Changing POU dimerization preferences converts Oct6 into a pluripotency inducer

Abstract

The transcription factor Oct4 is a core component of molecular cocktails inducing pluripotent stem cells (iPSCs), while other members of the POU family cannot replace Oct4 with comparable efficiency. Rather, group III POU factors such as Oct6 induce neural lineages. Here, we sought to identify molecular features determining the differential DNA-binding and reprogramming activity of Oct4 and Oct6. In enhancers of pluripotency genes, Oct4 cooperates with Sox2 on heterodimeric SoxOct elements. By re-analyzing ChIP-Seq data and performing dimerization assays, we found that Oct6 homodimerizes on palindromic OctOct more cooperatively and more stably than Oct4. Using structural and biochemical analyses, we identified a single amino acid directing binding to the respective DNA elements. A change in this amino acid decreases the ability of Oct4 to generate iPSCs, while the reverse mutation in Oct6 does not augment its reprogramming activity. Yet, with two additional amino acid exchanges, Oct6 acquires the ability to generate iPSCs and maintain pluripotency. Together, we demonstrate that cell type-specific POU factor function is determined by select residues that affect DNA-dependent dimerization.

Keywords: DNA binding; Oct4; POU factors; reprogramming to pluripotency.

Figure 1. Oct transcription factors exhibit differential preferences for composite DNA‐binding sites

1. A

   Position weight matrices obtained using de novo motif discovery (HOMER) and ChIP‐Seq summits of Oct4 in mouse ESCs 56, Brn2 in MEFs 48 h after transfections, and Brn2 in mouse NPCs 23.

2. B, C

   Molecular models based on the crystal structures previously published 44 represent configurations of two dimers on DNA: Sox2–Oct4 heterodimer (B) and Oct6–Oct6 homodimer (C). Oct4 POU domain, green; Oct4/Oct6 linker region, magenta; Sox2 HMG, cyan; Oct6 POU domain, orange. Individual helixes of POU domains are numbered. In both models, the DNA is represented as a light gray surface. Unless indicated otherwise, the representation is kept throughout the manuscript.

3. D

   EMSAs with a twofold dilutions series starting at 2 μM of the Oct4 POU (left) or the Oct6 POU (right) in the presence of Cy5‐labeled MORE DNA. The free DNA and DNA bound as monomer or dimer are indicated.

4. E, F

   Bar plots of EMSA‐derived cooperativity factors for Oct–Oct homodimers on the MORE (E) or Sox2–Oct heterodimers on the SoxOct element (F) are shown for a panel of six POU proteins. The mean is shown with standard deviation as error bars (n ≥ 3), and Tukey's multiple comparison of means was used for assessment of statistical significances (***P < 0.001). N.A.: not assessed.

5. G

   Properties of Oct4 and Oct6 homodimers from molecular dynamics (MD) simulations. The structural snapshot on the left illustrates Oct4 (first monomer in green, second in dark blue) and the three parameters used: drmsd = the root‐mean‐square deviation of interatomic distances measured for the backbone of the structured domains (excluding the linker); θ = the angle describing the kink in helix α₄ (data for both monomers 1 and 2 are shown); d = the distance between the centers of mass of the two helices α₅; w = the DNA minor groove width in the region between the two monomers.

Source data are available online for this figure.

Figure EV1. The POU factor Oct4 exhibits a unique DNA‐binding preference

1. To count the relative abundance of the OctOct (MORE) versus SoxOct motif, the presence of position weight matrices obtained was analyzed using HOMER (lower panel) 39. References for ChIP‐Seq data sets are in order (left to right): Oct2 in B cells 39, Brn2 after 48 h of its overexpression in MEFs 54, Brn2 in NPCs 54, Brn2 in NPCs 55, and Oct4 in ESCs 72. The upper panel was generated using word searches with IUPAC strings instead of pwms on the same set of ChIPseq peaks as in the lower panel.

2. Representative EMSAs used for omega calculations showing the formation of Oct–Oct homodimers on the MORE DNA element for a panel of seven POU proteins. Complexes were separated on native gels and subsequently imaged. Note, in Fig 1E and F, only data for Oct1 but not the closely related Oct11 are shown.

3. Representative EMSAs performed as single‐tube reactions containing the Sox2 HMG and POU domains of Oct transcription factors, in combination with Cy5‐labeled SoxOct and FAM‐labeled MORE DNA elements. Complexes were separated on native gels and sequentially imaged using FAM and Cy5 channels.

4. Bar plots representing differences in DNA‐binding preference as determined by single‐tube EMSA reactions for a panel of six POU proteins (classes I‐VI) represented as log2 ratios of hetero‐ and homodimer band intensities on Cy5‐OctSox and FAM‐MORE DNA elements, respectively. Individual data points are shown as gray jitter plot. The mean is shown with standard deviation as error bars (n = 9), and Tukey's multiple comparison of means was performed to assess statistical significance (***P < 0.001).

Source data are available online for this figure.

Figure EV2. Oct6 forms a long‐lived homodimer on MORE DNA

1. A–D

   Off‐rate EMSA using the mouse Oct6 POU (A, C) and the mouse Oct4 POU (B, D) along with MORE (A, B) or SoxOct (C, D) DNA elements, respectively. Dissociation was initiated after incubation of proteins with Cy5‐labeled DNA for 1 h by the addition of a 200‐fold excess of unlabeled competitor DNA of the same sequences as the Cy5‐labeled reporter DNA. Reactions were prepared in single tubes and directly loaded onto running gels after indicated time points. Concentrations of the reactants and sequences of the forward DNA strand are mentioned.

Figure 2. A single amino acid residue guides DNA‐binding preferences of Oct4 and Oct6 POU domains

1. A

   Zoom‐in view of the Oct6–Oct6 homodimer interface in the molecular models based on crystal structures of two POU homodimers on MORE DNA. Oct6 POU_S, orange; POU_HD of the other monomer, blue. Important residues are labeled and Met (Oct6) in position 151 that was identified to guide the DNA‐binding preference is marked with a star. Individual helixes of POU domains are numbered. Color representation as in Fig 1B and C. For clarity, the second monomer of Oct4/Oct6 is shown in dark blue.

2. B, C

   Network of hydrophobic interactions in Oct6 (B) and Oct4 (C). The numbers show the minimal interatomic distances (excluding hydrogens) between the residues involved calculated from the simulations. In blue are those distances that change significantly between Oct4 and Oct6 and in red those unchanged. The shown standard deviations represent a measure for the variability during the simulations and not for the uncertainty in the calculations.

3. D

   EMSA showing a differential binding of WT Oct4 and Oct6 POU domains as well as mutated Oct4^151M and Oct6^151S POU domains on MORE (OctOct) element. The free DNA and DNA bound by monomer or dimer are indicated.

4. E

   EMSA showing binding of WT Oct4 and Oct6 POU domains as well as mutated Oct4^151M and Oct6^151S POU domains to SoxOct element in the absence or presence of the Sox2 HMG. The free DNA and DNA bound by monomers or dimer are indicated. Quantifications of cooperativity measurements are shown in Fig EV3C.

5. F

   Bar plot based on single‐tube EMSAs using Cy5‐SoxOct and FAM‐MORE DNA elements showing the difference in DNA binding of WT Oct4, WT Oct6, Oct4^151M, and Oct6^151S POU domains. The mean log2 ratios of Cy5 (heterodimer) and FAM (homodimer) band intensities are depicted with standard deviation (n = 9). Individual data points are shown as gray jitter plots. Tukey's multiple comparison of means was performed to assess significance (***P < 0.001, n.s., P > 0.05). Compare to Figs EV1D and EV3D.

6. G

   A scheme illustrating the DNA‐binding preferences of Oct POU domains to MORE and SoxOct elements. On the left side, WT Oct4 (green) preferentially heterodimerizes with Sox2 HMG (cyan) over homodimerizing on MORE. In the second scenario, WT Oct6 (orange) preferentially forms homodimers on MORE DNA rather than heterodimers with Sox2 on SoxOct DNA elements. The thickness of the arrows illustrates the relative abundance of the microstates under equilibrium conditions. The binding behavior is swapped for the engineered Oct6^151S and Oct4^151M proteins.

Source data are available online for this figure.

Figure EV3. An exchange of Met and Ser residues in Oct4 and Oct6 POU domains is responsible for different DNA‐binding preferences

1. Sequence alignment of POU domains of Oct family TFs. POU_S, the linker region, and POU_HD, as well as individual helixes, are labeled. Notably, this structure‐based alignment differs slightly from the one used previously 44, which is explained by a low conservation of residues in the N‐terminal part of the POU linker. The alignment was prepared in T‐Coffee software and colors distinguish conservation and amino acid residue types. A protein surface interacting with the MORE DNA (as shown in 17) is highlighted. Each of three studied residues (B) is marked by a black asterisk.

2. An overview of tested Oct4 and Oct6 mutants with amino acid exchanges in their POU_HD. The EMSA below shows the low cooperativity on MORE DNA element caused by the mutations. As the single mutation of the 151 site had the same effect as double and triple mutations, the single mutation was selected for further study.

3. Bar plots of EMSA‐derived cooperativity factors show different preferences between formation of Oct–Oct homodimers on MORE (left) and Sox2–Oct heterodimers on the SoxOct element (right) for WT and mutated forms of Oct4 and Oct6 POU domains. Cooperativity values for WT Oct4 and Oct6 correspond to the values shown in Fig 1E and F and are shown again for comparison with the mutant proteins. The mean is shown with standard deviation as error bars (n = 3–13), and Tukey's multiple comparison of means was performed to assess statistical significance (***P < 0.001).

4. Representative EMSAs performed in single‐tube reactions containing the Sox2 HMG and Oct4^151M and Oct6^151S POU domains of Oct transcription factors. Related to the quantification in Fig 2D.

Figure 3. The residue guiding DNA‐binding preference is a key Oct4 determinant in pluripotent cell generation

1. A, B

   Molecular models show the position of mutated residues in the structures of two dimers on DNA: Sox2–Oct4 heterodimer (A) and Oct4–Oct4 homodimer (B). Ser in position 151 (asterisk) was mutated to Met in order to shift the preference of Oct4 for homodimerization. The individual helixes of the POU domains are numbered.

2. C

   On the left side, a schematic overview of Oct4 and its mutants used for iPSC generation. POU_S and POU_HD are shown as green bars connected by the linker. Sites mutated to the respective Oct6 residues and Oct6 linker are denoted in orange, and the mutation in the Sox–Oct interface is in black. The efficiency of these constructs for iPSC generation from MEFs is depicted as the absolute number of GFP‐positive colonies on the right. Error bars represent standard deviations of biological replicates (n = 3), and differences between selected samples were compared using ANOVA (P‐values: Oct4^WTxOct4^7D,22K P = 6.9e‐3; Oct4^WTxOct4^LinkO6 P = 7.3e‐4; Oct4^WTxOct4^151M P = 1.2e‐2; Oct4^151MxOct4^LinkO6,151M P = 1.1e‐2; Oct4^LinkO6,151M xOct4^{7D,22K,LinkO6,151M} P = 7.1e‐3) (***P < 0.001, **P < 0.01, *P < 0.05).

3. D

   GFP‐positive colonies of mouse iPSCs generated by Oct4 mutants in combination with Sox2, Klf4, and c‐Myc. Colonies were imaged 16 days after second viral infection, using a fluorescence microscope. Scale bars: 250 μm; 10× objective.

Source data are available online for this figure.

Figure 4. The synthetic Oct6 molecule contributes to epigenetic reprogramming of mouse embryonic fibroblasts

1. On the left side, a schematic overview of Oct6 and its mutants used for iPSC generation. POU_S and POU_HD are shown as orange bars connected by the linker. Sites mutated to the respective Oct4 residues and Oct4 linker are denoted in green, and the mutation in the Sox–Oct interface is in black. The efficiency of these constructs for iPSC generation from MEFs is depicted as the absolute number of GFP‐positive colonies on the right. Error bars represent standard deviations of two biological replicates run in parallel.

2. Oct4‐GFP‐positive colonies of two O6SKM iPSC lines, expanded from single colonies of pluripotent cells that were generated by the contribution of the synthetic Oct6 molecule (Oct6^{7K,22T,LinkO4,151S}). Scale bars: 250 μm.

3. Genotyping of two O6SKM iPSC lines. Two stable lines of iPSCs generated with the Oct6, Sox2, Klf4, and c‐Myc retroviral set are positive for these four transgenes, but negative for the Oct4 transgene. OSKM iPSCs and OG2‐MEFs were used as PCR controls.

4. Expression analysis of viral transgenes in two O6SKM iPSC lines done by qRT–PCR using specific primers. OG2‐MEFs were harvested 5 days after infection and used for comparison as a positive control.

5. Two O6SKM iPSC lines immunostained for pluripotency markers Sox2 and Nanog. Oct4 expression was confirmed by the Oct4‐GFP transgene. DNA stained by DAPI. Scale bars: 100 μm.

6. Bisulfite sequencing of genomic Oct4, Nanog, and Col1a1 promoter regions of O6SKM iPSCs, OSKM iPSCs, and OG2‐MEFs. White and black circles represent unmethylated and methylated CpG sites, respectively.

7. Pairwise scatter plots comparing the global gene expression profile of O6SKM iPSCs1 with OG2 ESCs (left) and O4SK iPSCs (right). Black lines represent a twofold change in gene expression levels between the paired cell lines. On the right side of the plots, the color bar indicates scattering density. Red and green dots represent up‐ and downregulated genes, respectively. Positions of selected pluripotency‐related genes are highlighted as orange points.

8. Hierarchical clustering of O6SKM iPSC lines with OG2 ESCs and O4SK iPSCs, illustrating the close relationship between their global gene expression profiles.

Source data are available online for this figure.

Figure EV4. O6SKM‐induced pluripotent cells have normal karyotype and global gene expression profile, similar to embryonic stem cells

1. The relative transcript levels of Oct6 and its mutants were analyzed by qRT–PCR. The mean values for biological replicates using viral supernatants from same reprogramming experiments are shown with the standard deviation as error bars (n = 2).

2. Oct4‐GFP‐positive colonies of mouse iPSCs generated by Oct6 mutants in combination with Sox2, Klf4, and c‐Myc. Colonies were imaged 16 days after viral infection, using a fluorescence microscope. Scale bars: 250 μm.

3. Karyotyping of two stable mouse O6SKM iPSC lines revealed that cells from both lines carry the correct number of chromosomes—40. Scale bars: 20 μm.

4. Pairwise scatter plots comparing global gene expression profiles of O6SKM iPSC lines with OG2 ESCs, O4SK iPSCs, and previously published O4SKM iPSCs and MEFs. Black lines represent a twofold change in gene expression levels between the paired cell lines. On the right side of the plots, color bar indicates scattering density. Red and green dots represent up‐ and downregulated genes, respectively. Positions of selected pluripotency‐related genes are highlighted as orange points. Principal component analysis highlighting the close relationship between global gene expression profiles of O6SKM iPSCs (green circles), OG2 ESCs (red circles), and O4SK iPSCs (blue circles) is shown in the upper right corner.

Figure 5. Cells reprogrammed using the synthetic Oct6 molecule show pluripotency in vitro and in vivo

1. In vitro differentiation of O6SKM iPSC lines into cells of all three germ layers as shown by immunochemistry: endoderm (α‐fetoprotein, AFP), mesoderm (α‐smooth muscle actin, SMA), and ectoderm (β‐tubulin, TUJ1). Nuclei (DNA) were stained by Hoechst (blue). Scale bars: 200 μm.

2. F1 offspring of chimera male mice with contribution from O6SKM iPSCs and CD1 female mice.

3. Pictures demonstrating germline transmission; endogenous Oct4‐GFP signal was detected in the gonads of F1 pups. Scale bars: 200 μm.

Figure EV5. Cells reprogrammed using the synthetic Oct6 molecule show pluripotency in vivo

1. Sections of teratomas stained 4 weeks after subcutaneous injection of nude mice with iPSCs generated using O6SKM factors. Teratomas contain all three embryonic germ layers: endoderm (epithelium, e), mesoderm (muscle, m), and ectoderm (keratin, k, and neural epithelium with rosettes, n). Scale bars: 100 μm.

2. Oct4‐GFP‐positive germ cells were detected in the male fetal gonads of E13.5 embryos, confirming the germline contribution of O6SKM iPSCs. Scale bars: 100 μm.

3. Oct4‐GFP‐positive germ cells were detected in the gonads of E19.5 pups. Scale bars: 250 μm.

4. Representative example of two chimeric mice generated by O6SKM iPSCs. The agouti coat color originated from O6SKM iPSCs. These two chimeras represent mice numbers 13 and 14 (see genotyping results below).

5. PCR analysis of DNA isolated from the tails of chimeric mice was performed in order to demonstrate germline contribution by the presence of Oct4‐GFP (OG2) and Oct6 viral transgenes. Contamination by Oct4 virus was also excluded by PCR analysis; M stands for DNA marker, ctrl (+) refers to positive iPSC control, and ctrl (−) refers to negative MEF control. Samples 1–14 refer to O6SKM iPSCs2; samples 15–20 refer to O6SKM iPSCs1. Samples labeled with “X” are not relevant to this study.

6. Genotyping of Oct4‐GFP‐positive pups after germline transmission confirmed the presence of Oct6 viral transgene. M stands for DNA marker. Two GFP‐negative mice were used as negative controls (ctrl1 and ctrl2).

7. The Oct4 complementation assay using ZHBTc4 embryonic stem cells was performed as was described previously 68. Oct4 and engineered Oct6 transgenes were used to rescue the ES cells upon loss of endogenous Oct4 after the addition of doxycycline (Tc, tetracycline antibiotics) and thus maintain pluripotency. Scale bars: 250 μm.

8. The relative transcript levels of selected pluripotency markers and also of relative lentiviral levels of WT Oct4 and engineered Oct6 were analyzed by qRT–PCR. The mean values are shown with the standard deviation as error bars (n = 2).