Conversion of Sox17 into a Pluripotency Reprogramming Factor by ...

EMBRYONIC STEM CELLS/INDUCED PLURIPOTENT STEM CELLS

Conversion of Sox17 into a Pluripotency Reprogramming Factor by

Reengineering Its Association with Oct4 on DNA

RALF JAUCH,a#

IRENE AKSOY,b#

ANDREW PAUL HUTCHINS,b#* CALISTA KEOW LENG NG,

a,cXIAN FENG TIAN,

b

JIAXUAN CHEN,b PAAVENTHAN PALASINGAM,a PAUL ROBSON,b,d LAWRENCE W. STANTON,b,d PRASANNA R. KOLATKARa,d

aLaboratory for Structural Biochemistry and bStem Cell and Developmental Biology, Genome Institute of

Singapore, Singapore; cSchool of Biological Sciences, Nanyang Technological University, Singapore; Department

of Biological Sciences, National University of Singapore, Singapore

Key Words. Sox transcription factors • Induced pluripotent stem cells • Reprogramming • Pluripotency • Endoderm differentiation

ABSTRACT

Very few proteins are capable to induce pluripotent stem(iPS) cells and their biochemical uniqueness remainsunexplained. For example, Sox2 cooperates with other

transcription factors to generate iPS cells, but Sox17, de-spite binding to similar DNA sequences, cannot. Here, weshow that Sox2 and Sox17 exhibit inverse heterodimeriza-

tion preferences with Oct4 on the canonical versus anewly identified compressed sox/oct motif. We can swapthe cooperativity profiles of Sox2 and Sox17 by exchang-

ing single amino acids at the Oct4 interaction interface

resulting in Sox2KE and Sox17EK proteins. The reengi-neered Sox17EK now promotes reprogramming of so-matic cells to iPS, whereas Sox2KE has lost this potential.

Consistently, when Sox2KE is overexpressed in embryonicstem cells it forces endoderm differentiation similar towild-type Sox17. Together, we demonstrate that strategic

point mutations that facilitate Sox/Oct4 dimer formationon variant DNA motifs lead to a dramatic swap ofthe bioactivities of Sox2 and Sox17. STEM CELLS

2011;29:940–951

Disclosure of potential conflicts of interest is found at the end of this article.

INTRODUCTION

The Sox and POU (Oct) families of transcription factors con-sist of 20 and 14 members, respectively, and often act synerg-istically during vertebrate development (reviewed in [1–3]).Despite their diverse biological roles, the specificity of Soxproteins for DNA elements is largely indistinguishable andthe amino acids involved in specific DNA contacts are highlyconserved [4]. Therefore, a transcription factor acting alonemay lack site-specific binding, though selective dimerizationwith binding partners may provide the means to achieve spec-ificity in transcriptional control [5]. Indeed, several distinctSox/POU pairs have been implicated as key regulators of cel-lular fates: Sox2/Oct4 are essential factors in embryonic stem(ES) cells [6–8]; Sox2/Brn2 is important in neural develop-ment [9]; Sox11/Brn1 pair regulates glial cells [10]; andSox17 has been shown to cooperate with Oct4 during mesen-doderm formation [11].

In early development, though both are capable of inter-action with Oct4, Sox2, and Sox17 bring about fundamen-

tally different developmental effects. Sox2 is required forthe development of the epiblast [12], whereas Sox17 isessential in the formation of the definitive gut endoderm[13]. In addition, while Sox2 is a pluripotency factor,Sox17, when forcibly expressed in mouse and human EScells pushes the cells toward an endoderm-like cell fate [14].Furthermore, Sox17 cannot replace Sox2 in reprogrammingsomatic cells into induced pluripotent stem (iPS) cells [15].This demonstrates that, despite binding essentially the sameDNA motif, Sox2 and Sox17 have highly divergent develop-mental capabilities.

We became interested in competitive interactions betweenSox2 and Sox17 for the binding to Oct4 for two reasons.First, our recent x-ray crystallographic analysis of the DNA-binding domain (DBD) of Sox17 revealed that the Oct4 inter-action surface displays a markedly different electrostatic inter-face when compared with Sox2. These differences couldpotentially impart distinct abilities on Sox2 and Sox17 tointeract with Oct4 [16]. Second, our single-cell gene expres-sion analysis of preimplantation development [17] indicatedthat prior to the formation of the Oct4þ/Sox17þ primitive

Author contributions : R.J., I.A., and A.P.H.: conception and design, collection and/or assembly of data, data analysis and interpretation,manuscript writing, final approval of manuscript; C.K.L.N., X.F.T., J.C., and P.P.: collection of data; P.R., L.W.S., and P.R.K.: financialsupport, administrative support, data analysis and interpretation, manuscript writing, final approval of manuscript.

#R.J., I.A., and A.P.H. contributed equally to this article.

*Present address: Immunology Frontier Research Centre, Osaka University, Osaka, Japan.

Correspondence: Lawrence W. Stanton, Ph.D., Genome Institute of Singapore, 60 Biopolis Street #02-01, Singapore 138672. Telephone:þ65-6-808-8006; Fax: 6-808-8291; e-mail: [email protected] or Prasanna R. Kolatkar, Ph.D., Genome Institute of Singapore, 60Biopolis Street #02-01, Singapore 138672. Telephone: þ65-6-808-8006; Fax: 6-808-8291; e-mail: [email protected] ReceivedFebruary 17, 2011; accepted for publication March 14, 2011; first published online in STEM CELLS EXPRESS April 6, 2011. VC AlphaMedPress 1066-5099/2009/$30.00/0 doi: 10.1002/stem.639

STEM CELLS 2011;29:940–951 www.StemCells.com

endoderm and Oct4þ/Sox2þ epiblast, individual early innercell mass cells coexpress all three factors, thus, representing abiological system where competition between Sox2 andSox17 could have real development consequences.

The role of the Sox2/Oct4 pair as an inducer of pluripo-tency is well established [8, 18], and there is evidence thatSox17 and Oct4 functionally cooperate during endoderm dif-ferentiation [11, 19, 20]. It is, therefore, conceivable thatSox2 and Sox17 can compete for Oct4 and form stable com-plexes on specific genomic regions characterized by distinctcomposite cis-regulatory motifs.

To test this, we set out to identify variant sox/oct motifconfigurations in genomic regions occupied by Sox2 and Oct4in mouse ES cells [21] and identified a novel compressed ele-ment. In vitro heterodimerization assays, however, revealedthat Sox17/Oct4, but not Sox2/Oct4, was able to cobind thiselement, whereas the Sox2/Oct4 complex predominates on thecanonical site. By designing mutations using structural mod-els, we generated point mutations that swapped the heterodi-merization preferences of Sox2 and Sox17. This change notonly affected the dimerization capability of Sox17/Oct4 ver-sus Sox2/Oct4, but concomitantly converted Sox17 into apotent reprogramming factor and Sox2 into an inducer of theendodermal fate.

MATERIALS AND METHODS

Computational Analysis

To search for different sox/oct motif configurations, we took themotif derived from the Oct4/Sox2 chromatin immunoprecipita-tion (ChIP)-seq data [21] and implemented a position–weight ma-trix (PWM) search tool that scans through sets of FASTAsequences (further details are contained in the Supporting Infor-mation procedures). To construct different configurations of thesox/oct motif, we generated variants of the motifs, insertingunbiased base pairs in between the sox/oct motif, or makingreverse complement versions. This way we constructed differenthypothetical configurations of the sox/oct motif, from the canoni-cal (soxf_0bp_octf), order (octf_0bp_soxf), convergent (soxf_0b-p_octr; f and r signify the strand of the motif element)and divergent (octr_0bp_soxf). We also removed or introducedspacer base pairs (soxf_-1bp_octf, soxf_0bp_octf, soxf_1bp_octf.. soxf_10bp_octf).

Recombinant Proteins

HMG domains of mouse Sox2, Sox7, and Sox17 were cloned andheterologously expressed and purified to homogeneity asdescribed [16, 22]. An extended version of the Sox2HMG(denoted Sox2HMGl spanning amino acids 33–141, swissprot-idP48432) was cloned using the TOPO and GATEWAYTM LRtechnologies (Invitrogen, Singapore, www.invitrogen.com) andpurified as described [22]. The POU domain of mouse Oct4 (resi-dues 126–289, swissprot id P20263) was produced as describedin the Supporting Information procedures.

Site-Directed Mutagenesis

Amino acid substitutions were introduced using the QuikChange-XL site-directed mutagenesis kit (Stratagene, Singapore, www.ge-nomics.agilent.com) using DNA oligos listed in Supporting Infor-mation procedures. Recombinant mutant proteins were expressedand purified as described above.

Electrophoretic Mobility Shift Assays (EMSAs)

All EMSAs were carried out using double-stranded 50 Cy5-la-beled DNA (Sigma Proligo, Singapore, www.sigmaaldrich.com;see Supporting Information Table S1) following published

procedures [23]. Binding buffer contains 20 mM Tris-HCl pH8.0, 50 lM ZnCl2, 100 mM KCl, 10% glycerol, 2 mM b-mercep-toethanol, 0.1 mg/ml bovine serum albumin (BSA), and 0.1% (v/v) Igepal CA-630. A total of 250 nM dsDNA probes were mixedwith proteins in binding buffer and incubated for 1-hour at 4�Cin dark. Samples were loaded into a prerun 12% (w/v) 1� Tris-glycine polyacrylamide gel in 1XTG (25 mM Tris, pH 8.3; 192mM glycine) buffer and imaged.

Reporter Assays

Approximately 105 human embryonic kidney 293 cells in 24-wellplates were transfected with 275 ng pGL4-TK luciferase reporterplasmids containing a dual repeat of idealized compressed (50-CGGCGCGGCATTGTATGCAAATCGGCGGCGCGGCGCGGCATTGTATGCAAATCGGCGGCG0-30) and canonical (50-GGCGCGGCATTGTCATGCAAATCGGCGGCGGGCGCGGCATTGTCATGCAAATCGGCGGCG-30) motif, 2 ng pRL-SV40Renilla transfection control and 360 ng pcDNA3.1/nV5 plasmidscontaning Sox or Oct4 coding sequences using Lipofectamineand Optimem reagents (Invitrogen). After overnight incubation,the transfection mix was replaced with Dulbecco’s modifiedEagle’s medium growth media containing 10% fetal bovineserum þ 2 mM L-glutamine. Cells were lysed after 3 days andthe luciferase and renilla activities were detected using the dualluciferase assay kit (Promega). The experiments were repeatedtwice with three technical replicates each.

Generation and Characterization of iPS cell

The iPS assay was carried out using procedures modified from[24]. Experimental details of the iPS generation and validatingexperiments such as quantitative (Q) RT-PCR, immunostaining,viral titer determination, teratoma, and chimera formation aredescribed in the Supporting Information procedures.

Generation of Stable Sox Variant ExpressingCell Lines

The stable introduction of V5 and Venus-tagged Sox proteinsinto E14 mouse ES cells is detailed in the experimentalprocedures.

RESULTS

Analysis of Sox/oct Motif Configurationsin Mouse ESC

We devised a PWM scanning tool to quantify the occurrenceof different configurations of the composite Sox2/Oct4 DNAbinding motif (see Experimental Procedures and SupportingInformation Figure S1). We systematically assessed sox/octmotif configurations (Fig. 1A). For motif scans, we interro-gated 200bp windows of the mouse genome that have beenshown by ChIP to be occupied by Sox2 and Oct4 in ES cells[21]. Motif frequencies for differently spaced sox/oct in thecanonical orientation (soxf_nbp_octf where f denotes the for-ward orientation, r the reverse complement, and n the numberof spacer base pairs) are detailed in Figure 1B (see Support-ing Information Figure S2 for a complete list). As expected,the canonical sox/oct motif was most strongly enriched(370% above background). Many of the other motifs withspacer nucleotides inserted were found to be only modestlyenriched.

The second most abundant motif detected in the Sox2/Oct4 dataset was a novel ‘‘compressed’’ motif (here labeledas: soxf_-1bp_octf, Figure 1C, a 54% increase over back-ground). The compressed motif differs from the canonicalmotif by deletion of a single nucleotide at position seven ofthe canonical motif that is only weakly specified in the

Jauch, Aksoy, Hutchins et al. 941

www.StemCells.com

canonical motif (Fig. 1C). We recovered the sequences identi-fied by this alternative, compressed PWM and generated aweblogo representation (Fig. 1C).

Next, we compared the motif frequencies in genomicregions bound by Sox2 or Oct4 alone with regions coboundby Sox2 and Oct4 (Fig. 1D). As anticipated, canonical andcompressed composite motifs were detected less frequently atsites occupied by individual transcription factors (TFs) ascompared with cobound sites.

Profiling of Sox/Oct4 Binding to DifferentiallyConfigured Motifs

To assess the preference of Sox2 and Oct4 to physicallyassemble on differentially configured sox/oct motifs, we con-ducted EMSAs using purified DBDs of Sox2 and Oct4. Wescreened the heterodimerization potential of Sox2/Oct4 DBDpairs on a panel of systematically modified sox/oct motifs(Fig. 1A). As expected, we observed a strong cooperativeinteraction of Sox2 and Oct4 on the canonical element (Fig.2A, lanes 5–7). However, the dimerization was substantiallydiminished if a spacer of one or two base pairs was intro-duced (Fig. 2A, lanes 8–13). Heterodimerization was enabledon elements with a spacer length between 3 and 10 nucleo-tides, albeit with reduced efficiency, suggesting an additive or

weakly competitive binding mode (Fig. 2A, lanes 15–39). Ifthe arrangement of the motifs is altered (Fig. 1A, 2A, lanes41–49), dimer formation is abrogated for a changed motiforder (octf_0bp_soxf), strongly diminished for the convergingmotif (soxf_0bp_octr) and reduced for the diverging orienta-tion (soxr_0bp_octf). Unexpectedly, the newly identified com-pressed motif did not enable heterodimer formation (Fig. 2A,lanes 2–4).

We reasoned that three scenarios could explain the abun-dance of a compressed motif in genomic regions cotargetedby Sox2 and Oct4 despite the inability of the two proteins toheterodimerize on this sequence. First, the over-representationof the compressed motif could be caused by independentChIP enrichment of Sox2 or Oct4 bound singly to this siteand the averaging over a large population of cells creates thenotion of co-occurrence. Second, compressed motifs could belocated in the proximity to canonical motifs that recruit func-tional Sox2/Oct4 heterodimers causing it to copurify in ChIPexperiments, whereas the actual binding event occurs at anearby canonical motif. Third, the apparent enrichment of thecompressed motif is inflated by its similarity to the canonicalmotif.

To explore the last two issues, we measured the co-occurrences of both canonical and compressed motifs in theSox2/Oct4 bound regions. There are 1,784 Sox2/Oct4-bound

Figure 1. In silico discovery of sox/oct variant motifs. (A): Schematic of the variant motifs generated in this study. All the combinations wereexplored with 0–10 bp of spacer between the sox and oct parts of the motif, also a ‘‘compressed’’ motif was created by deleting the base pairbetween the sox and oct motif. (B): Frequency of variant motifs with �1, 1 through 10 bp’s of spacer. The frequency of motif is expressed aspercentage increase over a random list of genomic coordinates. A second, independently, generated random list is included for comparison. (C):The WebLogos of motifs discovered by the position weight matrix matching tool. (D): Percentage increase over background of motif against thedifferent ChIP-seq lists derived from [21], ‘‘Oct4/Sox2’’ is an overlap list containing all of the overlapping Oct4/Sox2 bound sites. The ‘‘Oct4’’and ‘‘Sox2’’ lists contain all of the Oct4 and Sox2 binding sites, respectively. Abbreviation: ChIP, chromatin immunoprecipitation.

942 Reengineering Sox Transcription Factors

regions, and the canonical motif is present at least once in995 of these locations, whereas the compressed motif occurs548 times. A total of 425 regions contained both a com-pressed and a canonical motif, which was a statistically sig-nificant co-occurrence (p ¼ 8.6 e �13; Fig. 2B). To estab-lish that the compressed motif constitutes a genuine motifand not a cryptic motif hidden within the canonical motifwe recovered the locations of the motifs and analyzed theintersection of genomic coordinates (Fig. 2C). We foundthat 135 compressed/canonical motifs shared genomic coor-dinates within 1bp of each other’s coordinate centers, indi-cating motif overlap (Fig. 2C). However, the majority ofcompressed motifs (528) are distinct from the canonicalmotif suggesting that the compressed motif constitutes agenuine motif variant. Finally, we asked whether the com-pressed motif is actually a subset of the canonical motif. Tothis end, we conducted a careful analysis of the PWMs andthe sequences that match the PWM and found that the ma-jority of sequences retrieved by the PWMs for the com-pressed and canonical motifs are distinct (Supporting Infor-mation Procedures).

We previously have shown that the binding affinities ofSox2 and Sox17 to sox motifs are indistinguishable [16].To compare the Oct4 heterodimerization properties of Sox2and Sox17, we assessed their differential assembly using tosame panel of sox/oct motif configurations tested for Sox2and Oct4. We observed that the overall pattern of heterodi-merization on most motif configurations recapitulates obser-vations made for the Sox2/Oct4 pair (Fig. 2D, lanes 5–49).However, in contrast to the inability of Sox2 and Oct4 tocoassemble on the compressed motif, Sox17 and Oct4exhibited a cooperative-binding mode on this element (Fig.2D, lanes 2–4). This finding indicates a qualitative bindingdifference of the Sox2/Oct4 versus the Sox17/Oct4 hetero-dimers on DNA motifs, which might constitute the bio-chemical basis for their distinct roles in mammaliandevelopment.

Assembly of Distinct Sox/Oct4 Pairs Depends on thecis-Regulatory Context

To further dissect the differential assembly of the Sox2/Oct4versus Sox17/Oct4 pairs, we conducted competition bindingexperiments. Figure 3A lanes 3–5 indicate that all bindingpartners retarded equal amounts of DNA when added indi-vidually. To assess whether the sequence of the addition ofthe protein components affects the heterodimerization effi-ciency, we premixed the labeled DNA probe first with eitherSox2 or Sox17 or the Oct4-binding partners before addingthe remaining proteins to the reaction. Lanes 7–8 and 10–11reiterate that only the Sox17/Oct4 heterodimer was capableof assembling on the compressed sox/oct element, whereasthe cobinding of Sox2 and Oct4 was obstructed. The hetero-dimerization of Sox17 and Oct4 on the compressed elementwas not abrogated in the presence of Sox2 (lanes 16–19). Asimilar experiment was carried out to study the assemblybehavior on the canonical sox/oct motif (Fig. 3B). Sox2 andOct4 exhibited a cooperative binding mode on this elementas indicated by a complete supershift of the Oct4/DNA com-plex if Sox2 was present (lanes 7–8). However, when Sox17and Oct4 were incubated with the canonical element, thesupershift of the Oct4/DNA complex was incomplete. Fur-thermore, when the three factors were incubated together,the majority of supershifted Oct4 migrated in a Sox2/Oct4/DNA complex, whereas the Sox17/Oct4/DNA complex wasof markedly lower abundance. The bulk of the Sox17 pro-tein remained singly bound to DNA. The sequence of

Figure 2. Differential assembly of Sox2HMG plus Oct4POU (A)

and Sox17HMG plus Oct4 (D) on a series of different motif configu-rations. Motif configurations were systematically designed as outlinedin Figure 1A using 30-bp DNA elements containing identical Sox(CATTGTC) and Oct4 (ATGCAAAT) sequences. To minimize bind-ing anomalies due to cryptic elements at the periphery or within thespacer region, we used idealized motifs (CATTGTC for sox andATGCAAAT for oct) and introduced G’s and C’s as spacer andboundary nucleotides. Each DNA element was mixed with individualtranscription factor proteins as well as with both Sox2 and Oct4DBDs in combination. A total of 250 nM each cy5-labeled DNA ele-ment was incubated with 50 nM Sox2HMGl/Sox17HMG and 250 nMOct4POU proteins alone and in combination followed by PAGE toassess the formation of ternary Sox/Oct4/DNA complex. Using Oct4-POU in excess was necessary to ensure similar shifts in lanes whereonly Sox2/17 or Oct4 was added because the active fraction of Oct4was measured to be 5-fold lower than the active fraction ofSox2HMGl (data not shown). The position of the various proteinDNA complexes are marked by arrows. The number of spacer base-pairs is indicated above and below the gels. X ¼ �1 denotes the com-pressed and X ¼ 0 the canonical elements. TANDEM denotes twoconsecutive sox motifs preceding the oct motif. The co-occurrence ofcanonical and compressed motifs within genomic regions cobound bySox2 and Oct4 in mouse embryonic stem cells is depicted in (B). Thesignificance of the overlap as compared with a randomly generatedlist was established using a binomial test. (C): Shows the portion ofmotifs with shared genomic coordinates.


www.StemCells.com

addition did not significantly affect the experimental out-come. We conclude that Sox2 out-competes Sox17 on thecanonical sox/oct element. Together, although Sox17/Oct4binding was sterically possible on the canonical element(lanes 10–11) complex formation was strongly enhanced onthe compressed element for the Sox17/Oct4 pair (Fig. 3A,lanes 10–11). Conversely, Sox2/Oct4 dimerization wasoccluded on the compressed motif, presumably due to sterichindrance (Fig. 3C).

We also tested another F-group Sox protein, Sox7, ashorter version of the Sox2 protein restricted to the core ofthe HMG domain, and we altered the spacer residue of thecanonical motif to establish the significance of our findings(Supporting Information Figure S3). In summary, Oct/Soxassembly on the compressed element is a robust propertyof the HMG domain of F-group Sox proteins Sox7 andSox17, whereas Sox2/Oct4 dimerization is impaired on thismotif.

Point Mutations at the Oct4 Interaction SurfaceSwap the DNA-Dependent Dimerization Potential ofSox2 and Sox17

Next, we sought to identify the structural elements that equipSox2 and Sox17 with distinct Oct4 interaction surfaces. Struc-tural studies on Sox17 [16] and Sox2 [25, 26] revealed thattheir DNA-binding and DNA-bending mechanisms are virtu-ally identical. However, we observed a difference at helix three[16] that constitutes the presumed Oct4 contact interface [25,26]. Interestingly, B group Sox proteins contain a basic lysinewithin this helix that is replaced by an acidic glutamate in Fgroup Sox proteins (Fig. 4A) [16]. To test if this residue affectsthe differential assembly of Sox2 and Sox17 with Oct4, wereciprocally mutated lysine 95 in Sox2 into a glutamate(Sox2KE) and glutamate 122 in Sox17 into a lysine (Sox17EK,Fig. 4B). We also generated Sox2 and Sox17 constructs inwhich all eight helix-3 residues in proximity to the putative

Figure 3. Differential assembly of Sox2/Oct4 and Sox17/Oct4 on the (A) compressed and (B) canonical elements. Proteins were added to 250nM DNA elements in the sequences indicated above the lanes. To distinguish shifts caused by Sox2 and Sox17, a N- and C-terminally extendedSox2-HMG (Sox2HMGl) domain was used in these experiments. As a consequence, both, the Sox2/DNA and the Sox2/Oct4/DNA complex,migrate slower than the corresponding Sox17 complexes, allowing us to visualize a Sox2/Oct4 and Sox17/Oct4 DNA complex on the same gel.Protein–DNA mixtures were incubated for 10 minutes before the next protein component was added. Sox2 and Sox17 were kept at 50 nM andOct4POU at 250 nM final concentration. Positions of binary Sox/DNA and Oct4/DNA as well as of ternary Sox/Oct4/DNA complexes are indi-cated. (C): Model summarizing the differential assembly of Sox2/Oct4 and Sox17/Oct4 on the canonical versus the compressed element.


Oct4 contact interface were converted into their correspondingcounterparts found in the other Sox protein, leading to a com-plete swap of helix-3 between Sox2 and Sox17 (denotedSox2H3-17 and Sox17H3-2, Fig. 4B). As further controls, weproduced proteins with swapped helix-3 while retaining the

Glu122 in Sox17 and Lys95 in Sox2 (Sox17H3-2exE anddenoted Sox2H3-17exK; Fig. 4B). Next, we assessed thepotentials of the six mutated Sox proteins for coassembly withOct4. The ability of the mutated Sox2 constructs, Sox2KE andSox2H3-17, to dimerize with Oct4 on the canonical motif was

Figure 4. (A): Alignment of the amino acid sequence of all mouse Sox proteins. The Sox subfamilies [1] are indicated to the right. The num-bering corresponds to the Sox17 sequence. Alpha helices are marked with a red bar. The Phe-Met wedge is indicated with an orange bar belowthe alignment and other DNA interacting residues are marked by black closed circles. Highly conserved and similar sequences are shaded inblack or gray. Empty blue circles correspond to the additional Oct1/Sox2 interface contacts seen in 1o4x (Oct1/Sox2/DNA) excluding the criticalK95(sox2). (B): Sequences of helix3 of the wild-type and mutated Sox proteins under study. Mutated amino acids that deviate from the wild-type are depicted in red. The position of the highlighted sequences with the structure is indicated with a dotted line. Structural model preparedwith pymol [27] by using the structural coordinates for Sox17 [16] and the Sox2/Oct1 on DNA [26]. The van der Waals surface of the DNAderived from the Sox2/Oct1 structure is shown in light gray. Sox17 (blue) was superimposed onto Sox2 (gray). Oct1 is shown in black. The glu-tamate (Sox17) and the lysine (Sox2) mutated are shown as ball-and-sticks. The position of other helix-3 residues that differ between Sox2 andSox17 are indicated. (C): Point mutations at the Sox/Oct4 interface swap the differential assembly behavior of Sox2 and Sox17 on the canonicalversus the compressed element. Indicated Sox proteins were incubated with DNA individually and in combination with Oct4. Positions of binarySox/DNA and Oct4/DNA as well as of ternary Sox/Oct4/DNA complexes are indicated.


www.StemCells.com

substantially weakened as illustrated by an incomplete super-shift of the Sox-monomers (compare Fig. 4C lanes 4, 8, and10). However, when Lys95 was kept and only the other sevenamino acids of the Oct interaction region were mutated a wildtype-like binding was observed (Fig. 4C, lane 4 and 12). Con-versely, Sox17EK and Sox17H3-2, cooperated more stronglywith Oct4 than wild type (WT) Sox17 on the canonical ele-ment, whereas Sox17H3-2exE cooperates as weakly as the WTSox17 protein (lanes 6, 12, 16, and 18). Consistently, assess-ment of binding of the mutated Sox constructs to Oct4 on thecompressed element showed the reverse pattern. Mutating theSox2HMG to Sox2KE or Sox2H3-17 installed the ability ofthis construct to assemble with Oct4 that was denied to thewild-type domain or Sox2H3-17exK (lanes 22, 26, 28, 30).Conversely, the Sox17EK and Sox17H3-2 mutations elimi-nated a key feature necessary for coassembling with Oct4 onthe compressed element, whereas Sox17H3-2exE coboundwith Oct4 in a wild-type manner (lanes 24, 32, 34, 36). Theseresults establish that a single amino acid at the Oct4 contactinterface swaps the binding preferences of the Sox2/Oct4 ver-sus the Sox17/Oct4 transcription factor pairs on motif variants.

So far, only the impact of the spacing and orientationbetween idealized sox/oct motifs has been assessed. However,the bipartite POU domains of Oct proteins have been shown toadopt pronounced structural rearrangements on different bind-ing sites [28, 29] and subtle sequence variations within the soxand oct sites, as well as in the periphery, may also affect dime-rization of Sox and Oct proteins. To test whether Sox2 andSox17 also exhibit qualitative binding differences with Oct4 onnonideal elements, we conducted differential assembly experi-ments on sox/oct motifs known to recruit Sox2 and Oct4 invivo [8, 26, 30–32]. All tested motifs exhibit a canonicalarrangement of sox and oct subsites but vary within the recog-nition sequences and the periphery (Fig. 5A). Indeed, we foundthat the overall efficiency of dimer formation differs betweenthe tested elements (Fig. 5B). For example, the formation of aSox2/Oct4 dimer on the canonical Utf1 element is less efficientthan on most other elements. Nevertheless, the overall patterncorroborates the model that the canonical arrangements of sox/oct subsites favors binding of Sox2 and Oct4, whereas thecompressed versions of all tested motifs abrogates Sox2/Oct4assembly and Sox17/Oct4 bind in a highly cooperative manner.

Next, we wondered whether the rational mutagenesis ofSox2 and Sox17 also affects the binding to DNA in cells inthe context of the full-length proteins. To test this, weexpressed V5 tagged full-length Sox2, Sox2KE, Sox17, andSox17EK proteins in mouse ES cells and performed ChIPexperiments using V5 antibodies. We subsequently quantifiedthe enrichment of ChIP-ed DNA at two prominent genomicloci known to be bound by Sox2 and Oct4 and essential forthe expression of the Nanog and Pou5f1 genes in mouse EScells [16, 30]. Although Sox2 was detected at both genomicloci, Sox2KE was no longer recruited to this site (Fig. 5C).Conversely, only the Sox17EK mutation installed binding tothese sites but WT Sox17 was not. To investigate the impactof the differentially configured sox/oct motifs on gene expres-sion, we performed luciferase reporter assays in HEK293 cellswith transiently expressed Sox variants and Oct4. Sox2strongly activated luciferase expression in the presence of thecanonical motif, whereas Sox2KE and Sox17 showed a signif-icantly weaker response (Fig. 5D). Sox17EK, however, acti-vated the reporter more strongly than WT Sox2. The inversetrend was observed for the compressed motif, Sox17 activatedexpression, whereas the Sox17EK mutation had a significantlydiminished activity. In accordance with assembly experiments,Sox2 was incapable of activating expression from a com-

pressed element, but Sox2KE was now capable of activatingexpression. Together, these results concur with the biochemi-cal activities measured using purified component and indicatethat the differential recognition of the canonical and com-pressed motifs by rationally designed Sox variants and Oct4also takes place in a cellular context.

Generation of iPS cells Using a RationallyEngineered Sox17 Construct

Sox2 and Sox17 differ in their ability to generate iPS cells[15]. Although Sox2 in combination with Oct4, Klf4, and c-Myc can reprogram somatic cells, Sox17 cannot. To investi-gate the functional significance of the amino acid substitutionthat dramatically altered the biochemical properties of Sox2and Sox17, we used iPS cell generation to assess the capacityof our redesigned Sox variants to induce pluripotency.

Four cotransfected transcription factors are able to effi-ciently generate mouse iPS cells: Oct4, c-Myc, Klf4, andSox2 [24]. In our reprogramming assay, we used mouse em-bryonic fibroblasts (MEFs), isolated from C57Bl6 mice, con-taining a green fluorescence protein (GFP) under the controlof the Oct4 promoter to assess the capacity of the reengi-neered Sox factors to induce reprogramming. Oct4-GFPMEFs were infected with retroviral vectors expressing Oct4,c-Myc, and Klf4 (OCK) together with one wild-type ormutated Sox factor—Sox2, Sox2KE, Sox17, or Sox17EK. Inaccordance with previous reports [15, 24], we observed GFP-positive (GFPþ) colonies in MEFs transduced with OCK þSox2, whereas MEFs transduced with OCK þ Sox17 did notyield any (Fig. 6A). Sox17EK gave rise to GFPþ colonieswith a morphology and Oct4-GFP levels indistinguishablefrom those generated with Sox2. On the contrary, the Sox2KEmutant was unable to generate any GFPþ colonies (Fig. 6A).To compare the efficiency of Sox2 and Sox17EK in reprog-ramming experiments, we counted the total number of GFPþcolonies obtained per plate of MEFs in three independentexperiments. After 21 days of infection, an average of 78 iPSclones were induced with OCK þ Sox2, from initial 267,000transduced fibroblasts, whereas no GFPþ colonies appearedwhen Sox2 was omitted (Fig. 6B). Interestingly, when Sox2was replaced by Sox17EK, an average of 295 iPS colonieswas obtained (Fig. 6B). We verified that the quantitative dif-ferences are not due to viral titer variations (Supporting Infor-mation Figure S4). To confirm the integration of OCK þSox17EK transgenes in the iPS clones, we performed primer-specific PCRs and validated, by DNA sequencing, that theinsert indeed contained the Sox17EK mutation (Supporting In-formation Figure S5). After the formation of the iPS colonies,as expected, the transgenes were efficiently silenced (Support-ing Information Figure S6). Also, the Sox17EK-reprog-rammed cells were found to be karyotypically normal (Sup-porting Information Figure S7). Next, we studied thepluripotent nature of the iPS cells generated by Sox17EKusing a series of different experimental techniques. First, weexpanded several of the iPS colonies and measured theexpression of a set of marker genes by QRT-PCR (Fig. 6C).All the pluripotency markers tested, Eras, Nanog, Zfp206, andZic3, exhibited expression levels in OCK þ Sox17EK-derivediPS cells comparable with normal ES cells and OCK þ Sox2iPS cells, whereas the original MEFs showed very lowexpression (Fig. 6C). Next, immunostainings were carried outand the Sox17EK iPS clones showed clear signals for the plu-ripotency markers Nanog and SSEA-1, which were indistin-guishable from Sox2 iPS clones. Likewise, all iPS cloneswere positive for alkaline phosphatase (Fig. 6D). To demon-strate pluripotency in vivo, two independent Sox17EK iPSclones were injected into immunocompromised mice, and


their potential for teratoma formation was assessed. IndividualiPS clones induced by Sox2 and Sox17EK formed teratomathat were composed of tissues derived from all three germlayers: ectoderm (neural tissues), mesoderm (muscle, carti-lage), and endoderm (ciliated epithelium; Fig. 6E). For a de-finitive assessment of the pluripotentiality of Sox17EK clones,mouse blastocysts injections were carried out. Sox2 andSox17EK iPS cells, constitutively expressing the fluorescentmarker mCherry gave rise to E13.5 chimeric embryos (Fig.6F). Collectively, these data demonstrate that Sox17EK, incooperation with OCK, is able to induce reprogramming likewild-type Sox2. On the contrary, the mutated Sox2KE proteinlacks this activity.

Given the inverted efficiency of Sox17EK and Sox2KE tobiochemically cooperate with Oct4 on canonical sox/oct ele-ments, we conclude that Sox17EK acquired the iPS inducingpotential by dimerizing with Oct4 on enhancers of pluripo-tency genes, an ability lost by Sox2KE.

The Ability to Promote Endodermal DifferentiationIs Swapped Between Sox2 and Sox17

Several recent studies demonstrated that the overexpression ofSox17, but not Sox2, pushed ES cells to differentiate intoendodermal tissue [33–35]. Having established that Sox17EKis capable of inducing pluripotency, we wondered if the

Figure 5. Binding of Sox2 and Sox17 to functional sox/oct motifs in vitro and in vivo. (A): Sequences of sox/oct motifs that were shown torecruit Sox2 and Oct4 in vivo. The compressed version of the motifs was generated by deleting the central base pair and placing it at the 50 end.(B): Differential assembly of WT Sox2 and Sox17 HMG domains on actual sox/oct motifs from the indicated genes as performed for the ideal-ized sox/oct in Figure 4C. (C): Chromatin immunoprecipitation experiment using the indicated V5-tagged Sox proteins heterologously expressedin mouse ES cells as described in the Supporting Information procedures. Fold enrichments (relative to input DNA) were assessed at Sox2-bind-ing sites in the Pou5f1 and Nanog promoter regions. (D): Luciferase reporter assays using HEK-293 cells cotransfected with reporter plasmidscontaining none, canonical, or compressed DNA elements as well as indicated Sox effectors and Oct4. ‘‘E’’ denotes empty Sox effector plasmids,‘‘2’’ Sox2, ‘‘17’’ Sox17, ‘‘KE’’ Sox2KE, and ‘‘EK’’ Sox17EK.


www.StemCells.com

Figure 6. Characterization of induce pluripotent stem (iPS) clones reprogrammed by Sox17EK. (A): Combined bright-field and fluorescent photo-graphs of iPS colonies. Shown are representative clones of iPS cells derived from MEFs containing a GFP reporter driven from a minimal Oct4 pro-moter that were transduced with Oct4, cMyc, Klf4 plus Sox2, Sox2KE, Sox17, or Sox17EK. Scale bars ¼ 100 lm. (B): The numbers of iPS coloniesgenerated by wild-type and mutant version of Sox2 and Sox17 from three independent experiments are shown. The indicated versions of the Soxfactors were cotransduced with OCK into Oct4-GFP MEFs. Oct4-GFP-positive colonies appearing on each plate, performed in biological and techni-cal triplicates (average 6 SD), were counted 3 weeks postinfection. (C): Quantitative real-time PCR analysis on iPS clones generated with Sox2and Sox17EK. Gene expression levels of Nanog, Zfp206, Zic3, and Eras relative to nontransduced Oct4-GFP MEFs and mouse embryonic stem cells(E14) from three replicates are presented as average 6 SD. (D): Cells derived from two independent iPS clones obtained by OCK-Sox17EK (C5,C15) transduction and cells from one clone obtained by OCK-Sox2WT (C201) transduction were immunostained for expression of pluripotencymarkers SSEA-1 and Nanog. Oct4-GFP expression of the same cells is also shown. The rightmost column depicts AP expression of those cells.Scale bars ¼ 100 lm. (E): Teratomas derived from iPS clones generated from OCK-Sox2 and OCK-Sox17EK. iPS cells derived from Oct4-GFPMEFs that were transduced with Oct4, c-Myc, Klf4 plus Sox2 (clone C201), or Sox17EK (clone C15) were injected intramuscularly into immunodefi-cient SCID mice. Three to five weeks later teratomas were dissected, fixed, and stained. (F): Pictures of E13.5 chimeric mice generated with iPScells reprogrammed with OCK þ Sox17EK and infected with a lentiviral vector expressing constitutively the fluorescent protein mCherry. Abbrevia-tions: AP, alkaline phosphatase; GFP, green fluorescence protein; MEF, mouse embryonic fibroblasts; WT, wild type.


converse is true; that the Sox2KE mutant can induce endo-derm differentiation in ES cells in a manner similar to Sox17.To test this, we infected mouse ES cells with lentiviralvectors expressing Sox2, Sox17EK, Sox17, or Sox2KE trans-genes, coupled to a Venus selection marker for fluorescence-activated cell sorting sorting using the Venus marker, and

established cell lines stably expressing the Sox variants. Weobserved that cells expressing Sox17 and Sox2KE adopted anendodermal morphology, whereas cells expressing Sox2 andSox17EK maintained an ES cell-like phenotype (Fig. 7A). Tofurther analyze these cells, we quantified the expression ofpluripotency and differentiation markers by QRT-PCR. We

Figure 7. Characterization of mouse ES cells derived after overexpression of Sox variants. (A): Bright-field photographs of representative celllines derived from E14 mouse ES cells after infection with lentiviral vectors to overexpress Sox2, Sox2KE, Sox17, or Sox17EK. Pictures were takenfour passages after infection. Scale bars ¼ 50 lm. (B): Quantitative real-time PCR analysis were performed to determine expression level of pluri-potency markers Nanog, Zfp42, Klf4, and Oct3/4, (C) specific endoderm markers Sox7, Gata4, Gata6, and FoxA2 and (D) specific ectoderm Sox1,Nestin, and mesoderm markers, Mixl1 and T-brachyury in cells expressing Sox variants as indicated. Gene expression levels from three replicatesare presented as average 6 SD. For comparison, the expression levels of all markers in the original mouse ES cell line (E14) are also presented.(E): Expression of SSEA-1, Nanog, Gata4, and Dab-2 proteins in cells stably expressing Sox2, Sox2KE, Sox17, and Sox17EK. The cell lines wereimmunostained for expression of pluripotency markers SSEA-1, Nanog, and endoderm markers Gata4 and Dab-2. Merged images with DAPI areshown. Scale bars ¼ 50 lm. Abbreviations: DAPI, 40,6 diaminido-2-phenylindole; ES cell, embryonic stem cell.


www.StemCells.com

observed that the expression of the pluripotency markersZfp42, Nanog, Oct4, and Klf4 were reduced in Sox17 andSox2KE expressing cells, as compared with control ES cellsas well as to Sox2- and Sox17EK-expressing cells (Fig. 7B).Conversely, the endoderm-specific markers Gata4, Gata6,FoxA2, and Sox7 were strongly induced in cells expressingSox17 or Sox2KE, whereas they were virtually absent inSox2- and Sox17EK-expressing cells (Fig. 7C). The cellularphenotype was further analyzed by immunostaining revealingthat Sox2 and Sox17EK cells express the pluripotencymarkers Nanog and SSEA1, whereas Sox17 and Sox2KE cellsexpress the endoderm markers Gata4 and Dab2 (Fig. 7E). Totest if Sox2KE, like Sox17, selectively induces endodermaldifferentiation of ES cells or whether it causes a general lossof pluripotency and the formation of all three germ layers, weanalyzed the expression of mesoderm and ectoderm markersSox1, Nestin, Mixl1, and T-brachyury. Neither mesodermalnor ectodermal markers were found to be significantly upregu-lated in Sox17 and Sox2KE cells, as compared with pluripo-tent cells, suggesting that Sox2KE acquired Sox17-like prop-erties of forcing ES cells specifically toward an endodermalfate (Fig. 7D).

Together, these results indicate that strategically placedpoint mutations at protein-interaction surfaces swap the bioac-tivities of Sox2 and Sox17. The Sox2 mutant, Sox2KE, haslost the activity to maintain pluripotency and has insteadgained the ability to induce endoderm in ES cells. Conversely,mutating Sox17 to Sox17EK abolished its role as a driver to-ward endoderm but equips it with the ability to maintain EScell pluripotency.

DISCUSSION

In this study, we progress from the in silico identification of anovel compressed motif, to the biochemical demonstrationthat this DNA motif can recruit different Sox/Oct pairs incontrast to its canonical counterpart, to the reverse engineer-ing through structure-based mutagenesis of the differential as-sembly behavior. Ultimately, we demonstrate that a singlepoint mutation rationally introduced at a site that affects Oct4interaction on composite DNA motifs, with only subtly differ-ent motif spacings in vitro, drastically changes gene expres-sion programs and swaps the activities of Sox17 and Sox2 inbiological assays. This mutation gives rise to a fundamentalchange in the developmental outcomes triggered by the Soxproteins. Forced expression of wild-type Sox2 and Sox17EKhave no effect on ES cells, but expression of wild-type Sox17and Sox2KE in ES cells causes differentiation toward anendoderm fate. Sox17EK can now potently reprogram MEFsinto iPS cells, whereas the wild-type Sox17 is incapable ofthis feat. These data suggest that Sox17EK has gained theability to specify pluripotency, but lost its endoderm promot-ing potential, whereas Sox2KE has lost its reprogrammingbehavior and gained the ability to specify an endoderm phe-notype. To our knowledge, this is the first example of therational conversion of a protein into a reprogramming factorby rational mutagenesis. That these developmental phenotypesare reflected in the binding capability of Sox/Oct4 to variantDNA motifs demonstrates the crucial importance of the Sox2/Oct4 dimerization in specifying pluripotency.

It is surprising that a single amino acid change causessuch a dramatic functional swap. Although both, B1 and Fgroup Sox proteins contain a C-terminal transactivation do-main, actual sequence conservation outside the HMG domainis poor [36]. Nevertheless, both, Sox2 and Sox9 can recruitp300 to exert transcriptional control despite highly divergent

sequences within their transactivation domains [37, 38].Therefore, the binding to generic coactivators, such as p300,does not implicitly depend on conserved domains within theperipheral regions and Sox2 and Sox17 may be equally capa-ble to recruit this cofactor. Hence, the qualitatively distinctdevelopmental roles of Sox2 and Sox17 do not appear to bedetermined by their transactivation domains but by their abil-ity to team up with Oct4 on specific cis-regulatory modules(CRMs). Indeed, the Sox17EK mutation installs the activityto assemble at selected genomic loci that are known to beconstructively targeted by Sox2 and Oct4. On the contrary,Sox2KE has lost this activity. Further studies are required toassess the redistribution of Sox17EK and Sox2KE binding incomparison with WT proteins on a genomic scale.

It is possible that gene sets specifying a particular devel-opmental lineage are earmarked by distinct sox/oct motif con-figurations. Indeed, some evidence is consistent with the hy-pothesis that different configurations of sox/oct motifs recruitparticular combinations of members of Sox and POU family[9, 10, 39–41]. Whether or not the compressed sox/oct motifor a derivative thereof comprises a core component of, forexample, endodermal CRMs, remains to be investigated.

We believe that this work provides general insights intoTF function and could guide further studies on the combinato-rial control of gene expression. Gene regulation requires mul-tiple inputs involving the assembly of combinations of tran-scription factors on CRMs and deciphering some sort of‘‘regulatory code’’ is intensely sought after. It is unclearwhether a precise arrangement of TF-binding sites is neces-sary to execute a regulatory event [42]. By showing that theprotein contact interface encodes the potential to cobind dif-ferently spaced motif variants, we propose a biochemical ra-tionale for constrained motif configurations that might applyto other transcription factor combinations. We, furthermore,demonstrate that altering the differential assembly potentialhas pronounced functional consequences. Together, this sug-gests that individual sequence specificities and the ability torecruit the downstream regulatory machinery are conservedfeatures of Sox family TFs (and perhaps others), whereas thepotential to differentially assemble on distinct composite cis-regulatory elements determines their distinct biological roles.

Future work should systematically address the questionwhether distinct motif configurations indeed earmark genes spe-cific for a particular biological process and recruit specific TFcombinations. The variety of functionally important Sox/POUpairs provides a suitable model system to test this possibility.

ACKNOWLEDGMENTS

We thank Rory Johnson and Shyam Prabhakar for valuable com-ments on the article and Selina Poon Kwee Lan, Choo Siew Hua,Kee Yew Wong, and Siaw-Wei Teng for technical support. Weare grateful to Petra Kraus and V Sivakamasundari from theGIS-GAP for cell injections, teratoma removals, and histology.This work was supported by the Agency for Science, Technol-ogy and Research (A*STAR; www.a-star.edu.sg) Singapore.A.P.H. is currently affiliated with the Immunology FrontierResearch Centre, Osaka University, Suita, Osaka, Japan.

DISCLOSURE OF POTENTIAL CONFLICTS

OF INTEREST

The authors indicate no potential conflicts of interest.


REFERENCES

1 Bowles J, Schepers G, Koopman P. Phylogeny of the SOX family ofdevelopmental transcription factors based on sequence and structuralindicators. Dev Biol 2000;227:239–255.

2 Ryan AK, Rosenfeld MG. POU domain family values: Flexibility,partnerships, and developmental codes. Genes Dev 1997;11:1207–1225.

3 Wegner M. From head to toes: The multiple facets of Sox proteins.Nucleic Acids Res 1999;27:1409–1420.

4 Badis G, Berger MF, Philippakis AA et al. Diversity and complexityin DNA recognition by transcription factors. Science 2009;324:1720–1723.

5 Ravasi T, Suzuki H, Cannistraci CV et al. An atlas of combinatorialtranscriptional regulation in mouse and man. Cell 2010;140:744–752.

6 Boyer LA, Lee TI, Cole MF et al. Core transcriptional regulatory cir-cuitry in human embryonic stem cells. Cell 2005;122:947–956.

7 Loh YH, Wu Q, Chew JL et al. The Oct4 and Nanog transcriptionnetwork regulates pluripotency in mouse embryonic stem cells. NatGenet 2006;38:431–440.

8 Rodda DJ, Chew JL, Lim LH et al. Transcriptional regulation ofnanog by OCT4 and SOX2. J Biol Chem 2005;280:24731–24737.

9 Tanaka S, Kamachi Y, Tanouchi A et al. Interplay of SOX and POUfactors in regulation of the Nestin gene in neural primordial cells. MolCell Biol 2004;24:8834–8846.

10 Kuhlbrodt K, Herbarth B, Sock E et al. Cooperative function of POUproteins and SOX proteins in glial cells. J Biol Chem 1998;273:16050–16057.

11 Stefanovic S, Abboud N, Desilets S et al. Interplay of Oct4 with Sox2and Sox17: A molecular switch from stem cell pluripotency to speci-fying a cardiac fate. J Cell Biol 2009;186:665–673.

12 Avilion AA, Nicolis SK, Pevny LH et al. Multipotent cell lineages inearly mouse development depend on SOX2 function. Genes Dev2003;17:126–140.

13 Kanai-Azuma M, Kanai Y, Gad JM et al. Depletion of definitive gutendoderm in Sox17-null mutant mice. Development 2002;129:2367–2379.

14 Niakan KK, Ji H, Maehr R et al. Sox17 promotes differentiation inmouse embryonic stem cells by directly regulating extraembryonicgene expression and indirectly antagonizing self-renewal. Genes Dev2010;24:312–326.

15 Nakagawa M, Koyanagi M, Tanabe K et al. Generation of inducedpluripotent stem cells without Myc from mouse and human fibroblasts.Nat Biotechnol 2008;26:101–106.

16 Palasingam P, Jauch R, Ng CK et al. The structure of Sox17 bound toDNA reveals a conserved bending topology but selective protein inter-action platforms. J Mol Biol 2009;388:619–630.

17 Guo G, Huss M, Tong GQ et al. Resolution of cell fate decisionsrevealed by single-cell gene expression analysis from zygote to blasto-cyst. Dev Cell 2010;18:675–685.

18 Nichols J, Zevnik B, Anastassiadis K et al. Formation of pluripotentstem cells in the mammalian embryo depends on the POU transcrip-tion factor Oct4. Cell 1998;95:379–391.

19 Kurimoto K, Yabuta Y, Ohinata Y et al. An improved single-cellcDNA amplification method for efficient high-density oligonucleotidemicroarray analysis. Nucleic Acids Res 2006;34:e42.

20 Reim G, Mizoguchi T, Stainier DY et al. The POU domain proteinspg (pou2/Oct4) is essential for endoderm formation in cooperationwith the HMG domain protein casanova. Dev Cell 2004;6:91–101.

21 Chen X, Xu H, Yuan P et al. Integration of external signaling path-ways with the core transcriptional network in embryonic stem cells.Cell 2008;133:1106–1117.

22 Ng CK, Palasingam P, Venkatachalam R et al. Purification, crystalli-zation and preliminary X-ray diffraction analysis of the HMG domain

of Sox17 in complex with DNA. Acta Crystallogr Sect F Struct BiolCryst Commun 2008;64:1184–1187.

23 Jauch R, Ng CK, Saikatendu KS et al. Crystal structure and DNAbinding of the homeodomain of the stem cell transcription factorNanog. J Mol Biol 2008;376:758–770.

24 Takahashi K, Yamanaka S. Induction of pluripotent stem cells frommouse embryonic and adult fibroblast cultures by defined factors. Cell2006;126:663–676.

25 Remenyi A, Lins K, Nissen LJ et al. Crystal structure of a POU/HMG/DNA ternary complex suggests differential assembly of Oct4and Sox2 on two enhancers. Genes Dev 2003;17:2048–2059.

26 Williams DC, Jr, Cai M, Clore GM. Molecular basis for synergistictranscriptional activation by Oct1 and Sox2 revealed from the solutionstructure of the 42-kDa Oct1.Sox2.Hoxb1-DNA ternary transcriptionfactor complex. J Biol Chem 2004;279:1449–1457.

27 DeLano WL. The PyMOL Molecular Graphics System. San Carlos,CA, USA: DeLano Scientific, 2002.

28 Remenyi A, Tomilin A, Pohl E et al. Differential dimer activities ofthe transcription factor Oct-1 by DNA-induced interface swapping.Mol Cell 2001;8:569–580.

29 Tomilin A, Remenyi A, Lins K et al. Synergism with the coactivatorOBF-1 (OCA-B, BOB-1) is mediated by a specific Pou dimer configu-ration. Cell 2000;103:853–864.

30 Chew JL, Loh YH, Zhang W et al. Reciprocal transcriptional regula-tion of Pou5f1 and Sox2 via the Oct4/Sox2 complex in embryonicstem cells. Mol Cell Biol 2005;25:6031–6046.

31 Nishimoto M, Fukushima A, Okuda A et al. The gene for the embry-onic stem cell coactivator UTF1 carries a regulatory element whichselectively interacts with a complex composed of Oct-3/4 and Sox-2.Mol Cell Biol 1999;19:5453–5465.

32 Tomioka M, Nishimoto M, Miyagi S et al. Identification of Sox-2 reg-ulatory region which is under the control of Oct-3/4-Sox-2 complex.Nucleic Acids Res 2002;30:3202–3213.

33 Qu XB, Pan J, Zhang C et al. Sox17 facilitates the differentiation ofmouse embryonic stem cells into primitive and definitive endoderm invitro. Dev Growth Differ 2008;50:585–593.

34 Seguin CA, Draper JS, Nagy A et al. Establishment of endoderm pro-genitors by SOX transcription factor expression in human embryonicstem cells. Cell Stem Cell 2008;3:182–195.

35 Shimoda M, Kanai-Azuma M, Hara K et al. Sox17 plays a substantialrole in late-stage differentiation of the extraembryonic endoderm invitro. J Cell Sci 2007;120:3859–3869.

36 Lefebvre V, Dumitriu B, Penzo-Mendez A et al. Control of cellfate and differentiation by Sry-related high-mobility-group box(Sox) transcription factors. Int J Biochem Cell Biol 2007;39:2195–2214.

37 Nowling TK, Johnson LR, Wiebe MS et al. Identification of the trans-activation domain of the transcription factor Sox-2 and an associatedco-activator. J Biol Chem 2000;275:3810–3818.

38 Tsuda M, Takahashi S, Takahashi Y et al. Transcriptional co-activa-tors CREB-binding protein and p300 regulate chondrocyte-specificgene expression via association with Sox9. J Biol Chem 2003;278:27224–27229.

39 Ambrosetti DC, Basilico C, Dailey L. Synergistic activation of thefibroblast growth factor 4 enhancer by Sox2 and Oct-3 dependson protein-protein interactions facilitated by a specific spatialarrangement of factor binding sites. Mol Cell Biol 1997;17:6321–6329.

40 Dailey L, Basilico C. Coevolution of HMG domains and homeodo-mains and the generation of transcriptional regulation by Sox/POUcomplexes. J Cell Physiol 2001;186:315–328.

41 Yuan H, Corbi N, Basilico C et al. Developmental-specific activity ofthe FGF-4 enhancer requires the synergistic action of Sox2 and Oct-3.Genes Dev 1995;9:2635–2645.

42 Segal E, Widom J. What controls nucleosome positions? Trends Genet2009;25:335–343.

See www.StemCells.com for supporting information available online.


Conversion of Sox17 into a Pluripotency Reprogramming Factor by ...

Documents