dCas9-mediated dysregulation of gene expression in human induced pluripotent stem cells during primitive streak differentiation

CRISPR-based systems have fundamentally transformed our ability to study and manipulate stem cells. We explored the possibility of using catalytically dead Cas9 (dCas9) from S. pyogenes as a platform for targeted epigenetic editing in stem cells to enhance the expression of the eomesodermin gene (EOMES) during differentiation. We observed, however, that the dCas9 protein itself exerts a potential non-specific effect in hiPSCs, affecting the cell ’ s phenotype and gene expression patterns during subsequent directed differentiation. We show that this effect is specific to the condition when cells are cultured in medium that does not actively maintain the pluripotency network, and that the sgRNA-free apo-dCas9 protein itself influences endogenous gene expression. Transcriptomics analysis revealed that a significant number of genes involved in developmental processes and various other genes with non-overlapping biological functions are affected by dCas9 over-expression. This suggests a potential adverse phenotypic effect of dCas9 itself in hiPSCs, which could have implications for when and how CRISPR/Cas9-based tools can be used reliably and safely in


Introduction
The term pluripotency refers to a cell's ability to differentiate into any cell or tissue of the human body's three germ layers (Itskovitz-Eldor et al., 2000).To maintain stem cells in such a state, chromatin of developmentally important genes is arranged in a bivalent state that keeps the genes silenced yet poised for activation.This state is characterized by the co-enrichment of both repressive (H3K27me3) and active (H3K4me1 and H3K4me3) chromatin marks at corresponding regulatory regions (Azuara et al., 2006;Bernstein et al., 2006;Mikkelsen et al., 2007;Pan et al., 2007;Zhao et al., 2007).The repressive H3K27 methylations are added and maintained by Polycomb Repressive Complex 2 (PRC2), which is counteracted by members of the complex of proteins associated with SET1 (COMPASS)-like family writing activating H3K4 methylations (Piunti and Shilatifard, 2016).Once differentiation is triggered, the cellular transcription machinery will initiate the process of resolving bivalent chromatin in favour of either an activated or repressed gene expression state (Voigt et al., 2013).As cells progress through distinct developmental trajectories, bivalency is gradually resolved at key regulatory loci, and this is hypothesized to ensure tight regulation of tissue-specific genes to prime the cells for downstream lineage bifurcation steps.Indeed, depletion of key H3K27-and H3K4-specific histone methyltransferases has been shown to reduce self-renewal and impair proper stem cell differentiation (Collinson et al., 2016;Lubitz et al., 2007).Furthermore, inherent epigenetic variability within bivalent chromatin motifs positively correlates with heterogeneity observed within and across induced human pluripotent stem cell (hiPSC) lines, significantly contributing to variable differentiation propensities (Carcamo-Orive et al., 2017).
The type II CRISPR/Cas9 system from S. pyogenes is an RNA-guided DNA endonuclease that can be programmed to target and cleave a specific DNA sequence through an engineered single guide RNA (sgRNA) (Jinek et al., 2012).The ease with which this platform can be programmed to specifically and robustly cleave target DNA sequences has significantly advanced our ability to add or delete genetic material, or to edit the genome by replacing an existing segment with a desired DNA sequence (Bressan et al., 2017;Cong et al., 2013;Hsu et al., 2014;Mali et al., 2013).Additionally, the discovery that specific mutations within the Cas9 protein abolish its nucleolytic activity without impacting sgRNA binding or DNA targeting has greatly expanded the range of CRISPR/Cas9-based tools (Qi et al., 2013).By coupling various effector proteins to this catalytically dead Cas9 (dCas9), gene expression can be activated or suppressed in a sequence-targeted manner (Balboa et al., 2015;Chakraborty et al., 2014;Chavez et al., 2015;Genga et al., 2019;Gilbert et al., 2014;Kearns et al., 2014;Nunez et al., 2021;Tian et al., 2019).
Although modulation of gene expression has been achieved using dCas9 fused with synthetic trans-activators, these artificial systems typically only act as scaffolds to recruit the native transcriptional regulatory machinery.Therefore, the long-term effectiveness and robustness of these gene programming modules remain limited, since they lack the catalytic ability to fully activate the loci needed to establish lineagespecific gene regulatory networks.In this work, we explore the possibility of using dCas9 as a platform for targeted epigenetic editing of bivalent promoter motifs in hiPSCs.Specifically, we explore whether the epigenetic bias within the bivalent chromatin can be shifted towards a more active state.By writing active H3K4me marks using H3K4-specific methyltransferase (MLL2 and MLL4) a more efficient gene activation during differentiation should be facilitated/promoted.As a target gene, we choose eomesodermin (EOMES), a T-box transcription factor that is maintained in a bivalent state in stem cells and is involved in regulating the exit from pluripotency towards the primitive streak and subsequently definitive endoderm specification (Arnold et al., 2008;Tosic et al., 2019).However, we found that targeting the epigenetic modifiers to the EOMES loci by fusing dCas9 to the catalytic SET domain from MLL2 or MLL4, does not promote an increase in EOMES activation during primitive streak differentiation.Instead, we discovered a potential non-specific effect of dCas9 overexpression in hiPSCs that alters gene expression levels during primitive streak differentiation.We further show that this effect is specific to the situation when the cells are cultured in medium that does not actively maintain pluripotency, and that the effect is independent of sgRNAs targeting dCas9 to the genome.Transcriptomics analysis revealed that a significant number of genes involved in developmental processes and various other genes with non-overlapping biological functions are affected, suggesting a potential global phenotypic effect of dCas9 overexpression in hiPSCs.

H3K4me-specific epigenetic editing in hiPSCs fails to enhance EOMES gene activation during primitive streak differentiation
To specifically promote epigenetic writing of H3K4me marks at defined sites, we fused the minimal functional catalytic SET domain of MLL2 (MLL2 Pre-SET-Post ) to catalytically dead Cas9 (dCas9) from S. pyogenes (Li et al., 2016;Zhang et al., 2015).MLL2 is a member of the mixed lineage leukemia (MLL) protein family of histone methyltransferases.In pluripotent stem cells, MLL2 is responsible for writing and maintaining H3K4me3 of bivalent promoters (Denissov et al., 2014;Hu et al., 2013), and should hence be able to reinforce the writing of activating H3K4me3 marks in the promoter region of EOMES.We designed four single guide RNAs (sgRNAs) for a putative regulatory region proximal to the transcription start site of EOMES (Tables S1 and  S2) and tested their performance to target dCas9 fused to p300core to the region and promote EOMES expression in HEK293T cells (Fig. 1A).The epigenome editor p300core was selected for the sgRNA screening assay because p300core-mediated H3K27Ac has been shown to promote robust activation of promoters and enhancers (Hilton et al., 2015;Klann et al., 2017).The four gRNAs synergistically increased EOMES expression levels (Fig. 1A).Next, we generated stable hiPSC clones encoding (i) V. Haellman et al. an inducible dCas9-MLL2 Pre-SET-Post expression construct that can be controlled by exogenous doxycycline (Dox) administration and (ii) constitutive expression of all four EOMES targeting sgRNAs (idCas9-MLL2 Pre-SET-Post -sgEOMES-hiPSCs).
To assess whether dCas9-MLL2 Pre-SET-Post can promote higher levels of H3K4me3 within the EOMES promoter regions, the idCas9-MLL2 Pre-SET-Post -sgEOME-S-hiPSCs were cultured for 72 h in pluripotency medium (mTeSR) without or with Dox administration and then seeded as single cells before primitive streak (PS) differentiation was triggered (Fig. 1B).Evaluating the relative levels of H3K4me3 and H3K27me3 within the EOMES promoter regions after the 72 h Dox administration showed that dCas9-MLL2 Pre-SET-Post did not alter the ratio between activating (H3K4me3) and repressive (H3K27me3) histone modifications (Fig. 1C).Furthermore, no significant effect on EOMES expression levels following PS induction was observed in the dCas9-MLL2 Pre-SET-Post stimulated cells (Fig. 1D).
hiPSCs require the exogenous addition of transforming growth factor-β (TGFβ) and high levels of fibroblast growth factor (FGF) in order to actively maintain pluripotency and self-renewal (Amit et al., 2004;Chen et al., 2011;Ludwig et al., 2006;Xu et al., 2005).We hypothesized that the effect of dCas9-MLL2 Pre-SET-Post -mediated epigenetic editing might not be sufficient to overcome these exogenous signals.To promote higher levels of H3K4me modification, we isolated the minimal functional catalytic SET domain of MLL4 (MLL4 Pre-SET-Post ), which natively has a higher catalytic activity than that of MLL2 (Zhang et al., 2015).After generating a stable hiPSC line encoding Dox-inducible dCas9-MLL4 Pre-SET-Post expression and the sgEOMES expression construct (idCas9-MLL4 Pre-SET-Post -sgEOMES-hiPSCs), we first tested whether dCas9-MLL4 Pre-SET-Post can induce EOMES gene activation in a minimal base medium that does not actively maintain the pluripotency network.After 72 h of continuous Dox-induced dCas9-MLL4 Pre-SET-Post expression, we observed a clear increase in EOMES expression levels compared to the uninduced cells (Fig. 2A), demonstrating that MLL4 Pre-SET-Post can activate the EOMES loci.However, when the cells were propagated in pluripotency medium, dCas9-MLL4 Pre-SET-Post failed to promote an increase in EOMES activation following PS induction (Fig. 2B).To confirm expression and nuclear translocation of the MLL2 and MLL4 dCas9-fusion constructs, which are tagged with multiple nuclear localization signal (NLS) peptides, we performed a transient overexpression experiment in HEK-293T cells, and subsequently immunostained the cells for the dCas9 protein.Imaging revealed distribution of dCas9 between the cytosol and nucleus for both MLL2 and MLL4 fusion constructs (Fig. S1).Taken together, these results indicate that dCas9-MLL4 Pre-SET-Post cannot activate EOMES expression in the presence of the exogenous growth factors used for maintenance of the pluripotency network.

dCas9-mediated-sgRNA-independent dysregulation of gene expression in hiPSCs during primitive streak differentiation
Next, to circumvent the effects of the pluripotency medium, we devised a different procedure for dCas9-MLL4 Pre-SET-Post -mediated priming of the EOMES loci.Instead of inducing dCas9-MLL4 Pre-SET-Post expression during propagation in pluripotency medium, dCas9-MLL4 Pre- SET-Post was induced for 24 h in a minimal base medium before PS differentiation was triggered (Fig. 3A).As a control we included a stable hiPSC line encoding the sgEOMES expression construct and Doxinducible expression of dCas9 without any epigenetic modifier (idCas9-sgEOMES-hiPSCs). Intriguingly, both dCas9-MLL4 Pre-SET-Post (Fig. 3B) and dCas9 without an epigenetic modifier (Fig. 3C) were able to increase EOMES activation.Genotyping of the dCas9-expressing hiPSC line confirmed integration of the respective dCas9 construct, thus eliminating the possibility that the two cell lines could have been accidently mixed (Fig. S2).To confirm that the effect is not due to Dox administration, or simply a consequence of induced expression of an exogenous protein, we performed control differentiation experiments then seeded as single cells before primitive streak (PS) induction was triggered for 24 h with exogenous Wnt (GSK3 inhibition CHIR99021) and TGF-β (Activin A) induction.Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 uninduced (-Dox) idCas9-MLL4 Pre-SET-Post -sgEOMES-hiPSCs.Bars represent the mean ± sem of n = 4 biologically independent samples.***P < 0.001; ns, not significant.
V. Haellman et al. using untransformed hiPSCs and stable hiPSCs encoding Dox-regulated expression of EYFP (iEYFP-hiPSCs).While significantly increased EOMES mRNA levels were observed on day 1 of differentiation (Fig. S3A), we found no significant difference between Dox-treated and non-treated hiPSCs or iEYFP-hiPSCs (Fig. S3B).Collectively, these results support the idea that dCas9 itself has a direct influence on hiPSCs, enhancing EOMES gene expression during PS differentiation.
Repeating the same experimental setup with a newly generated idCas9-sgEOMES-hiPSCs stable line confirmed that dCas9 itself can promote a significant increase in EOMES activation following PS induction (Fig. 4A).These results indicate that the observed effect is likely not caused by a locus-specific effect due to random integration of the  Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 uninduced (-Dox) idCas9-sgEOMES-hiPSCs.Bars represent the mean ± sem of n = 4 biologically independent samples.(B) Scheme depicting the protocol used to evaluate the effect of dCas9 expression on PS differentiation.idCas9-sgEOMES-hiPSCs were cultured first for 24 h in pluripotency medium (mTeSR), then for 24 h in base medium consisting of RPMI supplemented with sodium ascorbate, B-27 supplement, followed by PS induction with CHIR99021 (1 μM).dCas9 was either maintained uninduced (-Dox) or induced on day − 1 for 24 h, day − 2 for 48 h, day − 1 for 48 h, or day − 2 for 72 h by the addition of Dox (0.5 μg/mL), as illustrated on the scheme.Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 uninduced (-Dox) idCas9-sgEOMES-hiPSCs.Bars represent the mean ± sem of n = 4 biologically independent samples.(C) Scheme depicting the protocol used to evaluate the effect of the base medium composition during dCas9 induction.idCas9-sgEOMES-hiPSCs were cultured for 24 h in either (1) pluripotency medium (mTeSR), or base medium RPMI supplemented with sodium ascorbate and B-27 supplement including (2) bFGF (25 ng/mL) and TGFB1 (0.5 ng/mL), ( 3) bFGF (25 ng/mL), or (4) without any additional supplement.Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 uninduced (-Dox) idCas9-sgEOMES-hiPSCs cultured in the respective media conditions.Bars represent the mean ± sem of n = 4 biologically independent samples.*P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant.
V. Haellman et al. stable expression constructs.To determine whether this effect is directly caused by dCas9 induction in base medium or if there is a carryover effect of residual dCas9 during PS differentiation, dCas9 was transiently induced on different days and for different durations (Fig. 4B).Induction of dCas9 in base medium produced the highest increase in EOMES expression following PS induction, and extended induction prior to the switch to base medium did not increase the expression level of EOMES.However, any extended induction of dCas9 during PS differentiation reduced the EOMES expression to the same level as in the un-induced control cells.This indicates that the observed effect on EOMES expression is specific to the situation when dCas9 is induced, while the cells are maintained in base medium.By inducing dCas9 in base medium supplemented with reduced amounts of either basic fibroblast growth factor (bFGF) and transforming growth factor-β (TGFβ), bFGF alone, or no supplementary growth factor, we further confirmed that the enhanced activation of EOMES is specific to cells cultured in reduced medium conditions (Fig. 4C).
The core gene regulatory network controlling PS specification consists of EOMES, mix paired-like homeobox 1 (MIXL1), and brachyury (T) (Izumi et al., 2007).Hierarchically, EOMES acts upstream of MIXL1 as a positive regulator, and together EOMES and MIXL1 both act as negative regulators of T expression (Fig. 5A).Surprisingly, when characterizing the relative expression levels of EOMES, MIXL1, and T, we observed a general upregulation of all three genes (Fig. 5B).This suggests that the observed effect of dCas9 induction is not specific to EOMES, but instead is a potential global effect.As a control, we generated a stable hiPSC line that only harbours the Dox-inducible dCas9 expression construct and not the sgRNA expression construct (idCas9 hiPSCs).Strikingly, in the same experimental setup as used for the idCas9-sgEOMES hiPSCs line, dCas9 alone in the idCas9 hiPSCs line produced comparable upregulation of EOMES, MIXL1, and T following PS induction (Fig. 5C).Together, these results suggest that the sgRNA-free apo-dCas9 protein might exert an as-yet-unknown unspecific effect in hiPSCs.
Next we analyzed the transcriptome of both idCas9-sgEOMES and idCas9 cell lines before (day 0) and after PS differentiation (day 1), without (no Dox) and with dCas9 induction (Dox-treated).Analysis of the relative mRNA levels of dCas9 following 24 h induction confirmed significant increases in dCas9 transcripts at day 0 with Log2 (foldchange) values of 6 and 11 for idCas9-sgEOMES-and idCas9-hiPSC lines, respectively.Following Dox withdrawal, the fold-change values were dramatically decreased at day 1 of primitive streak differentiation for both cell lines (Table S3).To identify genes that might be nonspecifically regulated by dCas9-induction and influence subsequent PS differentiation, we first characterized the differentially expressed genes (DEGs) in either the idCas9-sgEOMES or idCas9 hiPSC line on day 0 (Figs.6A and B).We detected 976 upregulated genes and 733 downregulated genes when dCas9 was induced with the co-expression of sgRNAs targeting the EOMES loci (idCas9-sgEOMES-hiPSC, Fig. 6A).746 upregulated genes and 582 downregulated genes were identified following dCas9 induction alone (idCas9-hiPSCs, Fig. 6B).Out of the 1709 DEGs altered by dCas9 induction in the idCas9-sgEOMES-hiPSC line and the 1328 DEGs altered by dCas9 induction in the idCas9-hiPSC line, 378 genes were shared in both conditions (Fig. 6C).Among these 378 shared DEGs, nearly all (371 of 378, 98.1%) were regulated in the same direction in both hiPSC lines (Fig. 6D).Gene ontology (GO) analysis showed that the 125 genes that were codownregulated did not exhibit a significant enrichment in a biological process (Fig. S4).In contrast, the 246 co-upregulated genes were highly enriched for GO terms related to developmental processes, including anatomical structure development and morphogenesis (Fig. 6E).Specifically, the co-upregulated genes were enriched in circulatory system and tube development-related genes such as NPPB, SERPINE1, GREM1, DACT1, and MYOCD (Fig. 6D), indicating specifications towards PSderived lineages.
Next, we characterized DEGs in the two hiPSC lines on day 1 following 24 h of PS differentiation to investigate how the dCas9induced changes influence subsequent PS commitment.We identified a total of 3048 and 417 DEGs altered by prior dCas9 induction in the idCas9-sgEOMES and idCas9 hiPSC lines, respectively (Figs. 7A and B).Among these DEGs, 198 genes were shared between the two hiPSC lines (Fig. 7C), and 158 of these genes (79.8%) were regulated in the same direction (Fig. 7D).GO analysis of the co-regulated genes showed that the 93 co-downregulated genes were enriched at a low level for biological process terms related to immune system processes (Fig. S5).Similarly to the day 0 DEGs, the 65 co-upregulated genes were enriched for biological process terms related to developmental processes (Fig. 7E), including heart and tube development.Notably, we observed sustained downregulation from day 0 of seven genes and sustained upregulation of thirteen genes (Fig. 7D), including eight upregulated genes related to developmental processes (ADAMTS1, ANXA2, CD44, DACT1, EPHA2, GREM1, NQO1, and SPP1).Induction of dCas9 was also associated with elevated expression of key genes involved in embryonic morphogenesis and stem cell differentiation, such as EOMES, MIXL1, T, GSC and WNT3A during PS differentiation, confirming our previous observations regarding dCas9 induction and its influence on differentiation (see Fig. 5).
Taken together, the transcriptome analyses show that dCas9 induction in hiPSCs cultured in base medium triggers upregulation of genes related to developmental processes.Following PS differentiation, we confirmed that the sgRNA-free apo-dCas9 protein does have a nonspecific effect in hiPSCs, inducing improved PS differentiation and commitment towards PS-derived lineages.In an attempt to identify the reason for this, we computed Bayesian networks for the codownregulated and co-upregulated genes on days 0 and 1, respectively (Figs.S6 and S7).The directed acyclic graph of the 246 coupregulated and 93 co-downregulated genes from day 0 revealed three and two potential hubs, respectively, including protein interaction scaffold ABTB2, integrin ITGB5, and cyclic AMP-dependent transcription factor ATF4 among the co-upregulated genes (Fig. S6A), and Gprotein coupled receptor ADRA2A and HSP90 family chaperone HSP90AA1 among the co-downregulated genes (Fig. S6B).Among the 65 co-upregulated and 93 co-downregulated genes on day 1, we identified cell-surface antigen SLC3A2, Wnt inhibitor APCDD1, tumor necrosis factor receptor TNFRSF10B, and tyrosyl-tRNA synthetase YARS1 among the co-upregulated genes (Fig. S7A), and Rho GTPases regulator CGNL1 and stress-induced apoptosis regulator ANKRD24 among the codownregulated genes as potential hubs in the respective networks (Fig. S7B).Interestingly, none of the genes that exhibited sustained down-or upregulation from day 0 to day 1 appeared as significant hubs in the Bayesian networks.Furthermore, the identified hubs do not appear to have overlapping biological functions, nor are they directly associated with developmental processes, suggesting that dCas9 might have additional non-specific targets that are not limited to developmentally important genes.
We were interested in exploring whether similar DEG profiles might also be present in the datasets of other studies that employed CRISPR/ dCas9-based tools.For this, we selected four recently published studies: two based on CRISPRi using dCas9 alone (Mandegar et al., 2016) or a dCas9-KRAB/Dnmt3A (Nunez et al., 2021) fusion construct, and two based on CRISPRa, using a dCas9-VP64 (Kwon et al., 2020) or a dCas9-dMSK1 (Li et al., 2021) fusion construct.We compared significantly up and down-regulated genes in their datasets with the DEGs identified in response to dCas9 in this work.Analyzing the two studies based on CRISPRi revealed that only a few of the DEGs were shared with  246), downregulated in both groups (125), upregulated in response to dCas9 induction in idCas9-sgEOMES-hiPSC and downregulated in idCas9-hiPSC (1), downregulated in response to dCas9 induction in idCas9-sgEOMES-hiPSC and upregulated in idCas9-hiPSC (6) are indicated.Upregulated genes related to developmental processes (blue) and selected upregulated DEGs are specified.(E) Top significant GO terms for the 246 co-upregulated genes in response to dCas9 induction in idCas9-sgEOMES-hiPSC and idCas9-hiPSC.
the study using dCas9 alone (Mandegar et al., 2016), and none was shared with the study utilizing the dCas9-KRAB/Dnmt3A (Nunez et al., 2021) fusion construct (Extended Data 2).However, in the CRISPRa dataset based on the dCas9-VP64 fusion construct (Kwon et al., 2020), we identified 111 and 60 genes shared with our DEGs identified in response to dCas9 on days 0 and 1, respectively, of which 61 and 17 were regulated in the same direction (Extended Data 2).Similarly, the dataset based on the dCas9-dMSK1 fusion construct (Li et al., 2021) yielded 281 and 125 genes shared with our DEGs identified in response to dCas9 on days 0 and 1, respectively, of which 173 and 43 were regulated in the same direction (Extended Data 2).GO analysis of the shared genes that were regulated in the same direction in the dCas9-alone dataset (Mandegar et al., 2016) showed no GO term enrichment (Data not shown), while similar analysis of the dCas9-VP64 dataset (Kwon et al., 2020) and the dCas9-dMSK1 dataset (Li et al., 2021) showed enrichment for GO terms related to Biological Process (Fig. S8A) and Cellular Component (Fig. S8b), respectively.As in the case of the DEGs identified in response to dCas9 in this work (see Figs. 6E  and 7E), the dCas9-VP64 dataset (Kwon et al., 2020) was enriched for GO terms related to multicellular organism development and anatomical structure morphogenesis (Fig. S8A), and the dCas9-dMSK1 dataset (Li et al., 2021) was enriched for GO terms related to cell junctions (Fig. S8B).As no enriched GO terms were common among the shared DEGs from those studies, we speculated that the enrichments might instead be a result of co-regulation by a common set of transcription factors (TFs) perturbed by dCas9 expression.To identify shared co-regulatory TFs we utilized a recent published database named hTFtarget that integrates data from various TF target resources and epigenetic modification information to predict TF-target regulations (Zhang et al., 2020).Based on the query results from the list of unique DEGs from the dCas9 alone (Mandegar et al., 2016), dCas9-VP64 (Kwon et al., 2020), and dCas9-dMSK1 (Li et al., 2021) datasets that were Fig. 7. | dCas9-primed hiPSCs exhibit upregulation of genes involved developmental processes during primitive streak differentiation.(A-B) RNA-seq volcano plots showing DEGs detected in (A) idCas9-sgEOMES-hiPSC and (B) idCas9-hiPSC on day 1 following 24 h PS differentiation of hiPSCs that had been cultured in base medium with dCas9 induction (+Dox) versus no induction (-Dox).The standard cutoff point is represented by the horizontal dotted line (FDR <0.05).Upregulated (blue) and downregulated (red) genes, including the respective counts, are indicated for each comparison.(C) Venn diagram representing 2850 DEGs after dCas9 induction in idCas9-sgEOMES-hiPSC (green), 219 DEGs after dCas9 induction in idCas9-hiPSC (blue), and 198 shared genes.(D) Scatter plot representing all shared 198 DEGs in response to dCas9 induction in idCas9-sgEOMES-hiPSC and idCas9-hiPSC.Genes upregulated in both groups (65), downregulated in both groups (93), upregulated in response to dCas9 induction in idCas9-sgEOMES-hiPSC and downregulated in idCas9-hiPSC (7), downregulated in response to dCas9 induction in idCas9-sgEOMES-hiPSC and upregulated in idCas9-hiPSC (33) are indicated.Sustained downregulated (red) and upregulated (blue) genes from day 0, and DEGs related to developmental processes are specified.(E) Top significant GO terms for 65 co-upregulated genes in response to dCas9 induction in idCas9-sgEOMES-hiPSC and idCas9-hiPSC.
shared with and regulated in the same direction as the DEGs identified in response to dCas9 on days 0 and 1 in this work, we searched for TFs that were involved in regulating ≥75% of the respective list of query genes and that were common (Extended Data 3).Among the resulting 16 TF that are involved in regulating the top 25% of query genes from the respective lists of query genes and that were common, high-level GO terms related to developmental processes (SPI1, KDM2B, TP53, GLIS1, CDK9, BCOR, FOXO1) and anatomical structure morphogenesis (SPI1, KDM2B, TP53, FOXO1, FLI1, BCOR) were among the most highly enriched.This result is consistent with the enriched GO terms from the DEGs identified in response to dCas9 in this work.

Discussion
In this work, we explored the application of epigenetic editing for the writing of activating H3K4me marks within the bivalent promoter region of EOMES to enhance gene activation during differentiation.In our first approach, we used a fusion construct consisting of dCas9 from S. pyogenes and the minimal catalytic SET domain of H3K4-specific methyltransferases to write activating H3K4me marks within a putative regulatory region proximal to the transcription start site of EOMES.However, when targeted to the EOMES loci during propagation in pluripotency medium, neither MLL2 nor MLL4 promoted increased EOMES expression during PS differentiation.In a second approach, we adopted a new protocol for epigenetic editing of the EOMES loci, wherein the expression of the dCas9-MLL4 fusion construct was induced in a base medium that does not actively maintain the pluripotency network.Intriguingly, we discovered that dCas9 by itself, without any epigenetic modifier, influenced the subsequent EOMES activation during PS differentiation.Notably, this effect is specific to dCas9 induction in cells cultured in base medium that does not extrinsically maintain the pluripotency network.We initially speculated that, as demonstrated in previous reports showing that Cas9 can increase chromatin accessibility at the target loci (Barkal et al., 2016;Liu et al., 2019), dCas9 increases the accessibility at the EOMES loci for transcription factors of adjacent transcriptional regulatory elements, thereby enhancing subsequent EOMES activation during PS differentiation.However, when we characterized the regulatory network of EOMES, MIXL1, and T, all three genes were significantly upregulated.Since EOMES acts upstream of MIXL1 as a positive regulator, and together EOMES and MIXL1 both act as negative regulators of T expression, this suggests that dCas9 may have an aberrant global effect on the cells.We confirmed these observations using a stable hiPSC line encoding inducible expression of dCas9 without a sgRNA targeting the genome.Thus, the sgRNA-free apo-dCas9 protein itself appears to have an as-yet-unknown unspecific effect in hiPSCs.
Transcriptome analysis of hiPSCs cultured in base medium showed that dCas9 induction causes upregulation of genes involved in developmental processes, including anatomical structure development and morphogenesis, indicating the dCas9 might directly influence the regulation of genes controlling differentiation.Following PS differentiation, dCas9 induction resulted in upregulation of genes involved in embryonic morphogenesis and stem cell differentiation, including EOMES, MIXL1, and T, thereby validating our initial observations.We also identified a small subset of genes that exhibited sustained down-or upregulation from day 0 to day 1, with over half of the upregulated genes being related to developmental processes.Although many dCas9induction-associated DEGs in the idCas9-sgEOMES and idCas9 hiPSC lines were shared, many others were uniquely associated with a particular hiPSC line, which could be due to the presence of the sgRNAs or to random intrinsic differences between the lines generated during selection.Bayesian network analysis of the shared co-downregulated or co-upregulated genes revealed several potential hubs with nonoverlapping biological functions and no direct link to development processes.This suggests that in addition to genes involved in developmental processes, dCas9 might have other non-specific targets encoding various biological functions.Thus, further experimental validation of selected dCas9-induced DEGs is needed to map the full extent of dCas9's influence in human pluripotent stem cells.
Comparing the DEGs identified in response to dCas9 in this work to those in recently published work using CRISPRi/a-based tools (Kwon et al., 2020;Li et al., 2021;Mandegar et al., 2016;Nunez et al., 2021), we identified many DEGs from the CRISPRa-based datasets that were shared with, and regulated in the same direction as, the DEGs from this study.Moreover, they also shared some GO terms with the DEGs identified in response to dCas9 on day 0 in this study, including multicellular organism development, anatomical structure morphogenesis, and cell junctions.Analysis of potential shared co-regulatory TFs from all shared DEGs identified from the dCas9-alone (Mandegar et al., 2016), dCas9-VP64 (Kwon et al., 2020), and dCas9-dMSK1 (Li et al., 2021) datasets also revealed enrichment of TFs related to developmental processes and anatomical structure morphogenesis.The enrichment of DEGs related to shared processes is particularly surprising given that the data is derived from different cell types, including immortalized human cells (Li et al., 2021), human myogenic progenitor cells (Kwon et al., 2020), and different human iPSC lines (this work) (Mandegar et al., 2016).While potentially coincidental given the limited number of published datasets analyzed, these results could indicate that non-specific effects caused by dCas9 might be quite general, possibly even occurring in cells other than stem cells and stem-cell-derived cells.
The emergence and adaptation of CRISPR/Cas9-based tools have made it relatively straightforward to study and manipulate the genome (Bressan et al., 2017), bringing targeted therapy of genetic diseases ever closer to realization.Moreover, programmable targeting to the genome of the catalytically dead Cas9 protein allows for direct control and perturbation of endogenous gene expression (Balboa et al., 2015;Chakraborty et al., 2014;Chavez et al., 2015;Genga et al., 2019;Kearns et al., 2014;Nunez et al., 2021;Tian et al., 2019), enabling extensive phenotypic manipulation and characterization of cells.To ensure the reliability of results derived from CRISPR/Cas9-based applications, the fidelity of Cas nuclease, i.e. on-target versus off-target DNA binding, is a critical consideration.While off-target effects are well-documented when using catalytically active Cas9, due to non-specific sgRNA targeting and DNA cleavage events (Cameron et al., 2017;Hsu et al., 2013;Pattanayak et al., 2013;Tsai et al., 2015), there was no previous report indicating that the sgRNA-free apo-dCas9 protein itself has significant off-target effects.It has been reported that dCas9 has minimal off-target effects (Gilbert et al., 2014).However, as native S. pyogenes Cas9 retains a low-level affinity for DNA even in its sgRNA-free apo state (Sternberg et al., 2014), it is plausible to consider that dCas9 might be able to bind non-specifically to the genome and interfere with gene regulation when overexpressed in cells.Interestingly, additional ex vivo analysis has been shown that apo-Cas9 can slide along DNA (Shibata et al., 2017), possibly as a mechanism to scan for PAM sequences with its PAM-interacting domain (Banerjee et al., 2021).While these observations are limited to ex vivo systems, similar DNA binding and sliding within cells could result in Cas9 disrupting various endogenous regulatory protein-DNA complexes, causing unpredictable phenotypic perturbations.Furthermore, during the revision of this work, a new study was published showing that Cas9 interacts with many endogenous human RNA transcripts, often increasing their stability (Smargon et al., 2022).Increased stability of transcripts mediated by dCas9 binding would explain the higher EOMES mRNA levels we observed under Dox-induced dCas9 (-MLL4) expression.Since primary and hPSCs are innately more susceptible to perturbations as compared to cell lines, any dCas9-mediated non-specific effect might be more pronounced, especially when the cells' identity is not actively maintained by sufficiently high levels of exogenous signals to mask the putative influence of dCas9.Moreover, depending on the application, the off-target effects presented here might extend to other scenarios in which the cells' identity is not, or is only weakly, maintained by exogenous signals, as well as scenarios using dCas9 fused to effector domains.Therefore, extensive cell-type-specific characterizations under different environmental conditions are needed to guarantee the reliability of different CRISPR/Cas9-based tools.
Taken together, these results suggest that overexpression of the sgRNA-free apo-dCas9 protein in hiPSCs cultured in reduced medium conditions can alter the cell phenotype and influence subsequent directed differentiation.Moreover, the similarity of DEG profiles identified in some recently published work and the present study may suggest that the observed effect could have broad implications for when and how CRISPR/Cas9-based tools can be used reliably and safely in both research and clinical applications without interfering with the normal physiology of human cells.

Plasmid preparation
Plasmids were generated by digestion with standard restriction enzymes (New England BioLabs), followed by dephosphorylation of the plasmid backbones with Antarctic phosphatase (New England BioLabs, Cat.no.M0289), and then ligation of the backbones and inserts with T4 DNA ligase (Thermo Scientific, Cat.no.EL0011).Phosphorylation of single oligonucleotides was performed using T4 polynucleotide kinase (New England BioLabs, Cat.no.M0201) according to the manufacturer's instructions.Phosphorylation and annealing of two oligonucleotides were performed using T4 polynucleotide kinase (New England BioLabs, Cat.no.M0201) in T4 DNA ligase buffer (Thermo Scientific, Cat.no.B69).PCRs were performed with Phusion high-fidelity DNA polymerase (Thermo Scientific, Cat.no.F530) according to the manufacturer's instructions.All plasmids used in this study are described in Table S4.

Transient transfection of immortalized human cell lines
For transient transfection, HEK-293T cells were seeded into multi-well plates 1 day prior to transfection.Cells were transfected for 6 h by using polyethyleneimidine (PEI, MW 40,000, stock solution: 1 mg/ mL in ddH 2 O, Polysciences, cat.no.24765-2).In brief, transfection solutions were prepared in FBS-free medium by using 3 μg PEI/1 μg DNA and incubated for 25 min at 22 • C before being added dropwise to the cells.After 48 h, cells were lysed and RNA was processed for qPCR as described below.

Transient transfection and generation of stable polyclonal hiPSC lines
To generate piggyBac transposon-mediated stable polyclonal hiPSC lines, cells were seeded as single cells and then transiently transfected by using Lipofectamine Stem transfection reagent (Invitrogen, Cat.no.STEM00003).In brief, 80% confluent hiPSCs were washed with PBS (without calcium and magnesium, Gibco, Cat.no.14190169) and dissociated to single cells using StemPro Accutase (Gibco, Cat.no.A1110501) at 37 • C for 5-10 min, or until the cells were completely dissociated.Cells were harvested by using 4 vol of DMEM/F12 (Gibco, Cat.no.31331028) and gently triturated to further dissociate them.The cell suspension was centrifuged at 150×g for 5 min, then the supernatant was aspirated, and the cells were gently resuspended in mTeSR1 supplemented with rho kinase inhibitor (Y-27632; 100 μM, StemCell Technologies, Cat. no.72304).Cells were counted and then seeded on Geltrex-coated multi-well plates.After 20-24 h, the spent medium was aspirated.The cells were washed with PBS and fresh mTeSR1 was added.Cells were then transiently transfected by mixing the required amount of DNA with Opti-MEM reduced serum medium (Gibco, Cat. no. 31985070).Lipofectamine Stem was added to the solution and mixed thoroughly.The transfection mixture was incubated for 10-12 min at room temperature and then added dropwise to the cells.For optimal transfection results, cells were transfected within 30 min after removal of the rho kinase inhibitor.At 24 h after transfection, the medium was replaced with fresh mTeSR1.At 48-72 h after transfection, selection medium, consisting of mTeSR1 supplemented with puromycin (0.25 μg/ mL, Invivogen, Cat.no.ant-pr-1), and geneticin (100 μg/mL, Gibco, Cat.no.10131027) was added and culture was continued for 7 days.Subsequently the selection pressure was increased (0.5 μg/mL puromycin and 200 μg/mL geneticin) for an additional 7 days.Then, the cells were assessed for transgene expression and differentiation potential.

Primitive streak differentiation of hiPSCs
hiPSCs cultured for 72 h in mTeSR without or with dCas9 induction were dissociated and seeded as single cells in mTeSR1 at a density of 2.0 × 10 5 cells/cm 2 on Geltrex-coated multi-well plates.At 20-24 h after seeding, cells were washed with PBS and PS differentiation was induced for 24 h in RPMI (Gibco, Cat.no.61870010) containing sodium ascorbate (0.25 mM) and B-27 supplement (0.25 × , Minus Insulin, Gibco, Cat.no.A1895601), Normocin™ (50 μg/mL), supplemented with CHIR99021 (3 μM) and activin A (100 ng/mL).For spontaneous differentiation, hiPSCs were dissociated and seeded as single cells in mTeSR1 at a density of 2.0 × 10 5 cells/cm 2 on Geltrex-coated multi-well plates.At 20-24 h after seeding, cells were washed with PBS and cultured for 72 h in RPMI containing sodium ascorbate (0.25 mM), B-27 supplement (0.25 × , Minus Insulin), Normocin™ (50 μg/mL), and bFGF (5 ng/mL) without or with dCas9 induction.Alternatively, for dCas9 induction in base medium, cells were seeded at a density of 1.0 × 10 5 cells/cm 2 on Geltrex-coated multi-well plates.At 20-24 h after seeding, spent medium was replaced with fresh mTeSR1.At 2 days after seeding, cells were washed with PBS and cultured for 24 h in base medium as described in the results section.Then PS differentiation was induced for 24 h in RPMI containing sodium ascorbate (0.25 mM), B-27 supplement (0.25 × , Minus Insulin), Normocin™ (50 μg/mL), and the indicated concentration of CHIR99021.dCas9 was induced as described in the results section.

Immunocytochemical analysis of dCas9
For immunofluorescence staining, HEK-293T cells transiently transfected with the dCas9-fusion constructs were seeded onto Lab-Tek removable chamber slides (ThermoFisher, Cat.No. 154534 PK) 24 h after transfection and induced with Dox (50 ng/L) 8 h later.After 24 h of Dox induction, the cells were fixed using 4% paraformaldehyde (Sigma, Cat.No. 1.00496), permeabilized with 0.1% Triton X-100 (Sigma, Cat.No. T8787), and blocked with PBS buffer containing 5% donkey serum (Bio-Rad, Cat.No. C06SBZ).The slides were incubated at 4 • C overnight with mouse anti-Cas9 primary antibody (Cell Signaling Technology, Cat.No. 14697) diluted in blocking solution (1:500).After incubation, the slides were washed with blocking solution and incubated for 1 h at room temperature in the dark with donkey anti-mouse Alexa Fluor 594 (1:250; ThermoFisher, Cat.No. A-21203).Finally, the slides were washed with blocking solution and mounted with DAPI-containing solution (ThermoFisher, Cat.No. 00-4959-52).The cells were imaged with a SP8 Falcon system Leica microscope and the images were processed using ImageJ software.

Chromatin immunoprecipitation from hiPSCs
Adherent hiPSCs were washed twice with PBS, fixed in 1% formaldehyde in PBS for 10 min, neutralized in 0.2 M glycine for 5 min, collected by scraping, washed once with cold PBS supplemented with 1x cOmplete Protease Inhibitor (Roche), pelleted, flash-frozen in liquid N 2 , and stored at − 80 • C. Prior to immunoprecipitation, fixed cell pellets were thawed, washed once with cold PBS supplemented with 1x cOmplete Protease Inhibitor, lysed in nuclear lysis buffer (10 mM EDTA, 0.5% N-lauroylsarcosine (w/v), 50 mM HEPES, pH 8.0) supplemented with 1x cOmplete Protease Inhibitor for 20 min at RT, and sonicated using a pre-cooled Nextgen Bioruptor at high intensity for 25 min (cycles of 30 s ON, 30 s OFF).Sonicated chromatin was diluted thirty times in RIPA buffer (0.1% SDS, 1% Triton-X100, 0.1% sodium deoxycholate, 140 mM NaCl, 1 mM EDTA, 10 mM Tris-HCl, pH 8.0) and centrifuged (>15.000rpm for 10 min at 4 • C) to clear any remaining debris.Next, chromatin was aliquoted and the input control was collected.For each individual ChIP, the samples were complexed with ChIP-qualified antibody for K4me3 or K27me3 IP (4 μg, Abcam, a8580; 5 μg Abcam ab6002, respectively) overnight at 4 • C. Next, for each individual ChIP, 50 μL Protein A Dynabeads was mixed with 1 mL RIPA buffer and equilibrated briefly before addition of the immunocomplexed chromatin samples, and incubated for 4 h at 4 • C. The samples were then washed five times with 800 μL RIPA buffer, once with 800 μL LiCl buffer (0.25 M LiCl, 1% NP-40, 1 mM EDTA, 10 mM Tris-HCl, pH 8.0), and twice with 800 μL TE buffer (1 mM EDTA, 10 mM Tris-HCl, pH 8.0).Thereafter, samples were treated with RNase and Proteinase K. Antibodies were eluted from the beads and formaldehyde cross-linking was reversed by mild heating (65 • C).Chromatin was purified using phenol-chloroform extraction in phase-lock tubes and then the percent input relative to the adjusted input control was determined by means of qPCR.All qPCR primers used in this study are listed in Table S5.

Gene expression profiling by RT-qPCR
A Quick-RNA mini prep kit (Zymo Research, Cat.no.R1054) was used for RNA isolation, and 250 ng of purified RNA was used as a template for cDNA synthesis (High-Capacity cDNA Reverse Transcription Kit, Applied Biosystems, Cat.no.4368814).Gene expression was quantified using gene-specific primers and SsoAdvanced Universal SYBR Green Supermix (Bio-Rad, Cat.no.1725271) with an Eppendorf ep realplex Mastercycler (Eppendorf AG).Results are presented as relative fold changes normalized to the expression of one (TBP) or two (TBP and GAPDH) endogenous reference genes by using the ΔΔCt method (Schmittgen and Livak, 2008).All qPCR primers used in this study are listed in Table S6.

RNA sequencing (RNA-seq)
RNA-seq was conducted at the Genomics Facility Basel at the ETH Zurich Department of Biosystems Science and Engineering.Briefly, RNA was isolated and purified using Quick-RNA mini prep kit (Zymo Research) with in-column DNase treatment according to the manufacturer's instructions.The quality of the isolated RNA was determined with a TapeStation automated electrophoresis analyzer (Agilent).The sequencing libraries were prepared using TruSeq stranded mRNA kit (Illumina) according to the manufacturer's instructions, using 200 ng total RNA as input for each sample.The quality of the libraries was validated using a Fragment Analyzer (Agilent).Samples were single-end 101 bp sequenced on a NovaSeq 6000 (Illumina) using an SP(100) flow cell (Illumina).

RNA-seq data analysis
Raw reads were trimmed with PRINSEQ (Schmieder and Edwards, 2011) to a length of 80 bases from 100.We allowed for a maximum of 4 Ns in a read, required a mean minimum read quality of 30, and employed a threshold of 30 for the trim quality of the 3 ′ and 5' ends and a sliding window of size 10 for the trim quality.The trimmed reads were aligned with BWA (Li and Durbin, 2009) to the hg38 reference genome, which was extended to include the bacterial gene Cas9.Finally, we used featureCounts (Liao et al., 2014) to compute gene expression count values from the aligned reads.
The expression level y of a gene across experiments was modelled using the linear equation y = β 0 + β 1 D0 + β 2 D1 + β 3 sgRNA + β 4 Dox D0 + β 5 Dox D1 + β 6 Dox D0,sgRNA + β 7 Dox D1,sgRNA with covariates D0 and D1 modelling the effects of day − 1 to 0 and day 0 to − 1, and sgRNA accounting for the second cell line.For example, β 4 is the effected on gene expression in the cell line 1 treated with Dox on day − 1 for 24 h compared to cell line 1 not treated with Dox on day − 1 and β 6 is the effect in cell line 2 treated with Dox on day − 1 for 24 h compared to cell line 2 not treated with Dox on day − 1. Dox D1 and Dox D1,sgRNA denote cells that have been treated with Dox on day − 1 for 24 h, but measured on day 1.P-values were corrected for multiple testing with the Benjamini-Hochberg method (Benjamini and Hochberg, 1995).We binarized the corrected p-values by using a cutoff of 10% (Extended Data 1).Gene ontology enrichment of significant genes was computed with the Bioconductor/R package 'topGO' and the classic Fisher's exact test ('topGO' suggested default p-value cut-off 1%) (Alexa and Rahnenfuhrer, 2020).
The Bayesian networks of significantly co-regulated genes were computed with the 'orderMCMC' function of the R package 'BiDAG'.All controls and treatments were included as observations (Suter et al., 2021).The raw counts were library size-corrected and log-normalized.
Shared DEGs present in previously published work using either dCas9 alone (Mandegar et al., 2016), a dCas9-KRAB/Dnmt3A fusion construct (Nunez et al., 2021), a dCas9-VP64 fusion construct (Kwon et al., 2020), or a dCas9-dMSK1 fusion construct (Li et al., 2021) were determined by pairwise gene expression analysis.In brief, the raw gene expression counts of the four studies (GSE76980, GSE168012, GSE145575, GSE156381) were downloaded from the GEO database.The Bioconductor/R package 'edgeR' (Robinson et al., 2010) was used for the pairwise gene expression analysis and to define significantly up/down-regulated genes based on an FDR of 0.1.The significance of an overlap between one of our signatures and one of the other four studies was determined by Fisher's exact test (Extended Data File 2).Unique genes with a significant overlap were tested for gene ontology enrichment using 'topGO'.

Shared co-regulatory transcription factor identification
Potential shared co-regulatory transcription factors (TFs) were identified using the recently published hTFtarget database (http://bioin fo.life.hust.edu.cn/hTFtarget) to search for co-regulatory TFs (Zhang et al., 2020).The list of unique DEGs from the dCas9-alone (Mandegar et al., 2016), dCas9-VP64 (Kwon et al., 2020), and dCas9-dMSK1 (Li et al., 2021) datasets that were shared with, and regulated in the same direction as, the DEGs identified in response to dCas9 on days 0 and 1 in this work were used a query input for the search (Extended Data 3).Top potential co-regulatory TFs from the query results were selected based on their involvement in regulating >75% of the respective list of query genes and that were common among the datasets.TF gene ontology enrichment for the top potential co-regulatory TFs was determined using the ShinyGO gene ontology enrichment analysis tool (http://b ioinformatics.sdstate.edu/go/).

Statistical analyses
All data represent the means ± SD of two to four biologically independent samples per condition.Group comparisons were made with Student's t-test (cut-off of P < 0.05) using GraphPad Prism 8 software (GraphPad Software Inc.).

Fig. 1 .
Fig. 1. dCas9-MLL2 Pre-SET-Post fails to promote increased levels of H3K4me3 within the bivalent EOMES and does not enhance EOMES activation during subsequent primitive streak differentiation.(A) Identification of sgRNAs targeting a putative regulatory region proximal to the TSS of EOMES in HEK-293T assessed in terms of relative EOMES mRNA expression levels 48 h after co-transfection of dCas9-p300core and the indicated sgRNAs.Values were normalized to un-transfected control cells.Bars represent the mean ± sem of n = 2 biologically independent samples.(B) Scheme depicting the dCas9-MLL2 Pre-SET-Post -mediated epigenetic editing protocol.idCas9-MLL2 Pre-SET-Post -sgEOMES-hiPSCs were cultured for 3 days in mTeSR without or with Dox (1 μg/mL) induction and then seeded as single cells before primitive streak (PS) induction was triggered for 24 h with exogenous Wnt (GSK3 inhibitor CHIR99021) and TGF-β (Activin A). (C) Quantification of H3K4me3 and H3K27me3 marks within the bivalent promoter region of EOMES on day − 1 after 72 h in culture without (-Dox) or with idCas9-MLL2 Pre-SET-Post induction (+Dox).Actin beta (ACTB) is the control for active loci and Chr13.2 is an intergenomic negative control from hiPSCs cultured without idCas9-MLL2 Pre-SET-Post induction (-Dox).Bars represent the mean ± sem of n = 3 biologically independent samples.(D) Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 un-induced (-Dox) idCas9-MLL2 Pre-SET-Post -sgEOMES-hiPSCs.Bars represent the mean ± sem of n = 4 biologically independent samples.ns, not significant.

Fig. 4 .
Fig. 4. dCas9-mediated priming of EOMES is specific to reduced medium conditions.(A) Validation of results from Fig. 3C using a new idCas9-sgEOMES-hiPSC clone.Relative mRNA expression levels of EOMES following PS induction on day 1 were determined by means of qRT-PCR.Values were normalized to day 1 uninduced (-Dox) idCas9-sgEOMES-hiPSCs.Bars represent the mean ± sem of n = 4 biologically independent samples.(B) Scheme depicting the protocol used to evaluate the effect of dCas9 expression on PS differentiation.idCas9-sgEOMES-hiPSCs were cultured first for 24 h in pluripotency medium (mTeSR), then for 24 h in base medium