The experiments were performed using the mouse ESC line W4 (provided by the Rockefeller University Core Facility), the YPet-OCT4 and YPet-SOX2 ESC lines previously generated in our laboratory from the W4 cell line (Verneri et al., 2020) and the eGFP-NANOG cell line that was generated in this work. ESCs were maintained on 0.1% gelatin-coated dishes, passed every 3 days using trypsin-EDTA (Gibco) and grown at 37°C in a 5% CO₂ (v/v) incubator. Cells were cultured in DMEM (Gibco) supplemented with 15% ESC qualified fetal bovine serum (Gibco), 100 mM MEM nonessential amino acids (Gibco), 2 mM l-alanyl-L-glutamine (Gibco), 0.5 mM 2-mercaptoethanol, 100 U/mL penicillin (Gibco), 100 mg/mL streptomycin (Gibco), leukemia inhibitory factor (LIF) and 2i [1 μM PD0325901 (Tocris) and 3 μM CHIR99021 (Tocris)].

To induce non-directed differentiation, ESCs were incubated in the absence of LIF and 2i during 4 h for induced early-S phase cells measurements and for 24 or 48 h for eGFP-NANOG stable cell line validation.

For microscopy measurements, 18-mm round coverslips were placed into the wells of a 12 multiwell plate, incubated for 1 h with a 100 μg/mL poly-D-Lysine (Sigma) and for 2 h with a solution 20 μg/mL laminin (Thermo Fisher) at 37°C in a 5% CO₂ incubator. Next, ∼75,000 cells were added in each well and incubated with culture medium.

Transient transfection of cells with PCNA-RFP and/or HP1α-eGFP vectors was performed using Lipofectamine 2000 (Thermo Fisher) and 1.6 µg of plasmid DNA in Opti-MEM medium (Thermo Fisher) after culturing the cells for 24 h. The transfection medium was replaced after 6 h with fresh culture medium. PCNA-RFP coding vector (Sporbert et al., 2005) was kindly provided by Dr. Cristina Cardoso. Microscopy experiments were run after 48 h from transfection.

For the induction of YPet-OCT4 and YPet-SOX2 expression, the corresponding ESCs were incubated with 5 μg/mL doxycycline (Dox) for 48 h prior to microscopy measurements (Verneri et al., 2020).

eGFP-NANOG coding sequence, kindly provided by Dr. Nicolas Platcha, was subcloned in a PiggyBac transposon vector generating the PB-eGFP-NANOG plasmid. Cells were plated during 24 h onto a 0.1% gelatin-coated dish and co-transfected, using Lipofectamine 2000 (Thermo Fisher) as described above, with PB-eGFP-NANOG and the corresponding transposase coding vector in a 3:1 relation. After 72 h of expression, puromycin (1 μg/μL) was added. eGFP-NANOG expression was verified by fluorescence microscopy after 72 h from selection.

Clone isolation was performed manually, picking the colonies and re-plating them separately in a 24-well plate. After amplification of each clone, validation was performed by analysis of the cell cycle and the expression of pluripotency and differentiation markers by RT-qPCR. Finally, the selected clone gave rise to the ESC line referred to as eGFP-NANOG.

Cell cycle was analyzed by flow cytometry as previously described (Waisman et al., 2017). Briefly, single cell suspensions were fixed in 70% ethanol, rehydrated in PBS, stained with 25 μg/mL Propidium Iodide (Sigma) and incubated for 30 min. Then, samples were analyzed in a FACS Aria II flow cytometer (BD Biosciences). Data was compiled using Floreada software.

Cells expressing the proliferating cell nuclear antigen (PCNA) fused to the red fluorescent protein (PCNA-RFP) were classified according to their PCNA nuclear distribution similarly to previous works in the field (Leonhardt et al., 2000; Sporbert et al., 2002; Pomerening et al., 2008; Leung et al., 2011; Barr et al., 2016; Wilson et al., 2016; Dutta et al., 2019; Velasquez et al., 2019; Velasquez et al., 2022; Xie and Bankaitis, 2022). Specifically, cells in early-S (E-S) phase present multiple small foci that distribute homogeneously within the nuclear space; mid-S (M-S) cells present foci close to the nucleoli or at the nuclear periphery; late-S (L-S) phase have fewer and bigger foci of the fusion protein and finally, cells in G1 and G2 have a homogeneous distribution (Supplementary Figure S1A). Those cells that did not show these characteristic PCNA features were discarded from further analyses.

Some previous works described methods based on machine learning to quantify PCNA distribution (Schonenberger et al., 2015) and we have also developed an image-based routine with this purpose (Presentation 2, Supplementary Materials and Methods). However, the performance of these methods is not better than that of the manual classification widely used in the literature (Leonhardt et al., 2000; Sporbert et al., 2002; Pomerening et al., 2008; Leung et al., 2011; Barr et al., 2016; Wilson et al., 2016; Dutta et al., 2019; Velasquez et al., 2019; Velasquez et al., 2022; Xie and Bankaitis, 2022) which constitutes the gold standard method to date.

S cells transitioning between phases usually present main features of a given phase combined with some features of a previous or consecutive phase. In these few cases, we used the following criteria for splitting cells between contiguous phases:

1. E-S to M-S transitioning cells: they show a combination of foci in the periphery (characteristic of M-S) and other in the nucleus interior (as observed in E-S). The cell is considered E-S or M-S if most of the foci are localized in the interior or the periphery, respectively. The cell is discarded from the analysis if the number of foci in both locations is similar.

2. M-S to L-S transitioning cells: they show a combination of foci in the periphery (characteristic of M-S) with bigger foci (as observed in L-S). We consider the cell as L-S if the cell shows few foci (although most of them are at the periphery) but bigger than the average size, otherwise, we classified the cell as M-S.

Miss-classifications between continuous phases result in smaller differences between the parameters calculated for these phases, thus the differences between these populations might be larger than those detected in this work.

Cells were plated on coated coverslips and transfected with a vector encoding PCNA-RFP as described above. Then, the samples were fixed with 4% PFA for 15 min, permeabilized with 0.1% Triton X-100 in PBS and stained with DAPI.

The cells were observed in a widefield fluorescence microscope set to collect DAPI fluorescence using a 10x air objective (NA = 0.3). We used the StarDist ImageJ plugin (Weigert et al., 2020) to segment all nuclei and quantify their DAPI integrated intensity. The integrated intensity of each cell was normalized to that of the mean of its colony (I_DAPI, _normalized) for correcting small variations of the illumination of the sample and/or in DAPI staining throughout the coverslip.

The same cell colonies were then observed by confocal microscopy to locate those transfected cells and collect PCNA-RFP images with high spatial resolution (60x oil objective, NA = 1.35). By following this procedure, we collected DAPI and PCNA images of the same cells in the widefield and confocal microscopes, respectively.

RT-qPCR was performed and analyzed as previously described (Waisman et al., 2017; Verneri et al., 2020). Briefly, total RNA was extracted with QuickZol (Kalium Technologies). Then, RevertAid Reverse Transcriptase (ThermoFisher) was used to reverse transcribed RNA. cDNA Quantitative Real time PCR amplification was carried out using FastStart SYBR Green Master (Roche) in a LineGene 9600 engine (BioER). 2 or 3 biological replicates were performed in all the experiments, with 2 technical replicates for each condition. Gene expression was normalized to the geometrical mean of Gapdh values. The list of primers is included in Supplementary Table S1.

Confocal microscopy experiments were run in two FV1000 microscopes (Olympus) with either spectral or filter-based detectors. We checked a similar performance of these microscopes after running identical imaging and FCS experiments in both setups.

YPet and eGFP fusion proteins were visualized using a multi-line Ar laser tuned at 488 nm as excitation source whereas a 543 nm He-Ne laser was used for RFP fusion proteins. The laser light was reflected with a dichroic mirror DM405/488/543/635 for dual-color experiments and focused through an Olympus UPLSAPO 60X oil immersion objective (NA = 1.35) into the sample.

For confocal imaging, fluorescence was split into two channels set to collect in sequential mode between 500–530 nm and 580–680 nm (spectral detectors; measurements of YPet-OCT4, YPet-SOX2 and HP1α-eGFP in undifferentiated ESCs) or between 505–525 nm and 572–642 nm (emissions filters; measurements of YPet-OCT4, YPet-SOX2 and HP1α-eGFP in induced cells). A stack of 5 images was collected in each field and its average was used in further quantitative analyses.

Single-point FCS experiments were run using a laser power of ∼1 µW and collecting fluorescence in the range 500–530 nm (spectral detector; measurements of YPet-OCT4, YPet-SOX2 and HP1α-eGFP in undifferentiated ESCs) or 505–525 nm (emission filter; measurements of eGFP-NANOG undifferentiated ESCs and YPet-OCT4, YPet-SOX2 and eGFP-NANOG in induced cells) with the detectors set in the pseudo photon-counting mode.

We only run a single measurement (imaging and FCS experiments) in each cell to minimize photobleaching and photodamage that could also produce alterations in the cell cycle (Magidson and Khodjakov, 2013).

The microscope was set to collect intensity in a single-point of the cell nucleus at 50,000 Hz during 2.7 min (OCT4 and SOX2 experiments) or at 25,000 Hz during 5.4 min (NANOG experiments). We only run a single experiment in each cell to minimize photodamage.

The auto-correlation function (ACF) data was calculated using the SimFCS program (LFD, Irvine, CA, United States) and fitted with Eq. 1 that considers the diffusion of the TFs and their binding to two populations of fixed sites (White et al., 2016): G τ = 1 2 3 / 2 N f D 1 + τ τ D − 1 1 + τ ω 2 τ D − 1 / 2 + f s h o r t e − τ τ s h o r t + f l o n g e − τ τ l o n g (1) where N is the mean number of fluorescent molecules in the confocal volume, τ_D is the characteristic diffusion time, ω is the ratio between axial and radial waists of the observation volume, and f_D is the freely diffusing population fraction. f_short and f_long are the population fractions bound to short-lived and long-lived targets, and τ_short and τ_long are their residence times, respectively. The reciprocal of the residence time corresponds to the dissociation constant k_off.

Images were analyzed as previously described (Verneri et al., 2020). Briefly, the coefficient of variation (CV) was calculated as the ratio between the standard deviation of the intensity and the mean intensity of the nucleus, both calculated excluding nucleoli. Foci were identified in binarized images of nuclei considering an intensity threshold of mean + 2 × Standard deviation. The number and mean intensity of these foci were then calculated using the ImageJ plugin “Analyze Particles.” We only considered those structures with sizes ≥ optical resolution.

Results from microscopy experiments were expressed as mean ± SEM of at least three biological replicates. Statistical significance between groups was analyzed using Linear Mixed Models. Residuals fitted normal distribution and homogeneity of variance. Each cell cycle phase was compared with its previous phase. Differences were considered as significant at p-value ≤ 0.05. Statistical analysis was performed using the gls package of RStudio.

The core pluripotency TFs present a heterogeneous distribution in the nucleus with few foci (Verneri et al., 2020) that, at least in the case of OCT4, seem to play a relevant role in chromatin organization during interphase (Wang et al., 2021). Here, we use the term condensates to refer to as these membrane-less foci without implications on the molecular mechanism driving the formation of these structures.

We first asked whether the distribution of SOX2 and OCT4, and particularly their condensates, change during S phase. With this aim, we used the ESC lines YPet-OCT4 and YPet-SOX2 which express either OCT4 or SOX2 fused to the fluorescent protein YPet upon induction by Dox (Verneri et al., 2020).

To identify ESCs in S phase, we transfected the cells with a vector that encodes the proliferating cell nuclear antigen (PCNA) fused to the red fluorescent protein (PCNA-RFP). This protein associates to the replication fork (Moldovan et al., 2007; Boehm et al., 2016) and it has been extensively used in the literature to identify cells in S phase (Pomerening et al., 2008; Barr et al., 2016; Zerjatke et al., 2017; Grant et al., 2018). PCNA changes its nuclear distribution during the different stages of S phase (Leonhardt et al., 2000; Somanathan et al., 2001; Sporbert et al., 2002; Essers et al., 2005; Held et al., 2010; Schonenberger et al., 2015) and thus it has been used as a reporter of these different stages in both fixed (Dutta et al., 2019; Velasquez et al., 2019; Lin et al., 2020; Velasquez et al., 2022; Xie and Bankaitis, 2022) and living cells (Ersoy et al., 2009; Piwko et al., 2010; Leung et al., 2011; Wilson et al., 2016).

In line with these previous data, we observed that cells in early-S (E-S) phase showed multiple small foci of PCNA-RFP that distributed homogeneously within the nuclear space; mid-S (M-S) cells presented foci close to the nucleoli or at the nuclear periphery whereas cells transiting late-S (L-S) phase showed fewer and bigger foci of the fusion protein (Figure 1A; Supplementary Figure S1A). As expected, E-S, M-S and L-S cells presented increasing levels of DAPI intensities (Figures 1B, C) confirming that PCNA distribution allows the identification of cells in different stages of S-phase. Furthermore, the area of the optical section of nuclei in M-S and L-S phases was significantly larger than those of E-S cells (Supplementary Figure S1B), as expected from the increase of the nuclear volume during DNA replication (Maeshima et al., 2011). Finally, cells in G1 and G2 (herein referred to as G cells) showed a homogeneous distribution of PCNA-RFP (Figure 1A; Supplementary Figure S1A). Cells in mitosis were identified by the recruitment of the pluripotent TFs to the condensed chromosomes (Supplementary Figure S2A) as previously reported (Deluz et al., 2016), and were not included in further analyses.

We should also mention that these single-cell microscopy observations require cells growing attached to a substrate and thus we could not use cell sorting to select cells in specific phases of the cell cycle. The attachment of cells to the substrate takes hours and, during this period, the cell cycle of the sorted cells progress and the population become more heterogeneous.

Next, we quantified the distributions of YPet-OCT4 and YPet-SOX2 in single, live cells by calculating the coefficient of variation (CV), that provides information of the overall distribution of the fluorescent TFs, the mean number of foci (N_foci) and their intensity relative to that of the nucleus (I_r,foci). These parameters were previously used to obtain a quantitative description of the heterogeneous distribution of nuclear proteins including TFs such as OCT4 and SOX2 (Htun et al., 1999; Wu et al., 2006; Stortz et al., 2020; Verneri et al., 2020).

Although these parameters were not significantly different between G and S cells (Supplementary Figures S2B, C), changes were evident during the different stages of S phase (Figure 1D; Supplementary Figures S2D). Specifically, the number of foci in YPet-OCT4 (N cells: 34, 26 and 33 for E-S, M-S and L-S, respectively) and YPet-SOX2 (N cells: 30, 27, 35 for E-S, M-S and L-S, respectively) cells significantly increased after the E-S to M-S phase transition (Figure 1D). This observation could be related to the early timing of replication of active genes (Rhind and Gilbert, 2013; Hiratani and Takahashi, 2019); we hypothesize that pluripotency-associated OCT4 and SOX2 target genes probably replicate during E-S phase in PSCs and thus, those pluripotency TFs molecules that detach from replicating DNA might be further recruited to condensates.

In a previous work (Verneri et al., 2020), we showed that OCT4 and SOX2 condensates colocalize with regions of high chromatin compaction in ESCs, as evidenced by fluorescent versions of the histone H2B and the heterochromatin protein 1 (HP1α), a protein commonly associated with silenced heterochromatin regions (Grewal and Jia, 2007). Therefore, we next asked if the formation of new TF condensates after the E-S to M-S transition occurs concomitantly to the formation of compacted chromatin domains.

We analyzed the distribution of HP1α in ESCs co-transfected with vectors encoding HP1α fused to the enhanced green fluorescent protein (HP1α-eGFP) and PCNA-RFP by following the procedures described above. In line with previous results (Dialynas et al., 2007; Hinde et al., 2015; Strom et al., 2017), HP1α-eGFP formed foci in the nucleus of these ESCs (Figure 2A).

We first compared S and G cells and found that both CV and I_r,foci were significantly higher in S phase suggesting that HP1α is recruited to foci during DNA replication (Supplementary Figure S3A). Next, we analyzed HP1α distribution during S phase (Figure 2B; N cells: 40, 41 and 57 for E-S, M-S and L-S, respectively) and observed an increase of the CV value after the M-S to L-S transition (Supplementary Figure S3B) indicating that this transition promotes a more heterogeneous distribution of this protein.

Previous works showed that gene-poor, HP1-associated heterochromatin replicates at L-S phase (Ryba et al., 2010; Julienne et al., 2013; Rhind and Gilbert, 2013; Wootton and Soutoglou, 2021) and thus, we speculate that those changes in HP1α distribution occurring after the M-S to L-S transition could be due to a reorganization of HP1α required for heterochromatin duplication.

We should emphasize that the parameters characterizing the distribution of HP1α were not significantly different in E-S and M-S cells (Figure 2B; Supplementary Figure S3B), indicating that those changes of OCT4 and SOX2 distributions observed after this transition (Figure 1D) cannot be directly attributed to a reorganization of chromatin structure sensed through HP1α distribution.

To get further insights into the reorganization of HP1α occurring through the cell cycle, we used fluorescence correlation spectroscopy (FCS), a technique that provides exquisite information on the dynamical distribution and interactions of nuclear proteins in cells and whole organisms (Clark et al., 2016; White et al., 2016; Stortz et al., 2017; Cosentino et al., 2019; Verneri et al., 2020). Supplementary Figure S3C shows the ACF data obtained at different stages of the cell cycle (N cells: 58, 35, 27 and 33 for G, E-S, M-S and L-S, respectively). In line with our previous work (Cosentino et al., 2019), we fitted these data with Eq. 1, derived from a model that considers the diffusion of this protein and its interactions with chromatin targets in two distinct temporal windows (White et al., 2016; Stortz et al., 2017; Verneri et al., 2020).

Whereas HP1α-eGFP presented similar dynamics in S and G cells, we detected an increase of the lifetimes of HP1α-chromatin interactions in M-S cells (Figures 2C, D). A previous work (Quivy et al., 2008) showed that, to progress through M-S and L-S stages and to achieve the replication of pericentric heterochromatin, HP1 must interact with the chromatin assembly factor 1 complex (CAF-1), which is fundamental for heterochromatin organization in ESCs (Houlard et al., 2006). Therefore, we speculate that the changes detected in M-S cells could be indirectly related to the required establishment of interactions between HP1 and CAF-1 for heterochromatin replication in later stages of S phase; further work needs to be done to assess this hypothesis. Relevantly, a previous work also reported a distinct distribution of these molecules in mid-late S cells in which CAF-1, together with newly synthesized DNA, concentrates at the periphery of pericentric heterochromatin domains (Quivy et al., 2004).

We also analyzed publicly available data from genome-wide ATAC-seq experiments to get insights in the chromatin accessibility of ESCs (Friman et al., 2019) and observed that chromatin accessibility is constantly changing during the cell-cycle (Supplementary Figure S4). Particularly, heatmap analyses of chromatin accessibility changes reveal important variations in both late G1 and S phases. We should highlight that these results provide a broad panorama of the accessibility changes and cannot be directly compared to or associated with our experimental results since cells in Friman et al. (2019) were classified as early G1, late G1, S and S + G2.

Taken together, our results show that HP1α distribution and dynamics change during S phase. Relevantly, these modifications are not parallel to those observed for the pluripotent TFs (Figure 1) indicating that other, yet unknown factors different from chromatin compaction define the redistribution of OCT4 and SOX2 observed during S phase in ESCs.

To get insights into the dynamical organization of OCT4 and SOX2 during S phase, we run FCS experiments in YPet-OCT4 and YPet-SOX2 cells also transfected with the PCNA-RFP vector. We have previously shown that this technique allows detecting changes in the dynamical organization of pluripotency TFs occurring at early stages of differentiation (Verneri et al., 2020) and those indirectly produced by a histone acetyltransferase that modify the compaction of chromatin (Cosentino et al., 2019).

Figures 3A, B show that the dynamics of both TFs was significantly different in S and G cells (N cells OCT4: 53 and 157 for G and S; N cells SOX2: 56 and 169 for G and S, respectively). Specifically, S cells showed a higher proportion of TF molecules attached to long-lived chromatin targets, but these interactions were faster than those observed in G cells. These observations suggest that OCT4 and SOX2 reorganize among chromatin targets during S phase.

FCS experiments also revealed subtle, albeit statistically different modifications in the dynamics of OCT4 (N cells: 47, 42 and 68 for E-S, M-S and L-S, respectively) and SOX2 (N cells: 60, 46 and 63 for E-S, M-S and L-S, respectively) during S phase. Particularly, OCT4 detached from chromatin targets after the E-S to M-S phase transition and those bound OCT4 molecules remained longer at short-lived targets (Figure 3C). On the other hand, SOX2 redistributed to more labile chromatin targets in M-S cells as the lifetime of SOX2-chromatin interactions decreased and its short-lived bound fraction increased after this transition (Figure 3D). These observations indicate that the E-S to M-S transition involves changes of TF-chromatin interactions that depend on the TF identity concomitant with a reorganization of TFs condensates (Figure 1). Figures 3C, D also show changes in the lifetimes of short-lived interactions of both TFs and a reduction in the fraction of SOX2 molecules bound to the short-lived targets during the M-S to L-S transition.

Taken together, our results show that the landscape of interactions of OCT4 and SOX2 with chromatin targets changes during the DNA replication. Although these core pluripotency TFs also act as heterodimers in the regulation of many genes (Remenyi et al., 2003; Boyer et al., 2005; Chew et al., 2005; Rodda et al., 2005), the different dynamics during S phase is consistent with their different chromatin-binding landscape (Rizzino, 2013) and interactions with other proteins (Buecker et al., 2014).

We mentioned before evidence showing that mouse ESCs respond to differentiation cues after cell division (Chaigne et al., 2020) and specifically, in G1 phase (Waisman et al., 2017). Also, we have previously reported that differentiation stimuli trigger early changes in the distribution and dynamics of SOX2 and OCT4 that precede their downregulation (Verneri et al., 2020). In this context, we next asked if differentiation cues received in G1 affect the distribution and/or dynamics of these pluripotency TFs in the subsequent E-S phase.

We used YPet-OCT4 and YPet-SOX2 ESCs expressing PCNA-RFP and induced differentiation by LIF/2i withdrawal (Verneri et al., 2020) during 4 h (Figure 4A). G1 lasts approximately 2–3 h (Waisman et al., 2017) and thus E-S cells observed after the treatment most likely received the differentiation signal in the preceding G1 phase.

Surprisingly, most E-S cells did not show TF condensates after differentiation induction (Figure 4B); indeed, only 24% of YPet-OCT4 (N cells: 58) and 21% of YPet-SOX2 (N cells: 39) E-S cells that received the differentiation signal (herein referred to as induced E-S cells) presented the characteristic TF foci. Consequently, N_foci was lower in induced E-S cells although the remaining foci were brighter than those observed in undifferentiated E-S cells (Figure 4C) suggesting a higher recruitment of TF molecules to these fewer foci. Also, induced E-S cells have higher CV (Supplementary Figures S5A, B) suggesting an overall, more heterogeneous distribution of OCT4 and SOX2 in the nucleus of these cells compared to undifferentiated E-S cells.

To explore if these changes in pluripotency TF condensates are related to evident modifications of chromatin condensation in this early time-window of differentiation, we analyzed HP1α-eGFP distribution in E-S cells exposed to the differentiation stimulus in G1 phase (N cells: 37). We detected HP1α foci in every induced E-S cell suggesting that the lower number of OCT4 and SOX2 condensates are not a direct consequence of a similar reorganization of HP1α. The intensity of HP1α foci as well as CV values were significantly lower than those observed in undifferentiated E-S cells indicating that the differentiation stimuli triggered a more homogeneous distribution of HP1α in E-S phase (Figure 4C; Supplementary Figure S5C). Although chromatin in pluripotent ESCs is globally decondensed and increases its compaction during differentiation (Azuara et al., 2006; Meshorer et al., 2006; Gokbuget and Blelloch, 2019), this last result could be interpreted considering that there is a short period of chromatin decondensation during the first stages of differentiation (Chalut et al., 2012; Petruk et al., 2017). Particularly, chromatin decondenses after 2–4 h of differentiation induction in mouse ESCs (Petruk et al., 2017). Despite the methodological differences, this previous work suggests that decondensation occurs in a temporal window similar to that analyzed in our experiments.

To get further insights into the reorganization of OCT4 and SOX2, we run FCS experiments similar to those described above (Figures 4D, E).

Figures 4D, E show that induced E-S cells presented significantly higher lifetimes of OCT4 (N cells: 52) and SOX2 (N cells: 77) interactions with chromatin in comparison to undifferentiated E-S cells. These results suggest that differentiation induction in G1 promotes stronger OCT4 and SOX2 interactions with chromatin in E-S cells. Also, these TFs reorganized among targets as revealed by the higher long-lived binding fraction of SOX2 and the smaller short-lived binding fraction of OCT4 determined in induced E-S cells.

In contrast to OCT4 and SOX2, NANOG is associated with a naïve pluripotency state (Martello and Smith, 2014), the induction of differentiation results in its fast downregulation (Kalkan et al., 2017; Waisman et al., 2017; Verneri et al., 2020) and its expression levels also diminish during mitosis (Liu et al., 2017). Therefore, we next explored NANOG dynamics during the cell cycle and asked if it also responds to differentiation cues received in G1 phase as SOX2 and OCT4.

We first generated an ESC line that stably expresses NANOG fused to eGFP (eGFP-NANOG) as described in Materials and methods. We verified eGFP-NANOG expression and observed that the cell and colony morphologies and the cell cycle were similar to those corresponding to the parental cell line (Supplementary Figure S6A–C). We also induced differentiation by LIF/2i withdrawal and analyzed gene expression. Supplementary Figure S6D shows that the naïve pluripotency markers Nanog, Esrrb and Klf4 are downregulated at early differentiation and that the marker of primed pluripotency Oct6, which is upregulated early during differentiation (Kalkan et al., 2017; Waisman et al., 2017; Waisman et al., 2019), is upregulated, confirming that the new cell line is capable to exit naïve pluripotency despite expressing the fusion protein eGFP-NANOG.

Then, we transfected these cells with PCNA-RFP and run FCS experiments to analyze the dynamics of NANOG during the cell cycle.

Figure 5A shows that the fraction of NANOG molecules in S cells involved in long-lived NANOG-chromatin interactions was higher while the lifetime of these interactions was lower in comparison to G cells (N cells: 41 and 156 for G and S, respectively), in line with the behavior observed for SOX2 and OCT4. Also, we studied the dynamics of NANOG during the different stages of S phase (N cells: 43, 47 and 69 for E-S, M-S and L-S, respectively) (Figure 5B) and found that the lifetime of NANOG-chromatin interactions increased after the E-S to M-S transition. These results show that NANOG also redistributes among chromatin targets during the cell cycle.

Next, we asked if NANOG dynamics in E-S phase responds to differentiation cues received in G1. Although Nanog expression levels decrease during differentiation, its protein levels do not change significantly after 6 h from naïve pluripotency exit (Yang et al., 2019). In line with this observation, RNA levels of Nanog are not significantly different after 5 h of receiving the differentiation stimuli in G1 (Waisman et al., 2017).

Surprisingly, NANOG behavior (Figure 5C) was almost the opposite of that observed for OCT4 and SOX2 (Figures 4D, E). Specifically, we found that the lifetimes of NANOG-chromatin interactions were significantly smaller in induced E-S cells (N cells: 45) while the short-lived fraction increased in comparison to undifferentiated E-S cells suggesting weaker NANOG-chromatin interactions in cells that were induced to differentiate in G1.

The different behavior between OCT4/SOX2 and NANOG could be related to the different functional roles of these pluripotency TFs discussed above. We hypothesize that differentiation cues trigger a redistribution of OCT4 and SOX2 among binding sites in the chromatin that might be required to switch to a differentiation program in line with their expected activity as pioneer factors (Friman et al., 2019). In contrast, the detachment of NANOG from chromatin sites in induced E-S cells probably constitutes one of the initial steps toward its downregulation and the consequent rapid repression of pluripotency-related genes.