Circuit organization and transcriptomic heterogeneity of sympathetic circuits innervating cranial structures

Gokhan Senturk; Shawn Driscoll; Peter J Osseward, II; Miriam Gullo; Karen Lettieri; Samuel L Pfaff

doi:10.1016/j.celrep.2026.117446

. Author manuscript; available in PMC: 2026 Jul 16.

Published in final edited form as: Cell Rep. 2026 Jun 3;45(6):117446. doi: 10.1016/j.celrep.2026.117446

Circuit organization and transcriptomic heterogeneity of sympathetic circuits innervating cranial structures

Gokhan Senturk ¹, Shawn Driscoll ¹, Peter J Osseward II ¹, Miriam Gullo ¹, Karen Lettieri ¹, Samuel L Pfaff ^1,^2,^*

PMCID: PMC13372007 NIHMSID: NIHMS2190723 PMID: 42241285

SUMMARY

The superior cervical ganglion (SCG) innervates multiple effector organs within cranial tissues and elicits responses including pupil dilation, piloerection, vasoconstriction, and inhibition of salivation. Coordinated activation of these targets is associated with the display of different emotions; however, the underlying circuit organization and cellular heterogeneity of SCG neurons remain unclear. Here, we combined neuronal tracing with single-cell and spatial transcriptomics to characterize the SCG circuitry and heterogeneity. We found that each SCG neuron innervates a single effector organ. SCG subtypes defined by their projection (P) targets form two major compartments within the ganglia, but individual P-types were intermingled. Mature SCG transcriptomic (T) types emerge postnatally and exhibit rostra-caudal biases. While some T-types were enriched in SCG populations with specific axon projections, they did not show a strict one-to-one correspondence with P-types. These results suggest that individual sympathetic cranial effectors mediating facial emotions are controlled combinatorially by multiple transcriptomic SCG types.

Graphical Abstract

graphic file with name nihms-2190723-f0001.jpg

In brief

Senturk et al. map the molecular and anatomical organization of superior cervical ganglion neurons targeting cranial structures mediating facial emotions. Individual neurons project to a single target, yet each effector is innervated by multiple molecular subtypes. This organization suggests sympathetic responses are mediated via combinatorial engagement of distinct molecular subtypes.

INTRODUCTION

The sympathetic autonomic nervous system is a key regulator of internal physiological homeostasis, helping animals react to their environments.^1,2 Although the sympathetic system is often conceptualized as a unitary system for fight-or-flight,³ it is being increasingly appreciated that it controls diverse structures in complex patterns that often decouple the activity of individual effector organs.^1,4-10 This is particularly evident in the study of emotions, where autonomic activity differentiates between emotions expressed through voluntary facial expressions (Figure 1A).¹³ Here, we examined the circuitry of the superior cervical ganglion (SCG), which contains sympathetic neurons that innervate distinct types of effector organs within the cranial tissues controlling pupil dilation, piloerection, vasoconstriction, and salivation, among others (Figure 1B).

Figure 1. — (A) Decision tree demonstrating that heart rate and skin temperature distinguishe among emotions (adapted from Ekman et al.¹³).

(B) Sympathetic innervation of cranial tissues via superior cervical ganglion (SCG).

(C) Stimulation of spinal ventral roots innervating SCG elicit differential responses in the sympathetic cranial tissues. Each row corresponds to a separate animal. Number of + indicates the strength of the output with 0 as no response (adapted from Njå and Purves,¹¹ but also see Langley¹²).

(D) Two extreme hypothetical models of SCG circuit architecture for effector organ innervation.

(E–G) CTB injections into pairs of effector organs. Bar plots show the proportions of neurons with the indicated projection patterns below. Dashed lines separate data for each effector organ. Each dot represents the colocalization rate for a single animal.

(H) Triple injection of CTB into different portions of the skin. The bar plot shows the proportions of neurons exhibiting each projection pattern indicated below.

All scale bars represent 100 μm. See also Figure S1.

The sympathetic system has parallels with the somatic motor system controlling muscles, in that both comprise efferent pathways with axons that navigate through peripheral tissues to innervate their targets. While this suggests that there may be parallels in the circuit organization of both efferent systems, there are critical differences. Namely, the spinal neurons controlling sympathetic responses use peripheral ganglionic relays to control their effector organs whereas somatic motor neurons directly innervate muscle targets. Cholinergic preganglionic neurons within the thoracic spinal cord (pre-ganglionic column [PGC]), which have a close developmental relationship with somatic motor neurons,^14,15 project to targets such as the SCG within the neck, which contains neural crest-derived norepinephrine neurons that contact sympathetic effector organs (Figure 1B).¹⁶ Notably, stimulation of spinal ventral roots projecting to the SCG evokes differential responses among its effector organs (Figure 1C),^11,12 implying that this efferent pathway has the necessary hardware to elicit complex patterns of sympathetic responses. However, the circuit architecture and cellular heterogeneity enabling this remain to be characterized.

Despite their common neural crest origin, sympathetic neurons exhibit molecular and morphological heterogeneity, characterized by selective neuropeptide expression and variable soma sizes.^4,16-21 In particular, NPY⁺ and ACh⁺ neurons have historically been associated with vasoconstrictor²¹ and sudomotor¹⁷ functions, leading to the “neurochemical coding hypothesis” , which posits that sympathetic effectors are innervated by neurochemically defined populations.^4,6-8 Intriguingly, recent work identified transcriptomic-types that preferentially innervate specific effector organs in various sympathetic ganglia.^22,23 Whether such associations extend to the SCG^24,25 and generalize to a true one-to-one correspondence between transcriptomic-types and effector organ innervation remains to be determined.

Here, we used circuit tracing with single cell and spatial transcriptomics to investigate the circuit organization and cellular heterogeneity of SCG neurons. While we considered the possibility that the coordination of sympathetic responses involving multiple effector organs within the head could be mediated by SCG neurons that branch to multiple targets, we found that individual SCG neurons had one-to-one connectivity relationships with their targets. This connectivity specificity allowed us to define SCG neuron subtypes based on their projection status (P-types). We found that P-types form two compartments within the SCG, defined largely by their axon exit points from the ganglia. However, P-types within each compartment were not topographically organized. We identified eight transcriptionally-distinct subtypes of SCG neurons (T-types). To identify the relationship between T- and P-types, we combined neuronal tracing and molecular labeling and found that many P-types are molecularly heterogeneous, i.e., individual P-types comprise multiple T-types. Thus, the control of facial sympathetic responses is likely mediated by combinatorial patterns of activity among T-types of SCG neurons.

RESULTS

SCG neuron projections

To investigate the peripheral circuit organization enabling the differential recruitment of cranial sympathetic effector organs (Figures 1B and 1C), we examined SCG neurons and their projection patterns. The layered architecture of this system enables the formulation of two extreme models of circuit organization (Figure 1D). To map projections of individual SCG neurons, we injected cholera toxin subunit B (CTB) retrograde tracers into pairs of effector organs and quantified SCG neurons as single- or double-labeled. In control experiments, co-injection of two separate CTBs into the same effector organ yielded >85% neuronal double-labeling, indicating high efficiency of targeting. (Figures S1A-S1E). Most cranial targets could be selectively labeled with CTB; however, skin injections consistently leaked into underlying tissues (Figure S1F). Therefore, we excluded the skin from pairing with other effector organs for double labeling. Importantly, we did not detect any double-labeled SCG neurons that branch to innervate multiple targets including the eye, tongue, and/or salivary gland (Figures 1E-1G). Notably, this organization is observed despite blood vessels being common to all effector organs, indicating that vasculature-innervating SCG neurons do not branch to innervate blood vessels in multiple organs. Similarly, injection of three different CTB tracers into different portions of the skin (forehead, cheek, and chin) yielded mostly single-labeled neurons (double-labeling ≤7%; Figure 1H), indicating that each portion of the skin is also innervated by distinct SCG populations.

Together, these findings show that the SCG is primarily organized with neurons that have simple one-to-one connectivity to individual effector organs, including different portions of the skin (see model II, Figure 1D). Thus, SCG P-types are analogous to spinal somatic motor pools, which likewise innervate single muscle targets.

The spatial organization of SCG P-types

To investigate the spatial organization of P-types and their relationship to each other, we used double labeling with CTB in whole mount preparations. We also visualized the principal nerve branches of the SCG to assess the routes through which axons exit the ganglion. The SCG has superior (sup.) and inferior (inf.) output branches and a single input branch from the sympathetic trunk (ST) (Figure 2B).²⁶ The ST branch is formed by thoracic preganglionic axons that control SCG neuron activity.

We first injected CTB into the eye and salivary gland, which are located dorsally and ventrally, respectively, in the head (Figure 2A). SCG neurons projecting to these targets occupied distinct anterior (eye) and posterior (salivary gland) compartments of the ganglion, with partial overlap at the boundary (Figures 2B and 2C). Their axons exited through different branches: the eye via the anterior sup. branch and the salivary gland via the posterior inf. branch (Figures 2D and 2E). Although these findings support a topographic organization, CTB injections into the tongue and salivary gland, both ventrally located effectors (Figure 2F), provided a counterexample. Despite the tongue’s position between the eye and salivary gland along the dorsal-ventral axis, tongue-projecting neurons were not arranged in an intermediate band. Instead, both tongue- and salivary gland-projecting neurons localized to the posterior compartment in an intermingled, salt-and-pepper pattern (Figures 2G and 2H), and their axons exited exclusively via the posterior inf. branch (Figures 2I and 2J).

These results indicate that SCG organization reflects both topographic and salt-and-pepper features. Rather than smooth transitions, the ganglion is divided into two compartments (Figure 2O): an anterior compartment projecting through the sup. branch to dorsal cranial targets and a posterior compartment projecting through the inf. branch to ventral targets. Within each compartment, multiple P-types are intermingled rather than spatially segregated, representing a “pseudo-topography” distinct from the fine-grained spatial maps typical of somatic motor neurons. These results are also consistent with previous nerve branch recordings of the SCG in reduced preparations.²⁶

CTB injections into distinct skin regions along the dorsal-ventral axis of the head revealed a similar pattern (Figure 2K): Neurons innervating the forehead (dorsal) localized anteriorly, those innervating the chin (ventral) localized posteriorly, and those innervating the cheek were dispersed throughout the ganglion (Figure 2K). Consistent with our observations for other effector organs, exit routes of skin-innervating axons matched their soma position: forehead neurons exit via the sup. branch, chin neurons via the inf. branch, and cheek neurons via both branches (Figures 2L-2N).

NPY marks a subset of SCG neurons

The neurochemical coding hypothesis posits that sympathetic effectors are innervated by neurochemically defined populations.^4,6-8 Given the selective targeting of SCG projections, we examined whether NPY selectively labels fibers associated with blood vessels.^21,27,28 Tyrosine hydroxylase (TH) was used to label all sympathetic fibers, and CD31 was used to mark vascular endothelial cells. Vasoconstrictor fibers were identified as TH⁺ fibers adjacent to blood vessels. In the skin, piloerector fibers ran parallel to hair shafts, whereas vasoconstrictor fibers localized subcutaneously (Figure 3A). NPY⁺ fibers, although sparse, were confined to subcutaneous regions near large blood vessels, consistent with a vasoconstrictor identity (Figure 3A’). In the salivary gland, TH⁺ fibers surrounded both secretory cells and the major CD31⁺ vessel entering the gland (Figure 3B). NPY⁺ fibers were restricted to vascular innervation (Figure 3B’), with secretory cells receiving TH⁺/NPY^- input (Figure 3B”). In the tongue, TH⁺ fibers were primarily localized near the ventral cavity surrounded by large vessels. Most fibers were NPY⁺, consistent with vascular innervation (Figure 3D’-D”).

Figure 3. — Immunostaining of cranial tissues showing sympathetic fibers (TH⁺/NPY⁺) and blood vessels (CD31⁺). Insets show separated color channels and the merged image of zoomed-in views highlighting distinct tissue compositions. (A) Facial skin: piloerector and vasoconstrictor fibers (A’); arrows mark TH⁺/NPY⁺ fibers around a subcutaneous vessel.

(B) Salivary gland: vasculature entering the gland (B′) with surrounding TH⁺/NPY⁺ perivascular fibers (arrows); secretory region (B″) lacks NPY⁺ fibers.

(C) Pineal gland: core region (C′) near a major blood vessel (C″); arrowheads mark TH⁺/NPY⁺ fibers surrounding the blood vessel, arrows indicate TH⁺/NPY⁺ fiber traversing the gland.

(D) Tongue: blood vessels surrounding the large cavity ventral to the tongue (D′); arrows mark TH⁺/NPY⁺ perivascular sympathetic fibers (D″).

(E) Iris: zoomed-in leaflet (E′ and E″) showing distributed TH⁺/NPY⁺ fibers throughout the organ, independent of vasculature.

(F) Summary table showing TH⁺/NPY⁺ fiber presence and vascular association in (A)–(E) (+, present; −, absent). All scale bars represent 100 μm.

Unlike skin, salivary gland, and tongue, where NPY selectively marks vascular fibers, in the pineal gland and iris, NPY expression was uniform across all TH⁺ fibers irrespective of the tissue domain (Figures 3C and 3E). In the pineal gland, NPY⁺ fibers were distributed throughout the tissue, following vessels as well as traversing secretory regions (Figure 3C”). In the iris, all fibers expressed NPY regardless of vascular association (Figure 3E”). Together, these results indicate that SCG neurons show molecular diversity, yet simple correlations with NPY are not apparent. For example, NPY does not differentiate vascular fibers from those innervating iris smooth muscle or pineal secretory cells (Figure 3F).

Transcriptomic heterogeneity of SCG neurons

To test whether the SCG transcriptome reveals genetically distinct neurons with distinct functional properties analogous to motor and sensory types that innervate defined targets,^15,29-31 we performed single-cell RNA sequencing (scRNAseq) and multiplexed error-robust fluorescence in situ hybridization (MERFISH).³² To avoid transient developmental programs, we used mice ≥ P14, thereby profiling transcripts of mature (adult-like) neurons.²² Since transcription factor Isl1 labels nearly all SCG neurons but not glial cells (Figures 4A and S2A), we crossed Isl1::Cre mice to a tdTomato reporter (Isl1^Cre;Ai14) and isolated tdTomato⁺ neurons by fluorescence-activated cell sorting (FACS) for scRNAseq (Figure 4A). MERFISH included a 140-gene panel, including NPY as an auxiliary channel. Genes were selected from the scRNAseq dataset, sampling enriched genes, gene modules, and candidate markers.

In the scRNAseq dataset, ∼2,800 neurons passed quality control (QC), yielding 8 transcriptomically distinct populations (T-types: Dv1–Dv8, Figure 4B) identified by a top-down divisive (Dv) clustering algorithm³³ (STAR Methods). In the UMAP plot, clusters appeared as a compact core with small protrusions rather than discrete islands (Figure 4B). Many of their differentially expressed genes (DEGs) also showed moderate expression across multiple clusters, limiting the utility of single-gene markers for defining SCG clusters (Figure 4D). Dv2 was an exception, defined by a largely restricted DEG set (Figure 4D). Accordingly, Dv2 appeared as the most distant cluster, showing the lowest Spearman’s correlation with other clusters (Dv2 vs. others: ρ = [0.26–0.29]; Dv1 and Dv3–Dv5: ρ = [0.37–0.60]; Figure 4C).

To quantify variable DEG expression patterns (Figure 4D) and facilitate marker selection, we developed a set of biological distinctiveness metrics that collectively evaluate how well each cluster is defined by marker-like genes and enable their ranking accordingly. These metrics included: the number of significantly enriched genes (n_sig_up); their average fold change (fc), normalized p values (e_score), and normalized mutual information (nmi); the Jaccard index of the best marker (ji); and cluster stability across iterations (cv_score) (STAR Methods). In particular, nmi measures how well the enriched genes mark their cluster, whereas ji captures the specificity of the single best marker. Rankings varied across metrics (Figure 4E), indicating that each metric captures a different aspect of cluster distinctiveness. We used their combined ranking to identify stable relationships. Dv2 ranked the highest, Dv3 the lowest, with the remaining clusters forming a gradient (Figure 4E). This quantitative assessment aligned with qualitative observations (Figure 4D), supporting our computational approach. These results show that SCG T-types differ in distinctiveness, with Dv2 emerging as the most prominent subtype at the gene-expression level.

We next combined statistical analysis with manual curation to generate minimal gene sets marking the Dv1–Dv8 clusters (STAR Methods; Figure 4F). Five clusters (Dv2–Dv5 and Dv7) had single markers specific to their cluster, whereas markers for Dv1, Dv6, and Dv8 were shared across multiple clusters (Figure 4F). To assess marker accuracy, we calculated false-positive (FPR) and false-negative (FNR) rates. As expected from scRNAseq dropout, FNRs were high, but several markers also showed elevated FPR (Figure S2B). Except for Dv3 (Cldn11⁺), adding a second positive marker (AND/OR logic) reduced error rates, yielding sets with FPR ≤0.1, FNR ≤0.35, and 0.30 ≤ precision ≤0.87 (Figures 4G and S2B). Adding negative markers conferred little benefit and increased FNR as more genes were added, thus they were excluded from the final sets (Figures 4G and S2C). The inability to improve the performance of Dv3 was consistent with its lowest distinctiveness rank (Figure 4E), further validating our distinctiveness metrics.

Taken together, scRNAseq revealed transcriptomically distinct SCG neurons clustered into 8 T-types. However, many T-types had close transcriptomic relationships.

Neurochemical profiles of SCG transcriptomic clusters

The majority of the cells with high levels of NPY expression were assigned to Dv6 and Dv7, although NPY⁺ cells were also present in other clusters (Figure 4H). Interestingly, Dv6 and Dv7 also contained many NPY⁻ cells, which could result in false negatives in scRNAseq due to dropout. To address this, we examined NPY expression across DV1–Dv8 clusters using MERFISH auxiliary channel. Consistent with the scRNAseq results, MERFISH showed that Dv6 and Dv7 contained the largest NPY⁺ fractions, yet more than half of their neurons were NPY⁻ (64% and 56%, respectively; Figures S2D-S2F). Both NPY⁺ and NPY⁻ cells within DV6 and Dv7 were readily identifiable within the tissue (Figure S2F). The co-clustering of NPY+ and NPY⁻ cells within these clusters suggests that NPY serves as a heuristic marker rather than a defining feature of cluster identity. Nonetheless, the identification of two T-types with high NPY⁺ cell content is notable, raising the possibility that pupillomotor and vasoconstrictor neurons, both of which express NPY (see Figure 3), may segregate into different T-types.

Given the heterogeneity of NPY, we next surveyed 120 neuro-chemical genes representing major neurotransmission pathways. Of the 101 genes with detectable transcripts, most were sparsely expressed at the single-cell level (53 in <10% of cells and 16 in 10–50%), while 7 were nearly ubiquitous (>90%) (Figures S2G and S2H). At the cluster level, 9 genes selectively marked a single cluster (expressed in >25% of cells within that cluster), and 27 marked multiple clusters (Figures 4I and S2I). These values were highly sensitive to the threshold used (data not shown). Differential expression of these genes indicates that there may be a combinatorial neurochemical code that distinguishes SCG neuron subtypes (Figure S2I), although the low expression of many of these genes raises questions about their functional relevance.

Spatial distribution of SCG transcriptomic clusters

Next, we utilized the MERFISH dataset to characterize the spatial organization of Dv1–Dv8 clusters. The large soma size of SCG neurons facilitated segmentation with Cellpose 2³⁴ (Figure S2J). After QC and removal of non-neuronal cells, ~9,800 neurons across 9 P14 SCG sections were analyzed. Using Seurat label transfer, we assigned Dv1–Dv8 identities from scRNAseq to MERFISH neurons. These clusters had high prediction scores (Figure S2K) and showed concordant expression with their scRNAseq counterparts (Figure S2L). Each MERFISH cluster highly expressed its assigned marker gene (Figures 4F, 4J, and S2L), further confirming marker accuracy. In short, scRNAseq clusters were reliably identified in tissue (Figures 4B and 4J), with the exception of Dv3, which was not detected with MERFISH, possibly reflecting their enrichment at early postnatal ages (see Figure 5).

Figure 5. — (A) UMAP plot of the integrated P0/P14 scRNAseq dataset and its L01–L12 clusters.

(B) Mapping between P14 SCG clusters (Dv1–Dv8) and integrated P0/P14 clusters (L01–L12). For each Dv cluster, the heatmap shows the percentage of its cells assigned to each L01–L12 cluster.

(C) UMAP plot of the integrated dataset with P0 and P14 cells, color-coded.

(D) Proportions of L01–L12 clusters in P0 and P14 SCG. Chi-squared test on 2 × 2 cluster-by-age counts; *p < 0.05, **p < 0.01, ***p < 0.001; N.S., non-significant.

(E) Changes in proportions of L01–L12 clusters from P0 to P14. Cluster identities and meta-cluster memberships are shown on the right. Gray arrows indicate the P14 clusters that each of the L01–L12 clusters maps to (B); “–” denotes no clear or diffuse correspondence.

(F) Schematic summarizing (D) and (E). Each circle represents a cluster, with meta-cluster membership indicated on the right. For simplicity, only three clusters per meta-cluster are shown. Each cluster is divided into P0 (red) and P14 (black) populations, with small icons representing individual cells. Shapes denote transcriptomic differences and colors indicate dataset origin. Relative proportions of red and black icons reflect differences in cluster proportions between P0 and P14 SCG.

(G) Left: ranking of L01–L12 clusters by distinctiveness metrics, ordered by average rank across all metrics (higher implies more distinct). Green and magenta dots indicate meta-cluster membership. Right: sum of ranks for each cluster, grouped by meta-clusters, with mean values shown as horizontal bars.

(H) nmi score ranking of P0 and P14 subpopulations constituting P14-dominant clusters (see F). y axis values indicate rank among all 12 clusters split into P0 and P14 populations (24 total). Only P14-dominant clusters are shown.

(I) Relative (fold-change) expression of selected marker genes, analyzed separately for P0 and P14. Wilcoxon rank-sum test, *p < 0.05, **p < 0.01, ***p < 0.001.

See also Figure S3.

MERFISH revealed distinct spatial distributions for several T-types: Dv1 and Dv4 were caudally biased; Dv5, Dv7, and Dv8 rostrally biased; and Dv6 uniformly distributed (Figures 4J and 4K). Clusters were uniform along the medio-lateral axis (data not shown). Dv2 was excluded from this analysis due to its small cell number. Although Dv1, Dv7, and Dv8 showed gradual transitions, Dv4 and Dv5 were sharply separated (Figures 4J and 4K), consistent with the compartmentalized SCG organization revealed by tracing (Figure 2O). Interestingly, Dv1 neurons localized more caudally than Dv4 (Figure 4K), raising the possibility of a third caudal compartment, potentially innervating thoracic organs via the ST branch. Within each compartment, T-types were intermingled (Figure 4L). Together, these results show that the spatial organization of T-types mirrors the pseudo-topographical organization of P-types (Figure 2O), with some clusters displaying more diffuse patterns.

To further validate our choice of markers, we examined the spatial patterns of transcripts irrespective of cluster identity. Transcripts fell into four distinct patterns: rostral, uniform, caudal, and distal-caudal (Figure 4M). Cluster markers followed the spatial distributions of their respective clusters (data not shown). We independently validated this by RNAscope for Areg (Dv4) and Sctr (Dv6) (Figure S2M). These four spatial transcript patterns likely arise because the MERFISH gene panel was enriched for DEGs, which tend to exhibit spatial distributions that mirror those of their corresponding T-types, yielding similar rostral-caudal biases.

We next investigated whether neuron location along the rostral-caudal axis is the primary factor driving such SCG clustering. Our analysis leveraged the hierarchical structure of the top-down divisive clustering algorithm,³³ which recursively split cells using K-means (K = 2) along the axis of greatest variance (Figure 4N; STAR Methods). If spatial position were the dominant factor, rostral and caudal T-types would separate at the top level, producing two daughter clusters, N1 and N2, each dedicated to rostral vs. caudal clusters (Figure 4N). While caudal-biased clusters were confined to N2 descendants, the most distinct rostral-caudal separation (Dv4 vs. Dv5) only appeared at the final N2 split (Figure 4N), indicating that spatial location is not the principal axis of variation among SCG neurons, despite its prominence.

In summary, SCG neurons exhibit transcriptomic heterogeneity that aligns with a compartmentalized yet intermingled organization. However, closely related T-types, such as those that diverge at later stages of hierarchical clustering (i.e., Dv4 and Dv5), are not necessarily confined to the same compartment.

Maturation of SCG T-types postnatally

To examine the ontogeny of mature T-types, we performed scRNAseq on P0 SCG neurons, labeled and purified using Isl1^Cre;Ai14. Since target innervation continues during the early postnatal period,²² P0 is likely to capture heterogeneity potentially associated with circuit assembly. The relatively low precision of marker genes for DV1–Dv8 clusters (Figure 4G) precluded the use of single marker genes to identify cell types across ages. Additionally, T-types are defined by transcriptomic states; therefore, relying on a small subset of genes to identify T-types can be misleading. Therefore, to relate the two datasets (P0 and P14), we applied Seurat’s integration algorithm, which merges datasets into a shared coordinate space, enabling their joint analysis.

Clustering of the P0/P14 integrated dataset yielded 12 clusters (L01–L12, Figures 5A and S3A). P14 clusters mapped predominantly to a single integrated cluster; however, Dv3 was dispersed among several clusters (Figure 5B). In the integrated UMAP plot, P0 and P14 cells occupied mostly distinct regions (Figure 5C), reflecting their differential contributions to the integrated clusters: most clusters were enriched for either P0 or P14 (Figure S3B). This result indicates that different clusters predominate P0 and P14 datasets.

Accordingly, the proportions of clusters within the P0 and P14 SCG differed significantly between ages (Figure 5D), with some increasing and others decreasing as the animals matured (Figure 5E). To systematically analyze these changes, we grouped L01–L12 into three meta-clusters: P0^dominant (L01–L06), NS (L07), and P14^dominant (L08–L12). The P0^dominant group comprised clusters whose proportions decreased from P0 to P14 (Figure 5E). As a result, these clusters constitute the majority of SCG neurons at P0, and become a marginal population at P14 (Figure S3D). In contrast, P14^dominant clusters increased in proportion from P0 to P14, constituting the majority of neurons at P14 (Figure S3D). A direct comparison of correlations among separately clustered P14 and P0 clusters is consistent with these observations (Figure S3F). In short, P0^dominant clusters predominate P0 SCG, whereas P14^dominant clusters predominate P14. However, nearly all clusters persist at both ages, but some persist at low proportions (Figures 5D and 5F).

Interestingly, P0^dominant clusters mapped diffusely to Dv3 (Figures 5B-5E and S3C), the least distinctive P14 cluster, whereas P14^dominant clusters mapped to P14 Dv clusters with higher distinctiveness scores (Figures 4E, 5B, and 5E). This suggested that the P0 SCG transcriptome may contain relatively less distinct clusters. To investigate this, we applied distinctiveness metrics to L01–L12 clusters. Many of the P14^dominant clusters consistently ranked higher in many of the distinctiveness metrics compared to the P0^dominant clusters (Figures 5G and S3E). Therefore, the P14 SCG comprises clusters that tend to be more distinct and better defined by marker-like genes than those predominant at P0.

Although present at low proportions, P14^dominant clusters were already detectable at P0 (Figures 5D-5F and S3B). Therefore, we investigated how the distinctiveness of P14^dominant clusters themselves change as they mature from P0 to P14. To do so, we split each P14^dominant cluster into its constituent P0 and P14 cells (e.g., P0_L12 vs. P14_L12), and investigated marker expression and nmi values of DEGs. With the exception of Sctr, marker genes became differentially expressed only at P14, with base-level expression at P0 (Figure 5I). In terms of nmi scores, all subpopulations of P14^dominant clusters with P14 cells ranked higher than all subpopulations with P0 cells (Figure 5H), indicating sharper molecular definition of P14^dominant clusters at P14. Thus, not only does the SCG shift toward P14^dominant clusters postnatally, but these clusters themselves become more distinctive with age.

Surprisingly, although Vip marks a P0^dominant cluster (L05), it became differentially expressed only at P14 (Figure 5I). This shows that Vip⁺ neurons, despite becoming a rarer population, also become more transcriptionally distinct as the animals mature, which may explain the top-ranked distinctiveness status of L05 (Figure 5G).

In summary, our results show that there is a systematic change in the SCG transcriptome during early postnatal development, characterized by a shift in cell type composition. During this period, a new set of transcriptomically defined clusters, which tend to be more distinct and well-defined, begins to dominate the SCG. Moreover, these new clusters themselves become progressively more distinct as the animal matures. Overall, these results suggest a divergent transcriptional trajectory for SCG neurons during early postnatal life.

Relationship between T- and P-subtypes

To investigate whether SCG T-types correspond to P-types, we used CTB retrograde labeling in combination with scRNAseq or MERFISH at P14 (Figures 6A and 6B). For scRNAseq, CTB was injected into the salivary gland or skin, and the labeled neurons were isolated by FACS (Figure S4A). For MERFISH, CTB was injected into the salivary gland, tongue, eye, and skin. CTB labeling on tissue prepared for MERFISH was more variable; therefore, CTB⁺ cells were assigned manually based on label intensity and the pattern of cell filling (Figure S4B). Dv1–Dv8 identities were then mapped onto CTB⁺ datasets by label transfer using a bootstrapping approach with thresholds to exclude unstable assignments (Figure S4C).

Figure 6. — (A and B) Schematics illustrating our central question (A) and strategy (B) to determine the relationship between T- and P-types of SCG. Numbers in circles indicate the sequence of experiments/analyses.

(C) Bar plots showing proportions of Dv1–Dv8 clusters in the whole ganglion and in salivary-gland or facial-skin-projecting neurons (scRNAseq). Numbers in parentheses indicate cluster cell counts. Error bars: 95% bootstrap CIs of mean proportions.

(D and F) Fold enrichment of each cluster in projection-defined populations relative to the whole ganglion for scRNAseq (D) and MERFISH (F). Dashed line marks no enrichment (y = 1). Error bars: 95% bootstrap CIs of mean enrichment. Chi-squared test, *p < 0.05, **p < 0.01, ***p < 0.001; ns, non-significant.

(E) Representative images of Dv1–Dv8 clusters within each P-type in the MERFISH dataset. Boxes marked with an “X” indicate absence of that cluster in the corresponding P-type. N/A, not available.

See also Figure S4

RNAseq of the salivary gland-projecting neurons (SCG^→Salivary) revealed that 80% of this P-type composed of cells from the Dv4 and Dv6 T-types, with small contributions from several additional clusters (Figure 6C). Labeling specificity was confirmed histologically, and the neurons exhibited the expected NPY profile (Figure S4D), indicating that multiple clusters contribute to salivary gland innervation. Likewise, sequencing of skin-projecting neurons (SCG^→Skin) revealed that this P-type contained both Dv6 and Dv7 cells (Figure 6C). Furthermore, comparison of cluster proportions revealed enrichment of multiple clusters within each P-type; however, different clusters were enriched for salivary gland and skin innervation (Figure 6D). Thus, SCG neurons that innervate a particular effector organ comprise multiple T-types. Conversely, individual T-types, such as the Dv6 neurons, innervate multiple effector organs.

A consideration with CTB labeling for scRNAseq was the cell recovery yield for analysis. Therefore, to confirm and extend our findings, we employed CTB tracing combined with MERFISH to examine the relationship between effector organ innervation and T-types. CTB-labeled cells corresponding to clusters constituting each P-type were readily detected in the tissue sections used for MERFISH (Figure 6E). Consistent with sequencing results, the SCG^→Salivary P-type contained Dv1, Dv4, and Dv6 neurons (Figure S4E), with particular enrichment for Dv4 (Figure 6F). The SCG^→Eye P-type consisted exclusively of Dv6 neurons, whereas the SCG^→Tongue P-type included both Dv6 and Dv7 neurons (Figures 6F and S4E). For skin innervation, MERFISH detected Dv1 and Dv4 neurons in addition to Dv6, while only a few Dv7 neurons were observed (Figures 6F and S4E). In general, the distribution of multiple SCG T-types innervating individual effector tissues was consistent, as observed by scRNAseq and MERFISH (Figures 6D and 6F). Under the specific conditions of our sampling, MERFISH was particularly sensitive and detected significant numbers of Dv1, Dv4, Dv6, and Dv7 neurons innervating the facial skin; however, it is unclear why the proportion of Dv7 neurons was reduced compared to scRNAseq (Figures 6D and 6F). Nonetheless, each P-type examined comprised multiple T-types, as observed by both scRNAseq and MERFISH.

Overall, our findings show that individual P-types of SCG neurons innervate a single sympathetic effector organ of the head. Remarkably, P-types comprisemultiple T-types, indicating that there is not a strict one-to-one correspondence. Individual T-types project to multiple effector organs (i.e., one-to-many) and individual effector organs are innervated by multiple T-types (i.e., many-to-one). These findings indicate physiological responses associated with the cranial sympathetic activity are mediated by patterned activity of multiple SCG T-types.

DISCUSSION

The SCG innervates multiple cranial tissues and contributes to diverse sympathetic responses associated with emotional and physiological states. Here, we combined scRNAseq, MERFISH, and retrograde tracing to define the organization of SCG neurons at single-cell resolution. We identified segregated output populations (P-types) for each effector organ, transcriptomic subtypes (T-types) with spatial biases, and a developmental transition toward more distinct transcriptional identities. Importantly, T-types were enriched, but not exclusively dedicated, to specific P-types. This indicates that transcriptomic identity does not map in a simple one-to-one manner onto effector organs. In the following text, we discuss these findings in the context of sympathetic development and circuit logic.

SCG target innervation

The existence of segregated SCG neuron populations dedicated to individual effector organs suggests that this organization may be a common motif in motor systems. With disjoint innervation of each effector, such an arrangement may offer advantages where selective recruitment of specific outputs is needed, as observed across emotional states.¹³ At the same time, in many ethological contexts, such as during fight-or-flight responses, sympathetic outputs are co-activated across multiple organs.^2,3 Our results imply that this coordination, much like in the somatic motor system, must occur upstream of the output layer of the network.^35-37

The SCG is divided into rostral and caudal compartments, corresponding to neurons exiting through the superior or inferior nerve branches. This organization may reflect early developmental guidance cues, secreted near future exit points that bias axons toward one branch or the other. This organization, in turn, constrains the set of possible targets available to neurons for innervation. Caudal neurons tend to leave via the inferior branch and project to ventral head structures, whereas rostral neurons, preferentially leaving via the superior branch, innervate dorsal regions. Despite this spatial parcellation, our top-down hierarchical clustering analysis of T-types did not identify spatial location as the primary axis of molecular variation. This suggests that greatest transcriptomic variation among SCG neurons arise before, or independently of, the establishment of rostral-caudal compartments.

Postnatal diversification of SCG subtypes

Although most SCG axons have exited the ganglion by birth, our transcriptomic analyses revealed substantial changes between P0 and P14. During this period of active target innervation and organ maturation, SCG neurons shift from relatively homogeneous early states into eight transcriptomically defined T-types. These P14 T-types display clearer marker-like genes than their P0 counterparts, indicating that transcriptional identities become more distinct as development proceeds postnatally. A similar pattern has been observed in dorsal root ganglion (DRG) neurons,³⁸ suggesting this may be a broader feature of neural crest-derived lineages. Since sympathetic neurons undergo little birth or death after P0,^22,39 these changes most likely reflect transitions of existing cells from immature P0 states into more differentiated P14 T-types, consistent with the presence of differentiating neurons in early postnatal sympathetic ganglia.²² The timing of these changes, coinciding with target innervation and organ maturation, suggests that peripheral interactions may instructively refine SCG transcriptomic identity. In line with this idea, several target-derived signals (i.e., NGF, BDNF, endothelins, etc.) are known to influence sympathetic neuron properties.^16,40-44 Together, these observations point to a prominent role for cell-extrinsic cues in shaping SCG neuron identity. In contrast, somatic motor neurons tend to rely more heavily on cell-autonomous genetic programs.

Spatial organization and input formation of SCG neurons

Somatic motor pools within the spinal cord are topographically organized into discrete columns relative to their muscle targets, which likely support accurate premotor input and proper circuit formation.⁴⁵ In contrast, SCG P-types exhibit a pseudo-topographical organization with a salt-and-pepper-like arrangement within each compartment. This arrangement implies that although spatial location alone may simplify the task of establishing the proper presynaptic inputs, it is insufficient for PGC neurons to discriminate among SCG P-types. Additional guidance cues may, therefore, help align PGC inputs with the appropriate postganglionic neurons. Interestingly, transcriptional studies of embryonic PGC neurons identified two transcriptomically distinct subtypes,⁴⁶ raising the possibility that each SCG compartments may be innervated by distinct class of PGCs.

SCG transcriptomics and circuit logic

By integrating tracing with transcriptomic analyses, we found that SCG T-types are differentially enriched across P-types, but do not map in a one-to-one fashion. This prompted us to consider whether T-types defined at P14 correspond to the identity of their post-synaptic targets rather than their axonal projection patterns. Many effector organs contain multiple tissue types including, for example, blood vessels present in multiple targets. SCG T-types possibly correspond to post-synaptic tissue types within the effector organs. This could account for the shared Dv6 and Dv7 clusters as putative vasoconstrictors. However, this model does not explain cases such as Dv1 and Dv4 neurons, which innervate both piloerectors and the salivary gland. Nor does it account for why the pupil, despite being distinct from blood vessels, maps exclusively to Dv6. A related idea suggests that T-types correspond to broad tissue classes, such as smooth muscle or secretory cells, but this is also inconsistent with our observations. Although piloerectors are smooth muscles, they are innervated by Dv1 and Dv4 rather than by Dv6 or Dv7, which likely target vascular smooth muscle.

Alternatively, T-types may reflect patterns of presynaptic input rather than output. If transcriptomic identity influences those subsets of PGC neurons that form synapses onto each SCG neuron, then T-types could represent units through which upstream circuits coordinate sympathetic responses across multiple organs. This framework would be consistent with classical reflex-arc descriptions of sympathetic neurons^8,10 and would position T-types as “functional reflex units”. It would also provide a biologically plausible explanation for why individual T-types project to more than one effector. However, the postnatal emergence of SCG T-types versus the embryonic establishment of PGC inputs^47,48 makes this model difficult to reconcile with developmental timelines. Nonetheless, since Dv6 is shared between vasoconstrictor and pupillomotor neurons, this model would predict that pupil dilation should be 100% coupled with facial blood vessel constriction in the absence of parasympathetic activity.

The complexity we uncovered in the SCG system represents a counterexample to circuit architectures in which one-to-one relationships between transcriptomic types of neurons are linked to a single biological feature. We found that even the transcriptomic relationships within the same T-type are complex. For example, several SCG T-types comprise both NPY⁺ and NPY⁻ neurons, despite the established role of this neuromodulator for promoting vasoconstriction. Our findings indicate that the reductionist approach of linking transcriptomic neuron types with biological functions may not scale to all circuits controlling complex behavioral repertoires.

Limitations of the study

Retrograde tracing was performed from only a subset of effector organs, and several targets yielded few labeled neurons. Because each organ comprises multiple tissue types, including shared components such as vasculature, selective tracing of specific compartments was not technically feasible. Combined with the inherent variability and low signal intensity of CTB labeling, these factors likely reduced sensitivity and introduced noise despite stringent filtering and QC measures.

Our transcriptomic analysis was restricted to postnatal stages to allow comparison with projection patterns. Although we observed progressive divergence of transcriptional programs from P0 to P14, we did not capture embryonic periods during which SCG neurons undergo substantial differentiation.^16,49 As in the spinal cord, key marker genes may be transiently expressed embryonically and diminish postnatally,^50,51 raising the possibility that embryonic T-types may align more closely with P-types. However, a more complete understanding will likely require conceptualizing cell types as dynamic trajectories across ontogeny rather than snapshots at individual ages.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

All animal procedures were approved by the Salk Institute Institutional Animal Care and Use Committee under protocol 23-00022 and were conducted in accordance with NIH guidelines. The Salk Institute maintains an approved PHS Animal Welfare Assurance (D16-00328), USDA registration (93-R-0066), and AAALAC International accreditation. Isl1 – Cre mice⁵² (JAX 024242) were crossed to Ai14 (JAX 007914) to generate Isl1-Cre;Ai14 animals used for scRNAseq at P0 and P14. CTB tracing, histology, and sequencing were performed on wild-type mice (P10–P14 or adult; mixed background). Both sexes were used, animals were randomized to experimental groups, and no sex differences were observed. Mice ranged from P0 to 2 months of age, were housed under standard temperature and light conditions (12 h light/12 h dark), and had ad libitum access to food and water.

METHOD DETAILS

General surgical considerations

All surgical procedures and injections were performed under general anesthesia using aseptic techniques. Eye ointment was applied prior to surgery. Adult mice were anesthetized with an intramuscular cocktail (0.05 mg/kg fentanyl citrate, 5 mg/kg midazolam, 0.5 mg/kg dexdormitor). P10 pups were anesthetized with 4% isoflurane for induction and 2.5% for maintenance. Following surgery, adults received antagonists intraperitoneally (0.5 mg/kg flumazenil, 2.5 mg/kg antisedan) and 1 mg/kg buprenorphine SR for analgesia. P10 pups recovered spontaneously after isoflurane cessation. Animals were maintained on heating pads until fully mobile; P10 pups were then returned to their parents. Ibuprofen was provided in drinking water for all mice, which had ad libitum access to food and water throughout.

Cholera toxin subunit b (CTB) injections

CTB conjugated to Alexa dyes (Invitrogen C22841, C22843, C34778) was injected into P10 pups or adult mice (6 weeks–2 months). Animals were processed for histology, scRNAseq, or MERFISH 4–5 days later. Eye: Intravitreal injections were performed under anesthesia using a stereotaxic frame. A 27-gauge needle was used to puncture the sclera, followed by insertion of a glass micropipette to deliver 3 μL of 0.5% CTB. Salivary gland: Animals were placed supine, a midline neck incision was made, and lobes were exposed by blunt dissection. CTB was injected directly into the gland (adults: 3 μL of 0.5%; P10 pups: 0.5 μL of 0.5%). Incisions were closed with 6-0 silk sutures. Facial skin: Intradermal injections were delivered with a Hamilton syringe. Adults received 2.5 μL of 0.25% CTB at 6–8 sites across the forehead, cheek, and chin. P10 pups received two injections (5 μL each, 0.5% CTB) restricted to forehead and cheek. Tongue: A 31-gauge Hamilton syringe was used to deliver 1 μL of 0.5% CTB into the tongue.

Tissue preparation, staining and imaging

Tissues were collected from adult mice or ~P15 pups. Adult samples included SCG, iris (albino ICRs), tongue, salivary gland, pineal gland, and facial skin; pup samples included SCG, salivary gland, facial skin, and tongue.

Mice were euthanized and transcardially perfused with ice-cold PBS followed by 4% PFA (Electron Microscopy Sciences 15713). Dissected tissues were post-fixed in 4% PFA for 4 h at 4°C (SCG post-fixed 2–4 h). SCG were dissected in two steps: gross isolation with adjacent structures (carotid artery, vagus nerve) followed by fine dissection to isolate the ganglion.⁵³ After post-fixation, tissues were washed in PBS, cryoprotected in 30% sucrose (2 h for SCG/pineal; 2 days for tongue, salivary gland, facial skin), embedded in OCT (Tissue-Tek 4583), frozen, and cryosectioned (20–30 μm). For whole-mount imaging, SCG and iris samples were kept in PBS without freezing; iris was always processed as a whole mount without sucrose treatment.

For immunohistochemistry, sections or whole mounts were blocked in 3% horse serum (Vector S-2000-20) in 0.3% Triton-X/PBS and incubated with primary antibodies at 4°C (overnight for sections; 3 days for whole SCG/iris). After PBS washes, tissues were incubated with fluorophore-conjugated secondary antibodies (2 h RT for sections; overnight 4°C for whole mounts) and counter-stained with DAPI (Invitrogen D1306) or NeuroTrace (Invitrogen N21483). Sections were mounted in Mowiol. Whole SCG and iris were dehydrated sequentially in glycerol (25%, 50%, 80%; 2 h each) and mounted in 80% glycerol.

Primary antibodies: rabbit anti-NPY (1:1000, Peninsula T-4070), sheep anti-TH (1:1000, Novus NB300-110), rabbit anti-TH (1:1000, Millipore AB152), rat anti-CD31 (1:1000, BD 550274). Species-matched secondary antibodies (Thermo Fisher) were used at 1:1000. All antibodies were diluted in 3% horse serum in 0.3% Triton-X/PBS. Isl1-Cre;Ai14 and CTB signals were detected via endogenous tdTomato or the CTB fluorophore.

Images were acquired on an Olympus FV3000 confocal microscope using 4x, 10x, or 20x objectives at 1024x1024 or 2048x2048 resolution with or without tiling. Images are shown as z-projections.

RNAscope

Fresh-frozen (unfixed) SCG were cryosectioned at 20 μm and processed using the RNAscope Multiplex Fluorescent V1 assay (ACD, 320851). Probes for Areg (ACD 430501) and Sctr (ACD 508511) were used. All steps followed the manufacturer’s protocol except that protease treatment was performed with Protease IV for 30 min.

Sample preparation for 10x scRNAseq

scRNAseq was performed on P0 and P14 pups. Whole-ganglion sequencing used Isl1-Cre;Ai14 mice (n=5 per age), with tdTomato-positive SCG neurons isolated by FACS. Projection-dependent sequencing used P14 wild-type mice that received CTB injections into the salivary gland (n=21) or facial skin (n=5) 4 days before dissection; CTB⁺ neurons were isolated by FACS.

Mice were euthanized with isoflurane and decapitated. SCG were dissected in ice-cold HBSS (Gibco 14175-095). Ganglia were enzymatically dissociated in a mixture of papain/DNase (Worthington LK003150) and collagenase IV (2.5–3 mg/mL; Worthington LS004188) for 20 min at 37°C with agitation. Following digestion, tissue was mechanically triturated in DMEM/F12 with 10% FBS, and the suspension was sequentially pelleted (200g, 5 min), washed (DMEM/F12/FBS) and then passed through 40-μm strainers to obtain a single-cell suspension. All solutions and pipettes were BSA-coated (1%) and kept on ice.

Cells were sorted on a BD FACS Diva into either (1) a 24-well plate for quality and viability assessment (DAPI), or (2) PBS for 10x processing. Sorted cells were pelleted (300g, 5 min), resuspended in ~50 μL PBS, and loaded onto the 10x Chromium instrument. Single-cell partitioning and library preparation followed the manufacturer’s protocol (10x Genomics). Libraries were sequenced on an Illumina HiSeq 4000 using paired-end reads at the UCSD IGM Genomics Center.

Sample preparation for MERFISH

MERFISH was performed on fixed-frozen SCG with or without prior CTB tracing. The reference dataset used one SCG section from each of nine P14 wild-type mice. For projection-dependent MERFISH, wild-type mice received CTB injections 4 days prior to dissection targeting the salivary gland (n=2, P10), facial skin (n=2, P10), tongue (n=3, P10), or eye (n=3, adult). Three sections per sample were collected (total 30 sections).

SCG were perfused, post-fixed in 4% PFA for 24 h at 4°C, embedded, and cryosectioned at 10 μm. MERFISH was performed on the Vizgen MERSCOPE platform following the manufacturer’s fixed-frozen protocol using a 140-gene panel. Cell boundary staining was included for segmentation. CTB protein was detected using goat anti-CTB (1:12,000; List Labs #703) and Anti-Goat Aux 6 protein stain as secondary. Sections were cleared overnight at 47°C and imaged on the MERSCOPE system according to Vizgen’s instructions.

QUANTIFICATION AND STATISTICAL ANALYSIS

GraphPad Prism software and R were used for basic statistical analyses. Data are presented as mean ± standard error of the mean (SEM) when applicable, except in Figures 5I, 6C, 6D, 6F, and S4E, where error bars represent 95% bootstrap confidence intervals (CI) of the mean.

Permutation KS test was used to compare spatial distribution differences among SCG transcriptomic types (Figure 4K). Chi-square test was used to compare the distribution of cluster proportions between P0 and P14 SCG (Figure 5D) and the fold enrichment of proportions of SCG transcriptomic types among different projection patterns (Figures 6D and 6F). Wilcoxon rank-sum test was used to assess the significance of fold changes in marker gene expression (Figure 5I). Mann–Whitney test was used to compare NPY⁺ cell percentages among salivary gland-projecting neurons in pups versus adults (Figure S4D). In all cases, the following statistical significance indicators were used: *p < 0.05, **p < 0.01, ***p < 0.001; ns, non-significant. See below for more details.

CTB detection, contour plots and colocalization

CTB⁺ cells were manually identified using the Cell Counter plugin in ImageJ. For contour plots, a coordinate system was defined for each ganglion by setting the origin at the entry point of the incoming nerve and the x-axis along the line connecting the origin to the rostral superior nerve exit. To examine the spatial distributions of labeled cells, one- and two-dimension kernel density estimations were calculated in MATLAB using the “kde” and “kde2d” functions (Matlab file exchange) for each individual animal and averaged to create mean density estimations. Colocalization was assessed manually by scoring single- and double-labeled cells.

Adult versus pup NPY+ SCG cells

NPY⁺ cells and their colocalization with CTB were quantified by manual counting as described above. The percentage of NPY⁺ neurons among CTB-labeled cells was compared between pups and adults using an unpaired nonparametric Mann–Whitney test (GraphPad Prism). n=3 animals in each group, p < 0.05 was used as a significance measure.

scRNAseq preprocessing and QC

Paired-end reads were aligned with Cellranger, and the filtered feature–barcode matrix was used as the UMI count table. Counts were imported into R for downstream analysis, including QC filtering, neuronal cell identification, highly variable gene selection, PCA, UMAP, and Leiden clustering (see scRNAseq cell type calls and clustering). Top-down divisive clustering followed a separate preprocessing workflow. In the P14 whole-SCG dataset, a major expression gradient driven by Malat1, Snhg11, and Meg3 was observed. To evaluate cellular organization independent of this effect, we used Seurat’s module scoring and regression to remove this pattern during data standardization prior to dimensionality reduction.

QC metrics included total UMIs, number of detected genes, percent mitochondrial UMIs, and percent expression of ontology-based gene sets.⁵⁴ Metrics were evaluated as histograms to remove outlier tails. For ambiguous cases, quantile-based thresholds were applied, where the lower threshold = median − 3 × (median − 25th percentile) and the upper threshold = median + 3 × (75th percentile − median). Cells failing any threshold were excluded. In the P14 whole-SCG dataset, two additional contaminant clusters (Apoe⁺ glial contaminated cluster and Slc17a6⁺ putative sensory neurons) were removed.

scRNAseq cell type calls and clustering

Cell type identification: Neurons were identified by comparing expression of known markers for neuron (Snhg11, Snap25, Elavl3, Tubb3), glia (Apoe, C1qa, C1qb, C1qc, Tyrobp), schwann (Mpz, Pmp22) and vascular (Myl9) types; cells expressing neuronal markers at higher levels than non-neuronal markers were assigned as neurons.

Top-down divisive clustering (P14 SCG): Clusters were defined using a top-down divisive algorithm in which each proposed split was evaluated using k-means (k=2).³³ For every split, standard scRNAseq preprocessing was performed, including highly variable gene selection, log-normalization, regression of the Malat1/Snhg11/Meg3 gradient (see scRNAseq preprocessing and QC), PCA, and automated selection of significant PCs. Because k-means is non-deterministic, each split was run 100 times; the solution with the lowest cluster index (within-cluster sum of squares/total sum of squares) was selected. Proposed splits were accepted only if the observed cluster index was significantly lower than expected under a null model: Gaussian data with matched variances were repeatedly generated and clustered, and p-values were computed from the resulting null distribution. Splits with p < 0.01 were retained. This produced a final set of 8 clusters. These clusters were cross-validated to produce the final clustering solution (see cross-validation of clustering solutions).

Cross-validation of clustering solutions

Clustering solutions for scRNAseq data were cross-validated to access cell-assignment confidence and cluster stability. 5-fold cross validation was performed using Seurat’s label transfer workflow as a classifier for 50 trials. The top 2000 highly variable genes were used across all iterations. At each train/test split we split the single-sample Seurat object into 2 Seurat objects. Highly variable genes were called on the training split and a new PCA reduction was calculated. The cluster assignments for the test split were then predicted using Seurat’s label transfer workflow with the training split as reference. From each trial we retained the predicted cluster assignment for all cells and their accompanying prediction scores. The average of each trial’s prediction score was calculated for a final confidence measure and each cell’s final cluster assignment was based on the majority predicted identity across trials.

Calculation of classification metrics

We calculated classification metrics for genes in scRNAseq data with respect to cluster assignments to assist in finding and evaluating markers for clusters. Metrics calculated were true positive rate (TPR, recall), positive predictive value (PPV, precision), false positive rate (FPR), false negative rate (FNR), F-measure (F1 score), Jaccard index, and normalized mutual information (NMI). For each cluster at each gene we binarize the gene expression (non-zero expression = 1) and the clustering assignments (cluster being evaluated = 1, all others = 0. The binarized clustering vector is used as the known positives and the binarized expression as the predicted positives. From these we count true positives (expression = 1, clustering = 1), false positives (expression = 1, clustering = 0), true negatives (expression = 0, clustering = 0) and false negatives (expression = 0, clustering = 1). These 4 counts are then used to calculate TRP, PPV, FPR, and FNR. The F-measure is the harmonic mean of TPR and PPV (recall and precision). The Jaccard index and NMI were calculated as described elsewhere from the binarized gene expression at each gene and the per-cluster binarized clustering vector.

scRNAseq cluster correlogram

Correlograms were generated using average of cell-to-cell Spearman correlations, computed either between clusters (off-diagonal) or within each cluster (diagonal). Cell-to-cell correlations were calculated across the highly variable genes selected during initial dimensionality reduction.

Differential gene expression testing

Differential expression (DE) was assessed using the Wilcoxon rank-sum test. Each cluster was compared against the background of all other clusters. Cluster and background pools were downsampled to a maximum of 100 cells and tested across multiple trials. For each gene, p-values from all trials were sorted, and the value at the 51st quantile was taken as the final p-value. Post-hoc adjustment was performed by scaling p-values by the total number of genes detected in the experiment, ensuring consistency across clusters with different numbers of testable genes. Genes with adjusted p < 0.05 were considered significantly differentially expressed.

Over-representation analysis was also performed using binarized expression (gene detected = 1; not detected = 0). For each gene, a 2×2 contingency table was constructed comparing the number of cells detecting the gene within the cluster versus the entire dataset, with complementary counts used to complete the table. Significance was evaluated using a chi-square test.

Distinctiveness metrics

Cluster distinctiveness was quantified using several feature-based metrics. For each cluster we computed the following metrics based on the cluster’s significantly enriched and over-represented genes (see differential gene expression testing and calculation of classification metrics). From those genes we computed the following:

n_sig_up: the number of genes
fc: the mean log₂ fold-change
nmi: the mean normalized mutual information
e_score: the mean −log10(post-hoc p-value) from Wilcoxon rank-sum test
ji: the maximum Jaccard index

cv_score: From cross-validation trials (see Cross-validation of Clustering Solutions) we used the average prediction scores from cells per cluster to calculate each cluster’s cv_score.

To summarize and compare distinctiveness metrics we ranked them (‘rank’ function in R). Sum of rank values (Figures 4E and 5G) are the sum of the individual ranked metrics described here. For P0/P14 integrated clusters, distinctiveness metrics were calculated separately for each P0 and P14 subset of each cluster. Ranks of P0 and P14 subsets were then averaged to yield a ranking score for each cluster (Figure S3E). For Figure 5H, P0 and P14 ranks kept separately for each cluster.

Marker identification

Marker genes for each cluster were identified by filtering results from differential expression and detection over-representation tests (cluster vs. all other cells). To prioritize genes that are both enriched within a cluster and sparse outside it, classification performance metrics were computed for each gene (see calculation of classification metrics). Candidate markers were ranked by F-measure (F1 score), and final selections were done by inspection of violin plots or heatmaps and, where relevant, biological considerations.

Combinatorial marker sets

Combinatorial marker pairs were identified by starting with the top single-gene marker for each cluster and evaluating secondary positive markers in AND and OR combinations. Classification performance metrics (false-positive rate, FPR; false-negative rate, FNR) were computed as described in calculation of classification metrics. For each gene pair, binarized expression vectors were combined such that a cell was scored as positive if both genes were detected (AND) or if either gene was detected (OR).

Gene pairs were evaluated against predefined thresholds of FPR ≤0.1 and FNR ≤0.345, the latter chosen based on the performance of Vip in cluster Dv2, which serves as a well-established benchmark marker. All significantly enriched genes (Wilcoxon rank-sum test, post-hoc p < 0.05) were considered as candidate secondary markers. For clusters whose primary marker exceeded the FPR or FNR thresholds, secondary markers were screened to identify combinations that brought both metrics within the acceptable range.

Negative combinatorial markers were evaluated using the eight single-gene markers defining the P14 scRNAseq clusters. For each cluster, the positive markers of the other seven clusters were treated as candidate negative markers. Its positive marker was then paired with 1–7 negative markers. Binarized expression of negative markers was inverted and combined with the positive marker using logical AND. For multiple-negative cases, all possible gene combinations were tested, and the combination yielding the maximal F-measure was selected. Classification performance (FPR, FNR, F-measure) was quantified for each cluster, and results were plotted as FNR or FPR versus the number of negative markers.

Npy and neurochemical expression analysis

MERFISH cells were classified as Npy⁺ or Npy⁻ by thresholding Npy signal intensity. The threshold was determined by sorting Npy values and identifying the elbow point, defined as the point with the maximum perpendicular Euclidean distance from a line connecting the first and last points (Figure S2E).

A list of 120 neurochemical genes was examined, and genes were considered detected if ≥1 UMI was present, yielding 101 detected genes. A neurochemical gene was defined as marking a cluster if it was expressed in >25% of that cluster’s cells. Histograms were used to summarize neurochemical gene expression patterns by showing how many clusters (Figure 4I) or what fraction of total cells (Figure S2H) were marked by what number of neurochemical genes.

P0 and P14 scRNAseq integration and clustering

P0 and P14 SCG neurons were integrated using Seurat’s integration workflow, applied iteratively to maintain a balanced representation of both datasets. In each iteration, 2,000 cells were randomly subsampled from P0 and P14, integrated using P14 as the reference, reduced with PCA, and clustered using a shared nearest-neighbor graph (20 neighbors) and the Leiden algorithm (resolution = 1). Cluster co-membership across iterations was aggregated into a pairwise consensus matrix and partitioned by hierarchical clustering to obtain a preliminary solution. The number of clusters was selected to maximize similarity between P14 consensus identities and their previously defined divisive-clustering identities, quantified with normalized mutual information (NMI). This solution was then cross-validated to yield the final integrated clustering (see Cross-validation of Clustering Solutions).

Integrated P0/P14 cluster assignments were compared to divisive-clustering assignments using a row-normalized contingency table (Figure 5B). P0 and P14 proportions within each integrated cluster were tested for dataset bias (clusters that significantly deviate from being equally proportioned) using chi-square tests on P0 versus P14 counts within each cluster against all other cells (Figure 5D). Log₂(P14/P0) proportion ratios (Figure 5E) were computed by subsampling 2,000 P0 and 2,000 P14 cells, counting cluster membership, adding a pseudocount of 1, and taking the log₂ ratio. This procedure was repeated for 100 trials and averaged. Clusters significantly biased toward P0 or P14 were labeled P0-dominant or P14-dominant. Differential gene expression (see differential gene expression testing) was performed per cluster within each dataset using the integrated clustering solution. Top single-gene markers for P14 divisive clusters were taken from these results, and square-root fold-changes were plotted for all integrated clusters in both datasets (Figure 5I; Wilcoxon rank-sum test).

Mapping P14 SCG clusters to projections datasets

Cluster identities from the P14 whole-SCG dataset (Dv1–Dv8) were transferred to projection-defined scRNAseq and MERFISH datasets using Seurat’s anchor-based label transfer with a 5-fold iterative cross-validation procedure with 50 trials. For each trial, reference and query datasets were each randomly partitioned into five bins and cluster labels were predicted at each bin using a different subset of reference cells. In each bin loop iteration we leave one reference bin out and use the remaining four reference bins as the training set. Cells were assigned confidence categories based on identity stability across trials.

Core: same identity in >90% of trials
Transient: multiple identities, but one identity >50% of trials
Low confidence: no identity >50% of trials

For each cell, the mean prediction score (Seurat output) was also computed across trials. Prediction-score histograms stratified by confidence category were used to determine score thresholds that maximized retention of core cells while minimizing inclusion of transient cells.

MERFISH gene-panel and data analysis

A custom 140-gene MERFISH panel (123 target genes) was designed based on the scRNA-seq dataset, including significantly enriched genes, gene modules, and candidate markers; Npy was included in an auxiliary FISH channel. MERFISH data were acquired on the MERSCOPE system and decoded using Vizgen software (v232).

Cell segmentation was performed using the Vizgen Postprocessing Tool (VPT) with CLAHE normalization and a custom Cellpose2 model³⁴ trained on Cellbound3-stained images. Segmentation was applied to all z-planes (0–6), and cell geometries were merged across planes using the VPT “Harmonize” strategy with a minimum cell area of 500 pixels. VPT sum-signals and derive-entity-meta-data were used to generate cell-by-gene count matrices, summed intensity values (including high-pass–filtered NPY and CTB intensities), and geometric features.

For spatial analysis, each section (9 reference SCG; 28 SCG with CTB) was aligned to a standardized coordinate system with the ST entry point as the origin and the rostral–caudal axis defined anatomically. Ganglia were size-normalized using a custom Python script by independently scaling the y-axis, left x-axis, and right x-axis such that each ganglion’s rostral–caudal and mediolateral extents matched the population averages.

MERFISH QC and neuron identification

MERFISH cells were retained if they expressed ≥25 genes. Neurons were identified using a combination of clustering and detection of positive neuronal markers from the gene panel, and assignments were further confirmed by inspecting cell morphologies in the MERFISH visualizer.

Mapping P14 SCG clusters to MERFISH

Cluster identities from the P14 whole-SCG dataset (Dv1–Dv8) were transferred to the MERFISH reference dataset (no CTB tracing) using Seurat’s anchor-based label transfer with SCT normalization and all MERFISH-detected genes. The procedure produced predicted identities and associated prediction scores (0–1). Cells with prediction scores ≥0.6 were retained as confidently annotated based on the observed transition between ambiguous and stable assignments; cells below this threshold were excluded.

MERFISH versus scRNAseq P14 cluster heatmap

Cluster-average normalized expression profiles were computed for P14 whole-SCG scRNAseq and P14 MERFISH datasets using the intersection of genes present in both assays. MERFISH cluster identities were assigned using the scRNAseq–derived labels (see previous section). Average expression profiles from both datasets were z-scaled and combined into a single heatmap. Rows (clusters) and columns (genes) were ordered by hierarchical clustering; the cluster dendrogram was cut into 7 groups, corresponding to the 7 P14 scRNAseq clusters recovered in the MERFISH dataset.

MERFISH cluster spatial patterns

For each cluster, spatial density profiles were computed by pooling aligned, normalized rostral–caudal positions of all cells and estimating density curves in R (density function). To control for variable ganglion sizes, density estimates were averaged over repeated subsampling of each ganglion to a common cell count.

Cluster-specific spatial distributions were tested for deviation from the overall neuronal distribution using a permutation-based Kolmogorov–Smirnov (KS) test. For each ganglion, rostral–caudal positions of all neurons (“control”) and of cluster-assigned neurons (“test”) were extracted. Control and test sets were repeatedly subsampled to match the smallest ganglion size. Null KS statistics were generated by randomly permuting control/test labels (1,000 iterations). Observed KS statistics were computed over 100 subsampled iterations. p-values were obtained by comparing observed statistics to the null distribution, and the final p-value for each cluster was taken as the median across iterations.

Spatial distribution of genes in MERFISH

For each gene, spatial expression profiles along the rostral–caudal axis were tested against a uniform (cell-density–matched) distribution. Gene-specific spatial distributions were generated by repeating each cell’s position according to its UMI count for that gene. Cell-position distributions from the same ganglia served as controls.

Significance was assessed with a permutation-based Kolmogorov–Smirnov test: expression and cell distributions were subsampled to a common size, an observed KS statistic was computed (100 iterations), and null statistics were generated by random shuffling (1,000 iterations). The median p-value from comparisons to the null distribution was used; genes with p < 0.05 were considered spatially biased.

Significant genes were grouped into four spatial classes (most caudal, caudal, rostral, uniform) by hierarchical clustering of normalized 31-bin rostral–caudal histograms averaged across ganglia.

Identification of CTB+ cells in MERFISH

CTB⁺ cells were manually identified in the MERSCOPE Visualizer based on label intensity and completeness of cell filling. Each candidate cell was assigned a confidence class (high, mid, low) to account for variability in CTB signal quality. CTB signal density (high-pass–filtered CTB intensity normalized to cell volume) was compared across classes, and only high- and mid-confidence cells were retained for downstream analyses.

Cluster proportion barplots

Cluster proportions were estimated by bootstrapping cluster identities within each dataset. For each bootstrap sample, cells were counted per cluster and converted to proportions by dividing by the total cell number. Bar heights represent the mean proportion across bootstrap iterations, and error bars denote the 95% confidence interval.

Fold enrichment of proportions barplots

Fold enrichment was computed by dividing bootstrapped cluster proportions from each query dataset by bootstrapped proportions of control cluster identities. Bar heights represent the mean fold-enrichment across bootstrap iterations, and error bars denote the 95% confidence interval. For scRNAseq analyses, projection-dependent sample’s cluster proportions were normalized with the whole SCG’s cluster proportions. For MERFISH analysis, CTB+ cells make up a subset of all cells captured in each sample. Same-projection samples were pooled and analysis was run on the cluster proportions of the pool of CTB+ cells and normalized by the pool of all cells in the samples. Projection-dependent cluster proportions were tested for significant deviation from control (whole SCG) proportions using the chi-squared test on counts of cells in clusters.

Supplementary Material

NIHMS2190723-supplement-1.pdf^{(4.1MB, pdf)}

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2026.117446.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
rabbit anti-NPY	Peninsula Laboratories	T-4070; RRID:AB_518504
sheep anti-TH	Novus Biologicals	NB300-110; RRID:AB_10002491
rabbit anti-TH	Millipore Sigma	AB152; RRID:AB_390204
rat anti-CD31	BD Pharmingen	550274; RRID:AB_393571
goat anti-CTB	List Labs	#703; RRID:AB_10013220
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 555	Thermo Fisher Scientific	A-31572; RRID:AB_162543
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 488	Thermo Fisher Scientific	A-21206; RRID:AB_2535792
Donkey anti-Rat IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 488	Thermo Fisher Scientific	A-21208; RRID:AB_2535794
Donkey anti-Sheep IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor^™ 647	Thermo Fisher Scientific	A-21448; RRID:AB_2535865
Chemicals, peptides, and recombinant proteins
Paraformaldehyde 20% Solution, EM Grade (PFA)	Electron Microscopy Sciences	15713
Cholera Toxin Subunit B, Alexa Fluor 647 conjugated	Invitrogen	Cat#C34778
Cholera Toxin Subunit B, Alexa Fluor 555 conjugated	Invitrogen	Cat#C34776
Cholera Toxin Subunit B, Alexa Fluor 488 conjugated	Invitrogen	Cat#C34775
Cholera Toxin B Subunit	List Labs	#104
Normal Horse Serum Blocking Solution	Vector Laboratories	S-2000-20
DAPI	Invitrogen	D1306
NeuroTrace	Invitrogen	N21483
Triton-X	Fisher Scientific	BP151-500
Glycerol	Fisher Scientific	BP229-1
Bovine Serum Albumin (BSA)	Sigma-Aldrich	A3059-50G
Collagenase IV	Worthington	LS004188
Critical commercial assays
Papain Dissociation System	Worthington Biochemical Corporation	LK003150
Chromium Next GEM Single Cell 3′ Kit v3.1	10x Genomics	1000269
RNAscope v1	Advanced Cell Diagnostics	320851
MERSCOPE Sample Prep Kit	Vizgen	10400012
MERSCOPE 140 Gene Panel	Vizgen	10400001
MERSCOPE 140 Gene Imaging Kit	Vizgen	10400004
MERSCOPE Cell Boundary Stain Kit	Vizgen	10400118
MERSCOPE Anti-Goat Protein Stain Kit	Vizgen	10400108
Deposited data
MERFISH datasets	This paper	GEO: GSE325797
scRNAseq datasets	This paper	GEO: GSE325798
Experimental models: organisms/strains
Isl1 – Cre mouse line	The Jackson Laboratory	024242
Adult mice mixed background	N/A	N/A
Ai14(RCL-tdT)-D	The Jackson Laboratory	007914
Oligonucleotides
Areg, RNAscope Probe	Advanced Cell Diagnostics	430501
Sctr, RNAscope Probe	Advanced Cell Diagnostics	508511
Software and algorithms
Cell Ranger	10x Genomics	N/A
R	The R project	https://www.r-project.org/
FIJI	NIH ImageJ software	https://fiji.sc/
MATLAB	MathWorks	https://www.mathworks.com
BD FACSDiva^™ Software v9.0	BD Biosciences	https://www.bdbiosciences.com/en-us/products/software/instrument-software/bd-facsdiva-software
MERSCOPE visualizer	Vizgen	https://vizgen.com/vizualizer-software/
Cell Pose 2	Nature Methods	Pachitariu and Stringe³⁴ 2022
Prism	GraphPad	https://www.graphpad.com
Other
Superfrost Plus Microscope Slides	Fisher Scientific	12-550-15
Tissue-Tek O.C.T. Compound	Sakura	4583
HBSS buffer	Gibco	14175-095
DMEM:F12	Gibco	11039-021
FBS serum	Gibco	A5256701
Falcon 40 μm Cell Strainer	Corning	352340
Small Animal Stereotaxic Instrument	Kopf Instruments	Model 962
Permahand Silk Suture	Ethicon	786G
Microliter Syringe	Hamilton	7632-01
Micro Pipettes (Borosilicate glass)	VWR	53432-706
Disposable Pasteur Pipets	Fisher Scientific	K883350-0009
Confocal microscope	Olympus	FV3000
MERSCOPE imager	Vizgen	N/A
HiSeq 4000 System	Illumina	N/A

Open in a new tab

Highlights.

Individual superior cervical ganglion (SCG) neurons innervate single sympathetic effectors
Populations of SCG neuron defined by their projection targets display pseudo-topography
Single cell transcriptomics reveal distinct SCG neuron types with rostral-caudal bias
Transcriptomic subtypes of SCG neurons do not map one-to-one onto projection types

ACKNOWLEDGMENTS

S.L.P. is the Benjamin H. Lewis Chair in Neuroscience. This research was supportedby the National Institute of Neurological Disorders and Stroke (NINDS) United States (grant nos. R01 NS123160 and R01 NS135055), the Sol Goldman Charitable Trust United States, and the NGS Core Facility (Salk) with funding from NIH-NCI CCSG United States: P30 014195. Carolina Thorn-Perez assisted with preparation of the Graphical Abstract.

Footnotes

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Samuel L. Pfaff (pfaff@ salk.edu).

Materials availability

Mouse lines used in this study are available from the lead contact with a completed materials transfer agreement.

Data and code availability

MERFISH dataset has been deposited at NCBI Gene Expression Omnibus (GEO) as GEO: GSE325797 and is publicly available as of the date of publication. scRNAseq datasets have been deposited at NCBI GEO as GEO: GSE325798 and are publicly available as of the date of publication.
This paper does not report original code.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

DECLARATION OF GENERATIVE AI AND AI-ASSISTED TECHNOLOGIES IN THE WRITING PROCESS

During the preparation of this manuscript, the authors used ChatGPT in order to revise documents for typos, grammatical mistakes, and improve clarity and flow. After using these tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

DECLARATION OF INTERESTS

The authors declare no competing interests.

REFERENCES

1.Jänig W. (2022). The Integrative Action of the Autonomic Nervous System : Neurobiology of Homeostasis, Second Edition (Edition: Cambridge University Press; ). [Google Scholar]
2.Cannon WB (1929). ORGANIZATION FOR PHYSIOLOGICAL HOMEOSTASIS. Physiol. Rev 9, 399–431. 10.1152/physrev.1929.9.3.399. [DOI] [Google Scholar]
3.Selye H. (1936). A Syndrome produced by Diverse Nocuous Agents. Nature 138, 32. 10.1038/138032a0. [DOI] [PubMed] [Google Scholar]
4.Wang T, Tufenkjian A, Ajijola OA, and Oka Y (2025). Molecular and functional diversity of the autonomic nervous system. Nat. Rev. Neurosci 26, 607–622. 10.1038/s41583-025-00941-2 [DOI] [PubMed] [Google Scholar]
5.Saper CB (2002). The central autonomic nervous system: conscious visceral perception and autonomic pattern generation. Annu. Rev. Neurosci 25, 433–469. 10.1146/annurev.neuro.25.032502.111311. [DOI] [PubMed] [Google Scholar]
6.Morrison SF (2001). Differential control of sympathetic outflow. Am. J. Physiol. Regul. Integr. Comp. Physiol 281, R683–R698. 10.1152/ajpregu.2001.281.3.R683. [DOI] [PubMed] [Google Scholar]
7.Jänig W, and McLachlan EM (1992). Specialized functional pathways are the building blocks of the autonomic nervous system. J. Auton. Nerv. Syst 41, 3–13. 10.1016/0165-1838(92)90121-v. [DOI] [PubMed] [Google Scholar]
8.Jänig W, and McLachlan EM (1992). Characteristics of function-specific pathways in the sympathetic nervous system. Trends Neurosci. 15, 475–481. 10.1016/0166-2236(92)90092-m. [DOI] [PubMed] [Google Scholar]
9.Jänig W. (1996). Spinal cord reflex organization of sympathetic systems. Prog. Brain Res 107, 43–77. 10.1016/s0079-6123(08)61858-0. [DOI] [PubMed] [Google Scholar]
10.Boczek-Funcke A, Dembowsky K, Häbler HJ, Jänig W, McAllen RM, and Michaelis M (1992). Classification of preganglionic neurones projecting into the cat cervical sympathetic trunk. J. Physiol 453, 319–339. 10.1113/jphysiol.1992.sp019231. [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Njå A, and Purves D (1977). Specific innervation of guinea-pig superior cervical ganglion cells by preganglionic fibres arising from different levels of the spinal cord. J. Physiol 264, 565–583. 10.1113/jphysiol.1977.sp011683. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Langley JN (1892). On the Origin from the Spinal Cord of the Cervical and Upper Thoracic Sympathetic Fibres, with Some Observations on White and Grey Rami Communicantes. Phil. Trans. Roy. Soc. Lond. B 183. xviii–124. [Google Scholar]
13.Ekman P, Levenson RW, and Friesen WV (1983). Autonomic nervous system activity distinguishes among emotions. Science 221, 1208–1210. 10.1126/science.6612338. [DOI] [PubMed] [Google Scholar]
14.Alaynick WA, Jessell TM, and Pfaff SL (2011). SnapShot: spinal cord development. Cell 146, 178–178.e1. 10.1016/j.cell.2011.06.038. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Dasen JS, and Jessell TM (2009). Hox networks and the origins of motor neuron diversity. Curr. Top. Dev. Biol 88, 169–200. 10.1016/S0070-2153(09)88006-X. [DOI] [PubMed] [Google Scholar]
16.Scott-Solomon E, Boehm E, and Kuruvilla R (2021). The sympathetic nervous system in development and disease. Nat. Rev. Neurosci 22, 685–702. 10.1038/s41583-021-00523-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Schotzinger RJ, and Landis SC (1990). Acquisition of cholinergic and peptidergic properties by sympathetic innervation of rat sweat glands requires interaction with normal target. Neuron 5, 91–100. 10.1016/0896-6273(90)90037-g. [DOI] [PubMed] [Google Scholar]
18.Luebke JI, and Wright LL (1992). Characterization of superior cervical ganglion neurons that project to the submandibular glands, the eyes, and the pineal gland in rats. Brain Res. 589, 1–14. 10.1016/0006-8993(92)91155-8. [DOI] [PubMed] [Google Scholar]
19.Lahtivirta S, Koistinaho J, and Hervonen A (1995). A subpopulation of large neurons of the sympathetic superior cervical ganglion innervates the NGF-rich submandibular salivary gland in young adult and aged mice. J. Auton. Nerv. Syst 50, 283–289. 10.1016/0165-1838(94)00099-6. [DOI] [PubMed] [Google Scholar]
20.Gibbins IL (1991). Vasomotor, pilomotor and secretomotor neurons distinguished by size and neuropeptide content in superior cervical ganglia of mice. J. Auton. Nerv. Syst 34, 171–183. 10.1016/0165-1838(91)90083-f. [DOI] [PubMed] [Google Scholar]
21.Ekblad E, Edvinsson L, Wahlestedt C, Uddman R, Håkanson R, and Sundler F (1984). Neuropeptide Y co-exists and co-operates with noradrenaline in perivascular nerve fibers. Regul. Pept 8, 225–235. 10.1016/0167-0115(84)90064-8. [DOI] [PubMed] [Google Scholar]
22.Furlan A, La Manno G, Lübke M, Häring M, Abdo H, Hochgerner H, Kupari J, Usoskin D, Airaksinen MS, Oliver G, et al. (2016). Visceral motor neuron diversity delineates a cellular basis for nipple- and pilo-erection muscle control. Nat. Neurosci 19, 1331–1340. 10.1038/nn.4376. [DOI] [PubMed] [Google Scholar]
23.Wang T, Teng B, Yao DR, Gao W, and Oka Y (2025). Organ-specific sympathetic innervation defines visceral functions. Nature 637, 895–902. 10.1038/s41586-024-08269-0. [DOI] [PubMed] [Google Scholar]
24.Ziegler KA, Ahles A, Dueck A, Esfandyari D, Pichler P, Weber K, Kotschi S, Bartelt A, Sinicina I, Graw M, et al. (2023). Immune-mediated denervation of the pineal gland underlies sleep disturbance in cardiac disease. Science 381, 285–290. 10.1126/science.abn6366. [DOI] [PubMed] [Google Scholar]
25.Mapps AA, Thomsen MB, Boehm E, Zhao H, Hattar S, and Kuruvilla R (2022). Diversity of satellite glia in sympathetic and sensory ganglia. Cell Rep. 38, 110328. 10.1016/j.celrep.2022.110328. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Lichtman JW, Purves D, and Yip JW (1979). On the purpose of selective innervation of guinea-pig superior cervical ganglion cells. J. Physiol 292, 69–84. 10.1113/jphysiol.1979.sp012839. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Kumari R, Pascalau R, Wang H, Bajpayi S, Yurgel M, Quansah K, Hattar S, Tampakakis E, and Kuruvilla R (2024). Sympathetic NPY controls glucose homeostasis, cold tolerance, and cardiovascular functions in mice. Cell Rep. 43, 113674. 10.1016/j.celrep.2024.113674. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Zhu Y, Yao L, Gallo-Ferraz AL, Bombassaro B, Simões MR, Abe I, Chen J, Sarker G, Ciccarelli A, Zhou L, et al. (2024). Sympathetic neuropeptide Y protects from obesity by sustaining thermogenic fat. Nature 634, 243–250. 10.1038/s41586-024-07863-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Qi L, Iskols M, Shi D, Reddy P, Walker C, Lezgiyeva K, Voisin T, Pawlak M, Kuchroo VK, Chiu IM, et al. (2024). A mouse DRG genetic toolkit reveals morphological and physiological diversity of somatosensory neuron subtypes. Cell 187, 1508–1526.e16. 10.1016/j.cell.2024.02.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Liau ES, Jin S, Chen YC, Liu WS, Calon M, Nedelec S, Nie Q, and Chen JA (2023). Single-cell transcriptomic analysis reveals diversity within mammalian spinal motor neurons. Nat. Commun 14, 46. 10.1038/s41467-022-35574-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Blum JA, Klemm S, Shadrach JL, Guttenplan KA, Nakayama L, Kathiria A, Hoang PT, Gautier O, Kaltschmidt JA, Greenleaf WJ, and Gitler AD (2021). Single-cell transcriptomic analysis of the adult mouse spinal cord reveals molecular diversity of autonomic and skeletal motor neurons. Nat. Neurosci 24, 572–583. 10.1038/s41593-020-00795-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Chen KH, Boettiger AN, Moffitt JR, Wang S, and Zhuang X (2015). RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090. 10.1126/science.aaa6090. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Osseward PJ 2nd, Amin ND, Moore JD, Temple BA, Barriga BK, Bachmann LC, Beltran F Jr., Gullo M, Clark RC, Driscoll SP, et al. (2021). Conserved genetic signatures parcellate cardinal spinal neuron classes into local and projection subsets. Science 372, 385–393. 10.1126/science.abe0690. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Pachitariu M, and Stringer C (2022). Cellpose 2.0: how to train your own model. Nat. Methods 19, 1634–1641. 10.1038/s41592-022-01663-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Jansen AS, Nguyen XV, Karpitskiy V, Mettenleiter TC, and Loewy AD (1995). Central command neurons of the sympathetic nervous system: basis of the fight-or-flight response. Science 270, 644–646. 10.1126/science.270.5236.644. [DOI] [PubMed] [Google Scholar]
36.Levine AJ, Hinckley CA, Hilde KL, Driscoll SP, Poon TH, Montgomery JM, and Pfaff SL (2014). Identification of a cellular node for motor control pathways. Nat. Neurosci 17, 586–593. 10.1038/nn.3675. [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Strack AM, Sawyer WB, Hughes JH, Platt KB, and Loewy AD (1989). A general pattern of CNS innervation of the sympathetic outflow demonstrated by transneuronal pseudorabies viral infections. Brain Res. 491, 156–162. 10.1016/0006-8993(89)90098-x. [DOI] [PubMed] [Google Scholar]
38.Sharma N, Flaherty K, Lezgiyeva K, Wagner DE, Klein AM, and Ginty DD (2020). The emergence of transcriptional identity in somatosensory neurons. Nature 577, 392–398. 10.1038/s41586-019-1900-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Gonsalvez DG, Cane KN, Landman KA, Enomoto H, Young HM, and Anderson CR (2013). Proliferation and cell cycle dynamics in the developing stellate ganglion. J. Neurosci 33, 5969–5979. 10.1523/JNEUROSCI.4350-12.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Makita T, Sucov HM, Gariepy CE, Yanagisawa M, and Ginty DD (2008). Endothelins are vascular-derived axonal guidance cues for developing sympathetic neurons. Nature 452, 759–763. 10.1038/nature06859. [DOI] [PMC free article] [PubMed] [Google Scholar]
41.Schotzinger RJ, and Landis SC (1988). Cholinergic phenotype developed by noradrenergic sympathetic neurons after innervation of a novel cholinergic target in vivo. Nature 335, 637–639. 10.1038/335637a0. [DOI] [PubMed] [Google Scholar]
42.Glebova NO, and Ginty DD (2004). Heterogeneous requirement of NGF for sympathetic target innervation in vivo. J. Neurosci. 24, 743–751. 10.1523/JNEUROSCI.4523-03.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Njå A, and Purves D (1978). The effects of nerve growth factor and its antiserum on synapses in the superior cervical ganglion of the guineapig. J. Physiol 277, 53–75. [PMC free article] [PubMed] [Google Scholar]
44.Sharma N, Deppmann CD, Harrington AW, St Hillaire C, Chen ZY, Lee FS, and Ginty DD (2010). Long-distance control of synapse assembly by target-derived NGF. Neuron 67, 422–434. 10.1016/j.neuron.2010.07.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Sürmeli G, Akay T, Ippolito GC, Tucker PW, and Jessell TM (2011). Patterns of spinal sensory-motor connectivity prescribed by a dorsoventral positional template. Cell 147, 653–665. 10.1016/j.cell.2011.10.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Amin ND, Senturk G, Costaguta G, Driscoll S, O’Leary B, Bonanomi D, and Pfaff SL (2021). A hidden threshold in motor neuron gene networks revealed by modulation of miR-218 dose. Neuron 109, 3252–3267.e6. 10.1016/j.neuron.2021.07.028. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Erickson AG, Motta A, Kastriti ME, Edwards S, Coulpier F, Théoulle E, Murtazina A, Poverennaya I, Wies D, Ganofsky J, et al. (2024). Motor innervation directs the correct development of the mouse sympathetic nervous system. Nat. Commun 15, 7065. 10.1038/s41467-024-51290-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Rubin E. (1985). Development of the rat superior cervical ganglion: ganglion cell maturation. J. Neurosci 5, 673–684. 10.1523/JNEUROSCI.05-03-00673.1985. [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Kameda Y. (2014). Signaling molecules and transcription factors involved in the development of the sympathetic nervous system, with special emphasis on the superior cervical ganglion. Cell Tissue Res. 357, 527–548. 10.1007/s00441-014-1847-3. [DOI] [PubMed] [Google Scholar]
50.Lai HC, Seal RP, and Johnson JE (2016). Making sense out of spinal cord somatosensory development. Development 143, 3434–3448. 10.1242/dev.139592. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Hayashi M, Hinckley CA, Driscoll SP, Moore NJ, Levine AJ, Hilde KL, Sharma K, and Pfaff SL (2018). Graded Arrays of Spinal and Supraspinal V2a Interneuron Subtypes Underlie Forelimb and Hindlimb Motor Control. Neuron 97, 869–884.e5. 10.1016/j.neuron.2018.01.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Yang L, Cai CL, Lin L, Qyang Y, Chung C, Monteiro RM, Mummery CL, Fishman GI, Cogen A, and Evans S (2006). Isl1Cre reveals a common Bmp pathway in heart and limb development. Development 133, 1575–1585. 10.1242/dev.02322. [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Savastano LE, Castro AE, Fitt MR, Rath MF, Romeo HE, and Muñoz EM (2010). A standardized surgical technique for rat superior cervical ganglionectomy. J. Neurosci. Methods 192, 22–33. 10.1016/j.jneumeth.2010.07.007. [DOI] [PubMed] [Google Scholar]
54.Ilicic T, Kim JK, Kolodziejczyk AA, Bagger FO, McCarthy DJ, Marioni JC, and Teichmann SA (2016). Classification of low quality cells from single-cell RNA-seq data. Genome Biol. 17, 29. 10.1186/s13059-016-0888-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

NIHMS2190723-supplement-1.pdf^{(4.1MB, pdf)}

[R1] 1.Jänig W. (2022). The Integrative Action of the Autonomic Nervous System : Neurobiology of Homeostasis, Second Edition (Edition: Cambridge University Press; ). [Google Scholar]

[R2] 2.Cannon WB (1929). ORGANIZATION FOR PHYSIOLOGICAL HOMEOSTASIS. Physiol. Rev 9, 399–431. 10.1152/physrev.1929.9.3.399. [DOI] [Google Scholar]

[R3] 3.Selye H. (1936). A Syndrome produced by Diverse Nocuous Agents. Nature 138, 32. 10.1038/138032a0. [DOI] [PubMed] [Google Scholar]

[R4] 4.Wang T, Tufenkjian A, Ajijola OA, and Oka Y (2025). Molecular and functional diversity of the autonomic nervous system. Nat. Rev. Neurosci 26, 607–622. 10.1038/s41583-025-00941-2 [DOI] [PubMed] [Google Scholar]

[R5] 5.Saper CB (2002). The central autonomic nervous system: conscious visceral perception and autonomic pattern generation. Annu. Rev. Neurosci 25, 433–469. 10.1146/annurev.neuro.25.032502.111311. [DOI] [PubMed] [Google Scholar]

[R6] 6.Morrison SF (2001). Differential control of sympathetic outflow. Am. J. Physiol. Regul. Integr. Comp. Physiol 281, R683–R698. 10.1152/ajpregu.2001.281.3.R683. [DOI] [PubMed] [Google Scholar]

[R7] 7.Jänig W, and McLachlan EM (1992). Specialized functional pathways are the building blocks of the autonomic nervous system. J. Auton. Nerv. Syst 41, 3–13. 10.1016/0165-1838(92)90121-v. [DOI] [PubMed] [Google Scholar]

[R8] 8.Jänig W, and McLachlan EM (1992). Characteristics of function-specific pathways in the sympathetic nervous system. Trends Neurosci. 15, 475–481. 10.1016/0166-2236(92)90092-m. [DOI] [PubMed] [Google Scholar]

[R9] 9.Jänig W. (1996). Spinal cord reflex organization of sympathetic systems. Prog. Brain Res 107, 43–77. 10.1016/s0079-6123(08)61858-0. [DOI] [PubMed] [Google Scholar]

[R10] 10.Boczek-Funcke A, Dembowsky K, Häbler HJ, Jänig W, McAllen RM, and Michaelis M (1992). Classification of preganglionic neurones projecting into the cat cervical sympathetic trunk. J. Physiol 453, 319–339. 10.1113/jphysiol.1992.sp019231. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Njå A, and Purves D (1977). Specific innervation of guinea-pig superior cervical ganglion cells by preganglionic fibres arising from different levels of the spinal cord. J. Physiol 264, 565–583. 10.1113/jphysiol.1977.sp011683. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R12] 12.Langley JN (1892). On the Origin from the Spinal Cord of the Cervical and Upper Thoracic Sympathetic Fibres, with Some Observations on White and Grey Rami Communicantes. Phil. Trans. Roy. Soc. Lond. B 183. xviii–124. [Google Scholar]

[R13] 13.Ekman P, Levenson RW, and Friesen WV (1983). Autonomic nervous system activity distinguishes among emotions. Science 221, 1208–1210. 10.1126/science.6612338. [DOI] [PubMed] [Google Scholar]

[R14] 14.Alaynick WA, Jessell TM, and Pfaff SL (2011). SnapShot: spinal cord development. Cell 146, 178–178.e1. 10.1016/j.cell.2011.06.038. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] 15.Dasen JS, and Jessell TM (2009). Hox networks and the origins of motor neuron diversity. Curr. Top. Dev. Biol 88, 169–200. 10.1016/S0070-2153(09)88006-X. [DOI] [PubMed] [Google Scholar]

[R16] 16.Scott-Solomon E, Boehm E, and Kuruvilla R (2021). The sympathetic nervous system in development and disease. Nat. Rev. Neurosci 22, 685–702. 10.1038/s41583-021-00523-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Schotzinger RJ, and Landis SC (1990). Acquisition of cholinergic and peptidergic properties by sympathetic innervation of rat sweat glands requires interaction with normal target. Neuron 5, 91–100. 10.1016/0896-6273(90)90037-g. [DOI] [PubMed] [Google Scholar]

[R18] 18.Luebke JI, and Wright LL (1992). Characterization of superior cervical ganglion neurons that project to the submandibular glands, the eyes, and the pineal gland in rats. Brain Res. 589, 1–14. 10.1016/0006-8993(92)91155-8. [DOI] [PubMed] [Google Scholar]

[R19] 19.Lahtivirta S, Koistinaho J, and Hervonen A (1995). A subpopulation of large neurons of the sympathetic superior cervical ganglion innervates the NGF-rich submandibular salivary gland in young adult and aged mice. J. Auton. Nerv. Syst 50, 283–289. 10.1016/0165-1838(94)00099-6. [DOI] [PubMed] [Google Scholar]

[R20] 20.Gibbins IL (1991). Vasomotor, pilomotor and secretomotor neurons distinguished by size and neuropeptide content in superior cervical ganglia of mice. J. Auton. Nerv. Syst 34, 171–183. 10.1016/0165-1838(91)90083-f. [DOI] [PubMed] [Google Scholar]

[R21] 21.Ekblad E, Edvinsson L, Wahlestedt C, Uddman R, Håkanson R, and Sundler F (1984). Neuropeptide Y co-exists and co-operates with noradrenaline in perivascular nerve fibers. Regul. Pept 8, 225–235. 10.1016/0167-0115(84)90064-8. [DOI] [PubMed] [Google Scholar]

[R22] 22.Furlan A, La Manno G, Lübke M, Häring M, Abdo H, Hochgerner H, Kupari J, Usoskin D, Airaksinen MS, Oliver G, et al. (2016). Visceral motor neuron diversity delineates a cellular basis for nipple- and pilo-erection muscle control. Nat. Neurosci 19, 1331–1340. 10.1038/nn.4376. [DOI] [PubMed] [Google Scholar]

[R23] 23.Wang T, Teng B, Yao DR, Gao W, and Oka Y (2025). Organ-specific sympathetic innervation defines visceral functions. Nature 637, 895–902. 10.1038/s41586-024-08269-0. [DOI] [PubMed] [Google Scholar]

[R24] 24.Ziegler KA, Ahles A, Dueck A, Esfandyari D, Pichler P, Weber K, Kotschi S, Bartelt A, Sinicina I, Graw M, et al. (2023). Immune-mediated denervation of the pineal gland underlies sleep disturbance in cardiac disease. Science 381, 285–290. 10.1126/science.abn6366. [DOI] [PubMed] [Google Scholar]

[R25] 25.Mapps AA, Thomsen MB, Boehm E, Zhao H, Hattar S, and Kuruvilla R (2022). Diversity of satellite glia in sympathetic and sensory ganglia. Cell Rep. 38, 110328. 10.1016/j.celrep.2022.110328. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] 26.Lichtman JW, Purves D, and Yip JW (1979). On the purpose of selective innervation of guinea-pig superior cervical ganglion cells. J. Physiol 292, 69–84. 10.1113/jphysiol.1979.sp012839. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Kumari R, Pascalau R, Wang H, Bajpayi S, Yurgel M, Quansah K, Hattar S, Tampakakis E, and Kuruvilla R (2024). Sympathetic NPY controls glucose homeostasis, cold tolerance, and cardiovascular functions in mice. Cell Rep. 43, 113674. 10.1016/j.celrep.2024.113674. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] 28.Zhu Y, Yao L, Gallo-Ferraz AL, Bombassaro B, Simões MR, Abe I, Chen J, Sarker G, Ciccarelli A, Zhou L, et al. (2024). Sympathetic neuropeptide Y protects from obesity by sustaining thermogenic fat. Nature 634, 243–250. 10.1038/s41586-024-07863-6. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Qi L, Iskols M, Shi D, Reddy P, Walker C, Lezgiyeva K, Voisin T, Pawlak M, Kuchroo VK, Chiu IM, et al. (2024). A mouse DRG genetic toolkit reveals morphological and physiological diversity of somatosensory neuron subtypes. Cell 187, 1508–1526.e16. 10.1016/j.cell.2024.02.006. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Liau ES, Jin S, Chen YC, Liu WS, Calon M, Nedelec S, Nie Q, and Chen JA (2023). Single-cell transcriptomic analysis reveals diversity within mammalian spinal motor neurons. Nat. Commun 14, 46. 10.1038/s41467-022-35574-x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R31] 31.Blum JA, Klemm S, Shadrach JL, Guttenplan KA, Nakayama L, Kathiria A, Hoang PT, Gautier O, Kaltschmidt JA, Greenleaf WJ, and Gitler AD (2021). Single-cell transcriptomic analysis of the adult mouse spinal cord reveals molecular diversity of autonomic and skeletal motor neurons. Nat. Neurosci 24, 572–583. 10.1038/s41593-020-00795-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R32] 32.Chen KH, Boettiger AN, Moffitt JR, Wang S, and Zhuang X (2015). RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science 348, aaa6090. 10.1126/science.aaa6090. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Osseward PJ 2nd, Amin ND, Moore JD, Temple BA, Barriga BK, Bachmann LC, Beltran F Jr., Gullo M, Clark RC, Driscoll SP, et al. (2021). Conserved genetic signatures parcellate cardinal spinal neuron classes into local and projection subsets. Science 372, 385–393. 10.1126/science.abe0690. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R34] 34.Pachitariu M, and Stringer C (2022). Cellpose 2.0: how to train your own model. Nat. Methods 19, 1634–1641. 10.1038/s41592-022-01663-4. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] 35.Jansen AS, Nguyen XV, Karpitskiy V, Mettenleiter TC, and Loewy AD (1995). Central command neurons of the sympathetic nervous system: basis of the fight-or-flight response. Science 270, 644–646. 10.1126/science.270.5236.644. [DOI] [PubMed] [Google Scholar]

[R36] 36.Levine AJ, Hinckley CA, Hilde KL, Driscoll SP, Poon TH, Montgomery JM, and Pfaff SL (2014). Identification of a cellular node for motor control pathways. Nat. Neurosci 17, 586–593. 10.1038/nn.3675. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R37] 37.Strack AM, Sawyer WB, Hughes JH, Platt KB, and Loewy AD (1989). A general pattern of CNS innervation of the sympathetic outflow demonstrated by transneuronal pseudorabies viral infections. Brain Res. 491, 156–162. 10.1016/0006-8993(89)90098-x. [DOI] [PubMed] [Google Scholar]

[R38] 38.Sharma N, Flaherty K, Lezgiyeva K, Wagner DE, Klein AM, and Ginty DD (2020). The emergence of transcriptional identity in somatosensory neurons. Nature 577, 392–398. 10.1038/s41586-019-1900-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R39] 39.Gonsalvez DG, Cane KN, Landman KA, Enomoto H, Young HM, and Anderson CR (2013). Proliferation and cell cycle dynamics in the developing stellate ganglion. J. Neurosci 33, 5969–5979. 10.1523/JNEUROSCI.4350-12.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] 40.Makita T, Sucov HM, Gariepy CE, Yanagisawa M, and Ginty DD (2008). Endothelins are vascular-derived axonal guidance cues for developing sympathetic neurons. Nature 452, 759–763. 10.1038/nature06859. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R41] 41.Schotzinger RJ, and Landis SC (1988). Cholinergic phenotype developed by noradrenergic sympathetic neurons after innervation of a novel cholinergic target in vivo. Nature 335, 637–639. 10.1038/335637a0. [DOI] [PubMed] [Google Scholar]

[R42] 42.Glebova NO, and Ginty DD (2004). Heterogeneous requirement of NGF for sympathetic target innervation in vivo. J. Neurosci. 24, 743–751. 10.1523/JNEUROSCI.4523-03.2004. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R43] 43.Njå A, and Purves D (1978). The effects of nerve growth factor and its antiserum on synapses in the superior cervical ganglion of the guineapig. J. Physiol 277, 53–75. [PMC free article] [PubMed] [Google Scholar]

[R44] 44.Sharma N, Deppmann CD, Harrington AW, St Hillaire C, Chen ZY, Lee FS, and Ginty DD (2010). Long-distance control of synapse assembly by target-derived NGF. Neuron 67, 422–434. 10.1016/j.neuron.2010.07.018. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] 45.Sürmeli G, Akay T, Ippolito GC, Tucker PW, and Jessell TM (2011). Patterns of spinal sensory-motor connectivity prescribed by a dorsoventral positional template. Cell 147, 653–665. 10.1016/j.cell.2011.10.012. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R46] 46.Amin ND, Senturk G, Costaguta G, Driscoll S, O’Leary B, Bonanomi D, and Pfaff SL (2021). A hidden threshold in motor neuron gene networks revealed by modulation of miR-218 dose. Neuron 109, 3252–3267.e6. 10.1016/j.neuron.2021.07.028. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R47] 47.Erickson AG, Motta A, Kastriti ME, Edwards S, Coulpier F, Théoulle E, Murtazina A, Poverennaya I, Wies D, Ganofsky J, et al. (2024). Motor innervation directs the correct development of the mouse sympathetic nervous system. Nat. Commun 15, 7065. 10.1038/s41467-024-51290-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] 48.Rubin E. (1985). Development of the rat superior cervical ganglion: ganglion cell maturation. J. Neurosci 5, 673–684. 10.1523/JNEUROSCI.05-03-00673.1985. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R49] 49.Kameda Y. (2014). Signaling molecules and transcription factors involved in the development of the sympathetic nervous system, with special emphasis on the superior cervical ganglion. Cell Tissue Res. 357, 527–548. 10.1007/s00441-014-1847-3. [DOI] [PubMed] [Google Scholar]

[R50] 50.Lai HC, Seal RP, and Johnson JE (2016). Making sense out of spinal cord somatosensory development. Development 143, 3434–3448. 10.1242/dev.139592. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R51] 51.Hayashi M, Hinckley CA, Driscoll SP, Moore NJ, Levine AJ, Hilde KL, Sharma K, and Pfaff SL (2018). Graded Arrays of Spinal and Supraspinal V2a Interneuron Subtypes Underlie Forelimb and Hindlimb Motor Control. Neuron 97, 869–884.e5. 10.1016/j.neuron.2018.01.023. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R52] 52.Yang L, Cai CL, Lin L, Qyang Y, Chung C, Monteiro RM, Mummery CL, Fishman GI, Cogen A, and Evans S (2006). Isl1Cre reveals a common Bmp pathway in heart and limb development. Development 133, 1575–1585. 10.1242/dev.02322. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R53] 53.Savastano LE, Castro AE, Fitt MR, Rath MF, Romeo HE, and Muñoz EM (2010). A standardized surgical technique for rat superior cervical ganglionectomy. J. Neurosci. Methods 192, 22–33. 10.1016/j.jneumeth.2010.07.007. [DOI] [PubMed] [Google Scholar]

[R54] 54.Ilicic T, Kim JK, Kolodziejczyk AA, Bagger FO, McCarthy DJ, Marioni JC, and Teichmann SA (2016). Classification of low quality cells from single-cell RNA-seq data. Genome Biol. 17, 29. 10.1186/s13059-016-0888-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

Circuit organization and transcriptomic heterogeneity of sympathetic circuits innervating cranial structures

Gokhan Senturk

Shawn Driscoll

Peter J Osseward II

Miriam Gullo

Karen Lettieri

Samuel L Pfaff

SUMMARY

Graphical Abstract

In brief

INTRODUCTION

Figure 1. SCG neurons innervate single effector organ.

RESULTS

SCG neuron projections

The spatial organization of SCG P-types

Figure 2. Compartmentalization of P-types within the SCG without strict topography.

NPY marks a subset of SCG neurons

Figure 3. NPY profile of SCG nerve fibers.

Transcriptomic heterogeneity of SCG neurons

Figure 4. Transcriptomic subtypes of P14 SCG neurons and their spatial organizations.

Neurochemical profiles of SCG transcriptomic clusters

Spatial distribution of SCG transcriptomic clusters

Figure 5. Divergent transcriptional programs drive the postnatal emergence of mature SCG T-types.

Maturation of SCG T-types postnatally

Relationship between T- and P-subtypes

Figure 6. Differential enrichment of T-types across P-types without strict one-to-one correspondence in P14 SCG.

DISCUSSION

SCG target innervation

Postnatal diversification of SCG subtypes

Spatial organization and input formation of SCG neurons

SCG transcriptomics and circuit logic

Limitations of the study

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

METHOD DETAILS

General surgical considerations

Cholera toxin subunit b (CTB) injections

Tissue preparation, staining and imaging

RNAscope

Sample preparation for 10x scRNAseq

Sample preparation for MERFISH

QUANTIFICATION AND STATISTICAL ANALYSIS

CTB detection, contour plots and colocalization

Adult versus pup NPY+ SCG cells

scRNAseq preprocessing and QC

scRNAseq cell type calls and clustering

Cross-validation of clustering solutions

Calculation of classification metrics

scRNAseq cluster correlogram

Differential gene expression testing

Distinctiveness metrics

Marker identification

Combinatorial marker sets

Npy and neurochemical expression analysis

P0 and P14 scRNAseq integration and clustering

Mapping P14 SCG clusters to projections datasets

MERFISH gene-panel and data analysis

MERFISH QC and neuron identification

Mapping P14 SCG clusters to MERFISH

MERFISH versus scRNAseq P14 cluster heatmap

MERFISH cluster spatial patterns

Spatial distribution of genes in MERFISH

Identification of CTB+ cells in MERFISH

Cluster proportion barplots

Fold enrichment of proportions barplots

Supplementary Material

Highlights.

ACKNOWLEDGMENTS

Footnotes

REFERENCES

Associated Data

Supplementary Materials

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases