1) – LIF targets in mouse ES cells: experimental design
To identify LIF targets in pluripotent and committed-derived ES cells as well as key genes regulated during the transition from pluripotent to early differentiated cells, we carried out an experimental strategy based on these three facts i) A complete change of cell culture medium (including serum and LIF) is required every 48 h to induce a boost in STAT3 phosphorylation allowing cells to remain pluripotent, ii) 24 h after LIF withdrawal, STAT3 is inactivated by dephosphorylation and if incubated for longer in the absence of LIF, ES cells will differentiate. However, if LIF is added back 24 h after its withdrawal, differentiation is prevented and cells maintain their pluripotent state. Cells generated following 24 h of LIF starvation can therefore be categorised as in a state of reversible commitment, iii) 48 h following LIF withdrawal, while remaining LIF-sensitive, cells are irreversibly committed to differentiate. Indeed, even if LIF is added back, cells differentiate and a proportion of them die by apoptosis [24
]. The LIF-dependent kinetic experiments of STAT3 activation showed a maximal stimulation at 30 min., which corresponds to the maximal level of phospho-tyr705 and phospho-ser727 STAT3 protein leading to the formation of active STAT3-DNA complexes (Figure ).
Figure 1 The experimental strategy. A1) Western blot analyses showing STAT3 and P-STAT3 and A2) band shift analyses of specific LIF-dependent complexes in ES cells depleted of LIF for 24 h and reinduced with LIF for 10 to 120 min. or in pluripotent cells (+ LIF) (more ...)
Gene profiling experiments were carried out with mRNA extracted from the CGR8 mouse ES cell line cultured under growth conditions as defined in Figure . Cells in Pluripotent (Pluri), reversible (Rev.) or irreversible (Irrev.) commitment states were treated with serum or serum and LIF, 25 min. before harvesting. Protein lysates were also prepared in parallel with RNA samples. Analysis of LIF-dependent induction of STAT3, as detected by the extent of its phosphorylation, confirmed LIF responsiveness of pluripotent, reversibly, and irreversibly committed cells. In addition, we observed a progressive decrease in the levels of STAT3 protein after 24 h and 48 h of LIF starvation, which correlates with the loss of cell pluripotency. In contrast, no changes in OCT4 protein level was observed at these early times of cell differentiation. As expected, the amount of ERK2 proteins was unaltered under these various conditions and therefore, it was used as a protein loading control (Figure ). The quality of mRNA hybridized to microarrays was checked by RT-PCR experiments on a control (Hprt
) and known LIF-induced genes as Socs3, JunB
and the pluripotent Klf4
], (Figure ).
2) Gene expression profiling in the absence or presence of LIF
Five independent total RNA preparations from each condition, (Figure ), were processed and hybridized on the Mouse Genome 430 2.0 Array (Affymetrix) which includes 45101 probe sets. After RMA normalisation [49
] and outlier removal, pairwise comparisons have been performed using the paired Student t-test (summarized in Figure ). This allows the classification of genes into 7 tables (Tables 1 to 7, see Additional file 1
). In addition, multiple testing correction has been performed using the Benjamini and Hochberg procedure [50
], as described in the Methods section.
Figure 2 Summary of pair-wise comparisons. A) LIF-induced (Lifind) or repressed (Lifrep) genes, in pluripotent cells (samples 1 and 2) or after 24 h (samples 3 and 4) or 48 h (samples 5 and 6) of LIF withdrawal, B) Pluri. genes and C) Markers from reversibly and (more ...)
Tables 1 and 2 show the sets of the LIF-regulated genes detected in ES cells reversibly and irreversibly committed to differentiate following acute treatment with LIF. Genes whose expression was induced by LIF, named hereafter Lifind
genes, belong to various gene families including the Early growth response genes (Egr1,2
), members of Immediate early response (Ier2
), and of Kruppel-related gene (Klf4 and 5
) families, Ras-dex1
and the PI3K regulator Dapp1
. In addition, an EST of unknown function, 1459961_at
, encoded by a gene localized at proximity of the Stat3
locus (5' to the promoter) has also been retrieved (named Stat3Loc
, see Additional file 2
). A subgroup of genes induced after 24 h of LIF withdrawal was also induced after a 48 h period of LIF starvation (compare Tables 1 and 2). In addition, few LIF-repressed genes (Lifrep
genes), with the extent of repression less than 2- fold, were retrieved 24 h after LIF withdrawal (Ddh2
and ESTs), and no repressed genes were identified after 48 h of LIF starvation. While our stringent criteria of gene selection by the t-test analyses do not lead to the identification of LIF-modulated genes in pluripotent cells (+ LIF conditions), we detected a LIF-dependent increased expression of some genes, including Ypel2
, in pluripotent cells, by semi-quantitative RT-PCR analysis (see below, Figure ). Altogether, these results showed that the LIF response was different in pluripotent, reversibly and irreversibly committed cell populations, suggesting that LIF outcomes do not rely on the sole activation of STAT3, similarily induced in these three conditions.
Figure 4 Highlights on Cluster 8. A) Tree view representation of Lifind genes, regulated similarly by LIF in r. com. (reversible commitment) and ir. com. (irreversible commitment) conditions and validation of a selection of genes by semi-quantitative RT-PCR, (more ...)
Tables 3 and 4 depict pair-wise comparisons of genes whose expression was significantly diminished after 24 h (Table 3) or 48 h of LIF starvation (Table 4) in comparison to pluripotent cells (+LIF conditions). In this analysis, Lifind
genes, as defined in Table 1, were also retrieved (as Socs3
) indicating that some cell fate-regulated genes could also be reinducible by LIF, a property not shared by the majority of genes retrieved in these Tables. This analysis allowed to identify the core "Pluri
" genes including known pluripotent markers such as Stat3
, Cd9, Esrrb, Tbx3, Tcl1
] and members of gene families not yet described to be regulated in the ES cell model. We identified genes encoding for proteins involved in cell adhesion (as Ceacam1
), Wnt (Fzd5
), Ras (Mras
), Notch (Notch4
), Interleukin1 (Irak3
) and Tgfb (Bmp4
) signalling pathways, in cytoskeleton organization (Mapt
, Myosin IF
), and in lipid (Abca1
) and superoxide anion metabolism (Sod2
). Many transcription factors (Tcfcp2like1
) and chromatin regulators (Myst4
) were also identified.
Tables 5, 6 and 7 gather genes whose expression increases upon LIF withdrawal. Twenty genes, whose expression level was up regulated in irreversibly committed cells versus pluripotent cells, were also expressed in reversibly committed cells (genes quoted as "YES" in Table 5, like Lef1, Wt1, Dnmt3a and b, Oct6 and Otx2). This findings suggest that irreversible commitment maybe linked to the sustained expression of a set of genes expressed at earlier time following LIF withdrawal. In table 5, we noted that the majority of genes (120 out of 140) has a particular expression profile characterized by low level of expression in pluripotent cells, high level of expression in reversibly committed cells and low level of expression again in irreversibly committed cells ("low-high-low" profile, quoted as "NO", in Table 5). These genes, including Bhc80, Bicd1, Six4, Bmi1, Aire, Jmjd1c, Jmjd2c and Tle4, define a new category of genes, whose expression is transiently induced in reversibly committed cells (see also below, in Figure ). Table 7 includes genes induced at 48 h versus 24 h of LIF starvation. Some of these genes are specific of the irreversibly committed cells.
Figure 3 Identification of eight independent gene clusters. A) Each probe set is represented by a single row of colored boxes and single column represents each microarray. Cells with increasing up regulation colored with yellow of increasing intensity and increasing (more ...)
3) Gene clustering of all conditions
We performed an ANOVA (F-test) on data obtained under the different cell growth conditions. The expression profiles of highly significant differentially expressed genes [pvalue < 10-6
with a FDR (False Discovery Rate) of 1.48. 10-4
; n = 292] have been categorized using hierarchical clustering. This allowed to group genes with similar expression profiles, an indication of potential co-regulation and common functions [52
]. Tree view representation of the clusters led to the identification of eight independent groups of genes (Figure ). The upper dendrogram shows the expected clustering of the different cell growth conditions with the pluripotent state in one branch (pluri.), and the reversibly (r. com.) and irreversibly (ir. com.) committed conditions in the other branch. The complete list of genes of the cluster analysis is available (see Additional file 3
Cluster 1 (similarity score = 0.7547) includes genes whose expression, induced at 24 h of LIF starvation, is sustained at 48 h (Foxp1, Enpp3, Otx2, Pak1, Wt1, Gja1, Lef1 and Fzd7). Cluster 3 (similarity score = 0.8354) defined the differentiated markers specific of the irreversibly committed cells, like Car4, Embigin, Fgf5, Ptpn13, Irx3, Fgfr11, Rbp7, some of them corresponding to known cell lineage markers such as Fgf5.
Clusters 2 (similarity score = 0.7938) and 7 (similarity score = 0.7109), are the mirror image from each other and define new atypical gene behaviors. Modulation in the expression level of these genes was only detected in the reversibly committed cell population. Indeed, the expression profile of these genes is: "high-low-high" (Cluster 2) or "low-high-low" (Cluster 7) in "pluri – r.com – ir.com" cell growth conditions. Among genes from Cluster 2 we identified Vimentin, Hspb8 and Jund2. Genes from Cluster 7 include transcription factors like Mlr2, Aire, Tle4 and chromatin regulators as Bhc80, Smarca2. Genes from these clusters might be necessary to regulate the transition from pluripotent to irreversibly committed cells and their identification would not have been possible without the clustering of all data obtained in our experimental conditions.
Clusters 4 (similarity score = 0.8109), 5 (similarity score = 0.8170) and 6 (similarity score = 0.7163) define genes highly expressed in pluripotent cells and whose expression level, was i) abruptly decreased 24 h following LIF starvation (in Cluster 5: Pim3, Fzd5, Sod2, Esrrb, Tcfcp2l1, Mras, Gjb3 and 5, Bcl3, Stat3, Smarcd3, Yes and Cd9), ii) progressively decreased, (in Cluster 4: Stmn1 and 2, Scarb2, Timp1, Pcolce, Tcl1 and Trps1) or iii) decreased only 48 h after LIF starvation (in Cluster 6: Lgals3, Cordon bleu, Vegfc, Zfp74, Hist1h1c). Genes from cluster 5 are named "Pluri" genes thereafter.
Cluster 8 (similarity score = 0.7488) includes genes whose expression was induced by LIF in both reversibly or irreversibly committed cells. Only one gene (probe set 1446583_at) was found repressed by LIF in this analysis.
The regulated expression of a selection of genes from each cluster and Tables 1 and 2 (see Additional file 1
) was validated by semi-quantitative RT-PCR (Figures , ).
4) Cell fate-independent and cell-specific Lifind genes
To determine whether the Lifind
genes from Table 1 (see Additional file 1
) are modulated by LIF in other cellular context, we analysed their expression profile with the heatmapper tool, under our set of conditions (Figure ), and in an embryoid body (EB) kinetic analysis following LIF withdrawal, from day 1 to day 10 (Figure ). Indeed, we had previously shown that some LIF targets (like Socs3
) are expressed, several days after LIF withdrawal, in differentiated cells [46
]. In this study, we extended this analysis and showed that almost all the genes induced by LIF in reversibly and irreversibly committed cells, are re-expressed, concomitantly with the re-expression of LIF and of its receptors (gp130 and gp190) at day 10 of EB differentiation (Figures and ). In contrast, genes marked with an asterisk in Figure are not re-expressed (see Figure ). Therefore, we have characterized two types of Lifind
genes: those induced by LIF at 24, 48 h and 10 days after LIF withdrawal and which correspond to cell fate-independant LIF targets [named hereafter "Pleio-Lifind"
genes and including Socs3, Fos
] and genes whose induction by LIF is cell-restricted [named hereafter "Spe-Lifind"
genes] and which were not known, so far, as LIF targets.
Figure 5 In silico analyses of Lifind genes in various microarray experiments. Tree view representation of Heatmapper analyses of Lifind genes retrieved from Table 1 and analysed with Affymetrix data from reversibly (rev.) and irreversibly (irrev.) committed cell (more ...)
5) Regulation of "Pluri" genes in embryoid bodies kinetic following LIF withdrawal
Genes from Cluster 5 (represented in Figure ), were also analysed with the heatmapper tool in the time course of EB differentiation (Figure ). Master genes, Oct4
, whose expression is stable up to day 2 (Sox2
) or day 3 (Oct4
) and which are therefore not present in cluster 5, were manually added for this analysis. Three groups of genes were identified: i) those strictly expressed in pluripotent cells (red line, Cluster a), including pluripotent cell marker like Esrrb
; ii) those re-expressed at day 10 of the EB diffentiation kinetic (blue line, Cluster b), including the LIF-dependent Cd9
], and iii) genes whose expression was higher in differentiated than in pluripotent cells (blue line, Cluster c), including Stat3
. This analysis helps to identify relevant new genes potentially involved in the maintenance of ES cell pluripotency, since known pluripotent markers are present in each of these clusters.
Figure 6 In silico analysis of Pluri genes. Heatmapper analysis of genes from Cluster 5 (as defined in Figure and represented here in panel A) has been performed with Affymetrix data taken from the kinetic of LIF withdrawal (d1 to d10 of EBs differentiation), (more ...)
6) Analyses of Lifind and Pluri gene expression in Oct4 and Nanog knock-down experiments
The overall expression level of Oct4
was unaltered up to three days upon LIF withdrawal in the kinetic of embryoid bodies (Figure ). However, the recent finding that ES cells are heterogeneous for the expression of many markers, like NANOG but no OCT4, prompted us to determine whether the expression level of genes identified in this study could be directly modulated by these master genes. This should help to find a specificity of these core "pluripotent master genes" [53
]. We have analysed expression profiles of Lifind
genes after knock-down of Nanog
by RNAi. Availability of Affymetrix data, from Loh et al, [19
], performed with mouse ES cells infected with lentivirus expressing shNanog
silencing RNAs, allowed such an analysis. Heatmapper analyses were performed with Lifind
(Figure ) or Pluri
genes (Figure ). The expression of Lifind
genes are either repressed like Ier2
, Socs3, klf5 or Egr1
(Figure , Cluster 1) or induced like Zfp36
(Figure , Cluster 2) by Oct4
knock down. Remarkably, Ier3
is the only Lifind
gene whose expression is repressed by Nanog
silencing. In addition, this analysis emphasizes the particularity of some Spe-Lifind
genes (like Pabpc1, Plscr1, Dapp1
) and of JunB
whose expression is altered neither by Oct4
nor by Nanog
Figure 7 In silico analysis of Lifind and Pluri genes: Regulation by OCT4 or NANOG. Heatmapper analyses performed respectively with LifindA) and PluriB) genes whose level of expression was analysed in microarray experiments taken from Loh et al, . Relevant (more ...)
A similar analysis has also been conducted with the Pluri genes (Figure ), highlighting their differential regulation by Oct4 or Nanog. Indeed, we identified 5 groups of genes whose expression is down (like Sox2, Irak3 and Susd2 in Figure , Cluster 1) or up (like Cd9, Gjb3 and Gjb5 in Figure , Cluster 5) regulated following Oct4 but not Nanog silencing or down regulated by independent silencing of both genes (like Esrrb, Ly6 and Sod2 in Figure , Cluster 2 and part of Cluster 4 including Slc11a1, Nanog and Fblim1). The expression of only one gene (Ceacam1) was increased by Nanog silencing (Figure , Cluster 3). These analyses define novel gene categories, not similarly regulated by master genes, which might be part of novel sub-networks involved in the regulation of ES cell pluripotency.
7) Effects of knock-down of a selection of Lifind and Pluri genes in pluripotent cells
analyses allowed us to select genes among the Lifind
genes for further functional studies. We focused our efforts on a selection of Spe
) and also included Ier3
owing to its particular expression profile in Oct4
silencing experiments. Concerning the Pluri
genes, we performed the experiments with genes specifically expressed in pluripotent cells such as Irak3, Susd2
and whose expression was affected by only one master gene (Oct4
), except Ly6
, which behaves as Esrrb
, a recently identified pluripotent marker [25
]. We have also included Tcfcp2l1
, a transcription factor strictly expressed in pluripotent cells (Figure , Cluster a).
We have tested the function of these genes using an RNA interference strategy. E
ndoribonuclease prepared (e) siRNAs [57
] were generated, targeting the different genes. ES cells were transfected twice with each esiRNA and stained with Alkaline phosphatase (ALP) five days after the first transfection (see Additional file 4
). Knock-down of Stat3
led to morphological cell differentiation (with loss of ALP staining), and proliferative defects similar to that observed when LIF was withdrawn for 4 to 5 days, consistent with their known function in maintaining pluripotency (positive controls). Knock-down of the various genes tested did not induce significant morphological changes, as shown for Ier3
as well as for the other genes tested (see Additional file 4
and results not shown). However, alteration in gene expression level (as shown in Figure ) occurred much earlier than morphological changes, which are observed at three days upon LIF withdrawal (see Additional file 4
A) and as morphological changes, which are observed at three days upon LIF withdrawal, as shown in Figure and as reported in other ES cell lines, [47
]. This prompted us to analyse the effects of gene silencing on the expression of early differentiation markers. Expression level of Lef1, Pak1, Car4
, all induced at 24 or 48 h of LIF withdrawal (see Figure ), was analysed after 5 days of transfection in ES cells grown with LIF. The expression of Nestin
, induced at later times of LIF withdrawal, was also studied. The efficiency of the silencing of each esiRNA was determined by RT-qPCR and shown to be in the range of 50 to 80% (see Additional file 5
and not shown). The expression of almost all the differentiation markers tested was modulated by the silencing of Stat3
. However, individual silencing of the selection of Lifind
genes led to no effects or modest up regulation of some of the tested differentiation markers (see Additional file 5