Research Article

Multipotent peripheral glial cells generate neuroendocrine cells of the adrenal medulla

See allHide authors and affiliations

Science  07 Jul 2017:
Vol. 357, Issue 6346, eaal3753
DOI: 10.1126/science.aal3753

Following the yellow brick road

The adrenal glands affect a variety of processes such as stress responses and metabolism. The mature adrenal gland is formed from multiple tissue sources, including cells of neural origin. Furlan et al. traced the origins of these cells. The cells first become Schwann cell precursors and follow along nerves to travel from the dorsal root ganglia of the spine to the adrenal gland. Once there, the cells differentiate into chromaffin cells. The authors used singlecell transcriptomics to reveal the shifts in functional programs during migration, development, and differentiation.

Science, this issue p. eaal3753

Structured Abstract


Circulating adrenaline can have profound effects on the body’s “inner world,” adjusting levels depending on demand to maintain organ and bodily homeostasis during daily living. In the more extreme fight-or-flight response, the surge of adrenaline is “energizing” through effects on organs and tissues, including increased heart rate and blood glucose levels, and redirecting oxygen and glucose to limb muscles. Chromaffin cells located in the adrenal medulla constitute the main hormonal component of the autonomic nervous system and are the principal source for release of catecholamines, including adrenaline, in the systemic circulation. Understanding the cellular origin and biological processes by which the adrenal medulla is formed during development is needed for mechanistic insights into how the hormonal component of the autonomic nervous system is formed and its relation to the rest of the autonomic nervous system.


Adrenergic chromaffin cells in the adrenal medulla are thought to originate from a common sympathoadrenal lineage close to the dorsal aorta, where these cells split in a dorsoventral direction, forming the sympathetic chain and adrenal medulla, respectively. Revisiting this dogma, we examined the cell type origin of chromaffin cells, lineage segregation of sympathoblasts and chromaffin cells, the gene programs driving specification of chromaffin cells from progenitors, and the proliferative dynamics by which the adrenal medulla is formed.


We found that chromaffin cells of the adrenal medulla are formed from peripheral glia stem cells, termed Schwann cell precursors. Genetic cell lineage tracing revealed that most chromaffin cells arise from Schwann cell precursors, and consistently, genetic ablation of Schwann cell precursors results in marked depletion of chromaffin cells. Genetic ablation of the preganglionic nerve, on which Schwann cell precursors migrate, similarly leads to marked deficiencies of chromaffin cells, and fate-tracing cells unable to differentiate into chromaffin cells reveal an accumulation of glia cells in the region of the adrenal medulla. Experiments reveal that sympathetic and adrenergic lineages diverge at an unexpectedly early stage during embryonic development. Embryonic development of the adrenal medulla relies on recruitment of numerous Schwann cell precursors with limited cell expansion. Thus, the large majority of chromaffin cells arise from Schwann cell precursors migrating on preganglionic nerves innervating the adrenal medulla. Unexpectedly, single-cell RNA sequencing revealed a complex gene-regulatory mechanism during differentiation of Schwann cell precursors to chromaffin cells, whereby Schwann cell precursors enter into a gene expression program unique for a transient cellular state. Subsequently, this gene program and chromaffin cell gene networks suppress glial gene programs, advancing cells into the chromaffin cell identity.


By revisiting development of the adrenergic sympathetic system, we discovered a new cellular origin of this nervous system component. The adrenergic medulla is built from both neural crest cells and Schwann cell precursors, with a major contribution from Schwann cell precursors in rodents. A cellular origin from Schwann cell precursors highlights the importance of peripheral nerves as a stem cell niche and transportation routes for progenitors essential for neuroendocrine development. These results and mechanisms of differentiation through a transient intermediate cell type may also be helpful in advancing our knowledge on neuroblastoma and pheochromocytoma, because these most often arise from the adrenal gland region.

Adrenal medulla largely originates from Schwann cell precursors.

Overview of adrenal medulla development resulting from lineage tracing and nerve ablation experiments. SCP, Schwann cell precursor; AG, adrenal gland; NT, neural tube; n, notochord; DRG, dorsal root ganglion; IML, intermediolateral column; NCC, neural crest cells; NC, neural crest; DA, dorsal aorta; SRG, suprarenal sympathetic ganglion. Red encodes early NCCs and their derivatives. Blue encodes late neural crest and SCP-derived cell types.


Adrenaline is a fundamental circulating hormone for bodily responses to internal and external stressors. Chromaffin cells of the adrenal medulla (AM) represent the main neuroendocrine adrenergic component and are believed to differentiate from neural crest cells. We demonstrate that large numbers of chromaffin cells arise from peripheral glial stem cells, termed Schwann cell precursors (SCPs). SCPs migrate along the visceral motor nerve to the vicinity of the forming adrenal gland, where they detach from the nerve and form postsynaptic neuroendocrine chromaffin cells. An intricate molecular logic drives two sequential phases of gene expression, one unique for a distinct transient cellular state and another for cell type specification. Subsequently, these programs down-regulate SCP-gene and up-regulate chromaffin cell–gene networks. The AM forms through limited cell expansion and requires the recruitment of numerous SCPs. Thus, peripheral nerves serve as a stem cell niche for neuroendocrine system development.

Chromaffin cells are neuroendocrine cells that produce catecholamines, which, once released into the bloodstream, mediate a stress response in health and disease by regulating bodily organs and tissues, including effects on metabolism. Furthermore, neuroblastoma, the most common extracranial tumor in children, originates from the sympathoadrenal compartment during development. Despite the functional importance of chromaffin cells, their origin and development are not well understood. Here, we revisited the embryonic origin of chromaffin cells and discovered a new chromaffin progenitor type, that is, the nerve-associated Schwann cell precursor (SCP).

Current opinion holds that the adrenergic chromaffin cells in the adrenal medulla (AM) originate from a migratory stream of neural crest cells that commit to a common sympathoadrenal lineage located close to the dorsal aorta. Upon their arrival at the dorsal aorta region, the committed sympathoadrenal cells proliferate and produce a spatial split in a dorsoventral direction. The more ventral group of cells forms the AM, whereas the dorsal portion coalesces into a sympathetic ganglion (SG) (1, 2). However, this view is challenged by the discovery of early differential expression of markers in the sympathoadrenal lineage (3, 4) and the presence of SOX10+ satellite glial cells in the early sympathetic system at the time of adrenal anlagen formation (5, 6). Direct evidence on how the AM is formed is missing, and furthermore, the current idea does not explain the cellular origin of the suprarenal sympathetic ganglion (SRG) in direct association with the adrenal gland. Thus, it has not been clear whether progenitors committed to the sympathoadrenal fate produce both sympathetic and chromaffin cells or whether SCPs and satellite glial cells also participate in the generation of chromaffin cells.

SCPs serve as multipotent stem cells, which can differentiate into numerous cell types, and most or all of the parasympathetic nervous system arises from these cells (710). It appeared possible that chromaffin cells also could arise from SCPs because, like parasympathetic neurons, they are located inside the organ where they function, and appear later in embryonic development in relation to neural crest stem cell migration.

SCPs build the AMs

To examine whether SOX10+ nerve-associated SCPs contribute to the generation of the main adrenergic sympathetic system, we fate traced nerve-associated SCPs using neural crest and glia-specific inducible Cre lines Sox10CreERT2 and Plp1CreERT2 coupled to the R26RYFP reporter. Genetic cell fate tracing was initiated in SCPs by tamoxifen (TAM)–induced recombination at E11.5 when neural crest cell migration is complete in the trunk (fig. S1) and multiple derivatives have been produced. The analysis of medullas at embryonic day 17.5 (E17.5) revealed large amounts of traced TH+ chromaffin cells (Fig. 1, A to C). Although initiating recombination at E12.5 demonstrated lower amounts of traced chromaffin cells, activation of recombination at E15.5 resulted in almost no contribution in E17.5 AM (Fig. 1, B and C). At the same time, the contribution of traced cells to sympathetic neurons of the SRG, ganglia of the sympathetic chain, and mesenteric or para-aortic ganglia was negligible (Fig. 1, A to C, and fig. S2), showing that the SCP compartment is restricted in terms of generating chromaffin cells versus sympathoblasts. Any neural crest recombination would be expected to result in fate-traced cells in the sympathetic system; thus, this finding also confirms the specificity of recombination in SCPs and not neural crest, consistent with the absence of freely migrating neural crest cells at these stages.

Fig. 1 Chromaffin cells of the adrenal gland originate from PLP1+ and SOX10+ SCPs at E11.5 and E12.5.

(A) Immunohistochemistry for yellow fluorescent protein (YFP) (recapitulating Plp1 expression) and TH on sections of the developing AM and suprarenal ganglion (SRG) after genetic tracing in Plp1CreERT2/+;R26RYFP/+ animals injected with TAM at E11.5 and analyzed at E17.5. The arrow points at the developing AM. (B and C) Quantification of the proportion of TH+/Plp1YFP+ cells traced in Plp1CreERT2/+;R26RYFP/+ (B) and of TH+/Sox10YFP+ cells in the AM, SRG, and SG of Sox10CreERT2/+;R26RYFP/+ mice (C) injected at E11.5, E12.5, or E15.5 and analyzed at E17.5. Note that the recombination efficiency (percentage of SOX10+/Plp1YFP+ cells out of all SOX10+ cells in embryonic nerves) at E12.5 is lower than that at E11.5. (D) Immunohistochemistry for SOX10 and TH on E13.5 sections of developing AM and SRG after TAM-induced cell ablation of SOX10+ cells at both E11.5 and E12.5 in Sox10CreERT2/+;R26RDTA/DTA mice. Note the marked decrease in SOX10+ and TH+ cell numbers in Sox10CreERT2/+;R26RDTA/DTA embryos as compared to Sox10CreERT2/+;R26R+/+. (E) Quantification of SOX10+ and TH+ cell numbers identified in the AM, SRG, and SG of Sox10CreERT2/+;R26RDTA/DTA and Sox10CreERT2/+;R26R+/+ embryos. Data are means ± SEM, two-tailed Student’s t test. In (A) to (C), for injection at E11.5, n = 3; at E12.5, n = 4; and at E15.5, n = 3. In (D) and (E), for Sox10CreERT2/+;R26R+/+ and Sox10CreERT2/+;R26RDTA/DTA AM, n = 3; SRG, n = 4; and SG, n = 3. ns, not significant. AM, adrenal medulla; SG, sympathetic ganglion; SRG, suprarenal sympathetic ganglion; DA, dorsal aorta.

Given the recombination efficiency (Fig. 1, B and C), we conclude from these experiments that at least half of all chromaffin cells are generated from nerve-associated SCPs between E11.5 and E15.5. In line with this, diphtheria toxin subunit A (DTA)–based ablation of SCPs performed in Sox10CreERT2;R26RDTA embryos injected twice with TAM at E11.5 and E12.5 and analyzed at E13.5 or E17.5 revealed a reduction of Sox10+ cells in both the SRG and the AM and TH+ chromaffin cell depletion in the AM in contrast to the unaffected sympathoblast numbers (Fig. 1, D and E, and fig. S3).

SCPs migrate along nerves, in contrast to the free migration of neural crest cells and common sympathoadrenal progenitors. Thus, if chromaffin cells arise from SCPs, then the formation of the AM should depend on the nerves innervating the gland. We therefore investigated the dependence of AM development on the presence of preganglionic motor nerve fibers. Preganglionic sympathetic motor neurons are located in the intermediolateral (IML) portion of the spinal cord. Fast blue retrograde tracing experiments previously identified NOS+/CHAT+ neurons as the source of AM innervation (11, 12). To achieve ablation of preganglionic nerves, we used intersectional genetics by breeding HB9Cre mice to the Isl2DTA mice. In these mice, DTA expression is driven by the Isl2 promoter. Isl2 and Hb9 expression is initiated in both somatic and visceral preganglionic neurons at E11.5 (13). HB9Cre mice were first bred to R26RTomato reporter mice to generate HB9Cre;R26RTomato embryos. Analysis of HB9Cre;R26RTomato embryos at E15.5 revealed that Hb9 is expressed by NOS+/ISL1+ preganglionic neurons and by their fibers in the adrenal gland (fig. S4, A and B) but expressed by only 5% of all developing chromaffin cells (fig. S4, C and D). Furthermore, we did not detect the expression of Isl2 and Hb9 in developing AM using single-cell transcriptomics. Thus, this approach allows for specific ablation of preganglionic motor neurons. At E14.5, only a few motor neurons were left in the spinal cord of Hb9Cre;Isl2DTA embryos (Fig. 2, A and B, and fig. S4E). Analysis showed a reduction by 78% in the total number of AM cells in Hb9Cre;Isl2DTA mice (Fig. 2, C to F) when compared to control mice, but the number of sympathetic neurons in the SRG was unaffected (Fig. 2E). The remaining chromaffin cells in Hb9Cre;Isl2DTA mice might be derived from the early neural crest streams or few remaining traversing visceral sensory afferent fibers, because both visceral and somatic motor nerves are eliminated in the Hb9Cre;Isl2DTA model.

Fig. 2 Preganglionic nerves are necessary for AM assembly.

(A and B) Immunohistochemistry for neuronal nitric oxide synthase (nNOS), CHAT, and ISL1 on E14.5 sections of spinal cords from control Isl2DTA/+ (A) and Hb9Cre/+;Isl2DTA/+ embryos (B) shows almost complete ablation of CHAT+/nNOS+/ISL1+ preganglionic neurons. (C and D) Immunohistochemistry for TH, SOX10, and CHAT on E14.5 sections of AM from control Isl2DTA/+ (C) and nerve-ablated Hb9Cre/+;Isl2DTA/+ embryos (D). Note the significant reduction of chromaffin cells in the AM but not of SRG sympathetic neurons in nerve-ablated (D) compared to control (C) embryos. (E) Quantification of (C) and (D). Data are means ± SEM, n = 3 for Isl2DTA/+ and n = 5 for Hb9Cre/+;Isl2DTA/+, two-tailed Student’s t test. (F) Graphical summary of the results. (G) Schematic showing the origin of chromaffin cells from nerve-associated SCPs. IML, intermediolateral cell column; CC, central canal; AG, adrenal gland; AM, adrenal medulla; NCC, neural crest cells; NC, neural crest; NT, neural tube; n, notochord; DA, dorsal aorta; DRG, dorsal root ganglion; SRG, suprarenal ganglion.

SCPs actively migrate along the nerves, because DiI tracing from the ventral root exit points in E11.5 embryos led to later emergence of DiI+ cells within the AM region (fig. S5A). Migration of SCP along nerves was also analyzed using genetic tracing. Krox20 (Egr2) is expressed exclusively in boundary cap cells located at the peripheral nervous system/central nervous system border (14, 15), and these cells lose Krox20 expression when detaching from the cap and migrating toward various locations in the embryos. We used Krox20Cre;R26RRFP mice to show that a few of these cells arrive at dorsal aorta and AM regions (fig. S5B). Combined, these findings suggest that, in contrast to sympathetic neuron origin directly from the migratory neural crest, 77.8% of the AM chromaffin cells are generated by recruitment of SCPs, schematically shown in Fig. 2G.

If chromaffin cells are formed by the initiation of a chromaffin gene expression program in nerve-associated SCPs, then abolishing this transcriptional differentiation program would be expected to result in the accumulation of SCPs that fail to differentiate. To test this hypothesis, we examined mice deficient in Ascl1—a critical factor driving chromaffin differentiation (16). Specifically, we generated Ascl1CreERT2/CreERT2;R26RTomato mice, in which the insertion of two CreERT2 alleles generates a knockout of Ascl1, allowing for genetic cell lineage tracing of Ascl1−/− cells. In control Ascl1 heterozygous mice (Ascl1CreERT2/+;R26RTomato), fate-traced Ascl1TOM+ cells generated low numbers of S100β+/SOX10+/Ascl1TOM+ Schwann cells and numerous S100β/SOX10/TH+/PHOX2B+/Ascl1TOM+ chromaffin cells, whereas in mice lacking Ascl1 (Ascl1CreERT2/CreERT2;R26RTomato mice), Ascl1TOM+ cells failed to down-regulate glial markers S100β and SOX10 (Fig. 3, A to H). At the same time, numerous traced cells were found to initiate expression of the transcription factor (TF) PHOX2B that is part of the chromaffin differentiation program but failed to become catecholaminergic (that is, SOX10/PHOX2B+/TH) (Fig. 3, H to J). Using serial sections, we quantified the frequency with which AM Ascl1TOM+-traced cells remained as glial-like cells (SOX10+), differentiated into chromaffin cells (TH+/PHOX2B+), or displayed the truncated chromaffin differentiation program with initiation of PHOX2B that, nevertheless, failed to become catecholaminergic (SOX10/PHOX2B+/TH). Ascl1CreERT2/CreERT2;R26RTomato mice displayed a fivefold increase in proportion and sixfold increase in numbers of traced SOX10+ cells, and the appearance of SOX10/PHOX2B+/TH cells with simultaneous 55% reduction of TH+ cell numbers (Fig. 3, I to K). The accumulation of glial cells in the absence of a neurogenic program is consistent with an SCP origin of chromaffin cells. Thus, our experiments using SCP fate tracing, SCP ablation, nerve ablation, and interference of chromaffin differentiation are all consistent with an SCP origin of large numbers of chromaffin cells in the AM.

Fig. 3 The role of Ascl1 in the SCP-to-chromaffin transition.

(A to H) Genetic ablation of Ascl1 prevents the glia-to-chromaffin transition. Most of the nerve-associated SOX10+/S100β+ cells expressing Ascl1 at E10.5 down-regulated glial markers in Ascl1CreERT2/+;R26RTOM/+ (control) embryos at E15.5 and have differentiated toward TH+/PHOX2B+ chromaffin cells (A to D) but failed to do so in Ascl1-deficient Ascl1CreERT2/CreERT2;R26RTOM/+ (mutant) embryos (E to H), whereas a new, intermediate PHOX2B+/TH/SOX10 population is observed (compare D and H). (I and J) Most of the Ascl1TOM+ cells in Ascl1CreERT2/+;R26RTOM/+ embryos are chromaffin cells with a minor contribution to SOX10+ glia, whereas in the Ascl1-deficient Ascl1CreERT2/CreERT2;R26RTOM/+ embryos, the larger population is represented by PHOX2B+/TH/SOX10 cells, with simultaneous increase in SOX10+ glia and reduction in TH+/PHOX2B+ chromaffin cells. Note the complete absence of the Ascl1TOM+-traced PHOX2B+/TH/SOX10 population in the control, which is observed in the mutant. (K) The proportion of Ascl1TOM+/SOX10+ glia over the total glia population is significantly increased in the mutant. Data are means ± SEM, n = 3 for all cases, two-tailed Student’s t test. In (A) and (E), the main differentiation trajectories derived from SCPs (marked with a pink halo) are shown with big arrows, whereas trajectories producing a minor or reduced cell population are shown by dashed arrows. White arrowheads point to Ascl1TOM+ glial cells. ChC, chromaffin cell; IC, intermediate cell.

Lineage segregation of sympathoblasts and chromaffin cells

Sympathetic neurons and chromaffin cells have been considered to originate from a common sympathoadrenal progenitor lineage located as a cluster of cells in the vicinity of the dorsal aorta. Our results indicate that large numbers of chromaffin cells are of a different cellular origin as compared to sympathetic neurons. This implies that the two lineages split at an earlier stage in development than previously thought. The primordium of AM forms around E12.5 and the earliest observations identifying the primary adrenal anlage are at E11.5 (1); hence, the lineage split is expected at some time before this. To examine this, we performed a number of genetic cell lineage tracing experiments. Ascl1 is expressed by both sympathetic and chromaffin cells, and the differential timing of its activation of expression can help to observe the lineage separation event. Ascl1CreERT2/+;R26RTomato mice were injected with TAM at E11.5 to fate trace Ascl1-expressing cells from this stage in development. Analysis of E17.5 embryos revealed that at E11.5, the lineages of sympathetic neurons and chromaffin cells were largely separated because most of the AM cells were traced, whereas only a few paravertebral or SRG sympathetic neurons were TOM+ (fig. S6A). This indicated that at E11.5, sympathoblasts down-regulated active transcription of Ascl1 as compared to chromaffin progenitors, and consequently, the lineages are separated at this time.

The tyrosine kinase RET is important for proper migration, proliferation, and survival of sympathetic cells (17) and is expressed by neural crest, sympathoblasts, and chromaffin cells at different developmental time points (18, 19). Whole-mount (fig. S6C) and cryosection-based immunohistochemical analysis (fig. S6, D and E) revealed RET expression in sympathetic progenitors and an absence of Ret expression in the vast majority of SCPs along the nerves of E11.5 embryos. RetCreERT2 mice were bred to R26RTomato mice for cell fate–tracing cells expressing Ret at E11.5 in RetCreERT2;R26RTomato mice with analysis at E15.5. Both paravertebral and SRG sympathetic neurons were traced, in contrast to minor tracing of AM cells (fig. S6B). The same result was obtained with TAM injections at E10.5 (Fig. 4, A and B), suggesting an early lineage split in progenitors generating sympathetic neurons and chromaffin cells.

Progenitors of sympathetic neurons express Sox10 and Plp1; however, the genes are down-regulated during differentiation, which starts at around E10.5 (20). Thus, tracing from the Plp1 locus at this stage should help to further corroborate the lineage split at this early stage. Although tracing in Plp1CreERT2/+;R26RYFP/+ embryos injected with TAM at E10.5 revealed more than 50% contribution to AM by E17.5, contribution to SG and SRG was no more than 10% (Fig. 4, C and D).

Fig. 4 Early specification and separation of sympathoadrenal lineages during development.

(A) Immunohistochemistry for TOMATO (recapitulating Ret expression) on RetCreERT2/+;R26RTOM/+ embryos injected with TAM at E10.5 and analyzed at E15.5. Arrowheads point to traced glia. (B) Analysis of Ret tracing in the AM, SG, and SRG in RetCreERT2/+;R26RTOM/+ embryos injected with TAM at E10.5 or E11.5 and analyzed at E15.5. Note the high recombination in the SG and SRG (for E10.5 injection, 51.77 ± 1.40% in the SG and 35.05 ± 1.19% in the SRG; for E11.5 injection, 64.96 ± 3.01% in the SG and 36.4 ± 3.08% in the SRG) in contrast to the AM (for E10.5 injection, 8.09 ± 0.71%; for E11.5 injection, 6.54 ± 0.72%). (C) Immunohistochemistry for YFP (recapitulating Plp1 expression) and TH on Plp1CreERT2/+/R26RYFP/+ embryos injected with TAM at E10.5 and analyzed at E17.5. Note the high numbers of Plp1YFP+/TH+ cells in the AM but only a few in the SG and SRG. (D) Quantification of (C). Note the high recombination in the AM (54.60 ± 3.01%) in contrast to the SG and SRG (9.23 ± 0.22% and 11.05 ± 1.37%, respectively). (E) Immunohistochemistry for YFP (recapitulating Chat expression), TH, and TuJ1 on ChatCre/+;R26RYFP/+ embryos analyzed at E17.5. Arrows point to ChatYFP+ cells in the SRG and SG. (F) Schematic showing the fate restriction of neural crest cells toward sympathetic neurons and SCPs, finally differentiating into chromaffin cells. In (B), (D), and (E), data are means ± SEM, n = 3 for all cases, two-tailed Student’s t test. MN, motoneurons; SN, sympathetic neurons; ChC, chromaffin cells; AM, adrenal medulla; SG, sympathetic ganglion; SRG, suprarenal sympathetic ganglion.

Because many early differentiating sympathoblasts are transiently exhibiting cholinergic phenotype (19), we used ChatCre/+;R26RYFP/+ embryos to rule out possible transition from sympathoblasts toward chromaffin cells of AM. We found no signs of conversion of maturing CHAT+ sympathoblasts into chromaffin cells (Fig. 4E). Together, all these data support an early lineage split between the sympathetic and the AM chromaffin cells, after the completion of neural crest migration (Fig. 4F).

Gene programs driving SCP specification into adrenergic cells

To obtain molecular insights into gene expression programs governing SCP–to–chromaffin cell transition, we performed single-cell RNA sequencing of the developing adrenomedullary cells at E12.5 and E13.5, stages when chromaffin cells are first appearing. We labeled the entire neural crest–derived compartment using Wnt1-Cre;R26RTomato mice and sorted fluorescent single cells from dissected medullae with associated SRGs into 384-well plates for individual sequencing using Smart-seq2 protocol (21). The SRG was included to allow comparison of AM cells to sympathetic neurons. Analysis using pathway and gene set overdispersion analysis (PAGODA) (22) identified coordinated expression variability signatures separating distinct subpopulations of cells in both E12.5 and E13.5 samples in a statistically significant way (Fig. 5, A and F, and fig. S7, A to C). The two time points also showed a common subpopulation structure (Fig. 5, B and G), with a prominent cluster of sympathoblasts (purple), chromaffin cells (green), and a cluster of SCPs (blue). The SCPs were marked by high levels of Foxd3, Sox10, Plp1, and Erbb3. ERBB3 is a part of the heteromeric receptor complex in glial cells that is necessary for survival and proliferation of SOX10+/PLP1+ SCPs receiving ligand interaction from membrane-anchored neuregulin-1 (NRG1) that is presented by the nerves (23). In the areas poor in NRG1, or after soluble NRG1 is not produced any longer (2), SOX10+ SCPs can maintain their survival and multiply only following the nerve branches that supply them with anchored NRG1. Reiprich et al. showed the presence of the SOX10+ cells in locations where the nerves normally descend to the AM, but the role of the nerve has not been explored (6). Consistently, nerves in the sympathoadrenal area were confirmed to be NRG1+ (fig. S8A). SOX10+ SCPs displayed expression of multiple genes participating in myelination (Pou3f1/Oct6, Sh3tc2, Lgi4, Dhh, Mal, etc.) and other glial cell markers (Fabp7, Mpz—and genes coding for S100β, laminins, Mag) (fig. S8, B to E). The sympathoblasts were distinguished by the presence of Slc18a3 (19) and Cartpt (coding for CART) (4) as well as by broader markers of sympathoadrenal differentiation such as Th, and chromaffin cells were identified by a Th+/Phox2b+/Chgb+/Slc18a3/Cartpt signature (Fig. 5, C and H, and fig. S7, B to D) (4).

Fig. 5 Transcriptional heterogeneity of the developing AM demonstrates SCP-to-chromaffin fate transition through a defined bridge state.

(A). Distinct subpopulations of neural crest–derived cells (columns) are seen within the E12.5 AM. Top five clusters of cells are given by the color bar below the dendrogram. Top seven statistically significant aspects (rows) of transcriptional heterogeneity are shown, labeled according to the main Gene Ontology (GO) category or a key gene driving each aspect. (B) Transcriptional profile–based subpopulations in the E12.5 sample are visualized using t-SNE embedding. The labels show interpretation of the distinct subpopulations based on the key marker genes below and a continuous bridge cell population (in red) connecting the SCPs and the chromaffin clusters. (C) Expression levels of representative marker genes within the E12.5 population (first 10 plots). Cell cycle dynamics (“dividing cells”: yellow, mitotic; green, interphase) are also shown. Finally, position of each cell along the SCP-to-chromaffin differentiation trajectory pseudotime is shown in the last plot. (D) Expression profiles of 1480 genes (rows) that are statistically significantly associated with the differentiation trajectory are shown as a function of pseudotime (cell position, x axis; individual cells colored according to their cluster membership are shown right below the axis). The genes can be categorized into those expressed early (close to SCPs), late (close to chromaffin), and transiently (within the bridge subpopulation). (E) Examples of pseudotime expression profiles for key genes within each category. Each point represents a cell, with the small horizontal line around it indicating uncertainty [95% confidence interval (CI)] of the cell’s position within the differentiation pseudotime. A smoothed regression line with the associated 95% CI is shown in red. (F to J) Analogous subpopulation and SCP-to-chromaffin differentiation bridge can be seen in the neural crest–derived cells of the E13.5 AM. In (A), the grayscale gradient represents z scores from 4.3 to 63, whereas in (F), the z scores are from 5 to 105. (K and L) t-SNE plots of developing medulla subpopulations normalized for the cell cycle–correlated genes at E12.5 (K) and E13.5 (L). Note that the yellow population aligns with the red bridge population. The graphs show the fraction of cells undergoing mitosis (y axis) as a function of pseudotime. Cells were classified as mitotic in the case of a positive score on the mitosis-driven aspect of heterogeneity.

The remaining subpopulations (red and yellow in Fig. 5, B and G) were intermediate types of cells spanning the expression state space between SCP-like to more chromaffin-like cells. The greatest observed overt difference between red and yellow subpopulations was the cell cycle signature present in the yellow cluster but absent from the red cluster (Fig. 5, C and H). To validate whether expression of cell cycle genes causes the separation of these apparently two different cell types, we normalized for cell cycle–associated genes and reanalyzed the data set. With an omission of the cell cycle–related genes, the two identified populations (red and yellow; Fig. 5, B and G) closely aligned with each other (Fig. 5, K and L). These intermediate cells form a continuous “bridge” between the chromaffin and SCP clusters, suggesting that they capture the transcriptional transition between SCP and chromaffin fates. In contrast, no such continuous bridge was observed between SCPs and sympathoblast clusters or between chromaffin cells and sympathoblasts (Fig. 5, B, G, K, and L).

To examine the transcriptional transition from SCP to chromaffin cells, we performed a pseudotime analysis (see Materials and Methods), positioning cells along a differentiation trajectory and identifying genes that show statistically significant expression variation along this reconstructed differentiation time course (Fig. 5, D and I). In addition to genes that are up- or down-regulated at the beginning or the end of the SCP-chromaffin path, we found a set of genes that were transiently increased specifically in differentiating cells. These were up-regulated at different intermediate time points along this differentiation trajectory (Figs. 5, D to J, and 6A; fig. S7, B and C; and tables S1 and S2), but were not expressed in SCPs or chromaffin cells (Fig. 5, D to J; fig. S7, B and C; and tables S1 and S2). This indicates the existence of an intermediate transient cellular state and that a complex regulatory cascade takes place within the bridge structure connecting SCPs and chromaffin cells. In the beginning of the pseudotime axis, SCP-specific genes Foxd3, Sox10, Erbb3, and others were down-regulated as cells moved into the Ascl1+/Htr3a+ intermediate bridge state toward chromaffin differentiation (Th+/Chga+) (Fig. 5, D, E, I, and J). Analysis of mitotic signatures showed prevalence of cell divisions only in the first half of the SCP-chromaffin trajectory (Fig. 5, C, H, K, and L). In addition, ~300 of bridge-characterizing, transiently expressed genes defined the transition state, including TF signature (Sox11, Hes6, Hipk2, Ascl1, Btg2, Aes, etc.) and signaling-related modules (Dll, PlxnA2, Htr3a, Htr3b, Cdkn1c, Kcnj12, etc.) (Fig. 6A and fig. S7B). Numerous TFs and signaling genes (SiGs) demonstrated affinity to the first (TFs: Tcf3, Smo, Id3, Ybx1, Sox4, etc.; SiGs: Notch1, Fzd2, Ptk7, etc.) or the second half (TFs: Hand1, Hand2, Phox2a, Eya1, Thra, Gata3, Insm1, Tbx20, Tlx2, etc.; SiGs: PlxnA3, Dll4, Amer2, Cxxc4, etc.) of the SCP-to-chromaffin trajectory. Toward the end of the pseudotime trajectory, we observed a group of specific genes indicative of the final steps of chromaffin differentiation beyond Th and Chga (TFs: FoxQ1, Egr1, Elf4, etc.; SiGs: Nrp2, etc.) (fig. S7, B and D). Furthermore, numerous genes showed expression specific to the chromaffin subpopulation and separated chromaffin cells from sympathoblasts (Bhlhe40, Sox1, Adora2) (fig. S7D and table S3). Overall, among the genes showing statistically significant association with the SCP-chromaffin differentiation trajectory, we found 139 TFs and 60 molecules from well-understood signaling pathways (fig. S7B).

Fig. 6 Mapping a novel bridge population of cells to the anatomical location and developmental timeline.

(A) Genes in the bridge cell population whose expression is dynamically changing in a statistically significant way. (B) Left: Immunohistochemistry for CHAT, PHOX2B, and ISL1 on sections of AM region from E11.5 embryos. Note that PHOX2B+ cells are associated with CHAT+ nerve. Middle: In 24-hour–traced Ascl1CreERT2/+;R26RTOM/TOM embryos analyzed at E11.5, nerve-associated cells of the intermediate stage express Ascl1 in addition to the glial markers ERBB3 and SOX10. Note that TH+/Ascl1TOM+ cells retain low levels of SOX10. Right: Immunohistochemistry at E11.5 for TH, ERBB3, and SF1 shows that TH+ cells near the assembling adrenal cortex (SF1+ cells) are associated with the nerve but do not express glial marker ERBB3. (C and D) Validation of the bridge cell population in SCP-to-chromaffin transition marked by the expression of Htr3aEGFP in E11.5 (C) and E12.5 (D) embryos. In (C), immunohistochemistry for GFP (recapitulating Htr3a expression), SOX10, and SF1 shows the anatomical position of the very first Htr3aEGFP AM cells within the developing adrenal gland region (marked by SF1 expression). Right: Immunohistochemistry for GFP (recapitulating Htr3a expression), SOX10, and TH shows that intermediate bridge Htr3aEGFP cells of AM rarely express TH but retained variable levels of SOX10. In (D), AM cells are either intermediate bridge TH/Htr3aEGFPhigh and SOX10+/Htr3aEGFPlow or chromaffin TH+/Htr3aEGFP cells. (E) Htr3aEGFP+ cell number in the medulla peaks at E12.5, declines at E13.5, and remains constant throughout development. Each dot represents counts from one section through medullary region (n = 2 embryos for all ages). (F) Left: Whole-mount immunofluorescence for PHOX2B and TH (magenta, autofluorescent blood vessels), CART, and Htr3aEGFP (middle) and immunohistochemistry on transversal cryosection (right). (G) Note that Cartpt (gene coding for CART) is not expressed in the intermediate Htr3aEGFP+ bridge cells or differentiating chromaffin cells, whereas it is present in sympathetic neurons of the SRG. (H and I) Corresponding 3D reconstructions of sympathoadrenal structures of E12.5 embryos. Large blood vessels were reconstructed on the basis of blood vessel autofluorescence of embryonic erythrocytes [see (F)]. Sympathetic structures were reconstructed on the basis of CART+ and PHOX2B+ signal. Note the consolidation of ventral sympathetic structures (presumably SRG, para-aortic, future mesenteric, and celiac ganglia) at this stage and the presence of the intermediate bridge cells outlined in the AM by Htr3aEGFP signal (white arrows).

These results show a mechanism involving a progressive down-regulation of SCP genes and up-regulation of chromaffin genes, starting with Phox2b and Ascl1, passing through transient stages, and culminating with the induction of effector genes such as Th and Chga. Studies have demonstrated that ASCL1 and PHOX2B are critical TFs during the formation of sympathetic (24) and parasympathetic (9) nervous systems, as well as adrenal gland development (16). The GO annotations showed statistically significant overrepresentation of differentially expressed genes from NOTCH, transforming growth factor–β (TGFβ), canonical WNT, and Sonic signaling pathways at various steps of SCP-to-chromaffin transition (fig. S9).

In situ existence of SCP cells in transit to generate adrenergic AM cells

Equipped with new and old markers for SCPs, bridge cells, and adrenergic chromaffin cells, we next examined the dynamics resulting in the formation of the AM cells within the tissue. The steroidogenic precursors that form the adrenal cortex are present at E11.5 and are revealed by steroidogenic factor-1 (Sf1) expression. SF1+ cells started at this stage to coalesce to form the adrenal cortex (Fig. 6, B and C, and fig. S6D). At this time, Ascl1TOM+, PHOX2B+, and TH+ cells (that is, chromaffin cell progenitors) first appear in association with preganglionic sympathetic CHAT+ nerves in the prospective location of AM (Fig. 6B). Injection of TAM in E10.5 Ascl1CreERT2/+;R26RTomato embryos and analysis at E11.5 showed that some Ascl1TOM+ cells were expressing Th, whereas other Ascl1TOM+ cells expressed the SCP markers Sox10 and Erbb3. Ascl1TOM+/TH+ cells were retaining some levels of Sox10 expression (Fig. 6B), suggesting that glial program (that is, SOX10) down-regulation precedes the commitment to the chromaffin cell fate, consistently with single-cell transcriptomics data. In line with this, nerve-associated TH+ cells were not ERBB3+ at E11.5 or E12.5 (Fig. 6, B and D), consistent with the single-cell transcriptomics analysis (Fig. 5, E and J) and confirming that Erbb3 expression is extinguished before a putative SCP-to-chromaffin fate switch.

We next identified cells in the SCP-to-chromaffin intermediate bridge stage [red and yellow clusters on t-distributed stochastic neighbor embedding (t-SNE) plots in Fig. 5, B and G] using the marker Htr3a. To detect Htr3a-expressing cells, we took advantage of the Htr3aEGFP reporter mouse line. Analysis of E11.5 embryos confirmed the existence of Htr3a-expressing [enhanced green fluorescent protein (EGFP+)] cells, most of which appeared to be nerve-associated and negative for TH as early as E11.5 in the AM region (Fig. 6C). The small number of weakly EGFP+ cells also stained for TH is likely caused by the slow degradation of EGFP molecules. The EGFP+ cells displayed variable levels of Sox10 expression at all analyzed stages (Fig. 6, C and D, and fig. S7E). To visualize cells in transition from SCPs to chromaffin cells between E12.5 and E13.5, we performed lineage tracing using Sox10CreERT2;R26RYFP mice. Numerous traced SOX10+ cells of AM were PHOX2B+, which is a key TF driving the adrenergic fate. By contrast, in SG, traced SOX10+ cells were negative for PHOX2B+ (fig. S7, F and G). This agrees with the single-cell transcriptomics data, where Phox2b expression is predicted early among SOX10+ cells entering the bridge stage (fig. S7B). Quantification of cells in transit to adopt a chromaffin cell fate by measuring the number of Htr3aEGFP+/TH cells revealed chromaffin cell differentiation of SCPs to occur at E11.5, with a peak at E12.5 and decline at E13.5 (Fig. 6, D and E, and fig. S7H), consistent with our cell fate tracing of SCPs. Taking advantage of this, we conducted whole-embryo three-dimensional (3D) rendering at E12.5 to identify the intermediate bridge cells in the intact tissue. For this purpose, we used a combination of various markers such as PHOX2B+ to detect sympathetic and chromaffin progenitors, TH for differentiated sympathetic and chromaffin cells, neurofilament to detect sensory nerves of DRG, RET to detect early sympathetic progenitors, and CART (coded by Cartpt) to detect the sympathetic component (chromaffin progenitors are CART). Finally, Htr3aEGFP+ was used to identify intermediate bridge cells.

Two separate rostrocaudal extensions of PHOX2B+/RetCFP+ sympathetic progenitor cells above and below dorsal aorta were identified at E11.5. The dorsal extension of cells represents the paravertebral sympathetic chain, whereas the ventral population largely reflects para-aortic and suprarenal ganglia, coalescing as early as E11.5 based on the RetCFP+ profile (fig. S6C). The identification of a parallel ventral sympathetic chainlike structure challenges the current idea that the ventral portion represents only chromaffin progenitors. On the basis of the mutually exclusive expression of Cartpt in sympathetic cells but not in the AM and also having Htr3aEGFP expressed by intermediate bridge cells, we used 3D reconstructions to identify the bridge intermediate cells within the ventral PHOX2B+/TH+/CART population at E12.5 (Fig. 6, F to I).

Bridge and differentiating chromaffin cells showed an increasing expression of nicotinic cholinergic receptors based on single-cell RNA sequencing. This could suggest the role of cholinergic signaling (25) in AM development (fig. S10, A and B). To demonstrate a putative role of cholinergic signaling for adrenergic differentiation, we isolated the AM from the embryos at E12.5 and E13.5 and cultivated them separately for 24 to 48 hours in the presence of activators or inhibitors of nicotinic receptors. The analysis (including 3D measurements) showed that SOX10+ and TH+ cells appear in denervated and isolated medullae equally well in all conditions, suggesting that cholinergic signaling is not a key for the early steps of medulla formation. However, the levels of Th expression were found higher in conditions with inhibitors (by 45% in the free-floating and 30% in the membrane-cultured AM explants; fig. S10, C to H), suggesting that the nerve activity may play a role in scaling the production of catecholamines by chromaffin cells by regulating the related enzymatic pathways.

Cell dynamics during AM organogenesis

The different cellular origin and molecular process of sympathetic neurons and chromaffin cells allows for possible differences in mechanisms of tissue expansion during organogenesis. Organs can be built by a few cells expanding massively or many cells expanding modestly.

We made use of the pTRE-H2BGFP;rtTA transgenic mouse line to examine proliferation of cells during the formation of the AM. In this mouse strain, histone H2B is fused to GFP. Because the number of histones is doubled at every cell division, a transient expression of H2BGFP is rapidly diluted if cells expand robustly; hence, the intensity of GFP correlates with the number of cell divisions after a doxycycline pulse (fig. S11A). In embryos injected at E11.5 and analyzed at E12.5, H2BGFP retention was high in already fated TH+ chromaffin cells and low in ERBB3+ SCPs, and the reverse was observed when embryos were injected at E13.5 and analyzed at E14.5 (fig. S11A). This finding was substantiated by immunofluorescence analysis of KI67 (Mki67) expression, a marker for dividing cells. At E12.5, most SOX10+ cells were largely KI67+, whereas TH+ cells were KI67, consistent with our analysis of single-cell RNA sequencing data (Fig. 5, C, H, K, and L). Conversely, at E14.5 and E17.5, the reverse was observed (fig. S11B). The transient arrest in the proliferation of early differentiated TH+ chromaffin cells is similar to that previously reported in developing SG (4, 20).

Thus, the AM is formed in two consecutive steps. The first phase of organogenesis involves an expansion of SCPs, which are deposited through differentiation of a number of quiescent chromaffin cells in the AM. In the second phase, chromaffin cells expand in numbers by proliferation.

This finding prompted us to establish the number of chromaffin cells deposited from individual progenitors into the AM by each progenitor cell, revealing how many cell divisions take place to generate the AM. To achieve this, we took advantage of clonal genetic tracing with color coding of neural crest clones using the R26RConfetti reporter (26) coupled to Sox10CreERT2 injected with TAM at E8.5. To achieve our objective, we used an experimental paradigm, which results in rare recombination events that, when combined with the stochastic generation of different traceable color codes after recombination in the Confetti mice, allow for clonal analysis. The analysis of AM and adjacent SRG from clonal density–traced embryos at E17.5 revealed the presence of color-coded clones (figs. S11C and S12) that comprised fewer than 32 cells per medulla (with mean at 14 cells ± 1.8 shown in fig. S11E). Spatial structure of the average clone demonstrated the presence of few cellular clusters, often separated from each other by some distance. Numerous clones were unique for each system, although some cells labeled by the same color code were shared between AM and SRG (fig. S12). This shows that chromaffin cells in medulla and sympathetic neurons in adjacent SRG might originate from different single neural crest cells traced from E8.5, although the interpretation of these results requires caution because neural crest progeny can migrate up to six segments in the rostrocaudal direction (27).

Next, we used Sox10CreERT2;R26RConfetti animals to trace SCPs in a low density–tracing mode starting from E11.5. Analysis at E17.5 showed that clones of chromaffin cells (originating from SCPs) were represented by less than 20 TH+ cells (with mean number of cells per medulla = 10 ± 1.2; fig. S11E). These cells often shared the clonal color code with local SOX10+ glial cells (fig. S11D, arrows) and were organized into discrete spatial clusters composed of two to six chromaffin and glial cells (fig. S11, F to H). Thus, the SCP–to–chromaffin cell conversion ratio is estimated to be 1 SCP cell labeled at E11.5 per about 10 identified traced chromaffin cells at E17.5. These ratios are valid only until E17.5 because chromaffin cell proliferation continues into postnatal stages (28). Therefore, our results suggest that numerous SCPs are required to build the AM during embryogenesis.


Our findings represent the discovery of the SCP-dependent origin of chromaffin cells and explain the fate split between the sympathetic and adrenergic lineages, especially considering previously published data on clonal tracing of neural crest in sympathoadrenal domain (29, 30). The difference in immediate origin and related timing of progenitor recruitment might define the split between sympathetic and chromaffin fates instead of a spatial heterogeneity of inductive signal.

Together, our results expand the diversity of SCP-derived cell types beyond previously known derivatives, such as parasympathetic (9, 10) enteric neurons (31, 32), melanocytes (8), endoneural fibroblasts (33), mesenchymal stem cells (34, 35), and adult glial cells (36). Our findings show that the peripheral nerves are niches and transportation routes for progenitors essential also for neuroendocrine development. The nerve ablation experiment demonstrates that high amounts of chromaffin cells are derived from the nerve, because the AM of nerve-ablated mice contained 78% less cells as compared to control. The knowledge gained in this study on cell types, their proliferative status, transcriptome, and how they contribute to the generation of chromaffin cells in the AM advances understanding of the origin of the adrenergic system and could also provide insight into neuroblastoma and pheochromocytoma because these most often arise from the adrenal gland region (37). We have discovered that SCPs generate chromaffin cells via an intermediate progenitor cell type, a bridge cell, characterized by a specific transcriptional program. Thus, the identification of novel adrenal progenitors, SCPs and bridge cells, might help in understanding the origin of neuroblastoma and pheochromocytoma (for MycN expression in SCPs and bridge cells, see fig. S7C).

In an evolutionary context, a direct contribution from neural crest cells together with indirect contribution through SCPs may explain how the AM can be adaptively scaled in a species-specific manner. This idea agrees with results by Green et al., showing a marked change through evolution in the contribution of SCPs as progenitors during the formation of the enteric nervous system (38). Thus, it is possible that SCPs represent an evolutionary substrate cell type essential for the advancement of the neuroendocrine system in the vertebrate lineage.


Materials and Methods

Figs. S1 to S12

Tables S1 to S3

References (3943)


Acknowledgments: I.A. was supported by Swedish Research Council, Bertil Hallsten Research Foundation, ERC Consolidator grant, and Ake Wiberg Foundation. P.E. was supported by the Swedish Research Council, Knut and Alice Wallenberg Foundation, Swedish Cancer and Söderberg Foundations, and ERC advanced grant (PainCells). P.V.K. was supported by NSF CAREER award (NSF-14-532) and NIH 1R01HL131768 from the National Heart, Lung, and Blood Institute. N.A. was supported from the Russian Science Foundation (RSF) grant no. 16-15-10273. V.D. was supported by VR (2015-03387), StratNeuro, Russian Foundation for Basic Research (16-04-01243), Stipend for Young Scientists # CП-2890.2016.4, and RSF (17-74-20037). F.L. and S.H. were supported by VR, KI, Ragnar Söderberg, and Wallenberg Foundations. J.P. was supported by VR International Postdoc Fellowship. A Deo lumen, ab amicis auxilium. We thank The Eukaryotic Single Cell Genomics facility at SciLifeLab and K. Wallenberg and S. Picelli for their invaluable support. We also thank J. Hjerling-Leffler and A. M. Manchado for reagents and mice. We also thank Karolinska Institutet FACS facility at CMB, more specifically J. Avila-Carino and B. Pannagel, for their help. We are grateful to U. Suter for sharing Plp1CreERT2 mice that we used in collaboration with U. Suter’s laboratory and K. Meletis for sharing ChatCre mice. We acknowledge V. Pachnis for providing Sox10CreERT2 mouse strain. Plp1CreERT2 and Sox10CreERT2 strains are available from U. Suter (ETH Zurich) and V. Pachnis (The Francis Crick Institute) laboratories correspondingly under material transfer agreements with their institutions. We are also grateful to O. Kharchenko for help with illustrations. Finally, we thank J. F. Brunet, C. Kalcheim, and H. Rohrer for critical discussions during recent years. Supplementary materials contain additional data. RNA sequencing results are available at Gene Expression Omnibus (GEO) under accession number GSE99933 ( and are available online through the interactive interface (for E13.5, see; for E12.5, see
View Abstract

Stay Connected to Science

Navigate This Article