human genomes was collected from The Jackson Lab Synteny Browser (syntenybrowser.jax.org). Visualization of connections between shared genes on chromosomes/segments duplicated in mouse and human PSCs was generated using the R/BioCircos package(Cui et al., 2016). Mouse Chromosomes 1, 6, 8, and 11 were identified by this study. Human Chromosomes 1, 1q32, 7, iso7p, 12, 12p13.3, 14, 17, 17q25, 20, and X were identified from(Maitra et al., 2005; Mitalipova et al., 2005; Catalina et al., 2008; Lefort et al., 2008; Spits et al., 2008; Mayshar et al., 2010; Amps et al., 2011; Halliwell et al., 2020). Differential expression analysis Differential expression was performed using R/Seurat package for single-cell RNA-seq. The function ‘FindMarkers’ was used with the default Wilcoxon rank-sum test parameter to find differentially expressed genes between each comparison. GO term enrichment was performed using the R/ClusterProfiler package. Biological process ontologies of differentially expressed genes with log2 fold-change > |0.3| between trisomy 8 and completely euploid cells were visualized using the ‘gseGO’ function in the R/clusterProfiler package (minGSSize = 10, maxGSSize = 800, pvalueCutoff = 0.1) showing 16 categories split between activated and suppressed. Genetic mapping of aneuploidy Genetic mapping of aneuploidy states of the samples in the bulk datasets were performed using the R/qtl2 package (Broman et al., 2019) using ploidy as a binary trait. Significance thresholds were calculated by running permutation analysis 1,000 times, generating a LOD = 8.74 for p = 0.01, LOD = 7.75 for p = 0.05, and LOD = 7.32 for p = 0.01. X-linked Gene analyses Correlations between transcript and protein abundance were calculated using the R/stats package. The 6 genes implicated in the sex differences in pluripotency of mESC were collected from(Schulz et al., 2014), the tumor suppressor genes on Chromosome X were collected from The Tumor Suppressor Gene Database (bioinfo.uth.edu/TSGene). 2CLC Gene expression in bulk dataset Lists of genes upregulated in Zscan4+ cells in mESC cultures were collected from (Macfarlan et al., 2012; Amano et al., 2013) (supplemental data). Gene lists were filtered for genes with expression data in our bulk RNA-seq dataset before calculating the expression z-scores for each remaining gene present across all samples. Significance between sum of z-scores for each chromosomal sex were calculated by t test using the R/stats package. LMO demultiplexing and preprocessing of isogenic scRNA-seq Data processing was performed mostly as described in the standard R/Seurat package tutorial. Only cells with less than 4% of total reads from mitochondrial genes, greater than 4,500 detected genes in the early dataset and 4,000 detected genes in the aged dataset were kept for further analysis. LMO barcode demultiplexing was performed using the ‘MULTIseqDemux’ function in the R/Seurat package. These cell identifications were found not to coincide with unsupervised clustering or XX biased gene expression in downstream analyses and were regarded as uninformative due to the low LMO barcode read counts. Cell cycle scoring of isogenic scRNA-seq Cell cycle scoring was performed the same as in the DO mESC scRNA-seq dataset – however, to ensure no subtle differences between chromosomal sexes due to cell cycling were observed, the difference between G2M and S phase scores were regressed out during scaling of the data. Identification and characterization of pluripotent state, differentiation primed, MEF, and 2CLC clusters Pluripotent clusters were identified using expression of several key genes associated with general, naive, and primed pluripotency. Additionally, several genes differentially expressed in XX and XY mESC were chosen to aid in identifying chromosomal sex. Genes were selected using information from (Schulz et al., 2014; Song et al., 2019; Wang et al., 2020; Genolet et al., 2021; Endoh and Niwa 2022; Lentini et al., 2022). MEF were identified using the expression of Thy1 and by predominant G1 cell cycle phase. 2CLC were identified using expression of the two Zinc finger and SCAN domain containing 4 (Zscan4) genes expressed in preimplantation development: Zscan4c and Zscan4d. Clustering resolution was set just sensitive enough that the 2CLC cluster separated out, as it was the smallest cluster of interest. Virtual karyotyping of isogenic scRNA-seq Cells with chromosomal CNV were identified using the same R/CONICSmat pipeline used for the DO mESC scRNA-seq data except that cells were not processed with the ‘SCTransform’ function. Visualization Most plots were generated using the R/ggplot2 package. QQ plots were generated using the R/ggpubr package, UpSet plots were generated using the R/UpSetR package, points underneath boxplots were plotted using R/ggplot2 with the ‘geom_quasirandom’ function from the R/ggbeeswarm package. Published: June 9, 2026 Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.stemcr.2026.102933. Supplemental information Document S1. Figures S1–S6 Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Figures S1–S6 Document S2. Article plus supplemental information Data Availability Statement •Diversity Outbred single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287621. Diversity Outbred single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287621. •C3H/HeSn-Paf/J scRNA-seq single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287717. C3H/HeSn-Paf/J scRNA-seq single-cell RNA-seq data have been deposited in https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE287717. •Analysis code and additional files to run analyses have been made available through https://github.com/StantonAR/mESC_Aneuploidy. Analysis code and additional files to run analyses have been made available through https://github.com/StantonAR/mESC_Aneuploidy. •Additional information required to analyze the data in this paper is available from the lead contact upon request. Additional information required to analyze the data in this paper is available from the lead contact upon request.