William Nordhaus explains the economic consequences of Greenland ice sheet melt.
New Research In
Physical Sciences
Featured Portals
Articles by Topic
Biological Sciences
Featured Portals
Articles by Topic
Edited by Ali H. Brivanlou, The Rockefeller University, New York, NY, and accepted by Editorial Board Member Brigid L. Hogan January 23, 2019 (received for review September 5, 2018)
Significance
While the dynamics of cellular signaling pathways have received increased attention recently, whether the dynamics of signaling vary depending on the context remains unclear. The WNT/β-catenin signaling pathway plays essential roles in embryonic development, control of cell proliferation, and maintenance of stem cell niches. Here we used quantitative microscopy to study the dynamics of the WNT pathway in pluripotent stem cells. Signaling dynamics were transient in the pluripotent state but sustained during differentiation, and this switch relied on synergy with other key developmental pathways. In addition, we show that β-catenin dynamics vary across several commonly used cell lines. This work highlights the changing dynamics of WNT signaling during early differentiation and provides a template for studying how signaling depends on cellular context.
Abstract
WNT/β-catenin signaling is crucial to all stages of life. It controls early morphogenetic events in embryos, maintains stem cell niches in adults, and is dysregulated in many types of cancer. Despite its ubiquity, little is known about the dynamics of signal transduction or whether it varies across contexts. Here we probe the dynamics of signaling by monitoring nuclear accumulation of β-catenin, the primary transducer of canonical WNT signals, using quantitative live cell imaging. We show that β-catenin signaling responds adaptively to constant WNT signaling in pluripotent stem cells, and that these dynamics become sustained on differentiation. Varying dynamics were also observed in the response to WNT in commonly used mammalian cell lines. Signal attenuation in pluripotent cells is observed even at saturating doses, where ligand stability does not affect the dynamics. TGFβ superfamily ligands Activin and BMP, which coordinate with WNT signaling to pattern the gastrula, increase the β-catenin response in a manner independent of their ability to induce new WNT ligand production. Our results reveal how variables external to the pathway, including differentiation status and cross-talk with other pathways, dramatically alter WNT/β-catenin dynamics.
During metazoan development, WNT/β-catenin signaling is critical for proper morphogenesis and patterning of tissues and cell types (1, 2). In adults, WNT plays a key role in maintaining homeostasis and regulating stem cell niches (3, 4). Mutations in WNT/β-catenin pathway components are frequently found in human cancers (5, 6).
Canonical WNT signals are transduced to the nucleus by β-catenin (Fig. 1A) (6⇓–8). WNTs are secreted lipid-modified proteins that bind a cell surface receptor complex consisting of Frizzled and LRP5/6 (7⇓–9). WNT receptor binding leads to inhibition of the β-catenin destruction complex (consisting of multiple scaffolds and kinases, including APC, AXIN, GSK3β, and CK1) (10), allowing β-catenin accumulation and translocation to the nucleus. In the nucleus, β-catenin complexes with TCF/LEF to regulate target genes in a context-specific manner (11). Along with being the primary effector of canonical WNT signaling, β-catenin also has a role in stabilizing adherens junctions at the cell membrane.
Fig. 1.
A CRISPR-Cas–mediated GFP knockin labels endogenous β-catenin and does not disturb signaling or differentiation. (A) Simplified canonical WNT/β-catenin pathway. Pharmacologic small molecules used in this study are shown in red. (B) Schematic of resulting mRNA transcribed from the labeled β-catenin allele after CRISPR-Cas–mediated PuroR-T2A-GFP knockin. The puromycin resistance protein (PuroR) facilitates selection of labeled cells. T2A is a self-cleaving peptide that enables separation of PuroR from GFP-β-catenin. (C) Confocal microscopy images of live hESCs with GFP-labeled β-catenin and the nuclear marker H2B-RFP. (Left) Before exogenous WNT3A addition. (Right) Nuclear GFP-β-catenin accumulation after 4 h of WNT3A treatment. (D) Representative image of an 800- μm-diameter micropatterned hESC colony stained with DAPI (magenta) and Brachyury (cyan) using immunofluorescence after 42 h of treatment with exogenous BMP4. (E) Quantification of BRA expression in the hESC and labeled reporter cell lines (n = 20 colonies for unlabeled cells; n = 11 for reporter cells).
During gastrulation, dramatic morphogenetic rearrangements occur simultaneously with patterning of the primitive streak (PS) by BMP, WNT, and NODAL signals (12). Given these cellular movements and the rapid changes in expression patterns of all these ligands, it is clear that cells will experience rapidly changing levels of all these morphogens. The coupling of patterning, growth, and morphogenesis, along with the lack of methods for temporally precise perturbation of signaling, makes systematically dissecting the contribution of signaling dynamics difficult in vivo. In contrast, in vitro, researchers can administer precise amounts of signaling ligands while inhibiting endogenous ligands. Similarly, the combinatorial effects of multiple ligands can be investigated directly. Finally, ligands can be provided dynamically, which enables testing the effects of various ligand dynamics, such as adding the same doses of ligand at different rates of change (13, 14). In addition, in vitro cell culture is highly amenable to live cell imaging techniques.
While numerous regulators of the WNT/β-catenin pathway have been identified (7, 15, 16), less is known about WNT/β-catenin signaling dynamics. Because of the variety of contexts in which β-catenin plays crucial roles, and the diversity of potential regulators, it is impossible to understand β-catenin dynamics in any particular setting without making explicit measurements. Here we created a fusion of GFP and β-catenin at the endogenous locus and used quantitative microscopy to measure signaling dynamics. We found that the response to WNT varies significantly by differentiation stage and cell type. β-catenin response to WNT was adaptive in human embryonic stem cells (hESCs) but sustained in many other cell types. Adaptation in hESCs is controlled at or upstream of GSK3β and confers sensitivity to the WNT rate of change at lower doses. However, when hESCs were subjected to a PS differentiation protocol (17), β-catenin was stably activated. Surprisingly, both TGFβ and BMP synergized with exogenously provided WNT via a mechanism independent of WNT ligand induction, and BMP could induce nuclear β-catenin independent of WNT ligands altogether. Our results reveal insight into how WNT/β-catenin signaling dynamics vary by context, and how WNT signaling synergizes with other key morphogens during early development.
Results
A CRISPR-Cas Mediated GFP Knockin Labels Endogenous β-Catenin Without Perturbing Signal Transduction or Differentiation.
To measure WNT/β-catenin signaling dynamics in single cells, we used CRISPR-Cas9 gene editing (18⇓⇓–21) to insert GFP at the N terminus of endogenous β-catenin in hESCs (Fig. 1 B and C). To facilitate computational nuclear identification and analysis, these cells also express RFP-H2B incorporated into the genome with the ePiggyBac transposable element system (22). GFP-β-catenin localization was confirmed with immunofluorescent antibody staining of β-catenin (SI Appendix, Fig. S1A), and correct expression of the GFP-β-catenin fusion without any other expression of GFP was confirmed by Western blot analysis (SI Appendix, Fig. S1B). GFP-β-catenin accumulated in the nucleus in response to exogenous WNT3A (Fig. 1C), and the transcriptional dynamics of β-catenin target genes in response to WNT3A was very similar in unmodified and modified hESCs (SI Appendix, Fig. S1C). As a stringent test of the potential of these cells, spatial patterning and differentiation were unaffected in a micropatterned gastruloid protocol (23, 24) (Fig. 1 D and E).
Human Pluripotent Cells Respond Adaptively to Exogenous WNTs, and Adaptation Is Controlled Upstream of GSK3.
In hESCs treated with exogenous WNT3A, initially GFP-β-catenin rapidly accumulates in both the nucleus and cell membrane (Fig. 2 A and B). However, while the increase in membrane GFP-β-catenin is sustained, nuclear GFP-β-catenin begins to decline at approximately 4 h after addition of WNT3A (i.e., hESCs adapt to constant WNT signals). Peak signaling is dose-dependent and occurs earlier at lower doses. Adaptation is complete at sufficiently low doses of WNT3A, and at saturating WNT3A (doses >300 ng/mL), cells adapt to approximately 40% of peak signal. A recently developed commercially available induced pluripotent stem cell (iPSC) line with a GFP-β-catenin fusion (Allen Institute) had identical dynamics (SI Appendix, Fig. S2). Signaling dynamics were verified in unmodified hESCs using immunofluorescent antibody staining for β-catenin (Fig. 2C).
Fig. 2.
WNT/β-catenin signaling is partially adaptive in stem cells, and adaptation is controlled upstream of GSK3β. (A) Representative images from time-lapse imaging of GFP-β-catenin–labeled hESCs treated with 100 ng/mL WNT3A at 0, 3, and 15 h. (B) Quantification of nuclear and membrane (Inset) levels of GFP-β-catenin in hESCs treated with various concentrations (ng/mL) of WNT3A. (C) Quantification of β-catenin dynamics by either GFP or immunofluorescence in the reporter cell line compared with immunofluorescence in the parental line. (D) Quantification of nuclear GFP-β-catenin in hESCs at indicated seeding densities with or without WNT3A, represented as a ratio to 50,000 cells/cm2. This is the seeding density for all other experiments. (D′) Same data as in D but represented as the ratio to mean signaling before WNT addition at that density. (E) GFP-β-catenin–labeled hESCs treated with various concentrations (µM) of CHIR99021, a pharmacologic GSK3β inhibitor. Error bars in all graphs indicate SEM of ≥644 cells. In this and all other figures, nuclei are computationally identified using a nuclear label (RFP-H2B if live cells, DAPI if not). The mean intensity of the indicated marker (GFP-β-catenin or anti-β-catenin) is normalized against the intensity of the nuclear label. In all panels except D′, “mean nuclear ßcat” is always defined as the ratio over the negative control for that experiment. Membrane (a.u.) is the mean pixel intensity of computationally identified membranes across at least eight images per condition after background subtraction. Note that the values shown for nuclei and membranes are not directly comparable, as only the former are normalized. (Scale bar: 20 µm.)
To assay possible effects of endogenous WNT ligands induced by exogenously supplied WNT3A, cells were treated with a small-molecule inhibitor of WNT secretion, IWP2 (25), along with either a high dose or an intermediate dose of exogenous WNT3A (SI Appendix, Fig. S3). Inhibition of WNT secretion lowered the final adapted level only at intermediate doses, indicating that additional WNT ligands are induced, but the high dose of WNT3A alone is capable of completely saturating the response and is not additive with endogenous WNT signaling. In addition, while increasing cell density increased the total GFP-β-catenin in cells (Fig. 2D), it did not affect the fold change in nuclear β-catenin in response to stimulation (Fig. 2D′), and dynamics were not affected by ROCK inhibition with the small-molecule Y-27632 (SI Appendix, Fig. S4). Pathway activation using the small-molecule GSK3β inhibitor CHIR99021, commonly used as a WNT substitute in differentiation protocols (17, 26⇓–28), resulted in nonadaptive dose-dependent increases in nuclear β-catenin (Fig. 2E). Sustained signaling in response to CHIR99021 demonstrates that the mechanisms that control adaptation must act upstream of GSK3β. This result also highlights that while small-molecule inhibitors of GSK3β are potent activators of β-catenin signaling, they induce very different dynamics than WNT ligands.
WNT Inactivation or Instability Does Not Account for Adaptation in hESCs.
Posttranslational palmitoylation increases the hydrophobicity of WNT proteins and is thought to make them prone to aggregation and increase the difficulty of purification (29, 30). In addition, recombinant WNT likely is not as potent as its endogenous counterpart, given the very high exogenous doses used in vitro (often >200 ng/mL) compared with other morphogens (Activin and BMP4 both saturate Smad signaling at 3 ng/mL) (14). This has led other researchers to question the stability of recombinant WNTs in culture media (31⇓–33). However, if the observed adaptation to WNT signals were due to WNT3A instability in culture media, then the time to adapt to WNT should be dose-dependent and would become very long at doses well above saturation, which is not the case. In addition, at saturating WNT3A, conditioned media taken from WNT3A-treated hESCs retains its ability to activate signaling in previously unstimulated cells over the entire course of adaptation (SI Appendix, Fig. S5), proving that adaptation at this dose is not due to WNT instability. Finally, a commercially available “WNT stabilizer” (34) had no effect on the adaptation of hESCs (SI Appendix, Fig. S6B). This WNT stabilizer caused cells to quickly aggregate (SI Appendix, Fig. S6A) and increased their nuclear area (SI Appendix, Fig. S6C), likely indicative of biological effects beyond simply stabilizing recombinant WNT.
We next aimed to investigate whether subsaturating doses of exogenous WNT lose potency over time. To do so, we compared the ability of media containing saturating and subsaturating doses of WNT3A to activate β-catenin after exposure to cells. We treated cells with varying doses of WNT3A for 8 h, by which time they had begun to adapt, and then swapped the media from these cells with that of previously untreated cells while measuring the resulting signaling in each group (Fig. 3A and SI Appendix, Fig. S5). The response was greatly decreased in the second group of cells only when subsaturating WNT was used. Thus, the media undergoes a net decrease in WNT/β-catenin activating potential, but only when a subsaturating dose of WNT is used. In a complementary experiment, we periodically replaced WNT3A-containing media with fresh media containing the same dose of WNT3A (thus replacing inactivated WNT and potentially removing secreted regulators). At subsaturated doses, replenishing the WNT led to a more sustained response, while cells treated with high WNT continued to adapt despite the constant replenishment of exogenous WNT (Fig. 3B). This finding demonstrates that a net decrease in WNT-activating potential occurs, but only affects signaling dynamics at subsaturating levels of WNT.
Fig. 3.
Adaptation to saturating WNT is not due to ligand decay. (A) Quantification of β-catenin dynamics with live cell imaging. At 0 h, cells were either left untreated (blue and black curves) or treated with 100 ng/mL WNT3A (Left) or 1,000 ng/mL WNT3A (Right) (red and green curves). After 8 h, medium was either swapped between a treated well (green) and an untreated well (black) or not (red and blue). The experiments shown on the left and right were performed simultaneously, so that a single no treatment control (blue) was used, as reproduced in each graph. (B) Cells were treated with indicated concentrations of WNT3A, then either left unperturbed (red) or had their medium replaced with fresh medium containing an identical concentration of WNT3A (black) at the indicated time points (dashed gray line; every 2 h for the first 12 h). Error bars in all graphs indicate SEM of ≥494 cells.
WNT activity loss could be due to biological effects, including secretion of WNT inhibitors (eg, DKKs, SFRPs, WIFs, and Notum) into the culture media or degradation of WNT ligands by the receiving cells. Alternatively, WNT activity could be reduced by cell-independent ligand decay or binding to the culture vessel. To test these hypotheses, we performed conditioned media experiments using different numbers of cells and various culture vessels for conditioning. To prevent endogenous WNTs from playing a role, IWP2 was included in all conditions. If loss is due to cell-dependent ligand decay or secretion of extracellular inhibitors, we would expect conditioned media prepared with higher cell densities to have the greatest loss of potency. In addition, the difference in loss of potency between conditioned media prepared with cells and media prepared under identical conditions but without cells would indicate the degree of cell-dependent potency loss.
We found that the loss of potency in conditioned media obtained from our standard seeding density (50,000 cells/cm2) was similar to the loss of potency that occurred when no cells were present (SI Appendix, Fig. S7A), while, paradoxically, high cell densities (400,000 cells/cm2) were actually associated with less potency loss (SI Appendix, Fig. S7B). We hypothesized that higher cell densities could effectively decrease the surface area of the culture vessel and prevent ligand binding. Consistent with this idea, conditioned media prepared in a low-retention Eppendorf tube had the lowest loss of potency. Thus, a major factor in the loss of potency in media transfer experiments is binding to the culture vessel; however, the relevance of this to the adaptation seen at low doses without media transfer is difficult to ascertain. It may be that ligands bound to the culture vessel are permanently inaccessible to the cells; alternatively, the bound population may remain in equilibrium with the population in the media, and thus the effect of binding during adaptation in a single culture dish may be less than that seen in transfer experiments. In any event, adaptation of WNT signaling at subsaturating doses likely is due primarily to some combination of the recombinant WNT’s half-life in culture media, its binding to the culture vessel, and the same media-independent mechanisms seen with higher doses.
Since WNT better retains its potency when the media is conditioned with cells, it is unlikely that the secretion of freely diffusing WNT inhibitors or cell-dependent ligand decay plays a significant role. The increased potency from higher cell density is likely due to lowering the culture vessel’s surface area, and thus the amount of WNT lost to plastic binding. The secretion of WNT-stabilizing proteins or inhibitors not transferred with media may also play a role (35).
hESCs Are Sensitive to WNT Dynamics.
Previous work with the TGFβ and Shh pathways has shown that ligand presentation dynamics can alter the signaling response and change cell fate decisions (13, 14, 36⇓–38). Thus, we tested whether hESCs are sensitive to WNT dynamics. A hallmark of signaling adaptation is sensitivity to the rate of ligand presentation (i.e., the time derivative of the ligand concentration); therefore, we hypothesized that administering WNT sufficiently slowly would attenuate the response. Indeed, the signaling response was lower in hESCs when the WNT3A was administered gradually (SI Appendix, Fig. S8A). However, given that cells are sensitive to WNT concentration over more than two orders of magnitude (Fig. 2), increasing to saturating doses while avoiding substantial response to a single step is impractical. Our experiments involved 10 steps of ligand increase, and at higher final doses of WNT, the rate of ligand increase was sufficient to saturate the response (SI Appendix, Fig. S8B). In the future, it might be possible to attenuate the response to higher doses by slowly administering WNT over more extended periods with automated fluidic systems, thus more precisely probing the parameter space in which hESCs are sensitive to ligand derivatives.
WNT Target Genes Have Various Response Profiles.
We next asked how WNT target gene activation dynamics might correlate with those of β-catenin. While some genes were transiently activated in response to WNT3A (e.g., DKK4, DKK1, AXIN2), others were sustained (e.g., EOMES, LEF1) (Fig. 4A). It is possible that transcription of the nonadaptive genes is maintained by mechanisms independent of β-catenin. For example, the WNT targets NODAL and BRACHYURY are also regulated by Nodal signaling, and thus we used the small molecule SB431542, a specific inhibitor for the Nodal receptor, to decouple the WNT response from the downstream induced Nodal response. Interestingly, NODAL induction by WNT is adaptive when its self-activation is inhibited (Fig. 4B). In addition to regulation independent of the WNT pathway, some nonadaptive WNT targets may be sufficiently sensitive to nuclear β-catenin such that the β-catenin level following adaptation is sufficient to saturate their response. To test whether transiently induced target genes result from the dynamics of β-catenin, we compared the dynamics of transcription in WNT3A-treated versus CHIR99021-treated cells. Induction of transcriptional targets with adaptive dynamics became sustained when CHIR99021 was used (SI Appendix, Fig. S9), indicating that sustaining nuclear β-catenin is sufficient to maintain the transcription of these targets.
Fig. 4.
WNT target genes have varied response profiles. qRT-PCR in hESCs treated with 300 ng/mL WNT3A (A) or 300 ng/mL WNT3A (B) with or without the small-molecule TGFβ pathway inhibitor SB431542 (10 µM) for the indicated duration.
β-Catenin Signaling Dynamics Are Context-Specific and Sustained in a PS Differentiation Protocol.
We next asked how the response to WNTs might vary across cell types. Immunofluorescent imaging using an antibody for β-catenin revealed a variety of dynamics in response to WNT3A across different mammalian cell lines (Fig. 5A), including both adaptive and sustained profiles, demonstrating that adaptation is not a universal feature of WNT signaling and that WNT/β-catenin dynamics can change with context and cell fate. Since the in vivo equivalent of hESCs is epiblast cells, which differentiate to PS fates in response to WNT, we examined β-catenin dynamics in hESCs differentiated to the PS-like cell fate. For PS differentiation, we adapted a previously published protocol (17). hESCs were stimulated with Activin, BMP4, and 1 μM CHIR 99021 for 24 h. The identity of these cells was confirmed by immunofluorescent staining showing down-regulation of E-cadherin and SOX2 and up-regulation of the PS-marker Brachyury (SI Appendix, Fig. S10).
Fig. 5.
Signaling is sustained during differentiation to PS fates, and the dynamic response to exogenous WNTs varies in other cell types. (A) Quantification of anti-β-catenin by immunofluorescent imaging in the cell lines indicated in the legend. (B and C) Quantification of time-lapse movies of GFP-β-catenin–containing hESCs. (B) Treated with or without PS differentiation media. (C) Response to exogenous WNT3A in pluripotent hESCs (Left) vs cells differentiated for 24 h to PS-like fate (Right). Error bars in all graphs indicate SEM of ≥267 cells.
Interestingly, the differentiation protocol induced a sustained β-catenin signaling profile (Fig. 5B). This finding was surprising, as the dose of CHIR99021 is insufficient to induce detectable nuclear β-catenin when presented in isolation (Fig. 2E). The PS cells were refractory to exogenous WNT, as adding WNT3A after 24 h of PS differentiation produced no response over the mock-treated control (Fig. 5C).
Activin and BMP Activate WNT-Signaling with a Delay.
Given the sustained induction of nuclear β-catenin in the PS-differentiation protocol, we sought to identify how Activin and BMP might individually contribute toward β-catenin signaling. We found that both Activin alone and BMP alone can induce sustained β-catenin signaling (Fig. 6 A and B). β-catenin activation with Activin and BMP was delayed compared with β-catenin activation with WNT3A (Fig. 2B) and SMAD activation with Activin or BMP (14), implying that β-catenin signaling is a downstream effect of the Activin and BMP signals. Since both BMP and Activin can induce the production of new WNT ligands (12), we used IWP2 or recombinant DKK1 (an inhibitor of the canonical WNT coreceptor LRP6) to inhibit the activity of WNTs induced downstream of Activin or BMP. We found that IWP2 completely blocked the increase in nuclear β-catenin in response to Activin, indicating that Activin requires the presence of endogenous WNTs to activate β-catenin signaling. However, even a combination of IWP2 and DKK1 only partially blocked the increase in nuclear β-catenin in response to BMP, and by the end of this experiment, nuclear β-catenin levels had risen to approximately 30% of the levels achievable with exogenously added WNT3A in the same experiment. This suggests, surprisingly, that while part of the β-catenin response to BMP is attributable to new synthesis and secretion of WNT ligands, another part of the response is WNT-ligand independent.
Fig. 6.
TGFβ ligands Activin and BMP4 increase β-catenin signaling independent of WNT-ligand induction. IWP2 (3 µM) is a small-molecule inhibitor of endogenous WNT secretion, and DKK1 (200 ng/mL) is a protein that blocks canonical WNT signaling at the LRP5/6 receptor. (A–C and E) Quantification of time-lapse imaging of GFP-β-catenin–labeled hESCs treated as indicated. All treatments were administered simultaneously. (D and F) Nuclear β-catenin over baseline at 20 h after the indicated treatments showing TGFβ ligand synergy with WNT independent of WNT secretion. Data are reproduced from A, C, and E. In all graphs, Activin A and BMP4 doses were 30 ng/mL and 10 ng/mL, respectively, and error bars represent SEM of ≥511 cells.
Activin and BMP Synergize with WNT Without the Requirement for WNT Ligand Induction.
We next asked how the presence of Activin or BMP might affect the response to exogenous WNT signals. When intermediate WNT3A (30 ng/mL) is added to cells at the same time as Activin (Fig. 6C) or BMP4 (Fig. 6E) along with IWP2, the initial response is similar to the WNT3A response alone. After 7 h, signaling continues to adapt in WNT3A-only–treated cells but begins to increase in WNT3A+Activin- and WNT3A+BMP-treated cells. The effect of Activin or BMP4 is only partly dependent on the induction of endogenous WNTs, since signaling increases despite the presence of IWP2. The total signaling after 20 h of Activin or BMP cotreatment with WNT was more than the sum of their individual effects (Fig. 6 D and F).
Activin and BMP no longer affect β-catenin when WNT3A saturates the response (300 ng/mL WNT3A; Fig. 6 C and E). In addition, there was little difference when either Activin or BMP was added 10 h before exogenous WNT3A rather than simultaneously (SI Appendix, Fig. S11). This demonstrates that neither Activin nor BMP requires induction of endogenous WNTs to increase β-catenin signaling. However, while Activin still requires synergy with another source of WNTs to increase β-catenin signaling, BMP increases β-catenin signaling even when all other sources of WNTs have been removed.
Discussion
Here we show that β-catenin signaling responds adaptively to constant WNT signaling in pluripotent stem cells, and that these dynamics change dramatically with cell context. At saturating doses of WNT, adaptation is partial, cell-autonomous, and controlled at or upstream of GSK3β. This is in contrast to TGFβ signaling, which we previously showed to be adaptive in pluripotent cells (14), where adaptation is near complete even at saturating doses. Finally, our results show that TGFβ signaling and BMP signaling increase β-catenin signaling independent of their ability to induce WNT ligands.
Our GFP-β-catenin knockin labeling strategy is quantitative, allows measurement with high temporal resolution, conserves spatial information, and, in contrast to β-catenin fluorescent protein overexpression reporters (39), maintains pathway stoichiometry. In the future, we will be able to combine this approach with self-patterning differentiation assays (e.g., organoids, gastruloids, embryoids) (40, 41), providing us with the signaling trajectories required for each cell fate. This approach is not limited to development and likely can be used just as effectively for dissecting signaling in models of diseased tissue.
Previously, Goentoro et al. (42) reported that total β-catenin levels were stably increased in a rectal carcinoma cell line (RKO) that contains a mutation in E-cadherin that prevents β-catenin localization to the membrane. This mutation facilitates biochemical analysis; however, whether these dynamics are representative of other cell types is unclear. More recently, Kafri et al. (39) measured β-catenin dynamics in HEK293 cells by overexpressing a YFP-β-catenin reporter. While the β-catenin dynamics for RKO and HEK293 cells that we measured here are consistent with those studies, examination of a larger number of cell lines and treatments showed that these dynamics are highly context-dependent. In addition, RKO cells were an outlier, in that in all cell lines tested, they exhibited the greatest fold change in nuclear β-catenin from pre- to post-WNT treatment (Fig. 5), which when combined with potential indirect effects of known mutations in RAF and PI3K pathways (43, 44) calls into question the use of RKO cells as a general Wnt model. Our results highlight the importance of studying β-catenin dynamics specifically in the context of interest, and caution against extrapolating between cell lines, organisms, or different differentiation states.
It is becoming increasingly appreciated that signaling pathways respond to stimulation with complex dynamics of signal transduction that influence how ligands are interpreted. For example, in hESCs, the SMAD4 response to BMP4 ligand is stable (14, 22, 45), while its response to Nodal is adaptive (14, 45). Adaptive signaling has also been observed in other pathways, including EGF (46), NF-κB (47, 48), and SHH (37). Pioneering work on ERK signaling in PC12 cells has shown that ERK signals can be either transient or sustained, depending on whether they are stimulated with EGF or with NGF (49⇓⇓–52), but whether pathway dynamics are typically intrinsic to a particular ligand and pathway or vary depending on the context remains unclear. The present study demonstrates that WNT/β-catenin signaling dynamics vary dramatically even in response to the same ligand, as we observed a range of adaptive and sustained responses depending on the stage of differentiation or the cell type. It will be interesting to revisit the dynamics of the pathways described above to determine whether context-dependence is a common feature of signal transduction.
Since WNT/β-catenin signaling controls a variety of processes in different contexts, it will be interesting to see whether certain dynamic properties are associated with distinct roles. For instance, does providing positional information during gastrulation require an adaptive response to WNTs? Are WNT/β-catenin dynamics adaptive in similar contexts, such as anterior-posterior patterning of the neural tube? Adaptive signaling dynamics sensitizes cells to the derivative of ligand concentration, and we (53, 54) and others (55, 56) have suggested that in certain contexts, more positional information can be gained by responding to ligand derivatives compared with concentration alone. In contrast, are there scenarios in which adaptation would be at odds with the role of the WNT pathway? For instance, WNT adaptation appears to be incompatible with the requirements for WNT in maintaining homeostasis in intestinal crypts (4, 57). In these crypts, constant WNT signaling maintains the stemness of highly proliferative multipotent cells. These cells differentiate when they move away from the WNT signal. Would adaptation not lead to the loss of stem cells in the crypt? If cells had adapted to the WNT signal, they would be unable to determine when the signal was lost. It seems that the dynamic requirements for maintaining a constant zone of stemness might differ from the requirements for patterning during gastrulation, and it will be interesting to compare WNT/β-catenin dynamics between these contexts in the future.
This work has revealed that additional β-catenin signaling can be provided from TGFβ and BMP ligands through a variety of mechanisms, including induction of new WNT ligands, sensitization of cells to exposure to WNT ligands, and, for BMP, a mechanism that is completely WNT ligand-independent. The mechanism of this ligand-independent β-catenin signaling by BMP is unclear, and future experiments should aim to identify the functional biochemical mechanisms for pathway cross-talk in pluripotent cells and evaluate how they might be modulated in other contexts.
During gastrulation, BMP signaling from the trophectoderm induces WNT signaling in the PS (12). The WNT signal induces the TGFβ pathway ligand Nodal, which in turn activates BMP signaling. In this way, these signals are thought to reinforce each other and pattern the gastrula. It will be interesting to see if Activin/BMP’s induction of β-catenin signaling outside of ligand induction is important in the context of gastrulation. One hypothesis is that TGFβ- or BMP-mediated stabilization of WNT dynamics might be required to drive certain PS fates, and so differentiation to these fates would occur only in regions of the embryo in which both signals are present. If true, it would be interesting to determine whether this synergy between pathways is independent of the induction of new WNT ligands by Nodal or BMP.
Why baseline nuclear β-catenin increases with cell density is unclear (Fig. 2D). As hESCs are known to produce endogenous WNTs (58, 59), it might be that the higher baseline is simply a consequence of endogenous WNT secretion. Alternatively, adherens junctions may be stabilized and total β-catenin levels at the membrane increased at higher densities, thereby increasing both total and nuclear β-catenin. This is consistent with the effects on WNT-induced mesoderm differentiation observed by modulating the stiffness of the culture surface (60). In addition, it is interesting that the fold change in nuclear β-catenin on addition of WNT was conserved despite the variations in total β-catenin. Future work could determine whether WNT targets are primarily sensitive to fold change or to absolute levels by varying the culture density and examining the dynamics of WNT target genes.
A more fine-grained investigation of the mechanisms that control WNT adaptation and WNT/TGFβ/BMP pathway cross-talk is needed. Our present work shows that adaptation is controlled at or upstream of GSK3β, but the possible role of the known pathway inhibitors remains unclear. Are the adaptation mechanisms always active or are they induced only at high doses of WNT? Which proteins are involved in regulating adaptation, and are their levels modulated to control the degree of activation? More generally, future investigation is needed to understand how the dynamics of the pathway are tuned to achieve divergent functions depending on the context.
Methods
Cell Culture, Treatments, and Differentiation.
hESCs (ESI BIO; ESI017) and iPSCs (Coriell Institute; AICS-0058-067) were maintained in pluripotency maintenance culture as described previously (22). For all experiments with hESCs and iPSCs, cells were seeded into mTeSR1 medium (STEMCELL Technologies) containing rock inhibitor Y27672 (10 μM; STEMCELL Technologies; 05875) at a density of 5 × 104 cells/cm2 except when noted otherwise. The medium was changed the following morning approximately 2 h before administering any treatments. Experiments with other cell lines were similar except for differences in seeding density and culture media. Seeding density was lowered by up to a factor of five for the largest cells, such as RKO and C2C12. DMEM (Corning; 10-017) with 10% FBS (Fisher Scientific; 16000044) culture medium was used for C2C12, RKO, HEK293, and MDCK cell lines. NOF151-hTERT cells were grown in a 1:1 mixture of MCDB 105 (Sigma-Aldrich; M6395) and 199 medium (Sigma-Aldrich; M0393) supplemented with 10% FBS and 1 ng/mL epidermal growth factor (R&D Systems; PRD236)
When exogenous ligands or small molecules were administered to cells, they were prediluted in a volume equal to 20% of the final culture medium volume to facilitate rapid mixing. The following recombinant proteins and small molecules were used: Activin A (R&D Systems; 338-AC; 30 ng/mL), BMP4 (R&D Systems; 314BP050; 10 ng/mL), CHIR99021 (MedChem Express; HY-10182), DKK1 (R&D Systems; 5439-DK-010; 300 ng/mL), IWP2 (Stemgent; 04-0034; 3 µM), SB431542 (STEMCELL Technologies; 72232; 10 µM), WNT3A (R&D Systems; 5036-WN), WNT3A packaged with WNT-stabilizer reagent (AMSBIO; AMS.rhW3aL-002-stab), and Y-27632 (STEMCELL Technologies; 72302; 10 µM)
For WNT ramps (SI Appendix, Fig. S8), WNT3A was administered every 2 h, so that the total administered WNT after each step was as indicated. Ramp duration and time between steps were chosen based on experimental feasibility.
The PS differentiation conditions (Fig. 5) are adapted from Loh et al. (61). Our modifications included using Gibco Essential 6 Medium (Fisher Scientific; A15165-01) as a base and decreasing the CHIR99021 dose to 1 μM. The other supplements were as described previously (61): 30 ng/mL Activin A, 20 ng/mL FGF2 (Thermo Fisher Scientific; PHG6015), and 40 ng/mL BMP4.
Ibidi µ-Slide 8 Well plates (Ibidi; 80826) were used for live imaging experiments. Low-rentention Eppendorf tubes (SI Appendix, Fig. S7) were from Thermo Fisher Scientific (80826).
Plasmids and Generation of GFP-β-Catenin–Labeled hESC Cell Line.
We followed a previously published protocol (18) for CRISPR-cas gene editing. Guide RNA expression was from a PCR-amplified gBlock (IDT) as described previously (18) but containing the β-catenin targeting sequence, cgtggacaatggctactca, located close to the ATG start codon. The homology donor template DNA was prepared by PCR amplification of PuroR-T2A-GFP over two rounds. The primers in the first round contained 5′ overhangs that add part of the β-catenin homology arms. The second round of PCR was done using the PCR product of the first reaction as a template and primers containing the remaining homology arm sequence in their 5′ overhangs. Primer sequences are listed in Table 1. Cas9 expression plasmid, homology donor DNA, and guide RNA were conucleofected in hESCs using the P3 Primary Cell 4D-Nucleofector X Kit (Lonza; V4XP-3012), and positive transformants were selected with puromycin (2 μg/mL; Thermo Fisher Scientific; A1113803). The RFP-H2B construct has been described previously (22). GFP and RFP double-positive cells were obtained by fluorescence-activated cell sorting.
View this table:
Table 1.
Primers used for GFP-β-catenin labeling
Sequence of gBlocks for β-Catenin GuideRNA.
AGTATTACGGCATGTGAGGGCCTATTTCCCATGATTCCTTCATATTTGCATATACGATACAAGGCTGTTAGAGAGATAATTGGAATTAATTTGACTGTAAACACAAAGATATTAGTACAAAATACGTGACGTAGAAAGTAATAATTTCTTGGGTAGTTTGCAGTTTTAAAATTATGTTTTAAAATGGACTATCATATGCTTACCGTAACTTGAAAGTATTTCGATTTCTTGGCTTTATATATCTTGTGGAAAGGACGAAACACCGcgtggacaatggctactcaGTTTAAGAGCTATGCTGGAAACAGCATAGCAAGTTTAAATAAGGCTAGTCCGTTATCAACTTGAAAAAGTGGCACCGAGTCGGTGCTTTTTTGTTTTAGAGCTAGAAATAGCAAGTTAAAATAAGGCTAGTCCGTTTTTAGCGCGTGCGCCAATTCTGCAGACAAATGGCTCTAGAGGTACGGCCGCTTCGAGCAGACATGATAAGATACATTGA
Immunofluorescent Antibody Staining.
Cells were fixed with 4% PFA and stained as described previously (22). Antibodies and dilutions used are listed in Table 2.
View this table:
Table 2.
Antibodies used in this study
Western Blot Analysis.
Western blot analyses were performed using standard protocols using the following antibodies: β-catenin (mouse, 1:20,000 dilution; BD Biosciences; 610154), GFP (rabbit, 1:1,000; Cell Signaling Technology; 2956S), peroxidase-β-actin (1:50,000; Sigma-Aldrich; A3854), peroxidase-rabbit-IgG (1:2,500; Jackson ImmunoResearch; 711–035-152), and peroxidase-mouse-IgG (1:5,000; Jackson ImmunoResearch; 711–035-150).
Quantitative RT-PCR.
Quantitative RT-PCR (qRT-PCR) was performed following the manufacturer’s instructions. In briefly, the RNAqueous-Micro Total RNA Isolation Kit (Thermo Fisher Scientific; AM1931) was used to prepare RNA, and the SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific; 11754-050) was used to prepare cDNA. PCR reactions were done in a StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific; 4367659). ATP5O was used to normalize all genes. Primer sequences are listed in Table 3. Primers created for this study were designed using the qPrimerDepot bank now located at https://pga.mgh.harvard.edu/primerbank/.
View this table:
Table 3.
qRT-PCR primers used in this study
Imaging and Analysis.
Imaging was performed on an Olympus/Andor spinning disk confocal microscope using either a 20×, 0.75 NA air or a 40×, 1.25 NA silicone oil objective. Most of the images displayed in the figures were taken at 40×, while the majority of movies were quantified at 20× (SI Appendix, Supplemental Text). Time-lapse imaging intervals were either 10 or 15 min, and Z-stacks were acquired in three planes spaced 2.5-μm apart. Image analysis was performed using Ilastik (64) (www.ilastik.org) and custom software written in MATLAB (MathWorks) and described previously (23). Analysis code is available from https://github.com/josephkm (63). In brief, maximum intensity projections were taken across Z-slices, and background was subtracted. Background was identified by minimum intensity projection across numerous images and was manually checked for consistency. Nuclear pixels were identified using Ilastik, and resulting masks were imported to MATLAB for segmentation of cells and image quantification. Nuclear intensities of each cell were normalized against that cell’s nuclear marker, and the mean and SEM of these cells were then normalized against the negative control for that experiment (typically a mock-treated control group). For live imaging datasets, a median filter in time over a window of 12 time steps was applied to the control condition before normalization, to avoid propagation of fluctuations from the control to the experimental condition. The control traced in each figure shows the control condition normalized to its own median filter to allow observation of the variability in this condition.
Acknowledgments
We thank Sapna Chhabra for help with micropatterned gastruloid experiments and members of A.W. lab for helpful discussions and comments on the manuscript. This work was funded by grants from the National Institute of General Medical Sciences (R01 GM126122), the National Science Foundation (MCB-1553228), the Cancer Prevention and Research Institute of Texas (RR140073), and the Simons Foundation (511079).
Footnotes
↵1Present address: Sainsbury Laboratory, Cambridge University, Cambridge CB2 1LR, United Kingdom.
↵2Present address: Centers for Disease Control and Prevention, Atlanta, GA 30333.
- ↵3To whom correspondence should be addressed. Email: aryeh.warmflash{at}rice.edu.
Author contributions: J.M. and A.W. designed research; J.M., Y.L., O.A., T.S., and M.S. performed research; M.S. contributed new reagents/analytic tools; J.M. and A.W. analyzed data; and J.M. and A.W. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission. A.H.B. is a guest editor invited by the Editorial Board.
Data deposition: The analysis code for this study has been deposited in GitHub, https://github.com/josephkm.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1815363116/-/DCSupplemental.
Published under the PNAS license.
References
Synergy with TGFβ ligands switches WNT pathway dynamics from transient to sustained during human pluripotent cell differentiation
Sign up for Article Alerts
Article Classifications
- Biological Sciences
- Developmental Biology
Jump to section
You May Also be Interested in
A study suggests that tropical cyclones have been increasing in intensity over the past four decades, consistent with predictions of physical theory and numerical simulations.
Image credit: NOAA/NASA.
Atomic-scale analysis of the Tagish Lake meteorite finds that its raspberry-like magnetite grains formed in a sodium-rich, high-pH environment, providing insight into the chemistry of the earliest fluids in the Solar System.
Image credit: Chi Ma.
Bayesian analysis of the chronology of life’s emergence and development on Earth suggests that if Earth’s history were repeated, the emergence of intelligence might be a sufficiently rare event that it would not be guaranteed to reoccur.
Image credit: NASA Earth Observatory image by Joshua Stevens, using Suomi NPP VIIRS data from Miguel Román, NASA's Goddard Space Flight Center.
Profile of NAS member and biochemist Edward H. Egelman.