|Home | About | Journals | Submit | Contact Us | Français|
Female human induced pluripotent stem cell (hiPSC) lines exhibit variability in X-inactivation status. The majority of hiPSC lines maintain one transcriptionally active X (Xa) and one inactive X (Xi) chromosome from donor cells. However, at low frequency, hiPSC lines with two Xas are produced, suggesting that epigenetic alterations of the Xi occur sporadically during reprogramming. We show here that X-inactivation status in female hiPSC lines depends on derivation conditions. hiPSC lines generated by the Kyoto method (retroviral or episomal reprogramming), which uses leukemia inhibitory factor (LIF)-expressing SNL feeders, frequently had two Xas. Early passage Xa/Xi hiPSC lines generated on non-SNL feeders were converted into Xa/Xa hiPSC lines after several passages on SNL feeders, and supplementation with recombinant LIF caused reactivation of some of X-linked genes. Thus, feeders are a significant factor affecting X-inactivation status. The efficient production of Xa/Xa hiPSC lines provides unprecedented opportunities to understand human X-reactivation and inactivation.
Female human induced pluripotent stem cell (hiPSC) and human embryonic stem cell (hESC) lines with two active X chromosomes (Xa) occur infrequently, and Xa/Xa hESC lines often become Xa/Xi (Bruck and Benvenisty, 2011; Cheung et al., 2011; Fan and Tran, 2011; Hanna et al., 2010; Hoffman et al., 2005; Lagarkova et al., 2010; Lengner et al., 2010; Marchetto et al., 2010; Pomp et al., 2011; Shen et al., 2008; Silva et al., 2008; Tchieu et al., 2010; Teichroeb et al., 2011). Some Xa/Xa hESC lines do not exhibit X-inactivation upon differentiation (Hoffman et al., 2005). The reasons for this variability are not fully understood, but derivation and culture conditions affect epigenetic features of X chromosomes (Hanna et al., 2010; Lengner et al., 2010; Pomp et al., 2011; Ware et al., 2009).
This study investigated the X-inactivation status of hiPSCs derived by the Kyoto method, which uses SNL feeder cells that produce high level of leukemia inhibitory factor (LIF) (McMahon and Bradley, 1990; Nakagawa et al., 2008; Takahashi et al., 2007). We report here that the Xi of donor fibroblasts was frequently reactivated in hiPSC lines generated on SNLs. Early passage hiPSC lines were Xa/Xi and converted into Xa/Xa lines upon continued passage on SNL but not non-SNL feeders. Lines cultured on non-SNL feeders supplemented with LIF had features of X-reactivation. These data indicate that feeder cells significantly affect X-inactivation status and that LIF contributes to reactivation. Reliably generating hiPSCs with the desired Xa/Xi or Xa/Xa pattern is useful in disease modeling and clinical applications.
We used microarrays to examine X-linked gene expression in hiPSC lines derived from differentiated H9 ESCs (H9-reporter) (Figures 1A and B, S1A–I and Table S1) or human fibroblasts (hFibs) (Figures 1D and E, S1J–Q and Table S1). Approximately 40% of X-linked genes were expressed at >1.5-fold higher levels in female hiPSC lines than in Xa/Xi or XY hESC lines (Figure 1C). Plotting the expression ratios of female hiPSCs and hESCs onto the human genome revealed that the X was the only chromosome with chromosome-wide upregulation in hiPSCs (Figure 1F). Thus, X-linked genes are specifically upregulated in female hiPSCs derived from differentiated hESCs or hFibs, suggesting X-reactivation in female hiPSCs.
We next examined expression of two X-linked genes, PGK1 and XIST, by fluorescent in situ hybridization (FISH). We found >60% of hiPSCs had two sites of nuclear transcript accumulation for PGK1, in contrast to a Xa/Xi hESC line, which had only one site in ~60% of cells (Figure 2A and B). XIST RNA coating and high expression were detected in Xa/Xi hESCs, but not in hiPSCs (Figure 2A and C). In addition, a majority of hiPSCs exhibited RNA polymerase II (polII) staining on both Xs, while hESCs exhibited staining on one X (Figure 2D), indicating that these hiPSC lines have two Xas.
By single nucleotide polymorphism (SNP) sequencing, the X-linked gene, WDR44, was expressed at a higher level in the hiPSCs derived from differentiated H9-reporter cells (H9r iPSCs) than in H9-reporter ESCs (Figure 2E, left) and detected from both alleles only in H9r iPSCs (Figure 2E, right). Bisulfite sequencing of the WDR44 promoter showed hESCs had a mixed methylation pattern characteristic of Xa/Xi cell lines (Heard and Disteche, 2006; Shen et al., 2008), while hiPSCs were hypomethylated (Figure 2F).
Finally, we used X to autosome expression ratios (X/A ratios): Xa/Xi cell lines have lower X/A ratios than Xa/Xa cell lines (Bruck and Benvenisty, 2011; Lin et al., 2007; Nguyen and Disteche, 2006). X/A ratios derived from deposited microarray data sets from hiPSC and hESC lines in which X-inactivation status is already characterized (Hanna et al., 2010; Lengner et al., 2010; Tchieu et al., 2010) were well correlated with X-inactivation status (Figure 2G, left three lanes). We found that X/A ratios from our female hiPSCs were comparable to reported Xa/Xa cells. These results confirmed X-reactivation in hiPSCs and indicate X/A ratios provide a useful method of identifying potential Xa/Xa hiPSC lines.
We analyzed a pure population of cells differentiated into endothelial cells. Xa/Xa hiPSC-derived endothelial cells exhibited low X/A ratios, comparable to primary endothelial cells and those differentiated from male or Xa/Xi hESCs (Figure 2H). XIST RNA was not detected, while only a single site of nascent PGK1 transcript accumulation was detected in > 60% of Xa/Xa-derived endothelial cells (Figure 2I and J). These results indicate that one X is silenced after differentiation of Xa/Xa hiPSCs.
We analyzed X-inactivation status in more hFib-derived hiPSC lines generated on SNLs by three- or four-factor viral reprogramming or integration-free episomal vector reprogramming (Okita et al., 2011) (Table S1). At early passage (passage (p)5), X/A ratios of all lines analyzed were consistent with reported Xa/Xi lines (Figure 3A, at p5 on SNLs). At late passage (>p15), 20 of 23 lines exhibited high X/A ratios (X/A > 0.3), comparable to those reported for Xa/Xa cells (Figure 3A, at > p15 on SNLs). Two cell lines at > p15 were further analyzed for X-inactivation by FISH, and the majority of cells showed two sites of nuclear transcript accumulation for PGK1 and no XIST RNA accumulation, indicating these lines are also Xa/Xa (Figure 3B). Further, Xa/Xa status was maintained with continued passage (Table S1). These results suggest our method frequently produces stable Xa/Xa hiPSCs.
By microarray analysis, XIST expression in all female hiPSCs on SNLs at early passage was similar to the donor Xa/Xi hFibs. At late passage, XIST was down regulated. We confirmed these findings by quantitative RT-PCR (Figure 3C). Also, SNP sequencing revealed that two X-linked genes (TSPAN6 and FRMPD4) are mono-allelically expressed at early passage but bi-allelically expressed at late passage of hiPSCs on SNLs (Figure 3D). These SNP sequencing results, together with the X/A ratios and XIST expression, indicate that the Xi is silent at early passage but is reactivated with continued propagation, concomitant with down-regulation of XIST.
Most female hiPSC lines reprogrammed with our protocol were Xa/Xa, but others laboratories reported Xa/Xi hiPSC lines derived using with the same reprogramming factors (Hanna et al., 2010; Pomp et al., 2011; Tchieu et al., 2010). A notable difference between protocols is the type of feeder cells employed. We used SNLs, which are immortalized mouse embryonic fibroblasts that express a LIF transgene (Takahashi et al., 2007), while other laboratories predominantly use mouse primary embryonic fibroblasts (MEFs) (Hanna et al., 2010; Pomp et al., 2011; Tchieu et al., 2010). Thus, we analyzed female hiPSC lines generated on non-SNLs, hFibs (Takahashi et al., 2009) or MEFs (Tchieu et al., 2010) (Table S1). None of the hiPSC lines derived on non-SNLs (0/12) had high X/A ratios (> 0.3) at > p15 (Figure 3A, at > p15 on non-SNLs). Three lines were analyzed by FISH, and the majority of cells had only one site of PGK1 accumulation with or without XIST RNA coating (Figure 3E and F). Therefore, hiPSCs generated on non-SNLs retain one Xi as reported (Pomp et al., 2011; Tchieu et al., 2010).
Since LIF is secreted by SNLs, we examined its role on X chromosome-wide gene expression by our protocol. Female hiPSC lines were initially generated on non-SNLs (hFibs or MEFs) and transferred to SNLs or non-SNLs plus recombinant (r)LIF (Figure 4A). For hiPSCs generated on hFibs, transfer occurred at p9. For hiPSCs generated on MEFs, transfer occurred at p1, when hiPSC colonies were initially picked. Female hiPSC lines transferred to SNLs had increased X/A ratios, concomitant with down regulated XIST and up-regulated X-linked genes. None of the sister lines continually cultured on non-SNLs had substantially increased X/A ratios, suggesting they remain Xa/Xi (Figures 4B–D and S3A–B). The two hiPSC lines generated on MEFs and transferred to SNLs had X/A ratios similar to those generated and cultured on SNLs and bi-allelic expression of TSPAN6 and FRMPD4 (Figure 4C and E). While the hiPSC lines generated on hFibs and transferred to SNLs had increased X/A ratios, the ratios were lower than when hiPSCs were derived exclusively on SNLs, which may be a consequence of initial reprogramming on hFibs and/or the later transfer to SNLs (Figure 4B). In support of the timing of transfer affecting X-inactivation status, one of four hiPSC lines generated on MEFs and transferred to SNLs at p4 or p7 had features of X-reactivation (Figure S3D and E). Thus, culture on SNLs can convert early passage Xa/Xi hiPSCs generated and cultured on non-SNLs into Xa/Xa hiPSCs.
Two of four hiPSC lines generated on non-SNLs and transferred to non-SNLs plus rLIF had increased X/A ratios, concomitant with down regulation of XIST and up-regulation of X-linked genes (Figures 4B, C, F and S3A and B). In a line derived on hFibs and transferred to hFibs plus rLIF at p9, the X/A ratio was comparable to that of the sister line after transfer to SNLs (Figure 4B). The line derived on MEFs and transferred to MEFs plus rLIF at p1 had a lower X/A ratio than its sister line that was transferred to SNLs at p1. While transfer to SNLs elicited bi-allelic expression of FRMPD4 and TSPAN6 (Figure 4E), transfer to MEFs plus rLIF caused bi-allelic expression of PGK1 (Figure 4G) and FRMPD4 but not TSPAN6 (Figures 4H), suggesting that rLIF promotes reactivation of a subset of X-linked genes.
Next we asked if the intermediate X/A ratio in hiPSCs cultured on non-SNLs plus rLIF reflected intermediate X-linked gene expression across the entire chromosome. We plotted expression levels of genes across the X from hiPSCs transferred to non-SNLs plus rLIF or SNLs, normalized to the expression of hiPSCs maintained on non-SNLs for the same number of passages (Figure 4I for MEF hiPSCs and S3C for hFib hiPSCs). There was variability across the X (Figure 4I). Some regions were more similar between the MEF plus rLIF hiPSCs and MEF hiPSCs and other regions were more similar between the MEF plus rLIF hiPSCs and SNL hiPSCs, suggesting that rLIF promotes full reactivation of a subset of X-linked genes. On the remainder of the X the MEF plus rLIF hiPSCs had expression intermediate to MEF hiPSCs and SNL hiPSCs, which may reflect up regulation of the Xa or X-reactivation upon addition of rLIF. FRMPD4 lies in a region of intermediate expression, suggesting that, while there is bi-allelic expression, the reactivated genes may not be expressed as highly in MEF plus rLIF hiPSCs than in SNL hiPSCs. Our results indicate reprogramming with SNL feeders promotes robust X-reactivation, and LIF may contribute to this epigenetic alteration of the Xi.
In this study, we showed that female hiPSCs derived by the Kyoto method, using retroviral or episomal vectors, frequently have two Xas. In female hiPSCs derived from hFibs, the Xi remained silent at early passage, but reactivated later, concomitant with down-regulation of XIST. Since episomal reprogramming (Okita et al., 2011) supported X-reactivation, neither viral integration nor continued expression of exogenous factors is necessary for this process. Xa/Xa hiPSCs were poised for X-inactivation upon differentiation. The high frequency X-reactivation and stable Xa/Xa status in hiPSCs derived with the Kyoto method (17/20 retroviral, 3/3 episomal) contrasts the sporadic or lack of X-reactivation of other methods (Cheung et al., 2011; Hanna et al., 2010; Kim et al., 2011; Lagarkova et al., 2010; Marchetto et al., 2010; Mekhoubad et al., 2012; Pomp et al., 2011; Tchieu et al., 2010). Thus, this method provides an unprecedented tool to understand epigenetic regulation of X chromosomes in human cells.
The Kyoto method uses SNL feeders to derive and maintain hiPSCs. When MEFs or hFibs feeders were used, we did not observe frequent production of Xa/Xa hiPSC lines. However, early passage Xa/Xi hiPSCs generated and cultured on non-SNLs could be converted into Xa/Xa after several passages on SNLs, implicating SNLs in X-reactivation.
SNLs supported production of hiPSCs with two Xas, suggesting a role for LIF in X-reactivation. Indeed, transfer into rLIF-containing medium caused up-regulation of X-linked genes and increased X/A ratios in two of four hiPSC lines initially generated on non-SNLs. While transfer to rLIF promoted up-regulation of X-linked genes, not all genes were up-regulated to the same extent as on SNLs, and some were not up-regulated at all. Also, not all genes assayed exhibited bi-allelic gene expression in hiPSCs cultured with rLIF. Thus, culture with rLIF does not have the same chromosome-wide effects that are seen with culture on SNLs. Thus, SNLs may have activities in addition to LIF that enable frequent and chromosome-wide X-reactivation. One possible activity is glycosylation of LIF: glycosylated LIF may have different roles from non-glycosylated LIF (Blanchard et al., 1998), and rLIF is not glycosylated. Identification of such activities, including LIF glycosylation, is an important future task.
Overexpression of OCT3/4, KLF2 and KLF4 in conjunction with MAPKK and GSK3b inhibitors and rLIF stochastically converts Xa/Xi hiPSCs into Xa/Xa hiPSCs (Hanna et al., 2010). In this study, we showed that expression of exogenous OCT3/4 and KLF4 during culture is not required for reactivation. While our results suggest that LIF contributes to SNL-mediated X-reactivation, the effect of SNLs on MAPK and GSK3b pathways should also be investigated.
The timing of transfer to SNLs may affect X-inactivation status, as transfer of initial hiPSC colonies directly onto SNLs resulted in X-reactivation. In contrast, transfer of four Xa/Xi hiPSCs to SNLs at p15 or later did not promote X-reactivation (data not shown). Since the reprogramming process continues during expansion of iPSC clones (Polo et al., 2010), perhaps early exposure to SNLs during this dynamic stage of reprogramming impacts X-linked gene expression at later passage. Epigenetic alterations acquired during culture without SNLs may render the Xi less responsive to the signals that trigger reactivation.
A small number of hiPSC lines fail to reactivate the Xi even on SNL feeders. While these Xa/Xi lines are indistinguishable from Xa/Xa lines in differentiation ability and global expression patterns of autosomal genes, these Xa/Xi lines might not be fully reprogrammed. In mouse, there is a relationship between X-reactivation and the “naive” pluripotent state, in which pluripotent cells efficiently contribute to chimeric embryos (Fan and Tran, 2011; Nichols and Smith, 2009; Payer et al., 2011). If the developmental potential of Xa/Xa hiPSCs is also greater, the insights obtained from the Kyoto method may be advantageous for reliable production of quality hiPSCs for future medical applications. However, when treating X-linked human monogenic diseases, such as Rett syndrome, Xa/Xi hiPSCs in which the Xi carries the mutation would be a more attractive source of material for cell-replacement therapies (Tchieu et al., 2010).
The efficient X-reactivation in our hiPSC lines is a useful tool for elucidating mechanisms of X-reactivation in human cells. Furthermore, a reliable source of Xa/Xa hiPSC lines poised for X-inactivation provides tools to study this process. X-reactivation and X-inactivation have mainly been examined in mouse systems. However, the mechanisms in human may differ from those in mouse (Maherali et al., 2007; Okamoto et al., 2011; Tchieu et al., 2010; van den Berg et al., 2009). The difficulty in obtaining human embryos and the unstable X-inactivation status in hESCs make it extremely difficult to study X-inactivation in humans (Okamoto et al., 2011; van den Berg et al., 2009). Our hiPSCs may overcome these challenges.
The role of XIST during X-inactivation is not clear in human cells. Our hiPSC lines exhibited X-inactivation upon differentiation. However, there was no detectable XIST expression in the resulting purified endothelial cells. While it is unusual for differentiated cells not to express XIST, this non-coding RNA is not necessary for maintenance of X-inactivation (Brown and Willard, 1994) and can be epigenetically silenced in cultured cells, including hESCs (Shen et al., 2008; Silva et al., 2008; Tchieu et al., 2010). Further analyses of our hiPSCs could provide new insights into regulation of XIST expression and X-inactivation in humans.
EXTENDED EXPERIMENTAL PROCEDURES is described in the Supplemental Information.
All hiPSC lines were generated by established protocols (http://www.cira.kyotou.ac.jp/e/research/protocol.html). All hESC lines were obtained from the National Stem Cell Bank (WiCell). hiPSC and hESC lines (Table S1) were maintained using standard protocols (Takahashi et al., 2007), with the exception that human insulin-like growth factor II (Chemicon; 33 ng/ml) was added into the ES medium at Gladstone. Recombinant human LIF (Millipore; 10 ng/ml) was added into the medium as indicated. All karyotyping was performed at StemCell Technology, US, or the Nihon Gene Research Laboratories, Japan. In all instances, passage 1 (p1) refers to when colonies are initially picked. SNL feeder cells are available at Health Protection Agency Culture Collection (http://www.hpacultures.org.uk/products/celllines/generalcell/detail.jsp? refId=07032801&collection=ecacc_gc).
Microarray (Whole Human Genome Microarray 4 × 44K or G3, Agilent) analyses were performed as described (Takahashi et al., 2007). All gene expression values were normalized by the 75% percentile shift method. All (Figure 1B and E) or probes that were used for calculation of X/A ratios (Figure S3A and B) were used for the heat maps. In Figures 1E, ,4I4I and S3C, IGV software (Broad Institute) was used. For Figure 1E, the expression ratios were calculated with averaged data from female hESC (ESI03, H7, H9 and H9-reporter) and hiPSC (K-3F-1, K-3F-2 and 3S-5F-4) lines. For Figures 4I and S3C, the expression ratios were calculated using selected data that were also used for Figure 4B or C. The microarray data for the deposited Xa/Xi and Xa/Xa lines used were downloaded from NCBI GEO (GEO number: GSE21222 and GSE22246). We have deposited the microarray data of hiPSC and hESC lines to GEO DataSets with the accession number GSE34527.
We thank members of Yamanaka laboratory for useful discussions, Tim Rand for plotting the results of bisulfite sequencing and X/A ratios, Kenta Nakamura for hiPSC differentiation and useful discussions, Laura Mitic for maintaining the confocal microscope, Lei Lue for the endothelial cell differentiation, Michiyo Koyanagi for microarray analyses, Mari Ohnuki for discussion of neural differentiation, Gary Howard and Anna Lisa Lucido for editorial review, Karena Essex for administrative supports, Stem Cell Core for providing stem cell culture services and Bioinformatics Core for conducting the statistic analyses. We also would like to thank Kathrin Plath and Sanjeet Patel for providing their hiPSC lines. K.T. is a scholar of the California Institute for Regenerative Medicine (CIRM). B.P. is funded by NIH R01 GM088506. This work was supported in part by grants from the Leading Project of MEXT (Japan, to SY), the Funding Program for World-Leading Innovative R&D on Science and Technology (FIRST Program) of the JSPS (Japan, to SY), Grants-in-Aid for Scientific Research of the JSPS and MEXT (Japan, to SY), and the Program for Promotion of Fundamental Studies in Health Sciences of NIBIO (Japan, to SY). These studies were also made possible by funding from the Gladstone Institutes, L.K. Whittier Foundation, from NHLBI/NIH (U01-HL100406, U01-HL098179) and the CIRM. The Gladstone Institutes received support from a National Center for Research Resources Grant RR18928-01.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.