Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, Ontario, Canada
Department of Chemical, Engineering and Applied Chemistry, University of Toronto, Toronto, Ontario, Canada
Institute of Biomaterials and Biomedical Engineering, 164 College Street, Room 407, Rosebrugh Building, University of Toronto, Toronto, Ontario, Canada M5S 3G9. Telephone: 416-978-8888; Fax: 416-978-4317
Highly ordered aggregates of cells, or niches, regulate stem cell fate. Specific tissue location need not be an obligatory requirement for a stem cell niche, particularly during embryogenesis, where cells exist in a dynamic environment. We investigated autoregulatory fixed-location-independent processes controlling cell fate by analyzing the spatial organization of embryonic stem cells (ESCs) using quantitative single-cell immunocytochemistry and a computational approach involving Delaunay triangulation. ESC colonies demonstrated radial organization of phosphorylated signal transducer and activator of transcription 3, Nanog, and Oct4 (among others) in the presence and absence of exogenous leukemia inhibitory factor (LIF). Endogenous self-renewal signaling resulted from autocrine non-LIF gp130 ligands, which buffered cells against differentiation upon exogenous LIF deprivation. Together with a radial organization of differential responsiveness to gp130 ligands within colonies, autocrine signaling produced a radial organization of self-renewal, generating a fixed-location-independent autoregulatory niche. These findings reveal fundamental properties of niches and elucidate mechanisms colonies of cells use to transition between fates during morphogenesis.
Specialized microenvironments, or niches, direct the types and levels of exogenous stimuli received by cells including both diffusible (autocrine/paracrine) and immobilized (cell surface/extracellular matrix [ECM]) signals. Somatic stem cells exhibit niche dependence for control of cell fate in vivo (reviewed in ). Untangling the regulatory aspects of the niche may provide strategies for the in vitro manipulation of stem cell fate. Most somatic stem cell niches described to date reside in specific tissue locations with defined architecture (reviewed in ) and have been considered stable components of tissues, as demonstrated by the ability of empty niches to be repopulated . Although niches may include multiple cell types and extrinsic signals organized within a highly structured architecture, there are two basic requirements for a stem cell niche: production of local extrinsic signal(s) and spatial organization of receptive cells relative to those signals to facilitate position-dependent choices between contrasting cell fates. Specific tissue locations need not be an obligatory requirement for the stem cell niche. A definition that encompasses “fixed-location-independent autoregulation”  is more broadly applicable to stem cells such as the epidermal stem cell, which has been postulated to reside in clusters in the basal layer of the interfollicular epidermis [5, 6] but which lacks “a clearly defined niche” by classic standards . Fixed-location-independent autoregulatory niches (FLIANs) may also play a prominent role in guiding development during embryogenesis, where cells persist in a spatially dynamic microenvironment.
Embryonic stem cells (ESCs) derived from the early embryo represent a tractable model system for investigating autoregulatory properties of stem cells. Mouse ESCs are derived from the inner cell mass (ICM) of the 3.5-day preimplantation blastocyst [8, 9]. Early differentiation and axial patterning within the postimplantation embryo appear to be induced and spatially organized by an asymmetrical arrangement of derivatives of the extraembryonic (trophectoderm and primitive endoderm) lineages (reviewed in ). In contrast, the signals and mechanisms regulating self-renewal of pluripotent cells in the preimplantation and early postimplantation embryo, as well as the mechanisms by which cells transition between responding to self-renewal- and differentiation-inducing signals, are less clear. Interleukin-6 (IL-6)-type cytokines including leukemia inhibitory factor (LIF) have been shown to maintain ESCs in vitro and to enable their derivation [11, 12], an effect attributable to signal transducer and activator of transcription 3 (STAT3) activation [13, –15]. LIF-induced dimerization of the gp130 and leukemia inhibitory factor receptors (LIFRs) phosphorylates receptor-associated Janus kinases (Jaks), causing recruitment and phosphorylation of STAT3. Subsequent nuclear translocation of tyrosine (705)-phosphorylated signal transducer and activator of transcription 3 (pSTAT3) results in altered gene expression and maintenance of self-renewal, a state typified by the expression of the key stem cell transcription factors Oct4  and Nanog [17, 18]. The Jak-STAT pathway is the only signaling pathway demonstrated to maintain pluripotency of the ICM, although a critical role for the pathway appears only to arise during a delay in blastocyst implantation known as diapause . The ability of Jak-STAT signaling to stabilize the in vivo stem cell compartment during diapause is reflected in vitro, where it acts to stabilize an otherwise transient ESC population.
The capacity of ESCs to recapitulate various features of in vivo development in vitro indicates strong autoregulatory control of cell fate. For example, when cultured as aggregates in the absence of LIF, ESCs form embryoid bodies (EBs) that appear to mimic aspects of development from blastocyst to egg-cylinder stages (embryonic day [E] 3.5–E6.5) . Furthermore, manipulation of ESCs in vitro has uncovered novel autoregulatory control mechanisms. The selective ablation of differentiated cells in EB culture has revealed endogenous self-renewal capacity, enabling the maintenance of entirely undifferentiated EBs in the absence of exogenous LIF . In addition, in adherent culture, paracrine signaling from newly differentiated cells has been shown to promote ESC self-renewal; this finding is correlated with increased expression of LIF by differentiated cells . These and other observations indicate that ESCs represent a valuable model system for understanding autoregulatory mechanisms controlling the ICM and suggest that ESCs may form FLIANs that support self-renewal in vitro.
To directly investigate autoregulatory properties of ESCs, it was necessary to develop an approach for measuring protein levels in single cells in conjunction with spatial analysis—and to do this quantitatively across heterogeneous populations of cells. By combining automated fluorescent microscopy and single-cell image analysis with a computational approach for interrogating spatial organization, we define an in vitro fixed-location-independent autoregulatory niche; a niche characterized by the radial organization of cell fate determinants and self-renewal within colonies. Furthermore, we present a novel mechanism for controlling self-renewal within the niche: autocrine (non-LIF) gp130 ligand expression combined with spatial organization of differential responsiveness to that ligand.
Materials and Methods
D1C2 (E14TG2a-derived LIF−/−) cells  (generously provided by Dr. A. Smith) and R1 cells  were routinely cultured as previously described .
Adherent cultures were washed twice with ice-cold phosphate-buffered saline (PBS) and harvested in lysis buffer (20 mM Tris, pH 8, 10% glycerol, 1% Nonidet P-40, 0.1% sodium dodecyl sulfate, 10 mM sodium phosphate, and 100 mM sodium vanadate) . Equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membrane. Membranes were incubated with anti-phospho-Stat3 (tyr705) (9131; Cell Signaling Technology, Beverly, MA, http://www.cellsignal.com) (1:1,000), and anti-glyceraldehyde-3-phosphate dehydrogenase (ab9485; Abcam, Cambridge, U.K., http://www.abcam.com) (1:1,000) primary antibodies and goat anti-rabbit IgG-HRP (A0545; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com) (1:2,000) secondary antibody in Tris-buffered saline solution containing 0.1% Tween 20 and 1% bovine serum albumin. Bands were visualized by enhanced chemiluminescence (ECL Plus; RPN2132; GE Healthcare, Chalfont St. Giles, U.K., http://www.gehealthcare.com).
Immunocytochemistry and Single-Cell Fluorescence Quantification
For all immunocytochemical studies, unless otherwise noted, 15% knockout serum replacement (10828-028; Invitrogen, Carlsbad, CA, http://www.invitrogen.com) was used. Cells were seeded at 5,000 cells per well (72-hour studies) or 10,000 cells per well (24-hour studies) in 96-well plates (6005182; PerkinElmer, Wellesley, MA, http://www.las.perkinelmer.com) coated overnight with a solution of fibronectin (12.5 μg/ml; F1141; Sigma-Aldrich) and gelatin (0.02%) (optimal for imaging). In the presence of fibronectin, we observed a flattened but compact colony morphology, similar growth rate, and similar self-renewal/differentiation capacity . Cells were treated with LIF, Jak inhibitor 1 (420099; Calbiochem, San Diego, http://www.emdbiosciences.com), anti-gp130 neutralizing antibody (AF468; R&D Systems Inc., Minneapolis, http://www.rndsystems.com), or controls, as indicated.
Immunocytochemistry was performed as previously described . The following primary antibodies were used: anti-β-catenin (9563), anti-phospho-Stat3 (tyr705) (9131), anti-phospho-p44/42 mitogen-activated protein kinase (thr202/tyr204) (9101), anti-phospho-Smad1 (ser463/465)/Smad5 (ser463/465)/Smad8 (ser426/428) (9511; Cell Signaling Technology), anti-actin (A 5060), anti-GATA4 (G8794; Sigma-Aldrich), anti-Oct3 (Oct4) (611202; BD Transduction Laboratories, Franklin Lakes, NJ, http://www.bdbiosciences.com), and anti-Nanog (generously provided by Dr. T. Yamanaka) antibodies diluted 1:200 (1:1,600 for anti-Nanog). The following secondary antibodies were used: AlexaFluor 488 (A-11034) and AlexaFluor 546 (A-11030; Molecular Probes Inc., Eugene, OR, http://probes.invitrogen.com) plus Hoechst (862096; Sigma-Aldrich) (0.1 μg/ml). Cells were scanned using the ArrayScan automated fluorescent microscope (Cellomics, Pittsburgh, http://www.cellomics.com). Average pixel intensity (total pixel intensity divided by number of pixels) within the nuclear area (as defined by Hoechst staining) of individual cells was determined for fluorophores of interest, along with x-y coordinates. Compensation between fluorophore channels was performed manually by determining the linear relationship between channels using appropriate controls (similar to flow cytometry). Oct4 subpopulations were determined by fitting two Gaussian distributions (OriginPro 7.5; OriginLab, Northampton, MA, http://www.originlab.com) to log-transformed data. Conditions were performed in triplicate wells within each experiment, and each experiment was performed at least three independent times unless otherwise stated.
A spatial analysis program was developed to analyze immunocytochemistry data to identify individual cells and colonies and to correlate fluorescence intensity with spatial organization such as position within-colonies. The spatial analysis program computes the Delaunay triangulation (a set of triangles where all points represent the vertices of triangles, and no point lies within a triangle) for a set of cell coordinates from a well (triangulation implementation by S. Fortune ) to produce the set of nearest “neighbors” for all cells. Next, cell-cell contact between the sets of neighbors is determined on the basis of distance (D), and sets with cells in common are combined to produce colonies. Radial position of cells within colonies is then determined relative to colony centroid. Cell clusters between 10 and 140 cells in number and with an x-width to y-width ratio no greater than 2:1 were selected for subsequent analysis to exclude colony aggregates. For radial analysis of protein levels (fluorescence intensity), D = 16.32 μm (12 pixels). Cells were binned into center (radius = D) and ring (width = D) compartments. For radial analysis of the Oct4− subpopulation, D = 18 pixels (24.4 μm) to reflect the larger footprint of Oct4− cells. To analyze protein levels in the Oct4+ subpopulation, all Oct4− cells were excluded from colony identification and radial analysis. Threshold cutoff was set to include the top 90% of the Oct4+ subpopulation, thereby excluding >99% of the Oct4− subpopulation, allowing us to compare equivalent subpopulations between treatments.
For radial protein analysis, >40,000 Oct4+ cells were identified, and 100–300 Oct4+ colonies were analyzed per experiment. Each ring average was normalized to the center to remove any colony size dependence. Ring averages across all colonies were calculated and normalized to actin control. For Oct4− subpopulation distribution analysis, >30,000 cells were analyzed per experiment. To determine subpopulation segregation by density, the distance between individual cells and their nearest neighbors was determined alongside Oct4 status (to enhance the stringency by which we identified differentiated cells, Oct4 cutoff set at 95% of Oct4− subpopulation). For radial analysis of Oct4− cells, the percentage of Oct4− cells in each ring was determined. One-hundred fifty to 300 colonies were analyzed per experiment. Ring averages were averaged across all colonies. Conditions were performed in triplicate within each experiment, and each experiment was performed at least three independent times.
Statistical analysis was performed using either GenStat (VSN International, Hemel Hempstead, U.K., http://www.vsni.co.uk) or SigmaStat (SysStat Software, Inc., Point Richmond, CA, http://www.systat.com) software. A paired Student's t test was performed to determine the prevalence of Oct4 radial colonies over random chance. One-way analysis of variance (ANOVA) with blocks (independent plates) was used to test for differences between rings. Two-way ANOVA with blocks (independent plates) was used to test for differences between rings and treatments with or without LIF. All pairwise comparisons were performed using Holm-Sidak multiple comparisons test following ANOVA.
Reverse Transcription-Polymerase Chain Reaction
Total RNA was extracted using GenElute Mammalian Total RNA Miniprep Kit (RTN-70; Sigma-Aldrich). One-Step reverse transcription-polymerase chain reaction (RT-PCR) kit (210210; Qiagen) was used according to manufacturers recommendations. Primers used were as follows: oncostatin m (OSM), 5′-gaagaaaacgggaggaggag-3′, 5′-gggctccaagagtgattctg-3′ (406 base pair [bp]); ciliary neurotrophic factor (CNTF), 5′-caaaatccacagccaggaat-3′, 5′-agtcgctctgcctcagtcat-3′ (400 bp); IL-11, 5′-tgctgacaaggcttcgagta-3′, 5′-ccccctccaagtaccaaaat-3′ (403 bp); cardiotrophin-1 (CT-1), 5′-cactgcaggcatcttctcag-3′, 5′-tcaacaacaagcaagcaagc-3′ (403 bp); cardiotrophin-like cytokine (CLC), 5′-agcttagctgggacctacctg-3′, 5′-agcagccagaagtcatccat-3′ (398 bp); LIF ; IL-27 p28 and, Epstein-Barr virus-induced gene 3 (EBI-3) (together making up functional IL-27) ; and IL-6 . All primers except CT-1 were designed to border introns; genomic contamination for CT-1 was eliminated by DNaseI treatment and verified by substituting Taq polymerase (D-4545; Sigma-Aldrich) for enzyme mix. The identities of all PCR products were verified by restriction digestion.
Spatial Organization of Cell Fate Determinants Within Adherent Cultures
In vitro, ESCs cluster together into tight refractile colonies to the exclusion of morphologically differentiated cells. If autoregulatory processes influence cell fate decisions, we reasoned, levels of self-renewal-associated proteins, such as pSTAT3 and Oct4, would be influenced by cell position. To test this, automated fluorescent microscopy of adherent cultures was combined with quantitative single-cell image analysis. This approach was found to be both sensitive and accurate for measuring levels of pSTAT3 (Fig. 1A) and Oct4 in response to LIF stimulation.
As an initial evaluation of spatial organization, we took raw single-cell fluorescence datasets and reconstructed the topography of protein expression and activation in individual wells of 96-well plates. Prominent spatial organization of pSTAT3 and Oct4 could be observed in cultures plated as single cells and maintained in either the presence or absence of LIF. This organization seemed to arise in concert with the formation of cell clusters. Representative reconstructions for 72-hour LIF-deprived cultures are shown (Fig. 1B, 1C). Interestingly, STAT3 phosphorylation was observed in the absence of exogenous LIF and in an apparently nonrandom fashion similar to Oct4, suggesting the presence of a spatially organized endogenous STAT3 activating signal. To investigate the observed spatial organization further, we developed an approach using Delaunay triangulation to identify single cells and cell clusters and to examine the relationship between protein levels and cell position within colonies relative to colony centroid. In particular, analyzing the spatial organization of signaling within colonies enabled us to discriminate between cell-autonomous signals (no organization in colonies), colony-autonomous signals (outward radial organization), and peripheral cell-dependent signals (inward radial organization).
Oct4, pSTAT3, and Nanog Expression Levels Are Radially Organized
Oct4 expression appears to decrease radially within Oct4+ (undifferentiated) colonies both by visual examination and upon spatial analysis of individual colonies (Oct4− cells were excluded from both colony identification and subsequent analysis) (Fig. 1D–1F). To determine the prevalence of the observed radial organization, we quantified the number of Oct4+ colonies demonstrating an uninterrupted radial organization in Oct4 expression (where each consecutive sampling ring out from the center has less average intensity per cell then the one previous). To do this, we costained for Oct4 and actin, identified colonies of Oct4+ cells, and then removed all colonies that demonstrated uninterrupted radial organization in actin. This approach eliminated possible confounding factors such as three-dimensional cell clumping (which by visual examination of images was marginal or absent under our optimized conditions) and other morphological concerns, such as differences in the columnar nature of cells at the center versus perimeter of colonies. Twenty percent (20.3% ± 0.7% SE) of the remaining colonies demonstrated an uninterrupted radial organization of Oct4 (greater than chance; p < .001), with no significant difference between LIF supplemented and nonsupplemented conditions. This organization was not limited to colonies of a defined size. Equivalent results were also obtained when analysis was limited to colonies exhibiting comparable cell density in all rings, and to cells of comparable size (data not shown). Notably, this stringent measure of spatial organization is a significant under-representation of overall organization. Aggregates of colonies, colonies falling at the edge of fields, and colonies with noncircular morphology, although not amenable to quantitative radial analysis, ubiquitously exhibited concentric organization.
To identify the underlying signaling events that may explain this radial organization, we analyzed Oct4+ colonies for a set of signaling factors known to be involved in either differentiation or self-renewal. These included pSTAT3, phospho-SMADs 1/5/8, and β-catenin (all of which are nuclear-translocating transcription factors known or believed to play some role in ESC self-renewal [33, 34]) and phospho-extracellular signal-related kinase (ERK) 1/2 (which is activated alongside STAT3 in response to LIF but promotes differentiation ). The self-renewal transcription factor (TF) Nanog and Hoechst staining as a measure of DNA content were also included in this analysis. Cells were costained for each protein with Oct4, and analysis was performed on radially organized Oct4+ colonies. Resulting radial profiles were normalized to actin. Significant radial organization was observed for Oct4 (p < .001), Nanog (p = .013), pSTAT3 (p < .001), Hoechst (p = .032), and pERK 1/2 (p < .001) in the presence of LIF, whereas in the absence of LIF, only Oct4, Nanog, and pSTAT3 demonstrated significant radial organization (all p < .001) (Fig. 2A–2E). Importantly, in all cases, the radial trends observed were found to continue in subsequent rings (when present). However, because few colonies were large enough to exhibit these rings, these data were not included. Interestingly, the extent of radial organization (relative to colony center) for pSTAT3 decreased in the absence of LIF, whereas it increased for Oct4 and Nanog (all p < .001). In contrast, β-catenin showed no radial organization in nuclear signaling levels in either the presence or the absence of LIF (although nuclear translocation could be induced by the glycogen synthase kinase 3β inhibitor BIO; unpublished observation) (Fig. 2F). Likewise, pSMADs 1/5/8 showed no organization (data not shown). Radial organization of Oct4, Nanog, and pSTAT3 levels within Oct4+ colonies in the presence and absence of exogenous LIF indicates the presence of a self-renewal promoting signal endogenous to the Oct4+ population and suggests STAT3 signaling may be the effecter of this signal.
Endogenous STAT3 Activation Is due to a Non-LIF Autocrine Factor
The outwardly decreasing radial organization of pSTAT3 observed in Oct4+ colonies in the absence of exogenous LIF suggests the presence of an endogenous STAT3 activator originating within colonies, rather than paracrine signaling from peripheral differentiated cells. To determine whether this STAT3 activation occurs in the absence of a significant differentiated subpopulation, ESCs plated and maintained for 24 hours in the absence of LIF (before notable differentiation) were examined. After 24 hours of LIF deprivation Oct4− cells composed 9.2% (±1.8 SE) of the population, and significant STAT3 activation was apparent in the Oct4+ subpopulation (Fig. 3A, 3B). Given the short half-life of pSTAT3 (<10 minutes; unpublished observation) this activity represents an endogenous signal. To establish whether this STAT3 activity was Jak-mediated, an inhibitor of the Jak family was used (Jak Inhibitor 1 [JakI1] ). Exposure to Jak inhibitor reduced STAT3 activity (Fig. 3A, 3B) in a dose-dependent fashion (data not shown). Jak inhibitor also successfully removed STAT3 activation and organization in colonies exposed at 72 hours after plating (data not shown).
Next, to determine whether the endogenous Jak-mediated signal was due to either autocrine or paracrine LIF secretion or was due to a component of the medium, LIF−/− D1C2 (E14Tg2a-derived) cells were cultured for 24-hour in control medium or in medium lacking serum replacement. Endogenous STAT3 activity was observed in both cases (p < .001 for all pairwise comparisons of +LIF, −LIF, and JakI1 treatments within each group) (Fig. 3A). As further confirmation, an analysis of LIF−/− D1C2 cultures revealed pSTAT3 spatial organization identical to R1 cells (data not shown). In addition, STAT3 activity could be removed by simple medium exchange 0.5 or 1 hour prior to fixing (Fig. 3B), indicating that it arises from a secreted factor. These results demonstrate that endogenous STAT3 activity is due to a secreted, Jak-dependent, and non-LIF autocrine factor.
Autocrine gp130 Ligand Buffers Stem Cells Against Differentiation
To determine whether this autocrine activity promotes self-renewal, ESCs were cultured for 48 hours in the presence or absence of 600 nM JakI1. ESCs maintained in the absence of LIF undergo increased differentiation in the presence of JakI1 (Fig. 4A) without noticeable effects on cell proliferation or survival (Fig. 4A, inset). Since the unidentified autocrine factor activates STAT3 in a Jak-dependent manner, we sought to determine whether this factor was an IL-6 family member. RT-PCR analysis was performed to assess expression of known IL-6-type cytokines. R1 cells expressed detectable levels of mRNA for all family members with the exception of IL-27 (one of the two subunits making up functional IL-27 was not expressed, although no positive control was available) (Fig. 4B, 4D), whereas LIF−/− D1C2 cells showed the same expression pattern with the exception of LIF (Fig. 4C, 3D). To determine whether secreted gp130 ligands could account for the autocrine self-renewal activity, cells were grown in the presence of a gp130 neutralizing antibody. Neutralization of gp130 resulted in increased differentiation of both R1 and LIF−/− D1C2 cells, equivalent to maintenance in the Jak inhibitor (Fig. 4E). These results suggest a complex regulatory network of secreted factors wherein the localized secretion of gp130-activating ligands may be differentially regulated to affect ESC fate.
Spatial Organization of Responsiveness to gp130 Ligands Organizes pSTAT3 in Colonies
With the identification of a functional autocrine gp130 ligand, we next wanted to determine how radial organization of STAT3 signaling is established in colonies. The simplest explanation would be that this pattern arises as a consequence of paracrine signaling of the endogenous gp130 ligand(s). Under the appropriate conditions (secretion rate, receptor number, etc.), cells in the center of colonies may receive more signal (combined autocrine and paracrine activation) than those at the perimeter, resulting in an outward radial organization of STAT3 activation. However, in light of the fact that radial organization of pSTAT3 occurs in the presence of high concentrations of exogenously added gp130 ligands such as LIF, other mechanisms may be responsible for the observed organization. We therefore tested the possibility that the radial organization of pSTAT3 observed in colonies arises from a correlation between cell position and responsiveness to gp130 ligands. To test this, we examined whether STAT3 responsiveness to a short pulse of exogenous LIF correlated with radial position. Cells maintained in LIF for 60 hours were deprived of LIF for 12 hours before being restimulated for 15 minutes (a duration of stimulation that produces maximal transient STAT3 activation; unpublished observation). Although the absolute magnitude of STAT3 activation increased 1.95-fold (±0.21 SE) for the stimulated over the unstimulated cells, no statistically significant change in radial organization relative to colony center was observed (Fig. 5A). This indicates that cell position correlates with responsiveness to gp130 ligands. Next, we determined whether cell responsiveness correlated with Oct4 level in the absence of colonies and the possibility of radial paracrine signaling (or other colony effects). To do this, cells plated as single cells and grown for 12 hours (before colony formation) were examined for a correlation between Oct4 and pSTAT3. A positive correlation was observed between Oct4 and pSTAT3 for cells maintained in the absence of LIF and presence of JakI1 (Fig. 5B), in the absence of LIF (Fig. 5C), in the presence of LIF (Fig. 5D), and upon 15-minute LIF stimulation of LIF-deprived cultures (Fig. 5E). Furthermore, with increasing levels of endogenous and exogenous ligand stimulation, the relationship between Oct4 and pSTAT3 became steeper. An analysis of LIF−/− D1C2 cells produced similar results (data not shown). The proportionately greater response to stimulation in high Oct4 versus low Oct4 expressers indicates that responsiveness of cells to both autocrine and exogenous gp130 ligands is correlated with “cell state” (Oct4 level). Responsiveness and cell state in turn correlate with position within the colony. Collectively, the presence of self-renewal promoting autocrine gp130 ligand(s), the radial organization of responsiveness to these ligands, and the corresponding radial organization of critical self-renewal proteins within the Oct4+ subpopulation are strong evidence of an autoregulatory niche.
Spatial Organization of Protein Levels Correlates with Cell Fate
A necessary tenet for any niche is position-dependent control of cell fate. We therefore analyzed the position of Oct4– cells to test for the cell fate organization predicted by our spatial analysis of protein levels within the Oct4+ subpopulation. Specifically, our analysis predicts that differentiated cells should be radially distributed within colonies, being found most frequently at colony perimeters. Subpopulation segregation, with differentiated cells apparent at the perimeter or otherwise excluded from Oct4+ colonies, is commonly observed upon visual examination of ESC cultures. To quantitatively verify this segregation, we examined the relationship between distance from individual cells to their nearest neighbor and Oct4 status (Oct4+ vs. Oct4−). As expected, the farther cells are from their nearest neighbor the more likely they are to be Oct4− (p < .001), in either the presence (Fig. 6A) or absence (Fig. 6B) of LIF. By examining the original cell images from which the datasets were obtained, we established that beyond approximately 24.4 μm (18 pixels) from their nearest neighbor, cells are most likely not in contact. Thus, Oct4+ cells are generally found in close association with other cells, whereas Oct4− cells are found much more frequently at a lower density and in isolation.
Next, as a direct test of the cell fate predictions made by our spatial analysis of protein levels within the Oct4+ subpopulation, we examined the location of differentiated cells within colonies. As predicted from our analysis, the proportion of cells that were Oct4− increased with distance from colony centroid (p < .001) (Fig. 6C). The discovery that Oct4− cells are not only found more frequently at the perimeter of colonies but are actually radially organized within colonies corroborates our protein analysis and confirms the presence of an autoregulatory niche.
In contrast to the typical highly structured niche of somatic stem cells, pluripotent cells in the developing embryo persist in a dynamic spatial environment. In the absence of a fixed environment, autoregulatory processes may be expected to play a primary role in directing cell fate. To investigate autoregulatory processes, we developed an approach to measure the spatial organization of cells and intracellular cell fate determinants within ESC cultures. ESCs plated as single cells were found to self-organize to form FLIANs that temporarily maintain self-renewal in the absence of exogenous LIF. The niche is composed of a compact body of cells characterized by radial organization of the transcription factors Oct4 and Nanog and the signal transducer/transcription factor pSTAT3 (Fig. 2A–2C). Through transgenic approaches these proteins have previously been demonstrated to support self-renewal in a dose-dependent manner [13, 17, 37, 38]. Herein, we report for the first time spontaneous heterogeneity in protein levels within the undifferentiated (Oct4+) subpopulation that correlate with cell fate decisions in both self-renewal and differentiation-promoting conditions. Specifically, Oct4, Nanog, and pSTAT3 levels in undifferentiated (Oct4+) cells decreased with increasing distance from colony centroid, whereas the proportion of differentiated (Oct4−) cells increased (Fig. 6C). This approach thereby provides unique insight into signaling hierarchies leading to cell fate decisions in unmanipulated cultures.
The presence and radial organization of pSTAT3 in the absence of exogenous LIF suggests that self-renewal in the niche is controlled by an endogenous STAT3-activating signal. Indeed, by using inhibitors of Jak and gp130 signaling, we demonstrated the presence of an endogenous gp130 ligand(s) that temporarily buffers cells against exogenous LIF deprivation. Specifically, the endogenous gp130 ligand(s) was shown to promote STAT3 activation (Fig. 3) and self-renewal (Fig. 4), in the absence of effects on proliferation or survival (Fig. 4A, inset). Gp130 ligands are known to play a role in maintaining the ICM in vivo during a delay in implantation (diapause). During normal nondiapause development, embryos lacking gp130 show lethality between E12 and E18 [39, 40], whereas LIFR mutants die around the time of birth [41, 42]. However, if implantation is delayed, these mutations result in a failure of the blastocyst to implant and loss of the ICM . The restricted expression of LIF transcripts to the trophectoderm and gp130/LIF receptor transcripts to the ICM implies a paracrine relationship for maintaining the ICM during diapause . Paracrine LIF signaling by differentiated cells has also been suggested to play a role in adherent cultures in vitro . In fact, using LIF−/− ESCs, it was previously demonstrated that differentiated cultures produce functional levels of both LIF and an unidentified ESC renewal factor, which acts independently of STAT3 . Interestingly, it was reported that LIFR−/− embryos survive longer during diapause than gp130−/− embryos, indicating the presence of an additional LIFR-independent gp130 ligand in vivo . The identical behavior of wild-type and LIF−/− cell lines in our assays suggests that the endogenous gp130 ligand identified is not LIF. Furthermore, endogenous pSTAT3 is present in newly plated cultures containing few differentiated cells and is organized within colonies in a radially diminishing gradient, suggesting that this factor arises within the Oct4+ subpopulation. Paracrine signaling from differentiated cells, given their position at the perimeter of colonies, would be predicted to produce an inverse radial organization (doughnut-shaped) as has been observed for primate embryonic stem cells grown in supportive conditions . Therefore, we present evidence for the first time, to our knowledge, of an autocrine self-renewal promoting signal in ESCs.
There are eight known members of the IL-6 family of cytokines; these include CT-1, CLC, CNTF, IL-6, IL-11, IL-27, LIF, and OSM. All members form receptor complexes that include the common gp130 receptor subunit, as well as ligand-specific subunits . By RT-PCR, we demonstrated the expression of all known IL-6 family members (with the exception of IL-27) in ESC cultures maintained in LIF (Fig. 4B–4D). The recent discovery of IL-11 and OSM in ESC-conditioned media corroborates this . It has been demonstrated that both CNTF and CT-1 can substitute for LIF in maintaining ESC self-renewal [46, 47], whereas mouse OSM (except at very high concentrations) and IL-6 cannot [46, 48]. These differences reflect differences in receptor subunit expression [48, 49] rather than cytokine-specific differences, as IL-6 combined with soluble IL-6Rα is effective at maintaining ESCs . It is therefore probable that with appropriate receptor and cofactor expression, all IL-6 family members could substitute for LIF. Additional studies will be needed to elucidate the specific factor or factors responsible for the autocrine activity observed.
We provide strong evidence that cell fate within the niche is controlled by the radial organization of STAT3 signaling. It is worth noting that radial organization was also observed for pERK 1/2 (Fig. 2E), albeit only in the presence of exogenous LIF. Similar to STAT3, ERK 1/2 phosphorylation occurs downstream of gp130 activation. However, ERK 1/2 activation has been shown to oppose STAT3-mediated self-renewal . Interestingly, the lack of pERK 1/2 radial organization in colonies cultured in the absence of exogenous LIF suggests either that some gp130-activating ligands are ineffective at activating ERK 1/2 compared with LIF or that different thresholds of gp130 signaling are needed to activate STAT3 and ERK 1/2. Importantly, such differential activation of STAT3 and ERK 1/2 by autocrine gp130 ligands could effectively reduce the threshold in pSTAT3 necessary for self-renewal by eliminating conflicting ERK 1/2 signaling.
Since pSTAT3 organization was observed in the presence of exogenous LIF (alongside pERK 1/2, which is also activated by gp130 signaling), we tested the possibility that this spatial organization arises from a spatial heterogeneity in responsiveness to gp130 ligands. Such heterogeneity could also explain the spontaneous differentiation observed in the presence of exogenous LIF . As anticipated, responsiveness to autocrine and exogenous gp130 ligands was found to correlate with Oct4 expression, as well as location in the niche (Fig. 5A, 5B).
The mechanism underlying differential responsiveness in the niche is unclear, but it may involve changes in pathway component expression. Differences in LIFR expression have been reported to exist in the ICM, and as a consequence, different ESC lines derived from the same ICM were shown to demonstrate different capabilities for differentiation . The hierarchy of signaling events leading to differential responsiveness is also unclear. Level of responsiveness does not appear to be a direct consequence of Oct4 expression level, as cells exhibiting decreased Oct4 expression (via short hairpin RNA interference) retain normal STAT3 signaling (E. Walker, R. Davey, M. Ohishi, P. Zandstra, and W. Stanford, unpublished results). The substantial variation we observed between Oct4 and pSTAT3 levels in single cells further indicates an indirect rather than direct link between Oct4 expression and gp130 ligand responsiveness. In fact, it is more likely that responsiveness of cells to gp130 ligands determines STAT3 activation, which in turn effects Oct4 expression [13, –15]. Nor does this spatial organization of responsiveness arise because of signaling through any of the major pathways believed to be involved in self-renewal and differentiation (β-catenin, pSMADs 1/5/8, and pERK 1/2), since spatial organization of these proteins was not observed or, in the case of pERK 1/2 (which supports differentiation), would predict an inverse organization. Interestingly, STAT3-binding sites have been found in the promoters of multiple gp130 pathway components [53, , , –57]. It is therefore possible that paracrine gp130 signaling establishes this organization of responsiveness via a positive feedback loop. Interestingly, since this organization occurs in the presence of supplemented LIF, this would require a greater efficacy of signaling for the secreted factor than for LIF. Such a possibility could arise from differences in ligand-receptor affinities or numbers of ligand-specific receptor subunits. How differential responsiveness arises and is spatially organized is an active area of investigation.
The existence of a self-organizing niche in ESC colonies demonstrates that autoregulatory processes can maintain stem cell populations independent of a fixed tissue location. The autoregulatory system described here combines an autocrine self-renewal signal with spatial organization of responsiveness to that signal. This system promotes self-renewal by concentrating both signal and responsiveness to the interior of colonies, thereby facilitating differentiation at the perimeter even in the presence of high concentrations of gp130 ligands (both exogenous and endogenous) (Fig. 7). Such an autoregulatory system suggests that modulation of ligand responsiveness may be an initiating step in differentiation, and one that may provide a mechanism of irreversibility to the cell fate decision. Since ESCs reintroduced into the blastocyst contribute normally to development (and thereby retain signaling fidelity in culture) these autoregulatory processes may be expected to play a role in early embryogenesis.
Interestingly, direct extrapolation of this niche in vivo would imply a role for STAT3 signaling during normal (nondiapause) development. STAT3−/− embryos progress to E6.0 (dying between E6.5 and E7.0), a result that has been interpreted as demonstrating a noncritical role for STAT3 in the establishment and maintenance of the ICM . However, these STAT3 knockouts showed reduced size at E6.0, and E3.5 blastocyst outgrowths were also substantially smaller than wild-type, indicating some important role for STAT3 during normal ICM development . A noncritical but supportive role for STAT3 in maintaining the ICM is paralleled in vitro, where we report a nonsufficient but supportive role for endogenous STAT3 signaling in maintaining ESC cultures in the absence of exogenous LIF. These results imply that self-renewal normally arises as a result of contributions from multiple signaling events. Indeed, our spatial analysis suggests that organizing systems (niches) rather than “master” signals control stem cell fate, an idea well-developed for the somatic stem cell niche.
The authors indicate no potential conflicts of interest.
We thank Dr. Boris Steipe, Dan Horner, and Raheem Peerani for helpful discussions regarding spatial analysis; Dr. William Stanford for comments; and Drs. Austin Smith and Shinya Yamanaka for providing reagents. This work is funded by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Program, as well as the Canadian Stem Cell Network. R.E.D. is a recipient of a NSERC postgraduate fellowship. P.W.Z. is the Canada Research Chair in Stem Cell Bioengineering.