Klf5 is involved in self-renewal of mouse embryonic stem cells

ESCs derived from the inner cell mass of the blastocyst can differentiate into primitive ectoderm, primitive endoderm and trophoectoderm cells, and in turn into all cell types present in the embryo. ESCs are maintained in the undifferentiated state during self-renewal by a complex regulatory network involving three transcription factors, namely Oct3/4 (Pou5f1) (Nichols et al., 1998), Nanog (Chambers et al., 2003; Mitsui et al., 2003) and Sox2 (Avilion et al., 2003). These factors regulate their own expression and that of many other genes (Boyer et al., 2005; Loh et al., 2006). Other transcription factors play important roles in ESC pluripotency and self-renewal (Niwa, 2007). These data illustrate the complexity of transcription regulation in ESCs, which is still not completely understood.

Recently, several results indicate that transcription factors belonging to the Kruppel-like family could have an important role in the regulation of ESCs. Ectopic expression of Klf4, together with Oct3/4, Sox2 and Myc, results in the conversion of differentiated cells into pluripotent ES-like cells (Takahashi et al., 2006; Okita et al., 2007; Werning et al., 2007; Maherali et al., 2007). Recent findings indicate that Klf2 or Klf5 can replace Klf4 in the gene combination inducing cell reprogramming (Nakagawa et al., 2008) and that triple knockdown (KD) of Klf2, Klf4 and Klf5 abolishes the undifferentiated phenotype of ESCs (Jiang et al., 2008).

In this study, we demonstrate that KD of even only Klf5 abolishes the ESC undifferentiated phenotype, whereas its constitutive expression prevents ESC differentiation.

The screening of a collection of short hairpin RNAs (shRNAs) designed to target mouse mRNAs allowed us to observe that the shRNA targeting Klf5 mRNA was able to interfere with ESC differentiation (see supplementary material Fig. S1). Klf5, also known as intestinal-enriched factor and basic transcription element binding protein 2, is a Zn-finger transcription factor belonging to the Sp/Kruppel-like family. In adults, it is expressed in the proliferating crypt cells of the intestinal epithelium and at low levels in the testis, uterus, placenta, lung and in the proliferating basal layer of the epidermis (Ohnishi et al., 2000). Klf5 knockout causes early embryonic lethality (Shindo et al., 2002), which suggests that this factor plays a key role during early development.

Klf5 mRNA and protein are present in undifferentiated ESCs and levels rapidly decrease after induction of ESC differentiation by two different approaches (Fig. 1A,B). Immunostaining demonstrated that Klf5 is expressed at various levels in almost all undifferentiated ESCs (Fig. 1C) and that it substantially colocalises with Oct3/4 and Nanog [964 out of 1100 cells (87.6%) were positive for both Oct3/4 and Klf5; 512 out of 678 cells (75.5%) were Nanog and Klf5 positive; 625 out of 720 cells (86.8%) expressed both Oct3/4 and Nanog]. After 3 days in differentiation conditions, the Klf5 signal disappeared from most cells, as observed for Oct3/4 and Nanog (supplementary material Fig. S2). The expression of Klf5 mRNA in vivo is in agreement with that observed during in vitro differentiation of ESCs. In fact, we found the Klf5 transcript in blastocysts at embryonic day (E)3.5 but not in the epiblast of E6.5 embryos (Fig. 1D). Immunostaining of pre-implantation embryos revealed nuclear Klf5 in the morula (Fig. 1E) and in many cells of the blastocyst at Theiler stage 4 (Fig. 1E,F), when the blastocoelic cavity is formed. At these stages, the cells also expressed Oct3/4 and Nanog. At a later stage, the expression of Klf5 persisted (supplementary material Fig. S2). Immunostaining of sections from E11.5 embryos revealed that Klf5 is also expressed in small groups of cells in the genital ridge, coincident with primordial germ cells expressing Oct3/4 (Fig. 1H).

The observation that Klf5 expression is restricted to undifferentiated ESCs both in vitro and in vivo and is tightly regulated when differentiation occurs suggests that it might be required to maintain ESCs in an undifferentiated state. To address this hypothesis, we explored the effects of Klf5 KD in undifferentiated ESCs. To this aim, ESCs were transfected with the previously used Klf5 shRNA or with a mixture of four short interfering RNAs (siRNAs), all targeting different regions of the Klf5 mRNA. Whereas non-silencing (CRL)-shRNA-transfected cells were indistinguishable from untransfected cells, Klf5 KD resulted in the appearance of clusters of enlarged flattened cells (supplementary material Fig. S3).

To verify that these morphological changes result from loss of the undifferentiated phenotype, we counted alkaline-phosphatase-positive colonies (APcs), as a marker of undifferentiated ESCs, in cells plated at clonal density (∼80 cells/cm²). Although grown in the presence of leukaemia inhibitory factor (LIF) and serum, Klf5-KD cells lost their undifferentiated phenotype, as witnessed by the drastic reduction of APcs (Fig. 2A). Accordingly, Oct3/4 and Nanog expression was significantly decreased, as demonstrated by immunostaining (Fig. 2B) and reverse-transcriptase (RT)-PCR (Fig. 2C).

Furthermore, Oct3/4 KD decreased levels of the cognate mRNA and protein by >50% and was accompanied by a significant reduction of Sox2 and Nanog mRNA levels. In these conditions, Klf5 mRNA was significantly decreased (Fig. 2D), which reinforces the concept that the expression of Klf5 is restricted to the ESC undifferentiated state.

To explore phenotypic changes induced by Klf5 KD, we analysed several markers of cell fate. As shown in Fig. 2E, markers of endoderm (Gata4, Hnf4 and Sox17), ectoderm (Sox1) and visceral endoderm (Afp) were undetectable in both Klf5- and CRL-shRNA-transfected cells; mRNAs for the mesoderm markers brachyury and Meox1, and for the trophoblast markers Cdx2, Eomes and PL-1, appeared only in Klf5 KD cells. Immunostaining of Klf5 KD cells showed that 12.3±0.2% of cells (n=920) expressed brachyury and 11.2±0.6% (n=950) expressed Cdx2 (supplementary material Fig. S4). The remaining cells did not express any examined differentiation markers, and some of these cells still expressed Nanog (28.5±3%) and/or Oct3/4 (37.8±3.5%).

We then examined ESC clones stably expressing Klf5 under the control of a constitutive β-actin promoter (see supplementary material Fig. S5). Mock-transfected ESC clones plated at ∼80 cells/cm² showed only a few APcs (60±3.8 APcs/100-mm plate, n=4) 6 days after LIF withdrawal, whereas, despite LIF withdrawal, ESCs stably expressing exogenous Klf5 showed 912.5±33 APcs/100-mm plate (n=4, Fig. 3A). As shown in Fig. 3B-D, after 5 days in differentiation conditions, exogenous Klf5 expression sustains Oct3/4 and Nanog expression whereas, in mock-transfected cells, Oct3/4 and Nanog expression is strongly decreased.

The transcription-factor network implicated in ESC self-renewal consists of positive- and negative-feedback loops involving Oct3/4 and Nanog. The effects of Klf5 overexpression or silencing on the expression of these genes and on the ESC undifferentiated phenotype suggest that this transcription factor could play a direct role in this regulatory network. To address this point, Klf5 was co-transfected with reporter vectors that drive the expression of luciferase under the control of Nanog or Oct3/4 promoters. Under these conditions, Klf5 overexpression was accompanied by a significant induction of the Nanog and Oct3/4 promoters (Fig. 4A). In addition, in Klf5 KD cells, transcription from both the Oct3/4 and Nanog promoters was about 50% lower, which confirms that Klf5 drives the activity of these two genes (Fig. 4A).

To evaluate whether Klf5 regulates Oct3/4 and Nanog by directly interacting with their promoters, we performed chromatin immunoprecipitation (ChIP) experiments, which demonstrated the direct interaction of Klf5 with the Nanog promoter (Fig. 4B). Also, in the case of the Oct3/4 promoter, a significant enrichment of chromatin that co-immunoprecipitated with Klf5 was observed. To confirm the presence of a Klf5 cis-element in the Nanog promoter, we performed electrophoretic mobility shift assay (EMSA) experiments. The probe on which the Oct4-Sox2 cis-element is present was not supershifted by the anti-Klf5 antibody and no changes in the band pattern were seen after challenging the probe with extracts from Klf5-overexpressing cells (supplementary material Fig. S6). On the contrary, by using a second probe covering a region downstream of the previous one, an additional shifted band appeared when it was challenged with extracts from 3×FLAG-Klf5-expressing cells (Fig. 4C). This band was erased by pre-treatment with anti-Klf5 antibody. Furthermore, one of the bands obtained by challenging the probe with wild-type ESC extracts was significantly decreased when the extract was pre-treated with anti-Klf5 antibody. The sequence of this region contains at least one element that is compatible with the known consensus of Klf factors. An oligonucleotide bearing mutations that disrupt this cis-element does not compete for either exogenous or endogenous Klf5 candidate bands (Fig. 4C).

Systematic analyses of Oct3/4 and Nanog target genes revealed numerous candidate genes, including Klf5 [see supplementary information from Boyer et al. and Loh et al. (Boyer et al., 2005; Loh et al., 2006)]. Thus, we examined the mouse Klf5 genomic region and found that it contains a Nanog candidate cis-element. Alignment of this region with all available genomic sequences revealed a 100% conservation of these cis-elements in Klf5 orthologues in all mammalian species and also in the chicken orthologue (supplementary material Fig. S7). ChIP experiments designed to determine whether Nanog interacts directly with the Klf5 promoter showed a significant enrichment for Nanog binding to Klf5 chromatin, comparable to that observed for Nanog binding to the Oct3/4 and Nanog promoters (Fig. 4D).

Experimental evidence indicates a regulatory crosstalk between Klf5 and the Sp/Kruppel-like factor Klf4. At least in adult cells, this crosstalk seems to be involved in cell proliferation (McConnell et al., 2007), but Klf5 KD or overexpression does not appear to significantly alter ESC proliferation, as demonstrated by the BrdU-incorporation assay (supplementary material Fig. S8). To address whether Klf4 is involved in Klf5 regulation, we analysed the effects of Klf4 KD or overexpression in ESCs, and found that changing Klf4 expression levels does not significantly modify the levels of Klf5, and also of Oct3/4 and Sox2, whereas Klf4 KD induces a significant decrease of Nanog mRNA (Fig. 4F). Furthermore, Klf4 and Klf2 mRNAs remained unchanged in Klf5-KD or -overexpressing cells (Fig. 4G). On the contrary, the KD of Oct3/4 significantly decreased Klf2 mRNA levels, whereas Nanog suppression led to a strong decrease of Klf4 mRNA. Taken together, these results suggest that Klf4 is not involved in the transcriptional regulation of Klf5 in ESCs. However, it is worth noting that Klf5, Klf4 and Klf2 mRNAs are all regulated as a consequence of Oct3/4 and/or Nanog suppression, thus further supporting the results indicating that all these proteins have a role in the regulation of ESCs.

Our results demonstrate that Klf5 is an essential factor of the core regulatory network that maintains ESCs in the undifferentiated state. Klf5 KD does not modify levels of Klf2 and Klf4, thus suggesting that the presence of these two factors cannot completely substitute for the reduction or absence of Klf5. The relevance of Klf5 is also supported by the observation that Klf5^–/– embryos die before E8.5 (Shindo et al., 2002), whereas Klf4^–/– mice die at birth (Segre et al., 1999) and Klf2^–/– embryos die between E12.5 and E14.5 (Kuo et al., 1997). Nevertheless, it is worth noting that Klf2 and Klf4 are more efficient than Klf5 in adult-cell reprogramming (Nakagawa et al., 2008), and their expression is modified by Oct3/4 or Nanog suppression, respectively (Fig. 4), thus indicating that these three Klf proteins are all important for ESC functions. The regulation of Klf5 is closely related to that of Oct3/4 and Nanog. In particular, Klf5 directly regulates both Nanog and Oct3/4 promoters. Furthermore, Klf5 KD impairs differentiation induced by removal of LIF and serum from the ES culture medium as observed for the KD of Nanog (Chambers et al., 2003), and both Klf5 and Nanog overexpressions are able to sustain the ESC undifferentiated state in the absence of LIF (Chambers et al., 2003; Mitsui et al., 2003).

The α1-tubulin–EGFP vector was derived from pEGFP (Clontech), containing the G418 resistance gene, by replacing the CMV promoter with the rat α1-tubulin promoter from nucleotide –1050 to +5 (Schmandt et al., 2005). Klf5 cDNA (NIH Mammalian Gene Collection, Invitrogen) was sub-cloned into p3×FLAG-CMV7.1 vector (Sigma-Aldrich). The 3×FLAG-tagged Klf5 cDNA was then subcloned into the pCBA-GFP vector (kind gift of Francesca Tuorto, CNR, Naples, Italy), by replacing GFP with 3×FLAG-Klf5. The CBA promoter, containing the CMV immediate early enhancer fused to the chicken β-actin promoter, is active in both undifferentiated and differentiated ESCs (Chung et al., 2002). A 379-bp fragment encompassing the promoter region of Nanog (Kuroda et al., 2005), and a 2.2-kb fragment encompassing the promoter region of Oct3/4 (Chew et al., 2005) were cloned into the promoterless luciferase vector pGL3-Basic (Promega).

E14Tg2a (BayGenomics) mouse ESCs were maintained as described elsewhere (Niwa et al., 2000). Plasmids and siRNA smart pools (Dharmacon) were transfected using either ArrestIn (Open Biosystems) or Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. The sequences of shRNA used are listed in supplementary material Table S1. To generate the reporter cell lines, we electroporated ESCs with α1-tubulin–EGFP plasmid. Recombinant clones were selected with 380 μg/ml G418 for 7 days (Invitrogen). One randomly selected clone (C7) was characterised for the co-expression of EGFP and β3-tubulin under differentiation conditions, and for self-renewal and in vitro pluripotency (see supplementary material Fig. S1). For luciferase assay, reporter plasmids were co-transfected with an internal control plasmid and, 24 hours later, Firefly and Renilla luciferase activities were measured using a dual-luciferase reporter system (Promega).

To induce neural differentiation, ESCs were plated onto gelatine-coated dishes at low density (1×10³-5×10³ cells/cm²) in the following differentiation medium: knockout DMEM supplemented with 10% KSR, 2 mM glutamine, 100 U/ml penicillin/streptomycin (Invitrogen), 0.1 mM β-mercaptoethanol (Sigma). Embryoid-body differentiation was obtained as described elsewhere (Maltsev et al., 1993).

shRNA-expressing plasmids (Open Biosystems) were transfected in α1-tubulin–EGFP cells plated in 96-microwell plates. Transfected cells were selected with puromycin (2 μg/ml, Sigma) for 3 days, then trypsinised and plated in differentiation medium. EGFP-positive cells appeared within 4-5 days. The presence/absence of EGFP was evaluated at different time points (4, 7 and 10 days of differentiation) by using a fluorescence-inverted microscope (Leica Microsystems).

Total RNAs were extracted using Trizol (Invitrogen). RNA (2 μg/reaction) was reverse-transcribed using M-MuLV reverse transcriptase (Biolabs). Details on semiquantitative and real-time PCR are reported in supplementary material Tables S2 and S3. Total cell lysates were obtained by using Laemmli lysis buffer and analysed by western blot using the following antibodies: Klf5 (KM1784), Oct3/4, GAPDH (Santa Cruz), anti-Flag (Sigma), Nanog (Calbiochem), secondary HRP-conjugated antibodies (Santa Cruz and Amersham Biosciences). KM1784 was raised in rabbits against human BTEB2 (orthologue of mouse Klf5). The specificity of this antibody was checked by immunostaining and western blot (supplementary material Fig. S9).

ESCs and embryos were fixed in 4% paraformaldehyde and stained with antibodies recognizing βIII-tubulin (1:400; Sigma), Oct3/4 (1:100; Santa Cruz), Nanog (1:100, R&D Systems), brachyury (1:100, Santa Cruz), Cdx2 (Biogenex), BrdU (1:10, Roche) and Klf5 (1:300, KM1784), and with appropriate secondary antibody (Molecular Probes). Images were captured with an inverted microscope (DMI4000, Leica Microsystems) or with a confocal microscope (LSM 510 META, Zeiss). Immunohistochemistry of E11.5 embryos was performed as described by Puelles et al. (Puelles et al., 2006).

Cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature and formaldehyde was then inactivated by the addition of 125 mM glycine. The chromatin was then sonicated to an average DNA-fragment length of 200-1000 bp. Soluble chromatin extracts were immunoprecipitated using anti-FLAG (Sigma, cat. F3165) or -Nanog (R&D Systems, cat. AF2729) antibodies. Supernatant obtained without antibody was used as input control. The amount of precipitated DNA was calculated by real-time PCR relative to the total input chromatin, and expressed as percent of total chromatin according to the following formula: 2^ΔCt × 10, where Ct represents the cycle threshold and ΔCt = Ct(input) – Ct(immunoprecipitation) (Frank et al., 2002). The Student's t-test was used to measure statistical significance. Oligonucleotide pairs are reported in supplementary material Table S4. EMSAs were performed as described by Bevilacqua et al. (Bevilacqua et al., 2005).

This work is supported by grants to T.R. from Ministry of Research, Regione Campania, Italy. We are grateful to Antonio Simeone, Dario Acampora, Caterina Missero and Riccardo Cortese for precious suggestions, and to Jean Ann Gilder for reading the manuscript.