A harmonized atlas of mouse spinal cord cell types and their spatial organization

In the paper

“With this approach, the cells and nuclei from all six studies were spatially interposed in a UMAP visualization of principal component space (Fig. 1c) and separated into groupings that each expressed a panel of well-established cell type markers such as Snap25 (neurons), Mbp (oligodendrocytes), Aqp4 (astrocytes), and Ctss (microglia) (Fig. 1d, e).”

What the sources say

• Partly supported

  The Study Design and Data Analysis section describes using Seurat’s canonical correlation analysis to integrate datasets and identify cell-type clusters. Its Figure 2, however, shows breast epithelial cells from four individuals in a tSNE plot—not cells and nuclei from six studies in a UMAP—and does not report the named brain-cell markers. No page number is available in our copy. (LLM, conf=0.86)

  cites [22] Nguyen, Q. H., Pervolarakis, N., Nee, K. & Kessenbrock, K. Experimental considerations for single-cell RNA sequencing approaches. Front Cell Dev. Biol. 6, 108 (2018). sourceExperimental — accuracy not yet measured
• Partly supported

  The STAR Methods section (page not provided) describes Seurat integration followed by PCA, UMAP visualization, and cell-type annotation, supporting the methodological part of the claim. It does not report the cited document’s result involving cells and nuclei from six studies or the specified Snap25, Mbp, Aqp4, and Ctss marker-defined groupings. (LLM, conf=0.86)

  cites [23] Stuart, T. et al. Comprehensive integration of single-cell data. Cell 177, 1888–1902.e21 (2019). sourceExperimental — accuracy not yet measured

In the paper

“The Prkcg dorsal, excitatory family was comprised of Excit-3 and Excit-4 and likely corresponded to neurons involved in light static touch and allodynic pain in pathological situations46,48–56.”

What the sources say

• Partly supported

  The Results section ('Spatial distribution and molecular identity define distinct spinal microcircuits'; page unavailable) identifies PKCγ-positive dorsal excitatory interneurons, and another Results section links an overlapping CCK-positive population to static paw-withdrawal reflexes. Our copy does not identify a Prkcg family comprising Excit-3 and Excit-4 or establish that those clusters mediate allodynic pain in pathological conditions. (LLM, conf=0.83)

  cites [46] Gatto, G. et al. A functional topographic map for spinal sensorimotor reflexes. Neuron 109, 91–104.e5 (2021). sourceExperimental — accuracy not yet measured
• Partly supported

  The Introduction identifies excitatory interneuron subtypes involved in innocuous touch processing, and the Methods describe analysis of PKCγ-positive neurons for excitatory markers (pages unavailable). Our copy does not identify a Prkcg family comprising Excit-3 and Excit-4 or link those neurons specifically to light static touch and allodynic pain in pathological conditions. (LLM, conf=0.78)

  cites [48] Abraira, V. E. et al. The cellular and synaptic architecture of the mechanosensory dorsal horn. Cell 168, 295–310.e19 (2017). sourceExperimental — accuracy not yet measured
• ?Could not verify

  The Results section on Y1Cre neurons reports a role in light punctate touch, but does not identify a Prkcg family comprising Excit-3 and Excit-4 or link that family to pathological allodynic pain. Searches for Prkcg, PKCγ, Excit-3/Excit-4, and allodynia found only a PKCγ antibody in Methods and an allodynia-related reference, not the attributed finding. No page numbers are available in our copy. (LLM, conf=0.75)

  cites [49] Acton, D. et al. Spinal neuropeptide Y1 receptor-expressing neurons form an essential excitatory pathway for mechanical itch. Cell Rep. 28, 625–639.e6 (2019). sourceExperimental — accuracy not yet measured
• Partly supported

  The Results section (page not available) identifies a dorsal PKCγ-positive subset of excitatory RORα interneurons and reports that ablating RORα interneurons impairs static light touch. It does not identify that subset as Excit-3 or Excit-4, test its specific role in static touch, or link it to allodynic pain in pathological conditions. (LLM, conf=0.85)

  cites [50] Bourane, S. et al. Identification of a spinal circuit for light touch and fine motor control. Cell 160, 503–515 (2015). sourceExperimental — accuracy not yet measured
• Partly supported

  The Results section ('Spinal neurons marked by VT3Cre expression'; page unavailable) places PKCγ-positive interneurons in an area receiving low-threshold mechanoreceptor input and says they have been implicated in neuropathic pain. The Introduction says they might mediate dynamic and punctate mechanical hypersensitivity, but distinguishes punctate from static allodynia. The source does not identify Excit-3 or Excit-4 or establish a role for these neurons in light static touch. (LLM, conf=0.84)

  cites [51] Cheng, L. et al. Identification of spinal circuits involved in touch-evoked dynamic mechanical pain. Nat. Neurosci. 20, 804–814 (2017). sourceExperimental — accuracy not yet measured
• Partly supported

  The Discussion (page unavailable) describes PKCγ-positive excitatory neurons in a spinal circuit linked to mechanical pain, and the Results report that static allodynia was reduced after ablation of SOM-lineage neurons. The paper does not identify a Prkcg family comprising Excit-3 and Excit-4 or establish that those groups mediate light static touch. (LLM, conf=0.82)

  cites [52] Duan, B. et al. Identification of spinal circuits transmitting and gating mechanical pain. Cell 159, 1417–1432 (2014). sourceExperimental — accuracy not yet measured
• Partly supported

  The Background proposes that PKCγ cells transmit low-threshold tactile input and contribute to tactile allodynia in chronic pain states; the Discussion likewise describes a suggested route from tactile inputs to nociceptive neurons. The paper does not identify a “Prkcg dorsal” family comprising Excit-3 and Excit-4, or specifically describe its role in light static touch. Page numbers are unavailable in our copy. (LLM, conf=0.86)

  cites [53] Gutierrez-Mecinas, M., Furuta, T., Watanabe, M. & Todd, A. J. A quantitative study of neurochemically defined excitatory interneuron populations in laminae I-III of the mouse spinal cord. Mol. Pain. 12, 174480691662906 (2016). sourceExperimental — accuracy not yet measured
• ?Could not verify

  Extended Data Figure 2 uses PKCγ only as a marker of a dorsal-horn region; it does not identify a Prkcg excitatory family or Excit-3 and Excit-4. The paper examines static mechanical sensitivity and tactile allodynia (Methods, Behavioral experiments), but its interneuron findings concern CCK+ cells (Extended Data Figure 8). Searches for Prkcg, Excit-3, Excit-4, and PKCγ interneuron function found no passage linking the claimed family to those sensations. (LLM, conf=0.75)

  cites [54] Liu, Y. et al. Touch and tactile neuropathic pain sensitivity are set by corticospinal projections. Nature 561, 547–550 (2018). sourceExperimental — accuracy not yet measured
• Not checked yetcites [55] Malmberg, A. B., Chen, C., Tonegawa, S. & Basbaum, A. I. Preserved acute pain and reduced neuropathic pain in mice lacking PKCgamma. Science 278, 279–283 (1997). source
• Partly supported

  The Discussion section on spinal cord circuits for mechanical hypersensitivity describes PKCγ neurons as part of a proposed pathway that can transform touch into pain after injury, and reports their activation in a neuropathic-pain model. It does not establish a role specifically in static touch. Excit-3 and Excit-4 are the citing authors’ classification, not populations identified in this source. (LLM, conf=0.82)

  cites [56] Peirs, C. et al. Dorsal horn circuits for persistent mechanical pain. Neuron 87, 797–812 (2015). sourceExperimental — accuracy not yet measured

In the paper

“The Maf dorsal, the excitatory family was comprised of Excit-5, Excit-6, and Excit-7 which expressed markers of neurons involved in light touch46,48,50,56,57.”

What the sources say

• Partly supported

  The Results section, “Spatial distribution and molecular identity define distinct spinal microcircuits” (page unavailable), reports cMaf-positive dorsal horn excitatory interneurons and places the overlapping CCK-positive population in the low-threshold mechanoreceptor recipient zone, supporting a connection to light touch. It does not identify a “Maf dorsal” family comprising Excit-5, Excit-6, and Excit-7 or establish marker expression for those three groups. (LLM, conf=0.82)

  cites [46] Gatto, G. et al. A functional topographic map for spinal sensorimotor reflexes. Neuron 109, 91–104.e5 (2021). sourceExperimental — accuracy not yet measured
• ?Could not verify

  The Results section, “A Dorsal Horn Molecular-Genetic Toolbox” (page unavailable), describes excitatory interneuron subtypes in a touch-processing region but does not identify a Maf dorsal family or Excit-5, Excit-6, and Excit-7. Searches for those labels and their proposed light-touch markers found no passage addressing the claimed grouping or marker expression. (LLM, conf=0.74)

  cites [48] Abraira, V. E. et al. The cellular and synaptic architecture of the mechanosensory dorsal horn. Cell 168, 295–310.e19 (2017). sourceExperimental — accuracy not yet measured
• Partly supported

  The Results section (‘RORα identifies a distinct population of excitatory neurons’; page unavailable) identifies MafA+/c-Maf+/RORα+ excitatory interneurons, and the paper links RORα interneurons to light touch. It does not identify a ‘Maf dorsal’ family or classify it as Excit-5, Excit-6, and Excit-7, so that part of the claim is unsupported by this source. (LLM, conf=0.87)

  cites [50] Bourane, S. et al. Identification of a spinal circuit for light touch and fine motor control. Cell 160, 503–515 (2015). sourceExperimental — accuracy not yet measured
• ?Could not verify

  The Results sections on VGLUT3-positive interneurons and mechanical hypersensitivity describe dorsal horn excitatory neurons and responses to low-threshold mechanical stimulation (page unavailable). Searches for “Maf dorsal,” “Excit-5,” “Excit-6,” and “Excit-7” found no passage identifying those clusters or their light-touch markers, so our copy does not establish the attributed classification. (LLM, conf=0.74)

  cites [56] Peirs, C. et al. Dorsal horn circuits for persistent mechanical pain. Neuron 87, 797–812 (2015). sourceExperimental — accuracy not yet measured
• Partly supported

  The Results section on deep dorsal horn neurons (page unavailable) reports Maf expression in a subset of CCK-lineage neurons involved in dynamic mechanical allodynia. It does not identify a “Maf dorsal” excitatory family, assign Excit-5, Excit-6, and Excit-7 to it, or establish that those clusters express markers of light-touch neurons. (LLM, conf=0.76)

  cites [57] Peirs, C. et al. Mechanical allodynia circuitry in the dorsal horn is defined by the nature of the injury. Neuron 109, 73–90.e7 (2021). sourceExperimental — accuracy not yet measured

In the paper

“The mid inhibitory cell types were grouped as one family comprised of Inhib-15, Inhib-16, Inhib-17, Inhib-18, Inhib-19, Inhib-20, and Inhib-21, which expressed the glycinergic marker Slc6a5 (with the exception of Inhib-21) and also the gaba-ergic marker Gad2, implicating these neurons in sensorimotor processing84–86.”

What the sources say

• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [84] Hilde, K. L. et al. Satb2 is required for the development of a spinal exteroceptive microcircuit that modulates limb position. Neuron 91, 763–776 (2016). sourceExperimental — accuracy not yet measured
• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [85] Koch, S. C. et al. RORβ spinal interneurons gate sensory transmission during locomotion to secure a fluid walking gait. Neuron https://doi.org/10.1016/j.neuron.2017.11.011 (2017). sourceExperimental — accuracy not yet measured
• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites [86] Fink, A. J. P. et al. Presynaptic inhibition of spinal sensory feedback ensures smooth movement. Nature 509, 43–48 (2014). sourceExperimental — accuracy not yet measured

In the paper

“We validate a combinatorial marker code for each neuronal cell type and map their spatial distributions in the adult spinal cord.”

In the paper

“Here, we generate an atlas of spinal cell types based on single-cell transcriptomic data, unifying the available datasets into a common reference framework.”

In the paper

“The three studies that FACS sorted neurons from the spinal cord (Hayashi, Haring, and Baek datasets) predominantly gave rise to cells in the neuronal sub-clusters, as well as non-neural cells that most likely represented doublets.”

In the paper

“We now present SeqSeek, a web-based resource for querying this data by gene or cell type and for accessing automated classification algorithm of any spinal cord cell or nucleus from raw sequencing data.”

In the paper

“Co-integration with embryonic cell types allowed us to infer putative lineage relationships for each postnatal cell type and uncovered complex convergent contributions from multiple lineages to many cell types.”

In the paper

“The motoneuron (MN) family includes alpha motoneurons (MNa) which had relatively higher levels of Poln and Spp134, gamma motoneurons (MNg) which had relatively higher levels of Esrrg and Htr1f35, and the related preganglionic cells (PGC) which expressed Gfra3, Nos1, and Fbn236,37.”

What the sources say

• Not checked yetcites [35] Friese, A. et al. Gamma and alpha motor neurons distinguished by expression of transcription factor Err3. Proc. Natl Acad. Sci. USA106, 13588–13593 (2009). source
• Not checked yetcites [36] Schaller, S. et al. Novel combinatorial screening identifies neurotrophic factors for selective classes of motor neurons. Proc. Natl Acad. Sci. USA 114, E2486–E2493 (2017). source
• Not checked yetcites [37] Springell, D. A., Powers-Martin, K., Phillips, J. K., Pilowsky, P. M. & Goodchild, A. K. Phosphorylated extracellular signal-regulated kinase 1/2 immunoreactivity identifies a novel subpopulation of sympathetic preganglionic neurons. Neuroscience 133, 583–590 (2005). source

In the paper

“Dorsal clusters showed very high robustness with this measure, whereas mid and ventral clusters showed low to moderate robustness.”

In the paper

“Moreover, the integration-adjusted gene expression values markedly improved the ability to identify relationships between the clusters of the original studies and also improved the top correlation score for each original cluster (Supplementary Fig. 1c, d).”

In the paper

“This analysis also revealed that putative dorsal clusters were well separated from each other by long dendrogram branches, while putative mid and ventral clusters were much closer to each other in this reduced gene expression space (Fig. 3a).”

In the paper

“This work revealed striking differences between dorsal and ventral neuronal cell types, both in their cell-type relationships and molecular trends.”

In the paper

“Finally, we tested a range of automated classification algorithms and identified a two-tiered model based on label transfer and neural networks as the best method for classifying spinal cord cell types.”

In the paper

“A comparison across studies revealed that these high throughput studies detected fewer genes per cell/nucleus than studies that used single well technical approaches (the Hayashi, Haring, and Baek datasets), and studies that used cells (the Hayashi, Haring, Zeisel, and Baek datasets) detected more genes per cell/nucleus but had higher levels of immediate early gene and stress gene expression than”

In the paper

“As a result, this merged dataset contains more cells and nuclei than were analyzed in the original studies and a uniform set of genes.”

In the paper

“The only study to dissect the spinal cord including the dorsal and ventral spinal roots (the Sathyamurthy dataset) was the only source of Schwann and peripheral glia cells that would be located in these roots.”

In the paper

“This family was only comprised of nuclei from the Sathyamurthy and Rosenberg datasets but we could not detected refined sub-populations of motoneurons. However, this included only 565 MNs.”

In the paper

“Moreover, among the three studies that examined all cell types, the early postnatal Rosenberg study showed enrichment of immature cells of oligodendrocyte lineage relative to the adult Sathyamurthy study, while the adolescent Zeisel study showed an intermediate distribution.”

In the paper

“We observed 20 dorsal excitatory clusters, 14 dorsal inhibitory clusters, 10 deep dorsal/mid excitatory clusters, 7 deep dorsal/mid inhibitory clusters, 8 ventral excitatory clusters, 6 ventral inhibitory clusters, 3 cholinergic motoneuron clusters, and 1 cluster of the cerebrospinal fluid contacting neurons (CSF-cN).”

In the paper

“Unfortunately, the cells or nuclei from each study were segregated from each other almost completely, indicating that the study of origin is a major source of variability in the dataset (Fig. 1b).”

In the paper

“We validated the differentially expressed patterns of several of these genes using in situ hybridization and also at the protein level for CAMK2α and WFA-lectin to reveal perineuronal nets, thereby confirming predicted gene expression signatures that would differentially regulate plasticity in the dorsal and ventral horns of the spinal cord (Fig. 3c–f).”

In the paper

“When either all genes or the top 500 highly variable genes were analyzed, there were weak overall correlations and very few alignments between clusters from different studies (Supplementary Fig. 1a, b, d).”

In the paper

“Excit-17 included almost exclusively nuclei from the Rosenberg dataset and showed enriched expression of the immature neuron marker Dcx.”

In the paper

“Excit-5 also expressed Pvalb and Cck, Excit-6 expressed Rorb and Cpne4, and Excit-7 was distinguished by having only nuclei from the Rosenberg dataset and expressed the immature neuron marker Dcx, suggesting an immature phenotype.”

In the paper

“These clusters expressed enriched levels of Car12 (in particular in Excit-9 and Excit-10), the neuropeptide receptors Trhr (Excit-8), Npr1 (Excit-9 and Excit-10), and Nmur2 (Excit-10), and the neuropeptide Grp (Excit-9).”

In the paper

“Inhib-6 and Inhib-7 expressed Cdh3 and were distinguished by co-expression of Kcnip2 and Pvalb in Inhib-7.”

In the paper

“Prkcg is a classic marker gene in the spinal cord and defined this family together with the neuropeptides Cck and Trh (Excit-3) and Nts (Excit-4).”

In the paper

“While Inhib-8 contained only low levels of Cdh3 in this analysis, Cdh3 expression was confirmed by in situ hybridization and this cluster was included in this family based on proximity in principal component space as reflected in the UMAP and dendrogram analysis. Inhib-8 also expressed Klhl14.”

In the paper

“The Cpne4 dorsal, excitatory family was comprised of Excit-1 and Excit-2. Excit-1 was a rare subset, both in the harmonized clusters and in the in situ counts, that also expressed Dach2. Excit-2 was more prevalent and co-expressed Prkcg as well as Cbln2.”

In the paper

“The Megf11 cluster (Excit-20) displayed features of dorsal excitatory neurons and mid excitatory neurons, being located in lamina 4/5 and being grouped with mid neurons in principal component space in the UMAP and dendrogram analysis.”

In the paper

“Of note, there were two discrepancies in this family between the sequencing data and the in situ hybridization data: Cck was present at high levels in the Excit-3 in the sequencing data but we did not detect Cck in most Prkcg-expressing cells of the adult spinal cord and Prkcg was not enriched in Excit-4 in the sequencing data but was readily detected with Nts and Calb1 in this cluster in adult t”

In the paper

“In addition to Rorb, Inhib-1 expressed Sorcs3, Inhib-3 expressed Nppc as well as Nrgn, Inhib-4 expressed Rxfp2, and Inhib-5 did not express these other genes. Inhib-2 expressed Sorcs3 and Adamts5.”

In the paper

“Both subsets also expressed Calb1, although it was not specific to these clusters.”

In the paper

“This cluster was very distinct from other neuronal populations, inhibitory, and also expressed the early neuron marker Sox2 and the V2b lineage markers Gata2 and Gata3, suggesting an immature phenotype.”

In the paper

“The Rreb1 dorsal, excitatory family was comprised of Excit-12 and Excit-13. These clusters also expressed Satb1 and either Zim1 (Excit-12) or Nmur2 and Crh (Excit-13).”

In the paper

“PGCs were clustered into 23 subtypes that varied by spinal cord level and MNa were clustered into 14 subtypes that also varied by spinal level.”

In the paper

“Within this family, Excit-14 and Excit-15 were slightly separated and also similar to the Rreb1 family clusters and expressed Col5a2 (Excit-14) or Col5a2 and Enpp1 (Excit-15).”

In the paper

“Excit-16, Excit-18, and Excit-19 expressed the neuropeptides Tac1 (Excit-16), Nmu-hi/Tac2-lo (Excit-18), and Tac2hi/Nmu-lo (Excit-19).”

In the paper

“Each of these clusters expressed Pdyn, while Inhib-10 also expressed Gal and Mlxipl and Inhib-11 also expressed Gal only. Of note, the clusters in this family also expressed Rorb and Nrgn.”

In the paper

“Dual contributions of dI5.5 and dI2.1 to Excit-20 and of V1 and V2b to Inhib-27 were also observed.”

In the paper

“The model showed very high confidence scores for correct predictions; however, performance varied with cell type prevalence suggesting a target for improving the model in the future (Supplementary Fig. 11).”

In the paper

“Excit-35 also expressed Vamp1, Pou3f1, Shox2, and Pou6f2 and Excit-36 expressed Esrrg.”

In the paper

“Of note, nearly all of the cells and nuclei in this family were from the Rosenberg and Sathyamurthy datasets and Excit-26 in particular was predominantly from the Rosenberg dataset.”

In the paper

“A neural network model with one hidden layer (256 nodes) and an SGD optimizer showed the best performance (overall accuracy of 80%) and was selected for further work.”

In the paper

“For example, Pou6f2 was expressed in the deep dorsal horn and in the dorsal part of the ventral horn and was enriched in mid-excitatory (Excit-21, Excit-28, and Excit 30), ventral excitatory (Excit-34 and Excit-35), and ventral inhibitory (Inhib-24) clusters that are located within this domain.”

In the paper

“The Chat dorsal inhibitor cluster Inhib-14 was a deep dorsal (lamina 4), inhibitory and cholinergic population and also expressed Nos180–82.”

What the sources say

• Not checked yetcites [80] Laing, I., Todd, A. J., Heizmann, C. W. & Schmidt, H. H. Subpopulations of GABAergic neurons in laminae I-III of rat spinal dorsal horn defined by coexistence with classical transmitters, peptides, nitric oxide synthase or parvalbumin. Neuroscience 61, 123–132 (1994). source
• Not checked yetcites [81] Kluchová, D. et al. Partial colocalization of NADPH-diaphorase and acetylcholinesterase positivity in spinal cord neurons. Physiol. Res. 49, 151–155 (2000).
• Not checked yetcites [82] Mesnage, B. et al. Morphological and functional characterization of cholinergic interneurons in the dorsal horn of the mouse spinal cord. J. Comp. Neurol. 519, 3139–3158 (2011). source

In the paper

“Of note, nearly all of the cells and nuclei in this family were from the Rosenberg and Sathyamurthy datasets.”

In the paper

“Similarly, Nfib was expressed in the medial deep dorsal horn (mid) spinal cord and was enriched in both excitatory (Excit-23, Excit-25, and Excit-30) and inhibitory (Inhib-15 and Inhib-21) clusters.”

In the paper

“These clusters expressed Npy and were distinguished by low levels of Vgf (Inhib-12) or by expression of Qrfpr (Inhib-13).”

In the paper

“Other markers sub-divided the clusters in a combinatorial manner, including Nms (Excit-21), Bcl11a (Excit-22 through Excit-25), Satb1 and Cdh23 (Excit-23, Excit-24, and Excit-25), Cep112 (Excit-23 and Excit-24), and Prox1 (Excit-25).”

In the paper

“All three strategies performed well, with label transfer showing the best performance (overall accuracy of 89%), followed by the neural network (83%), and then the SVM (80%) (Fig. 6a and Supplementary Table 6).”

In the paper

“We found that long trajectories of embryonic cells led towards particular cluster families, which allowed us to infer developmental relationships (Fig. 5b).”

In the paper

“The molecular identities of the cardinal classes were confirmed by classic marker analysis (Fig. 5c) and these marker genes often extended into the postnatal cell type domains as well.”

In the paper

“Inhib-15 expressed Prox1, Gabra1, and Nfib, Inhib-16 expressed Gpc3 and Sema5b, Inhib-17 expressed Satb2, Inhib-18 expressed Sema5b, Inhib-19 expressed Ccbe1 and Pou6f2, Inhib-20 expressed higher levels of Tfap2b as well as Zfhx3, and Inhib-21 expressed Nfib and was distinguished by having only Gad2 and not Slc6a5 and was mainly derived from the Rosenberg dataset.”

In the paper

“Excit-26 expressed Nfib, Excit-27 expressed Adamts2, Excit-28 expressed Chat and Pitx2 and thus likely corresponds to V0c neurons, and Excit-29 expressed Pmfbp1.”

In the paper

“The Clarke’s column marker gene, Gdnf, was expressed at very low levels in the merged dataset, but was present in several Excit-30 cells.”

In the paper

“This analysis also supported a division of deep dorsal horn excitatory neurons into two overall groups (with the ME/Lmx1b family being more closely related to dI5, while the other ME clusters are more closely related to dI1-dI3) and a division of ventral excitatory neurons into two groups (with Excit-31 through Excit-34 being closer to dI1-3 and Excit-35 through Excit-37 being closer to V0, V2a, a”

In the paper

“We found that the model performance varied with the co-clustering frequency of each cluster and with the ability to identify cell types in situ and we propose that these measures can be used to set a reasonable expectation for neural network performance.”

In the paper

“Excit-37 expressed low levels of the V3 marker gene Sim1 as well as Rnf220.”

In the paper

“These clusters could be distinguished by Gm26673, Syt2, and Prlr (Excit-31), Mdga1 and Prlr (Excit-32), and Bnc2 and Pou6f2 (Excit-34).”

In the paper

“We found that a linear model (with no regularization and with an SGD optimizer) showed the best performance, with an overall test accuracy of 85% (Fig. 6b and Supplementary Table 6).”

In the paper

“In contrast, ventral clusters are much more similar to each other, with close or overlapping distributions in principal component space and overlapping gene expression patterns.”

In the paper

“Dorsal clusters are distinct from each other with clearly separated individual cell types that can be grouped loosely into families.”

In the paper

“For example, three very different methods – single-cell data clustering, multiplexed in situ hybridization, and an artificial intelligence neural network – all showed a relatively weak ability to classify ventral cell types into discrete types and a relatively strong but still imperfect ability to classify dorsal cell types.”

In the paper

“Differential gene expression and gene ontology analysis revealed that learning-related genes are enriched in the dorsal horn while structural stability-related genes are enriched in the ventral horn.”

In the paper

“The neural network model was at least as accurate as other methods such as Seurat-based clustering and high-content in situ hybridization and was orders of magnitude faster.”

In the paper

“Of note, several cell types were not expected to be present in the independent dataset, including Schwann cells, peripheral glia, and meninges 2 (based on the surgical dissection method used that did not include spinal roots or outer layers of meninges) and including PGC, Excitatory-7, and Excitatory-17 (based on the lumbar region and adult age that was used). As expected, these cell types were”

In the paper

“In particular, several mid/ventral cell types were not detected in the independent dataset while two ventral clusters (Excitatory-31 and Inhibitory-27) were over-represented (Fig. 6 and Supplementary Fig. 13).”

In the paper

“This work also led to the surprising discovery of gene expression signatures for plasticity in the dorsal horn and for structural stability in the ventral horn.”

In the paper

“For example, the deep dorsal horn excitatory neurons can be divided into two families, with Excit-21 through Excit-25 in one family that is likely derived from dI5 precursors and Excit-26 through Excit-29 in a looser family that includes dI1/2 and V0 precursors.”

In the paper

“Within the ventral horn, this included components of perineuronal nets (which are thought to restrict plasticity and had been previously detected in the ventral spinal cord110,111) such as Tnr (Tenascin-R), as well as adhesion molecules such as Sdk1 and intracellular signaling components such as Ptpn5 (STEP) which de-phosphorylates CAMK2α, NMDA receptors, and ERK kinases33.”

What the sources say

• Not checked yetcites [110] Sorg, B. A. et al. Casting a wide net: role of perineuronal nets in neural plasticity. J. Neurosci. 36, 11459–11468 (2016). source
• Not checked yetcites [111] Galtrey, C. M., Kwok, J. C. F., Carulli, D., Rhodes, K. E. & Fawcett, J. W. Distribution and synthesis of extracellular matrix proteoglycans, hyaluronan, link proteins and tenascin-R in the rat spinal cord. Eur. J. Neurosci. 27, 1373–1390 (2008). source
• Not checked yetcites [33] Won, S. & Roche, K. W. Regulation of glutamate receptors by striatal-enriched tyrosine phosphatase 61 (STEP61). J. Physiol. 599, 443–451 (2021). source

In the paper

“Building on these preliminary findings, the scope of the harmonized analysis here afforded a much deeper characterization of cell type relationships and robustness, the validation of these molecular distinctions in tissue, and through machine learning.”

In the paper

“Adding regularization helped to lower the gap between the training and validation accuracy”

In the paper

“Whereas the training was faster, validation accuracy improved by around 5% when we lowered the batch size to 32.”

In the paper

“No additional improvement was seen at a batch size of 16, so the batch size was set to 32 for the rest of the study.”

In the paper

“GO analysis revealed that all of the enriched gene annotation clusters were associated with basic cell metabolism terms such as the ribosome, metabolic pathways, and proton transport (Supplementary Table 6).”

In the paper

“This pipeline achieved an overall accuracy of 80% on the validation data.”

In the paper

“Indeed, it is not clear whether the cell types from the original studies are comparable in their current forms, resulting in a fragmented set of incomplete and conflicting atlases for the spinal cord.”

In the paper

“This general feature was consistent with previous observations and likely reflects similar patterns of gene expression amongst mid and ventral clusters1,4.”

What the sources say

• Not checked yetcites [1] Sathyamurthy, A. et al. Massively parallel single nucleus transcriptional profiling defines spinal cord neurons and their activity during behavior. Cell Rep. 22, 2216–2225 (2018). source
• Not checked yetcites [4] Rosenberg, A. B. et al. Single-cell profiling of the developing mouse brain and spinal cord with split-pool barcoding. Science 360, 176–182 (2018). source

In the paper

“To best compare the data from these studies, we began with the raw sequencing reads from each study and performed our own data processing with uniform methods and filters.”

In the paper

“Together, these analyses demonstrated that integration of publicly available datasets can be used to harmonize spinal cord sequencing data, preserve important biological differences between studies, and uncover a robust set of shared cell types.”

In the paper

“Thus, we discovered general differences in the relationships between clusters in the dorsal versus the ventral horn and molecular trends that could confer differential plasticity control in these two regions.”

In the paper

“The Npy dorsal, the inhibitory family was comprised of Inhib-12 and Inhib-13. Studies suggest this family’s markers identify neurons involved in mechanical itch and pain73,78,79.”

What the sources say

• Not checked yetcites [73] Boyle, K. A. et al. A quantitative study of neurochemically defined populations of inhibitory interneurons in the superficial dorsal horn of the mouse spinal cord. Neuroscience 363, 120–133 (2017). source
• Not checked yetcites [78] Bourane, S. et al. Gate control of mechanical itch by a subpopulation of spinal cord interneurons. Science 350, 550–554 (2015). source
• Not checked yetcites [79] Tashima, R. et al. A subset of spinal dorsal horn interneurons crucial for gating touch-evoked pain-like behavior. Proc. Natl. Acad. Sci. USA 118, e2021220118 (2021). source

In the paper

“The Reln dorsal, excitatory family was comprised of Excit-8, Excit-9, Excit-10, and Excit-11 and expressed markers indicative of a role in chemical itch sensation58–62.”

What the sources say

• Not checked yetcites [58] Bell, A. M. et al. Expression of green fluorescent protein defines a specific population of lamina II excitatory interneurons in the GRP::eGFP mouse. Sci. Rep. 10, 13176–14 (2020). source
• Not checked yetcites [59] Christensen, A. J. et al. In vivo interrogation of spinal mechanosensory circuits. Cell Rep. 17, 1699–1710 (2016). source
• Not checked yetcites [60] Fatima, M. et al. Spinal somatostatin-positive interneurons transmit chemical itch. Pain 160, 1166–1174 (2019). source
• Not checked yetcites [61] Huang, J. et al. Circuit dissection of the role of somatostatin in itch and pain. Nat. Neurosci. 21, 707–716 (2018). source
• Not checked yetcites [62] Mishra, S. K. & Hoon, M. A. The cells and circuitry for itch responses in mice. Science 340, 968–971 (2013). source

In the paper

“The Pdyn dorsal, inhibitory family was comprised of Inhib-9, Inhib-10, and Inhib-11 and expressed markers suggestive of a role in chemical itch52,61,73–77.”

What the sources say

• Not checked yetcites [52] Duan, B. et al. Identification of spinal circuits transmitting and gating mechanical pain. Cell 159, 1417–1432 (2014). source
• Not checked yetcites [61] Huang, J. et al. Circuit dissection of the role of somatostatin in itch and pain. Nat. Neurosci. 21, 707–716 (2018). source
• Not checked yetcites [73] Boyle, K. A. et al. A quantitative study of neurochemically defined populations of inhibitory interneurons in the superficial dorsal horn of the mouse spinal cord. Neuroscience 363, 120–133 (2017). source
• Not checked yetcites [74] Brewer, C. L., Li, J., O’Conor, K., Serafin, E. K. & Baccei, M. L. Neonatal injury evokes persistent deficits in dynorphin inhibitory circuits within the adult mouse superficial dorsal horn. J. Neurosci. 40, 3882–3895 (2020). source
• Not checked yetcites [75] Brewer, C. L., Styczynski, L. M., Serafin, E. K. & Baccei, M. L. Postnatal maturation of spinal dynorphin circuits and their role in somatosensation. Pain 161, 1906–1924 (2020). source
• Not checked yetcites [76] Kardon, A. P. et al. Dynorphin acts as a neuromodulator to inhibit itch in the dorsal horn of the spinal cord. Neuron 82, 573–586 (2014). source
• Not checked yetcites [77] Serafin, E. K. et al. Transcriptional profile of spinal dynorphin-lineage interneurons in the developing mouse. Pain 160, 2380–2397 (2019). source

In the paper

“The Rorb and Adamts5 dorsal, the inhibitory family was comprised of Inhib-1, Inhib-2, Inhib-3, Inhib-4, and Inhib-5, with markers of neurons involved in the dampening of dynamic touch48,68,69.”

What the sources say

• Not checked yetcites [48] Abraira, V. E. et al. The cellular and synaptic architecture of the mechanosensory dorsal horn. Cell 168, 295–310.e19 (2017). source
• Not checked yetcites [68] Cui, L. et al. Identification of early RET+ deep dorsal spinal cord interneurons in gating pain. Neuron 91, 1137–1153 (2016). source
• Not checked yetcites [69] Gutierrez-Mecinas, M., Polgár, E., Bell, A. M., Herau, M. & Todd, A. J. Substance P-expressing excitatory interneurons in the mouse superficial dorsal horn provide a propriospinal input to the lateral spinal nucleus. Brain Struct. Funct. 223, 2377–2392 (2018). source

In the paper

“The Cdh3 dorsal, inhibitory family was comprised of Inhib-6, Inhib-7, and Inhib-8 and are likely to be involved in the dampening of dynamic touch and therefore in mechanical allodynia70–72.”

What the sources say

• Not checked yetcites [70] Boyle, K. A. et al. Defining a spinal microcircuit that gates myelinated afferent input: implications for tactile allodynia. Cell Rep. 28, 526–540.e6 (2019). source
• Not checked yetcites [71] Hughes, D. I. et al. Morphological, neurochemical and electrophysiological features of parvalbumin-expressing cells: a likely source of axo-axonic inputs in the mouse spinal dorsal horn. J. Physiol. 590, 3927–3951 (2012). source
• Not checked yetcites [72] Petitjean, H. et al. Dorsal horn parvalbumin neurons are gate-keepers of touch-evoked pain after nerve injury. Cell Rep. 13, 1246–1257 (2015). source

In the paper

“The Sox5 dorsal, the excitatory family was comprised of Excit-14, Excit-15, Excit-16, Excit-17, Excit-18, and Excit-19, and expressed markers suggestive of a role in coping pain and mechanical nociception46,52,56,63–67.”

What the sources say

• Not checked yetcites [46] Gatto, G. et al. A functional topographic map for spinal sensorimotor reflexes. Neuron 109, 91–104.e5 (2021). source
• Not checked yetcites [52] Duan, B. et al. Identification of spinal circuits transmitting and gating mechanical pain. Cell 159, 1417–1432 (2014). source
• Not checked yetcites [56] Peirs, C. et al. Dorsal horn circuits for persistent mechanical pain. Neuron 87, 797–812 (2015). source
• Not checked yetcites [63] Huang, T. et al. Identifying the pathways required for coping behaviours associated with sustained pain. Nature 565, 86–90 (2019). source
• Not checked yetcites [64] Moriyama, M. et al. Presynaptic modulation by neuromedin U of sensory synaptic transmission in rat spinal dorsal horn neurones. J. Physiol. 559, 707–713 (2004). source
• Not checked yetcites [65] Petitjean, H. et al. Recruitment of spinoparabrachial neurons by dorsal horn calretinin neurons. Cell Rep. 28, 1429–1438.e4 (2019). source
• Not checked yetcites [66] Polgár, E. et al. Substance P-expressing neurons in the superficial dorsal horn of the mouse spinal cord: insights into their functions and their roles in synaptic circuits. Neuroscience 450, 113–125 (2020). source
• Not checked yetcites [67] Smith, K. M. et al. Calretinin positive neurons form an excitatory amplifier network in the spinal cord dorsal horn. Elife 8, 1750 (2019). source

In the paper

“All of the clusters except Excit-25 expressed Tacr1 and Excit-21 also expressed Lypd1, suggesting that these are candidate ascending populations3.”

What the sources say

• Not checked yetcites [3] Häring, M. et al. Neuronal atlas of the dorsal horn defines its architecture and links sensory input to transcriptional cell types. Nat. Neurosci. 21, 869–880 (2018). source

In the paper

“Together, the joint analysis of spinal cord neurons from embryonic through adult stages began to align these two perspectives on cell types while also highlighting the complex relationships that exist between genetically defined cellular origins and mature transcriptional signatures.”

In the paper

“Excit-33 was comprised almost entirely of nuclei from the Rosenberg dataset and may represent an immature cell type.”

In the paper

“Excit-35, Excit-36, and Excit-37 are likely derived from the V2a lineage, as they expressed Vsx2 (Chx10) and included many cells from the Hayashi dataset that sorted cells based on Chx10 genetic expression and thus may play a role in skilled reaching87.”

What the sources say

• Not checked yetcites [87] Azim, E., Jiang, J., Alstermark, B. & Jessell, T. M. Skilled reaching relies on a V2a propriospinal internal copy circuit. Nature 508, 357–363 (2014). source

In the paper

“Of note, several cluster markers of ventral cell types, such as Sim1, were not observed in adult spinal cord tissue by in situ hybridization and while they are detected in the harmonized sequencing data, they likely represent lingering RNA from developmental samples.”

In the paper

“These clusters expressed Lmx1b, suggesting a dI5/dILB embryonic origin.”

In the paper

“Excit-28 and Excit-29 also express Onecut2 and Pou6f2, potentially revealing a link with ventral cell types.”

In the paper

“In general, we found that the ventral clusters had less distinct gene expression patterns and were less robust than dorsal and mid clusters; therefore, the final identities of these clusters should be considered with caution.”

In the paper

“The remaining ME family was comprised of mid, excitatory clusters were comprised of Excit-26, Excit-27, Excit-28, and Excit-29. These clusters do not express Lmx1b, in contrast to the other mid excitatory family and are likely derived from dI1-3 or ventral embryonic lineages.”

In the paper

“As the in situ hybridization experiments were performed on mid/lower lumbar spinal cord sections, we did not validate markers for this cluster.”

In the paper

“Inhib-22 and Inhib-27 also expressed the gaba-ergic marker Gad2 as well as Pax2 and Pou6f2. They were distinguished by low levels of Gata3 expression in Inhib-27, which may represent V2b lineage.”

In the paper

“They were identified by expression of Gm26673 and Pvalb in Inhib-23, which may suggest that this cluster included Ia-inhibitory neurons89.”

What the sources say

• Not checked yetcites [89] Alvarez, F. J. et al. Postnatal phenotype and localization of spinal cord V1 derived interneurons. J. Comp. Neurol. 493, 177–192 (2005). source

In the paper

“These results must be interpreted with caution because each of these datasets were included in the overall training data which could lead to artificially high performance.”