Author response:
The following is the authors’ response to the original reviews.
Reviewer #1 (Public review):
Summary:
This is an interesting study on the role of FGF signaling in the induction of primitive streak-like cells (PS-LC) in human 2D-gastruloids. The authors use a previously characterized standard culture that generates a ring of PSLCs (TBXT+) and correlate this with pERK staining. A requirement for FGF signaling in TBXT induction is demonstrated via pharmacological inhibition of MEK and FGFR activity. A second set of culture conditions (with no exogenous FGFs) suggests that endogenous FGFs are required for pERK and TBXT induction. The authors then characterize, via scRNA-seq, various components of the FGF pathway (genes for ligands, receptors, ERK regulators, and HSPG regulation). They go on to characterize the pFGFR1, receptor isoforms, and polarized localization of this receptor. Finally, they perform FGF4 inhibition and use a cell line with a limited FGF17 inactivation (heterozygous null) and show that loss of these FGFs reduces PS-LC and derivative cell types.
Strengths:
(1) As the authors point out, the role of FGF signaling in gastrulation is less well understood than other signaling pathways. Hence this is a valuable contribution to that field.
(2) The FGF4 and FGF17 loss-of-function experiments in Figure 5 are very intriguing. This is especially so given the intriguing observation that these FGFs appear to be dominating in this model of human gastrulation, in contrast to what FGFs dominate in mice, chicks, and frogs.
(3) In general this paper is valuable as a further development of the Human gastruloid system and the role of FGF signaling in the induction of PS-CLs. The wide net that the authors cast in characterizing the FGF ligand gene, receptor isoforms, and downstream components provides a foundation for future work. As the authors write near the beginning of the Discussion "Many questions remain."
We thank the reviewer for these positive comments.
Weaknesses:
(1) FGFs are cell survival factors in various aspects of development. The authors fail to address cell death due to loss of FGF signaling in their experiments. For example, in Figure 1E (which requires statistical analysis) and 1G (the bottom FGFRi row), there appears to be a significant amount of cell loss. Is this due to cell death? The authors should address the question of whether the role of FGF/ERK signaling is to keep the cells alive.
Indeed, FGF also strongly affects cell survival and it is an interesting question to what extent this depends on ERK. Our manuscript focuses instead on the role of FGF/ERK signaling in cell fate patterning. As mentioned in our discussion, figure 1de show that doxycycline induced pERK leads to more TBXT+ cells than the control without restoring cell number, suggesting the role of FGF in controlling cell number is independent of the requirement for FGF/ERK in PS-LC differrentiation. To further support this, we have added data showing low doses of MEKi are sufficient to inhibit differentiation without affecting cell number (Supp. Fig. 1i).
To address the reviewers question regarding the cause of cell loss, we now stained for BrdU and cleaved Cas3 to assess proliferation and apoptosis in the presence and absence of MEK and FGFR inhibition (new Supp. Fig.
1ef). This shows that the effect of these inhibitors on cell number is primarily due to a reduction in proliferation. We have also included statistical analysis in Fig.1e.
(2) Regarding the sparse cells in 1G, is there a reduction in cell number only with FGFRi and not MEKi? Is this reproducible? Gattiglio et al (Development, 2023, PMID: 37530863) present data supporting a "community effect" in the FGF-induced mesoderm differentiation of mouse embryonic stem cells. Could a community effect be at play in this human system (especially given the images in the bottom row of 1G)? If the authors don't address this experimentally they should at least address the ideas in Gattoglio et al.
Indeed, FGFRi reproducibly affects cell number more than MEKi, in line with the fact that pathways other than MAPK/ERK downstream of FGF (e.g. PI3K) play important roles in cell survival and growth. However, we think the lack of differentiation in MEKi and FGFRi in Fig.1g cannot be attributed to a loss of cells combined with a community effect. This is because without FGFRi or MEKi cells efficiently differentiate to primitive streak at much lower densities than those originally shown, consistent with the data we discuss in response to (1) arguing against a primarily indirect effect of FGF on PS-LC differentiation through cell density. In the context of directed differentiation (rather than 2D gastruloids), we have now shown in a controlled manner that the effect of MEKi and FGFRi does not depend on a community effect by repeating the experiment in Fig.1g while adjusting cell seeding densities to obtain similar final cell densities in all three conditions (new Fig.1g, new Supp Fig.1g). Furthermore we have included new data showing extremely sparse cells without MEKi or FGFRi still differentiate without problems (new Supp Fig 1h). We have also include Gattoglio et al in our revised discussion.
(3) Do the FGF4 and FGF17 LOF experiments in Figure 5 affect cell numbers like FGFRi in Figure 1?
We did not observe major changes in cell number in the FGF4 and FGF17 loss of function experiments. This is in line with our observation that low levels of ERK signaling are sufficient to maintain proliferation (new Supp. Fig. 1i), and the fact that low levels of ERK signaling are maintained in the absence of FGF4 and FGF17 (Fig.5), likely by FGF2 (Fig. 2). In contrast, FGFRi treatment in Fig.1 leads to a nearly complete loss of FGF signaling (ERK and other pathways) that has a dramatic effect on cell number.
Why examine PS-LC induction only in FGF17 heterozygous cells and not homozygous FGF17 nulls?
We were unable to obtain homozygous FGF17 nulls, it is not clear if there is a reason for this. In the absence of homozygous nulls, we have now further corroborated our findings with additional knockdown data (described in response to other comments below).
(4) The idea that FGF8 plays a dominant role during gastrulation of other species but not humans is so intriguing it warrants deeper testing. The authors dismiss FGF8 because its mRNA "...levels always remained low." (line 363) as well as the data published in Zhai et al (PMID: 36517595) and Tyser et al (PMID: 34789876). But there are cases in mouse development where a gene was expressed at levels so low, that it might be dismissed, and yet LOF experiments revealed it played a role or even was required in a developmental process. The authors should consider FGF8 inhibition or inactivation to explore its potential role, despite its low levels of expression.
We thank the reviewer for this suggestion. We have now analyzed the role of FGF8 using FISH to visualize its expression and siRNA to understand its function (Fig.5d,f,h; Supp.Fig.5e,g,6e). We found that FGF8 expression is higher earlier in differentiation, preceding most expression of TBXT. Our scRNA-seq only analyzed samples at 42h so did not capture this. Furthermore, FGF8 expression localized inside the PS-like ring rather than coinciding with it like FGF4. Surprisingly, FGF8 knockdown led to an increase in primitive streak-like differentiation, suggesting it may counteract FGF4. The results are shown in the revised Fig. 5 and Supplemental Fig. 5. While this certainly merits further investigation, understanding the role of FGF8 in more detail is beyond the scope of the current work.
(5) Redundancy is a common feature in FGF genetics. What is the effect of inhibiting FGF4 in FGF17 LOF cells?
Further siRNA and shRNA experiments showed that FGF17 knockdown had a much smaller effect than FGF4 knockdown on expression of primitive streak markers (Fig.5i, Supp.Fig.6f-i) but that FGF17 knockdown did lead to a complete loss of the mesoderm marker TBX6 (Fig.5j, Supp.Fig.6j). A double knockdown of FGF4+FGF17 looked similar to FGF4 alone (Supp.Fig.6k). Thus, we now think the more likely scenario is that FGF17 is downstream of FGF4-dependent PS-differentiation and although this may have a positive feedback effect whereby this FGF17 can then enhance further PS-differentiation, which we previously interpreted as partial redundancy, the primary role of FGF17 may be later, in mesoderm differentiation.
(6) I suggest stating that the authors take more caution in describing FGF gradients. For example, in one Results heading they write "Endogenous FGF4 and FGF17 gradients underly the ERK activity pattern.", implying an FGF protein gradient. However, they only present data for FGF mRNA , not protein. This issue would be clarified if they used proper nomenclature for gene, mRNA (italics), and protein (no italics) throughout the paper.
Thank you for the suggestion. We have edited the paper to more clearly distinguish protein and mRNA. We do think our data provide substantial indirect evidence for a protein gradient which is what the results heading is meant to convey. Receptor activation is high where ERK activity is high (Fig.3), and receptor activation is limited by ligands, since creating a scratch to let exogenous FGF reach the basal side of cells in the center leads to receptor activation (Fig.4). This strongly suggests ERK activity reflects an FGF protein gradient.
Reviewer #2 (Public review):
Summary:
The role of FGFs in embryonic development and stem cell differentiation has remained unclear due to its complexity. In this study, the authors utilized a 2D human stem cell-based gastrulation model to investigate the functions of FGFs. They discovered that FGF-dependent ERK activity is closely linked to the emergence of primitive streak cells. Importantly, this 2D model effectively illustrates the spatial distribution of key signaling effectors and receptors by correlating these markers with cell fate markers, such as T and ISL1. Through inhibition and loss-of-function studies, they further corroborated the needs of FGF ligands. Their data shows that FGFR1 is the primary receptor, and FGF2/4/17 are the key ligands for primitive streak development, which aligns with observations in primate embryos. Additional experiments revealed that the reduction of FGF4 and FGF17 decreases ERK activity.
Strengths:
This study provides comprehensive data and improves our understanding of the role of FGF signaling in primate
primitive streak formation. The authors provide new insights related to the spatial localization of the key components of FGF signaling and attempt to reveal the temporal dynamics of the signal propagation and cell fate decision, which has been challenging.
Weaknesses:
Given the solid data, the work only partially clarifies the complex picture of FGF signaling, so details remain somewhat elusive. The findings lack a strong punchline, which may limit their broader impact.
We thank this reviewer for their valuable feedback and compliment on the solidity of our data. The punchline of our work is that FGF4 and FGF17-dependent ERK signaling plays a key role in differentiation of human PS-like cells and mesoderm, and that these are different FGFs than those thought to drive mouse gastrulation. A second key point is that like BMP and TGFβ signaling, FGF signaling is restricted to the basolateral sides of pluripotent stem cell colonies due to polarized receptor expression, which is crucial for understanding the response to exogenous ligands added to the cell medium. Indeed, many facets of FGF signaling remain to be investigated in the future, such as how FGF regulates and is regulated by other signals, which we will dedicate a different manuscript to.
Reviewer #3 (Public review):
Jo and colleagues set out to investigate the origins and functions of localized FGF/ERK signaling for the differentiation and spatial patterning of primitive streak fates of human embryonic stem cells in a well-established micropattern system. They demonstrate that endogenous FGF signaling is required for ERK activation in a ringdomain in the micropatterns, and that this localized signaling is directly required for differentiation and spatial patterning of specific cell types. Through high-resolution microscopy and transwell assays, they show that cells receive FGF signals through basally localized receptors. Finally, the authors find that there is a requirement for exogenous FGF2 to initiate primitive streak-like differentiation, but endogenous FGFs, especially FGF4 and FGF17, fully take over at later stages.
Even though some of the authors' findings - such as the localized expression of FGF ligands during gastrulation and the importance of FGF/ERK signaling for cell differentiation in the primitive streak - have been reported in model organisms before, this is one of the first studies to investigate the role of FGF signaling during primitive streak-like differentiation of human cells. In doing so, the paper reports a number of interesting and valuable observations, namely the basal localization of FGF receptors which mirrors that of BMP and Nodal receptors, as well as the existence of a positive feedback loop centered on FGF signaling that drives primitive-streak differentiation. The authors also perform a comparison of the role of different FGFs across species and try to assign specific functions to individual FGFs. In the absence of clean genetic loss-of-function cell lines, this part of the work remains less strong.
We thank the reviewer for emphasizing the value of our findings in a human model for gastrulation. We agree more loss-of-function experiments would provide further insight into the role of different FGFs. While we did not manage to create knockout cell lines, we have now performed both siRNA and shRNA knock-down of all FGF4, and FGF17 in two different hPSC lines, performed siRNA knockdown of FGF8, and also made a FGF4+FGF17 shRNA double knockdown cell lines to more completely test the functions of the individual FGFs (Fig.5, Supp.Fig.5,6). Our data suggest FGF17 may be downstream of FGF4 and primarily required for mesoderm differentiation while FGF8 appears to counteract FGF4. In doing this we have added a large amount of new data to the manuscript and we have removed the heterozygous knockout data in the first version of the manuscript which we felt added little to the new data. Further experiments are still needed to solidify our interpretation but those are beyond the scope of the current work.
Reviewer #1 (Recommendations for the authors):
(1) FGF2 is added to culture experiments (e.g. Figure 4), but the commercial source is not mentioned in Methods. For example, it could be added to "Supplementary Table 1: Cell signaling reagents."
We apologize for this oversight and have now added the information to Supplementary Table 1.
(2) Line 117-118: "For example, by controlling the expression of Wnt or Nodal which are both required for PS-like differentiation". It is clear what the authors mean, but this is not a complete sentence.
We edited this for clarity, it now reads: “First, is FGF/ERK signaling required directly for PS-like differentiation, or does it act indirectly? These possibilities are not mutually exclusive. For example, FGF/ERK could be required directly but also act indirectly by controlling Wnt or Nodal expression, as both Wnt and Nodal signaling are required for PS-like differentiation.”
(3) Line 246 "...found its spatial pattern to strongly resembles that of pERK..." either remove "to" or change "resembles" to "resemble"
Thank you for catching this. We removed “to”.
(4) Lines 391- 393 seem to be missing a word in the last phrase: "...with FGF17 more important continued differentiation to mesoderm and endoderm." Maybe "during" after the word "important"?
Thank you for catching this, indeed the word “during” was missing and we have now added it.
(5) Please define acronyms in Figure 3D (PS-LC was defined previously, but not others).
We apologize for the oversight, we have now defined the acronyms.
(6) The three blue lines in Figure 5B (right) are hard to discern (and I'm not colorblind). I suggest also using a variety of dotted lines in a subset of these FGFs.
Thanks you for the suggestion. We have now given all the FGFs colors that are more clearly distinct and made the TBXT and TBX6 lines dashed.
Reviewer #2 (Recommendations for the authors):
(1) The reviewer acknowledges that FGF signaling is complex, particularly when dynamics and its correlation with cell fates are considered. To improve the clarity of the findings, the authors are encouraged to provide an additional schematic figure that clearly delineates the main findings of this study.
Thank you for the suggestion. We have now added a summary figure (Fig.6) to our discussion, which we hope helps present our findings more clearly.
(2) The data suggest that FGF signaling may function differently in mice compared to primates, and their stem cell model aligns more closely with the latter. While the authors discuss this in the contents only based on sequencing data, it would be valuable to conduct some experiments with mouse embryos to validate the key differences.
It is unclear to us which experiments the reviewer has in mind. There is ample data on FGF expression in the mouse literature, as are many knockout phenotypes. Furthermore, verifying loss of function phenotypes (e.g. FGF17 knockout) in mouse is beyond our expertise.
(3) Heparan sulfate proteoglycan (HSPG) is mentioned as an important component of FGF signaling; however, the only data related to HSPG is single-cell sequencing results. The authors should consider performing immunostaining or other assays to validate HSPG expression and spatial distribution, similar to the approach they used for other signaling components.
Our scratch experiments in Fig. 4 strongly argue against HSPGs as being responsible for the spatial pattern of FGF receptor activation: after a scratch across the colony the response is strong all along the scratch as expected if presence of FGF (an FGF gradient) controls the level of activity. If HSPGs were limiting, FGF flowing in from the media show not be able to uniformly activate receptors around the scratch.
In addtion, we have now included an immunostain for HS in a newly added Supp. Fig. 4 which does not explain the observed pattern of ERK signaling.
(4) In the scratch experiment, particularly high PERK expression is observed at the edge of the scratch. The authors should provide an explanation for why this expression is significantly higher compared to the edges of the colony. Additionally, it would be interesting to investigate the fate of the cells with super high PERK expression.
We have now determined that adaptive response to FGF is the reason that the response around the scratch is initially much higher than in the ERK activity ring that overlaps with the primitive streak-like cells. We have added figures showing that although the intial response to FGF exposure after scratching is very high, the response around the scratch adapts to levels similar in those in the ERK ring over the course of 6 hours (Fig.4ij).
(5) For some of the key experiments, multiple cell lines should be used to ensure that the findings are reproducible and applicable across different human stem cell lines.
We have now checked FISH stainings and knockdown phenotypes for different FGFs in two different cell lines: ESI17 (hESC, XX) and PGP1 (hiPSC, XY). These results are shown in Supplementary Figures 6. We found all results to be consistent.
(6) Where applicable, the meaning of error bars needs to be more clearly presented, including details on the number of independent experiments or samples used.
Thank you for pointing this out. Where error bar definitions were missing we have now added them to the figure captions.
Reviewer #3 (Recommendations for the authors):
(1) The authors only analyze the ppERK ring in micropatterns of a single size. What was the motivation for the choice of this size? Can the authors how the ppERK ring is expected to depend on colony size?
Much smaller patterns lose the interior pluripotent regions while much larger patters have a much larger pluripotent region, which requires larger tilings to image without providing additional insight. The colony sizedependence of cell fate patterning was described in the paper that established the 2D gastruloids model (Warmflash Nat Methods 2014) and we later showed this due to a fixed length scale of the BMP and Nodal signaling gradients from the colony edge (Jo et al Elife 2022). We have now included data showing that the ERK patterns behaves similarly, with a fixed length scale of the pattern implying that in smaller colonies the ERK ring becomes a disc and the entire center of the colony has high ERK signaling (Supp Fig 1a).
(2) The scRNAseq is somewhat confusing - why do the two datasets not overlap in the PHATE representation? This is unexpected, because the two samples have been treated similarly, and the authors have integrated their data to iron out possible batch effects. This discrepancy should be discussed. The authors should also specify from which reference exactly the first dataset comes from.
The two datasets do overlap nicely, the same fates are well mixed in the same place and the gene expresison profiles for the integrated data (e.g., Fig.2e) look smooth, so we believe the integration is good, but different cell fates are represented to different degrees. In particular, sample 2 shows much more mesoderm differentiation making the mesoderm branch mostly orange. Occassionally samples differentiate faster or slower than average which we see here, and these samples were collected far apart in time. We do not believe this affects our conclusions, if anything, we think performing the analysis on two samples that differ this much should make the conclusions more robust.
(3) If find it intriguing that exogenous FGF2 is important early on for primitive streak-like differentiation, although the authors show that it does not reach the center of the colony. The authors may want to discuss this conundrum. Does the FGF2 effect propagate from the outside to the inside, or does it act at an early stage when the cells have not yet formed a tight epithelium on the micropattern?
The cells in the experiment in Fig. 5a were given 24h to epithelialize, so we we do believe it acts from the edge. We believe this may be due to FGF2 modulating the early BMP response on the edge and are working on a manuscript that further explores this pathway crosstalk.
(4) The authors' statement that FGF4 and FGF17 have partially redundant functions is not very strong, mainly because the study lacks a full FGF17 loss-of-function cell line. If the authors wanted to improve on this point, they could knock down FGF4 in the FGF17 heterozygous line, or produce a homozygous FGF17 KO line. If there are specific reasons why FGF17 homozygous lines cannot be produced, this could be interesting to discuss, too. Finally, I noticed that the methods list experiments with an FGF17 siRNA, but these are not shown in the manuscript.
We agree our evidence was previously not as strong as it could be. While there is no reason we know of why homozygous knockout lines cannot be produced, we failed to produce on. To strengthen our evidence we have therefore included substantial new knockdown data. We have now performed both siRNA and shRNA knockdown of all FGF4, and FGF17 in two different hPSC lines, performed siRNA knockdown of FGF8, and also made a FGF4+FGF17 shRNA double knockdown cell lines to more completely test the functions of the individual FGFs (Fig.5, Supp.Fig.5,6). These experiments showed that FGF17 knockdown had a much smaller effect than FGF4 knockdown on expression of primitive streak markers (Fig.5i, Supp.Fig.6f-i) but that FGF17 knockdown did lead to a complete loss of the mesoderm marker TBX6 (Fig.5j, Supp.Fig.6j). A double knockdown of FGF4+FGF17 looked similar to FGF4 alone (Supp.Fig.6k). Thus, we now think the more likely scenario is that FGF17 is downstream of FGF4-dependent PS-differentiation and although this may have a positive feedback effect whereby this FGF17 can then enhance further PS-differentiation, which we previously interpreted as partial redundancy, the primary role of FGF17 may be later, in mesoderm differentiation. Furthermore, our new data suggests FGF8 may counteract FGF4 and limit PS-like differentiation.
Minor
(5) Line 63: Reference(s) appear to be missing.
This whole paragraph summarizes the results of the references given on line 55, we have now repeated the relevant references where the reviewer indicated.
(6) Supplementary Figure 1a,b does not show ppERK, unlike stated in lines 102 - 104.
Indeed, the data described in lines 102-104 is shown in Fig.1a and we have removed the original Supplementary Figure 1ab since it did not provide relevant information.
(7) Line 201: It is not clear whether this is a new sequencing dataset, or if existing datasets have been reanalyzed.
We agree our description was unclear. We have edited the text, which now explicitly states that our analysis is based on one dataset we collected previously and a replicate that was newly collected and deposited on GEO for this manuscript.
(8) Figure 2f; Supplementary Figure 2b, c: The colors need to be explained in scale bars. How has this data been normalized to allow for comparison between very different sample types?
We have now added color bars indicating the scale for each of these figure panels. As the caption stated, the interspecies comparison was normalized within each species, so the highest FGF level for any FGF at any time within each species is normalized to one. We are thus comparing between species the relative expression of different FGFs within each species. Indeed there is no good way to compare absolute expression between species. For extra clarity we have expanded our description of the interspecies comparison analysis and normalization in the methods section.
(9) Line 232: Where is the expression of SEF shown?
It is shown in Fig. 2i, under the official gene name IL17RD.
(10) Supplementary Figure 4 seems to be missing.
Thank you for pointing this out. We have now added a supplementary Fig.4.
(11) Line 437: Citation needed.
We have included citations now.
(12) Line 439: A similar feedback loop has been proposed to operate during mesoderm differentiation in mouse ESC (pmid: 37530863 ). The authors may consider citing this work.
Thank you for the suggestion, we have now included this work in the discussion. The feedback loop proposed in that work involves FGF8, while we were trying to explain why FGF4 and not FGF8 appears to be conserved across species by invoking an FGF4 feedback loop. Thus, it becomes even harder to explain differences in FGF4 and FGF8 expression between human and mouse gastrulation.
(13) Supplementary Figure 6 is not described in the main text.
We have removed the original Supplementary Figure 6 and corresponding heterozygous knockout data in the main figure which we felt added little to the extensive knockdown data we now present. We did create a new Supplementary Figure 6 showing additional knockdown data which is described in the main tekst.
(14) Submission of sequencing data to GEO needs to be updated.
We have now made the GEO data public.
