LIF-mediated control of embryonic stem cell self- renewal ... - CiteSeerX

The FASEB Journal • Research Communication

LIF-mediated control of embryonic stem cell selfrenewal emerges due to an autoregulatory loop Ryan E. Davey,* Kento Onishi,* Alborz Mahdavi,*,† and Peter W. Zandstra*,†,1 *Institute of Biomaterials and Biomedical Engineering, †Department of Chemical, Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario, Canada Stem cells convert graded stimuli into all-or-nothing cell-fate responses. We investigated how embryonic stem cells (ESCs) convert leukemia inhibitory factor (LIF) concentration into an all-or-nothing cell-fate decision (self-renewal). Using a combined experimental/computational approach we demonstrate unexpected switch-like (on/off) signaling in response to LIF. This behavior emerges over time due to a positive feedback loop controlling transcriptional expression of LIF signaling pathway components. The autoregulatory loop maintains robust pathway responsiveness (“on”) at sufficient concentrations of exogenous LIF, while autocrine signaling and low concentrations of exogenous LIF cause ESCs to adopt the weakly responsive (“off”) state of differentiated cells. We demonstrate that loss of ligand responsiveness is reversible and precedes loss of the ESC transcription factors Oct4 and Nanog, suggesting an early step in the hierarchical control of differentiation. While endogenously produced ligands were insufficient to sustain the “on” state, they buffer it, influencing the timing of differentiation. These results demonstrate a novel switch-like behavior, which establishes the LIF threshold for ESC self-renewal.—Davey, R. E., Onishi, K., Mahdavi, A., Zandstra, P. W. LIF-mediated control of embryonic stem cell self-renewal emerges due to an autoregulatory loop. FASEB J. 21, 2020 –2032 (2007)

ABSTRACT

Key Words: positive feedback 䡠 cell signaling 䡠 Jak-STAT 䡠 threshold 䡠 systems biology

Stem cells must decide between contrasting cell fates: proliferation vs. quiescence, self-renewal vs. differentiation, survival vs. apoptosis. These fates are generally considered to be all-or-nothing (yes/no) responses. All-or-nothing type responses arise from either a switch in exogenous signals presented to the cell (digital input), or a cell-intrinsic switchlike response (digital response) to a graded signal. The latter is considered to arise from the activation of an intracellular cell-fate effector in a sigmoidal (“ultrasensitive”) manner (1). Such ultrasensitive behavior is observed during Xenopus oocyte maturation where the mitogenactivated protein kinase (MAPK) cascade effectively switches between “off” and “on” states over a narrow range of progesterone concentrations (2). 2020

Numerous exogenous signals are known to alter the propensity of mouse embryonic stem cells (mESCs) to self-renew and differentiate (3– 8). At some level, complimentary and contradictory signals must be biochemically integrated and a cell fate decision made. Although key components of the cell fate machinery have been elucidated, it is not known what constitutes the biochemical all-or-nothing switch(es) for this cell fate decision. Switchlike mechanisms may arise at the level of cell signaling pathways or in downstream cell-fate machinery (transcription factor networks), each with unique implications on the nature of stem cell fate control. We investigated whether the Janus kinase— signal transducer and activator of transcription (JakSTAT) pathway produces switchlike behavior in response to LIF. Leukemia inhibitory factor (LIF) is the archetypal self-renewal factor for mESC. LIF belongs to the family of interleukin (IL)-6-type cytokines, which signal through the common receptor subunit gp130, along with a ligand-specific receptor subunit (such as the LIF receptor [LIFR␤]). Gp130 is ubiquitously expressed (9) and plays a role in a variety of cellular processes [reviewed in (10)]. Numerous gp130 ligands are produced during early stages of development, contributing not only to the self-renewal of pluripotent cells but also to the development of numerous differentiated cell types (10 –12). In vitro, gp130 ligands, including LIF are produced by both ESCs and their differentiated progeny (13–15). LIF induces heterodimerization of LIFR and gp130. This results in activation of receptor-associated Janus kinases (Jaks), phosphorylation of receptor docking sites, and recruitment of Src homology-2 (SH2) domain containing proteins such as STAT3. Subsequent phosphorylation of STAT3 (pSTAT3) induces nuclear translocation and activation of transcription. STAT3 activation has been shown to be necessary and sufficient to maintain self-renewal of mouse ESC in the presence of serum or serum replacement (16 –18). To investigate how stem cells convert a graded stimulus into an all-or-nothing cell fate decision, we exam1 Correspondence: Terrence Donnelly Centre for Cellular and Biomolecular Research, 160 College St., 11th Floor, University of Toronto, Toronto ON, Canada, M5S 3E1. Email: [email protected] doi: 10.1096/fj.06-7852com

0892-6638/07/0021-2020 © FASEB

ined the response of ESCs to LIF concentration. Although simulations of the canonical Jak-STAT pathway indicated that the pathway responds in an analog (nonswitchlike) fashion to ligand concentration, ESC signaling was found to evolve toward switch-like (‘on/ off’) behavior in a ligand-concentration-dependent manner. Using a joint experimental and computational approach, we demonstrated that this signaling behavior arose because of the presence of a positive feedback loop, whereby STAT3 activation controlled the expression of STAT3, gp130, and LIFR. ESCs were found to reversibly switch between “on” and “off” pathway states (preceding changes in cell fate), whereas differentiated cells remained fixed in the weakly responsive “off” state. The magnitude and timing of this switchlike behavior were modulated by the buffering effect of endogenously produced ligands. These results identify a new ESC all-or-nothing fate control system, whereby a novel switchlike behavior of the Jak-STAT pathway in conjunction with endogenous ligands establishes the kinetics and threshold for LIF-mediated self-renewal.

MATERIALS AND METHODS

pensation between fluorophore channels was performed using a linear equation generated from controls (similar to flow cytometry). Oct4 positive and negative subpopulations were identified by fitting two Gaussian distributions (OriginPro 7.5; OriginLab) to log-transformed data. Oct4⫺ analysis was restricted to 95% of the Oct4⫺ subpopulation, while Oct4⫹ analysis was restricted to the top 50% of Oct4 expressers. Mathematical model We describe the Jak-STAT pathway as a series of 21 nonlinear ordinary differential equations (Supplemental Fig. 1). This model reflects the basic structure (components and interactions) known to exist within the Jak-STAT pathway. The rate constants used were either taken from similar published systems or determined experimentally (Supplemental Table 1). The set of equations was solved using the Runge Kutta method as implemented by the dde23.m function in MatLab (R2006a; The MathWorks). To this canonical model, we added endogenous ligand (receptor binding and trafficking identical to LIF), pSTAT3-dependent STAT3 transcription, and STAT3 degradation, to examine the possible consequences of autocrine signaling and positive feedback on Jak-STAT pathway behavior. Levels of endogenous ligand were fit to experimental results, and the rate of pSTAT3dependent STAT3 transcription was set to balance STAT3 degradation at 500 pM LIF.

ESC culture R1 (41) ESCs and EF4 (22) (E14Tg2a parental) ESCs constitutively expressing Nanog (generously provided by Dr. Ian Chambers) were routinely cultured in 15% FBS (900 –108; Gemini Biological Products, Calabasas, CA) with 500 pM LIF (mLIF; ESG11–7; Chemicon, Temecula, CA) on gelatincoated tissue-culture flasks, as described previously (42). For all studies, 15% knockout serum replacement (10828 – 028; Invitrogen, Carlsbad, CA) was used unless otherwise noted. Immunocytochemistry and single cell fluorescence microscopy Cells were seeded at 5 ⫻ 103 cells/well in 96 well plates (6005182; Packard) coated with fibronectin (12.5 ␮g/ml; F1141; Sigma-Aldrich) and gelatin (0.02%). The presence of fibronectin does not significantly alter the growth rate or self-renewal/differentiation capacity of ESCs but produces a flattened but compact colony morphology optimal for imaging (15, 43). Cells were treated with LIF, IL-6 (406-ML) plus soluble IL-6 receptor (1830-SR; R&D Systems, Minneapolis, MN), Jak Inhibitor 1 (420099; Calbiochem, San Diego, CA), or controls as indicated. Cells were washed twice before treatment. Each experiment was performed in triplicate wells and was repeated at least three different times unless otherwise noted. Immunocytochemistry was performed as described previously (29). The following antibodies were used: antiphosphoStat3 (tyr705) (9131; Cell Signaling Technology, Danvers, MA); anti-Oct3 (Oct4) (611202; BD Biosciences, San Jose, CA); anti-Nanog (generously provided by Dr. T. Yamanaka); goat anti-rabbit AlexaFluor 488 (A-11034), and goat antimouse AlexaFluor 546 (A-11030; Molecular Probes, Eugene, OR). Antibodies were diluted 1:200 (1:1600 for anti-Nanog). Hoechst was used to identify nuclei (862096; Sigma-Aldrich, St. Louis, MO) (0.1 ␮g/mL). Cells were analyzed using the ArrayScan II automated fluorescent microscope (Cellomics, Pittsburgh, PA). Average pixel intensity (total pixel intensity divided by number of pixels) within the nuclear area of individual cells was determined for each fluorophore. ComAUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

Semiquantitative RT-PCR Total RNA was extracted using GenElute Mammalian Total RNA Miniprep Kit (RTN-70; Sigma-Aldrich). Reverse transcription of the RNA into cDNA was performed using SuperScript II RNase H Reverse Transcriptase kit (18064 – 014; Invitrogen) supplemented with random hexamers (27–216607; Amersham, Little Chalfont, Buckinghamshire, UK), RNAguard (37– 0915-01; Amersham), and 10 mM dNTP mix (184727– 013; Invitrogen). Semiquantitative PCR was subsequently performed for: stat3 (forward: 5⬘-CTT GGG CAT CAA TCC TGT GG-3⬘, reverse: 5⬘-TGC TGC TTG GTG TAT GGG TCT AC-3⬘; 513 base pairs) (44); gp130 (forward: 5⬘-GAT TGC CAG TCA AAC GAT GG-3⬘, reverse: 5⬘-AGC TGA TTT GCC CAC CTT GT-3⬘; 713 bp); socs3 (forward: 5⬘-TGC GCC ATG GTC ACC CAC AGC AAG TTT-3⬘, reverse: 5⬘-GCT CCT TAA AGT GGA GCA TCA TAC TGA-3⬘; 638 bp); oct4 (forward: 5⬘-AAG GTG TCC CTG TAG CCT CA-3⬘, reverse: 5⬘-GAG GAG TCC CAG GAC ATG AA-3⬘; 512 bp); lif-r (forward: 5⬘-CTT CGA TCC TCA ACA CAG AGC-3⬘, reverse: 5⬘-TGG TTA GTG CAC CCA TAG AGG-3⬘; 359 bp) (45); and normalized to ␤-actin. Relative signal strength was determined using densitometry (ChemiImager 5500; Alpha Innotech, San Leandro, CA, USA). Flow cytometry Flow cytometry analysis of STAT3, gp130, and LIFR was performed using IntraPrep reagents (IM2389; Beckman Coulter) as described previously (46). Cells were costained with anti-Oct3 (Oct4) and either anti-STAT3 (9132; Cell Signaling Technology), antigp130 (sc-656) or anti-LIFR antibodies (sc-659; Santa Cruz Biotechnology). Anti-mouse IgG1-PE (550083; BD Biosciences) and anti-rabbit IgG-FITC (IM0833; Beckman Coulter, Fullerton, CA, USA) secondary antibodies were used. Cells were run on a flow cytometer (XL; Beckman Coulter). 2021

Figure 1. Switchlike response to LIF does not appear to arise from a switchlike activation of STAT3: A) Mean Oct4 intensity per cell as measured using quantitative single-cell fluorescence microscopy for R1 ESCs cultured for 72 h in various concentrations of LIF. Mean Oct4 intensity in the absence of LIF was set as zero. Inset: Stimulus-response curves comparing the Michaelian response (nH⫽1) to the more sigmoidal (switchlike) responses of oxygen binding hemoglobin (nH⫽2.8), and MAPK activation in oocytes (nH⫽42). B) Representative histograms of Oct4 levels (log-transformed) at 72 h in response to various concentrations of LIF by quantitative single-cell fluorescence microscopy. C) Representative images from quantitative single-cell fluorescence microscopy showing a) nuclei, b) image analysis masks identifying cells based on nuclear stain, c) Oct4, d) identification of Oct4⫹ and Oct4⫺ subpopulations, e) pSTAT3, f) Oct4⫹ subpopulation masks applied to pSTAT3. Scale bar ⫽ 20 ␮m. D) Mean pSTAT3 intensity per cell within the Oct4⫹ subpopulation for cells cultured for 72 h in various concentrations of LIF. Mean pSTAT3 intensity in the absence of LIF was set as zero. Inset: mathematical simulation of STAT3 activation in response to LIF concentration. E) Representative histograms of pSTAT3 levels (log-transformed) at 72 h in response to various concentrations of LIF. All results are representative of at least three independent experiments. Error bars represent sd of three wells, each comprising 5–10 ⫻ 103 single-cell measurements.

RESULTS The Jak-STAT pathway does not appear to exhibit classic switchlike behavior ESCs convert LIF concentration into an all-or-nothing cell fate decision (self-renewal vs. differentiation). As a first step toward determining the mechanism responsible for converting LIF concentration into an all-ornothing fate decision, we examined the stimulus-response profile of LIF on Oct4 expression (a molecular determinant of the ESC state (19, 20)). Given that self-renewal is an all-or-nothing fate decision, we would expect this relationship to appear steeply sigmoidal (switchlike): some concentrations of LIF should support self-renewal (Oct4 “on”) and some concentrations 2022

Vol. 21

July 2007

should not (Oct4 “off”). However, after 72 h of culture, the stimulus-response profile of LIF on Oct4 expression shows a graded relationship with an apparent Hill coefficient (nH) of approximately one (nH⫽1.02⫾0.07) (Fig. 1A), typical of a Michaelian-type response (nH⫽1). Compared to a Michaelian stimulus-response profile switchlike responses are more sigmoidal (nH⬎1), effectively changing between off and on states over a narrower range of concentrations [Fig. 1A inset compares the Michaelian response with the sigmoidal profiles observed for oxygen binding hemoglobin (nH⬇2.8 (21)) and MAPK activation in Xenopus oocytes (nH⬇42 (2))]. The reason for the graded relationship between Oct4 expression and LIF concentration is that this measure represents a population average. At the population level, an apparently Michaelian response could

The FASEB Journal

DAVEY ET AL.

be generated by either a graded change in the level of Oct4 expression in individual cells, or alternatively, a change in the proportion of “on” (Oct4⫹) to “off” (Oct4⫺) cells. Indeed, LIF induced self-renewal is an all-or-nothing cell-fate decision; an examination of Oct4 levels in single cells demonstrates a bimodal profile where cells tend to be either “on” or “off”, reflective of the ultrasensitive switchlike relationship (Fig. 1B,C). For this switchlike response to appear Michaelian at the population level indicates variability in the sensitivity of cells to stimuli, or in the timing of individual cell responses, as not all cells switch from the Oct4⫹ to the Oct4⫺ state in concert. To determine whether the observed switchlike relationship between LIF and self-renewal could be generated by a switchlike activation of the Jak-STAT pathway in response to LIF, we measured STAT3 activation in the Oct4⫹ subpopulation in response to LIF concentration. For this analysis, the Oct4⫹ subpopulation was stringently defined as the top 50% of Oct4 expressing cells in order to exclude differentiating cells in transition between the Oct4⫹ and Oct4⫺ states (which varies with LIF concentration). The resulting subpopulation response profile demonstrates a Michaelian relationship closely matching that observed for Oct4 (nH⫽1.06⫾0.19) (Fig. 1D), illustrating a strong correlation between LIF concentration, STAT3 activation, and Oct4 expression. However, in contrast to the switchlike relationship between LIF concentration and Oct4 expression in single cells, examination of pSTAT3 levels in single Oct4⫹ cells suggests that the relationship between LIF concentration and STAT3 activation is in fact graded, with no obvious bias toward low (off) and high (on) STAT3 signaling states (Fig. 1E,C). Similar results were also observed on analysis of the total population, or the top 95% of the Oct4⫹ subpopulation (not shown). These results suggest that switchlike behavior responsible for switchlike expression of Oct4 (and the all-or-nothing nature of this cell-fate decision) occurs downstream of STAT3 activation. To investigate whether the known structure of the Jak-STAT pathway produces switchlike behavior, we simulated the response of an individual cell to a range of LIF concentrations using a mathematical representation of the pathway (description available in Supplemental Online Material). In agreement with the above results, simulations suggest that the Jak-STAT pathway would not necessarily be expected to respond in a switchlike fashion to LIF (nH⫽0.78⫾0.02) (Fig. 1D, inset). These results suggest that the Jak-STAT pathway does not contribute to the all-or-nothing nature of the cell-fate response to LIF, but instead passes information to downstream processes in an analog fashion. To verify this, we examined the response of the Jak-STAT pathway to LIF concentration at a time preceding overt differentiation to ensure that the observed STAT3 response to LIF concentration reflects normal pathway behavior rather than a consequence of differentiation processes. AUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

Response of the Jak-STAT pathway to LIF has a temporal dependency To investigate the response of the Jak-STAT pathway to LIF at a time prior to changes in Oct4 expression and cell fate, we examined cells after 6 h exposure to various concentrations of LIF. At this time, no differences in Oct4 expression or cell fate are apparent between LIF concentrations (demonstrated below). As was seen at 72 h, an examination of the distribution of STAT3 activation levels in Oct4⫹ cells at 6 h shows no obvious switchlike behavior as the STAT3 distribution appears unimodal (Fig. 2A). However, STAT3 activation is greater at low LIF concentrations for cells stimulated for 6 h compared with 72 h. A comparison of pSTAT3 population averages for 6 h and 72 h indicates significantly greater STAT3 activation for low and zero LIF concentrations at 6 h (Fig. 2B). The difference in STAT3 activation between these time points in the absence of exogenous LIF cannot be explained by the kinetics of STAT3 dephosphorylation, as the half-life of pSTAT3 is ⬃10 min (our observation). These results therefore indicate the presence of endogenous STAT3 activation that is greater at 6 h than 72 h. Recently, we demonstrated that ESCs express functional levels of autocrine gp130 ligands that temporarily maintain STAT3 activation and self-renewal in the absence of exogenous LIF (15). By using a Jak inhibitor to inhibit all endogenous Jak-dependent STAT3 activity, it is apparent that significant endogenous activity is indeed present at 6 h in the absence of LIF, while less endogenous activity is present at 72 h (Fig. 2B green and blue points). This endogenous activity alters the response of the Jak-STAT pathway to exogenous LIF at 6 h by propping up STAT3 activation at low and zero LIF concentrations, while not affecting STAT3 activation at high levels of exogenous LIF. The role of this autocrine factor(s) in altering the dose-response profile was examined by including an autocrine gp130 ligand (identical to LIF) in our simulations. This analysis suggests that an endogenous gp130 ligand could alter the shape of the LIF response profile in the manner observed experimentally (Fig. 2C). Interestingly, these results illustrate a temporal dependency in the LIF-pSTAT3 dose-response profile. Between 6 h and 72 h, the pathway appears to respond in a more switchlike manner; differences in STAT3 activation between LIF concentrations become exaggerated, and substantial levels of endogenous activity are lost. This behavior may arise from a change in the level of endogenous signal over time or a change in the response of the pathway to the endogenous signal. Although both possibilities may account for the apparent switchlike behavior of Jak-STAT3 signaling, they represent fundamentally different control mechanisms. As a first step toward distinguishing between these possibilities and to investigate the consequences of this change in STAT3 signaling behavior, we examined the kinetics of STAT3 signaling on LIF deprivation. 2023

Figure 2. Endogenous STAT3 activity alters the LIF-pSTAT3 dose-response in a temporally dependent manner. A) Representative histograms of pSTAT3 levels (log-transformed) at 6 h in response to various concentrations of LIF. B) Mean pSTAT3 intensity per cell within the Oct4⫹ subpopulation for cells cultured for 6 h (red) or 72 h (black) in various concentrations of LIF, or cultured in the absence of LIF and presence of Jak inhibitor 1 (1000 nM) for 6 h and 72 h (cyan and blue points). Mean pSTAT3 intensity in the absence of LIF at 72 h was set as zero. Error bars represent sd of three wells, each comprising 5–10 ⫻ 103 single-cell measurements. All results are representative of at least three independent experiments. C) Mathematical simulations of STAT3 activation in response to LIF concentration in the presence (red) and absence (black) of endogenous ligand (identical to LIF).

ESCs gradually lose STAT3 activation in the absence of LIF and begin to signal like differentiated cells To understand the mechanism underlying the appearance of switchlike behavior between 6 h and 72 h we first examined the kinetics of STAT3 signaling in the Oct4⫹ and Oct4⫺ subpopulations in the absence of LIF. On exogenous LIF deprivation pSTAT3 levels in the Oct4⫹ subpopulation undergo a rapid decrease to the (nonzero) endogenous level (Fig. 3A). This rapid decrease occurs over the first hour of LIF deprivation and reflects the short half-life of STAT3 phosphorylation. Over the next 72 h, pSTAT3 levels undergo a second much more gradual decrease (Fig. 3A). This gradual decrease in pSTAT3 levels in the Oct4⫹ subpopulation occurs in concert with a reduction in expression of both Oct4 and Nanog [a molecular determinant of the ESC state (22, 23)] (Fig. 3B,C) and precedes an increase in the number of Oct4⫺ cells (Fig. 3D). Interestingly, in contrast to the Oct4⫹ subpopulation, the Oct4⫺ subpopulation was generally weakly responsive to both endogenous and exogenous gp130activating stimuli (Fig. 3A, blue and light blue). Furthermore, endogenous STAT3 activity in the Oct4⫹ subpopulation appears to decrease to that of the Oct4⫺ subpopulation in the absence of exogenous LIF. Examination of representative pSTAT3 histograms at 72 h illustrates this further (Fig. 3E–G). Although there is variability in the response of cells in each subpopulation, distinct Oct4⫹ (red) and Oct4⫺ (blue) STAT3 signaling signatures are apparent at 72 h in the presence of LIF (Fig. 3F) but not in the absence (Fig. 3G). In the absence of LIF STAT3 activity in the Oct4⫹ subpopulation decreases to the level exhibited by Oct4⫺ cells (Fig. 3G). Cells falling between the Oct4⫹ and Oct4⫺ subpopulation cutoffs (a population consisting of both undifferentiated and differentiating cells) 2024

Vol. 21

July 2007

demonstrate an intermediary level of STAT3 activation in the presence of LIF and a similar drop in its absence [Fig. 3G,F (olive green)]. Collectively, these observations raise the intriguing possibility that ESCs may gradually lose responsiveness to Jak-STAT-activating ligands in the absence of LIF and adopt the weakly responsive state of differentiated cells prior to the loss of Oct4 and Nanog. Such a change in responsiveness could explain the increase in switchlike behavior observed between 6 h and 72 h and would suggest that this switchlike behavior establishes the ligand threshold for self-renewal-supportive (ESClike) and nonsupportive (differentiated cell-like) levels of STAT3 activation. To investigate the possibility that responsiveness to Jak-STAT activating ligands is controlled in a ligand-dependent fashion, we tested the responsiveness of cells to gp130 ligands after various durations of LIF deprivation. Responsiveness to gp130 ligands is regulated by prior ligand availability To examine whether the Oct4⫹ subpopulation downregulates responsiveness to gp130 ligands in the absence of LIF, we deprived cells of LIF for various lengths of time before LIF reintroduction. As anticipated, with increasing duration of LIF deprivation we observed a decrease in the response of Oct4⫹ cells to LIF restimulation (Fig. 4A, black) (this was also verified by flow cytometry). The response of the Oct4⫹ subpopulation became more similar to that of the weakly responsive Oct4⫺ subpopulation with time (Fig. 4A, red). No equivalent decrease in LIF responsiveness was observed in the Oct4⫺ subpopulation. These results indicate that on LIF deprivation, the Oct4⫹ subpopulation down-regulates LIF responsiveness, becoming

The FASEB Journal

DAVEY ET AL.

Figure 3. ESCs gradually lose STAT3 activation in the absence of LIF and respond like differentiated cells: A) Mean pSTAT3 intensity per cell within the Oct4⫹ subpopulation for cells cultured for 72 h in the presence (black) or absence (red) of LIF, and for the Oct4⫺ subpopulation in the presence or absence of LIF (blue and light blue). Results normalized to Oct4⫹ subpopulation level in the presence of LIF. B) Mean Oct4 intensity per cell of the Oct4⫹ subpopulation for cells cultured over 72 h in the presence (black) or absence (red) of LIF. Mean Oct4 intensity per cell in the Oct4⫺ subpopulation was set as zero. Results at each time were normalized to the ⫹LIF condition. C) Mean Nanog intensity measured in the Oct4⫹ subpopulation the same as in (B). D) Number of Oct4⫺ cells in cultures maintained over 72 h in the presence (black) or absence (red) of LIF. E) Representative Oct4 histogram illustrating subpopulation selection criteria based on fitting Gaussian distributions (red) to log-transformed data: Oct4⫹ (⫹, top 50% of the Oct4⫹ subpopulation), Oct4⫺ (–, bottom 95% of the Oct4⫺ subpopulation), and intermediate (in between ⫹ and –). Subpopulations were used in (F) and (G). F and G) Representative pSTAT3 histograms for the total population (black), Oct4⫺ subpopulation (blue), Oct4⫹ subpopulation (red), and for intermediate cells (olive green) in the presence (F) or absence (G) of LIF. All results are representative of at least three independent experiments. Time points were performed in parallel. Error bars represent sd of three wells, each comprising at least 5–10 ⫻ 103 single-cell measurements for the Oct4⫹ subpopulation and at least 1 ⫻ 103 for the Oct4⫺.

like the weakly responsive Oct4⫺ subpopulation before the complete loss of Oct4. Similar results were observed for E14Tg2a ESCs and for cells cultured in the presence of serum (data not shown). The diminishing response of cells to the endogenous signal (Fig. 3A and Fig. 4A “0 min” time points) mirrors that of exogenous LIF, indicating that loss of responsiveness affects endogenous gp130 ligands as well. To test this, we examined the response of cells to IL-6 plus soluble IL-6 receptor, which signals through the common gp130 receptor subunit without requiring additional ligand-specific receptor subunits. Testing the response of cells to IL-6 plus soluble IL-6 receptor thereby allows us to examine the response of cells to all IL-6-type cytokines. As expected, the response of cells to IL-6 demonstrated a similar decrease with increasing duration of LIF deprivation (Fig. 4A, blue). AUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

The above results suggest that the switchlike behavior generated between 6 h and 72 h arises from a LIF concentration-dependent change in gp130 ligand responsiveness with time. Furthermore, this change precedes loss of the stem cell state. To examine whether the change in responsiveness occurs in a LIF concentration-dependent manner, cells were cultured in various LIF concentrations for 66 h, followed by 6 h LIF deprivation, and 15 min stimulation with 500 pM LIF. The results demonstrate a clear dose-dependent relationship between the LIF concentration cells, which were originally cultured in (for 66 h), and their subsequent ability to respond to saturating levels of stimulation (500 pM LIF for 15 min) (Fig. 4B). These results indicate that Jak-STAT pathway components may be autoregulated in a ligand concentration-dependent manner. 2025

Figure 4. Responsiveness to gp130 ligands is regulated by ligand concentration: A) ESCs were deprived of LIF for 24 h, 48 h, or 72 h before restimulation with either 500 pM LIF or IL-6 (50 ng/mL) plus soluble IL-6 receptor (1 ␮g/mL) for 1 h. Mean pSTAT3 intensity was measured in the Oct4⫹ (black) and Oct4⫺ (red) subpopulations in response to LIF, and in the Oct4⫹ subpopulation in response to IL-6 plus sIL-6 receptor (blue). LIF controls are shown in orange. B) Cells were cultured for 66 h in various concentrations of LIF, deprived of LIF for 6 h, and restimulated with 500 pM LIF for 15 min. Time points were performed in parallel. All results are representative of at least three independent experiments. Error bars represent the sd of three wells, each comprising 5–10 ⫻ 103 single-cell measurements.

Autoregulation controls responsiveness to gp130 ligands The mechanism mediating autoregulatory control over gp130 ligand responsiveness may involve pSTAT3-dependent expression of JAK-STAT pathway components. Functional pSTAT3 binding sites have been identified in the promoters of both STAT3 and gp130 and are predicted in LIFR (24 –26). Furthermore, a positive feedback control loop where pSTAT3 regulates expression of STAT3 and gp130 was recently demonstrated to arise during astrogliagenesis (27). To examine whether autoregulation occurs in ESCs, we examined the expression of STAT3; gp130; LIFR; the pSTAT3-regulated inhibitor suppressor of cytokine signaling 3 (SOCS3) (28); and Oct4; in the presence and absence of LIF, as well as in the presence of JAK inhibitor. As shown above (Fig. 2B), these conditions result in different levels of STAT3 activation, thereby, allowing us to investigate 2026

Vol. 21

July 2007

the possibility of pSTAT3 dose-dependent control over gene expression. Semiquantitative RT-PCR results at 6-h exposure demonstrate a dose-dependent relationship between STAT3 activation and expression of STAT3, gp130, and LIFR. Expression was highest in the presence of LIF, weaker in the absence of LIF, and weakest in the presence of JAK inhibitor (Fig. 5A, B). Dose dependency was also observed at both 3 h and 12 h (data not shown). A similar dose dependency was observed for SOCS3, a known pSTAT3 transcriptional target (28), while no differences in Oct4 expression were observed. Examination of STAT3, gp130, and LIFR protein levels in the Oct4⫹ subpopulation by flow cytometry confirmed the observed changes in gene expression. Differences in STAT3 levels between ⫹LIF and –LIF were apparent as early as 12 h and increased with time (Fig. 5C). We have also observed a small decrease in LIF responsiveness at as early as 12 h (data not shown). Interestingly, although differences in gp130 and LIFR levels exist, they were first apparent much later, around 48 h to 60 h (Fig. 5D, E). Further to this, the Oct4⫺ subpopulation demonstrated reduced expression of STAT3, gp130, and LIFR relative to their Oct4⫹ counterparts (Fig. 5F–H, respectively). Changes in protein levels were confirmed via immunocytochemistry and single cell image analysis (data not shown). Together with the gene expression data, these findings suggest the presence of a positive feedback loop controlling the expression of JAK-STAT pathway components, including STAT3, LIFR, and gp130 (Fig. 5I) and that this positive feedback loop may be responsible for the change in the LIF-STAT3 stimulus-response profile over time. To explore the effects of positive feedback on the response of the Jak-STAT pathway to ligand concentration, we simulated this new pathway structure. Positive feedback enhances the switchlike behavior of the Jak-STAT pathway creating a reversible switch To establish that positive feedback can account for the difference in the LIF dose-response between 6 h and 72 h, we performed simulations where we included both autocrine signaling and positive feedback control over STAT3 expression (pSTAT3-dependent STAT3 transcription) (description of adapted model available in supplemental online material). Resulting simulations suggest that in the presence of a constant autocrine signal, positive feedback can indeed alter the behavior of the Jak-STAT pathway and cause a gradual steepening of the LIF-STAT3 stimulus response curve with time (Fig. 6A, lines). These results fit well with what we observe experimentally at 6 h, 24 h, 48 h and 72 h (Fig. 6A, points). By performing these simulations in the absence of an endogenous signal, it is more readily apparent that the Jak-STAT pathway is evolving a more switchlike response to LIF concentration with time (Fig. 6B). The degree of switchlike (sigmoidal) behavior achieved in these simulations necessarily depends

The FASEB Journal

DAVEY ET AL.

Figure 5. Autoregulatory control of STAT3, gp130, and LIFR expression: A) Semiquantitative RT-PCR for ␤-actin (control), Oct4, SOCS3, STAT3, gp130, and LIFR in the presence of LIF, absence of LIF, or absence of LIF and presence of JakI1 (1000 nM). B) Densitometry for (A). C–E) Representative flow cytometry histograms for STAT3 (C), gp130 (D), and LIFR (E) protein levels in the Oct4⫹ subpopulation in the presence (shaded) or absence (outline) of LIF over time. Time (in h) is inlaid. F–H) Histograms showing STAT3 (F), gp130 (G), and LIFR (H) protein levels in the Oct4⫹ (shaded) or Oct4⫺ (outline) subpopulation in the presence of LIF. Oct4⫹ and Oct4⫺ subpopulations were defined as shown in (i). Nonspecific staining control is shown in (ii). I) Schematic representation of positive feedback loop controlling expression of pathway components. All results are representative of at least three independent experiments.

on the rates of STAT3 production and degradation, as well as the duration of simulation. These results suggest that the Jak-STAT pathway is responding in a switchlike manner to LIF. The apparent Hill coefficient determined for the LIF-pSTAT3 dose-response profile at 72 h (Fig. 1) is obfuscated by not considering the fact that the pathway at 72 h has become unresponsive to substantial levels of endogenous signal. To estimate the actual degree of switchlike behavior attained in vitro by 72 h, we approximated the level of autocrine activity and refit the 72 h data. We approximated the level of autocrine ligand to be equivalent to 5–10 pM exogenous (bulk) ligand by comparing the dose-response profiles in Fig. 6A. This autocrine ligand approximation also matches the concentration used in our simulations in Fig. 6A (10 pM), which fit well with experimental data. By estimating the autocrine concentration and refitting the data, the LIFpSTAT3 dose-response profile at 72 h exhibits an apparent Hill coefficient of ⬃3.0 to 4.7 (the binding of oxygen to hemoglobin has an nH of 2.8). Irrespective of the true extent of switchlike behavior attained by the pathway, the evolution of this behavior AUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

has a significant impact on STAT3 activity. The switch effectively establishes a ligand threshold above which STAT3 activation is maintained and below which STAT3 activity is gradually lost (effectively filtering out substantial levels of STAT3 activation arising from endogenous signal and low levels of exogenous LIF), causing cells to adopt a STAT3 activation state similar to differentiated cells. Interestingly, because ESCs under these conditions double approximately every 16 h (29), this switchlike behavior is established over a time period where cells have divided multiple times. Positive feedback has been shown in other systems to lead to irreversible changes in pathway activation (30). Indeed, differentiated cells appear locked in the poorly responsive state, indicating that at some point loss of responsiveness is irreversible. Such signaling behavior could be a causative factor in the irreversibility of ESC differentiation. To address this question, we examined whether positive feedback in the Jak-STAT pathway could generate irreversible changes in responsiveness. Model simulations indicated that loss of STAT3 on complete ligand removal is reversible by ligand reintroduction provided the total number of STAT3 molecules 2027

Figure 6. Positive feedback enhances switchlike behavior over multiple cell divisions creating a reversible switch: A) Mean pSTAT3 intensity per cell within the Oct4⫹ subpopulation for cells cultured in various concentrations of LIF for 6 h (black points), 24 h (red points), 48 h (blue points) or 72 h (green points). Mean pSTAT3 intensity in the absence of LIF was set as zero. Means were normalized to plus LIF values at each time point. Results are representative of at least three independent experiments. Curves represent mathematical simulations illustrating the effects of endogenous ligand and positive feedback on STAT3 activation over time. Simulations included pSTAT3-dependent STAT3 transcription (positive feedback), STAT3 degradation, and an endogenous ligand (10 pM). STAT3 production and degradation were balanced such that a constant level of STAT3 molecules was maintained in the presence of 500 pM LIF. B) Simulations illustrating the switchlike effects of positive feedback on the shape of the dose-response profile over time in the absence of endogenous ligand at 6 h (red) and 72 h (black), compared to simulations lacking both endogenous ligand and positive feedback (blue). C) Simulations illustrating the effects of decreasing endogenous ligand concentration [100 pM (green), 10 pM (blue), 1 pM (red), 0 pM (black)] on the rate and magnitude of STAT3 loss on LIF withdrawal (%). Positive feedback was included in the simulations. 500 pM LIF was returned in the 0 pM case (magenta) to examine reversibility. D) Mean pSTAT3 intensity per cell within the Oct4⫹ subpopulation for cells constitutively expressing Nanog. Cells were cultured in LIF for 10 days (black), deprived of LIF for 10 days (green), cultured in JakI1 (3 ␮M) for 10 days (pink), cultured in LIF for 5 days before 5 days LIF withdrawal (olive green), or cultured for 5 days in the absence of LIF before 5 days LIF reintroduction (red), or cultured in the absence of LIF and presence of JakI1 (3 ␮M) for 5 days before 5 days LIF reintroduction (blue). Cells were then cultured in the absence of LIF for 6 h before restimulation with 500 pM LIF for 1 h. E) Light micrographs of cells cultured in LIF for 5 days before 5 days LIF withdrawal (⫹LIF to –LIF), or cultured for 5 days in the absence of LIF before 5 days LIF reintroduction (–LIF to ⫹LIF). Results are representative of two independent experiments. Error bars represent the sd of three wells, each comprising 5–10 ⫻ 103 single-cell measurements.

remains above zero [Fig. 6C (black) shows number of STAT3 molecules dropping to 1 before being restimulated and recovering (magenta)]. If the STAT3 count falls to zero, irreversibility ensues (due to a disruption of the positive feedback loop). The presence of autocrine signal prevents this possibility by restricting the magnitude of STAT3 loss in the absence of LIF (Fig. 6C green, blue, red). Not only does the autocrine signal determine the magnitude of the difference between ‘on’ and ‘off’ states, it also determines the rate of change between states (Fig. 6C green, blue, red). To verify the model prediction that loss of responsiveness is reversible, as well as to confirm that switchlike behavior is not a consequence of differentiation, we utilized cells constitutively expressing Nanog (22). Nanog-expressing cells remain undifferentiated in the absence of LIF, thereby enabling us to decouple STAT3 signaling from cell fate control and in doing so examine switch behavior and reversibility in ESCs after prolonged LIF deprivation. Nanog expressing cells cultured in the presence of LIF maintained high LIF responsiveness (Fig. 6D black). On LIF deprivation, responsiveness was lost (Fig. 6D, green and olive green); however, subsequent culturing in the presence of LIF recovered responsiveness (Fig. 6D, red). Responsiveness could also be recovered for cells treated with Jak inhibitor to remove all measurable endogenous STAT3 activity (Fig. 6D, pink and blue). Despite switches in responsiveness, undifferentiated morphol2028

Vol. 21

July 2007

ogy (Fig. 6E) and high Oct4 expression (not shown) were maintained. These results demonstrate that irreversibility is not a consequence of the positive feedback loop itself but arises instead because of some event associated with differentiation.

DISCUSSION The biochemical basis by which stem cells convert graded stimuli into all-or-nothing cell-fate decisions is not understood. The all-or-nothing nature of this response could be established by switchlike (“on/off”) behavior of cell signaling pathways and/or downstream cell-fate machinery. Moreover, multiple switches may be used to control a cell-fate decision. Recently, a computational analysis of the feedback loops controlling Nanog, Oct4, and Sox2 expression suggested switchlike behavior is theoretically possible within this network (31). We investigated how ESCs generate an all-or-nothing cell fate decision in response to LIF. Although simulations of Jak-STAT signaling suggested that known pathway structure would not produce switchlike behavior (Fig. 1D, inset), subsequent experimental and computational analysis indicated that switchlike behavior gradually evolves over time due to the presence of a positive feedback loop and the buffering effect of endogenous ligands. We have demonstrated that ESCs express functional

The FASEB Journal

DAVEY ET AL.

Figure 7. Emergence of switchlike behavior in ESCs: STAT3 activation is buffered at low concentrations of exogenous LIF (solid circles) by endogenous gp130 ligands (open circles). However, endogenous ligands and low levels of exogenous LIF are insufficient to maintain expression of Jak-STAT pathway components, and pathway responsiveness is lost with time. Consequently, the pathway switches from a self-renewal supportive (ON) to an unsupportive (OFF) state.

levels of gp130 activating ligands that act to temporarily support STAT3 activation and self-renewal in the absence of exogenous LIF (15). Here, we report that this endogenous activity props up STAT3 activation in the absence (or at low concentrations) of exogenous LIF. However, endogenous STAT3 activity was found to decrease with time in the absence of LIF, and ESCs gradually adopted the weak STAT3 activation state characteristic of differentiated cells prior to the loss of Oct4 and Nanog. This decrease in endogenous STAT3 activity arose from the presence of a positive feedback loop controlling the expression of STAT3, gp130, and LIFR (Fig. 5), and thereby, responsiveness to gp130 ligands in a ligand concentration-dependent manner (Fig. 4B). The emergence of switchlike behavior is described in Fig. 7. In support of a positive feedback loop controlling pathway responsiveness, functional STAT3-binding elements have been identified in the promoters of STAT3 (24) and gp130 (25) and are predicted to exist in LIFR (26) and JAK 1 (27). Furthermore, a positive feedback loop linking STAT3 activation to pathway responsiveness was recently demonstrated to up-regulate LIF responsiveness during astroglialgenesis, an effect attributed to the up-regulation of STAT3, gp130, and Jak1 expression (27). The observed decrease in STAT3 protein in the absence of LIF correlated well with the decrease in LIF responsiveness and preceded changes in gp130 and LIFR, indicating that STAT3 expression may be the prime determinant of ligand responsiveness during the initial phase of LIF deprivation. As STAT3 levels and ligand responsiveness decreased on LIF deprivation, so too did expression of the stem AUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

cell transcription factors Oct4 and Nanog, suggesting a tight correlation between level of STAT3 activation and self-renewal processes. Levels of STAT3 activation and Oct4 expression also correlate in single cells in the presence and absence of LIF (15). Importantly, cells constitutively expressing Nanog [which maintain an undifferentiated state independent of STAT3 (22, 23)] also experience a drop in responsiveness in the absence of LIF, indicating that this control loop can occur independent of changes in stem cell fate. These results imply that decreased ligand responsiveness is a cause rather than a consequence of the differentiation process. The finding that ESCs actively regulate responsiveness to gp130 ligands by autoregulating STAT3 expression provides insight into the role of Wnt signaling in self-renewal. Wnt/␤-catenin signaling has been shown to promote ESC self-renewal, but this requires the presence of low levels of LIF (32, 33). Wnt and LIF actually synergize to promote self-renewal; low levels of LIF that are otherwise insufficient to maintain selfrenewal can do so in the presence of Wnt or constitutively active ␤-catenin. Interestingly, a 2- to 3-fold increase in STAT3 expression and protein is observed in the presence of Wnt or constitutively active ␤-catenin (32). Our findings suggest that in the absence of threshold levels of LIF, STAT3 protein becomes limiting. Indeed, STAT3 activity arising from endogenous ligands is lost with time due to insufficient maintenance of STAT3 expression. Our results would therefore predict that Wnt/␤-catenin- mediated STAT3 transcription would enhance STAT3 activation, enabling subthreshold levels of ligand to maintain self-renewal. In fact, even in the presence of the GSK3␤ inhibitor BIO, which supports self-renewal by releasing ␤-catenin from GSK3␤ inhibition (6), increased STAT3 reporter activity was observed in the absence of any exogenous LIF (33). It is likely that this is a result of increased activity of endogenous gp130 ligands due to increased STAT3 transcription. To explore the consequences of a positive feedback loop on Jak-STAT pathway behavior we performed simulations. Simulations indicated that the presence of a positive feedback loop could enable the Jak-STAT pathway to evolve over time in the manner observed (Fig. 6A) by enhancing the switchlike behavior of the pathway (Fig. 6B). In other systems, positive feedback has been shown capable of creating irreversibility in switchlike responses (30). However, irreversibility does not arise as a natural consequence of positive feedback in this system (Fig. 6C); undifferentiated cells are capable of reversibly switching between responsive and unresponsive states (Fig. 6D). These results fit well with previous work demonstrating that LIF responsiveness could be up-regulated from a weakly responsive state during astrogliagenesis (27). Interestingly, the Oct4⫺ subpopulation retains low responsiveness in the presence of LIF, implying that the switch eventually becomes fixed in a suppressed state. Such suppression could result from repression of STAT3, gp130, and 2029

LIFR expression through epigenetic mechanisms (i.e., changes in methylation and acetylation), or expression of antagonistic transcription factors such as neurogenic basic-helix-loop-helix (bHLH) factors, as has been observed in other cell contexts (27, 34, 35). Although this switch-like behavior does not represent a “perfect” on/off switch, it has a profound effect on STAT3 signaling, capable of filtering out considerable levels of stimulus. At subthreshold levels of ligand (i.e., endogenous ligand levels and low levels of exogenous LIF) responsiveness cannot be maintained and ESCs adopt the weak STAT3 activation state observed for differentiated cells. This low level of activation is insufficient to maintain self-renewal, and differentiation proceeds. The gradual loss of STAT3 activity at subthreshold concentrations of ligand may explain why autocrine ligands only temporarily support self-renewal (buffer against differentiation) in the absence of exogenous LIF (15). Even though the endogenous signal generates considerable levels of STAT3 activation (Fig. 2B), this activity does not fully maintain STAT3, gp130, and LIFR expression (Fig. 5A,C), and STAT3 activation consequently decreases with time (Fig. 3A and Fig. 6A). Indeed, simulations suggest that the autocrine signal acts as a buffer, influencing both the magnitude of the switch and rate of switching between “on” and “off” states but is not sufficient to prevent switching (Fig. 6C). Combined with the buffering effect of autocrine ligands, the switchlike behavior thereby establishes the LIF threshold for self-renewal supportive levels of STAT3 activation (“on”), as well as the kinetics of switching between “on” and “off” states. Differences in STAT3 activation between Oct4⫹ and Oct4⫺ subpopulations suggests that the different effects of gp130 ligands on undifferentiated and differentiated subpopulations may be a consequence of STAT3 signal strength. Examination of STAT3-dependent gene expression for ESCs in the “on” and “off” Jak-STAT signaling states would be informative in this regard. Positive feedback in the Jak-STAT pathway produces a switchlike response to LIF in ESCs. This mechanism establishes the ligand threshold for ESC-supportive “on” and nonsupportive “off” STAT3 activation states. On/off information is presumably then passed on to downstream processes to be integrated with additional signals before a final cell-fate decision is made. The biochemical switch(s) for self-renewal may therefore exist, at least in part, at the level of cell signaling rather than exclusively in downstream cell-fate processes. This has implications on the hierarchy of events leading to changes in cell fate. By ⬃36 h, LIF deprivation STAT3 activation in the Oct4⫹ subpopulation becomes switched to the “off” state. This precedes loss of Oct4 and Nanog. Recently, it was shown that STAT3 mediates self-renewal via activation of c-myc expression (36). In agreement with our results, the authors describe a similar loss of c-myc expression by 36 h in the absence of LIF. Interestingly, stem cell commitment to differentiation also appears to occur around 36h LIF deprivation, as differentiation of a subset of cells proceeds even 2030

Vol. 21

July 2007

on LIF reintroduction [unpublished observations, and (37)], and ESCs deprived of LIF for 48 h lose their ability to contribute to chimeras (38). Furthermore, a recent study identifying putative downstream targets of STAT3 in ESCs showed changes in the expression of STAT3 targets appearing after 36 h of LIF (or conditionally active STAT3) withdrawal (39). One verified target, the putative zinc-finger protein jumonji domain containing protein 1A (Jmjd1a), showed a decrease in expression beginning between 24 h and 48 h and was completely lost by 72 h following LIF withdrawal. Jmjd1a was also downregulated during differentiation of human ESCs, and its expression appeared restricted to testis and ESCs, further suggestive of a role in pluripotency. The kinetics of these processes suggests that the Jak-STAT switch may serve to establish the timing of commitment to differentiation by setting the window of self-renewal supportive gene expression. Our data suggest that cell-fate decisions are initiated and restricted by progressive changes in cell sensitivity to exogenous stimuli. Such a scenario may be advantageous; for example, adoption of an “off” Jak-STAT signaling state in ESCs (or their pluripotent founder cells) may be necessary to initiate differentiation and stabilize it in the face of multiple gp130 ligands produced by both undifferentiated and differentiated cells in vivo (11), and in vitro (13–15). This work also has implications on the kinetics of differentiation. Generation of switchlike behavior (and differentiation) occurs over a length of time where cells have divided multiple times, thereby enabling stem cell populations to expand prior to differentiation. Interestingly, our results suggest that the timing and magnitude of the switchlike behavior appear to be set by the interaction between the rate of positive feedback and the buffering effect of endogenous signal. Microenvironmental heterogeneity in endogenous signaling may be compounded by these processes to create the spatial organization of pSTAT3 and cell fate observed within ESC colonies (an “autoregulatory ESC niche”) (15). Given the variety of exogenous factors affecting self-renewal and differentiation, and the elusiveness of a conserved stem cell transcriptional program (40), ordering the variety of biochemical processes affecting stem cell fate into networks of switchlike modules (such as the Jak-STAT module we describe) may be a useful abstraction for dissecting and controlling the decisionmaking process. Considering the limitless permutations of protein and gene activation states available to cells and the all-or-nothing character of cell-fate decisions, such organization may, in fact, reflect the true nature of stem cell commitment. We thank Dr. Ian Chambers for providing EF4 cells. This work is funded by NSERC Discovery Program, as well as the Canadian Stem Cell Network. R.E.D. is a recipient of a Natural Sciences and Engineering Research Council of Canada postgraduate fellowship. P.W.Z. is the Canada Research Chair in Stem Cell Bioengineering.

The FASEB Journal

DAVEY ET AL.

REFERENCES 1. 2. 3.

4.

5.

6.

7.

8. 9.

10. 11.

12. 13.

14.

15.

16.

17. 18. 19. 20.

Goldbeter, A., and Koshland, D. E., Jr. (1982) Sensitivity amplification in biochemical systems. Q. Rev. Biophys. 15, 555–591 Ferrell, J. E., Jr., and Machleder, E. M. (1998) The biochemical basis of an all-or-none cell fate switch in Xenopus oocytes [see comments]. Science 280, 895– 898 Smith, A. G., Heath, J. K., Donaldson, D. D., Wong, G. G., Moreau, J., Stahl, M., and Rogers, D. (1988) Inhibition of pluripotential embryonic stem cell differentiation by purified polypeptides. Nature 336, 688 – 690 Williams, R. L., Hilton, D. J., Pease, S., Willson, T. A., Stewart, C. L., Gearing, D. P., Wagner, E. F., Metcalf, D., Nicola, N. A., and Gough, N. M. (1988) Myeloid leukaemia inhibitory factor maintains the developmental potential of embryonic stem cells. Nature 336, 684 – 687 Ying, Q. L., Nichols, J., Chambers, I., and Smith, A. (2003) BMP induction of Id proteins suppresses differentiation and sustains embryonic stem cell self-renewal in collaboration with STAT3. Cell 115, 281–292 Sato, N., Meijer, L., Skaltsounis, L., Greengard, P., and Brivanlou, A. H. (2004) Maintenance of pluripotency in human and mouse embryonic stem cells through activation of Wnt signaling by a pharmacological GSK-3-specific inhibitor. Nat. Med. 10, 55– 63 Dani, C., Chambers, I., Johnstone, S., Robertson, M., Ebrahimi, B., Saito, M., Taga, T., Li, M., Burdon, T., Nichols, J., and Smith, A. (1998) Paracrine induction of stem cell renewal by LIFdeficient cells: a new ES cell regulatory pathway. Dev. Biol. 203, 149 –162 Gadue, P., Huber, T. L., Nostro, M. C., Kattman, S., and Keller, G. M. (2005) Germ layer induction from embryonic stem cells. Exp. Hematol. 33, 955–964 Saito, M., Yoshida, K., Hibi, M., Taga, T., and Kishimoto, T. (1992) Molecular cloning of a murine IL-6 receptor-associated signal transducer, gp130, and its regulated expression in vivo. J. Immunol. 148, 4066 – 4071 Kamimura, D., Ishihara, K., and Hirano, T. (2003) IL-6 signal transduction and its physiological roles: the signal orchestration model. Rev Physiol Biochem. Pharmacol. 149, 1–38 Nichols, J., Davidson, D., Taga, T., Yoshida, K., Chambers, I., and Smith, A. (1996) Complementary tissue-specific expression of LIF and LIF-receptor mRNAs in early mouse embryogenesis. Mech. Dev. 57, 123–131 Bader, A., Al-Dubai, H., and Weitzer, G. (2000) Leukemia inhibitory factor modulates cardiogenesis in embryoid bodies in opposite fashions. Circ. Res. 86, 787–794 Rathjen, P. D., Nichols, J., Toth, S., Edwards, D. R., Heath, J. K., and Smith, A. G. (1990) Developmentally programmed induction of differentiation inhibiting activity and the control of stem cell populations. Genes Dev. 4, 2308 –2318 Guo, Y., Graham-Evans, B., and Broxmeyer, H. E. (2006) Murine embryonic stem cells secrete cytokines/growth modulators that enhance cell survival/anti-apoptosis and stimulate colony formation of murine hematopoietic progenitor cells. Stem Cells 24, 850 – 856 Davey, R. E., and Zandstra, P. W. (2006) Spatial organization of embryonic stem cell responsiveness to autocrine Gp130 ligands reveals an autoregulatory stem cell niche. Stem Cells 24, 2538 – 2548 Matsuda, T., Nakamura, T., Nakao, K., Arai, T., Katsuki, M., Heike, T., and Yokota, T. (1999) STAT3 activation is sufficient to maintain an undifferentiated state of mouse embryonic stem cells. EMBO J 18, 4261– 4269 Niwa, H., Burdon, T., Chambers, I., and Smith, A. (1998) Self-renewal of pluripotent embryonic stem cells is mediated via activation of STAT3. Genes Dev. 12, 2048 –2060 Raz, R., Lee, C. K., Cannizzaro, L. A., d’Eustachio, P., and Levy, D. E. (1999) Essential role of STAT3 for embryonic stem cell pluripotency. Proc. Natl. Acad. Sci. U. S. A. 96, 2846 –2851 Scholer, H. R., Balling, R., Hatzopoulos, A. K., Suzuki, N., and Gruss, P. (1989) Octamer binding proteins confer transcriptional activity in early mouse embryogenesis. EMBO J. 8, 2551–2557 Niwa, H., Miyazaki, J., and Smith, A. G. (2000) Quantitative expression of Oct-3/4 defines differentiation, dedifferentiation or self-renewal of ES cells. Nat. Genet. 24, 372–376

AUTOREGULATORY CONTROL OF ESC SELF-RENEWAL

21. 22.

23.

24.

25. 26.

27.

28.

29.

30. 31. 32.

33.

34.

35.

36.

37.

38.

39.

Stryer, L. (1995) Biochemistry, Freeman, New York Chambers, I., Colby, D., Robertson, M., Nichols, J., Lee, S., Tweedie, S., and Smith, A. (2003) Functional expression cloning of Nanog, a pluripotency sustaining factor in embryonic stem cells. Cell 113, 643– 655 Mitsui, K., Tokuzawa, Y., Itoh, H., Segawa, K., Murakami, M., Takahashi, K., Maruyama, M., Maeda, M., and Yamanaka, S. (2003) The homeoprotein Nanog is required for maintenance of pluripotency in mouse epiblast and ES cells. Cell 113, 631– 642 Ichiba, M., Nakajima, K., Yamanaka, Y., Kiuchi, N., and Hirano, T. (1998) Autoregulation of the Stat3 gene through cooperation with a cAMP-responsive element-binding protein. J. Biol. Chem. 273, 6132– 6138 O’Brien, C. A., and Manolagas, S. C. (1997) Isolation and characterization of the human gp130 promoter. Regulation by STATS. J. Biol. Chem. 272, 15003–15010 Zhan, M., Miura, T., Xu, X., and Rao, M. S. (2005) Conservation and variation of gene regulation in embryonic stem cells assessed by comparative genomics. Cell Biochem. Biophys. 43, 379 – 406 He, F., Ge, W., Martinowich, K., Becker-Catania, S., Coskun, V., Zhu, W., Wu, H., Castro, D., Guillemot, F., Fan, G., de Vellis, J., and Sun, Y. E. (2005) A positive autoregulatory loop of JakSTAT signaling controls the onset of astrogliogenesis. Nat. Neurosci. 8, 616 – 625 Auernhammer, C. J., Bousquet, C., and Melmed, S. (1999) Autoregulation of pituitary corticotroph SOCS-3 expression: characterization of the murine SOCS-3 promoter. Proc. Natl. Acad. Sci. U. S. A. 96, 6964 – 6969 Viswanathan, S., Davey, R. E., Cheng, D., Raghu, R. C., Lauffenburger, D. A., and Zandstra, P. W. (2005) Clonal evolution of stem and differentiated cells can be predicted by integrating cell-intrinsic and -extrinsic parameters. Biotechnol. Appl. Biochem. 42, 119 –131 Bagowski, C. P., and Ferrell, J. E., Jr. (2001) Bistability in the JNK cascade. Curr. Biol. 11, 1176 –1182 Chickarmane, V., Troein, C., Nuber, U. A., Sauro, H. M., and Peterson, C. (2006) Transcriptional dynamics of the embryonic stem cell switch (Online). PLoS Comput. Biol. 2, e123 Hao, J., Li, T. G., Qi, X., Zhao, D. F., and Zhao, G. Q. (2006) WNT/beta-catenin pathway up-regulates Stat3 and converges on LIF to prevent differentiation of mouse embryonic stem cells. Dev. Biol. 290, 81–91 Ogawa, K., Nishinakamura, R., Iwamatsu, Y., Shimosato, D., and Niwa, H. (2006) Synergistic action of Wnt and LIF in maintaining pluripotency of mouse ES cells. Biochem. Biophys. Res. Commun. 343, 159 –166 Karpf, A. R., Peterson, P. W., Rawlins, J. T., Dalley, B. K., Yang, Q., Albertsen, H., and Jones, D. A. (1999) Inhibition of DNA methyltransferase stimulates the expression of signal transducer and activator of transcription 1, 2, and 3 genes in colon tumor cells. Proc. Natl. Acad. Sci. U. S. A. 96, 14007–14012 Blanchard, F., Kinzie, E., Wang, Y., Duplomb, L., Godard, A., Held, W. A., Asch, B. B., and Baumann, H. (2002) FR901228, an inhibitor of histone deacetylases, increases the cellular responsiveness to IL-6 type cytokines by enhancing the expression of receptor proteins. Oncogene 21, 6264 – 6277 Cartwright, P., McLean, C., Sheppard, A., Rivett, D., Jones, K., and Dalton, S. (2005) LIF/STAT3 controls ES cell self-renewal and pluripotency by a Myc-dependent mechanism. Development 132, 885– 896 Boeuf, H., Merienne, K., Jacquot, S., Duval, D., Zeniou, M., Hauss, C., Reinhardt, B., Huss-Garcia, Y., Dierich, A., Frank, D. A., Hanauer, A., and Kedinger, C. (2001) The ribosomal S6 kinases, cAMP-responsive element-binding, and STAT3 proteins are regulated by different leukemia inhibitory factor signaling pathways in mouse embryonic stem cells. J. Biol. Chem. 276, 46204 – 46211 Burdon, T., Stracey, C., Chambers, I., Nichols, J., and Smith, A. (1999) Suppression of SHP-2 and ERK signalling promotes self-renewal of mouse embryonic stem cells. Dev. Biol. 210, 30 – 43 Ko, S. Y., Kang, H. Y., Lee, H. S., Han, S. Y., and Hong, S. H. (2006) Identification of Jmjd1a as a STAT3 downstream gene in mES cells. Cell Struct. Funct. 31, 53– 62

2031

40. 41.

42. 43.

44.

2032

Mikkers, H., and Frisen, J. (2005) Deconstructing stemness. EMBO J. 24, 2715–2719 Nagy, A., Rossant, J., Nagy, R., Abramow-Newerly, W., and Roder, J. C. (1993) Derivation of completely cell culture-derived mice from early-passage embryonic stem cells. Proc. Natl. Acad. Sci. U. S. A. 90, 8424 – 8428 Chang, K. H., and Zandstra, P. W. (2004) Quantitative screening of embryonic stem cell differentiation: endoderm formation as a model. Biotechnol. Bioeng. 88, 287–298 Prudhomme, W., Daley, G. Q., Zandstra, P., and Lauffenburger, D. A. (2004) Multivariate proteomic analysis of murine embryonic stem cell self-renewal versus differentiation signaling. Proc. Natl. Acad. Sci. U. S. A. 101, 2900 –2905 Ebong, S., Yu, C. R., Carper, D. A., Chepelinsky, A. B., and Egwuagu, C. E. (2004) Activation of STAT signaling pathways

Vol. 21

July 2007

and induction of suppressors of cytokine signaling (SOCS) proteins in mammalian lens by growth factors. Invest. Ophthalmol. Vis. Sci. 45, 872– 878 45. Fouladi Nashta, A. A., Andreu, C. V., Nijjar, N., Heath, J. K., and Kimber, S. J. (2004) Role of leukemia inhibitor factor (LIF) in decidualisation of murine uterine stromal cells in vitro. J. Endocrinol. 181, 477– 492 46. Viswanathan, S., Benatar, T., Rose-John, S., Lauffenburger, D. A., and Zandstra, P. W. (2002) Ligand/receptor signaling threshold (LIST) model accounts for gp130-mediated embryonic stem cell self-renewal responses to LIF and HIL-6. Stem Cells 20, 119 –138

The FASEB Journal

Received for publication December 11, 2006. Accepted for publication December 25, 2006.

DAVEY ET AL.