Schwann cell precursors contribute to skeletal formation during embryonic development in mice and zebrafish

Significance

Multipotent Schwann cell precursors (SCPs) generate numerous cell types. Here, in both mouse and zebrafish, SCPs contributed to the generation of mesenchymal, chondroprogenitor, and osteoprogenitor cells during embryonic development. These findings reveal a source of cartilage and bone cells and previously unanticipated interactions between the nervous system and skeleton during development.

Abstract

Immature multipotent embryonic peripheral glial cells, the Schwann cell precursors (SCPs), differentiate into melanocytes, parasympathetic neurons, chromaffin cells, and dental mesenchymal populations. Here, genetic lineage tracing revealed that, during murine embryonic development, some SCPs detach from nerve fibers to become mesenchymal cells, which differentiate further into chondrocytes and mature osteocytes. This occurred only during embryonic development, producing numerous craniofacial and trunk skeletal elements, without contributing to development of the appendicular skeleton. Formation of chondrocytes from SCPs also occurred in zebrafish, indicating evolutionary conservation. Our findings reveal multipotency of SCPs, providing a developmental link between the nervous system and skeleton.

Cartilage is elastic tissue containing specialized chondrocytes, promoting endochondral bone growth and facilitating articulation of skeletal elements. Bone is constantly remodeled: Osteoblasts secrete the matrix required; osteocytes generated from osteoblasts modulate turnover; and osteoclasts resorb the bone. During embryonic development, bones form via either endochondral ossification, where long bones form on a cartilage template, or intramembranous ossification, where flat bones, including the skull and certain craniofacial bones, form by direct differentiation of mesenchymal cells into osteoblasts, with no cartilaginous intermediate.

Chondrocytes and bone-forming cells originate from 3 embryonic lineages: neural crest cells (NCCs) produce facial skeleton (1); paraxial mesodermal cells contribute to cranial and axial skeleton (2); and lateral plate mesodermal cells generate the skeleton of arms and legs (3). At designated locations, these cells form mesenchymal condensations that differentiate into chondrocytes (endochondral ossification) or differentiate into osteoblasts (intramembranous ossification) (4). Formation of transient, multipotent, embryonic NCCs in specific domains of the neuroectoderm during neurulation involves interactions between the neural plate and epidermal ectoderm (5). Thereafter, these cells undergo epithelial-to-mesenchymal transition, delaminate, migrate to the head and trunk, and differentiate into numerous cell types (6). In the head, they become neurons and glia, as well as ectomesenchymal tissues such as cartilage, bone, and other connective tissues of the craniofacial prominence and pharyngeal arches (5).

Schwann cell precursors (SCPs), direct descendants of NCCs, provide trophic and functional support for sensory and motor neurons and their axons (7, 8). SCPs navigate along peripheral axons to various sites, detach, and generate several cell types, including pigment cells (9), parasympathetic and enteric neurons (10, 11), dental mesenchymal cells (12), endoneurial fibroblasts (13), neuroendocrine chromaffin cells of the adrenal medulla (14), and Zuckerkandl organ (15). Thus, SCPs are multipotent, providing cells for various tissues and organs during development, when NCCs are no longer available (16). However, the role of SCPs in forming skeletal elements remains unknown.

Here, we demonstrate that, during murine embryonic development, SCPs leave peripheral nerves to contribute a few, but significant number of cells to cartilage and bone, both in the craniofacial region and trunk, including the scapula and ribs. A similar phenomenon occurs during zebrafish development.

Results

Specific Genetic Labeling of Embryonic SCPs in the Plp1^CreERT2 Mouse Line.

Proteolipid protein-1 (PLP1), the major myelin protein, is expressed initially by NCCs (17, 18). However, after NCCs have delaminated and migrated [embryonic day 9.5 (E9.5) for the cranium and E10.5 for the trunk (17, 19)], this protein is expressed predominantly by SCPs (9). Accordingly, the SCP lineage can be traced genetically in Plp1^CreERT2 mice (10–12, 14). To determine whether SCPs contribute to the formation of cartilage and bones, tamoxifen-inducible Plp1^CreERT2 mice were crossed with a R26R^YFP reporter or R26R^Confetti multicolor reporter strain (20) by inducing Cre-dependent genetic recombination at either E11.5 or E12.5, when SCPs could be labeled without labeling NCCs (9–12, 14).

The glial specificity of Plp1^CreERT2 mice was confirmed utilizing SOX10 as a panglial marker (21) and neuron-specific class III β-tubulin (TUJ1) to visualize peripheral nerves (22). The YFP signal visualized with anti-GFP antibodies and immunolabeling of SOX10 were both localized along TUJ1⁺ nerves in the face and trunk upon short-term tracing from E11.5 to E12.5 (SI Appendix, Fig. S1 A and B), as confirmed by scanning the entire mouse embryo (Movie S1; neurofilaments visualized with 2H3 antibody). Covisualization of Cre protein and GFP signal revealed recombination efficiencies of 57.8 ± 14.5% (211 GFP⁺Cre⁺ per 378 total Cre⁺ cells) when Plp1Cre^ERT2:R26R^YFP/+ mice were traced from E12.5 to E13.5 and 51.22 ± 5.5% (102 double-positive per 199 Cre⁺ cells) when traced from E15.5 to E16.5 (SI Appendix, Fig. S1 C and D). Similar experiments, employing Plp1^CreERT2 crossed with R26R^{Confetti/Confetti} mice, confirmed overlap between Cre and Sox10 expression along Tuj1⁺ nerves (SI Appendix, Fig. S1E), as well as between Sox10 and Confetti proteins, the latter visualized with anti-GFP antibody, which does not recognize RFP among 4 different confetti proteins, when traced from E12.5 to E13.5 (SI Appendix, Fig. S1F).

To prove that Plp1^CreERT2-labeled cells generate Schwann cells, we combined immunodetection of Protein Zero (P0, a Schwann cell marker) with YFP signal retrieval by anti-GFP antibody in Plp1Cre^ERT2:R26R^YFP/+ mice traced from E11.5 to either E12.5 or E17.5. The percentage of P0⁺ among GFP⁺ cells associated with TUJ1⁺ nerves rose from 12.3 ± 8% at E12.5 to 56.6 ± 18.6% at E17.5 (SI Appendix, Fig. S2 A–D), indicating that Schwann cells arose from SCPs.

Arising of murine sympathetic chain ganglia entirely from NCCs (23, 24) allowed confirmation that the latter were not traced. Predictably, neurons (staining for tyrosine hydroxylase [TH]) in sympathetic ganglia were not genetically labeled in Plp1^CreERT2;R26R^YFP embryos upon tracing from either E11.5 or E12.5 to E17.5 (SI Appendix, Fig. S2 E and F). All sympathetic ganglia were scanned for the YFP signal in TH⁺ sympathetic neurons to exclude leakage of Plp1^CreERT2 into NCCs.

Altogether, when recombination is induced at E11.5 or E12.5 in Plp1^CreERT2 mice, SCPs, but not NCCs are genetically labeled, as reported previously (10, 12, 14). Since 12.33 ± 8% of YFP⁺ cells express a marker for mature Schwann cells after 24-h tracing, some more mature Schwann cells may also be labeled, although these cells can be generated from SCPs during this period.

Spontaneous recombination without tamoxifen was excluded by scanning 7 whole Plp1^CreERT2;R26R^YFP embryos at E11.5 (representative scanning of Cre⁺ embryo in SI Appendix, Fig. S2G and Movie S2; 6.5% maximal laser power in Movie S1 versus 50% for Movie S2, which produces autofluorescent signals from internal organs and blood vessels). Furthermore, when the heads of 11 Cre⁺ Plp1^CreERT2;R26R^YFP embryos from 3 independent litters, either uninjected or injected with corn oil only, were scanned at E17.5, no spontaneous recombination was detected (representative sagittal facial sections in SI Appendix, Fig. S2 H and I). Since more spontaneous recombination is possible with 2 reporter copies, we tested Plp1^CreERT2 coupled with either 1 or 2 copies of the R26R^YFP allele (heterozygotes or homozygotes). Although no recombination occurred without tamoxifen, we minimized the risk of accidental recombination before tamoxifen by performing all subsequent experiments with R26R^YFP heterozygous mice (unless otherwise specified). Two copies of the R26R^Confetti allele were used to generate more color recombination options.

SCPs Give Rise to Mesenchymal Cells.

Since cells of the chondrocyte and osteoblast lineages arise from mesenchymal cells (25), we examined the potential SCP contribution to mesenchymal cells with color-coded genetic tracing of Plp1^CreERT2;R26R^{Confetti/Confetti} mice from E12.5 to E17.5. Clonal appearance was based on confetti color (Fig. 1A), and association with nerves revealed by staining neurons and confetti clones in the same area with the PGP9.5 marker and an anti-GFP antibody, respectively (Fig. 1B). At E17.5, some descendants of SCPs initially labeled remained on the nerve, maturing into Schwann cells (arrowheads in A′, A′′, B′, and B′′ in Fig. 1 A and B). Other progenies of the same color, most likely derived from the same progenitors, had detached from the nerves and exhibited typical mesenchymal cell morphology (arrows in A′, A′′, B′, and B′′ in Fig. 1 A and B). These cells formed small, tightly packed clonal patches, distinct from the large clones observed previously with NCC tracing in this Plp1^CreERT2;R26R^{Confetti/Confetti} strain (26). Note that the orange clone in A′ arises from the YFP and RFP signals combined, an occurrence so rare that the likelihood that these cells in such proximity are from different clones is essentially zero (see SI Appendix for statistical analysis). At E17.5, 4.6% of the genetically labeled cells with mesenchymal morphology were not associated with nerves (Fig. 1B). Immunostaining for PDGFRa, a marker for mesenchymal cells and endoneurial fibroblasts (27), in Plp1^CreERT2;R26R^{Confetti/Confetti} mice traced from E12.5 to E17.5 revealed that traced cells with mesenchymal appearance stain positively (Fig. 1C), confirming their mesenchymal nature.

Fig. 1.

SCPs generate mesenchymal cells during murine embryonic development. (A–C) Genetic tracing of Plp1^CreERT2;R26R^{Confetti/Confetti} embryos revealed SCP contribution to proximal mesenchymal cells. Confetti clones expressing RFP, YFP, and/or CFP proteins are shown in A. The same tissue section was immunostained with PGP9.5 (for neuron) and GFP (for confetti) in B. (C) The traced cells off TUJ1⁺ nerves were positive for mesenchymal marker, PDGFRa. In A–C, arrowheads indicate traced cells on nerves and arrows indicate traced cells that become mesenchymal cells.

To exclude direct induction of recombination in mesenchymal cells by Plp1^CreERT2, we analyzed Plp1^CreERT2;R26R^{Confetti/Confetti} mice traced from E12.5 to E13.5. Of 1,279 Cre⁺ cells, only 14 cells expressed PDGFRa and were located outside TUJ1⁺ nerves (1.09%; 11 embryos quantified). Similarly, of 1,391 GFP⁺ cells, only 19 were positive for PDGFRa and located outside TUJ1⁺ nerves (1.37%, 11 embryos analyzed; SI Appendix, Fig. S1 G and H). However, no Cre⁺YFP⁺ cells were detected outside TUJ1⁺ nerves (18 Cre⁺ or YFP⁺ cells in 11 embryos), suggesting that 1.09% of Cre⁺ cells outside nerves do not cause recombination and probably reflect variation due to antibody unspecificity or sporadic Cre too low to induce recombination.

Since nerve-associated PDGFRa⁺ endoneurial fibroblasts generate mesenchymal cell types during digit regeneration (27), we showed that 2.8% of Cre⁺ cells along TUJ1⁺ nerves also express PDGFRa⁺ (35 of 1,265; SI Appendix, Fig. S1G). Staining of recombined cells with anti-GFP antibody gave similar results (3.4%; 46 GFP⁺PDGFRa⁺ among 1,371 GFP⁺ cells; SI Appendix, Fig. S1H), suggesting that recombined cells include only 2.8 to 3.4% potential endoneurial fibroblasts.

SCPs Contribute to Chondrogenesis.

Chondrogenic mesenchymal condensations appear in the murine craniofacial compartment at E12.5, enlarge at E13.5, and become cartilage at E14.5 (28). At E12.5 these condensations express SOX9 (Fig. 2 A and B and SI Appendix, Fig. S3), a marker for both chondroprogenitors and mature chondrocytes (29, 30). Tracing Plp1^CreERT2;R26R^YFP/+ mice from E11.5 to E12.5 revealed virtually no overlap between YFP⁺ cells and SOX9⁺ chondroprogenitors (4 double-positive among 16,776 SOX9⁺ cells in the craniofacial region; none among 6,929 SOX9⁺ cells in the ribs and scapula) (Fig. 2 A and B and SI Appendix, Fig. S3A). At E12.5 in wild-type nontraced mice as well, no cells expressed both SOX9 and the SCP marker, SOX10 (SI Appendix, Fig. S3 B and C).

Fig. 2.

SCPs generate chondroprogenitors in craniofacial region and trunk during murine embryonic development. Genetic labeling in Plp1^CreERT2;R26R^YFP/+ embryos from E11.5 to E12.5 revealed no overlap between SCPs and SOX9⁺ chondroprogenitors (A and B). Prolonged tracing revealed appearance of SCP progeny in cartilage at E15.5 (C and D) and E17.5 (E and F). Orange arrowheads and white arrows indicate YFP⁺ cells on TUJ1⁺ nerves and close to nerves and cartilage, respectively, and white arrowheads indicate YFP⁺SOX9⁺ chondrocytes in C′ and D′. The orange arrow in E′ indicates an YFP⁺ perichondrial cell. These images represent at least 5 Cre⁺ embryos from independent litters. A total of 5 facial and 2 trunk skeletal elements in each embryo were checked (see SI Appendix, Fig. S4 for details).

At E15.5, TUJ1⁺ nerves and associated YFP⁺ glial cells (labeled with tamoxifen at E11.5) grew close to cartilaginous elements (orange arrowheads in Fig. 2 C′ and D′, and C and D). YFP⁺ cells dissociated from nerves and some YFP⁺SOX9⁺ chondrocytes (white arrows and arrowheads, respectively, in Fig. 2 C′ and D′, and C and D) were also observed. At E17.5, these double-positive cells had separated completely from the neurites, constituting various parts of the facial and trunk skeletons (Fig. 2 E and F and SI Appendix, Fig. S4 A–G and V). Some YFP⁺ cells generated perichondrium cells (orange arrows in Fig. 2 E′ and F and SI Appendix, Fig. S4 A–C). In craniofacial cartilage, perichondrial cells differentiate into chondrocytes arranged in characteristic transversal columns (28), as occasionally observed here when tracing SCP cells (white arrowheads in SI Appendix, Fig. S4 B and C). In the trunk, SCPs contributed to formation of chondrocytes in the rib and scapula (SI Appendix, Fig. S4 F, G, M, N, and V). Interestingly, these rib chondrocytes originate from the mesoderm (2), suggesting that, by migrating along nerves, SCPs might help form structures other than neural crest derivatives.

Differentiation of SCPs into mature chondrocytes (directly or via mesenchymal or perichondrial intermediate) was confirmed by visualizing YFP⁺ cells in regions expressing type II collagen (SI Appendix, Fig. S4X). Genetic tracing from E12.5, as from E11.5, revealed SCP contribution to various cartilage elements (SI Appendix, Fig. S4 H–N and V), in contrast to tracing from E15.5 (SI Appendix, Fig. S4 O–U and V), suggesting that SCPs differentiate into cartilage transiently, similar to their formation of parasympathetic nerves (10). No contribution to long bone cartilage was observed (SI Appendix, Fig. S4W). To estimate the SCP contribution, we examined embryos carrying 1 (R26R^YFP/+) or 2 copies (R26R^YFP/YFP) of the YFP reporter, where the frequency of recombination might differ. YFP⁺ cells generate up to 10% of the SOX9⁺ chondrocytes in homozygous embryos (SI Appendix, Fig. S4 Y and Z). In light of 58% efficiency of Cre recombinase (see above), SCPs might contribute ≥10% of all chondrocytes, rendering this chondrogenesis of considerable biological significance.

SOX10 is a marker of SCPs. We employed Sox10^CreERT2 crossed with R26R^{Confetti/Confetti} mice to further verify the SCP contribution to chondrocytes. Short-term tracing before cartilage is induced (E10.5 to E11.5) showed recombination in SOX10⁺ cells along nerves, with no recombination in the SOX9⁺ chondrogenic domain (SI Appendix, Fig. S5 A–D). In total, 1.47% of cells in the SOX9⁺ domain of the craniofacial region were GFP⁺ (57 of 3,878) and 0.23% in the trunk were (22 of 9,430), whereas 99% of recombined cells were associated with TUJ1⁺ nerves, indicating their glial origin. Prolonged tracing (E10.5 to E17.5) revealed contribution to cartilage in both facial and trunk regions (SI Appendix, Fig. S5 E–J), similar to Plp1^CreERT2 genetic tracing. Clearly, Sox10 tracing must be interpreted carefully, since at E9.5 Sox10 labels late NCCs in the craniofacial region, whereas at E12.5 early chondroprogenitors are labeled (26, 28), leaving a narrow time window for specific SCP labeling.

To understand whether the fate of specific SCP populations is restricted, we utilized the Dhh^Cre line. DHH is first expressed by SCPs around E12.0 in a patchy manner (13), suggesting that the DHH⁺ subpopulation matures only into Schwann cells and endoneurial fibroblasts (13, 31). Support is provided by only partial overlap between SOX10⁺ and DHH⁺ populations along TUJ1⁺ neurites at E12.5 (11.6%; range, 2.2 to 21.9%) Tomato⁺SOX10⁺ cells within the SOX10⁺ population in the face and 6.9% (2.0 to 17.8%) in the trunk (SI Appendix, Fig. S6 A and B). In addition, DHH⁺ cells did not contribute to the sympathoadrenal anlage (marked with TH antibody), which is formed by SCPs (14) (SI Appendix, Fig. S6C). As expected, Dhh⁺ SCPs did not contribute to cartilage and instead generated scattered Schwann cells within nerve bundles at E17.5 (SI Appendix, Fig. S6 D–H). This suggests that SCPs becomes heterogeneous around E12.5, and not all late subpopulations give rise to mesenchymal tissue, including cartilage.

Altogether, these observations provide proof-of-principle that embryonic SCPs contribute to formation of various cartilage elements of the skeleton.

SCPs Contribute to Osteogenesis.

Next, we examined potential contribution of peripheral glial cells to formation of the osteoblast lineage, beginning with osteoblast progenitors and utilizing the transcription factors Osterix (OSX) and Runx2 as markers. Genetic tracing of Plp1^CreERT2;R26R^YFP/+ embryos from either E11.5 or E12.5 to E17.5 revealed colocalization of YFP with either marker in bones (Fig. 3 A–C and SI Appendix, Fig. S7 A–C, D, and E). The greatest contribution to osteoprogenitors, in the mandibular region, was not dramatic (2.1± 0.22%; Fig. 3D). No YFP⁺ osteoprogenitors were observed during tracing from E15.5 to E17.5 (Fig. 3D). DMP1 and E11, markers for osteocytes (32), showed that some YFP⁺ cells differentiate into DMP1-expressing osteocytes (Fig. 3 E and F) in E11⁺ and DMP1⁺ bone regions (SI Appendix, Fig. S7 F–I). Importantly, ossified elements in the craniofacial region appear initially only at E15.5 (28), several days after our genetic labeling.

Fig. 3.

SCPs generate osteoprogenitor cells and osteocytes in facial region and trunk during murine embryonic development. (A–C) SCPs progeny in Plp1^CreERT2;R26R^YFP/+ embryos traced from E11.5 to E17.5 were positive for osteoprogenitor marker OSX in the ossified parts of mandible (A), rib (B), and scapula (C). The white arrowheads indicate double-positive cells. (D) Quantification of YFP⁺ cells among the OSX⁺ population. Data represent mean ± SEM where at least 3 embryos from independent litters were analyzed. (E and F) The same traced embryos were positive for Dmp1 RNA probe. A′′′′, B′′′′, and C′′′′ depict the same tissue sections as A–C but stained with von Kossa and Alcian blue (vK and Ab). The white dashed lines outline the mineralized portion of the bone.

Altogether, these experiments suggest that during embryonal bone formation an SCP subpopulation contributes to the osteoblast lineage, including preosteoblasts and osteocytes.

The SCPs’ Contribution to Skeletogenesis Has Been Conserved Evolutionarily from Zebrafish to Mice.

The contribution of glial cells to skeletal development was also explored in Sox10^CreERT2;Ubi:zebrabow zebrafish (33), which express YFP or CFP upon 4-hydroxytamoxifen-induced recombination. Only YFP⁺ cells could be monitored because of greater signal and imaging constraints.

During development of wild-type AB zebrafish larva from 24 to 48 h postfertilization (hpf), cells stained with anti-SOX10 antibody were associated closely with nerves, i.e., they were SCPs and peripheral glial cells (SI Appendix, Fig. S8 A–D). Tracing in Sox10^CreERT2;Ubi:zebrabow fish from 24 to 48 hpf revealed overlap between YFP and SOX10 along nerves (SI Appendix, Fig. S8 E and F), indicating that Cre-dependent recombination results in genetic tracing of SCPs. Although cartilage elements in the facial region of zebrafish larvae start to form at about 60 hpf, the area between the otic vesicle and first somite showed traces of alcian blue staining at 48 hpf (SI Appendix, Fig. S8G), and immunohistochemistry revealed the first SOX9a⁺ chondroprogenitors around the otic vesicle (SI Appendix, Fig. S8H). During 24-h tracing (24 to 48 hpf), some SOX10-traced cells detached from the nerves but did not express SOX9a (SI Appendix, Fig. S8I; 6 double-positive of 395 YFP⁺ cells). Surprisingly, occasional YFP⁺ cells located along the nerves expressed SOX9a (SI Appendix, Fig. S8J).

With recombination at 24 or 48 hpf and tracing until 44 d postfertilization (dpf), late juvenile stage, SCPs contributed to various craniofacial skeletal structures in a spatially compact manner, suggesting clonal origin of labeled chondrocytes and of rare cartilage patches in pharyngeal (ceratohyal), mandible (palatoquadrate), and chondrocranial (parachordal) regions and the ear (otic vesicle) (5.3 ± 1.1% and 8.3 ± 1.5% of the cells present during the 2 tracing periods, respectively) (SI Appendix, Fig. S9 A–H). As in mice, patches of YFP⁺ chondrocytes were often located adjacent to flat YFP⁺ perichondrial cells (orange arrows in SI Appendix, Fig. S9 A, B, D, F, and G). Visualization of cartilage with SOX9a antibody confirmed the contribution of traced cells to various skeletal elements (SI Appendix, Fig. S9 I–P). Thereafter, crossing Sox10^CreERT2;Ubi:zebrabow with Col2-mCherry zebrafish showed early chondrogenesis simultaneously with tracing of SCPs. Live imaging of these fish when chondrogenesis began and several days later revealed no overlap between YFP⁻ and mCherry⁺ cells when Sox10⁺ cells were traced from 24 to 54 hpf or from 48 to 72 hpf, whereas YFP⁺ chondrocytes appeared at 120 hpf (SI Appendix, Fig. S9 Q and R). Importantly, chondrocytes in zebrafish express SOX10, a common marker for cartilage. We labeled SCPs before chondrocyte formation, i.e., 24 or 48 hpf (at 48 hpf chondrogenesis starts around the otic vesicle; see above). However, to compare glial contribution to that of genetically labeled chondrocytes, we exposed Sox10^CreERT2;Ubi:zebrabow fish to 4-hydroxytamoxifen at 5 dpf (when pharyngeal cartilage forms), 7 dpf (this formation is complete), or 11 dpf (metamorphosis), and traced until 44 dpf. Labeling at 5 dpf resulted in large clusters of YFP⁺ chondrocytes (SI Appendix, Fig. S8K), with almost all cartilage elements labeled at 7 or 11 dpf (SI Appendix, Fig. S8 L and M). In all cases, the labeling pattern was distinct from that of SCPs, where only small, probably clonal clusters of several YFP⁺ chondrocytes were observed (SI Appendix, Fig. S9 A–H).

Altogether, these experiments suggest that, as in mice, SCPs contribute to skeletogenesis in zebrafish.

Discussion

Novel, fundamentally noncanonical roles of the peripheral nervous system discovered recently include tissue regeneration (34), controlling stem cell niches (35), relaying fine positional information (36), and delivering multipotent cells to various locations (14). In the latter connection, SCPs are emerging as a new class of multipotent progenitor cells active during both embryogenesis and postnatal life (10–12, 14), although their potential remains to be elucidated in detail. Here, genetic tracing demonstrated that a small, but significant fraction of SCPs generate mesenchymal cells and chondroprogenitors and osteoprogenitors, expanding their destinies and emphasizing their multipotency. This capacity was limited to embryonic days E11.5 to E15.5 in the mouse, a time window similar to SCP-dependent generation of chromaffin cells in the adrenal medulla (14) and parasympathetic neurons (10). This window closes as SCPs commit to their fate, becoming phenotypically intermediate, immature Schwann cells (14).

NCCs are immediate progenitors of craniofacial ectomesenchymal cells, as well as multipotent SCPs (37). We show that SCPs generate craniofacial mesenchymal cells, resembling their differentiation into dental mesenchyme (12). These are strong indications that, after leaving nerves, SCPs retain the capacity of their developmental ancestors to produce various mesenchymal populations. One potential explanation might involve developmental constrains posed by the transience of NCCs. Condensation of cartilaginous mesenchymal (i.e., SOX9⁺) in murine craniofacial regions begins around E12.5, well after NCC migration (E9.5) (19, 26, 28). Thus, SCPs potentially provide additional plastic, local, targeted source of mesenchymal progenitors during late skeletogenesis. Although in mice SCP contribution to craniofacial development might not be significant, other animals, which develop longer or larger animals, such as humans, might need these additional cells for proper skeletogenic development. This resembles the strategy apparently involved in late generation of chromaffin cells in the adrenal medulla (14) or parasympathetic ganglia (10).

This SCP-derived chondrogenesis might be directly related to the condensing mesenchyme, consisting of SCP progeny. If so, mesenchymal cells derived from SCPs can differentiate into cartilage or bone, depending on location, analogous to the position-based differentiation of mesenchymal cells derived from NCCs in the craniofacial region (26). Alternatively, the perichondrium, a thin layer of mesenchymal-type cells surrounding embryonic cartilage elements, may be an intermediate stage. Although cartilage is not innervated, the surrounding perichondrium becomes highly innervated after formation of the skeletal elements (ref. 38 and own observations), i.e., after E12.5, when NCC migration is complete. Accordingly, SCPs may be delivered to the perichondrium during innervation and generate additional mesenchymal cells that produce clusters of labeled chondrocytes. Indeed, generation of chondrocytes by perichondrial cells does occur (28). In support of this intermittent role, we observed no SOX10⁺ SCPs near SOX9⁺ chondrogenic condensations at E12.5, whereas SCP contribution to cartilage elements begins when the nerves extend closer to these elements. In many cases, labeled cells remain present in both the perichondrium and underlying chondrocytes. Furthermore, many clones were transversal, strongly resembling formation of craniofacial cartilage from the perichondrium by intercalation (28). This might fine-tune cartilage shape by spatially constraining delivery of progeny cells from the perichondrium (28). However, the relatively few chondrocytes derived from SCPs and the lack of any specific spatial distribution argue against this.

The SCP contribution to the perichondrium is especially pronounced in zebrafish, indicating evolutionary conservation, as with SCP differentiation into melanocytes (16). This ancient SCP multipotency is further demonstrated by their capacity to generate the enteric nervous system in cyclostomes (lamprey) (39). However, the evolutionary origin of this multipotency remains unclear (40). In interpreting our findings with zebrafish, one must remember that Sox10, employed for genetic tracing of SCPs, is also expressed by mature chondrocytes. Although we labeled SCPs before chondrogenesis onset and observed no overlap between Sox10-labeled cells and the early marker of chondroprogenitors Sox9a outside nerves, some labeled mesenchymal cells other than SCPs, may have contributed to skeletal elements.

In bony structures, OSX⁺ and Runx2⁺ cells are considered to be osteoprogenitors (41). We found that SCPs generate such cells as well as osteocytes. For endochondral ossification, OSX⁺ osteoprogenitors from the perichondrium migrate into skeletal elements, which are initially cartilaginous, along with blood vessels (42, 43). Thus, like cartilaginous elements, the initial SCP contribution to the perichondral mesenchyme may be followed by development and migration of OSX⁺ osteoprogenitors from the perichondrium inside ossified skeletal elements. With intramembranous ossification, the origin of OSX⁺ cells is less clear, presumably involving direct differentiation of mesenchymal cells into osteoprogenitors. Our findings suggest that at least some OSX⁺ osteoprogenitors originate from peripheral glial cells of associated nerves. Although this contribution is limited, Plp1^CreERT2 recombination is 50 to 60% (ref. 10 and own observations) and the YFP signal was retrieved with an antibody as well, which underestimates the actual SCP contribution to osteogenesis. Thus, although relatively minor, the supply of SCPs to both chondroprogenitors and osteoprogenitors could be at least twice that estimated here.

We observed that SCPs contribute to formation of multiple skeletal parts arising either via intramembranous ossification (mandible, subscapular fossa) or endochondral ossification (rib, acromion, and coracoid processes of the scapula), as well as nasal and otic cartilage. Such contribution of neural crest derivatives to trunk skeletal elements contradicts the classical view, although a mixture of mesodermal and NC derivatives do contribute to the scapula (44). This contribution to the ribs, which originate from paraxial mesoderm (44), excluding any direct NCC contribution, is particularly interesting. This suggests that SCPs migrating along developing nerves generate cells that form skeletal elements largely of nonneural crest origin. The reason for mixing descendants of different developmental origins in the same structures remains unexplained. We observed no SCP contribution to long bones of limbs or paws. Another report suggests that NCCs and their progeny (including SCPs) contribute to stromal cells in the marrow of limb bones late during embryonic development and at early postnatal age (45). That study involved noninducible Wnt1^Cre2 mice, utilizing the Wnt1 promoter in NCCs, from E8.5 to E9.5. However, several recent reports indicate that this promoter is active in bone cells (46–49), so secondary activation of Wnt1^Cre2 in those differentiated cells might explain the labeling observed (45). In support of this, we detected no NCC contribution to the limbs upon inducing recombination in these cells by injecting tamoxifen into Sox10^CreERT2;R26R^Confetti mice at E8.5, in line with the classical view concerning tissues to which NCCs contribute (50). As we discussed recently, genetic experiments with noninducible Cre mice must be interpreted carefully, due to potential nonspecific promoter activity during ontogenesis (51).

Various interactions between the nervous system and skeleton have long been proposed (12, 45, 52). We demonstrate that during murine embryogenesis, glial progenitors associated with peripheral nerves differentiate directly (although rarely) into skeletal progenitors. Whether these progenitors retain intimate interactions with the nervous system or differ from other skeletal progenitors remains unknown. At present, it is unclear whether glia contribute to skeletogenesis during the postnatal period. Such contribution was not observed after E15.5 here, but this does not exclude reactivation under pathological conditions or following trauma, e.g., bone fracture. Although we observed no contribution by endoneurial fibroblasts to skeletal cells during embryonic development, this clearly happens during trauma (27). Finally, although less likely, we cannot exclude involvement of mature Schwann cells in skeletogenesis, since 12% of them were labeled during genetic tracing. However, this does not alter our conclusions concerning interplay between the nervous system and developing skeleton, since both SCPs and Schwann cells belong to the peripheral nervous system. In summary, precursors of peripheral glial cells can differentiate into skeletal progenitors during embryonic development. This glial skeletogenesis represents a previously unanticipated interaction between the nervous and skeletal systems.

Materials and Methods

Mice and Zebrafish.

All animal work was permitted by the Ethical Committee on Animal Experiments (Stockholm North Committee) and conducted according to The Swedish Animal Agency’s Provisions and Guidelines. The Plp1^CreERT2, R26R^YFP, Sox10^CreERT2, R26R^Confetti, and Dhh^Cre mice (18, 20, 27, 53) and Sox10^CreERT2 (54) and Ubi:zebrabow (33) zebrafish have been described. Col2:mCherry zebrafish were a gift from Chrissy Hammond, School of Physiology and Pharmacology, University of Bristol, Bristol, UK (55).

Immunohistochemistry.

Immunohistochemistry and whole-mount embryo staining were performed as described (14, 19, 51).

RNAscope.

RNAscope was performed according to ACD Bio’s manual using the Dmp1 probe (ACD Bio; no. 441171), followed by overnight incubation with primary antibody.

See extended methodological details in SI Appendix, SI Material.