Overlapping functions of Krüppel-like factor family members: targeting multiple transcription factors to maintain the naïve pluripotency of mouse embryonic stem cells

Krüppel-like factors (Klfs) encode zinc-finger transcription factors and have a pivotal role in maintaining self-renewal in pluripotent stem cells (PSCs). Functions related to pluripotency were identified in three of the Klf family members. Klf4 was identified as one of the Yamanaka factors to reprogram somatic cells into PSCs in cooperation with Oct3/4 (Pou5f1 – Mouse Genome Informatics), Sox2 and Myc (Takahashi and Yamanaka, 2006), as well as the target of the leukemia inhibitory factor (LIF) signal in mouse embryonic stem cells (mESCs) (Hall et al., 2009; Niwa et al., 2009). Klf2 was reported to be a target of Oct3/4 (Hall et al., 2009) and to mediate the action of the mitogen-activated protein kinase kinase (MEK) inhibitor in mESCs to support pluripotency in defined culture condition (Yeo et al., 2014). Klf5 was reported to govern the rapid proliferation and stable self-renewal of mESCs (Ema et al., 2008). The overlapping functions of these three Klf family members in self-renewal of mESCs were examined by partial loss-of-function analyses with small interference RNA (siRNA)-mediated knockdown (Jiang et al., 2008). Their functions in reprogramming of somatic cells (Nakagawa et al., 2008) and primed PSCs (Jeon et al., 2016) into naïve PSCs have also been characterized. However, their functions in maintenance of pluripotency have not been well assessed by complete loss-of-function assays in mESCs. Here, we applied a combinatorial inducible knockout strategy to Klfs and assessed their chimera contribution abilities, confirming that Klf2, Klf4 and Klf5 share overlapping functions, and that Klf2 and Klf4 are not essential to maintain pluripotency in mESCs.

We generated a series of inducible knockout mESCs by serial modifications of the endogenous alleles of Klf2, Klf4 and Klf5 with introduction of loxP sites and a tamoxifen (Tx)-inducible form of Cre recombinase (Fig. S1). Inducible knockout of Klf2, but not Klf4, resulted in significant reduction of efficiency in stem cell colony formation (Figs 1A and 2A). However, constitutive knockout of each gene in Klf2- or Klf4-null mESCs showed stable self-renewal with a normal proliferation ratio comparable to that of wild-type mESCs in either conventional culture medium with fetal calf serum (FCS) or very low FCS with knockout serum replacement (KSR) (Figs 1B and 2B). The complete loss of the targeted gene product was confirmed by western blot analysis (Fig. S2). Quantitative PCR analysis of reverse-transcribed mRNA (RT-qPCR) revealed that they expressed comparable levels of the transcripts of functionally verified pluripotency-associated transcription factors, such as Oct3/4, Sox2, Tbx3, Nanog, Esrrb, Gbx2, Nr5a2, Tfcp2l1 and Nr0b1 (Figs 1C and 2C). They also expressed the remaining Klf family members (Klf4 and Klf5 in Klf2-null mESCs, and Klf2 and Klf5 in Klf4-null mESCs) at similar levels to wild-type mESCs, suggesting that there was no obvious dosage compensation within the three Klf family members. The expression levels of the floxed alleles were lower than those of wild-type alleles without induction of Cre activity, which might be caused by the leaky activation of Cre without induction, as well as the faint effect of the deletion for self-renewal of these mESCs. It has been known that the pluripotency-associated transcription factors show distinct expression patterns in self-renewing populations, either homogeneous (Oct3/4 and Sox2) or heterogeneous in various degrees (Nanog, Klf2, Klf4, Klf5 and Tbx3) (Niwa et al., 2009). When the expression patterns of the pluripotency-associated transcription factors were examined by immunostaining, both Klf2- and Klf4-null mESCs showed similar patterns to those in wild-type mESCs except for the deleted genes (Figs 1D and 2D). When a single Klf2- or Klf4-null mESC carrying the ubiquitous EGFP expression vector (CAG-EGFP-IZ) was injected into blastocysts followed by transfer into pseudo-pregnant mice, they gave rise to chimeric embryos with high levels of contributions (Figs 1E and 2E), indicating that both Klf2- and Klf4-null mESCs retain pluripotency.

Inducible knockout of both Klf2 and Klf4 simultaneously resulted in dramatic reduction of efficiency in stem cell colony formation (Fig. 3A). However, Klf2/Klf4-null mESCs with constitutive knockout were able to be established. They continued self-renewal with slower proliferation ratio than wild-type mESCs, with loss of Klf2 and Klf4, irrespective of the culture condition (Fig. 3B, Fig. S2). They showed comparable levels of the transcripts of pluripotency-associated transcription factors (Fig. 3C). The expression of Klf5 was slightly but significantly increased (Fig. 3C), suggesting its compensatory function. The expression patterns of pluripotency-associated transcription factors in cell populations were also similar to those of wild-type mESCs (Fig. 3D). When a single Klf2/Klf4-null mESC carrying the ubiquitous EGFP expression vector was injected into blastocysts, followed by transfer into pseudo-pregnant mice, they gave rise to chimeric embryos with high levels of contributions (Fig. 3E), indicating that Klf2/Klf4-null mESCs retain pluripotency. Because these knockout mESCs retain only Klf5 among the three Klf family members known to function in mESCs, Klf5 alone might be sufficient to support pluripotency in mESCs.

It was reported that Klf5-null mESCs showed destabilized self-renewal and slow proliferation (Ema et al., 2008). When knockout of Klf5 was induced using the same strategy, a significant proportion of the Klf5-null mESCs formed stem cell colonies (Fig. 4A). The established Klf5-null mESCs showed self-renewal with slightly slower proliferation ratio than wild-type mESCs (Fig. 4B). They showed comparable levels of the transcripts of pluripotency-associated transcription factors, as well as expression patterns in cell populations similar to those of wild-type mESCs, with complete loss of Klf5 protein (Fig. 4C,D). However, when a single Klf5-null mESC carrying the ubiquitous EGFP expression vector was injected into blastocysts, then transferred into pseudo-pregnant mice, they never gave rise to chimeric embryos, although the parental flox mESCs efficiently produced chimeric embryos. In previous reports, conventional injection strategy with multiple Klf5-null mESCs into blastocysts gave low contribution to chimeras (Ema et al., 2008), suggesting that loss of Klf5 caused reduction of pluripotency. Reduction of pluripotency might be detected more sensitively by our single mESC injection assay. Alternatively, it could be simply caused by their proliferation capacity after implantation. Inducible knockout of Klf2 and Klf5 caused significant reduction of efficiency in stem cell colony formation (Fig. 4A), but the Klf2:Klf5 and Klf4:Klf5 double-knockout mESCs were able to be established (Fig. 4B). These double-knockout mESCs expressed comparable levels of the transcripts of pluripotency-associated transcription factors, and displayed similar expression patterns in cell populations to those of wild-type mESCs (Fig. 4D,E,F), but lost the ability to contribute to chimeric embryos after single-cell injection, as in the case of Klf5-null mESCs. Klf2 expression was slightly elevated in Klf4:Klf5-null mESCs, whereas Klf4 expression was unchanged in Klf2:Klf5-null mESCs. These data suggested that the double-knockout mESCs are able to maintain the mESC-like state with sufficient expression of key pluripotency-associated transcription factors.

To confirm the overlapping function of these three Klf family members, we generated inducible triple-knockout mESCs for Klf2, Klf4 and Klf5. Induction of a triple-knockout event with Tx resulted in complete cease of self-renewal. Very few differentiated colonies were observed after induction of the triple knockout, and increased apoptosis and cell death were observed by fluorescence-activated cell sorting (FACS) analysis (Fig. S3); therefore, the removal of the three Klf family members seemed to induce cell death rather than differentiation. Induction of Klf2:Klf4:Klf5 triple knockout resulted in rapid downregulation of naïve-state-specific transcription factors (Nanog, Esrrb, Tbx3, Gbx2, Tfcp2l1, Nr0b1 and Nr5a2), followed by gradual downregulation of general pluripotency-associated transcription factors (Oct3/4 and Sox2) (Fig. 5A), suggesting indirect (or parallel) regulation of the genes in the latter category by the Klf family. These data suggested that the overlapping functions of the Klfs are essential to maintain self-renewal of mESCs via activation of naïve-state-specific genes.

To test the functional specificity of the Klf family members to maintain the ES-like state in mESCs, we performed phenotypic rescue experiments using the inducible triple-knockout mESCs. The parental triple-floxed mESCs retained pluripotency as confirmed by their chimera formation ability (Fig. 5C). When we assessed the rescue ability of various pluripotency-associated transcription factors, we found that Klf2, Klf4 and Klf5 were able to restore stem cell colony formation in the triple-knockout mESCs, while others (Oct3/4, Sox2, Nanog, Esrrb, Tbx3, Gbx2, Nr5a2, Nr0b1 and Tfcp2l1) were unable to, confirming the functional specificity of the Klf family members (Fig. 5D and data not shown). As none of the single naïve-state-specific transcription factors were able to maintain the ES-like state in the triple-knockout mESCs, it could be interpreted that the functions of the Klf family do not converge on a particular single target gene among these candidates.

The data shown above indicated that the three Klf family members Klf2, Klf4 and Klf5 share a unique function to maintain the ES-like state. However, the analysis of our previous RNA sequencing (RNA-seq) data of mESCs (Yamane et al., 2015) revealed that many other Klf family members are also expressed at comparable levels to Klf2, Klf4 and Klf5 in mESCs (Fig. 5B), suggesting a possible contribution of other Klf family members. The phenotype of the Klf2:Klf4:Klf5 triple knockout might be caused by the reduction of the overlapping functions below the threshold level as a sum, to which other Klf family members also contribute. To address this possibility, we tested the abilities of other Klf family members to maintain the ES-like state of the triple-knockout mESCs. The mouse genome contains 17 Klf family members that are categorized into three groups (or unclassified) based on the amino acid sequence of their three tandem zinc fingers located at the C-terminus (McConnell and Yang, 2010). Klf2, Klf4 and Klf5 all belong to Group 2. We chose nine members, excluding Klf2, Klf4 and Klf5, from all categories: Klf3 and Klf8 from Group 1, Klf6 and Klf7 from Group 2, Klf9, Klf10 and Klf16 from Group 3; and Klf15 and Klf17 from unclassified (Fig. 5E). In addition, we also included Sp1 and Sp5 in this assay because of the similarity of their amino acid sequence in the three tandem zinc fingers to that of the Klf family, and their high level of expression in mESCs (Fig. 5B). The coding sequences were inserted into expression vectors with or without a fusion of Ty1-tag, and their abilities to maintain the mESC-like state of the triple-knockout mESCs were assessed. Proper protein expression from these vectors was confirmed by western blot analysis of the protein lysates prepared from the pools of stable transfectants obtained in the absence of Tx with anti-Ty1 antibody (Fig. S5A). In the culture with Tx, we found that the rescue ability was quite specific to Klf2, Klf4 and Klf5 with or without Ty1 (Fig. 5F). The sole exception was Klf17, although the number of rescued clones was lower than those of canonical Klfs. These transfectants continued propagation, and proper rescue events were confirmed by western blot and RT-qPCR analyses of the pooled transfectants grown in the presence of Tx (Fig. S2). It was recently reported that KLF17 is expressed in the epiblast of human pre-implantation embryos (Blakeley et al., 2015) as well as in human naïve-like PSCs (Takashima et al., 2014). Interestingly, these human PSCs lack the expression of KLF2, suggesting functional replacement by KLF17 (Blakeley et al., 2015). The rescue ability of human KLF17 was comparable to that of mouse Klf17 and lower than that of human KLF4. These data suggested a possible role of KLF17 that might share an overlapping function with KLF4 and KLF5 in human naïve PSCs as a replacement for KLF2.

Taking advantage of the rescue system, we next addressed the molecular basis that specifies the function of the Klf family members in mESCs. The rescue assay with a series of mutant forms of Klf4 indicated that the zinc-finger domain is required for its specific function. However, the N-terminal region might also possess a specific function because complete deletion resulted in abolishment of the rescue ability, while deletion of the known functional domains, transactivation domain [TAD; amino acids (aa) 91-109 of 1-483 aa for Klf4] (Geiman et al., 2000) and serine-rich domain (SRR; aa 124-148) for phosphorylation by ERK1/2 (MAPK3/MAPK1) and binding to βTrCP2 (FBXW11) for protein degradation (Kim et al., 2012) caused partial loss (Fig. 5G). From the domains swapped between Klf4 and other Klf family members [Klf3 of Group 1, Klf7 of Group 2 and Klf9 of Group 3 (Fig. 5I)] we could see that the zinc-finger domain of Klf7, but not that of Klf3 and Klf9, was capable of replacing that of Klf4 (Fig. 5H), indicating the conserved function of the zinc-finger domain in Group 2 Klf family members. This was confirmed by successive replacement with the zinc-finger domain of Drosophila luna (Fig. 5H) that shows the highest homology to Group 2 mouse Klf family members (Fig. 5E, Fig. S4). Drosophila luna and dar1, but not Krüppel, seem to be the ancestral genes of the Klf family, considering their highest homology among the Drosophila genome to the mouse Klf family members at the three tandem zinc-finger domains and the conserved exon boundary between the first and second zinc-finger domains (Fig. S4). We confirmed that both wild-type Krüppel and the Klf4 mutant carrying the zinc-finger domain of Krüppel lack the rescue ability (Fig. 5J). Therefore, Klf is not a Krüppel-like factor in this revised homology search. These data also suggested the contribution of the N-terminal region to the specific function in PSCs because the native form of mouse Klf7 and Drosophila luna failed to restore the ability to maintain the ES-like state in the triple-knockout mESCs, indicating insufficient function of their N-terminal domain.

These data indicated that Klf2, Klf4 and Klf5 share specific functions to support pluripotency, mainly based on the specific zinc-finger sequence. To identify the responsible region within the zinc-finger domain, we performed subdomain swapping in the three tandem zinc-finger domains between Klf4 and Klf3 in Klf4/3, a chimeric molecule that consists of the N-terminal region of Klf4 and the zinc-finger domain of Klf3 (Fig. 5I). When a series of the swapping mutants was tested, we found that the mutant carrying the third zinc finger of Klf4 restored the ability to maintain the mESC-like state in the triple-knockout mESCs (Fig. 5H). These data indicated that the specific function of the zinc-finger domain depends on the unique third zinc finger shared among Group 2 Klf family members.

Triple knockout of Klf2, Klf4 and Klf5 resulted in the downregulation of multiple naïve-specific transcription factors, while the restoration of any one of them was able to maintain the mESC-like state of the triple-knockout mESCs. It is possible that the overlapping functions of the Klf family are connected to multiple targets downstream. To address this, we chose five naïve-specific transcription factors (Nanog, Esrrb, Tbx3, Gbx2 and Nr5a2) to test the ability of their combinations to support the mESC-like state of the triple-knockout mESCs. When all 32 combinations were tested in the inducible triple-knockout mESCs carrying a Rex1 (Zfp42)-mCherry reporter as a monitor of the naïve state (Kalkan et al., 2017; Toyooka et al., 2008), we found that particular combinations showed the ability to support stem cell colony formation with Rex1-mCherry expression (Fig. 6A). The combination of Nanog and Tbx3 showed significant ability, which was further enhanced by the addition of Esrrb. The combination of Nanog, Esrrb and Gbx2 also showed the synergistic potential. In most cases, the primary stem cell colonies were able to be expanded and serially passaged more than ten times while keeping mESC morphology and Rex1-mCherry expression (Fig. 6B), although they showed slower proliferation ratio and higher incidence of differentiation than wild-type mESCs. These data indicated that forced expression of particular combinations of transcription factors were able to maintain the ES-like state in the absence of the canonical Klf family members.

Then, we examined the expression levels of the endogenous pluripotency-associated genes in the rescued clones by RT-qPCR (Fig. 6C). These rescued clones maintained the expression of Oct3/4, Sox2, Tfcp2l1, Sall4 and Foxd3 at comparable levels to those of wild-type mESCs. In contrast, the expression levels of Utf1, Nr0b1, Tcl1 and Dppa3 were reduced, although their expression was maintained in the triple-knockout mESCs rescued by Klf2. Exceptionally, Dppa3 expression level was restored in the rescue clone with all five transgenes, suggesting their synergistic effect. These data suggested that the former set of genes was indirectly activated by the Klf family via at least three naïve transcription factors to maintain the ES-like state, whereas the latter set was directly activated by Klf family members, although dispensable for maintenance of the ES-like state.

It was suggested that the functions of the Klf family members in PSCs are specific to the naïve state based on their high level of expression in the naïve state as well as their potential to reprogram primed PSCs into the naïve state (Jeon et al., 2016). For strict confirmation of this hypothesis, we assessed whether the triple-knockout mESCs adapt to conversion to the primed state in vitro. The inducible triple-knockout mESCs were cultured in the condition for epiblast stem cells (EpiSCs) (Sugimoto et al., 2015), which contained activin, FGF2 and IWP2 (inhibitor of porcupine to block autocrine Wnt protein signaling) instead of LIF, with or without Tx for 5 days, followed by passages in the same condition. After three passages, we found that the inducible triple-knockout mESCs cultured with Tx showed flattened colony morphology and lost the expression of Rex1-mCherry, whereas those without Tx retained a dome-like colony morphology with strong Rex1-mCherry signal (Fig. 7A). Then, we analyzed the gene expression patterns in these cells by RT-qPCR (Fig. 7B,C). The triple-knockout mESCs cultured without Tx maintained the expression of several naïve markers (Nanog, Tbx3, Esrrb, Nr5a2, Tfcp2l1 and Nr0b1) at comparable levels to those of wild-type mESCs, with faint induction of primed markers [Fgf5, Otx2, Wnt3, Dnmt3b, Oct6 (Pou3f1 - Mouse Genome Informatics), Sox3 and Sall2] (Smith, 2017; Tesar et al., 2007). In contrast, the triple-knockout mESCs cultured with Tx showed a similar expression pattern of naïve and primed markers to that of embryo-derived EpiSCs (Tesar et al., 2007). These triple-knockout cells continued propagation up to ten passages. This was in contrast to the triple-knockout mESCs cultured in the condition for naïve PSCs with LIF, which suddenly ceased proliferation. These data indicated that the functions of Klf2, Klf4 and Klf5 are dispensable for self-renewal in primed PSCs.

The overlapping function of the three Klf family members Klf2, Klf4 and Klf5 in mESCs to maintain self-renewal was first reported by Jiang et al., who observed partial loss of function using siRNA-mediated knockdown (Jiang et al., 2008). Among them, the functions of Klf2 and Klf4 have been further emphasized in relation to pluripotency because of their higher potential for reprogramming than that of Klf5 (Jeon et al., 2016; Nakagawa et al., 2008), tight links to the known signal pathways to maintain pluripotency (Niwa et al., 2009; Yeo et al., 2014) and unique potential to reset naïve state in human PSCs (Takashima et al., 2014). Here, we employed the conventional strategy of inducible knockout for these three Klf family members. The cell biological characterization revealed that both Klf2-null and Klf4-null mESCs retained pluripotency as evaluated by their abilities to contribute to chimeric embryos after single-cell injection into the blastocyst. Moreover, Klf2:Klf4 double-knockout mESCs maintained pluripotency, indicating that both Klf2 and Klf4 are not essential to maintain pluripotency in mESCs. These results are consistent with the phenotypes of Klf2 and Klf4 knockout embryos reported previously. Klf2-null embryos die between embryonic day (E) 12.5 and E14.5, owing to defects in tunica media formation and blood vessel stabilization (Kuo et al., 1997), whereas Klf4-null mice die immediately after birth as a result of the defects in the epithelial barrier in skin and colon (Katz et al., 2002; Segre et al., 1999). Both show no phenotype in pluripotent cell population at pre- and peri-implantation stage. We also demonstrated that the specific function of the Klf family members in maintaining pluripotency is restricted to these three Klfs and to Klf17 to a weaker degree. Therefore, Klf5 is the sole Klf family member working in Klf2:Klf4-null mESCs for the maintenance of pluripotency, and its function is sufficient because the expression level of Klf17 is negligible in mESCs (Fig. 5B). The overlapping potential of Klf17 with the canonical Klfs might suggest the significance of human KLF17 in human naïve PSCs, because in these cells the expression of KLF17 is high, whereas the expression of KLF2 – which has a significant role in the maintenance of pluripotency in naïve mESCs – is low (Blakeley et al., 2015). The functional redundancy of these Klf family members is a typical example of their overlapping functions (Niwa, 2018).

The loss of Klf5 resulted in the loss of the ability to contribute to chimeric embryos in our strict assay system. Highly contributed chimeras might die after implantation because Klf5-null embryos show lethal phenotype at this timing (Ema et al., 2008). However, it was previously reported that Klf5-null mESCs were able to give rise to chimeric embryos by the conventional strategy of blastocyst injection, although the contribution was poor (Ema et al., 2008). Here, we did not perform the chimera assay with injection of multiple mESCs into blastocysts for our Klf5-null mESCs, so we could not rule out the possibility that a minor proportion of Klf5-null mESCs retain the ability to contribute to chimeric embryos. The poor ability might be caused by the slow proliferation of Klf5-null mESCs that can cause elimination from chimeric embryos after implantation when the epiblasts expand very rapidly (Snow, 1977). Alternatively, they might have reduced ability to respond to the external signal in the developmental context, which could be masked in vitro by excess amount of external signals, such as LIF, in the culture medium, although essential in the developmental context as found in the Klf5-null embryos (Lin et al., 2010).

The domain swapping experiment revealed that the third zinc-finger domain of the Klfs is crucial for their specific function in mESCs. The comparison of the amino acid sequences of this domain between Klf4 and Klf3 elucidated the difference by only four amino acids (64Q, 65K, 69A and 80M in Fig. S6). Interestingly, these four residues cluster out of the third zinc-finger domains when Klf4 binds to the target sequence with the first and second zinc fingers (Hashimoto et al., 2016), suggesting their function in the interaction with other protein(s) rather than the target DNA sequence, or with an additional target DNA sequence besides the one bound by the three zinc-finger domains of Klf4 (Schuetz et al., 2011). The specific function of the zinc-finger domain in Klf4 was shared by that of Drosophila luna, which shows the highest amino acid homology with conserved exon-intron organization in the Drosophila genome, suggesting that the pluripotency-associated function of the zinc-finger domain is a co-optional use of the evolutionarily conserved function as in the case of the evolution of Sox2 (Niwa et al., 2016). Among the genes sharing homology in the three tandem zinc-finger domains, Drosophila luna and daf, as well as Caenorhabditis elegans klf2, share the common exon boundary between the first and second zinc fingers with most of the mouse Klf family members (Fig. S4). These genes could be the descendants of the ancestral gene of the vertebrate Klf family. They are in contrast to Drosophila Krüppel, which shows lower homology without this boundary, of which the zinc finger domain does not support self-renewal of mESCs. Therefore, Klf family members are not like Krüppel in the evolutional sense.

The rescue experiment with multiple naïve-specific transcription factors revealed that the common function of Klf family members is mediated by multiple target genes. The potent targets include Nanog and Tbx3, both of which are well known for their function to support LIF-independent self-renewal of mESCs (Mitsui et al., 2003; Niwa et al., 2009). They could be the direct targets activated by the Klf family, considering their rapid downregulation in the triple-knockout mESCs. The rescue effect might be based on their parallel function in activating the core transcription factors, such as Oct3/4 and Sox2 (Niwa et al., 2009). In addition to them, Esrrb and Gbx2 showed contribution to support the ES-like state in the triple-knockout mESCs. These genes are also known for their function in supporting LIF-independent self-renewal of mESCs (Martello et al., 2012; Tai and Ying, 2013). When these transgenes were introduced into the triple-knockout mESCs, they supported the expression of a set of pluripotency-associated transcription factors, including Oct3/4, Sox2, Sall4, Foxd3 and Tfcp2l1. These genes, except Tfcp2l1, are known as core components of the pluripotency-associated transcription factors because they are expressed in both naïve and primed PSCs (Factor et al., 2014). Oct3/4 and Sox2 are absolutely essential for maintaining self-renewal of mESCs (Masui et al., 2007; Niwa et al., 2000), whereas Sall4 is required for rapid and stable self-renewal (Sakaki-Yumoto et al., 2006) and Foxd3 is known for its function to control transition from naïve to primed state (Respuela et al., 2016; Krishnakumar et al., 2016). Tfcp2l1 is specifically expressed in naïve PSCs and possesses the ability to support LIF-independent self-renewal in mESCs (Martello et al., 2013), suggesting its functional importance in maintaining self-renewal. Because these functionally verified pluripotency-associated genes are activated in the rescued triple-knockout mESCs without canonical Klfs, the involvement of the canonical Klfs in activating this set of pluripotency-associated genes is dispensable. In contrast, there is another set of genes that shows decreased levels of expression in the rescued triple-knockout mESCs. This set includes Utf1, Nr0b1, Tcl1 and Dppa3, none of which are known for their essentiality to maintain self-renewal. However, their function to modulate stable self-renewal has been reported. For example, Utf1 is reported to function in executing pluripotency by regulation of the bivalent genes (Jia et al., 2012). We previously demonstrated that Nr0b1-null mESCs show destabilized self-renewal by de-repression of Zscan4c (Fujii et al., 2015). Tcl1 is known to be involved in rapid proliferation (Miyazaki et al., 2013). The lack of proper activation of these genes might be a cause of the slow proliferation of the rescued triple-knockout mESCs with high incidence of differentiation. Therefore, the overlapping function of the canonical Klfs could be to upregulate multiple transcription factors to activate the core transcription factors, as well as to activate unique target genes to coordinate rapid and stable self-renewal (Fig. 7D).

The function of the Klf family members is strictly specific to the naïve state because the triple-knockout mESCs are successfully captured at primed state. These data confirmed that the primed state PSCs possess a transcriptional network to maintain the expression of core transcription factors, which is completely distinct from those in the naïve state PSCs, as proposed previously (Weinberger et al., 2016). Rapid and complete exit from the naïve state by removal of the Klf family members will provide opportunity to analyze the formative state recently proposed to be a transient state between the naïve and primed states (Smith, 2017). The extended analysis in the context of the transcription factor network will be required for further elucidation of the functions of these Klf family members in maintaining naïve pluripotency. The cell biological evidence shown here will provide a solid basis to systemic omics analysis in future.

EB5 mESCs (derived from male E14tg2a mESCs, Oct3/4^{IRES-BSD-pA/+}, deposited to RIKEN Cell Bank) and their derivatives were cultured on a gelatin-coated surface in Glasgow minimum essential medium (G-MEM) (Wako Pure Chemical, 078-05525) supplemented with 10% KSR (Thermo Fisher Scientific, 10826-028), 1% FCS (Hyclone), 1× sodium pyruvate (Nacalai Tesque, 06977-34), 1× nonessential amino acids (Nacalai Tesque, 06344-56), 10⁻⁴ M 2-mercaptoethanol (Nacalai Tesque, 21417-52) and 1000 U/ml LIF (home-made conditioned medium produced by COS cells transiently transfected with pCAGGS-LIF). For the proliferation assays, the mESCs were also cultured in G-MEM supplemented with 10% FCS, 1× sodium pyruvate, 1× nonessential amino acids, 10⁻⁴ M 2-mercaptoethanol and 1000 U/ml LIF. All mESC lines described in this article will be available from RIKEN Cell Bank.

To assess the transition to primed state, the mESCs were cultured in Dulbecco's modified Eagle medium/Ham's F-12 (Wako Pure Chemical, 048-29785), supplemented with 15% KSR, 1× sodium pyruvate, 1× nonessential amino acids, 10⁻⁴ M 2-mercaptoethanol, 20 ng/ml activin A (human/mouse/rat recombinant, R&D Systems, 338-AC-010), 12 ng/ml FGF2 (human recombinant, Wako Pure Chemical, 067-04031) and 2 µM IWP2 (AdoQ Bioscience, A12707-5) on mitomycin C-treated mouse embryonic fibroblast cells (Sugimoto et al., 2015).

For generation of the Klf2 knockout (KO) vector, genomic DNA fragments for 5′ and 3′ homology arms (Chr:8, 72317939-72319206 and 72321033-72324220 in GRCm38), as well as the floxed region containing exons 2 and 3 (Chr:8, 72319207-72321032 in GRCm38), were amplified from EB5 genomic DNA using the primers Klf2-5′, Klf2-3′ and Klf2-flox, respectively, shown in Table S1. The PCR fragments of 5′ and 3′ homology arms were assembled in the ClaI-NotI sites of pBR-blue II, resulting in pBR-Klf2 5′+3′. The floxed region was subcloned in the NheI site of ploxP-BNA. Then, the SpeI fragment carrying the floxed genome was introduced into the SpeI site between the 5′ and 3′ homology arms of pBR-Klf2 5′+3′, resulting in pBR-Klf2 5′+flox+3′. Finally, the XhoI fragment of the Frt-IRES-pac-pA-Frt cassette was introduced into the BamHI site flanking the 5′ end of the 3′ loxP sequence of pBR-Klf2 5′+flox+3′ using the partial fill-in method, resulting in pBR-Klf2 floxKO.

For generation of the Klf4 KO vector, genomic DNA fragments for 5′ and 3′ homology arms (Chr:4, 55532490-55531410 and 55528124-55524067 in GRCm38), as well as the floxed region containing exons 3, 4 and 5 (Chr:4, 55531419-55528125 in GRCm38), were amplified from EB5 genomic DNA using the primers Klf4-5′, Klf4-3′ and Klf4-flox, respectively, shown in Table S1. These PCR fragments of 5′ and 3′ homology arms were assembled in the ClaI-NotI sites of pBR-blue II, resulting in pBR-Klf4 5′+3′. The floxed region was subcloned in the NheI site of ploxPP2272. Then, the SpeI fragment carrying the floxed genome was introduced into the SpeI site between the 5′ and 3′ homology arms of pBR-Klf4 5′+3′, resulting in pBR-Klf4 5′+flox+3′. Finally, the XhoI fragment of the Frt-IRES-pac-pA-Frt cassette and Frt-IRES-Hytk [fusion of hph and HSV-tk (Lupton et al., 1991)]-pA-Frt was introduced into the AgeI site flanking the 3′ end of the 3′ loxP sequence of pBR-Klf4 5′+flox+3′ using adaptor DNA, resulting in pBR-Klf4 floxKO pac and pBR-Klf4 floxKO Hytk, respectively.

For generation of the Klf5 KO vector, genomic DNA fragments for 5′ and 3′ homology arms (Chr:14, 99299127-99300223 and 99303877-99307091 in GRCm38), as well as the floxed region containing exons 2 and 3 (Chr:14, 99300224-99303885 in GRCm38), were amplified from EB5 genomic DNA using the primers Klf5-5′, Klf5-3′ and Klf5-flox, respectively, shown in Table S1. These PCR fragments were assembled in the SacII-XhoI sites of pBR-blue II, resulting in pBR-Klf5 5′+3′. The floxed region was subcloned in the BamHI site of ploxP-BNA. Then, the SpeI fragment carrying the floxed genome was introduced into the AgeI site between the 5′ and 3′ homology arms of pBR-Klf5 5′+3′ using adaptor DNA, resulting in pBR-Klf5 5′+flox+3′. Finally, the XhoI fragment of Frt-SA-IRES-neo-pA:PGK-pacΔtk [fusion of pac and HSV tk (Chen and Bradley, 2000)]-pA-Frt cassette was introduced into the NheI site flanking the 3′ end of the 5′ loxP sequence of pBR-Klf5 5′+flox+3′ using the adaptor DNA, resulting in pBR-Klf5 floxKO.

For generation of the Rex1-mCherry knock-in vector, mCherry-IRES-pac-pA cassette was introduced into the SpeI-EcoRI sites of Rex1 5′+3′ homology arms as described previously (Toyooka et al., 2008).

The PiggyBac vectors pPBCAG-MerCreMer-IH, pPBCAG-EGFP-IZ, pBRPBCAG-cHA-IN and pBRPBCAG-Ty1linker-cHA-IN were constructed based on pGG131 (Guo et al., 2009). Ty1 linker consists of 3× Ty1 (EVHTNQDPLD) and glycine linker (GGSGGGGSGGGSSSS). For construction of the expression vectors of the Klf family members, the entire coding regions were amplified from the complementary DNA (cDNA) of mESCs using the primers shown in Table S1, and the PCR products were digested and inserted into the XhoI-NotI site of either pBRPBCAG-cHA-IN or pBRPBCAG-Ty1linker-cHA-IN. Drosophila luna and Krüppel were amplified from the cDNA of Drosophila embryos or Drosophila genomic DNA.

EB5 mESCs (10⁷) were electroporated with 100 µg linearized KO vector DNA at 800 V and 3 µF in a 0.4-cm cuvette using a Gene Pulser (Bio-Rad), followed by culture with 1.5 µg/ml puromycin (for Klf2 and Klf4 KO pac vector), 200 µg/ml Hygromycin B (for Klf4 KO Hytk vector) or 1.5 µg/ml of puromycin and 200 µg/ml of G418 (for Klf5 KO vector) for 8 days. The resulting stem cell colonies were picked up, expanded and genotyped by PCR using the primer pairs KO-PCR1 and KO-IRES.

For the removal of the Frt cassette, the correctly targeted clones were seeded in a well of a 48-well plate at 10⁴ cells per well, and transfected with 1 µg circular pCAG-FLPe-IP plasmid using Lipofectoamine 2000 (Thermo Fisher Scientific, 11668027), followed by culture for 3 days. Then, these transfected cells were re-plated and cultured for 8 days at clonal density (for Klf2 KO and Klf4 KO pac) or normal density with 1 µM gancyclovir (for Klf4 KO Hytk and Klf5 KO) for 8 days. The resulting stem cell colonies were picked up, expanded and genotyped by PCR.

In the case of Klf2 KO and Klf4 KO, the selection of the gene conversion event was applied to isolate homozygous KO clones (Nakatake et al., 2013). Heterozygous mESCs (10⁶) carrying the Frt-IRES-pac-pA-Frt cassette were seeded on a 90-mm dish with medium containing 6 µg/ml puromycin and cultured for 3 days. Then, the medium was changed to reduce the concentration of puromycin to 1.5 µg/ml and the cells were cultured for 8 days. The resulting stem cell colonies were picked up, expanded and genotyped by PCR.

To generate the inducible mESC lines, the clones in which the loxP sites were correctly introduced in both alleles and the Frt cassette were removed, seeded in a well of a 48-well plate at 10⁴ cells per well, and transfected with 0.25 µg circular pPB-CAG-MerCreMer-IH, 0.25 µg circular pPB-EGFP-IZ and 0.5 µg circular pCAG-PiggyBac transposase (PBase) plasmids using Lipofectoamine 2000, followed by culture for 2 days. Then, these transfected cells were re-plated and cultured with 200 µg/ml Hygromycin B and 20 µg/ml Zeocin for 8 days. The resulting stem cell colonies were picked up, expanded and assessed for the expression of Egfp by fluorescent microscopic analysis, and the function of MerCreMer was assessed by PCR genotyping of the cells cultured with 1 µg/ml 4-hydroxy tamoxifen (Tx).

For the establishment of the KO clones, the inducible KO mESCs were cultured with 1 µg/ml Tx for 3 days. Then, these cells were dissociated and seeded at clonal density (1000 cells per 90-mm dish), followed by culture for 8 days. The resulting stem cell colonies were picked up, expanded and genotyped by PCR. A PCR reaction with three primers was applied to ensure a proper deletion event.

A dissociated single mESC was injected into a C57BL6 blastocyst by microinjection. The blastocyst was then transferred into the uterus of a pseudo-pregnant female ICR mouse. Embryos were collected at 13.5 days postcoitum to evaluate the chimera contribution ability of mESCs by analyzing with fluorescence microscopy. The efficiency of chimera production with wild-type mESCs was shown in our previous reports (Ohtsuka et al., 2012; Ohtsuka and Niwa, 2015). All animal experiments conformed to Institute Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Committee for Laboratory Animal Experimentation (RIKEN Kobe Institute).

Monoclonal antibody against Klf2 was obtained by immunization of a mouse with purified fusion protein of GST and full-length Klf2, followed by the standard method. Cells were fixed with 4% paraformaldehyde in PBS for 30 min at 4°C, followed by permeabilization with 0.2% Triton X-100 in PBS for 10 min at room temperature. These cells were incubated with the following primary antibodies overnight at 4°C: mouse monoclonal anti-Oct3/4 (Santa Cruz Biotechnology, sc-5279), 1:1000; rabbit polyclonal anti-Sox2 (antisera), 1:1000; rat monoclonal anti-Nanog (e-Bioscience, MLC-51), 1:1000; mouse monoclonal anti-Klf2 (home-made), 1:1000; goat polyclonal anti-Klf4 (R&D Systems, AF3640), 1:1000; rabbit polyclonal anti-Klf5 (Abcam, ab24331); and rabbit polyclonal anti-Tbx3 (antisera, Toyooka et al., 2008). After washing, cells were incubated with appropriate Alexa Fluor 488-conjugated secondary antibodies for 1 h at room temperature with Hoechst 33258, and fluorescent images were taken on an Olympus OX-71 equipped with a CCD camera. Western blotting was performed with the antibodies for immunostaining as well as mouse monoclonal anti-Ty1 (Diagnode, MAb-054-050, 1:10,000) and rabbit polyclonal anti-Cdk2 (Santa Cruz Biotechnology, sc-163, 1:1000) for 8 µg of total cell lysates prepared in TDL buffer [150 mM NaCl, 50 mM Tris-HCl (pH 8.0), 0.5% deoxycholic acid, 1% NP-40, 0.1% SDS]. The signal was detected by SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific) using the CCD camera system LAS 4000 Mini (Fuji Film).

The inducible triple-KO mESCs (Klf2^fl/fl:Klf4^fl/fl:Klf5^fl/fl: CAG-MCM-IH:CAG-Egfp-IZ) were seeded in a well of a 48-well plate at 10⁴ cells per well, and transfected with 0.5 µg circular expression vector (derived from pBRPBCAG-cHA-IN or pBRPBCAG-Ty1linker-cHA-IN) and 0.5 µg circular pCAG-PiggyBac transposase (PBase) plasmid using Lipofectoamine 2000, followed by culture for 2 days with or without Tx. Then, the transfected cells were re-plated and cultured with 200 µg/ml G418 with or without Tx for 8 days. For evaluation of the rescue ability, the colonies were stained with Leishman staining solution, the numbers of the colonies with compact morphology were counted under a microscope, and the ratio between the presence and absence of Tx was calculated. For assessing the rescue event, the pool of colonies was cultured for preparation of RNA for quantitative PCR and protein lysate for western blotting. For the combinatorial transfection of five transcription factors (Nanog, Tbx3, Esrrb, Gbx2 and Nr5a2), these cDNAs were subcloned into pBRPBCAG-cHA-IN. Then, 0.15 µg of each of the expression vectors and 0.25 µg of circular PBase vector were transfected. For the combination of fewer than five factors, the total amount of the expression vector was adjusted with pBRPBCAG-cHA-IN.

The open reading frame sequences for the members of mouse Klf family (Klf1-17), mouse Sp1 and Sp5, and Drosophila luna were collected from Ensembl genome database. The amino acid sequences of the zinc-finger domains consist of the three tandem zinc fingers shown in Fig. S4, and were analyzed with the UPGMA program of GENETYX-MAC to obtain the evolutionary tree.

For apoptotic assay, Annexin V-APC (BD Biosciences) staining was performed according to the manufacturer's protocol using inducible triple-KO mESCs and wild-type mESCs.

First-strand DNA was synthesized from 500 ng of the total RNA prepared by a QuickGene RNA cultured cell HC kit (Kurabo) in 20 µl of the reaction mixture containing oligo-dT primers using a ReverTra Ace first strand synthesis kit (Toyobo). Real-time PCR was performed with THUNDERBIRD SYBR qPCR Mix (Toyobo) using a CFX384 Real-Time System (Bio-Rad). Primer sequences are listed in Table S1.

We thank Dr Futatsugi-Nakai (RIKEN CDB) for critical editing of the manuscript.