Early anteroposterior regionalisation of human neural crest is shaped by a pro-mesodermal factor
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
This manuscript addresses the important question of how cell types acquire regional identity during embryonic development. The authors study the role of TBXT in the establishment of posterior identity and show unexpected temporally restricted and cell-specific modes of acquisition of posterior identities in neural crest and spinal cord cells. They conclude that Wnt signaling influences posterior identity acquisition in neural crest cells whereas FGF is the main driver for spinal cord axial patterning.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Developmental Biology (eLife)
Abstract
The neural crest (NC) is an important multipotent embryonic cell population and its impaired specification leads to various developmental defects, often in an anteroposterior (A-P) axial level-specific manner. The mechanisms underlying the correct A-P regionalisation of human NC cells remain elusive. Recent studies have indicated that trunk NC cells, the presumed precursors of childhood tumour neuroblastoma, are derived from neuromesodermal-potent progenitors of the postcranial body. Here we employ human embryonic stem cell differentiation to define how neuromesodermal progenitor (NMP)-derived NC cells acquire a posterior axial identity. We show that TBXT, a pro-mesodermal transcription factor, mediates early posterior NC/spinal cord regionalisation together with WNT signalling effectors. This occurs by TBXT-driven chromatin remodelling via its binding in key enhancers within HOX gene clusters and other posterior regulator-associated loci. This initial posteriorisation event is succeeded by a second phase of trunk HOX gene control that marks the differentiation of NMPs toward their TBXT-negative NC/spinal cord derivatives and relies predominantly on FGF signalling. Our work reveals a previously unknown role of TBXT in influencing posterior NC fate and points to the existence of temporally discrete, cell type-dependent modes of posterior axial identity control.
Article activity feed
-
-
Author Response
Reviewer #1 (Public Review):
“The authors suggest that they uncovered two distinct phases of how the posterior axial identity is controlled; the first involving TBXT/Wnt to generate posterior 'uncommitted progenitors', which then go on to generate NCCs, and the second involving FGF to impart posterior axial identity onto CNS/spinal cord cells.”
Based on our new data we have slightly modified our model: (i) TBXT controls posterior axial identity acquisition in NMP precursors and both their trunk NC and CNS spinal cord derivatives; (ii) this early, TBXT-driven posteriorisation phase appears to be WNT dependent; (iii) a subsequent TBXT/WNT-independent phase of Hox cluster regulation occurring during the transition of NMPs towards their NC/spinal cord derivatives is controlled predominantly by FGF signalling. This model …
Author Response
Reviewer #1 (Public Review):
“The authors suggest that they uncovered two distinct phases of how the posterior axial identity is controlled; the first involving TBXT/Wnt to generate posterior 'uncommitted progenitors', which then go on to generate NCCs, and the second involving FGF to impart posterior axial identity onto CNS/spinal cord cells.”
Based on our new data we have slightly modified our model: (i) TBXT controls posterior axial identity acquisition in NMP precursors and both their trunk NC and CNS spinal cord derivatives; (ii) this early, TBXT-driven posteriorisation phase appears to be WNT dependent; (iii) a subsequent TBXT/WNT-independent phase of Hox cluster regulation occurring during the transition of NMPs towards their NC/spinal cord derivatives is controlled predominantly by FGF signalling. This model is shown in Figure 9 in the revised manuscript.
“I am not convinced that their data show this; it is equally possible that NMPs are heterogeneous and the effects observed simply reflect a differential response of cells or selection. Since the authors largely analyse their data by qPCR it is difficult to disentangle this.”
We believe that the inclusion of new data defining the emergence of NMP derivatives at the single cell level through analysis of key trunk lineage-specific markers (HOXC9, SOX10, SOX1, SOX2) via immunostaining and image analysis/flow cytometry (see Figure 3-figure supplement 1, Figure 4C-D, Figure 5-figure supplement 1, Figure 7D-E in revised manuscript) should address the reviewer’s point. See also our response to the editorial comments above. It should be note that the vast majority of day 3 hESC-derived NMPs (>95%) is positive for TBXT protein expression based on antibody staining and thus the starting population for the generation of trunk NC/spinal cord progenitors can be considered largely homogeneous when it comes to the expression of this transcription factor.
“The authors include some expression data in mouse to support their in vitro findings. However, these need to be explained and integrated better.”
We hope that breaking down figure 4 and the related text into two parts has improved the integration of the in vivo data in the revised version of the manuscript.
Reviewer #2 (Public Review):
“The fact that the regimes are distinct makes the comparisons of neural crest versus spinal cord difficult to interpret as the cells have been exposed for different amounts of time to WNT and FGF when they asses the Hox code in neural crest or spinal cord cells. Specially because the spinal cord induction protocol involves four additional days of culture with FGF and CHIR, and the cells after seven days are not mature neural progenitors.
To address this point, we employed “neutral”, extrinsic signal-free culture conditions that drive NMPs towards a mixture of early pre-neural spinal cord progenitors and mutually exclusive SOX1+HOXC9+ CNS spinal cord and SOX10+HOXC9+ NC populations. This facilitated the effective assessment of cell fate and posterior axial identity acquisition simultaneously in both NMP-derived spinal cord and NC cells, during discrete time windows of TBXT knockdown (Figure 4 in revised manuscript). For details see our response above.
Likewise, the authors have previously shown that such a treatment induces the expression of dorsal neural tube/early neural crest markers”.
Although we have no evidence of SOX10 expression in cultures generated from NMPs following WNT and FGF agonist treatment for 4 days indicating absence of definitive NC cells, we opted to remove the “CNS” references when describing this cell population to accommodate for the possibility that it may be NC-potent given its previously described dorsal neural tube/early NC character (Cooper et al, 2022; Wind et al., 2021).
“It would be good to see some quality controls on the percentages of neural crest progenitors or spinal cord neural progenitors that they get in each signalling regime. Can the authors separate neural progenitor cells and neural crest cells (for example by FACS sorting with specific markers) to confirm the cell-type specific expression of the HOX genes in these experiments?”.
As mentioned above, we have now included immunostaining data quantifying thoroughly the induction of trunk SOX1+HOXC9+ CNS spinal cord and SOX10+HOXC9+ NC cells under different culture conditions/TBXT levels (see Figure 4C-D, Figure 5-figure supplement 1, Figure 7 and Figure 7-figure supplement 1).
“In the neural crest differentiation protocol, there is a slight, non-significant upregulation of neural progenitor markers following TBXT knockdown, can the authors quantify the percentage of neural cells in their cultures to see how much of the observed effect is specific to neural crest cells?”
We have quantified the emergence of SOX1+ CNS spinal cord progenitor cells in NMPderived trunk NC cultures using both FACS/intracellular staining and immunostaining/image analysis but their numbers are too small (2-3% of total cells with no statistically significant difference between control and TBXT knockdown cells, see Figure 3-figure supplement 1) to extract any meaningful conclusions on the effect of TBXT depletion on them. However, quantification of SOX1+HOXC9+ cells generated from NMPs upon culture in “neutral” basal conditions revealed that TBXT depletion results in a decrease in their number in addition to its established impact on trunk NC (see Figure 4C-D in revised manuscript).
“Previous work from the lab showed that a 3-day FGF/CHIR treatment of hESCs followed by a two-day incubation on basal medium is sufficient to induce neural progenitors that express Hox genes of posterior identity (PMID: 25157815). Can the authors draw the same conclusions for the spinal cord cells with this protocol if they deplete TBXT during the first three days and assay at day 7 the cells on basal medium, or if they deplete TBXT during the last four days of the protocol? The comparison of the 3-day FGF/CHIR regime followed by basal medium treatment versus the continuous FGF/CHIR for a 7-day period may help clarify the temporal and cell-type specific effects of the HOX code via TBXT/FGF on the neural crest and/or spinal cord cells”.
We have carried out this experiment as suggested by the reviewer (Figure 4C-D/line numbers 226-256 in the revised manuscript), for details see our responses above.
“In their data, it seems that anterior HOX genes (PG1-5) as well as other posterior HOX (PG6-9) are expressed in wild-type posterior neural crest and early spinal cord cells. Can HOX genes that mark posterior cranial, vagal or trunk identities be co-expressed in trunk neural crest or spinal cord cells? Is it possible that the differentiations generate cells that have different axial identities? I wonder if this interpretation comes from the normalization. Perhaps the authors could clarify if the levels of expression of the 3' Hox genes are higher or lower than 5' Hox genes in their differentiations”.
Co-expression of HOX paralogous group (PG) (1-5) and (6-9) transcripts does occur in the posterior part of the mouse embryo around E9.5, both in the NMP-containing tailbud region (Gouti et al, 2017) as well as in differentiated posterior neural/neural crest cells e.g. for Hoxb1 expression in E9.5 mouse embryos see (Arenkiel et al, 2003; Glaser et al, 2006); for Hoxc9 expression see (Bel et al, 1998). Thus, the presence of HOXPG(1-5) transcripts in HOXC9+ trunk NC cells is not surprising and in line with what has been reported previously in other studies describing the generation of posterior NC/spinal cord cell types from hESC/NMPs (Frith et al., 2018; Hackland et al, 2019; Lippmann et al, 2015; Mouilleau et al, 2021). Alternatively, the simultaneous detection of transcripts belonging to both HOXPG(1-5) and HOXPG(6-9) could indicate the co-emergence of a separate population of posterior cranial/cardiac/vagal NC cells during trunk NC differentiation. Moreover, the detection of HOX transcripts does not always correlate with corresponding protein positivity (Faustino Martins et al, 2020) pointing to the existence of post-transcriptional/-translational mechanisms controlling HOX protein expression. Unfortunately, we have not identified reliable (in our hands) antibodies against HOXPG(1-5) members that we can use together with HOXC9 in order to distinguish between these possibilities.
“In the experiments where the authors asses if TBXT binds directly or indirectly to the HOX clusters, the authors compare pluripotent cells with hNMPs. This data confirms that TBXT acts as an activator in hNMPs and that it binds to regions in the HOX clusters. Do the HOX regions overlap with known enhancers for the HOX genes for neural crest or spinal cord?”
We have included new ATAC-seq data mapping chromatin accessibility in day 8 trunk NC cells generated from TBXT-depleted and control hESC-derived NMPs. These data, combined with the ATAC-seq and TBXT ChIP-seq analyses from day 3 hESC-derived NMPs, indicate that TBXT controls chromatin accessibility in trunk NC-specific enhancers within HOX clusters, both directly through genomic binding, and indirectly possibly by influencing expression of other key transcriptional regulators such as CDX2. For details see Figure 8-figure supplement 2 and Appendix Table S9 and line numbers 458-482 in the revised manuscript.
“As they see distinct temporal phases of TBXT activity on spinal cord progenitors versus neural crest cells, the authors should test if there are changes in accessibility or TBXT binding in neural crest and spinal cord cells in the HOX locus and/or genome-wide. This comparison may help identify cell-type specific TBXT targets (perhaps acting with distinct coactivators) that are key in the two distinct phases of posterior axial identity control”.
As mentioned above, we have added new ATAC-seq data from analysis of trunk NC cells derived from TBXT knockdown shRNA hESC-derived NMPs in the presence and absence of Tet. These data can be found in Figure 8-figure supplement 2 and Appendix Table S9 in the revised manuscript. As expected, ATAC-seq analysis of pre-neural CNS spinal cord progenitors generated from TBXT knockdown shRNA hESC-derived NMPs in the presence and absence of Tet showed no significant differences in chromatin accessibility between the two conditions again our gene expression data (Figure 6 in revised manuscript). These data were not included in the new manuscript version but they are publicly available as part of our revised GEO submission (GSE184227). Mapping of TBXT genomic binding in NMP-derived trunk NC cells/spinal cord progenitors is not feasible due to the very low/absent expression of TBXT protein in these cell populations. See also our response to the editor’s suggestions.
“In the experiments where the authors examine the signalling pathway dependence of HOX expression during the transition in the neural crest differentiation protocol, it appears that CHIR/LDN treatment induces the highest levels of HOX expression (FIG 3F). Also, there is an increased expression of SOX1 while SOX10 expression is not detected "pointing to a role for BMP signalling in steering NMPs/dorsal pre-neural progenitors toward a NC fate in agreement with previous observations". The results may indicate that WNT and BMP inhibition may induce HOX gene expression in neural cells irrespective of FGF. How do the authors interpret this? How does it affect their final model where FGF (and not WNT) drives the expression of HOX genes in late pre-neural spinal cord progenitors?”.
Based on our data and published work, we speculate that during the transition of hESCderived NMPs towards trunk NC cell, cultures still exhibit autocrine and/or paracrine FGF signalling even in the absence of exogenous FGF agonist supplementation. This is supported by previous reports showing the expression of the active, phosphorylated version of the FGF effector ERK1/2 in differentiating pluripotent stem cells cultured in FGF-free media (Diaz-Cuadros et al, 2020; Stavridis et al, 2007; Ying et al, 2003). This endogenous FGF activity is probably sufficient for the maintenance of HOX gene expression in these cells, while exogenous BMP signalling stimulation is required for the induction of a NC fate. Given the reported antagonism between these two pathways during early neural/NC induction (Anderson et al, 2016; Marchal et al, 2009), treatment with the BMP inhibitor LDN193189 results in FGF signalling potentiation, which in turn leads to increased HOX gene expression and a switch toward a CNS neurectodermal fate at the expense of NC. Further work is needed to mechanistically dissect this hypothesis, which is beyond the scope of this manuscript.
“The identity of the cells in the inhibition of WNT or FGF treatments during the final four days towards spinal cord cells experiments is unclear. It would be very useful if the authors could characterize what cell types emerge after the treatments. In principle, I would expect that these treatments would generate different progenitor types (FGF inhibition may presumably give rise to mesoderm cells, whereas WNT inhibited may be pre-neural). Why would the authors expect these different cell types to have similar levels of expression of WNT targets or Hox genes?”
The inclusion of the new immunostaining data and the quantification of the proportions of SOX2+HOXC9+ emerging upon various WNT/FGF inhibitor treatments (Figure 7D-E in revised manuscript) has now enabled us to define the role of these signalling pathways in controlling HOX gene expression specifically in pre-neural spinal progenitors thus confirming our conclusions from the qPCR data without any bias introduced from contaminating, nonneural HOXC9+ cells.
-
Evaluation Summary:
This manuscript addresses the important question of how cell types acquire regional identity during embryonic development. The authors study the role of TBXT in the establishment of posterior identity and show unexpected temporally restricted and cell-specific modes of acquisition of posterior identities in neural crest and spinal cord cells. They conclude that Wnt signaling influences posterior identity acquisition in neural crest cells whereas FGF is the main driver for spinal cord axial patterning.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)
-
Reviewer #1 (Public Review):
In this paper the authors explore how trunk neural crest cells (NCCs) acquire regional identity in human ESC differentiation. Following from earlier findings that NMPs in vivo and NMP-like cells in vitro give to trunk neural crest, they now show that the transcription factor TBXT is required for the acquisition of posterior identity of NMPs and their derivative NCCs. When TBXT is reduced in hESCs they do not activate Hox gene expression or the expression of Wnt targets. Using a combination of TBXT ChIPseq in NMPs and ATACseq in control and TBXT depleted NMPs, they show that TBXT binds close to the TSS of genes whose expression is downregulated in the absence of TBXT and that in the absence of TBXT such regions lose their accessibility. These data suggest that TBXT mediates chromatin opening and subsequent …
Reviewer #1 (Public Review):
In this paper the authors explore how trunk neural crest cells (NCCs) acquire regional identity in human ESC differentiation. Following from earlier findings that NMPs in vivo and NMP-like cells in vitro give to trunk neural crest, they now show that the transcription factor TBXT is required for the acquisition of posterior identity of NMPs and their derivative NCCs. When TBXT is reduced in hESCs they do not activate Hox gene expression or the expression of Wnt targets. Using a combination of TBXT ChIPseq in NMPs and ATACseq in control and TBXT depleted NMPs, they show that TBXT binds close to the TSS of genes whose expression is downregulated in the absence of TBXT and that in the absence of TBXT such regions lose their accessibility. These data suggest that TBXT mediates chromatin opening and subsequent activation of these transcripts. Finally, the authors also suggest that acquisition of posterior character in NCCs is largely dependent on Wnt signalling, while posterior spinal cord cells largely depend on FGF signalling.
The role of FGF and Wnt signalling in establishing anterior-posterior identity is well documented and the authors explore these pathways and the role of TBXT in this process using differentiation of human ESCs. The finding that TBXT is required for NMPs and NMP-derived NCCs to acquire posterior identity is interesting, and the authors show that this is likely to involve chromatin accessibility mediated by TBXT and activation of target genes. The involvement of TBXT/Wnt loop in the acquisition of posterior NCC identity is a new finding, and the authors provide an underlying molecular mechanism.
The authors suggest that they uncovered two distinct phases of how the posterior axial identity is controlled; the first involving TBXT/Wnt to generate posterior 'uncommitted progenitors', which then go on to generate NCCs, and the second involving FGF to impart posterior axial identity onto CNS/spinal cord cells. I am not convinced that their data show this; it is equally possible that NMPs are heterogeneous and the effects observed simply reflect a differential response of cells or selection. Since the authors largely analyse their data by qPCR it is difficult to disentangle this.
Some conclusions rely on the changes in expression of just a handful of markers; since gene expression changes dynamically during development it is important to acknowledge that the interpretation is very dependent on the stage examined.
The authors include some expression data in mouse to support their in vitro findings. However, these need to be explained and integrated better.
-
Reviewer #2 (Public Review):
How cell types acquire regional identity during embryonic development remains largely unknown. In this manuscript, Gogolou et al. study the role of TBXT in the establishment of posterior identity of neural crest and spinal cord cells derived from human neuromesodermal progenitors (hNMPs). Previously, it was shown that activation of human pluripotent stem cells with FGF and Wnt/β-catenin, establishes a progressive and full colinear HOX activation in human axial progenitors in vitro. In this manuscript, the authors confirm these findings and show unexpected temporally restricted and cell-specific modes of acquisition of posterior identities in neural crest and spinal cord cells. Specifically, they find that TBXT depletion impedes posterior identity acquisition in neural crest cells whereas it does not impact …
Reviewer #2 (Public Review):
How cell types acquire regional identity during embryonic development remains largely unknown. In this manuscript, Gogolou et al. study the role of TBXT in the establishment of posterior identity of neural crest and spinal cord cells derived from human neuromesodermal progenitors (hNMPs). Previously, it was shown that activation of human pluripotent stem cells with FGF and Wnt/β-catenin, establishes a progressive and full colinear HOX activation in human axial progenitors in vitro. In this manuscript, the authors confirm these findings and show unexpected temporally restricted and cell-specific modes of acquisition of posterior identities in neural crest and spinal cord cells. Specifically, they find that TBXT depletion impedes posterior identity acquisition in neural crest cells whereas it does not impact spinal cord regionalization. Instead, they find that FGF is the main driver for spinal cord axial patterning.
This work addresses a very important question and the results they provide support their final model. The work opens up new interpretations on how cells define their axial identity and sets the ground to investigate how TBXT cooperates with other transcription factors to establish posterior identities prior to the acquisition of a neural crest or spinal cord fate. Further, this mechanistic insight may help explain the impairment of neural crest specification and HOX dysregulation in neural tube defects.
A rigorous characterization of the cell types that are generated during this differentiation under various signaling regimes is essential to separate the cell-specific effect of WNT and FGF on the HOX code versus the temporally restricted windows in which this can happen. In their experiments towards neural crest or spinal cord differentiation, the starting hNMP population is homologous and it emerges after a 3-day treatment of the cells with FGF and CHIR. Then, the authors use specific signalling regimes for the generation of neural crest or spinal cord cells. The fact that the regimes are distinct makes the comparisons of neural crest versus spinal cord difficult to interpret as the cells have been exposed for different amounts of time to WNT and FGF when they asses the Hox code in neural crest or spinal cord cells. Specially because the spinal cord induction protocol involves four additional days of culture with FGF and CHIR, and the cells after seven days are not mature neural progenitors. Likewise, the authors have previously shown that such a treatment induces the expression of dorsal neural tube/early neural crest markers (PMID: 33658223).
- It would be good to see some quality controls on the percentages of neural crest progenitors or spinal cord neural progenitors that they get in each signalling regime. Can the authors separate neural progenitor cells and neural crest cells (for example by FACS sorting with specific markers) to confirm the cell-type specific expression of the HOX genes in these experiments?
- In the neural crest differentiation protocol, there is a slight, non-significant upregulation of neural progenitor markers following TBXT knockdown, can the authors quantify the percentage of neural cells in their cultures to see how much of the observed effect is specific to neural crest cells?
- Previous work from the lab showed that a 3-day FGF/CHIR treatment of hESCs followed by a two-day incubation on basal medium is sufficient to induce neural progenitors that express Hox genes of posterior identity (PMID: 25157815). Can the authors draw the same conclusions for the spinal cord cells with this protocol if they deplete TBXT during the first three days and assay at day 7 the cells on basal medium, or if they deplete TBXT during the last four days of the protocol? The comparison of the 3-day FGF/CHIR regime followed by basal medium treatment versus the continuous FGF/CHIR for a 7-day period may help clarify the temporal and cell-type specific effects of the HOX code via TBXT/FGF on the neural crest and/or spinal cord cells.
- In their data, it seems that anterior HOX genes (PG1-5) as well as other posterior HOX (PG6-9) are expressed in wild-type posterior neural crest and early spinal cord cells. Can HOX genes that mark posterior cranial, vagal or trunk identities be co-expressed in trunk neural crest or spinal cord cells? Is it possible that the differentiations generate cells that have different axial identities? I wonder if this interpretation comes from the normalization. Perhaps the authors could clarify if the levels of expression of the 3' Hox genes are higher or lower than 5' Hox genes in their differentiations.
- In the experiments where the authors asses if TBXT binds directly or indirectly to the HOX clusters, the authors compare pluripotent cells with hNMPs. This data confirms that TBXT acts as an activator in hNMPs and that it binds to regions in the HOX clusters. Do the HOX regions overlap with known enhancers for the HOX genes for neural crest or spinal cord?
- As they see distinct temporal phases of TBXT activity on spinal cord progenitors versus neural crest cells, the authors should test if there are changes in accessibility or TBXT binding in neural crest and spinal cord cells in the HOX locus and/or genome-wide. This comparison may help identify cell-type specific TBXT targets (perhaps acting with distinct co-activators) that are key in the two distinct phases of posterior axial identity control.
- In the experiments where the authors examine the signalling pathway dependence of HOX expression during the transition in the neural crest differentiation protocol, it appears that CHIR/LDN treatment induces the highest levels of HOX expression (FIG 3F). Also, there is an increased expression of SOX1 while SOX10 expression is not detected "pointing to a role for BMP signalling in steering NMPs/dorsal pre-neural progenitors toward a NC fate in agreement with previous observations". The results may indicate that WNT and BMP inhibition may induce HOX gene expression in neural cells irrespective of FGF. How do the authors interpret this? How does it affect their final model where FGF (and not WNT) drives the expression of HOX genes in late pre-neural spinal cord progenitors?
- The identity of the cells in the inhibition of WNT or FGF treatments during the final four days towards spinal cord cells experiments is unclear. It would be very useful if the authors could characterize what cell types emerge after the treatments. In principle, I would expect that these treatments would generate different progenitor types (FGF inhibition may presumably give rise to mesoderm cells, whereas WNT inhibited may be pre-neural). Why would the authors expect these different cell types to have similar levels of expression of WNT targets or Hox genes?
-
