Correlations Between the Levels of Oct4 and Nanog as a Signature for Naïve Pluripotency in Mouse Embryonic Stem Cells§

Authors

  • Silvia Muñoz Descalzo,

    Corresponding author
    1. Department of Genetics, University of Cambridge, Cambridge, United Kingdom
    • Silvia Muñoz Descalzo, Department of Genetics, University of Cambridge, Cambridge CB2 3EH, U.K

      Alfonso Martinez Arias, Department of Genetics, University of Cambridge, Cambridge CB2 3EH, U.K

    Search for more papers by this author
    • Telephone: +44 (0)1223 766742; Fax: +44 (0)1223 333992

  • Pau RuÉ,

    1. Departament de Física i Enginyeria Nuclear, Universitat Politècnica de Catalunya, Terrassa, Spain
    Search for more papers by this author
  • Jordi Garcia-Ojalvo,

    1. Departament de Física i Enginyeria Nuclear, Universitat Politècnica de Catalunya, Terrassa, Spain
    Search for more papers by this author
  • Alfonso Martinez Arias

    Corresponding author
    1. Department of Genetics, University of Cambridge, Cambridge, United Kingdom
    • Silvia Muñoz Descalzo, Department of Genetics, University of Cambridge, Cambridge CB2 3EH, U.K

      Alfonso Martinez Arias, Department of Genetics, University of Cambridge, Cambridge CB2 3EH, U.K

    Search for more papers by this author
    • Telephone: +44 (0)1223 766742; Fax: +44 (0)1223 333992


  • Author contributions: S.M.D.: conceived the project, carried out the experiments, the measurements, and the analysis of the data, and wrote the manuscript; A.M.A.: conceived the project and wrote the manuscript; P.R. and J.G.O.: contributed to the analysis of the data.

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

  • §

    First published online in STEM CELLSEXPRESS November 7, 2012.

Abstract

The pluripotent state is traditionally associated with large absolute levels of certain transcription factors such as Nanog and Oct4. Here, we present experimental observations using quantitative immunofluorescence that pluripotency in mouse embryonic stem cells (mESCs) is established by specific ratios between Oct4 and Nanog. When cells are grown in 2i conditions, they exhibit uniform levels of pluripotency and this is associated with a high correlation between the levels of Oct4 and Nanog in individual cells. The correlation is lost when cells differentiate. Our results suggest that the correlation between these two factors and the distribution of Oct4/Nanog ratios can be used as quantifiers to distinguish between three subpopulations in an mESC culture: pluripotent, lineage-primed, and differentiating cells. When we apply these quantifiers to cells with lower levels of Nanog or mutant for β-Catenin or Tcf3, the results suggest that these cells exhibit higher probability of differentiation. STEM CELLS 2012;30:2683–2691

INTRODUCTION

The term pluripotency refers to the ability of a cell to give rise to all somatic elements of an organism, including the germ line. The best characterized example of pluripotent cells is mouse embryonic stem cells (mESCs), clonal populations derived from single cells of mouse blastocysts that can be maintained in culture indefinitely [1]. Understanding the molecular basis of pluripotency represents a significant challenge as it will provide insights into fundamental processes of embryonic development and allow a better use of these cells in biomedicine.

The last few years have revealed that in mESCs pluripotency relies on a small network of transcription factors centered around Nanog, Oct4, and Sox2, with a significant contribution from Klf4 [2, 3]. The activity of this network is counteracted by Tcf3 and is modulated by extracellular signals, principally by the controlled activity of the Fgf and Wnt pathways [4–6]. The significance of interactions between these molecular devices has been made evident through the effect they have on somatic cells: a cocktail containing Oct4, Sox2, and Klf4 is able to reprogramme any somatic cell into an ESCs, as long as it leads to the expression of Nanog [7, 8]. The frequency of this event is modulated by the activity of both Fgf and Wnt/β-Catenin signaling [9, 10]. Notably, the simultaneous inactivation of Fgf and activation of Wnt/β-Catenin signaling (2i conditions) in a growing culture of ESCs promotes and stabilizes pluripotency, maintaining cells in what has been termed the “ground state” [11].

Although we know much about the signals and transcription factors that control pluripotency, how cells register this state remains unclear. One of the intriguing aspects of the pluripotent state is the heterogeneity in the expression of proteins associated with it, which is a hallmark of ESC cultures and has been suggested to be an intrinsic component of stemness [12]. In 2i, where cells exhibit robust pluripotency, heterogeneities are severely reduced and all cells exhibit high, quasihomogeneous levels of Nanog and Oct4 [13]. This is often interpreted to suggest that the absolute levels of these transcription factors determine pluripotency. However, there is evidence that pluripotency requires the joint and balanced activity of Nanog and Oct4. For example, in cell fusion-induced reprogramming, while Nanog on its own has no impact on the frequency of reprogramming, it does have a significant effect when a cell overexpressing Nanog is fused with an established ESC, that is, a cell with an active pluripotency network (and thus an appropriate level of Oct4) [14]. Also, cells coexpressing Nanog and Oct4 show a more stable pluripotent state than cells overexpressing either of these factors individually [15]. These observations suggest that it is the relationship between the pluripotency factors rather than their presence, absence, or absolute levels that determines pluripotency. Here, we explore this further and reveal that pluripotency is associated with specific ratios between Oct4 and Nanog, and that β-Catenin, which has been shown to be important in the maintenance of pluripotency [6, 16], plays a role in the regulation of these ratios. Our quantitative analysis provides some insights into the state of pluripotency, its structure, and dynamics.

MATERIALS AND METHODS

Cell Culture

The cell lines used in this study were E14Tg2A (129-derived wild-type cells); targeted nanog GFP, clone A (TNGA) (which contain a green fluorescent protein insertion in the Nanog locus [17]); β-Cateninfl/-, β-Catenin−/−, and β-Catenin−/−C [6]; and Tcf3+/+ and Tcf3−/− [18]. For routine maintenance, all mESC lines were cultured on gelatin in serum and leukemia inhibitory factor (LIF) (for details see [19]). β-Cateninfl/-, β-Catenin−/−, and β-Catenin−/−C cells were cultured on fibronectin in 2i (N2B27, StemCells, Inc., supplemented with 1 μM PD0325901 and 3 μM, Chiron, Cambridge, U.K., http://www.stemcellsinc.com) and LIF.

Immunofluorescence

Immunostainings were performed as previously described [19]. The experiments and posterior analysis were performed using monoclonal antibodies: α-Nanog (eBiosciences 14-5761, 1:100, 5 μg/ml, Hatfield, U.K., http://www.ebioscience.com) and α-Oct4 (Santa Cruz 5279, 1:100, 2 μg/ml, Santa Cruz, CA, http://www.scbt.com). The specificity of these antibodies was tested against cells mutant for either for Oct4 (ZHBTc4, [20]) or Nanog [17]. Vectashield (with or without 4′,6-diamidino–2-phenylindole (DAPI)) was used as mounting medium.

Another combination of polyclonal antibodies was also used for out analysis (data not shown). These were: α-Nanog raised in rabbit (from Chemicon 5731,1:150, 6.7 μg/ml, Chemicon: Billerica, MA, http://www.millipore.com) and Oct4 raised in goat (Santa Cruz 8628, 1:100, 2 μg/ml). Analysis of mESCs stained using this combination of antibodies also reveals a high degree of linear correlation between Oct4 and Nanog (r = 0.80) and a similar slope (α = 0.43). These results show that the pluripotency quantifiers we have determined in our study using monoclonal Nanog and Oct4 antibodies are a general feature of mESCs associated with these proteins and not to specific antibodies. However, as this Nanog antibody has a nonspecific component, we performed our mathematical analysis with the monoclonal set of antibodies.

Imaging and Image Analysis

Samples were prepared for confocal microscopy and 512 × 512 images were acquired using a Zeiss LSM 510-Meta or Zeiss 700 confocal microscope. All images were obtained using the sequential scanning mode, with the same conditions of laser intensity, gain, and pinhole, and were processed in exactly the same way. Range indicator palette option was used to ensure that no oversaturated images were taken. Imaged colonies were randomly selected by size among those that would fit completely within the imaging field. Confocal images from an optical section located 4–5 μm from the coverslip were saved independently in grayscale for each of the channels. An automated image processing methodology based on mathematical morphology operators was designed for single-cell identification (segmentation) using membrane and nuclei information for quantification of the immunostainings. The algorithms were implemented in Matlab and integrated in a custom-made script. The removal of the oversegmented objects from the analysis was implemented in the script, while undersegmented ones were removed manually. The accuracy of the segmentation (considering the small, oversegmented, and big, undersegmented, objects) ranged from 78% to 87%, depending on the experiment. Scatterplots, line plot distributions, linear best fits, statistical, and Principal Component analysis (PCA) of the data were done using Matlab.

The quantitative immunofluorescence (QIF) analysis contains a number of assumptions: (a) that the antibody binds to all available specific antigens. As we use paraformaldehyde (PFA) as a fixative, which works as a crosslinker, a permeabilization step was included in the protocol to allow the access of the antibody to the antigens and increase its chances of binding. (b) That the antibody is specific to the antigen. To ensure this, the Nanog and Oct4 antibodies used were tested for specificity using Nanog and Oct4 mutant cell lines. (c) That there might be variability across samples. To reduce this variability, when different cell lines were tested for their Oct4/Nanog correlation, the whole experiment was performed in parallel. This includes the seeding of the cells, the complete protocol for fluorescent immunohistochemistry as well as the imaging steps. (d) That the intensity of the fluorescent signal is proportional to the concentration of the antigen.

RESULTS

Correlations and Ratios of Oct4 and Nanog as Parameters of an ESC Population

Fluorescent immunocytochemistry allows the quantification of the amounts of different antigens at the single-cell level. While there are some caveats to this approach, under a number of assumptions (see Materials and Methods) QIF provides information such as distributions of protein levels in single cells across populations or relative amounts of two or more proteins within a single cell, which are not accessible through other techniques involving population averages such as Western blot analysis. Here, we have used QIF of some landmark pluripotency factors to assess the state of mESCs (Fig. 1).

Figure 1.

Single-cell and population analyses of Oct4 and Nanog levels in Tg2A cells cultured under different growing conditions. (A, B): Representative confocal images of Nanog and Oct4 expression in E14Tg2A cells grown in serum + LIF (A) or 2i + LIF (B). Segmented images with nuclei and membrane channels, and identified objects are also shown; (see also Figure 2). Scale bar = 50 μm. (C, D): Correlation plots of Oct4 and Nanog levels (in fluorescence A.U., here and in subsequent similar graphs) in single Tg2A cells under standard self-renewal (serum + LIF, in C) and ground-state (2i + LIF, in D) conditions. r is the Pearson correlation coefficient between Oct4 and Nanog (here and in subsequent similar graphs). The blue line represents the lower limit for the Oct4/Nanog “allowed” values. The slope of this line is the slope of the linear regression for those cells with a value of Nanog larger than 10 (the low-expressing Nanog cells were removed from the analysis under the assumption that they are differentiating). The intercept of this line is the smallest value obtained by calculating the intercepts of the linear regression passing through every point. For comparison, the same line has been drawn in (D). (E): Statistical parameters of the distributions of Nanog and Oct4 in the population. (F–H): Distributions and parameters in five analyzed individual colonies; each color represents a colony. Abbreviation: LIF, leukemia inhibitory factor.

Figure 2.

Ratio of Oct4 and Nanog in Tg2A cells cultured under different growth conditions. (A, B): Correlation plots of the Oct4/Nanog ratio relative to the levels of Nanog in Tg2A single cells in serum + LIF (A) and 2i + LIF (B) conditions. The blue line is the slope of the linear regression calculated in Figure 1C. Insets show a zoomed area around this straight lower limit. (C): Distribution of the values of the Oct4/Nanog ratio across the population under the different culture conditions. The distribution frequency was obtained by binning the Oct4/Nanog ratio values in 20 logarithmically spaced categories. In standard self-renewing conditions, cells span a wider range of Oct4/Nanog ratio (black line, coefficients of variation [CV] = 2.04) than in 2i conditions (red line, CV = 0.53). (D–F): Principal component analysis (PCA) of Nanog and Oct4 data in serum + LIF (D) and 2i + LIF (E) displayed on top of the correlation plots. Arrows represent the detected components of the PCA. (F): Fraction of the total variance accounted for both components in serum + LIF (black boxes) and 2i + LIF (white boxes) conditions. In both datasets (D, E), the first component, PC1, approximately matches the absolute levels of Oct4 and Nanog and accounts for 80% and 93.4% of the variance, respectively. PC2, shown in color code, represents the difference between Nanog and Oct4 and accounts for the remaining variance. Notice that PC2 magnitudes (color) cannot be compared between both figures, but its accounted variance (20% and 6.6%, respectively) can. Abbreviations: LIF, leukemia inhibitory factor; SL, serum+LIF.

In standard self-renewing conditions (serum and LIF, Fig. 1A), the levels of Oct4 and Nanog exhibit a certain degree of correlation at the level of single cells (which can be measured by the Pearson correlation coefficient r), as shown in Figure 1C (r = 0.55 in our experiments, see also [21, 22] and Materials and Methods). This correlation is reflected in a feature that may be functionally significant, namely the existence of a sharp diagonal boundary (blue line in Fig. 1C) that defines a lower limit to the values of Nanog and Oct4 that can coexist in the same cell: no cells exist with a ratio Oct4 and Nanog below this limit. It is also worth noting that the distributions of expression values for Oct4 and Nanog across the population, as well as their statistical moments (black bars in Fig. 1E), are representative of the values within each colony, that is, the distributions, averages, and coefficients of variation (CV, standard deviation normalized by the mean) of the population and of each colony are similar to each other (compare Fig. 1C with 1F, and 1E with 1H, left panels).

These observations suggest that, in a self-renewing population of ESCs, there are constraints in the relationship between Oct4 and Nanog that are common to all colonies. In particular, the absence of cells with correlated values of Oct4 and Nanog below the diagonal in Figure 1C indicates that such values are not compatible with pluripotency. This becomes clear if instead of graphing the levels of Oct4 against the levels of Nanog in a population, we graph the ratio of the levels of Oct4 and Nanog relative to the levels of Nanog (Oct4/Nanog vs. Nanog) (Fig. 2A). This plot reveals a horizontal line (blue line in the figure) that underscores a lower limit to the ratio Oct4/Nanog which cells can have in a culture (Table 1). This line corresponds to the slope of the linear regression of a corrected Oct4 versus Nanog plot (Fig. 1C and Materials and Methods), which has a value of around 0.42 in this case. We shall call this line and the Oct4/Nanog ratio it represents the Ground State (GS) value or slope (Ground State, see below). When the Oct4/Nanog ratio is close to this level, cells in serum and LIF can have a wide spread of values of Nanog (Fig. 2A). The graph also shows that when the levels of Nanog drop below a certain value, cells display a large variation in the Oct4/Nanog ratio that tends to increase for decreasing values of Nanog and results in a relatively wide distribution of Oct4/Nanog values (black line in Fig. 2C, CV = 2.04).

Table 1. Quantifiers of mESCs
  1. a

    The percentage of cells in GS are those that are within one SD from the mean value of Oct4/Nanog ratio in cells grown under 2i conditions.

  2. b

    The slope is that of a linear regression for those cells with a value of Nanog larger than 10 (the low-expressing Nanog cells were removed from the analysis under the assumption that they are differentiating).

  3. c

    The slope is that of a linear regression for those cells with a value of Nanog larger than 5 (the low-expressing Nanog cells were removed from the analysis under the assumption that they are differentiating). Abbreviations: GS, ground state; LIF, leukemia inhibitory factor; TNGA, targeted nanog GFP, clone A.

original image

A PCA of the data provides additional information about the population as it allows us to characterize each cell in terms of two orthogonal components according to the distribution of the data variance. The first component, PC1, which describes the direction of maximum variance, identifies variations in absolute levels of Oct4 and Nanog; and PC2, roughly corresponds to a weighted difference between Nanog and Oct4 (Fig. 2D). According to this analysis, PC2 can be construed as an indicator of the degree of pluripotency in the population.

The observables that we have defined above (correlation between Oct4 and Nanog, distribution of Oct4/Nanog ratios around the GS slope, and PCA, Figs. 1, 2) constitute a statistical description of pluripotency (Table 1). Even though this description reveals a continuous degree of pluripotency, the PCA analysis (Fig. 2D) allows us to approximately distinguish three subpopulations: one with high Oct4/Nanog correlation levels clustered around the GS line (blue-green dots in Fig. 2D), a second one with low levels of Nanog and high levels of Oct4 (yellow-red dots), and a third one with low levels of both Oct4 and Nanog below the GS line (absent from the analysis presented above, which excludes differentiated cells). The lack of discontinuities between these classes reflects the dynamic nature of these states (reviewed in [12]). As the levels of Nanog are related to the probability of differentiation [17], we surmise that the first subpopulation that we identify corresponds to the pluripotent (blue-green dots), the second to the lineage-primed (red-yellow dots), and the third to the differentiating cell subpopulations (Fig. 2D).

These observations suggest that a population of ESCs can be characterized by quantifiers extracted from single-cell measurements of the correlation between the levels of Nanog and Oct4 (Table 1). Specifically, we surmise that the correlation between Oct4 and Nanog, the number of cells around the GS line, and the distribution of the Oct4/Nanog ratio together provide an objective description of the state of a population of ESCs (Table 1). The higher the value of r and the more cells at or above the GS line, the more pluripotent the population, that is, the lower its probability of differentiating over time. This is corroborated by a QIF analysis of TNGA cells, with only one copy of Nanog (Fig. 3) [17]. The correlation between Nanog and Oct4 in these cells is lower than the Tg2A control (.35 vs. .55; Fig. 3A) and the population contains many cells below the GS line (Fig. 3B). On the interpretation that the GS line defines a limit for robust pluripotency, our observations suggest that Nanog heterozygous cells have, as shown experimentally, a higher tendency to differentiate [17].

Figure 3.

Effects of Nanog dosage on the statistics of Oct4 and Nanog. Correlation plots of the Oct4 and Nanog levels (A) and Oct4/Nanog ratio relative to the levels of Nanog (B) in single TNGA cells. The green line in (B) is the slope of the linear regression calculated in Figure 1C for comparison. (C): Distribution of the values of the Oct4/Nanog ratio across the population. The results from the Tg2A cells grown in serum + LIF (from Fig. 2) are shown for comparison in (C). Abbreviations: LIF, leukemia inhibitory factor; SL, serum+ LIF; TNGA, targeted nanog GFP, clone A.

Effects of Signaling on Correlations

To test further the value of our measurements to describe the pluripotency of an ESC population, we performed an analysis of cells in the process of differentiation. Tg2A cells in self-renewing conditions were transferred to serum without LIF and the values of Oct4 and Nanog were determined in single cells 24 and 48 hours after removal of LIF (Fig. 4). Both the correlation and the Oct4/Nanog versus Nanog graphs behaved in agreement with our interpretation: during differentiation, the correlation values decreased (from r = .50 at the beginning of the experiment to r = .06 after 48 hours) and the ratio Oct4/Nanog was displaced to those related to the low levels of Nanog.

Figure 4.

Statistics of differentiating Tg2A cells after LIF removal. Correlation plots of the Oct4 and Nanog levels (A, C, E) and Oct4/Nanog ratio relative to the levels of Nanog (B, D, F) in single Tg2A cells. (A) and (B) represent the starting population (lower passage number cultured cells than those shown in Figure 1A, this explains the tighter Oct4/Nanog ratio distribution); (C) and (D), after culturing the cells for 24 hours without LIF; and (E) and (F), after culturing the cells for 48 hours without LIF. The blue correlation line was calculated for the Tg2A cells in serum + LIF as in Figure 1, and the same line has been drawn in (C–F) for comparison. (G): Distribution of the values of the Oct4/Nanog ratio across the population under the different culture conditions. Abbreviations: LIF, leukemia inhibitory factor; SL, serum+LIF.

After examining the status of our observables in a differentiating cell population, we explored their values at the opposite limit in which pluripotency is enhanced and differentiation inhibited (Fig. 1D). To do this, we use 2i conditions that have been suggested to take cells to a ground state where they reportedly exhibit reduced heterogeneity of Oct4 and Nanog and more stable pluripotency [11, 13]. In these conditions, we observe a high correlation between the levels of Oct4 and Nanog in single cells (r = 0.87 in 2i + LIF vs. 0.55 in serum and LIF; Figs. 1D, 2B), with most cells in the population lying within the GS line, and correspondingly the population exhibiting a low CV in the Oct4/Nanog ratio distribution (Fig. 2C). This homogeneity is reflected in the PCA analysis where the PC2 encompasses a heterogeneity of 6.6% compared to 20% in serum and LIF (Fig. 2E, 2F). This behavior is emphasized in the distribution of Oct4/Nanog versus Nanog values, which reveals a dramatic reduction in the presumed differentiating population (high Oct4/Nanog ratio) and an overrepresentation of cells in the GS line, as also indicated by an increase in the fraction of variance accounted by PC1 (Fig. 2F). Combined with the behavior of our measurements during differentiation (Fig. 4), this supports the notion that the GS line defines the ground state, and that cells in and around this line are pluripotent. This is in agreement with the observation that the proportion of cells within one standard deviation from the mean of Oct4/Nanog is 82.11% for the population in 2i conditions, while that number is only 30.45% in serum + LIF conditions (Table 1).

An interesting property of the 2i conditions is that the distributions of values within a colony do not match those of the population. Instead, each colony occupies a position in the correlation plot (Fig. 1G), and all cells within that colony have similar relative values of CV (Fig. 1H, right panels). This observation challenges the assumption, based on qualitative analysis, that all cells in 2i have homogeneous levels of Nanog and Oct4 [13]. While heterogeneity within a colony is indeed greatly reduced, it would appear that each colony behaves as a single cell and the intercolony differences that we observe raise the possibility that there are functional differences between them.

β-Catenin Is Required for the Maintenance of the Ground State

There is evidence that β-Catenin plays a role in the maintenance of pluripotency. For example, GSK3 inhibitors, which are a key component of 2i, have been shown to have an important effect on the amount and activity of β-Catenin and impact on the heterogeneities in Nanog expression within the culture [11, 23]. More significantly, loss of β-Catenin compromises pluripotency [6, 16] and increases the frequency of differentiation. This led us to analyze the effect that the loss of β-Catenin has on the parameters that we have established here (Fig. 5).

Figure 5.

Effects of β-Catenin on the statistics of Oct4 and Nanog. (A, D): Analysis of β-Cateninfl/-, (B, E) β-Catenin−/−, and (C, F) β-Catenin−/−C cells. (A), (B), and (C) are correlation plots of the Oct4 and Nanog levels in individual colonies from the different cell lines (the data from each colony is represented in a different color). A′, B′, and C′ show the mean values for Nanog and Oct4 in each colony. (D–F): are correlation plots of the Oct4/Nanog ratio relative to the levels of Nanog from the different cell lines. The blue line was calculated as in Figure 1 for the β-Cateninfl/- cells, for comparison the same line has been drawn in (D–F). (G): Distribution of the values of the Oct4/Nanog ratio across the population. Abbreviation: LIF, leukemia inhibitory factor.

Since loss of β-Catenin destabilizes pluripotency, it provides a test for the significance of our measurements. Furthermore, as β-Catenin mutant cells need to be grown in 2i conditions [6], this allows us to probe into the significance of the different pluripotency quantifiers. Correlation plots of Oct4 versus Nanog in β-Catenin mutant cells (Fig. 5B) show that although the levels of these proteins are low (when compared with the heterozygous parental line, Fig. 5A), they exhibit correlations within single cells similar to those of Tg2A cells in 2i conditions. Furthermore, each colony is located in a particular position along the correlation line, that is, 2i intercolony variation is preserved in the absence of β-Catenin (Fig. 5A′, 5B′). On this account, these cells are pluripotent, but analysis of the Oct4/Nanog versus Nanog plots (Fig. 5D, 5E) reveals distributions without a clear GS line and very skewed toward the differentiation region (with low levels of Oct4 and Nanog), confirming the instability of these cells and our suggestion that each pluripotency quantifier contains different but related information. One copy of β-Catenin missing the transcription activation domain stabilizes pluripotency and restores the parameters toward Tg2A levels (Fig. 5C-C′, 5F, and see below) and this provides a basis for the observation that this protein rescues the compromised pluripotency of β-Catenin mutant cells [6, 16].

Our results also suggest that each of the parameters that we are measuring reveals a different component of pluripotency. Thus, r (Oct4/Nanog correlation) is related to the degree of pluripotency of a given cell, a measure of its potential to differentiate at the time of the measurement: the higher the value of r, the less likely it is to differentiate. The GS line, particularly in the Oct4/Nanog versus Nanog graph defines a limit for stability of the pluripotent state at the population level, that is, how likely it is to differentiate over time: at or above 0.43 a population is stably pluripotent, below they are unstable and have a higher probability of differentiation. Furthermore, the further that a cell lies to the right side in this graph, the further away it is from differentiation, and this is the reason why the β-Catenin mutant cells, although correlated in the Oct4 versus Nanog plot, show that this correlation is not very stable as they lie very close to the differentiation region. This is further supported by the decrease in the proportion of cells in the GS state, from 92.21% for β-Cateninfl/- to 71.43% for β-Catenin−/−. This proportion can be rescued (to 97.52%) by expressing a truncated β-CateninC (β-Catenin−/−C), which lacks the C-terminal transactivation domain (Table 1). Bearing in mind that all these measurements are performed in 2i conditions, the observed changes are significant.

Tcf3 Mutants and the Ground State

The activity of the pluripotency network is repressed by Tcf3, and consequently Tcf3 mutant ESCs exhibit a very low rate of differentiation in the absence of LIF and are deemed to be pluripotent [5, 18, 21, 24]. These cells have low levels of Oct4, high levels of Nanog (Fig. 6), a low r (0.19 vs. 0.36 in their parental wild-type cell line), and all cells lie below the .46 GS mark. These parameters suggest that Tcf3 mutant cells are not in a robust pluripotent state (only 10.27% of the cells are in the GS state, vs. 34.07% of the parental cell line, Table 1), rather, like TNGA cells, they are prone to differentiation, although the loss of Tcf3 leads to a dramatic slow-down of the process. This might be due to the high levels of Nanog inducing the expression of other pluripotency genes [25] and is consistent with the delayed appearance of lineage specific markers in these cells under differentiation conditions [5].

Figure 6.

Statistics of Tcf3 mutant cells. Correlation plots of the Oct4 and Nanog levels (A, B) and Oct4/Nanog ratios relative to the levels of Nanog (C, D) in single Tcf3+/+ (A, C) and Tcf3−/− (B, D) cells. The blue line was calculated as in Figure 1 for the Tcf3+/+ cells, for comparison the same line has been drawn in (D). (E): Distribution of the values of the Oct4/Nanog ratio across the population.

DISCUSSION

We have shown that the degree of pluripotency of an ESC population can be quantified and represented by a statistics centered around relative measurements of the ratio between Oct4 and Nanog in single cells (Table 1). Together these measurements provide information about the fraction of cells that are in the ground state within a population and the stability of this state. Our results suggest that the correlation between Oct4 and Nanog is inversely related to the probability of differentiation and therefore is a measure of pluripotency. In a population under self-renewal conditions, there is a lower limit to the Oct4/Nanog ratio, which in our experimental setup has a value around 0.5. This limit becomes the most frequent value of the ratio in 2i conditions. It is possible that the existence of such a ratio reflects stoichiometric conditions between Oct4 and Nanog complexes that are optimal for pluripotency, and indeed there have been suggestions that the stoichiometry between Oct4 and Nanog plays a role in the maintenance of pluripotency [26, 27] and also in reprogramming [28].

One important conclusion from our analysis is that while pluripotency, as the sustained ability to generate all differentiated cell types of an embryo, can be traced to a single cell [1], it can only be measured as the tendency of a population to differentiate, that is, pluripotency is a statistical attribute of a population and is inversely related to its differentiation potential [12, 19]. Furthermore, at the single-cell level, pluripotency is tightly associated with the dynamics of a gene regulatory network that allows an exploration of a space of relative Oct4/Nanog levels [17, 19, 29, 30]. Our observation that the statistics of a population reflects the statistics of individual colonies supports the notion that there are rules to the dynamics of the network at the level of the population [12, 19, 30]. In self-renewal conditions, at any given time, there is a small fixed proportion of cells in the ground state and the occupancy of this state is constantly turned over under the control of the culture conditions. The 2i conditions constrain this dynamics and maintain the population in the ground state. The notion that there is a strong relationship between pluripotency and differentiation is underpinned by our analysis of mutants suggests that the pluripotency associated with Tcf3 mutants is due not to the maintenance of a pluripotent state but to a very slow rate of differentiation. Conversely, the absence of β-Catenin results in an instability of the state even in 2i where cells have a high and widespread Oct4/Nanog correlation. This observation suggests that β-Catenin is an important determinant of the dynamics of population and, although it might not be required for pluripotency [6, 16], it is required for its stability and robustness.

An intriguing feature of a population of ESCs is that under self-renewal conditions, the distribution within a colony mirrors that of the population as a whole, while this is no longer the case in ground-state conditions, where each colony has a position within the line. This suggests that, in the ground state, each colony is behaving as a cell and, as every colony is equally pluripotent, this reinforces our conclusion that it is the relative levels rather than the absolute levels of Oct4 and Nanog what determines pluripotency. This observation also suggests the existence of heterogeneities in the ground state, which have however a strong intercolony rather than intracolony component. Heterogeneities in the ground state have been described before [30] and have been used to suggest that cells is this state exhibit some form of differentiation priming. Our observations support this possibility and should encourage further experimental tests of cells in this state.

In summary, our results suggest that the quantifiers that we introduce might provide a hallmark for pluripotency and an objective criterion for the definition of its state. These observations should be of use in the interpretation and evaluation of the state of ESC cultures.

Acknowledgements

We want to thank I. Chambers, A. Smith, R. Kemler, and B. Merril for cell lines; Jeremy Gunawardena for discussions and comments on the manuscript; and Adrian Friday for invaluable discussions on the statistics of populations and statistical analysis in general. We would also like to thank Miguel Angel Luengo Oroz for the script to quantify fluorescence in immunostainings. This work was funded by an Advanced ERC grant to A.M.A. and a Royal Society International Joint Project to A.M.A. and J.G.O. J.G.O. also acknowledges financial support from the ICREA foundation. P.R. is supported by a FI grant from the Generalitat de Catalunya.

DISCLOSURE OF POTENTIAL C ONFLICTS OF INTEREST

The authors indicate no potential conflicts of interest.

Ancillary