A direct physical interaction between Nanog and Sox2 regulates embryonic stem cell self‐renewal

1Embryonic stem (ES) cell self-renewal efficiency depends on the level of expression of components of the pluripotency gene regulatory network. Among these, Oct4, Sox2 and Nanog play central roles. While the levels of Oct4 and Sox2 are relatively uniform in undifferentiated ES cells, the levels of Nanog vary considerably with high levels of Nanog directing efficient self-renewal. However, the mechanisms by which Nanog delivers this function in ES cells are not fully understood. In particular, although Nanog has been reported to interact with several proteins, the full extent of the Nanog interactome is not known.

2In the past few years, proteomic approaches have been employed to characterize and begin to understand the network of biochemical interactions controlling pluripotent cell function. This has resulted in the identification of additional proteins that interact with the key transcriptional factors Nanog, Sox2 and Oct4 to control and maintain the pluripotent state. Pioneering studies by identified a Nanog-centred interactome of 17 proteins that extended to other transcription factors including Oct4, Zfp281, Nac1, Rex1 and Nr0b1. This list of Nanog-interacting proteins has since been extended with a recent interactome identifying a total of 27 Nanog interactors. This relatively small number is in contrast to the larger number of interactors identified in recent Oct4 and Sox2 interactomes. Interactome studies have the potential to contribute to the elucidation of the mechanisms by which specific factors function. Central to this is the identification of the interacting amino-acid side chains on partner proteins as well as the functional significance of their association. To date, biochemical characterization of protein–protein interactions in pluripotent cells has been relatively sparse with most effort analysing the interaction between Sox2 and Oct4. From a biochemical perspective, little is known about how Nanog fits into the tight relationship between Oct4 and Sox2.

3Previously, we described a method to identify partner proteins interacting with nuclear proteins of interest in ES cells and used this to identify an extensive interaction network for the transcription factor Oct4. Here, this technique is applied to Nanog, resulting in identification of a Nanog interactome which includes over 130 Nanog partners in ES cells. From this, the direct interaction between Nanog and Sox2 was selected for further characterization, pinpointing individual residues required for the interaction and defining the functional consequences of elimination of the interaction between these central pluripotency regulators.

4An ES cell line expressing epitope-tagged Nanog protein was obtained by transfection of E14Tg2a cells with a construct in which the constitutive CAG promoter directs expression of (FLAG)3Nanog, linked via an IRES to puromycin resistance. Puromycin-resistant colonies were expanded and the resulting cell lines were analysed by immunoblotting. A cell line was identified (hereafter called F-Nanog) that expressed (FLAG)3Nanog at close to endogenous levels (Figure 1A). A qRT-PCR analysis of F-Nanog and E14Tg2a wild-type cells showed no significant difference in the expression level of the ES cell-specific genes Oct4, Sox2 and Rex1 (Figure 1B). In agreement with recent reports of autorepression by Nanog, F-Nanog cells show a strong decrease in expression of endogenous Nanog, which has the fortuitous consequence of maximizing the proportion of Nanog protein immunoprecipitated by anti-FLAG reagents.

5Nuclear extracts were prepared from F-Nanog cells and parental E14Tg2a cells and used for FLAG-affinity purifications as previously described. A Coomassie-stained gel of the eluted fractions showed several bands absent from the control E14Tg2a sample, indicating good signal-to-background ratio (Figure 1C). Mass spectrometry analysis was then performed on two independent affinity purifications from both F-Nanog and E14Tg2a control cells. An extensive set of Nanog partners was identified that could be grouped into several functional categories (Table I; Supplementary Table I; Figure 1D). The group with the highest representation is transcription factors, other groups present being chromatin modification complexes (e.g., NuRD and NcoR), proteins involved in phosphorylation or ubiquitination, basal transcriptional machinery members and RNA processing proteins. Mass spectrometric analysis of an independent cell line generated by expressing the same (Flag)3Nanog expression cassette in a Nanog-null ES cell line (RCNβH-B(t)) was used to verify candidate Nanog-interacting proteins (Table I; Supplementary Table I). Only the proteins identified in two out of three purifications are included in Table I and Supplementary Table I. Interactions between Nanog and Sox2, RNA polymerase II (RNAPolII), Nac-1, Sall4 and the NuRD subunit Mta2 were also observed by immunoblotting (Figure 1E). The Nanog interactome substantially overlaps with the published interactome of Oct4, Esrrb, Sall4, Nr0b1 and Tcfcp2l1 (Figure 2). Interestingly, Chd7 and the Ncor1 complex, which are not part of the Oct4/Esrrb/Sall4/Nr0b1/Tcfcp2l1 interactome, do interact with Nanog (Figure 2). This may reflect the robust interaction of Nanog with Sox2 (Table I; Supplementary Table I) as both Chd7 and the Ncor1 complex interact with Sox2.

6Due to the key role of Sox2 in ES cell biology, further characterization of the interaction between Nanog and Sox2 was undertaken. To determine whether the interaction between Nanog and Sox2 could be detected in wild-type ES cells, E14Tg2a nuclear extract was incubated either with an anti-Sox2 antibody and immunoprecipitates examined for the presence of Nanog or with an anti-Nanog antibody and immunoprecipitates examined for Sox2. Nanog was detected in Sox2 immunoprecipitates (Figure 3A) and Sox2 was also detected in Nanog immunoprecipitates (Figure 3B). To map the sites of interaction, co-transfections of (Flag)3Sox2 with (HA)3Nanog or Nanog deletion mutants were performed (Figure 3C) in E14/T cells. Nuclear extracts from ES cells transfected with (HA)3Nanog and (HA)3-tagged mutants lacking the N-terminus, the DNA binding homeodomain (HD) or the C-terminus of Nanog were immunoprecipitated with the HA antibody and after SDS–PAGE, immunoblots were probed for the presence of interacting Sox2 using a Flag antibody. (Flag)3Sox2 does not interact with a Nanog mutant lacking the C-terminal domain but the interaction between Sox2 and Nanog variants lacking either the N-terminus or the HD remained intact (Figure 3C). To identify the subregion of the Nanog C-terminal domain responsible for the interaction with Sox2, (Flag)3Sox2 was co-transfected with (HA)3Nanog variants carrying mutations within the C-terminal domain. Co-immunoprecipitations showed that deletion of the tryptophan repeat (WR) region, within the C-terminal domain of Nanog, but not residues C-terminal to the WR, abrogated the interaction with Sox2 (Figure 3D). Importantly, a Nanog mutant in which all 10 tryptophan residues in the WR region were mutated to alanines, (HA)3Nanog WRW10-A, also failed to interact with Sox2, pinpointing the tryptophan residues as critical determinants of the interaction with Sox2. To determine whether the interaction of Nanog and Sox2 was direct, Sox2 was co-expressed in E. coli alongside a fusion between Maltose Binding Protein and, either the Nanog tryptophan repeat, or the Nanog tryptophan repeat in which all the tryptophans were replaced by alanines (MBP-WR or MBP-WRW10-A) (Figure 3E). The MBP-fusion proteins were then purified on an amylose column and any interacting Sox2 was detected by immunoblotting with a Sox2 antibody. Only MBP-WR but not MBP-WRW10-A was able to co-precipitate Sox2 (Figure 3E). Taken together, these experiments indicate that Nanog and Sox2 interact directly, that the interaction with Sox2 can be mediated by the Nanog WR domain alone and that tryptophan residues within the WR are required for interaction with Sox2. In addition, the ability of these proteins to interact in E. coli implies that post-translational modifications are not required for interaction between Nanog and Sox2.

7To identify the region of Sox2 involved in the interaction with Nanog, we investigated mutants carrying deletions within the C-terminal domain, the HMG DNA binding domain or residues at the N-terminus of Sox2 (Figure 4A). Each of these mutants was co-expressed with (HA)3Nanog in E14/T cells, nuclear extracts prepared and the HA antibody used to co-immunoprecipitate (HA)3Nanog and interacting proteins. Samples were then analysed by SDS–PAGE and immunoblotting. (Flag)3Sox2 mutants lacking the N-terminal region, the DNA binding domain or the C-terminal 56 amino acid residues [(Flag)3Sox2 1-263] were still able to interact with Nanog (Figure 4A). However, (Flag)3Sox2 1-204 does not interact with Nanog, suggesting that the serine-rich region is involved in the interaction with the Nanog WR.

8The persistence of the Nanog–Sox2 interaction in nuclear extracts that have been treated with the nuclease, benzonase, to eliminate interactions mediated via DNA bridging, suggests that DNA binding is not required for the Nanog–Sox2 interaction. Moreover, the above results indicate that Nanog and Sox2 can interact in the absence of a DNA binding domain on either of the proteins (Figures 3C and 4A). To consolidate the notion that Nanog–Sox2 interaction is fully DNA independent, we show by co-immunoprecipitation of (Flag)3Sox2ΔHMG and (HA)3NanogΔHD that Nanog and Sox2 molecules that lack the DNA binding domains can still interact (Figure 4B).

9Our analysis of the ability of (HA)3Nanog to co-immunoprecipitate Sox2 mutants (Figure 4) suggested that the serine-rich region, from residues 205 to 263, plays a key role in the Nanog interaction. To narrow down the region of Sox2 interacting with Nanog, further deletion mutants within this region were generated (Figure 5A). Co-immunoprecipitation analyses show that while a Sox2 mutant truncated after residue 233 retained the ability to interact with Nanog, Sox2 mutants with deletion of residues between 205 and 233, or truncated after residue 212 were unable to interact with Nanog (Figure 5B). These analyses identify a critical Nanog-interacting region in Sox2 between residues 212 and 233. This sequence is highly enriched for hydroxyamino acids (12/21 residues) and, like the WR of Nanog, is devoid of acidic and basic side chains. Moreover, careful examination of this 21 amino-acid region highlighted three repeats of the sequence S X T/S Y that may be responsible for mediating the interaction with Nanog. To determine the potential importance of these motifs for the Nanog–Sox2 interaction, additional truncations were made after residues 218 and 226, which truncate Sox2 after repeat 1 or 2, respectively. This indicates that repeat 1 is sufficient for interaction with Nanog but that together repeats 1 and 2 interact with Nanog with an efficiency approaching that of wild-type Sox2 (Figure 5B). To examine the sequences required on Sox2 in more detail, a series of point mutations were generated within the repeats (Figure 6). Individual or combinatorial contributions of each of the three repeats to Nanog binding were initially examined (Figure 6A and B). Mutation of individual repeats suggests an order of importance for Nanog interaction of repeat 1>repeat 3>repeat 2 (Figure 6A). This is supported by analyses of the combinatorial mutants where mutation of repeats 1+3 almost entirely eliminates the ability of Sox2 to interact with Nanog (Figure 6B). Mutations of the amino acids at positions 1, 3 or 4 in all the three repeats were next analysed. Combined mutations at positions 1 and 3 had negligible effects (Figure 6C), implying that residues at these positions are not required for Nanog interaction. In contrast, the combined mutation of the tyrosines at position 4 indicates that these residues play a key role in the interaction with Nanog (Figure 6C). Together, these experiments suggest that the tyrosines are the residues directly interacting with the Nanog WR, with the tyrosines in repeats 1 and 3 being more important in this regard than the tyrosine in repeat 2.

10The Sox2 tyrosine residues could interact with Nanog partially via the hydroxyl, the phenyl ring or both. Since the tryptophans in WR are critical for the Nanog/Sox2 interaction this raises the hypothesis that hydrophobic stacking of the aromatic rings in the Sox2 tyrosines and the Nanog WR tryptophans mediate the interaction. If these were the case, then the tyrosines hydroxyl groups should be unimportant for the interaction between Nanog and Sox2. To test this hypothesis, the tyrosines were mutated to phenylalanine. The direct comparison of the interaction between (HA)3Nanog and (Flag)3Sox2:YYY>A or (Flag)3Sox2:YYY>F by co-immunoprecipitations clearly shows that substitution of the tyrosine residues with phenylalanines rescues the Nanog interaction, indicating that it is the benzene ring of these amino-acid residues that is required for the interaction to occur (Figure 6D).

11To investigate possible DNA sequences bound by the Nanog/Sox2 complex, (His)6-tagged Nanog and unmodified Sox2 were co-expressed in E. coli for use in Systematic Evolution of Ligands by Exponential Enrichment (SELEX). As controls, MBP-Nanog and (His)6-Sox2 were expressed individually. Purification from bacterial lysate containing co-expressed proteins on a nickel column followed by elution with imidazole yielded two proteins of the expected size for Nanog and Sox2. These were recognized by α-Nanog and α-Sox2 antibodies (Figure 7A), with N-terminal sequencing establishing the identities of the two bands as Nanog and Sox2. The Nanog–Sox2 interaction is robust, since the proteins co-purify through subsequent ion exchange (Figure 7B). The Nanog–Sox2 complex bound to the Ni-agarose, MBP-Nanog bound to amylose resin and (His)6-Sox2 bound to Ni-agarose were used for SELEX, the bound oligonucleotides cloned and the sequences determined (Figure 7C) used to derive the motifs shown (Figure 7D). The motif obtained from Nanog alone has a TAAT core sequence followed by CG, consistent with the motif obtained previously by SELEX and the nucleotide preferences of the isolated Nanog HD in EMSAs. Sox2 also gives a motif highly similar to that determined by SELEX (CA/TTTGA/T). The motif obtained from the Nanog/Sox2 complex is bipartite with bases 10–15 similar to the motif obtained by us and others for Sox2 alone and bases 5–7 showing similarity to the central core of the Nanog motif identified by SELEX (TAAT) in this work and by others. However, the published Nanog motif has a high degree of confidence over a four base sequence (TAAT) while the Nanog–Sox2 binding sequence shows high certainty for only three bases (TAA) with the preference for the 3′-flanking CG no longer apparent. This difference may reflect an alteration in the binding specificity of Nanog when in complex with Sox2. Interestingly, the SELEX motif shows high similarity to a Nanog/Sox2 motif identified by de novo methods from ChIP-Seq data, which notably also contains a 2-bp gap between the major binding nucleotide groups (Figure 7D). Therefore, a combined motif was generated and used to search available ChIP-Seq data sets. Analysis of three independent ChIP-Seq data sets identified 3257 Nanog/Sox2 overlapping peaks, which are common to the three data sets (out of a total of 16 454 from all Nanog/Sox2 overlapping peaks in the three data sets). Of these 3257 high confidence peaks, 29.1% (948 peaks) contain the motif. The motif occurs in a significantly smaller fraction of the Nanog only or Sox2 only peaks (4898 peaks out of a total of 31 271 peaks (15.7%; hypergeometric P-value <1 × 10−10). Examples of occurrences of the motif relative to the nearest gene are shown (Figure 7E; Supplementary Table II).

12To investigate the functional significance of the interaction between Nanog and Sox2, we took advantage of ES cells carrying a conditional Sox2 knock-out allele (Sox2CKO). In this cell line, one of the Sox2 alleles is flanked by loxP sites, while the other Sox2 allele has been replaced with a β-geo cassette. These cells also have a constitutively expressed CreERT2-IRES-Puro transgene integrated randomly in the genome. Upon addition of tamoxifen, CreERT2 is translocated to the nucleus and excises the Sox2 gene between the loxP sites (Figure 8A). As ES cells from which Sox2 activity has been removed are unable to self-renew and differentiate into trophectoderm-like cells, this cell line was used to test whether Sox2 mutant molecules impaired in Nanog binding could rescue the Sox2 null phenotype (Figure 8B). Sox2CKO cells expressing a GFP control plasmid completely differentiate upon Tamoxifen treatment (Figure 8B). As expected, cells transfected with an unmutated (Flag)3Sox2 cDNA rescued this differentiation phenotype. In contrast, cells expressing (Flag)3Sox2:YYY>A transgene showed a decrease in self-renewal activity with 50% fewer undifferentiated colonies compared to wild-type Sox2 (Figure 8C). In accordance with the interaction data, expression of (Flag)3Sox2:YYY>F fully rescued the differentiation phenotype (Figure 8C). To examine the possibility that the reduced colony formation by the Sox2:YYY>A cells was due to a reduced expression level, an immunoblot for Sox2 was performed. However, the amount of Sox2 expressed is comparable between Sox2:YYY>A and other lines and does not differ from the endogenous Sox2 level expressed by the parental line (Figure 8D). These data suggest that the interaction with Nanog is a key component in the function of Sox2 in ES cell self-renewal.

13To further investigate the effect of disrupting the Nanog–Sox2 interaction, the expression of genes present in the ChIP-Seq data sets was examined in cell lines expressing wild-type or mutant Sox2 (YYY>A). Of 13 genes analysed, 5 showed consistent differences by qRT-PCR when the Nanog/Sox2 complex was disrupted (Figure 8E). The genes that show altered expression include transcription factors reported to be important for ES cell identity (Rex1 and Klf5, the gene encoding the chromatin remodelling protein Myst4 as well as the cell-surface markers Ncam and Itga9. In addition, Oct4, which does not contain the Nanog/Sox2 motif, did not change expression level in absence of a Nanog/Sox2 functional complex. It is therefore likely that the effect of disrupting the Nanog/Sox2 complex on self-renewal is a consequence of the misregulation of the genes controlled by the two proteins in complex.

14By taking advantage of improved methodology the Nanog interactome has been expanded to over 130 proteins which can be subdivided into a number of different categories (Table I; Supplementary Table I). Many of the proteins identified in the interactome are components of large multi subunit complexes involved in chromatin modification, for several of which, all the known subunits are detected. Most of these are considered to be transcriptional repressors (NuRD, Polycomb Group protein (PcG), the atypical Polycomb complex E2F6, Sin3a and N-CoR) that bind to genomic sites adjacent to differentiation-specific genes to mediate repression. Emerging evidence suggests that NuRD and PcG complexes are also found at sites that are actively transcribed. How the NuRD complex is directed to target genes is not fully understood but Nanog and/or other NuRD-interacting transcription factors may target the complex to the relevant sites in the genome. In this respect, it is interesting that inducing Nanog protein results in enhanced binding of both Nanog and NuRD to the Nanog enhancer.

15Another proposed role for the chromatin modification complexes is to maintain repressed genes in a state that allows a rapid response to external cues. Evidence for this comes from the co-localization of enzymatically active PRC complexes and the paused form of RNA PolII at a large number of developmentally important genes. This could allow alterations in the signalling environment to promptly increase the level of gene expression. The interaction of Nanog with both PRC and RNA PolII may reflect this poised state of some genes. The association with the chromatin modification machinery is common to transcription factors involved in maintenance of ES cell pluripotency. However, the range of complexes binding to individual factors differs with SWI/SNF not directly connecting to Nanog (this study; ) but interacting with other transcription factors. Recent data showing that Esrrb can substitute for Nanog function in ES cells could in part be explained by the fact that Esrrb and Nanog bind to a number of the same chromatin modification complexes.

16The Nanog interactome includes a number of proteins that have not previously been identified in an ES cell transcription factor interactome. In addition to TET-1, which has also been shown to interact with Nanog, these include the RNA processing proteins Ilf3, Rbm9, Pum1/2 and the transcription factors Zfp326, Arid5b, Zfp609. Examining the function of these molecules and the significance of their interaction with Nanog will provide further detail on how the extensive protein interaction network functions to control pluripotency.

17In this study, we have focussed on the interaction between Nanog and Sox2 because of the central role of these proteins in the pluripotency gene regulatory network. Sox2 has been shown to interact with another key pluripotency factor, Oct4 by interaction of side chains within the DNA binding domains. In the case of Nanog and Sox2, interaction occurs through sequences outwith the DNA binding domains. Nevertheless, the sequence of the SELEX motif suggests that this interaction results in a specific spatial relationship of DNA binding domains of both proteins on DNA. The sequence of Sox2 that mediates interaction with Nanog is a triple repeat of the sequence S X S/T Y. Experiments analysing Sox2 mutants for their ability to rescue ES cells from differentiation induced by Sox2 deletion demonstrate the importance of the interaction of Nanog with Sox2. Mutation of the tyrosines in the S X S/T Y motifs to alanines reduces the formation of undifferentiated ES cell colonies to 50% of the level achieved using a non-mutant Sox2 cDNA in the absence of any difference in protein levels expressed by the transgene. Therefore, the 50% drop in undifferentiated colonies observed in the presence of Sox2:YYY>A is a result of the misregulation of Nanog/Sox2 gene targets. The use of the SELEX motif identified as a Nanog/Sox2 target sequence together with a previously published de novo target sequence allowed potential target genes of the Nanog/Sox2 complex to be identified. A number of these genes show altered expression upon abrogation of the Nanog/Sox2 interaction (e.g., Ncam, Itga9, Klf5 and Myst4). However, not all the genes tested are sensitive to loss of the interaction between Nanog and Sox2 (Supplementary Table II). This could suggest that in such cases the hydrophobic interaction of Nanog and Sox2 proteins is not required for chromatin binding, or that only in some cases is the associated gene sensitive to disruption of the interaction. The latter is reminiscent of our finding that only a subset of loci that bind Nanog respond to the presence of Nanog by modulating expression of a nearby gene.

18In ES cells, composite Oct/Sox binding sites have been proposed to be redundantly regulated by Sox4, Sox11 and Sox15. However, this redundancy does not extend to blockade of differentiation caused by Sox2 deletion. Consistent with this, Sox4, Sox11 and Sox15 are not present in the Nanog interactome and none of these Sox proteins contains a sequence that matches the S X S/T Y motif.

19The three copies of the S X S/T Y motif in Sox2 occur within a 15-residue sequence in which 9 residues are hydroxyamino acids. Despite this preponderance of hydroxyamino acids, it is the aromatic rings of the tyrosine residues that are critical mediators of the interaction with Nanog. This conclusion is derived from the fact that alanine substitution of all three serines at position 1 of the repeats or all three serines/threonines at position 3 of the repeats allowed continued efficient binding to Nanog, whereas alanine substitution of all three tyrosines decreased the Nanog interaction severely. Moreover, the fact that the Nanog interaction could be rescued when the tyrosines were substituted by phenylalanines indicates that the tyrosine hydroxyl groups are not required for the interaction and is highly suggestive that the two proteins interact by stacking of the aromatic rings. This is consistent with the fact that tyrosine and tryptophan residues cluster at protein–protein interaction ‘hot spots’. Functionally relevant stacking of tryptophan and tyrosine residues has also been demonstrated in the formation of an aromatic gate in apo flavodoxin, in the regulation of galactose oxidase activity and, of particular relevance to this study, in the interlocking of tyrosine and tryptophan residues at the interaction interface of human nuclear receptor pregnane X receptor (PXR) that mediates protein homodimerization.

20The lack of a requirement for the hydroxyl groups on Sox2 for the Nanog interaction is underscored by experiments using bacterially expressed recombinant proteins that demonstrate that the Nanog–Sox2 interaction occurs in the absence of post-translational modifications. However, this does not mean that post-translational modifications might not affect the interaction between Nanog and Sox2. The interaction between the two proteins occurs through polypeptide stretches devoid of strongly charged amino-acid side chains. Potential modification of hydroxyl groups on the Sox2 interaction surface, whether on the tyrosine or on the neighbouring serine and threonine residues, would introduce charged moieties that would be expected to interfere with the interaction between the hydrophobic interacting residues. Moreover, recent work indicates that hydroxyl groups on Sox2 can also be modified by addition of N-acetylglucosamine, although the effect on Sox2 function is unclear. The fact that Nanog interacts with proteins that mediate post-translational modifications such as phosphorylation and ubiquitination is consistent with the observation that Nanog is phosphorylated and ubiquitinated. In addition, Nanog partners could also be affected by such modifications because of physical proximity to the relevant enzymes. The role for these modifications and how they influence interactions between transcription factors and/or transcription factor function in ES cells is an important area for future investigation.

21The high number of Nanog-interacting proteins identified in this study suggest that Nanog acts as a ‘hub’ protein. The ability of individual partner proteins to interact with a hub protein like Nanog depends on the affinity of the interaction and the availability of the binding sites on both the hub protein and the partner, as has been discussed previously. Since both competitive and non-competitive interactions are simultaneously possible, it will be important to determine which factors compete for the same regions of Nanog. Of particular relevance will be whether factors that bind through the WR interact through a precise subregion of the WR or if there is variability in the exact sequence bound by a specific partner. To date, only Sox2 and Nac1 have been demonstrated to interact directly with the WR. Loss of the tryptophans of the WR has also been demonstrated to abrogate the interaction of Nanog with Sall4, Nr0b1, Zfp198 and Zfp281 but a direct interaction has not yet been shown for these proteins. In the situation where multiple factors bind the WR it is possible that binding of one factor could increase the affinity of another protein for interaction with an adjacent site in the same region resulting in co-operative binding of two or more factors. A clear potential example of this could be Nac-1, which has been reported to bind the C-terminal WR subunit. It is possible that both competitive and non-competitive binding to distinct sites on Nanog occurs simultaneously, allowing the assembly of large, functionally active complexes. An additional level of complexity arises from the possibility that the Nanog/Sox2 interaction may occur with either monomeric or dimeric Nanog. Potential mechanisms that affect the Nanog dimerization equilibrium, such as covalent modifications, could thereby play an important part in regulating interactions and subsequent downstream events.

22Mouse ESC lines were cultured on gelatin-coated dishes without feeders in GMEM/β-mercaptoethanol/10% FCS/LIF (GMEMβ/FCS/LIF) as described. Nanog null RCNβH-B(t) cells have been described: briefly, these cells have an IRES-HygromycinR-pA or an IRES-βgeo-pA replacement of Nanog sequences from intron I through to the 3’UTR. Sox2 conditional knock-out cells were obtained by re-targeting ES cells heterozygous for a Sox2flox allele with a Sox2-β-geo ‘knock-in’ targeting vector. This was followed by stable transfection of a pPyCAG-CreERT2IP construct (Figure 8A). Puromycin-resistant clones were screened for efficient deletion of the Sox2flox allele following tamoxifen treatment to select the Sox2CKO clone used here.

23E14Tg2a (Flag)3-Nanog and RCNβH-B(t) (Flag)3-Nanog cells were generated by electroporating E14Tg2a and RCNβH-B(t) cells with pPyCAG (Flag)3NanogIP linearized with ScaI. Electroporated cells were plated in GMEMβ/FCS/LIF and after 30 h, 1 μg/ml puromycin (Sigma, P9620) was added. Medium was replaced every 2 days and after 12 days, puromycin-resistant colonies were picked and Nanog expression levels were determined by immunoblotting with α-Nanog antibody. Sox2CKO cells expressing (Flag)3-Sox2, (Flag)3-Sox2:YYY>A, (Flag)3-Sox2: YYY>F or eGFP control were generated by electroporating Sox2CKO cells with pPyCAG(Flag)3-Sox2IH, pPyCAG(Flag)3-Sox2:YYY>AIH or pPyCAG(Flag)3-Sox2:YYY>FIH linearized with FspI. Electroporated cells were plated in GMEMβ/FCS/LIF medium and after 30 h, 100 μg/ml hygromycin B (Roche, 10843555001) was added. Medium was replaced every 2 days and after 12 days, hygromycin-resistant colonies were pooled to generate populations for rescue assays.

24Colony-forming assays were as described.

25(HA)3-Nanog, (HA)3-NanogΔN, (HA)3-NanogΔHD, (HA)3-NanogΔC, (HA)3-NanogΔC49 and (HA)3-NanogΔWR have been described. (HA)3-NanogWRW10A was generated by inserting synthetic DNA in which all 10 tryptophan codons were replaced by alanine codons between the two SexAI sites in the Nanog ORF. Flag-Sox2 was generated by cloning a Sox2 PCR product between the BamHI and NotI sites of pPyCAG(FLAG)3IP. (Flag)3-Sox2 1-263, (Flag)3-Sox2 1-204, (Flag)3-Sox2 1-233, (Flag)3-Sox2 1-226, (Flag)3-Sox2 1-218 and (Flag)3-Sox2 1-212 were generated by PCR using a forward primer containing a XhoI site 5′ to the (Flag)3 tag and reverse primers containing a stop codon at the desired position followed by a NotI site. (Flag)3-Sox2 ΔHMG and (Flag)3-Sox2 Δ205–233 were generated by PCR by overlap extension. PCR products were cloned between the XhoI and NotI sites in pPyCAGIP. (Flag)3-Sox2 ΔN120 was generated by PCR using a forward bipartite primer, containing a BamHI site, which anneals to the linker between the (Flag)3 tag and the Sox2 sequence and a sequence starting at codon 121 and a reverse primer containing a NotI site after the Sox2 stop codon. The PCR product was cloned between the BamHI and the NotI sites in pPyCAGIP(Flag)3-Sox2 vector. The primers used for the above PCRs are listed in Supplementary Table III.

26(Flag)3-Sox2:Rep1>A, (Flag)3-Sox2:Rep2>A, (Flag)3-Sox2: Rep3>A, (Flag)3-Sox2:Rep1+2>A, (Flag)3-Sox2:Rep1+3>A, (Flag)3-Sox2:Rep2+3>A, (Flag)3-Sox2:Rep1+2+3>A, (Flag)3-Sox2:SSS>A, (Flag)3-Sox2:TTS>A, (Flag)3-Sox2:YYY>A and (Flag)3-Sox2:YYY>F were generated by introducing synthetic DNA encoding the desired sequences (Genscript) between AgeI and ClaI sites in the Sox2 ORF. These two sites were introduced into Sox2 by silent mutation of C(693)>G (AgeI) and C(810)>A, C(813)>G (ClaI) (NM_011443.3). The coding sequence of all plasmids generated was checked by sequencing for absence of unwanted mutations.

27Sox2CKO cell lines, stably expressing (Flag)3-Sox2, (Flag)3-Sox2:YYY>A, (Flag)3-Sox2:YYY>F or eGFP were preselected for 3–5 days with 200 μg/ml G418 (PAA, P31-011), 0.75 μg/ml puromycin and 100 μg/ml hygromycin B. Cells were plated at clonal density (3 × 103 cells/63 cm2 dish) in hygromycin B (100 μg/ml) for 24 h. 4-OH Tamoxifen (1 μM) (Sigma, H7904) was added (12 h) to delete the endogenous Sox2 allele. Cells were cultured in 100 μg/ml hygromycin B for 14 days before alkaline phosphatase staining (Sigma, 86R). The rescue efficiency was determined from the ratio of alkaline-positive ES cell colonies (plus Tamoxifen/minus Tamoxifen).

28Preparation of nuclear extracts and purification of Flag-tagged proteins were performed as described. Briefly, nuclear extract was prepared from cells and Flag-tagged protein purified using 60 μl Flag-agarose beads per 1.5 ml of nuclear extract, during which samples were treated with 150 U/ml DNase Benzonase (4°C, 3 h) to decrease spurious protein purification due to DNA bridging. Nanog and interacting proteins were then eluted using Flag peptide (0.2 mg/ml). For production of proteins in E. coli MBP-WR/Sox2 and MBP-WRW10>A/Sox2 were cloned into pET Duet (Novagen) and expressed in BL21(DE3) cells. MBP-tagged proteins were lysed in 10 mM Tris pH 8.0, 100 mM NaCl, passed over amylose resin (NEB, E8021S), washed and proteins eluted with 10 mM maltose. Co-purifying protein was detected by immunoblotting. For SELEX, Nanog was cloned into pMalc2e (NEB) in frame with MBP and expressed in BL21 cells by addition of 1 mM IPTG. Cells were lysed in 10 mM Tris pH 8.0, 200 mM NaCl and purified on amylose resin. Sox2 was cloned into pET15b (Novagen) and expressed in BL21(DE3) by addition of 1 mM IPTG. Protein was purified by lysing cells in 25 mM Tris pH 8.0, 30 mM imidazole, 500 mM NaCl and passing lysate over nickel resin (His-select, Sigma, P6611). For co-expression, Nanog and Sox2 were cloned into pET Duet, to encode (His)6-Nanog and unmodified Sox2 and expressed in BL21(DE3) cells induced with 1 mM IPTG. Cells were lysed in 25 mM Hepes pH 7.6, 1 M NaCl, 5 mM imidazole and lysate incubated in batch mode with nickel resin. Ion-exchange purification of Nanog/Sox2 was performed at pH 7.6 on a 1-ml CM Sepharose FF column (GE Healthcare, 17-5056-01). Bound protein was eluted using a gradient of 0–1 M NaCl over 20 column volumes.

29For validation of Nanog interactors from E14Tg2a F-Nanog, 5 μg of Flag antibody (Sigma, F3165) or mouse IgG control (Santa Cruz, sc-2015) was added to 200 μl of nuclear extract and incubated (4°C, 3 h). Immunocomplexes were purified by addition of Protein G, washed in 20 mM Hepes pH 7.6, 10% glycerol, 100 mM KCl, 1.5 mM MgCl2, 0.2 mM EDTA pH 8, 0.2% NP-40, 0.5 mM DTT and eluted in Laemmli buffer. For probing extracts from RCNβH-B(t) and RCNβH-B(t) (Flag)3-Nanog, FLAG agarose (Sigma, A2220) was used to purify proteins. Complexes were eluted in Laemmli buffer and analysed by immunoblotting.

30For the characterization of the Nanog–Sox2 interaction, E14/T cells were transfected using Lipofectamine 2000 (Invitrogen, 11668-019) with (HA)3-Nanog or Nanog mutants and (Flag)3Sox2 or Sox2 mutants and nuclear extract was prepared. In all, 5 μg of HA antibody (MMS-101P, Covance) was added (4°C, 3 h) and immunocomplexes purified as above. Immunocomplexes were fractionated on NuPage-Novex 10% Bis-Tris gels (Invitrogen, NP0301BOX) and co-immunoprecipitating proteins detected by immunoblotting using the following antibodies: α-Sall4 (gift of Matthias Treir), α-Mta2 (8106, Abcam), α-Sox2 (sc-17320, Santa Cruz), α-Nanog (A300-397A, Bethyl Laboratories), α-Nac-1 (29047, Abcam), α-RNAPolII (PB-7C2, Euromedex) and α-Flag antibody (Sigma, F3165).

31Cells were lysed in 20 mM Hepes pH 7.6, 20% glycerol, 250 mM KCl, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 0.5% NP-40 and 1 × protease inhibitor cocktail (Roche). Protein extract was treated with 150 U/ml DNase Benzonase (Novagen) (30 min, 4°C), 60 μg of lysate run on SDS–PAGE, transferred onto nitrocellulose (Whatman(R) Protran(R)) and blots probed with antibodies diluted in 5% non-fat dry milk/TBS/0.01% Tween-20. Membranes were developed with Super-Signal West Pico (Pierce) and exposed to Hyperfilm (Amersham). Primary antibodies were anti-Nanog, 0.5 μg/ml, anti-Sox2 (sc-17320, Santa Cruz) 0.2 μg/ml and anti-LaminB (sc-6216, cs-6217 Santa Cruz) 0.2 μg/ml.

32Criteria for inclusion in Table I and Supplementary Table I as a Nanog-interacting protein are present in two out of the three purifications from E14Tg2a:F-Nanog and RCNβH-B(t):F-Nanog cells, with a Mascot score of >50 and at least three-fold higher than in the corresponding control experiment. The Mascot score is a statistical measure of confidence of correct identification of a protein from its peptides.

33Recombinant proteins used in SELEX assays were purified as described above. Purified protein identities were established by N-terminal sequencing (performed at the LIGHT Laboratories, Faculty of Biological Sciences, Leeds University by Edman degradation). Resin was washed with purification buffer supplemented with imidazole to 50 mM and used directly in SELEX. The oligo library contained 25 bp of random sequence flanked by ACGTGGATCCACTGACGG and GCTAGCGCCTCGAGACTTG. The double stranded library was synthesized by annealing the single-stranded library to a reverse primer and incubating in a Klenow fragment reaction. The initial round of SELEX consisted of incubation of 10 pmol of protein (on bead) with 20 pmol library in 20 mM Hepes pH 7.6, 200 mM KCl, 10% glycerol for 1 h. After extensive washing, bound oligonucleotides were eluted with imidazole containing buffer. PCR amplification of bound oligonucleotide was performed and the enrichment cycle repeated with 300 ng of total DNA. Five enrichment rounds were performed after which PCR products were cloned into TopoTA (Invitrogen). Individual clones were picked, DNA prepared and sequenced. Obtained sequences were submitted to MEME for motif searching.

34To identify joint Nanog–Sox2 (NS) DNA-binding events, enriched binding events (‘peaks’) based on three ChIP-seq experiments were combined using GeneProf (http://www.geneprof.org; ). After centring peaks on ±50 bp surrounding the highest point in the alignment, those overlapping in at least 1 bp were taken forward as joint NS peaks. A regular expression (A[TGA]T..[TC][AT]TT[GCT][AT]) was used to find occurrences of the NS motif. Peaks were linked to the closest transcription start site (max. distance≤30 kb). The full analysis workflow is accessible at gpXP_001309 (http://www.geneprof.org/show?id=gpXP_001309).

35RNA was extracted with TRIZOL (Invitrogen), DNase treated (Qiagen) and reverse transcribed with SuperScriptIII (Invitrogen). (Q)PCRs were performed in 384-well plates with a 480 LightCycler (Roche) using LightCycler 480 SYBR Green I Master (Roche). All primer sequences are listed in Supplementary Table III.