Differential Roles for Sox15 and Sox2 in Transcriptional Control in Mouse Embryonic Stem Cells* □ S

Sox family transcription factors play essential roles in cell differentiation, development, and sex determina-tion. Sox2 was previously thought to be the sole Sox protein expressed in mouse embryonic stem (ES) cells. Sox2 associates with Oct3/4 to maintain self-renewal of ES cells. In the current study, digital differential display identified transcripts for an additional Sox family member, Sox15 , enriched in mouse ES cells. Reverse tran-scription-PCR confirmed that Sox15 expression is highest in undifferentiated ES cells and repressed upon differentiation. Sox15 is expressed at low levels in several tissues, including testis and muscle. In vitro studies showed that Sox15, like Sox2, associated with Oct3/4 on DNA sequences containing the octamer motif and Sox-binding site. Gel mobility shift assays and SELEX analyses showed that Sox15 binds similar DNA sequences as Sox2 but with weaker affinity. In contrast to the early embryonic lethality observed in Sox2-null mice, Sox15-null ES cells and mice were grossly normal. DNA microarray analyses revealed that Otx2 , Ctgf , Ebaf , and Hrc are dysregulated in Sox15-null ES cells, however. Chromatin immunoprecipitation showed that Sox15, but not Sox2, bound to a Sox consensus binding site within the Hrc gene. Taken together, these data demon-strate differential roles for Sox15 and Sox2 in transcriptional control in mouse ES cells.

Sox family transcription factors play essential roles in cell differentiation, development, and sex determination. Sox2 was previously thought to be the sole Sox protein expressed in mouse embryonic stem (ES) cells. Sox2 associates with Oct3/4 to maintain self-renewal of ES cells. In the current study, digital differential display identified transcripts for an additional Sox family member, Sox15, enriched in mouse ES cells. Reverse transcription-PCR confirmed that Sox15 expression is highest in undifferentiated ES cells and repressed upon differentiation. Sox15 is expressed at low levels in several tissues, including testis and muscle. In vitro studies showed that Sox15, like Sox2, associated with Oct3/4 on DNA sequences containing the octamer motif and Soxbinding site. Gel mobility shift assays and SELEX analyses showed that Sox15 binds similar DNA sequences as Sox2 but with weaker affinity. In contrast to the early embryonic lethality observed in Sox2-null mice, Sox15null ES cells and mice were grossly normal. DNA microarray analyses revealed that Otx2, Ctgf, Ebaf, and Hrc are dysregulated in Sox15-null ES cells, however. Chromatin immunoprecipitation showed that Sox15, but not Sox2, bound to a Sox consensus binding site within the Hrc gene. Taken together, these data demonstrate differential roles for Sox15 and Sox2 in transcriptional control in mouse ES cells.
Embryonic stem (ES) 1 cells are derived from the inner cell mass of blastocysts and proliferate indefinitely while maintaining pluripotency, the ability to differentiate into all cell types within an organism (1). Mouse ES cells were first established in 1981 and led to the development of knockout mouse technology (2,3). Pluripotent stem cells were subsequently generated from human blastocysts in 1998 (4). Their rapid growth and pluripotency make human ES cells attractive sources for cell therapy in the treatment of diseases resulting from cell dysfunction, such as type 1 diabetes and Parkinson disease.
Sox family proteins arise from a group of genes related to the mammalian testis-determining factor Sry (15,16). Sox proteins are characterized by the highly conserved high mobility group (HMG) domains that consist of 79 amino acids and are involved in DNA recognition and binding (17). The HMG domains are also involved in the association with partner proteins (18,19). Sox proteins play essential roles in cell differentiation, development, and organogenesis (20,21).
In both mice and humans, homology-based screens have identified 20 distinct Sox proteins. Sequencing of the entire human and mouse genomes failed to identify additional members of this family (22). Based on primary sequence analysis as well as the presence of other structural elements, the 20 different Sox genes can be divided into eight groups (A-H).
To isolate Sox proteins specific for EC cells, Yuan et al. (23) designed oligonucleotide primers complementary to the most conserved least degenerate regions within the HMG box. Low stringency PCR was performed with cDNA from EC cells, and amplified products were subcloned. Thirty clones were sequenced, and all arose from the HMG domain of Sox2, a group B Sox family protein. An EC cell cDNA library was screened with this fragment, and all 15 clones identified encoded Sox2. No other Sox factor cDNA clones were obtained during this screening process, even when low stringency conditions were employed. Northern blot analyses showed high expression of Sox2 in mouse EC and ES cells. Taken together, these data indicated that Sox2 is the sole Sox protein expressed in mouse EC and ES cells.
Sox2 forms a complex with the POU family transcription factor Oct3/4 on the enhancer of the fibroblast growth factor 4 gene (23). Sox2 forms similar complexes on the enhancers of other target genes, such as UTF1 (24) and Fbx15 (25). In addition, both Oct3/4 (26) and Sox2 (27) are autoregulated by the Oct3/4-Sox2 complex. Sox2 binds to AACAA(A/T)G motifs, whereas Oct3/4 binds to ATT(A/T)GCAT motifs of the enhancers of these genes. Sox2-or Oct3/4-null embryos are lethal at the peri-implantation stages, and Sox2-or Oct3/4-null ES cells could not be established (14). Thus, Sox2 and its association with Oct3/4 are essential for the establishment and maintenance of pluripotent embryonic cells.
To better understand the processes governing mouse ES cell self-renewal, we analyzed expressed sequence tag (EST) data bases with digital differential display and identified several genes that are highly enriched in early mouse embryos and ES cells (12,25,28). Interestingly, ESTs encoding Sox15 (29) were found specifically in ES cells, indicating that Sox15 is expressed in ES cells in addition to Sox2. Sox15 expression has also been reported in muscle (30) and testis (31). Sox15 is the single member of Sox group G, but its HMG domain is closely related to the group B Sox proteins, including Sox2 (32), suggesting that the two Sox proteins may play similar roles in ES cells.
In the current study, we compared the expression, proteinprotein interactions, and DNA binding of Sox15 and Sox2 in mouse ES cells. Our data showed that Sox15 and Sox2 behave similarly in vitro. However, Sox15-deficient mice developed normally, and Sox2 and Sox15 appear to regulate different sets of genes in vivo. Our results highlight the unique DNA recognition mechanisms for each Sox family transcription factor.
Chromatin Immunoprecipitation (CHIP) Assays-Formaldehyde was added directly to the culture medium to a final concentration of 1% (v/v), and the dishes were gently shaken on a shaker at room temperature for 8 min. Glycine was added to a final concentration of 125 mM, and the dishes were returned to the shaker. After 5 min, dishes were washed three times with ice-cold phosphate-buffered saline and harvested by scraping into 3 ml of cold phosphate-buffered saline. Cells were collected by centrifugation at 2000 rpm for 5 min at 4°C, and the supernatants were discarded. Cell pellets were resuspended in 10 ml of cold phosphate-buffered saline plus 200 M phenylmethylsulfonyl fluoride. Cells were collected by centrifugation as above and resuspended in 5 ml of an ice-cold solution containing 5 mM HEPES, pH 8.0, 85 mM KCl, 0.5% (v/v) Nonidet P-40, and protease inhibitors (200 M phenylmethylsulfonyl fluoride, 1.4 g/ml pepstatin, 1 g/ml leupeptin). Samples were allowed to swell on ice for 10 min before homogenization with three strokes of a glass Dounce homogenizer to release nuclei. Nuclei were collected by centrifugation as above, and the supernatants were discarded. Nuclei were resuspended in 50 mM Tris, pH 7.6, 10 mM EDTA, 1% SDS, and the protease inhibitors (50 l/15-cm dish of cells). Samples were incubated on ice for 10 min, and immunoprecipitation dilution buffer A (0.01% SDS, 1.1% (v/v) Triton X-100, 1.2 mM EDTA, 16.7 mM Tris, pH 7.6, 167 mM NaCl) was added to a final volume of 750 l. Samples were sonicated for 30 s 10 times with 1-min intervals using a Bioruptor (Cosmo bio). Chromatin samples (500 l) were first precleared with normal mouse IgG (5 l) in the presence of protein G-Sepharose bead slurry (60 l of a 50/50 slurry of beads in TBS (16.7 mM Tris, pH 7.6, 167 mM NaCl) supplemented with 1 mg/ml bovine serum albumin and 200 g/ml salmon sperm DNA). Samples were incubated for 2 h at 4°C on a rotator, and beads were collected by centrifugation in a microcentrifuge at 2000 rpm. The unbound material (chromatin) was transferred to a new tube, and 5 g of each antibody were added. Samples were incubated overnight at 4°C on a rotator, 60 l of blocked protein G slurry were added, and incubation was continued on the rotator for an additional 2 h at 4°C. Beads were collected by centrifugation in a microcentrifuge at 12,000 rpm for 1 min. Beads were first washed twice with 500 l of ice-cold buffer B (0.05% (w/v) SDS, 1% (v/v) Triton X-100, 20 mM Tris, pH 7.6, 2 mM EDTA, 150 mM NaCl) and then washed once sequentially with buffer D (0.05% (w/v) SDS, 1% (v/v) Triton X-100, 20 mM Tris, pH 7.6, 2 mM EDTA, 500 mM NaCl), buffer 3 (0.25 M LiCl, 1.0% (v/v) Nonidet P-40, 1.0% deoxycholate, 10 mM Tris, pH 7.6, 1 mM EDTA), and buffer C (0.1% (v/v) Triton X-100, 150 mM NaCl, 20 mM Tris, pH 7.6, 2 mM EDTA). Beads were transferred to a microtube, and bound material was eluted by incubating the beads with 75 l of elution buffer (0.1 mM sodium bicarbonate, 1.0% (w/v) SDS) with vigorous shaking for 10 min. The elution was repeated four times. The four elutants were pooled, 30 l of 3 M NaCl and 10 g of RNase were added, and samples were heated at 65°C for 6 h to reverse the Schiff base linkage. DNA was collected by ethanol precipitation and resuspended in 20 l of water. A portion of this (0.5 l) was used for PCR amplification with ExTaq polymerase. Primers used were Hrc-R1-U (GTCTACCACCAACCTTCCCACTCACAAC) and Hrc-R1-L (GG-TGGGTCTGGCAGAGGGTCACA) for the fragment S1, Hrc-R2-U (TA-AGAGAGGGACCCAGAGAAAGAAAAG) and Hrc-R2-L (CTGTCTCTC-CTCTCTGAATCTGGGACTC) for S2, Hrc-R3-U (AGACAGACAGACA-CACAGAGAGACAGGC) and Hrc-R3-L (TCACTGTTCTGAGCTTCCG-TGTTTC) for S3, Hrc-R4-U (TAGATCTGGAGGTGGTTGGTTTGGG-TTC) and Hrc-R4-L (TTCAAAAGCTCTGGTGGTGACCAGCCTT) for S4, and Hrc-R5-U (CCGCCCCCTTTCCGCAGGTGCAG) and Hrc-R5-L (ATACATACCAGCCGGGTGTGCGTGTTCC) for S5.
Systematic Evolution of Ligands by Exponential Enrichment (SELEX)-pENTR-Sox15 or pDONR-Sox2 was recombined with pIH1119-gw to construct pIH1119-Sox15 and pIH1119-Sox2 for producing a fusion protein containing maltose-binding protein and Sox15 or Sox2. Purification of maltose-binding protein fusion proteins and SELEX were performed as previously described (12).
Construction of Sox15 Targeting Vectors-To disrupt the mouse Sox15 gene, we inserted a cassette carrying the internal ribosome entry site and a fusion of the ␤-galactosidase and neomycin resistance genes (␤geo) into the single exon of the gene, upstream of the HMG domain (34). A 1.5-kbp 5Ј arm of the targeting vector was amplified by the Expand long template PCR system (Roche Applied Science) with primers sox15-5arm-s-NotI (5Ј-GCGGCCGCAAGACAGGATTATTAGAC-3Ј) and sox15-5arm-as-SpeI (5Ј-ACTAGTCCCCCAGACGCTCCA-3Ј). A 3.9-kbp 3Ј arm was amplified with primers sox15-3arm-s-BamHI (5Ј-GGATCCGTCCCCTTTAGCCAAGAA-3Ј) and sox15-3arm-as-XhoI (3Ј-CTCGAGTTGGTGCTCTTAACCTCT-5Ј). The internal ribosome entry site-␤geo cassette was ligated between the two PCR fragments. A diphtheria toxin A cassette was placed downstream of the 3Ј arm. The resulting targeting vector was linearized with SacII and introduced into RF8 ES cells by electroporation (35). Genomic DNAs from G418-resistant colonies were screened for homologous recombination by Southern blot analyses.
Genotyping of ES Cells and Mice-For 5Ј recombination, genomic DNA was digested with EcoRI, separated on a 1% agarose gel, and transferred to nylon membrane. A 550-bp 5Ј probe was amplified with sox15-5Јsouth-s (5Ј-CAGAGACGAAATCGCAGCCGC-3Ј) and sox15-5Јsouthern-as (5Ј-CGCCTTCCTTCTCACTACAGA-3Ј). Hybridization with this probe resulted in an 8.0-kbp band from the wild-type locus and a 4.3-kbp band from the targeted locus.
After identifying ES cell clones that were correctly targeted, we determined genotypes of mice and ES cells with three-primer PCR. A sense primer, ␤geo-screening1 (5Ј-AATGGGCTGACCGCTTCCTCGT-GCTT-3Ј), was designed from the ␤geo cassette to amplify the targeted locus. Another sense primer, sox15-3Ј-m-tail-s (5Ј-GATGGCGCAGCA-GAACCCCAAGATGCAC-3Ј), was designed from the HMG domain to amplify the wild-type locus. An antisense primer, sox15-3Ј-m-tail-as (5Ј-GGGGCTCCAGCAAGGGAAGTATTATATG-3Ј), was designed to amplify both the wild-type and targeted loci. Amplification with these three primers produced a 1036-bp band from the wild-type locus and a 1371-bp band from the targeted locus.
Generation of Sox15-null and Rescued ES Cells-ES cells deficient in Sox15 were obtained by culturing the heterozygous ES cells with a high concentration (2-6 mg/ml) of G418 (36). To obtain rescue cells, pCAG-IP-Sox15 was transfected into Sox15-deficient ES cells by electroporation. To identify clone expressing Sox15, we screened colonies resistant to 2 g/ml puromycin by Northern blot and Western blot analyses.
DNA Microarray-Total RNA from Sox15 heterozygous ES cells and homozygous ES cells were labeled with Cy3 and Cy5 hybridized to Mouse Development Microarray (Algilent) according to the manufac-turer's protocol. The arrays were scanned with G2565BA Microarray Scanner System (Agilent). Hybridization was repeated with different clones. Data were analyzed with GeneSprings (Silico Genetics).

RESULTS
To identify candidate ES cell-specific genes, we compared EST libraries derived from mouse ES cells (three libraries, 33,077 clones) and various somatic tissues (103 libraries, 1,040,493 clones) by digital differential display. Twenty Unigene clusters were found exclusively in ES cell-derived libraries. One of them (Mm. 176369) encoded Sox15. This was unexpected, since Sox2 had been recognized as the sole Sox family protein expressed in mouse ES cells. Sox15 expression is found in testis and muscle, but its expression in ES cells had not been studied.
Sox15 expression was highest in two independent ES cell lines, RF8 and MG1.19, as assessed by RT-PCR (Fig. 1A). Its expression decreased following retinoic acid-induced differentiation. Weaker expression was also detected in stomach and skin. We confirmed the expression of Sox15 in undifferentiated ES cells at a protein level (Fig. 1B). These data show that Sox15, in addition to Sox2, is expressed in mouse ES cells.
Despite occupying different Sox family groups, Sox2 and Sox15 have a high degree of sequence homology. Thus, we first examined whether Sox15 associates with Oct3/4 as well. We introduced Myc-tagged Oct3/4 into MG1.19 mouse ES cells together with either HA-tagged Sox2 or HA-Sox15. We then precipitated HA-Sox proteins with anti-HA antibody and examined whether Myc-Oct3/4 was co-precipitated. Western blot analysis with anti-Myc antibody detected the association of Myc-Oct3/4 with both HA-Sox2 and HA-Sox15 (Fig. 2, upper  panel). When Myc-Oct3/4 alone was introduced into cells, it was not precipitated with anti-HA antibody. The association between Oct3/4 and the Sox proteins was also observed when Oct3/4 was precipitated with anti-Myc antibody (Fig. 2, lower  panel). Thus, both Sox15 and Sox2 associate with Oct3/4 in mouse ES cells.
We next performed a gel mobility shift assay to examine whether Sox15 binds to the enhancer regions of known Sox2 target genes (Fig. 3A). The mouse Fgf4 enhancer contains the consensus octamer motif (ATTAGCAT) and Sox binding site (AA-CAAAG), to which Sox2 and Oct3/4 bind synergistically (23). When we incubated 32 P-labeled Fgf4 enhancer fragments with COS7 cell extracts expressing Oct3/4, Sox2, or both, shifted bands corresponding to an Oct3/4 monomer, a Sox2 monomer and an Oct3/4-Sox2 complex were observed, respectively.
When we incubated 32 P-labeled Fgf4 probe with COS7 cell extracts expressing Sox15, a shifted band was also observed. When COS7 cell extracts expressing both Sox15 and Oct3/4 were used, the shifted band migrated more slowly than Oct3/4 alone, demonstrating that Sox15 and Oct3/4 synergistically bind to the Fgf4 promoter.
With probes containing point mutation in either the octamer motif or Sox-binding site, the band corresponding to the Sox15-Oct3/4 complex did not appear, indicating that the association between Sox15 and Oct3/4 is dependent on DNA binding of the two transcription factors to the octamer motif and Sox-binding site, respectively.
The band corresponding to the Sox15 monomer is less intense than that corresponding to the Sox2 monomer, suggesting a weaker affinity of Sox15 for the Fgf4 enhancer. In contrast, the band corresponding to the Oct3/4-Sox15 dimer was similar in intensity to the corresponding Oct3/4-Sox2 dimer. This suggests that synergism with Oct3/4 compensates for the weaker affinity of Sox15 for the DNA binding element.
In addition to bands corresponding to Oct3/4, we observed a more slowly migrating band even with extracts from control COS7 cells transfected with a mock plasmid (Fig. 3A and Supplementary Figure). This band is likely to correspond to Oct1, as previously shown by Yuan et al. (23). We found that this band was supershifted with the addition of Sox2 or Sox15, indicating that the two Sox proteins associate not only with Oct3/4 but also with Oct1.
We next examined binding to the mouse Fbx15 enhancer, which contains the consensus Sox binding site and a motif two nucleotides different from the octamer motif (TTTATCAT) (Fig.  3B). Oct3/4 can bind to this site only with the presence of Sox2 (25). We found that Oct3/4 alone or Sox15 alone barely bound to the Fbx15 enhancer, consistent with the weaker affinity of Sox15 for the Sox consensus site than Sox2. However, the Oct3/4-Sox15 complex can bind to the Fbx15 enhancer as effectively as the Oct3/4-Sox2 complex. These data further support the synergistic DNA binding of Oct3/4 and Sox15.
Despite similarities between the HMG domains of Sox15 and Sox2, some differences exist that may explain the decreased DNA binding of Sox15. Thus, we constructed expression vectors encoding chimeric proteins in which the HMG domains of Sox2 and Sox15 were exchanged. Sox2-15-2 contains the HMG domain of Sox15 within Sox2, and Sox15-2-15 is the opposite. We introduced these constructs into COS7 cells and confirmed protein expression (Fig. 4A).
We then performed gel mobility shift assays with these COS7 cell extracts and either the Fgf4 enhancer (Fig. 4B) or the Fbx15 enhancer (Fig. 4C). With both enhancers, we found that Sox2-15-2 showed weaker affinity than Sox2, whereas Sox15-2-15 showed stronger affinity than Sox 15. When Oct3/4-expressing extracts were included in the reaction, the differences between these Sox proteins became smaller. These data indicate that the weaker affinity of Sox15 for the Fgf4 and Fbx15 enhancer is, at least in part, attributable to small differences within the HMG domain compared with Sox2.
The decreased binding of Sox15 compared with Sox2 to the Fgf4 and Fbx15 enhancers suggests that the recognition sequences may differ between these proteins. To test this possibility, we performed SELEX analysis. We prepared affinity purification columns with recombinant maltose-binding protein-tagged Sox15 or maltose-binding protein-tagged Sox2. Oligonucleotides with random sequences were purified on these affinity columns, amplified by PCR, and reapplied to the affinity column. After repeating this procedure five times, the eluted DNA was subcloned and sequenced. As expected, Sox2 bound to AACAATG (Fig. 5), and Sox15 preferentially bound to nearly identical sequences. Thus, the decreased affinity of Sox15 for the examined enhancers is not attributable to an increased preference of an alternative binding site.
We have shown that Sox15 binds to DNA in a synergistic manner with Oct3/4, but we next wished to examine the functional outcomes of this interaction. We used a reporter gene in which luciferase cDNA was driven by the Fgf4 enhancer and promoter (23). This reporter gene was introduced into COS7 cells together with expression vectors encoding Oct3/4 and/or Sox proteins.
When both Sox15 and Oct3/4 were introduced, the Fgf4 enhancer was activated ϳ4-fold (Fig. 6A), but neither Sox15 nor Oct3/4 alone had an effect. However, luciferase activity induced by Sox15 and Oct3/4 was lower than that induced by Sox2 and Oct3/4, which showed ϳ7-fold enhancement. We obtained a similar result with the Fbx15 reporter construct (25), in which the luciferase cDNA is driven by the minimum thymidine kinase promoter plus five copies of the Fbx15 enhancer (Fig. 6B). These data indicate that Sox15 synergistically activates the Fgf4 and Fbx15 enhancers with Oct3/4 but to a lesser extent than Sox2.
To study the possible function(s) of Sox15 in ES cells, we inactivated Sox15 by homologous recombination. The Sox15 gene consists of two exons, and we constructed a targeting vector in which a cassette consisting of internal ribosome entry site and ␤-geo (a fusion of ␤-galactocidase and the neomycinresistant gene) replaced the HMG domain (Fig. 7A). This vector was introduced into RF8 ES cells by electroporation. Screening of 250 G418-resistant clones identified three positive clones by both PCR and Southern blot analyses (Fig. 7B).
One of the positive clones was injected into blastocysts of C57/BL6 mice, and germ line transmission was obtained. Sox15-null mice were born with the expected Mendelian ratio (ϩ/ϩ:ϩ/Ϫ:Ϫ/Ϫ ϭ 30:57:25). They were grossly normal in ap-pearance and fertile. This is in great contrast with the observed peri-implantation embryonic lethality of Sox2-null mice.
To better study the function of Sox15 in ES cells, we established homologous mutant ES cells by selecting heterozygous cells with a high concentration of G418. We obtained 48 colo-  Fbx15 (B). A, cell extracts were prepared from COS7 cells expressing Oct3/4 and/or Sox proteins and analyzed by electrophoresis mobility shift assay. As a control, the FGF4-WT probe was incubated with F9 EC cell extracts. Shown on the left are the positions of bands corresponding to each transcription factor(s). nies with 2-3 mg/ml of G418. PCR and Southern blot analyses revealed that 4 of 48 clones were homozygous for the Sox15 deletion. Northern blot analysis demonstrated that Sox15 transcripts were absent in these clones (Fig. 7C). This was again in great contrast to Sox2 mutants, of which homozygous mutant ES cells could not be obtained.
Sox15-null ES cells were normal in morphology and proliferation when maintained in an undifferentiated state on STO feeder cells (Fig. 7D). They were also competent in differentiation after LIF removal, retinoic acid treatment, and teratoma formation (not shown). Northern blot analyses showed that the expression levels of Sox/Oct target genes, such as Fgf4, UTF1, and Fbx15, were indistinguishable between wild-type, Sox15heterozygous, and Sox15-null ES cells (Fig. 7E).
We next performed DNA microarray analyses to study the effect of Sox15 deletion on gene expression. We used Agilent mouse development arrays that contain ϳ20,000 genes ex-pressed in early embryos. Comparison between Sox15-heterozygous ES cells and Sox15-null cells showed that Fgf4 and Fbx15 are normally expressed in Sox15-null ES cells, consistent with the results of Northern blot analyses (not shown). In addition, ES cell-specific genes, such as Nanog and Oct3/4, are also normally expressed in Sox15-deficient cells.
To study whether the expression of these genes was directly regulated by Sox15, we performed chromatin immunoprecipitation (CHIP) analysis. We identified six putative Sox binding sequences (S1-S5; Fig. 8B) in the flanking regions of the mouse Hrc gene. We precipitated formalin-fixed nuclear extracts of ac.uk/cgi-bin/seqlogo/logo.cgi). The height of each letters is sorted so that the most common one is on top. The height of the entire stack is adjusted to signify the information content (measured in bits) of the sequence at that position.
RF8 ES cells with anti-Sox15 antibody and performed PCR to amplify fragments S1-S5.
All five fragments were amplified from extracts before CHIP (Fig. 8C). However, only S1 and S5 were amplified from the anti-Sox15-precipitated extracts. When we performed CHIP with nuclear extracts from Sox15-null ES cells, S1, but not S5, was amplified. This suggests that Sox15 antibody may crossreact with other protein(s) with the binding and washing conditions we used. However, our data did show that S5 was specifically co-purified with Sox15.
When CHIP was performed with anti-Sox2 antibody, none of the five fragments were amplified (Fig. 8C). In contrast, the proximal enhancer of the mouse Nanog gene was co-precipitated with anti-Sox2 antibody, but not with anti-Sox15 antibody. Thus, Hrc is regulated specifically by Sox15 in vivo, whereas Nanog is regulated by Sox2, but not Sox15.
The S5 sequence does not contain the octamer motif. ChIP analysis showed that S5 was not co-purified with anti-Oct3/4 antibody (Fig. 8C). In contrast, the Nanog enhancer containing the octamer motif was co-precipitated. These data indicated that Oct3/4 does not associate with Sox15 on the S5 sequence of the Hrc gene.
In order to better understand Sox15 binding to the Hrc S5 sequence, we performed gel mobility shift assays with a 32 Plabeled probe containing this sequence. When this probe was incubated with ES cell nuclear extracts, a shifted band was observed (Fig. 8D). This band was identical to that observed with nuclear extracts of COS7 cells expressing Sox2. In contrast, a shifted band corresponding to Sox15 was only observed with extracts of COS7 cells expressing Sox15, but not with ES cell nuclear extracts. These data demonstrated that the S5 Sox binding site is preferentially bound by Sox2 in vitro, in contrast to the in vivo situation revealed by CHIP.

DISCUSSION
Sox2 was previously thought to be the sole Sox protein expressed in mouse ES and EC cells (23). However, we now report that Sox15 is also expressed in mouse ES cells. Our study demonstrated that Sox15 and Sox2 regulate different sets of genes in vivo, despite similar protein-protein interactions and DNA recognition in vitro; Sox2 regulates Fgf4 and Fbx15, whereas Sox15 regulates Hrc, Otx2, Ctgf, and Ebaf.
It remains elusive how the specificity of each Sox protein is determined. Sox15 and Sox2 recognize and bind similar DNA sequences, and the binding of both is enhanced by the presence of Oct3/4 binding. This is not wholly unexpected, since the HMG domains, which bind to DNA and Oct3/4, are 78% identical between the two Sox proteins.
In contrast, the identity outside the HMG domains is less than 30%. Some Sox proteins bind to other proteins through non-HMG domains (18). For example, the non-HMG motif PLNLSSR is required for binding of Sox6 to the co-repressor CtBP2 (41). Sox15 and Sox2 may bind to different transcription regulators through non-HMG domains and therefore regulate unique sets of genes. Consistent with this notion, we found that Oct3/4 did not associate with Sox15 on the Hrc gene.
The finding that Sox15 shares the same DNA recognition sequence with Sox2 suggests that Sox15 may interfere binding of Sox2 to its target genes such as Fgf4 and Fbx15. However, we found that the expression levels of Fgf4 and Fbx15 were not increased in Sox15-deficient ES cells (Fig. 7E). Furthermore, overexpression of Sox15 in ES cells did not decrease the expression of these genes (not shown). These data showed that Sox15 does not affect DNA binding of Sox2.
Recently, Kuroda et al. (42) reported that Nanog contains an adjacent octamer motif and Sox-binding site in the 5Ј-flanking region. They showed that Sox2 bound to this site in EC cells and embryonic germ cells. However, an undefined factor preferentially bound to the same site in ES cells they examined. Further studies are required to determine whether Sox15 binds to the Nanog gene in those cells.
During the preparation of this manuscript, the generation of Sox15-null mice was reported (43). Consistent with our data, FIG. 7. Targeted disruption of the mouse Sox15 gene. A, Structure of the Sox15 genomic locus, a targeting vector, and the targeted locus generated by homologous recombination. The targeting vector contains the ␤-geo cassette in place of the HMG domain. The length of the diagnostic EcoRI (E) or MluI (M) restriction fragments and the locations of the 5Ј or 3Ј probes for Southern blot analysis are shown. B, Southern blot analysis. Specific hybridization with the 5Ј probe produces an 8.0-kb band from the wild-type locus and a 4.3-kb band from the target locus. Hybridization with the 3Ј probe produces an 18.0-kb band from the wild-type locus and a 3.8-kb band from the target locus. ϩ/ϩ, ϩ/Ϫ, and Ϫ/Ϫ, genotypes of Sox15ϩ/ϩ, Sox15ϩ/Ϫ, and Sox15Ϫ/Ϫ cells, respectively. C, Northern blot analysis. Total RNA was isolated from indicated ES cells and hybridized with either Sox15 probe or Sox2 probe. D, morphology of Sox15-null ES cells. ES cells of the indicated genotypes were cultured at a low density to produce single cell-derived colonies. E, expression of Fgf4, Fbx15, and UTF1 in Sox15-knockout ES cells. Total RNA from ES cells of the indicated genotypes was analyzed with Northern hybridization for the expression of known target genes of Oct3/4 and Sox2. IRES, internal ribosome entry site. these Sox15-deficient mice were normal in development, gross appearance, and fertility. However, cultured Sox15-deficient myoblasts displayed a marked delay in differentiation in vitro (43). Expression of the early myogenic regulated factors MyoD and Myf5 was altered in Sox15-deficient myoblasts. These results suggest another specific role for Sox15 that cannot be compensated by other Sox family members.
Sox15 is the sole member of Sox family group G, and it is only found in mammals (32). Other organisms, including Fugu, Drosophila melanogaster, and Caenorhabditis elegans, do not have Sox15 orthologs. These data indicate that the Sox15 gene evolved relatively recently. This might account for its relatively minor roles compared with other members of the Sox family transcription factor.
FIG. 8. Identification of direct and specific targets of Sox15. A, total RNA from ES cells of the indicated genotypes was analyzed with RT-PCR for the expression of Otx2, Ctgf, Ebaf, and Hrc. In "rescued" cells, an expression vector for Sox15 was introduced into Sox15-null ES cells. B, five putative Sox binding sites (S1-S5) in the mouse Hrc gene. C, formalin-fixed nuclear extracts from wild-type or Sox15-null ES cells were precipitated with anti-Sox15 antibody or anti-Sox2 antibody. DNA fragments containing the putative Sox binding sites in the mouse Hrc gene were amplified by PCR. A DNA fragment containing the Nanog proximal enhancer was also amplified. D, gel mobility shift assay showing the binding of Sox2 and Sox15 to the S5 sequence of the mouse Hrc gene. The 32 P-labeled Hrc-S5 oligonucleotide was incubated with nuclear extracts from RF8 ES cells or COS7 cells expressing Sox2 or Sox15. As a control, the Fgf4-WT oligonucleotide was used. Ab, antibody.