First published online in STEM CELLSExpress January 22, 2009.
Author contributions: S.Y.: collection and assembly of data, data analysis and interpretation, manuscript writing; S.F.: collection and assembly of data; K.N.: collection and assembly of data; N.T.: collection and assembly of data; Y.T.: collection and assembly of data; Y.-I.F.: collection and assembly of data; H.A.: collection and assembly of data; K.U.: data analysis and interpretation; H.K.: data analysis and interpretation; H.N.: Provision of study material, data analysis and interpretation; R.N.: conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing.
Sall4 is a mouse homolog of a causative gene of the autosomal dominant disorder Okihiro syndrome. We previously showed that the absence of Sall4 leads to lethality during peri-implantation and that Sall4-null embryonic stem (ES) cells proliferate poorly with intact pluripotency when cultured on feeder cells. Here, we report that, in the absence of feeder cells, Sall4-null ES cells express the trophectoderm marker Cdx2, but are maintained for a long period in an undifferentiated state with minimally affected Oct3/4 expression. Feeder-free Sall4-null ES cells contribute solely to the inner cell mass and epiblast in vivo, indicating that these cells still retain pluripotency and do not fully commit to the trophectoderm. These phenotypes could arise from derepression of the Cdx2 promoter, which is normally suppressed by Sall4 and the Mi2/NuRD HDAC complex. However, proliferation was impaired and G1 phase prolonged in the absence of Sall4, suggesting another role for Sall4 in cell cycle control. Although Sall1, also a Sall family gene, is known to genetically interact with Sall4 in vivo, Sall1-null ES cells have no apparent defects and no exacerbation is observed in ES cells lacking both Sall1 and Sall4, compared with Sall4-null cells. This suggests a unique role for Sall4 in ES cells. Thus, though Sall4 does not contribute to the central machinery of the pluripotency, it stabilizes ES cells by repressing aberrant trophectoderm gene expression. STEM CELLS 2009;27:796–805
The first lineage commitment of the mammalian embryo leads to the development of two distinct lineages, designated as the inner cell mass and the trophectoderm. Cells of the inner cell mass are pluripotent and give rise to the epiblast and eventually to all the fetal tissues, whereas trophectoderm cells differentiate into the trophoblast cell layers of the placenta. Embryonic stem (ES) cells established from the inner cell mass differentiate into all three embryonic germ layers, but rarely into trophectoderm lineages, although trophectoderm differentiation can be initiated by genetic manipulation of key genes involved in trophectoderm development, such as overexpression of Cdx2 . Maintenance of ES cell pluripotency requires Oct3/4, which is abundantly expressed in ES cells . A one-half-fold reduction in Oct3/4 expression results in Cdx2 upregulation and promotes differentiation of ES cells into the trophectoderm, that is, the placental lineage . Therefore, strict regulation of Oct3/4 dosage, which leads to repressed Cdx2 expression, is essential for preventing ES cells from differentiating toward the trophectoderm.
The spalt (sal) gene family is conserved from Drosophila to humans and its members encode proteins characterized by multiple double zinc finger motifs of the C2H2 type. Humans and mice each have four known Sal-like genes (SALL1-4 in humans and Sall1-4 in mice). Mutations in SALL4 cause an autosomal dominant disorder known as Okihiro syndrome, which is characterized by limb deformities, eye movement deficits and, less commonly, anorectal, ear, heart, and kidney anomalies [4, 5]. Some of these phenotypes are likely to be caused by SALL4 haploinsufficiency because a reasonable proportion of Sall4 heterozygous mice show anal and heart anomalies, as we previously described . Sall4/Sall1 heterozygotic mice exhibit increased incidences of anorectal and heart anomalies, suggesting that Sall4 and Sall1 have a genetic interaction. Indeed, endogenous binding of Sall4 and Sall1 has been demonstrated, indicating that these proteins could function in a heterodimeric form, at least in some settings of organ formation . This could explain the phenotypic differences between SALL1 mutations in humans and Sall1-null mice. Specifically, Sall1-null mice show only kidney defects , whereas SALL1 mutations in humans are associated with Townes-Brocks syndrome, an autosomal dominant disease characterized by dysplastic ears, preaxial polydactyly, imperforate anus and, less commonly, kidney and heart anomalies . Several lines of evidence suggest that truncated SALL1 proteins resulting from human mutations function in a dominant-negative manner and that some symptoms of Townes-Brocks syndrome are likely to be caused by inhibition of SALL4 functions . Indeed, mice expressing a truncated Sall1 protein show similar phenotypes to those observed in human patients . The truncated Sall1 protein can also bind to Sall2 and Sall3 in vitro , but no potent genetic interactions have yet been detected in vivo. Sall2-deficient mice show no apparent phenotypes, and mice lacking both Sall1 and Sall2 show kidney phenotypes comparable with those of Sall1 knockout mice . Sall3-null mice die on the first postnatal day and exhibit cranial nerve deficiencies and oral structure abnormalities . However, Sall1/Sall3 double heterozygous mice have no apparent organ defects, in contrast to Sall1/Sall4 heterozygotes with severe phenotypic exacerbation .
Sall4-null mice, unexpectedly, die shortly after implantation into the uterus . Although Sall4-null blastocysts showed no commitment defects of the inner cell mass or trophectoderm in vivo, the Sall4-null inner cell mass showed retarded proliferation in vitro. As ES cells are derived from the inner cell mass, it is a reasonable proposition that Sall4 is essential for the maintenance of ES cells . Sall4-null ES cells cultured on feeder cells grew poorly but retained their pluripotency, as assessed by marker expression including the presence of Oct3/4 and absence of Cdx2, and by in vivo injection into blastocysts, in which Sall4-null ES cells contributed highly to the chimeras. However, a subsequent report indicated that shRNA-mediated Sall4 inhibition in ES cells led to a severe reduction in Oct3/4 and a secondary increase in Cdx2, which resulted in complete differentiation into the trophectoderm . These effects were only observed when feeder-free ES cells were used. The Sall4-reduced ES cells still contributed to the trophectoderm when aggregated with eight-cell embryos, and Sall4 siRNA injection into the one-cell stage embryos also increased Cdx2 in the inner cell mass. These results suggest that Sall4 may be involved in lineage commitment between the inner cell mass and trophectoderm and that Sall4 is essential for the pluripotency of ES cells. As these findings are in sharp contrast with our genetic data , we decided to reanalyze Sall4-null ES cells under feeder-free conditions to avoid potential off-target effects of RNAi experiments. In addition, based on systematic immunoprecipitation followed by mass-spectrometry, several reports have suggested the involvement of Sall4, as well as Sall1, in the transcriptional network in ES cells; although evidence of the functional importance of Sall1 was not provided [13, 14]. As Sall4 and Sall1 could form a heterodimer as mentioned earlier, we also analyzed the roles of Sall4 and Sall1 in ES cells by generating cells lacking both Sall4 and Sall1. Here, we show that feeder-free Sall4-null ES cells express Cdx2, but are maintained for a long period in an undifferentiated state with minimally affected Oct3/4 expression. These cells contributed solely to the inner cell mass and epiblast in vivo, indicating that they still retain their pluripotency and do not commit fully to the trophectoderm. Furthermore, we present distinct roles for Sall4 and Sall1 in ES cells.
MATERIALS AND METHODS
Introduction of a Blasticidin-Resistance Gene into the Oct3/4 Locus
Sall4 heterozygous (flox/−) ES cells  were electroporated with a targeting vector containing Oct3/4 genomic regions and an IRES-blasticidin-resistance gene (BSD) , and cultured on blasticidin-resistant fibroblasts (kindly provided by Dr. K. Araki, Kumamoto University). Southern blotting analysis revealed that 18 of 22 clones obtained were homologous recombinants (data not shown), and all 18 formed compact colonies. One of the clones (no. 14-3) was infected with an adenovirus-expressing Cre under the CAG promoter , as previously described . The primers used for reverse transcription-polymerase chain reaction (RT-PCR) are shown in supporting information Table 1, and RT(−) controls gave no signals. All the recombinant DNA experiments were carried out in accordance with the guidelines of the Japanese government.
Immunocytochemistry and Confocal Microscopy
The following antibodies were used for staining blastocysts and ES cells: monoclonal anti-Sall1  (PPMX Perseus Proteomics, Tokyo, Japan, http://www.ppmx.com); monoclonal anti-Sall4  (PPMX Perseus Proteomics); monoclonal anti-Oct3/4 (Santa Cruz Biotechnology Inc., Santa Cruz, CA, http://www.scbt.com); rabbit antisera against Oct3/4 ; and monoclonal anti-Cdx2 (BioGenex, San Ramon, CA, http://www.biogenex.com).
Manipulation of Embryos
Injection of ES cells with eight-cell and blastocyst stage embryos and RNAi injection into one-cell embryos were performed as previously described [6, 12]. Seven to 10 ES cells were injected into each eight-cell embryo. Chimerism was calculated as the ratio of green fluorescent protein (GFP)-positive areas measured by ImageJ software (NIH) against those of the inner cell mass topologically identified with DAPI staining. In the case of epiblast injection, the average of GFP signal intensity in the epiblast was divided by the maximum signal. All animal experiments were performed in accordance with institutional guidelines and review committees.
A GeneChip Mouse Genome 430 2.0 Array (Affymetrix, Santa Clara, CA, http://www.affymetrix.com) was used to compare the gene expression profiles of Sall4 flox/− and Δ/− ES cells according to the manufacturer's protocols. Data were analyzed using the Affymetrix GeneChip software.
Targeted Disruption of Both Alleles of Sall1
The Sall1-Hyg targeting vector for the first round of targeting was constructed by incorporating a 5′ SmaI-EcoRI 5.4-kb fragment and a 3′ HindIII-ClaI 2.8-kb fragment into a vector containing a hygromycin-resistance gene (Hygr) and a thymidine kinase gene in tandem (supporting information Fig. 1A). The second round of targeting was carried out as a vector containing a neomycin-resistance gene (Neor; supporting information Fig. 1B) as previously described .
Generation of Floxed Sall1 Mice
A Sall1-flox vector was constructed by incorporating 5′ HindIII-SphI 2.1-kb, SphI-SalI 0.8-kb, and 3′ SalI-KpnI 5.4-kb fragments into a vector containing the Neor gene (pGK-Neo), three loxP sequences and the diphtheria toxin A subunit (pMC1DTA), as shown in supporting information Figure 2A. The loxP sequences were placed so that exon one containing the starting codon was removed on Cre-mediated excision. E14.1 ES cells cultured on embryonic fibroblasts were electroporated with the targeting vector and three of 192 colonies were found to be homologous recombinants containing all three loxP sites. The latter analysis was performed by Southern blotting using EcoRI digestion with probe 1, HindIII digestion with probe three and HindIII digestion with probe two (supporting information Fig. 2B and data not shown). It should be noted that the HindIII site was created just downstream of the 3′-most loxP site. Next, we introduced pIC Cre into one of the positive clones (no. 166) and examined the neomycin-sensitivity of the resultant colonies. Two clones (no. 166-25 and no. 166-29) contained exon one flanked by two loxP sites, whereas 13 contained only one loxP site without exon one and Neor, as evaluated by Southern blotting analysis (data not shown). Both of the former clones generated germline chimeras, and homozygous floxed mice were healthy and fertile. When flox/+ mice were bred with mice ubiquitously expressing Cre , exon one was excised (Δ/+) and homozygotes (Δ/Δ) showed kidney defects that were comparable with those observed in conventional Sall1-null mice . Genotyping was carried out by either Southern blotting using HindIII digestion and probe two (supporting information Fig. 2C) or by PCR using the following primers: 5′-CCTCTGCCCGAGAGATCG-3′, 5′-GGCGCGTCTG ATTTTATTTC-3′, and 5′-AGGAACACTCACGAAATGGG-3′ (wild-type: 220 bp; floxed: 280 bp; Δ: 350 bp).
Derivation of Sall1/4-Null ES Cells from Blastocysts
Blastocysts obtained from intercrosses of Sall1/4 double floxed homozygous mice were cultured in DMEM containing 14% Knockout Serum Replacement (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), 1% serum, and 103 U/ml LIF for 10–14 days in a gelatin-coated 48-well plate. The cells were then suspended using 0.125% trypsin-EDTA and expanded on feeders. The resulting ES cells were infected with an adenovirus-expressing Cre. Genotyping of floxed Sall4 was carried out by PCR using the following primers: 5′-CCTCCCGGAATTGCTTATCT-3′, 5′-AGGACAAGGATCGTTCTACAGC-3′, 5′-CTGTCCATC TGCACGAGACT-3′, and 5′-GCTTCTGCCTCTGGTATTGC-3′ (wild-type: 200 bp; floxed: 450 bp; Δ: 220 bp).
Identification of Sall4-Binding Proteins
ES cells stably expressing FLAG-tagged Sall4 were lysed in lysis buffer (50 mM HEPES-KOH pH 7.8, 420 mM NaCl, 0.1 mM EDTA, 5 mM MgCl2, 1% NP40, 20% glycerol, 0.5 mM DTT, protease inhibitor cocktail) for 30 minutes. After adjustment of the final NaCl concentration of the supernatants to 150 mM, the extracts were precleared by agarose beads (Sigma) and mixed with beads conjugated with an anti-FLAG antibody (Sigma) at 4°C overnight. The beads were washed four times with lysis buffer and boiled for 5 minutes. The eluents were analyzed by Silver Quest (Invitrogen), and the candidate bands were subjected to mass spectrometry.
Sall4-heterozygous and Sall4-null ES cells, as well as Sall4-null ES cells with exogenously introduced FLAG-Sall4 showing normal phenotypes, were used for native chromatin immunoprecipitation (ChIP) analyses. Isolated nuclei were resuspended in a buffer containing 200 mM NaCl and digested into mono-nucleosomes or di-nucleosomes by micrococcal nuclease (Takara, Otsu, Japan, http://www.takara.co.jp). The extracts were precleared with agarose bead-conjugated normal IgG (Santa Cruz Biotechnology) for 2 hours, and then incubated overnight with an anti-H3Ac antibody (GE Healthcare) conjugated with protein G-Sepharose beads (GE Healthcare) or an anti-FLAG-M2 gel (Sigma). After washing, the eluted ChIP products and input DNA were treated with 25 μg/ml RNaseA at 37°C for 30 minutes, incubated with 60 μg/ml proteinase K at 55°C for 1 hour, and extracted with phenol/chloroform. The samples were analyzed by quantitative PCR using the PCR primers shown in supporting information Table 1.
Oct3/4 Expression is Maintained in the Absence of Sall4
We previously reported that Sall4-null ES cells still express stem cell markers, including Oct3/4 and Nanog, suggesting that Sall4 does not affect the pluripotency of ES cells . By contrast, shRNA-mediated Sall4 inhibition in ES cells is reported to lead to a reduction in Oct3/4 and complete differentiation toward a trophectoderm lineage, but only under feeder-free conditions . To examine whether Sall4 is indeed involved in Oct3/4 expression, a blasticidin S-resistance gene was introduced by homologous recombination into the Oct3/4 locus of ES cells containing one floxed and one disrupted allele of Sall4 (Sall4flox/−; Fig. 1A). As a result, 18 of 22 clones were correctly targeted and successfully maintained for long periods in the presence of blasticidin, suggesting that Sall4 heterozygous cells retain Oct3/4 expression. These ES cells were transfected with Cre recombinase using an adenovirus and cultured in the absence of blasticidin to avoid preferential selection of Oct3/4-positive cells. By day 2, most ES cells were Sall4-negative, as evaluated by western blot analysis, and no apparent apoptotic cell death was observed throughout the experiment (data not shown). On clonal selection, 17 of 24 picked-up colonies were Sall4-null (Sall4Δ/−; Fig. 1A). Thus, it is unlikely that we selected a minor population of successfully proliferating undifferentiated cells. When Sall4-null ES cells were cultured on feeders, they formed compact colonies that were positive for Oct3/4 (Fig. 1B), consistent with our previous findings . When cultured under feeder-free conditions, some of the Sall4-null ES cells formed flattened and scattered colonies with prominent cytoplasmic protrusions whereas others still formed compact colonies (Fig. 1B). Despite the morphological changes, feeder-free Sall4-null ES cells were positive for Oct3/4, as evaluated by immunostaining, and could be maintained for long periods in the presence of blasticidin. As an one-half-fold reduction in Oct3/4 proteins leads to differentiation of ES cells toward a trophectoderm lineage , Sall4 heterozygous (Sall4flox/−) and null (Sall4Δ/−) ES cells were further analyzed by fluorescence-activated cell sorting (FACS), but the Oct3/4 protein amounts did not significantly differ between the two cell types (Fig. 1C). These results demonstrate that Oct3/4 is maintained in the absence of Sall4 with or without feeders.
Cdx2 is Upregulated in Feeder-Free Sall4-Null ES Cells
Next, the gene expression profiles in Sall4 heterozygous and null ES cells in the presence or absence of feeders were examined by RT-PCR. Sall4-null ES cells expressed stem cell markers, such as Oct3/4, Sox2, and Nanog, although Nanog expression was slightly reduced (Fig. 2A, 2B, 2D). By contrast, Cdx2, a trophectoderm marker, was significantly increased in Sall4-null ES cells only in the absence of feeder cells. This tendency held true even in the blasticidin-positive stem cell fraction (Fig. 2A, lane 1; Fig. 2D), suggesting that the Cdx2 increase may not be merely secondary to the reduction in Oct3/4, a key repressor of Cdx2. Gata6, a primitive endoderm marker, remained unchanged in Sall4-null ES cells, indicating that Sall4 absence does not lead to general upregulation of alternative lineages (Fig. 2D). As we found Gata6 expression in the feeders, analysis using feeder-dependent ES cells was impractical. Fgf4 and Utf1, which are directly regulated by Oct3/4 and Sox2, were downregulated in Sall4-null ES cells, suggesting that Sall4 may be necessary for proper activity of the Oct/Sox complex, although further analyses are required. To more comprehensively examine the gene expression profiles, microarray analysis was performed using feeder-free ES cells cultured in the presence of blasticidin (Gene Expression Omnibus Accession numbers: GSE14219). The data obtained were consistent with those shown in Figure 2A and 2B, but almost all the trophectoderm markers, except for Cdx2, remained unaltered in Sall4-null ES cells (supporting information Fig. 3). These findings suggest that Sall4 absence does not result in full commitment toward the trophectoderm. This phenomenon was not observed in a microarray analysis using ES cells cultured on feeder cells (Gene Expression Omnibus Accession numbers: GSE14219). Interestingly, when the feeder-free Sall4-null ES cells were replated on feeders, the increased Cdx2 expression returned to the normal level (Fig. 2C), suggesting that this increase was reversible and that Sall4-null cells were not completely committed to the trophectoderm. Taken together, Sall4 absence leads to reversible derepression of Cdx2 in the absence of feeders.
G1 Phase is Prolonged in Feeder-Free Sall4-Null ES Cells
To examine the self-renewal of feeder-free Sall4-null ES cells at the single cell level, blasticidin-resistant (Oct3/4-positive) stem cells were enriched and then cultured at a clonal density for 6 days without blasticidin, which would allow the survival of spontaneously differentiated cells (Fig. 3A). Under these conditions, most Sall4 heterozygous cells formed large alkaline phosphatase (AP)-positive colonies, suggesting efficient self-renewal of stem cells (Fig. 3B, 3C). Although the plating efficiencies examined on the next day after plating did not differ between Sall4 heterozygous and null ES cells, the resultant colony numbers, as well as the percentages of AP-positive colonies, from the Sall4-null cells were significantly smaller that those of the heterozygous ones (Fig. 3B, 3C). Typical Sall4-null colonies comprised small clusters of Oct3/4-positive cells surrounded by scattered Cdx2-positive cells, whereas most Sall4 heterozygous colonies were only positive for Oct3/4 (Fig. 3D), consistent with the data in Figure 2. As self-renewal activity of ES cells can be divided into maintenance of pluripotency and proliferative potential, we next examined the latter activity. Serial passage experiments clearly showed that feeder-free Sall4-null ES cells had reduced proliferation (Fig. 3E). The population doubling time of Sall4-null ES cells was 26.1 hours, whereas that of Sall4-heterozygous ES cells was 20.7 hours. Cell cycle analysis revealed that Sall4-null cells showed a decreased S-phase and increased G1 phase compared with heterozygous cells (Fig. 3F). These data suggest that Sall4 absence in ES cells leads to delayed G1/S transition. We next investigated microarray data to search for candidate genes that could regulate G1/S transition. Cyclin D1 was decreased, whereas some cyclin-dependent kinase inhibitors, such as p21 and p57, were upregulated in Sall4 absence (Fig. 3G). Thus, Sall4 is essential for efficient proliferation, possibly by controlling these cell cycle regulators. Although the reason for the increase in Cyclin D2 remains unknown, it would be unlikely to explain the extended G1 phase in Sall4-null cells.
Feeder-Free Sall4-Null ES Cells Remain Pluripotent and Do Not Differentiate into the Trophectoderm In Vivo
To examine whether feeder-free Sall4-null ES cells differentiate into the trophectoderm or remain pluripotent with derepressed Cdx2, ES cells marked with GFP were injected into eight-cell embryos or blastocysts. In these experiments, feeder-free ES cells were cultured in the absence of blasticidin to avoid preferential selection of Oct3/4-positive cells. In eight-cell embryo injection, both Sall4 heterozygous and null ES cells contributed only to the inner cell mass that represents the origin of ES cells and never to the trophectoderm (Fig. 4A, 4B). Of 94 embryos injected with Sall4 heterozygous ES cells, 89 (94.7%) were chimeric. Chimerism within the inner cell mass was 69.7% ± 16.5%. Of 113 embryos injected with Sall4-null cells, 108 (95.6%) were chimeric, and the chimerism was 71.1% ± 16.7%. When injected into blastocysts, both Sall4 heterozygous and null ES cells efficiently contributed to the epiblast, into which the inner cell mass differentiates, but never localized in the extra-embryonic tissues (Fig. 4C). Of 38 embryos (84.2%) injected with heterozygous cells and of 25 (76.0%) injected with Sall4-null cells, 32 and 19, respectively, were chimeric. Chimerism was not remarkably different (26.4% ± 13.2% for heterozygous cells and 23.4% ± 15.8% for null cells). These data indicate that Sall4-null cells have comparable pluripotency to heterozygous ones in vivo. As ES cells overexpressing Cdx2 efficiently contribute to the trophectoderm in vivo , we conclude that feeder-free Sall4-null ES cells are not fully committed to the trophectoderm lineage but retain their pluripotency despite the aberrant Cdx2 expression. As we previously showed by blastocyst injection that Sall4-null ES cells cultured on feeders solely contribute to the epiblast , our data unequivocally demonstrate that Sall4-null ES cells retain their pluripotency in vivo.
Next, we investigated whether the Sall4-null inner cell mass differentiates into the trophectoderm in vivo. This was not observed in conventional Sall4-null blastocysts, in which Oct3/4 only localized in the inner cell mass with Cdx2 in the trophectoderm . To further rule out possible maternal RNA effects, we injected Sall4 siRNA into one-cell embryos and cultured them until the blastocyst stage. In the Sall4-knockdown blastocysts, Cdx2 was expressed only in the trophectoderm and not in the inner cell mass (supporting information Fig. 4), suggesting that lineage conversion toward the trophectoderm does not occur in vivo, similar to the case for ES cells.
Sall1 Has a Minimal Role in the Maintenance of ES Cells
Sall1, another Sall family member, is essential for kidney development, and Sall1 and Sall4 interact genetically with each other during the formation of various organs, such as the anus, heart, and brain [6, 7]. Sall1 is expressed in ES cells and binds to Sall4, as shown by immunoprecipitation . Other studies have reported that Sall1, in addition to Sall4, is present in the functional protein network in ES cells [13, 14]. To investigate the roles of these Sall family genes and their functional redundancy in ES cells, we first generated Sall1-deficient ES cells by tandem targeting. When a vector containing a neomycin-resistance gene was introduced into heterozygous cells containing a hygromycin-resistance allele (supporting information Fig. 1), four of 88 clones obtained were homozygous for Sall1. This frequency was comparable with that of the first-round targeting using hygromycin resistance (four of 96), suggesting a minimal effect from the absence of Sall1 in ES cells. Indeed, isolated Sall1-null ES cells expressed stem cell markers, but not Cdx2, and their proliferation capacity was not impaired (data not shown).
Next, we established Sall1/4-null ES cells to verify the redundancy of these Sall family members. As Sall1/4 double heterozygotes die shortly after birth , it was impossible to obtain blastocysts from double homozygotes. Therefore, we first generated mice containing a floxed allele of Sall1 (supporting information Fig. 2). Homozygous mice were born healthy and fertile. When crossed with mice ubiquitously expressing Cre recombinase , kidney aplasia was observed, consistent with the findings for conventional Sall1 mutant mice (data not shown). These data indicate that Sall1 in the floxed allele becomes functionally null on Cre excision. We also generated floxed Sall4 mice using ES cells, as we have reported . Again, homozygous mice were healthy and fertile, and on Cre excision, homozygotes showed lethality at peri-implantation, mimicking the conventional Sall4 mutants. Next, ES cells were obtained from blastocysts of Sall1/4 double homozygotes, and Cre was introduced using an adenovirus. By day 2, most colonies became Sall1/4-null, similar to the case for Sall4-null ES cell generation. We successfully obtained Sall1/4-null ES cells at a high frequency (14 of 30 colonies), confirmed by western blotting analysis and immunostaining (Fig. 5A, 5B). Proliferation of Sall1/4-null ES cells was impaired, but these cells could be maintained for a long period, similar to the case for Sall4-null ES cells (data not shown). Oct3/4 and Nanog expressions were retained, although the latter was slightly reduced, whereas Cdx2 was only upregulated in feeder-free Sall1/4-null ES cells (Fig. 5C, 5D). These results suggest that Sall1 absence does not significantly exacerbate the phenotypes of Sall4-null ES cells. Furthermore, reintroduction of Sall4, but not Sall1, into the Sall1/4-null ES cells restored Cdx2 expression to the normal level (data not shown). Thus, Sall4, but not Sall1, is essential for the maintenance of ES cells.
Sall4 Binds to the HDAC Complex and Suppresses the Cdx2 Promoter
To determine the binding partners of Sall4 and the molecular mechanisms underlying the phenotypes of Sall4-null ES cells, FLAG-tagged Sall4 was introduced into wild-type ES cells and their lysates immunoprecipitated with an anti-FLAG antibody and analyzed by mass spectrometry. Figure 6A shows the binding of Sall4 to Mi2β, MTA1-3, p66, and histone deacetylase (HDAC), all of which belong to the Mi2-NuRD complex. We confirmed that three of these endogenous proteins (Mi2β, MTA1, and HDAC2) bound to Sall4 (Fig. 6B). Sall1 also bound to Sall4 (Fig. 6A, band II), consistent with our previous findings . ChIP analysis revealed that the histones in the Cdx2 promoter were hyperacetylated in Sall4-null ES cells (Fig. 6C), and that Sall4 was bound to the Cdx2 promoter (Fig. 6D). Thus, Sall4 is likely to recruit the HDAC complex, thereby repressing the Cdx2 promoter in a hypoacetylated inactive state, which could lead to stabilization of ES cells against aberrant differentiation into the trophectoderm.
The phenotype of Sall4-null ES cells are similar, to some extent, to that of ES cells lacking Nanog . In the Nanog-null state, ES cells retain their pluripotency but are prone to differentiate, and their efficiency of self-renewal is impaired. Thus, Nanog does not mark commitment but safeguards self-renewal by countering the effects of differentiation inducers. It is likely that Sall4 absence would also lead to a similar alternative state of ES cells that retain their pluripotency but exhibit impaired self-renewal. It is possible that Sall4 and Nanog control some overlapping sets of differentiation-associated genes, although Cdx2 upregulation has not been reported in Nanog-null ES cells. It is noteworthy that Sall4 is reported to bind to Nanog [19, 20]. Although the physiological significance of this interaction needs to be proved genetically, these proteins could prevent progression to commitment possibly by reversing the nascent epigenetic modification, including the mechanism shown in the present article. Interestingly, both Nanog and Sall4 were found to have reprogramming activity in fusion experiments using ES and somatic cells [21, 22], and both proteins are expressed in human ES and induced pluripotent stem (iPS) cells . Thus, Sall4 may also play a role in the establishment of iPS cells downstream of Oct3/4, Sox2, and Klf4.
All Sall family proteins have conserved 12-amino acid motifs in their N-terminus , and this region binds to the NuRD complex. Our data are not only consistent with this earlier report but also contribute toward the better understanding of Sall-mediated mechanisms. This is because we identified the NuRD complex by a nonbiased method, demonstrated endogenous Sall4-NuRD interactions and revealed the Cdx2 promoter to be a bona-fide target of the Sall4 protein complex. Recently, Nanog was shown to bind to a unique HDAC complex named Node, which contains MTA1/2 and HDAC1/2 but lacks Mbd3 and Rbbp7 . As Sall4 was also detected in the Nanog immunoprecipitates, it is tempting to speculate that Sall4 binds to Nanog and the Node complex, thereby repressing the target promoters. Sall4 is also reported to bind to Cyclin D1 and that they synergize for transcriptional repression . Furthermore analysis of Sall4-containing protein complex will be necessary.
The polycomb repressor complex (PRC), including Ring1A/1B, is known to be essential for the maintenance of ES cell identity and acts by repressing differentiation-associated genes . Although the phenotypes and gene expression profiles of Sall4-null ES cells display some similarities with those of Ring1A/1B-null cells, our preliminary ChIP analysis indicated that the binding sites of Sall4 and Ring1A/1B do not significantly overlap, suggesting PRC-independent repression of differentiation-related genes (data not shown).
Cdx2 is also repressed by Oct3/4, which binds directly to the Cdx2 promoter . Even a one-half-fold reduction in the Oct3/4 level leads to upregulation of Cdx2 and complete differentiation toward the trophectoderm . Thus, it is possible that the Cdx2 increase in Sall4-null ES cells is secondary to the Oct3/4 reduction. Indeed, Zhang et al.  reported that Sall4 knockdown by shRNA leads to a severe reduction in Oct3/4 and a subsequent Cdx2 increase and that Sall4 binds directly to the Oct3/4 promoter. However, our data indicated that Oct3/4 expression was minimally affected in Sall4-null ES cells. Moreover, Sall4-null cells selected based on Oct3/4 expression using blasticidin resistance still showed an increase in Cdx2, suggesting an Oct3/4-independent mechanism. Zhang et al.  also reported that ES cells with reduced Sall4 could not be maintained and differentiated completely into the trophectoderm because of the loss of Oct3/4. As well, these ES cells contributed to the trophectoderm when aggregated with eight-cell embryos. These results contradict our genetic data, in which Sall4-null ES cells could be maintained for long periods and never contributed to the trophectoderm. We did not observe any immediate apoptosis or differentiation on Sall4 removal, and the majority of cells behaved as described. Thus, it is unlikely that our data are solely based on a minor cell population with milder phenotypes, and off-target effects of the shRNA could be one of the reasons for the discrepancy.
Cdx2 upregulation was not observed when Sall4-null ES cells were cultured on feeders. Although the molecular nature of the factors supplied by the feeders remain unknown, it is clear that Sall4 is not required for Cdx2 suppression when ES cells are cultured on feeders. Analysis of Sall4 binding to the Cdx2 locus in the feeder-dependent condition should give a hint to resolving this issue. It is possible that similar supportive signals may be present in vivo. Sall4-null blastocysts never showed aberrant differentiation toward the trophectoderm lineage in vivo . Moreover, Sall4 siRNA injection into one-cell stage embryos did not result in a Cdx2 increase in the inner cell mass. Thus, it is likely that Sall4 plays only a permissive role in the maintenance of the inner cell mass in vivo. However, the situation may be similar to that of feeder-free ES cells, for example, after implantation into the uterus. At least, the Sall4-null inner cell mass cultured in vitro without feeders expressed Cdx2 . Detailed analyses following epiblast-specific knockout of Sall4 would address these issues and are under way in our laboratory.
Self-renewal activity of ES cells can be divided into maintenance of pluripotency and proliferative potential. Although feeder-free Sall4-null ES cells exhibit aberrant Cdx2 expression, they retain pluripotency. By contrast, proliferation was impaired and G1 phase was prolonged in Sall4 absence, which is consistent with our previous report in feeder-dependent conditions . We found several G1/S-specific cell cycle inhibitors were upregulated in Sall4-null ES cells, thus Sall4 could possibly directly inhibit some of these inhibitors by recruiting the HDAC complex. By contrast, Cyclin D1 was decreased in the absence of Sall4, indicating that Sall4 may also function as a positive regulator. Indeed Lim et al. recently showed that Sall4 binds to multiple loci and activates downstream targets . Detailed ChIP analysis, as well as functional studies, would elucidate the physiological relationship between Sall4 and cell cycle inhibitors.
We also demonstrated that Sall1, another Sall family gene, has a minimally overlapping role with Sall4 in ES cells. Sall1 has been proposed to be involved in the transcriptional network in ES cells [13, 14]. Endogenous binding between Sall1 and Sall4 and their in vivo genetic interaction have also been shown . However, Sall1-null ES cells had no apparent defects and even Sall1/4-null ES cells showed similar phenotypes to those of Sall4-null cells. These observations indicate that Sall1 has a minimal role in ES cell maintenance and that Sall4 has a function that is distinct from Sall1. Indeed, forced expression of Sall1 could not rescue the phenotypes of Sall4-null ES cells, suggesting qualitative differences in the signals evoked by Sall4 and Sall1. We recently identified AT-rich sequences that are bound by the C-terminal zinc finger domain of Sall1 . The sequences of the zinc finger domains are well conserved between Sall1 and Sall4, and Sall4 also binds to the AT-rich sequences, at least in vitro (unpublished results). Despite these similarities, Sall4 should regulate different targets from Salll1 in ES cells. As the numbers of zinc finger domains differ between the two proteins (Sall4 has eight and Sall1 has 10), the tertiary structure of the proteins may differ, which could lead to different affinities for target stimulation. We previously showed that Sall1-high cells in the developing kidney represent the progenitors of renal epithelia, including renal tubules . It is tempting to speculate that Sall1 maintains the renal progenitors by antagonizing derepression of differentiation-related genes, similar to the case for Sall4 in ES cells, by regulating different sets of target genes.
Taken together, we have shown that Sall4 is essential for efficient proliferation, as well as for stabilization, but not pluripotency, of ES cells by repressing aberrant trophectoderm gene expression. Furthermore studies will be required to elucidate the molecular nature of the cell stemness, which could be applicable to regenerative medicine using ES and iPS cells. It will be important to determine the extent to which similar mechanisms are conserved in organ formation, because Sall4 is expressed in a variety of organs during the later stages of development and is a causative gene for Okihiro syndrome.
In the present study, we have shown that Cdx2 is upregulated in feeder-free Sall4-null ES cells, whereas Oct3/4 expression is minimally impaired. These findings could arise from derepression of the Cdx2 promoter, which is normally suppressed by Sall4 and the Mi2/NuRD HDAC complex. However, Sall4-null cells retained their pluripotency and did not commit to the trophectoderm, as assessed by in vivo experiments. Thus, Sall4 does not contribute to the central machinery of the pluripotency, but stabilizes ES cells by repressing aberrant trophectoderm gene expression.
We thank K. Shinmyozu and A. Nakamura for mass spectrometry analyses, T.A. Endoh, Y.L. Yamaguchi, H. Meguro, and C. Kobayashi for technical assistance, and S.S. Tanaka for helpful discussions. This study was supported in part by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) and by the Global COE Program (Cell Fate Regulation Research and Education Unit), MEXT, Japan.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest.