Single cell transcriptomics reveals spatial and temporal dynamics of gene expression in the developing mouse spinal cord

The data provide a compendium of gene expression for classifying spinal cord cell types.

“provide a compendium of gene expression for classifying spinal cord cell types”

The single-cell mRNA sequencing dataset accurately recapitulates neural tube development.

“We confirmed that the data accurately recapitulates neural tube development, allowing us to identify new markers for specific progenitor and neuronal populations.”

The study recovered 21,465 cells from cervical and thoracic regions of the mouse neural tube across five developmental timepoints from e9.5 to e13.5.

“We recovered 21,465 cells from cervical and thoracic regions of the mouse neural tube across five developmental timepoints from e9.5 to e13.5.”

Using the dataset, the authors inferred developmental trajectories that lead to distinct neural cell fates.

“Using this dataset, we inferred the developmental trajectories that lead to distinct neural cell fates”

Using the dataset, the authors identified cohorts of co-regulated genes involved in specific developmental processes.

“identified cohorts of co-regulated genes that are involved in specific developmental processes”

The analysis identified new markers for specific progenitor and neuronal populations.

“allowing us to identify new markers for specific progenitor and neuronal populations”

Plotting the proportion of progenitors and neurons over time revealed a sharp decrease in the overall proportion of progenitors from >65% of cells at e9.5 to <15% of cells at e13.5.

“Plotting the proportion of progenitors and neurons over time revealed a sharp decrease in the overall proportion of progenitors from >65% of cells at e9.5 to <15% of cells at e13.5”

The combinatorial expression of a curated gene list allowed classification of 33,754 cells, 87% of the total cells.

“This allowed the classification of 33,754 cells, 87% of the total cells.”

The analysis suggested previously unknown expression patterns of TFs, including Hmx2 and Hmx3 in V1 and dI2 neurons.

“This analysis also suggested multiple previously unknown expression patterns of TFs, for example: expression of Hmx2 and Hmx3 in V1 and dI2 neurons”

Motor neurons are the most prominent class of neurons at early developmental stages (e9.5 and e10.5).

“Motor neurons (MNs) are the most prominent class of neurons at early developmental stages (e9.5 and e10.5)”

Plotting the resulting average levels of marker gene expression for the assigned cell identities revealed the well-known patterns of progenitor- and neuron-specific gene expression.

“Plotting the resulting average levels of marker gene expression for the assigned cell identities revealed the well-known patterns of progenitor- and neuron-specific gene expression”

Visualising the dataset with tSNE separated cells into multiple groupings that reflected the anticipated cell types.

“Visualising the resulting dataset with t-distributed stochastic neighbour embedding (tSNE) dimensionality reduction separated cells into multiple groupings that reflected the anticipated cell types”

Between e11.5 and e12.5, the rate of neurogenesis in the spinal cord appears to slow.

“Between e11.5 and e12.5, the rate of neurogenesis in the spinal cord appears to slow”

At later developmental stages (after e11.5), the proportions of inhibitory dI4 and excitatory dI5 neurons increased markedly.

“At later developmental stages (after e11.5), the proportions of inhibitory dI4 and excitatory dI5 neurons increased markedly”

Interrogation of the data recovered nine differentially expressed cell adhesion molecules: Cldn3, Cbln1, Cdh8, Pcdh11x, Clstn2, Pcdh7, Megf11, Nrxn3 and Ret.

“Interrogation of the data recovered nine differentially expressed cell adhesion molecules (Cldn3, Cbln1, Cdh8, Pcdh11x, Clstn2, Pcdh7, Megf11, Nrxn3 and Ret).”

The analysis identified the three main types of spinal cord neurons: cholinergic MNs, inhibitory GABAergic and glycinergic interneurons, and excitatory glutamatergic interneurons.

“These identified the three main types of spinal cord neurons: cholinergic MNs, inhibitory GABAergic and glycinergic interneurons, and excitatory glutamatergic interneurons.”

Most neural crest and mesoderm cells originated from e9.5 and e10.5 embryos.

“Conversely, most of the neural crest and mesoderm cells originated from e9.5 and e10.5 embryos”

From the 4096 combinatorial models applied to 12 neuronal populations, 147 distinct patterns were predicted.

“From the 4096 combinatorial models that were obtained, we predicted 147 distinct patterns.”

Most neurons were contained in the datasets obtained at later developmental stages.

“Most neurons were contained in the datasets that were obtained at later developmental stages”

There were obvious differences in the proportions of cell types from different timepoints.

“By contrast, there were obvious differences in the proportions of cell types from different timepoints.”

Nrxn1 is specifically expressed in excitatory neurons (dI1, dI2, dI3 and dI5).

“Further investigation identified Nrxn1 as specifically expressed in excitatory neurons (dl1, dl2, dl3 and dl5) (data not shown).”

Cldn3 was predicted to be specifically expressed in MNs before, but not after, e11.5.

“was predicted to be specifically expressed in MNs before, but not after, e11.5”

At e13.5, the fraction of progenitors in dorsal domains decreased relative to ventral.

“24 h later, at e13.5, the fraction of progenitors in dorsal domains decreased relative to ventral”

The analysis suggested Sp9 is expressed in V1 neurons.

“Sp9 in V1 neurons”

Approximately 1% of the transcriptomes were from a mixture of neural and mesodermal cell types.

“This indicated that ∼1% of the transcriptomes were from a mixture of neural and mesodermal cell types”

The analysis suggested Bhlhe23 is expressed in dI2 neurons.

“Bhlhe23 in dI2 neurons”

Nrxn3 is specific for inhibitory neurons in the dI4, dI6, V1 and V2b domains.

“We also found Nrxn3 to be specific for inhibitory neurons in the dI4, dI6, V1 and V2b domains.”

Further subclustering the unclassified population did not reveal cell populations with spinal cord identity.

“Further subclustering the unclassified population did not reveal cell populations with spinal cord identity.”

Modules containing Pou2f2 and Nfib were expressed in subpopulations of most neuronal classes.

“Modules containing Pou2f2 and Nfib were expressed in subpopulations of most neuronal classes (Fig. 4B-D and Fig. S4).”

Clade V2a.1 was characterised by expression of Neurod2, Nfia and Nfib.

“The three other neuronal subtypes were characterised by the expression of Neurod2, Nfia and Nfib (clade V2a.1);”

The remaining four clades corresponded to more mature neurons.

“The remaining four clades corresponded to more mature neurons.”

The analysis suggested that Pou3f1 is expressed in dorsal dI1-dI3 neurons, which had not been described before.

“Our analysis also suggested that Pou3f1 is expressed in dorsal dI1-dI3 neurons (Fig. 3A), which, to our knowledge, had not been described before.”

Clade V3.3 expressed the TFs Pou2f2, Zfhx2, Zfhx3, and Zfhx4.

“Clade V3.3 expressed the TFs Pou2f2, Zfhx2, Zfhx3, and Zfhx4”

At e10.5, expression of Onecut2 and Pou2f2 was mutually exclusive.

“At e10.5, expression of both markers was mutually exclusive, with Onecut2-positive neurons typically occupying a more lateral position in the spinal cord than Pou2f2 neurons (Fig. 6C).”

At e10.5, Onecut2-positive neurons typically occupied a more lateral position in the spinal cord than Pou2f2 neurons.

“At e10.5, expression of both markers was mutually exclusive, with Onecut2-positive neurons typically occupying a more lateral position in the spinal cord than Pou2f2 neurons (Fig. 6C).”

Pou2f2 and Nfib were co-expressed with multiple domain-specific markers at the respective stages.

“Furthermore, Pou2f2 and Nfib were co-expressed with multiple domain-specific markers at the respective stages (Fig. S5C-E).”

Cells in clade V3.4 expressed Neurod2, Neurod6 and Nfia/b/x.

“whereas cells in clade V3.4 expressed Neurod2, Neurod6 and Nfia/b/x.”

Cells in clade MN.7 expressed higher levels of Hoxc9, which is characteristic of the thoracic regions.

“By contrast, cells in clade MN.7 expressed higher levels of Hoxc9, which is characteristic of the thoracic regions (data not shown).”

Hierarchical clustering of V3 neurons revealed at least four different subtypes.

“Hierarchical clustering of V3 neurons revealed at least four different subtypes (Fig. 4B).”

Clade V2a.2 was characterised by expression of Shox2, Pou3f1, Pou2f2, Zfhx2, Zfhx3 and Zfhx4.

“Shox2, Pou3f1, Pou2f2, Zfhx2, Zfhx3 and Zfhx4 (clade V2a.2);”

Clade V2a.5 expressed late neural progenitor markers Sox9 and Fabp7, indicative of progenitor identity.

“a second population that expressed late neural progenitor markers Sox9 and Fabp7, which is indicative of progenitor identity (clade V2a.5)”

The transcriptome analysis suggested that Pou2f2 is expressed in a group of V3 interneurons complementary to the Olig3- and Onecut2-expressing populations.

“Our analysis suggested that these TFs are expressed in specific subsets of neurons: Pou2f2 in a group of V3 interneurons that is complementary to the Olig3- and Onecut2-expressing populations;”

Assaying expression in e11.5 and e12.5 cervical spinal cord confirmed each of the three predictions.

“Assaying expression in e11.5 and e12.5 cervical spinal cord (Fig. 4E-G) confirmed each of the predictions.”

Pou3f1, Onecut2 and Maf also demarcated subpopulations of multiple neuronal classes.

“Furthermore, Pou3f1, Onecut2 and Maf also demarcated subpopulations of multiple neuronal classes, which is consistent with the expression of these TFs in multiple types of neurons (Francius et al., 2013).”

• cites Francius et al., 2013 sourceNot checked yet

  Not checked in the free run.

The composition of gene modules containing these TFs was well conserved between neuronal domains.

“The composition of the gene modules that contain these TFs was well conserved between neuronal domains.”

Sets of TFs are used recurrently to define subpopulations of neurons in multiple domains.

“We first identified sets of TFs that are used recurrently to define subpopulations of neurons in multiple domains (Fig. 5).”

Subclustering of the identified MNs revealed six clades.

“Subclustering of the identified MNs revealed six clades (Fig. 4C)”

The proportion of neurons expressing Onecut2 and Pou2f2 peaked at early developmental stages (before e11.5).

“The proportion of neurons that expressed Onecut2 and Pou2f2 peaked at early developmental stages (before e11.5), whereas Nfib and Neurod2 expression were induced at e12.5 or e13.5.”

In most cases the Pou2f2 gene module also contained Zfhx2, Zfhx3 and Zfhx4.

“In most cases the Pou2f2 gene module also contained Zfhx2, Zfhx3 and Zfhx4.”

Clade MN.1 comprised early differentiated MNs characterised by the expression of Pou2f2 and Cldn3.

“Clade MN.1 comprised early differentiated MNs that were characterised by the expression of the TF Pou2f2 and the tight junction component Cldn3.”

Two clades (MN.6 and MN.7) of Foxp1-positive cells were identified.

“Two clades (MN.6 and MN.7) of Foxp1-positive cells were identified.”

Examining average expression of Onecut2, Pou2f2, Zfhx3, Zfhx4, Nfia, Nfib, Neurod2 and Neurod6 for each neuronal class between e9.5 and e13.5 revealed a temporal stratification of neuronal subtypes conserved between domains.

“Examining the average expression level of Onecut2, Pou2f2, Zfhx3, Zfhx4, Nfia, Nfib, Neurod2 and Neurod6 for each neuronal class between e9.5 and e13.5 revealed a temporal stratification of neuronal subtypes that was conserved between domains (Fig. 6A).”

Nfib and Neurod2 expression were induced at e12.5 or e13.5.

“The proportion of neurons that expressed Onecut2 and Pou2f2 peaked at early developmental stages (before e11.5), whereas Nfib and Neurod2 expression were induced at e12.5 or e13.5.”

Clade V2a.3 was characterised by expression of Maf and Onecut2.

“and Maf and Onecut2 (clade V2a.3).”

Hierarchical clustering of the single cell transcriptome data revealed four major subtypes of V2a neurons.

“Consistent with this, hierarchical clustering of the single cell transcriptome data revealed four major subtypes of V2a neurons (Fig. 4D).”

The Nfib gene module often included Nfia, Nfix, Neurod2, Neurod6 and Tcf4.

“By contrast, the Nfib gene module often included Nfia, Nfix, Neurod2, Neurod6 and Tcf4.”

Staining for Zfhx3 confirmed its expression in subsets of neurons at e10.5.

“staining for Zfhx3 confirmed its expression in subsets of neurons at e10.5 (Fig. S5A)”

Nfib and Neurod2 were not expressed in neurons until after e11.5.

“Nfib and Neurod2 were not expressed in neurons until after e11.5 (Fig. 6E and Fig. S5B).”

Clade V3.1 consisted of newly differentiating V3 neurons that expressed the neurogenic markers Neurog3, Hes6 and Neurod4.

“Clade V3.1 consisted of newly differentiating V3 neurons that expressed the neurogenic markers Neurog3, Hes6 and Neurod4.”

Pou3f1 had previously been reported in multiple neuronal subtypes of the spinal cord, e.g. in MNs, V2a and V3 interneurons, which the analysis correctly identified.

“This TF had previously been reported in multiple neuronal subtypes of the spinal cord, e.g. in MNs, V2a and V3 interneurons (Dasen et al., 2005; Francius et al., 2013), which we correctly identified.”

• cites Dasen et al., 2005 sourceNot checked yet

  Not checked in the free run.

• cites Francius et al., 2013 sourceNot checked yet

  Not checked in the free run.

Clades V3.3 and V3.4 were characterised by the expression of Pou3f1 and Olig3.

“Furthermore, clades V3.3 and V3.4 were characterised by the expression of Pou3f1 and Olig3.”

Clade MN.7 comprised thoracic preganglionic motor column (PGC) neurons.

“whereas clade MN.7 comprised thoracic preganglionic motor column (PGC) neurons.”

Hierarchical clustering independently on each neuronal class resulted in the identification of 59 subpopulations.

“This resulted in the identification of 59 subpopulations (Fig. 4A).”

Median motor column (MMC) neurons and phrenic motor column (PMC) neurons grouped in clade MN.4, which had high levels of Lhx3, Mecom, Pou3f1 and the PMC marker Alcam.

“Median motor column (MMC) neurons and phrenic motor column (PMC) neurons grouped in clade MN.4, which had high levels of Lhx3, Mecom, Pou3f1 and the PMC marker Alcam”

The transcriptome analysis suggested that Nkx6.2 is expressed in a subpopulation of V2a interneurons that does not overlap with Shox2 and Nfib.

“and Nkx6.2 in a subpopulation of V2a interneurons that does not overlap with Shox2 and Nfib.”

The transcriptome analysis suggested that Nr2f1 is expressed in Foxp1-positive MNs.

“Nr2f1 in Foxp1-positive MNs;”

The authors were able to detect all the progenitor populations and major neuronal subtypes described in cervical and thoracic regions of the spinal cord.

“Despite these limitations, we were able to detect all the progenitor populations and major neuronal subtypes that are described in cervical and thoracic regions of the spinal cord.”

Levels of the p3 marker Nkx2.2 were predicted to increase at the onset of neurogenesis.

“Moreover, levels of the p3 marker Nkx2.2 were predicted to increase at the onset of neurogenesis (Fig. S6B).”

The authors documented the transcriptional signatures of 21,465 cells isolated from cervical and thoracic regions of the mouse neural tube during the developmental period in which neuronal subtypes are generated.

“We have documented the transcriptional signatures of 21,465 cells that were isolated from cervical and thoracic regions of the mouse neural tube during the developmental period in which neuronal subtypes are generated.”

The first principal component aligned with neurogenesis, indicated by downregulation of Sox2 and upregulation of Tubb3.

“The first principal component aligned with neurogenesis, which was indicated by the downregulation of Sox2 and upregulation of Tubb3, hence we used PC1 coordinates as a pseudotemporal ordering for neurogenesis.”

Pseudotemporal reconstruction predicted expression changes in a set of 36 regionally patterned genes involved in neurogenesis.

“This predicted the expression changes in a set of 36 regionally patterned genes that are involved in neurogenesis (Fig. 7C).”

Dbx1 levels were markedly elevated in Neurog1-positive p0 progenitors.

“Assays confirmed these predictions: Dbx1 levels were markedly elevated in Neurog1-positive p0 progenitors (Fig. 7E), whereas Nkx2.2 levels were increased immediately adjacent to the p3 progenitor domain in Olig3-positive V3 neurons (Fig. S6E).”

The proportions of cell types recovered matched previously determined proportions in the spinal cord.

“Reassuringly, the proportions of the cell types recovered matched previously determined proportions in the spinal cord (Kicheva et al., 2014), which suggests that the methods did not substantially bias the composition of the transcriptomes recovered.”

• cites Kicheva et al., 2014 sourceNot checked yet

  Not checked in the free run.

Analysis of the single cell transcriptome dataset revealed a set of gene expression modules that is activated synchronously at characteristic developmental timepoints in multiple neuronal classes.

“Analysis of the single cell transcriptome dataset revealed a set of gene expression modules that is activated synchronously at characteristic developmental timepoints in multiple neuronal classes.”

Nkx2.2 levels were increased immediately adjacent to the p3 progenitor domain in Olig3-positive V3 neurons.

“Assays confirmed these predictions: Dbx1 levels were markedly elevated in Neurog1-positive p0 progenitors (Fig. 7E), whereas Nkx2.2 levels were increased immediately adjacent to the p3 progenitor domain in Olig3-positive V3 neurons (Fig. S6E).”

Plotting Olig3 expression dynamics along the differentiation trajectory from dp2 to dI2 neurons or from dp3 to dI3 neurons revealed that maximal expression of Olig3 coincided with expression of Neurog1 and Neurog2.

“Plotting Olig3 expression dynamics along the differentiation trajectory from dp2 to dI2 neurons or from dp3 to dI3 neurons revealed that the maximal expression of Olig3 coincided with the expression of Neurog1 and Neurog2 (Fig. S6C,D).”

Pseudotemporal reconstruction of the p0 to V0 transition predicted that Dbx1 levels increase during neurogenesis.

“Similar reconstruction of the p0 to V0 transition predicted that Dbx1 levels increase during neurogenesis (Fig. 7D).”

Examining gene expression changes as progenitors transitioned to neurons revealed transient upregulation of several TFs, including Atoh1, Neurog1, Neurog2, and several domain-specific TFs.

“Examining gene expression changes as progenitors transitioned to neurons revealed the transient upregulation of several TFs. These included known neurogenic factors, including Atoh1, Neurog1 and Neurog2, and several domain-specific TFs.”

Relatively small subclasses of several neuronal subtypes, such as V3, V2a and dI6, and specific subtypes of MNs, including LMC and MMC MNs, were readily detected.

“Consistent with this, relatively small subclasses of several of the neuronal subtypes, such as V3, V2a and dI6, and specific subtypes of MNs, including LMC and MMC MNs, were readily detected.”

Pseudotemporal ordering of the progenitor-to-neuron transition for multiple progenitor domains predicted the transient upregulation of domain-specific TFs during neurogenesis.

“Pseudotemporal ordering of the progenitor-to-neuron transition for multiple progenitor domains predicted the transient upregulation of domain-specific TFs during neurogenesis (Fig. 7C).”

Only a few dI6 neurons were recovered (95 out of 16450 neurons).

“Because only a few dI6 neurons were recovered (95 out of 16450 neurons), we excluded this population for the following subclustering analysis.”

In conclusion, the authors document the molecular diversity and cellular composition of the developing mouse neural tube.

“In conclusion, we document the molecular diversity and cellular composition of the developing mouse neural tube.”

The data provide an unbiased classification of neural tube cell populations and their associated gene expression profiles.

“The data provide an unbiased classification of neural tube cell populations and their associated gene expression profiles.”

The analysis highlighted a previously underappreciated temporal component to the mechanisms that generate neuronal diversity.

“the analysis highlighted a previously underappreciated temporal component to the mechanisms that generate neuronal diversity”

The analysis revealed common features in the sequence of transcriptional events that lead to differentiation of specific neuronal subtypes.

“revealed common features in the sequence of transcriptional events that lead to the differentiation of specific neuronal subtypes”

The partitioning algorithm correctly assigns identities to the cells in the dataset.

“We therefore conclude that the partitioning algorithm correctly assigns identities to the cells in our dataset.”

Bulk expression profiling studies do not have the resolution or sensitivity to provide detailed insight into gene expression or dynamics in specific cell lineages because of spatial complexity and lack of developmental synchrony.

“The spatial complexity of the neural tube and the lack of developmental synchrony between cells means that bulk studies do not have the resolution or sensitivity to provide detailed insight into gene expression or dynamics in specific cell lineages.”

The single cell transcriptomic atlas captures the domain dynamics of progenitor and neuronal populations.

“Based on these observations, we conclude that the single cell transcriptomic atlas captures the domain dynamics of progenitor and neuronal populations.”

The average number of unique molecular identifiers (UMIs) and detected genes in the retained cells was similar between all samples analysed.

“The average number of unique molecular identifiers (UMIs) and detected genes in these cells was similar between all samples analysed”

The presence of neural crest and mesoderm cells at e9.5/e10.5 was because of the difficulty of cleanly dissecting the neural tube without including any accompanying adjacent tissue at these developmental stages.

“This was because of the difficulty of cleanly dissecting the neural tube without including any accompanying adjacent tissue at these developmental stages.”

The analysis correctly predicted that Pcdh7, Megf11 and Ret are expressed in MNs.

“Consistent with previous work, the analysis correctly predicted that Pcdh7, Megf11 and Ret are expressed in MNs”

• cites Catela et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Hanley et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Lin et al., 2012 sourceNot checked yet

  Not checked in the free run.

The proportions of cells recovered from the assignment of cell identities closely match the growth dynamics of the spinal cord quantified in Kicheva et al. (2014).

“This revealed a close match (Fig. 2E), which allowed us to extend the analysis.”

• cites Kicheva et al., 2014 sourceNot checked yet

  Not checked in the free run.

The analysis correctly predicted expression in the different neuronal populations: cholinergic MNs; inhibitory GABAergic and glycinergic dI4, V1 and V2b interneurons; and excitatory glutamatergic dI1, dI2, dI3, dI5, V2a, and V3 interneurons.

“The analysis correctly predicted expression in the different neuronal populations: cholinergic MNs; inhibitory GABAergic and glycinergic dI4, V1 and V2b interneurons; and excitatory glutamatergic dI1, dI2, dI3, dI5, V2a, and V3 interneurons”

The prominence of motor neurons at early stages is consistent with the initially high differentiation rate of MN progenitors.

“which is consistent with the initially high differentiation rate of MN progenitors”

• cites Ericson et al., 1992 sourceNot checked yet

  Not checked in the free run.

• cites Kicheva et al., 2014 sourceNot checked yet

  Not checked in the free run.

• cites Novitch et al., 2001 sourceNot checked yet

  Not checked in the free run.

Cells from replicate embryos, whether of the same or different sex, tended to intermingle within the embedding, suggesting minimal batch variation between embryos.

“Cells from replicate embryos, whether of the same or different sex, tended to intermingle within the embedding, which suggested minimal batch variation between embryos”

Single cell transcriptome analysis enables the reconstruction of gene expression dynamics along differentiation trajectories.

“Single cell transcriptome analysis enables the reconstruction of gene expression dynamics along differentiation trajectories (Setty et al., 2016; Shin et al., 2015; Trapnell et al., 2014).”

• cites Setty et al., 2016 sourceNot checked yet

  Not checked in the free run.

• cites Shin et al., 2015 sourceNot checked yet

  Not checked in the free run.

• cites Trapnell et al., 2014 sourceNot checked yet

  Not checked in the free run.

The spinal cord atlas implicates novel TFs in specification of neuronal subtypes.

“We therefore conclude that the spinal cord atlas provides sufficient resolution to detect distinct neuronal subtypes and implicates novel TFs in their specification.”

Onecut TFs have been shown to be expressed early in V1 and MNs.

“onecut TFs have been shown to be expressed early in V1 and MNs (Roy et al., 2012; Stam et al., 2012)”

• cites Roy et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Stam et al., 2012 sourceNot checked yet

  Not checked in the free run.

The observations suggest a switch in neuronal subtype identity from Onecut2 to Pou2f2 between e9.5 and e10.5.

“These observations are consistent with the single cell sequencing data and suggest a switch in neuronal subtype identity from Onecut2 to Pou2f2 between e9.5 and e10.5.”

Neuronal subtype diversification in the spinal cord is at least partly driven by transcriptional programmes that are shared between multiple domains and sequentially induced in each of the domains.

“Based on these observations, we conclude that neuronal subtype diversification in the spinal cord is at least partly driven by transcriptional programmes that are shared between multiple domains and sequentially induced in each of the domains.”

• AssumptionRests on an assumption: The developmental stages sampled (e9.5–e13.5) encompass the period in which the neuronal subtypes studied are generated.

The spinal cord atlas provides sufficient resolution to detect distinct neuronal subtypes.

“We therefore conclude that the spinal cord atlas provides sufficient resolution to detect distinct neuronal subtypes and implicates novel TFs in their specification.”

The data correctly predicted the different subtypes of V2a interneurons.

“Thus, the data correctly predicted the different subtypes of V2a interneurons.”

The lateral V2a subgroup diversifies further into subgroups labelled by Shox2 and Maf/onecut family TFs.

“Furthermore, the lateral subgroup appears to diversify further into subgroups that are labelled by Shox2 and Maf/onecut family TFs (Francius et al., 2013; Harris et al., 2018 preprint).”

• cites Francius et al., 2013 sourceNot checked yet

  Not checked in the free run.

• cites Harris et al., 2018 preprint sourceNot checked yet

  Not checked in the free run.

Nfia/b are induced in MNs at later timepoints.

“Nfia/b are induced in MNs at later timepoints (Deneen et al., 2006; Kang et al., 2012).”

• cites Deneen et al., 2006 sourceNot checked yet

  Not checked in the free run.

• cites Kang et al., 2012 sourceNot checked yet

  Not checked in the free run.

There is a temporal component to the specification of neuronal subtype identity that complements the well described spatial axis of patterning.

“This suggests a temporal component to the specification of neuronal subtype identity that complements the well described spatial axis of patterning.”

Only two subpopulations were incorrectly assigned in the hierarchical clustering.

“Of these, initial inspection indicated that only two subpopulations were incorrectly assigned”

The timepoint of Renshaw cell generation correlates with expression of the RA-synthesising enzyme Aldh1a2 in the adjacent somites.

“Retinoic acid (RA) promotes Renshaw cell identity in V1 neurons (Hoang et al., 2018), and the timepoint of their generation correlates with expression of the RA-synthesising enzyme Aldh1a2 in the adjacent somites (Niederreither et al., 1997).”

• cites Niederreither et al., 1997 sourceNot checked yet

  Not checked in the free run.

The data highlighted a previously underappreciated systematic temporal component to the specification of neuronal subtype identity in the neural tube.

“The data highlighted a previously underappreciated systematic temporal component to the specification of neuronal subtype identity in the neural tube.”

• AssumptionRests on an assumption: The developmental stages sampled (e9.5–e13.5) encompass the period in which the neuronal subtypes studied are generated.

The onset of expression of Sox9 coincides with the switch of V1 neurons from Renshaw cells (onecut-expressing) to Foxp2-expressing neurons.

“The onset of expression of Sox9 coincides with the switch of V1 neurons from Renshaw cells (onecut-expressing) to Foxp2-expressing neurons (Kang et al., 2012; Stam et al., 2012; Stolt et al., 2003).”

• cites Kang et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Stam et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Stolt et al., 2003 sourceNot checked yet

  Not checked in the free run.

Olig3 expression is upregulated at the onset of neurogenesis.

“These observations confirm that Olig3 expression is upregulated at the onset of neurogenesis and validate the predictions that were made by the pseudotemporal ordering.”

In the spinal cord, a transcriptional cascade of Sox9 and Nfia/b in progenitors underlies the progressive activation of gliogenesis.

“In the spinal cord, a transcriptional cascade of Sox9 and Nfia/b in progenitors underlies the progressive activation of gliogenesis (Deneen et al., 2006; Kang et al., 2012; Stolt et al., 2003).”

• cites Deneen et al., 2006 sourceNot checked yet

  Not checked in the free run.

• cites Kang et al., 2012 sourceNot checked yet

  Not checked in the free run.

• cites Stolt et al., 2003 sourceNot checked yet

  Not checked in the free run.

The methods did not substantially bias the composition of the transcriptomes recovered.

“Reassuringly, the proportions of the cell types recovered matched previously determined proportions in the spinal cord (Kicheva et al., 2014), which suggests that the methods did not substantially bias the composition of the transcriptomes recovered.”

It is likely that the analysis underestimates the diversity of cell types in the spinal cord.

“Even though some of this diversity may arise at developmental stages later than those analysed in our study, it is likely that our analysis underestimates the diversity of cell types in the spinal cord.”

The dataset will not represent cell types that are unique to lumbar and sacral spinal cord.

“We restricted our analysis to cervical and thoracic regions of the neural tube, and therefore our dataset will not represent cell types that are unique to lumbar and sacral spinal cord.”

There is a large diversity of cell types in the spinal cord.

“There is a large diversity of cell types in the spinal cord. More than 50 distinct pools of MNs have been documented at limb levels of the spinal cord (Dasen et al., 2005; Landmesser, 2001; Philippidou and Dasen, 2013).”

• cites Dasen et al., 2005 sourceNot checked yet

  Not checked in the free run.

• cites Landmesser, 2001 sourceNot checked yet

  Not checked in the free run.

• cites Philippidou and Dasen, 2013 sourceNot checked yet

  Not checked in the free run.

The data form the basis of a molecular atlas of the developing mammalian spinal cord.

“This sheds light on the changing gene expression profiles that characterise neural tube development and forms the basis of a molecular atlas of the developing mammalian spinal cord.”

• AssumptionRests on an assumption: The mouse neural tube is representative of the developing mammalian spinal cord.

Retinoic acid promotes Renshaw cell identity in V1 neurons.

“Retinoic acid (RA) promotes Renshaw cell identity in V1 neurons (Hoang et al., 2018), and the timepoint of their generation correlates with expression of the RA-synthesising enzyme Aldh1a2 in the adjacent somites (Niederreither et al., 1997).”

• cites Hoang et al., 2018 sourceNot checked yet

  Not checked in the free run.

This study makes available detailed information on the transcriptional state of the entire population of neural tube cells.

“This study makes available detailed information on the transcriptional state of the entire population of neural tube cells and will likely accelerate efforts to map comprehensively the transcriptional network.”

The data allow the identification of genes and regulatory modules that define specific cell types.

“The data allow the identification of genes and regulatory modules that define specific cell types.”

The analysis suggests a temporal axis that contributes to the neuronal diversity that accompanies the well-characterised spatial patterning of neural progenitors.

“The analysis suggests a temporal axis that contributes to the neuronal diversity that accompanies the well-characterised spatial patterning of neural progenitors.”

The dynamics of the gene regulatory network result in progenitors undergoing a succession of changes in TF expression during their specification, mediated by repressive interactions.

“The dynamics of the gene regulatory network result in progenitors undergoing a succession of changes in TF expression during their specification, which is mediated by repressive interactions.”