Unmyelinated sensory neurons use Neuregulin signals to promote myelination of interneurons in the CNS

SUMMARY

The signaling mechanisms neurons use to modulate myelination of circuits in the central nervous system (CNS) are only partly understood. Through analysis of isoform-specific neuregulin1 (nrg1) mutants in zebrafish, we demonstrate that nrg1 type II is an important regulator of myelination of two classes of spinal cord interneurons. Surprisingly, nrg1 type II expression is prominent in unmyelinated Rohon-Beard sensory neurons, whereas myelination of neighboring interneurons is reduced in nrg1 type II mutants. Cell-type-specific loss-of-function studies indicate that nrg1 type II is required in Rohon-Beard neurons to signal to other neurons, not oligodendrocytes, to modulate spinal cord myelination. Together, our data support a model in which unmyelinated neurons express Nrg1 type II proteins to regulate myelination of neighboring neurons, a mode of action that may coordinate the functions of unmyelinated and myelinated neurons in the CNS.

In brief

Signals coordinating myelination of diverse neurons in the CNS are not well understood. Lysko et al. report that nrg1 type II regulates myelination of different neuronal classes in the zebrafish spinal cord. These nrg1 signals are expressed in unmyelinated sensory neurons, where they act to control myelination of neighboring interneurons.

Graphical Abstract

INTRODUCTION

The staggering complexity of neuronal circuits in the central nervous system (CNS) requires precise regulation of axonal conduction to coordinate the transmission of action potentials. Emerging evidence indicates that modulation of myelination can tune axonal conduction velocity to coordinate circuit activity, thereby enabling coherent input, output, and behavior.^1–5 Specific patterns of myelin sheaths have been described on distinct neurons in the brain and spinal cord,^6–9 but the mechanisms that encode these patterns are only starting to be understood. Increasing evidence shows that myelination can be modulated by neuronal activity^7,10–13 as well as a wide variety of signaling molecules, including PDGF, BDNF, ATP, and Neuregulin, among others,^14–16 but our understanding of how these factors coordinate myelination within complex neuronal assemblies remains incomplete.

Neuregulin1 (Nrg1) signals have a well-defined role in peripheral nervous system (PNS) myelination,^17–22 but their role in CNS myelination is less clear. The human NRG1 gene encodes more than 30 isoforms that act as extracellular ligands, binding to ErbB family receptor tyrosine kinases on the surface of target cells.^23–25 In the PNS, the transmembrane isoform Nrg1 type III is the key axonal signal that controls the proliferation, migration, and myelination of ErbB-expressing Schwann cells.^18,26,27 In CNS myelination, the function of the Nrg1 type III isoform is much more limited; myelin is present but reduced in the brain of ErbB receptor mutants.^28,29 However, these and other studies²⁰ report normal myelination in the spinal cord in Nrg1 and ErbB mutants, despite evidence from pioneering in vitro and ex vivo studies^30,31 that established Nrg1 signals as regulators of development, migration, or proliferation of oligodendrocytes, the myelinating cells of the CNS. The roles of other Nrg1 isoforms in CNS myelination are not well-known, although different Nrg1 isoforms regulate synapse formation, heart development, and other processes.^17,32,33

To investigate the roles of different Nrg1 isoforms in the CNS, we created a series of isoform-specific mutations in the zebrafish nrg1 gene and analyzed myelination in the spinal cord using live imaging. Our results indicate that the Nrg1 type II isoform modulates the myelination of distinct neuronal classes in the spinal cord. Analysis of Nrg1 type II expression and cell type-specific CRISPR experiments provide evidence that unmyelinated sensory neurons expressing Nrg1 type II control the myelination of neighboring interneurons. Our results indicate that, in contrast to the PNS, Nrg1 type II signals through ErbB2 receptors on neurons, rather than oligodendrocytes, to modulate myelination in the CNS.

RESULTS

nrg1 type II is required for normal myelination in the spinal cord

The nrg1 gene generates many isoforms through multiple promoters and alternative splicing. A unique promoter and 5′ exon distinguish each main class of isoforms, resulting in distinct tissue-specific expression patterns and N-terminal protein sequences: types I–VI are present in humans, while types I, II, and III are common to all vertebrates.^24,25 To determine which Nrg1 isoforms have functions in myelin sheath formation in the CNS, we used CRISPR-Cas9 to create a series of mutations in the zebrafish nrg1 gene. We used a single guide RNA (sgRNA) targeting the common EGF domain to eliminate the function of all nrg1 isoforms and sgRNAs targeting each isoform’s unique exons to create isoform-specific mutants (Figures 1A, 1B, and S1A–S1C).³⁴ To broadly assess myelination in these mutants, we performed in situ hybridization on whole mount embryos for myelin basic protein (mbp), a marker of myelinating glia. In the PNS, animals homozygous for mutations in the EGF exon or type III-specific exon lacked mbp expression, in accord with previous studies of zebrafish and mouse mutants,^18,27 whereas mpb expression in the PNS appeared normal in the type I and type II homozygous mutants (Figures S1D–S1I). All nrg1 mutants expressed mbp in the CNS (Figures S1D–S1I), consistent with previous analyses.^20,26,27

Figure 1. nrg1 type II signals are required for normal myelination in the spinal cord.

(A) Diagram of the zebrafish nrg1 locus. Colored rectangles indicate coding exons; open rectangles indicate UTRs. Arrows indicate alternative transcriptional start sites. Exon length is to scale; double slashes indicate intronic regions that are not to scale. Red arrowheads indicate sgRNA targets in 5′ isoform-specific exons and locations of corresponding mutations. Blue arrowhead indicates sgRNA target used to generate st150 mutation in the EGF domain, which is predicted to eliminate function of all nrg1 isoforms.

(B) Diagram of the mRNA splicing of nrg1 isoforms, illustrating the isoform-specific nature of 5′ exons and chosen sgRNAs.

(C) Confocal images of oligodendrocytes in the dorsal spinal cord at 3.5 dpf labeled using the claudink:GFP transgene. Wild-type fish have normal sheath parameters, while oligodendrocytes (OL) in EGF and type II mutants have reduced myelin. A dotted circle indicates the OL cell body, while brackets indicate the breadth of a single OL. Scale bar, 20 μm.

(D) Quantification of total myelin sheath length produced by individual oligodendrocytes in nrg1 EGF and isoform-specific lines. OLs in nrg1 EGF and type II mutants (mut) make significantly less myelin per OL than in wild-type (wt) animals, while no difference is observed in nrg1 type I or III mutant animals. Animals were genotyped after imaging. Error bars represent mean ± SD. A t test was used to assess significance, and p values were adjusted for multiple comparisons. Each point represents one OL; ≥ 30 OLs per category from ≥4 animals per category; one representative experiment shown from three replicates. Figures 1A and 1B modified from Lysko et al.³⁴ See also Figures S1 and S2.

To analyze myelin sheath formation at higher resolution, we visualized oligodendrocytes and their myelin sheaths by crossing our nrg1 EGF and isoform-specific lines to the stable transgenic reporter cldnk:GFP-CAAX,³⁵ which expresses membrane-bound GFP in all myelinating oligodendrocytes (Figure 1C). We analyzed myelin sheath formation in the dorsal spinal cord at 3.5 dpf (days post fertilization), when individual oligodendrocytes can be imaged. There was a significant reduction in the total length of myelin sheaths produced by oligodendrocytes in both the EGF and type II-specific mutants, but there was no significant difference in either the type I- or type III-specific mutants (Figure 1D). This reduction in myelin production per oligodendrocyte was a result of a significant reduction in sheath length in the EGF and type II-specific mutants, and we observed a similar reduction in erbb2 mutants (Figures S2A–S2C). While we did observe a reduction in number of sheaths per cell in the EGF mutants, the effect size is small, and we did not observe a significant reduction in sheath number in nrg1 type II or erbb2 mutants at 3.5 dpf. There was no significant reduction in dorsal oligodendrocyte number in either the EGF or type II-specific mutants (Figure S2D). These results indicate that a reduction in sheath length is the main factor reducing myelin production in nrg1 EGF and type II-specific mutants at this early stage of myelination.

To determine if these defects in myelin formation persisted at later stages, we also measured myelin sheath parameters at 5 dpf, when a greater proportion of oligodendrocytes have completed myelination.³⁶ The extensive myelination in the spinal cord at 5 dpf precludes analysis of individual oligodendrocytes with the cldnk:GFP-CAAX transgenic line. Therefore, to observe and quantitate myelination of individual oligodendrocytes at 5 dpf, we scatter-labeled individual cells by injecting the mbp:GFP-CAAX transgenic construct³⁷ into nrg1 and erbb2 mutants and their wild-type siblings. We again observed a reduction in total myelin per oligodendrocyte in nrg1 EGF, nrg1 type II-specific, and erbb2 mutants (Figures 2A–2D). In contrast to 3.5 dpf, we did not observe a reduction in sheath length, but instead we found that a reduction in sheath number was the main factor resulting in the overall myelin reduction in these mutants (Figures 2B’–2D’). In contrast to the requirement for nrg1 type III in the PNS, these results indicate that the nrg1 type II isoform is a key modulator of myelination in the developing spinal cord.

Figure 2. nrg1 type II is required for normal myelin sheath number in the spinal cord at 5 dpf.

Myelin sheaths analyzed using transient transgenic expression of mbp:GFP construct to scatter label oligodendrocytes in the dorsal spinal cord at 5 dpf.

(A) Myelin sheath number is significantly reduced in nrg1 EGF and type II mutants, similar to the reduction seen in erbb2 mutants. Scale bar represents 25 μm.

(B and C) Total sheath length per oligodendrocyte (OL) is significantly reduced in nrg1 EGF and nrg1 type II mutants.

(D) A similar reduction in total sheath length is observed in erbb2 mutants. (B’–D’) This reduction in total sheath length is driven by a reduction in sheath number per oligodendrocyte, rather than a reduction in sheath length. Animals were genotyped after imaging. Error bars represent mean ± SD. A t test was used to assess significance, and p values were adjusted for multiple comparisons. Each point represents one OL; ≥ 12 OLs per category from ≥4 animals per category; one representative experiment shown from three replicates.

nrg1 type II is required for normal myelination of diverse neuronal classes

In light of the reduction in spinal cord myelination in nrg1 type II mutants, we investigated which classes of neurons were insufficiently myelinated. In the larval zebrafish spinal cord, the most frequently myelinated neuronal classes are CoPA (commissural primary ascending) and RS (reticulospinal) neurons.⁷ CoPA interneurons send ascending projections along the dorsal spinal cord to the hindbrain, while RS neurons project from the hindbrain to targets within the dorsal and ventral spinal cord (Figure 3A).^38,39 To assess the myelination of these neuronal classes, we visualized the transgenic contactin fusion protein myelin reporter⁷ in neurons in nrg1 mutants. Transmembrane contactin (cntn) is localized to the axonal membrane, where it becomes excluded from ensheathed areas of the axon upon glial contact, revealing the “footprints” of myelin sheaths (Figure 3B; green segments). We expressed the contactin reporter using pan-neuronal drivers in transient transgenic experiments to randomly label spinal cord neurons. Neurons expressing the reporter were classified by their characteristic positions, morphologies, and projections, allowing us to analyze the myelination of the major myelinated neuronal classes in nrg1 mutants at 5.5–6 dpf. The major finding from this analysis was that both CoPA and reticulospinal neurons were myelinated significantly less frequently in nrg1 type II mutants than in wild-type siblings (Figures 3C–3E). In the minority of RS and CoPA neurons that were myelinated in nrg1 type II mutants, we observed no significant decrease in sheath number or length, and indeed sheath length was increased in the few myelinated CoPA neurons found in the nrg1 type II mutants (Figure S3). These results indicate that nrg1 type II regulates the myelination of both CoPA and reticulospinal neurons, the two most highly myelinated neuronal classes in the spinal cord.

Figure 3. nrg1 type II is required for normal myelination of diverse neuronal classes.

(A) lllustration of spinal cord neurons assessed with the myelin reporter. RS neuron cell bodies reside in the hindbrain and project posteriorly toward the tail(orange). CoPA neurons reside in the dorsal spinal cord and project anteriorly to the hindbrain (blue). RB sensory neurons reside in the dorsal spinal cord, receive touch stimulation via their sensory arbor, and project both anteriorly to the hindbrain and posteriorly (green).

(B) The UAS:GFP-2A-tdTomato-contactin myelin reporter construct expresses GFP along the entire axon, allowing characterization of neuronal type, while tdTomato-contactin is excluded from ensheathed segments of the axon, allowing assessment of myelination status; thus myelin sheaths appear green in merged images.

(C) Quantification of myelination status in CoPA and RS neurons. Percent of neurons that are myelinated indicated by bars, inset numbers indicate number of neurons assessed as myelinated or unmyelinated. For RS neurons, ≥21 neurons per genotype, ≥15 animals per genotype, 99 animals total; for CoPA neurons, ≥14 neurons per genotype, ≥11 animals per genotype, 49 animals total. Fisher’s exact test was used to assess significance.

(D) Myelin sheath pattern comparison between wild-type and nrg1 type II mutant CoPA neurons. Arrowheads indicate myelin sheaths along the CoPA subject neuron. Dotted white lines indicate the position of CoPA cell body on the contralateral side of the spinal cord, which is not visible in this image projection. The asterisk indicates CoPA cell body in nrg1 type II mutant image, while open arrowheads indicate the unmyelinated axon. Dashed gray lines indicates dorsal and ventral bounds of the spinal cord.

(E) Myelin sheath pattern comparison between wild-type and nrg1 type II mutant RS neurons. Arrowheads indicate myelin sheaths along the subject wild-type RS neuron, while open arrowheads indicate an unmyelinated RS axon in a nrg1 type II mutant. A dorsal unmyelinated RB neuron (cell body, #; sensory arbor, arrow) sends projections posteriorly and also anteriorly to the hindbrain. Animals were genotyped after imaging. Scale bar, 50 μm. See also Figure S3.

Unmyelinated Rohon-Beard sensory neurons express nrg1 type II

In the PNS, the level of Nrg1 type III expressed by a neuron regulates its myelination.^21,40 By analogy to the PNS, our analysis of myelination in the spinal cord of nrg1 type II mutants raised the possibility that myelinated neurons might express Nrg1 type II signals, perhaps at different levels corresponding to their myelination rates. nrg1 type II expression has been detected by in situ hybridization in spinal cord neurons,⁴¹ but it was not possible from these studies to determine which neuronal classes were labeled. To examine nrg1 type II expression with cellular resolution, we used two approaches: RNAScope in situ hybridization and construction of a transgenic reporter in which 3 kb of sequence directly upstream of the nrg1 type II isoform-specific exon drives expression of membrane-bound GFP (Figure 4A). We used RNAScope probes to detect both endogenous nrg1 type II mRNA and GFP mRNA in stable transgenic nrg1 type II reporter embryos (Figures 4B and 4C). Probes for endogenous nrg1 type II mRNA localized to three distinct locations at 21–24 hpf: along the dorsal spinal cord, in the notochord, and in the region of motor neurons in the ventral spinal cord. nrg1 type II expression in the ventral spinal cord has been previously detected by traditional in situ hybridization.⁴¹ We observed close co-expression of endogenous nrg1 type II mRNA and GFP mRNA from the nrg1 type II reporter in both the dorsal spinal cord and notochord, indicating that the nrg1 type II transgenic reporter accurately reflects endogenous nrg1 type II expression in these areas. We did not detect GFP mRNA or protein from the reporter in the region of the motor neurons (Figures 4B’, 4C, and 4D), indicating that other regulatory sequences must activate nrg1 type II expression in these cells. Confocal microscopy of the nrg1 type II reporter line at this stage revealed prominent expression of GFP in Rohon-Beard (RB) sensory neurons (Figures 4D–4F), but no GFP expression in CoPA, RS, or other spinal cord neurons. RB neurons are specified early in neurogenesis (Figure 4D) and have large somas located in the dorsal spinal cord (Figures 4D and 4E). In the embryo, RB neurons form characteristic sensory arbors (arrow, Figure 4F), and they send descending projections toward the tail and ascending projections to the hindbrain (Figure 4H) via a dorsal axonal tract (Figures 4E and 4G) called the dorsal longitudinal fasciculus.^42,43 Interestingly, almost all (>95%) RB neurons are unmyelinated, and the small number of others are myelinated by a single short sheath per axon.⁷ RB axons expressing the nrg1 type II reporter pass near oligodendrocytes in the dorsal spinal cord (Figure 4G), but course up to 50 micrometers away from the ventral spinal cord. In nrg1 type II mutants labeled randomly with the contactin myelination reporter, we observed no significant change in the number of RB neurons (wild-type: 19 RBs; mutant: 21 RBs per 100 fish analyzed) or their myelination status (wild-type, 96% unmyelinated; mutant, 91% unmyelinated; p = 0.64, Fisher’s exact test). In addition, nrg1 type II mutant larvae are responsive to trunk touch stimuli, a reaction dependent on the function of RB neurons.^43,44 These data indicate that nrg1 type II is not required for RB development or myelination but is instead required for the normal myelination of both neighboring and distant axons that do not themselves express nrg1 type II.

Figure 4. Unmyelinated RB sensory neurons express nrg1 type II.

(A) Diagram indicating location of 3 kb regulatory element and reporter construct design. Colored rectangles indicate coding regions; open rectangles indicate UTRs. Arrows indicate transcriptional start sites.

(B) Endogenous nrg1 type II mRNA detected by RNAScope in situ hybridization at 24 hpf. White arrowheads mark expression in dorsal spinal cord. Black arrowheads mark expression in the region of motor neurons in the ventral spinal cord. Expression is also evident in the notochord (nc). Dotted lines mark the dorsal and ventral extent of the spinal cord in the posterior trunk and anterior tail. Scale bar, 100 μm. (B’) In the same animal as (B), GFP mRNA from the nrg1 type II reporter transgene was also detected by RNAScope in situ hybridization. (B”) Merge of (B) and (B’). Co-expression of GFP mRNA with endogenous nrg1 type II mRNA is observed in both the dorsal spinal cord and notochord. Representative images from five replicate experiments.

(C) Co-expression of endogenous nrg1 type II mRNA and GFP mRNA from the nrg1 type II reporter transgene. White arrowheads indicate putative cell bodies of three GFP-expressing Rohon-Beard (RB) sensory neurons, which also express endogenous nrg1 type II mRNA. Black arrowheads indicate two areas of endogenous nrg1 type II mRNA expression in the region of motor neurons. Scale bar, 50 μm.

(D) GFP from the nrg1 type II reporter is expressed prominently in early-born Rohon-Beard (RB) neurons in the dorsal spinal cord (arrowheads; colocalized with the pan-neuronal marker NBT:dsRed⁴⁵) and transiently in the notochord (nc) at 22 h postfertilization. Scale bar, 200 μm.

(E) Large, GFP-expressing RB neurons send axons (arrowhead) anteriorly and posteriorly via the dorsal longitudinal fasciculus (DLF). Reporter expression alsooccurs in the posterior lateral line nerve (asterisk), which grows posteriorly at 1.5 dpf. Scale bar, 50 μm.

(F) The nrg1 type II reporter is expressed in neurons with large cell bodies and sensory arbors (arrow) characteristic of RB neurons at 1.5 days post fertilization (1.5 dpf). White arrowheads indicate the DLF as in (E). Scale bar, 50 μm.

(G) GFP-expressing axons from RB neurons (white arrowhead) are near oligodendrocytes (black arrowheads) in the dorsal spinal cord. Scale bar, 50 μm.

(H) GFP-expressing axons in the DLF (arrowhead) project anteriorly into the hindbrain (hb). GFP-expressing lateral line nerve descends from the posterior lateral line ganglion (#). Scale bar, 200 μm.

nrg1 type II is required in RB neurons to regulate myelination of other neurons in the spinal cord

To determine if nrg1 type II is required in RB neurons to regulate myelination in the spinal cord, we adopted a cell type-specific CRISPR approach,^46–48 in which an sgRNA is expressed ubiquitously and Cas9 is expressed in a cell type of interest, with the intent to eliminate nrg1 function specifically in RB neurons. In transient transgenic experiments with fluorescent reporters, we compared the specificity of three elements expressed in RB neurons, isl1(ss),⁴⁹ isl1(zCREST3),⁵⁰ and ngn1(LSEC).⁵¹ At developmental timepoints preceding oligodendrocyte myelination in the spinal cord (Figure 5A), we determined that the zCREST3 element is highly specific to RB neurons at 1, 2, and 3 dpf (≥96% of neurons labeled were RB neurons, Figure S4).

Figure 5. nrg1 type II signals are required in RB neurons for normal myelination in the spinal cord.

(A) lllustration showing the location of the zCREST3 upstream of the islet1 gene. In transient transgenic experiments, zCREST3:dsRed shows exclusive expression in RB neurons with characteristic sensory arbors. Scale bar, 100 μm.

(B) Diagram of cell type-specific CRISPR construct design, allowing RB expression of Cas9, and ubiquitous expression of EGF or isoform-specific nrg1-targeted sgRNAs.

(C) Confocal images of oligodendrocytes (OL) in the dorsal spinal cord labeled using the claudink:GFP transgene. A dotted circle indicates the OL cell body. Scale bar, 20 μm.

(D) Quantification of total myelin sheath length produced by individual OLs in animals expressing RB-specific Cas9, with indicated nrg1 EGF and isoform-specific sgRNAs. Each point represents an individual OL; at least 30 OLs quantified from ≥3 animals per category; one representative experiment shown from three replicate experiments. Error bars represent mean ± SD. A t test was used to assess significance, and p values were adjusted for multiple comparisons. See also Figures S4, S5, and S6.

To determine if nrg1 type II function is required in RB neurons for myelination of other neuronal classes, we used the zCREST3 element to express Cas9 in RB neurons, while simultaneously expressing guide RNAs targeting the common or isoform-specific exons of nrg1 (Figure 5B). We injected these constructs into wild-type fish expressing cldnk:GFP and measured dorsal oligodendrocyte myelin sheaths, as in our analysis of stable nrg1 mutant lines (Figure 1). We observed normal amounts of myelin per oligodendrocyte in animals injected with a negative control construct and in animals injected with constructs targeting nrg1 type I and III in RB neurons. In contrast, there was a significant reduction in myelin per oligodendrocyte in fish injected with constructs targeting either the common EGF exon or the nrg1 type II-specific exon in RB neurons (Figures 5C and 5D, compare with Figure 1D). We also performed this analysis using the ngn1(LSEC) element to express Cas9 and observed very similar results (Figure S6), indicating that different approaches to eliminate nrg1 type II function in RB neurons caused a similar reduction in myelination in the developing spinal cord. These results demonstrate that nrg1 type II function is required in RB neurons for myelination of other neuronal classes in the developing spinal cord.

erbb2 receptor is required in neurons, but not oligodendrocytes, for normal myelination in the spinal cord

The foregoing results indicate that myelination of CoPA and RS neurons is regulated by Nrg1 type II expressed in RB neurons. To determine which cells may be responding to Nrg1 type II, we sought to disrupt ErbB receptor tyrosine kinases, which are activated by Nrg1 signals. We have previously characterized the PNS in zebrafish erbb2^st61 mutants and erbb3^st48 mutants, in which Schwann cell migration and myelination are blocked.^26,27 In the CNS, we found a significant reduction in myelin sheath length per oligodendrocyte in erbb2^st61 mutants at 3.5 dpf and 5 dpf (Figures 2D, 6A, and 6B), similar to the phenotypes of nrg1 EGF and type II mutants. In contrast, erbb3 mutants had normal myelination in the dorsal spinal cord (Figure S7). We used the cell-specific CRISPR approach to test if erbb2 function was required in neurons or oligodendrocytes. We ubiquitously expressed an erbb2 sgRNA, while simultaneously expressing Cas9 either pan-neuronally, in RB neurons with the LSEC element as in Figure S6, or in oligodendrocytes (Figure 6C). Interestingly, there was a significant reduction in myelin per oligodendrocyte in fish injected with the pan-neuronal construct targeting erbb2, but not with the constructs driving Cas9 expression in RB neurons or oligodendrocytes (Figure 6D and 6E). Together these data indicate that Nrg1 type II produced by RB neurons signals through ErbB2 in other neurons to regulate spinal cord myelination.

Figure 6. erbb2 receptor is required in neurons, but not oligodendrocytes, for normal myelination in the spinal cord.

(A) Confocal images show that claudink:GFP-labeled oligodendrocytes (OL) in erbb2 mutants generate less myelin than wild-type controls. Scale bars, 20 μm.

(B) Quantification of myelin produced by individual OLs in animals wild-type, heterozygous, or mutant for erbb2^st61. OLs in animals lacking erbb2 produce significantly less myelin. Each point represents one OL; at least 40 OLs analyzed from ≥6 animals per genotype; one representative experiment shown from three replicates. Animals were genotyped after imaging.

(C) Diagram of cell type-specific CRISPR construct design, allowing expression of Cas9 in OLs (claudin), all neurons (NBT), or RBs (LSEC), and ubiquitous expression of erbb2-targeted sgRNA.

(D) Confocal images of oligodendrocytes in the dorsal spinal cord labeled using the claudink:GFP transgene. Dotted circle indicates the OL cell body, while brackets indicate the breadth of a single OL.

(E) erbb2 cell type-specific CRISPR quantification. Total myelin sheath length produced by individual oligodendrocytes in animals expressing Cas9 in neurons is significantly less than in the other conditions. No significant change in myelin production occurred in animals with erbb2 knockdown in RB neurons or oligodendrocytes. Each point represents one OL; at least 15 OLs quantified from ≥4 animals per category; one representative experiment shown from three replicates. Error bars represent mean ± SD. A t test was used to assess significance, and p values were adjusted for multiple comparisons. See also Figures S5 and S7.

DISCUSSION

Our results reveal a previously unappreciated function for Nrg1 type II signals and their ErbB2 receptors in myelination of axons in the CNS. RB sensory neurons in the spinal cord prominently express nrg1 type II mRNA, and cell type-specific CRISPR experiments indicate that nrg1 type II signals are required in these unmyelinated RB neurons to regulate the ensheathment of neighboring neurons. In vivo imaging indicates that myelination is reduced, but not eliminated, in two different classes of myelinated interneurons in nrg1 type II mutants. Thus, our analysis demonstrates that nrg1 type II signals are required in unmyelinated sensory neurons to regulate the extent of myelination of other neuronal classes in the spinal cord.

Our study defines a function for nrg1 type II signals that is distinct from prior work on Nrg1 isoforms in CNS myelination.^20,28 One prior study reported that Nrg1 type III heterozygous mutant mice have thinner myelin in the brain but not the spinal cord, identifying a potential role for Nrg1 type III signals in myelination in the CNS.²⁸ Nrg1 type III signals have also been implicated in increasing myelin in response to social cues,²⁹ but the exact location and mode of action of Nrg1 type III has not been defined. Our analysis of the zebrafish spinal cord demonstrated a requirement for nrg1 type II signals, which have not been examined in published mouse knockout studies of CNS myelination. In addition, our analysis revealed that nrg1 expression in one class of neurons affects the myelination of another—an observation enabled by the use of cell type-specific knockdown approaches and in vivo myelination reporters. Consistent with prior work,^20,28 our analysis revealed normal myelination in the spinal cord of nrg1 type III mutants and nrg1 type I mutants, but future studies of isoform-specific mutants may investigate possible roles for these signals in other areas of the CNS.

The mode of action that we define for Nrg1 type II proteins in the spinal cord is also quite distinct from the well-studied function of Nrg1 type III proteins in the PNS. In the PNS, Nrg1 type III is essential for many steps of Schwann cell development and myelination, including proliferation, migration, radial sorting, and myelin formation.^17,21,27,52 Whereas Nrg1 type III mutants have little or no myelin in the PNS, our analysis shows that myelination is only partly reduced in the spinal cord of nrg1 type II mutants. Thus, our evidence indicates that nrg1 type II modulates CNS myelination. Functional analyses of ErbB receptors reveal another important difference between the CNS and PNS: in the PNS axonal Nrg1 type III signals activate an ErbB2-ErbB3 heteromeric receptor on closely associated Schwann cells, whereas in the spinal cord, we show that myelination requires ErbB2 receptors not in oligodendrocytes, but in other neurons. In addition, our analysis revealed no role for ErbB3 in spinal cord myelination. Despite these differences, Nrg1 signals appear to exert exquisite control on myelination in both the CNS and the PNS; in both cases heterozygous mutants myelinate in proportion to Nrg1 dosage, indicating marked sensitivity to the levels of Nrg1 signals (Figures 1D and 2B–2D).^20,21,28

Based on nrg1 type II expression and our cell-specific CRISPR experiments, we propose that Nrg1 type II proteins expressed by RB neurons activate ErbB2 receptors on other neurons. Knockdown of nrg1 type II signals in RB neurons produces a similar phenotype to loss of erbb2 function in other neurons, suggesting that ErbB2 may be the main receptor transducing Nrg1 type II signals in this context. ErbB2 does not directly bind to Nrg1 proteins,^53,54 so it is likely that the active receptor is a heterodimer of ErbB2 with ErbB4. Although our results highlight a role for erbb2 in neuron-to-neuron signaling in myelination, we do not discount the possibility that Nrg1 type II signals might act directly on oligodendrocytes as well, through other ErbB receptor combinations.

A key outstanding question is the mechanism by which Nrg1 type II signals are transmitted to promote spinal cord myelination. Our results indicate that dorsally located sources of Nrg1 type II signals (i.e., RB neurons) promote myelination of both dorsal and ventral axons that project from distantly located neurons in the hindbrain and spinal cord. Because the Ig-like domain of Nrg1 type II binds extracellular matrix at synapses in other contexts, limiting diffusion and potentially increasing the signaling potency,^33,55,56 it seems likely that Nrg1 type II acts to promote myelination via a short-range mechanism. For example, RB, CoPA, and RS neurons all participate in the spinal cord escape response circuit, and it is possible that RB neurons express Nrg1 type II signals to regulate the myelination status of their synaptic partners. Touch activates RB sensory neurons to trigger motor neuron activation and a characteristic escape behavior.^44,57 RB axons ascend to the hindbrain, where RS cell bodies reside^58,59 and are activated in response to touch on the trunk.^60,61 RB axons also synapse onto CoPA neurons,^59,62–64 which modulate the escape response circuit.^64,65 Thus, as the circuit forms, Nrg1 type II from RB neurons could signal RS and CoPA neurons to increase their myelination. Myelination is modulated to coordinate the output of circuits of disparate length in different contexts, including the cerebellum, retina, and fish electromotor systems.^1,5,66,67 RB neurons along the spinal cord project varying distances to their targets in the hindbrain, so perhaps Nrg1 type II signals modulate myelination to optimize escape response circuits.

This study provides evidence that Nrg1 type II signals modulate myelination of distinct neuronal classes in the spinal cord. Our nrg1 type II reporter and cell-specific CRISPR experiments support a model in which unmyelinated neurons express Nrg1 type II signals to regulate myelination of neighboring interneurons, a mode of action that may coordinate function of circuits in the CNS involving both unmyelinated and myelinated neurons.

Limitations of the study

In this study we identify a previously unappreciated role for nrg1 type II signals in modulating spinal cord myelination, but different mechanisms may regulate myelination at other stages and elsewhere in the CNS. Although myelination is reduced in the spinal cord of nrg1 type II mutants by early larval stages, we do not know precisely when the nrg1 type II signals are required in RB neurons, nor do we know the cellular mechanism through which signaling occurs. Cell type-specific Cas9 expression, which supported our conclusion that nrg1 type II is required in RB neurons, has limitations, including both possible off-target effects and false-negative results. We took several measures to mitigate these concerns, including assessing the activity of regulatory elements used to express Cas9 at multiple stages, using different elements to drive expression in RB neurons, ensuring that all sgRNA employed were highly active, and comparing all cell type-specific CRISPR phenotypes to well-characterized germline mutants.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, William Talbot (wtalbot@stanford.edu).

Materials availability

Reagents generated in this study will be made available on request to the lead contact.

Data and code availability

We have deposited DNA sequences to GenBank: OP690417-OP690424; OP680564-OP680578. Additional DNA sequence information and microscopy images are available upon request. This paper does not report original code. Any additional information required to repeat experiments are available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Zebrafish lines and maintenance

All zebrafish experiments were conducted under protocols approved by the Stanford University institutional animal care and use committee and conforming to appropriate city, state, and national regulations. Prior to experimental procedures embryos and larvae were anesthetized with 0.016% Tricaine. To inhibit pigmentation, embryos and larvae were treated with 1-phenyl 2-thiourea (PTU, 0.003%). Embryos and larvae were analyzed up to 6 dpf, before the onset of sexual differentiation.

METHOD DETAILS

nrg1 mutations and genotyping

The nrg1 alleles st150, st151, st152, and st153 were generated by injecting sgRNAs (Table S1) and Cas9 mRNA into wildtype embryos at the one-cell stage. Injected fish were raised to adulthood and outcrossed to establish stable lines bearing variants in nrg1. The nrg1 type III st153 allele has been previously described.³⁴ The first EGF exon common to all known isoforms was targeted to create st150; a 4 bp deletion in the EGF domain that introduces a frameshift and premature stop (P264R_fs*270) predicted to disrupt the function of all nrg1 isoforms. Similar approaches were used to target isoform-specific exons, as illustrated in Figure 1. Sequence data for lesions are shown in Figure S1, G-I: (st151: 29 bp deletion[−32 + 3], A6E_fs*8), (st152: 31 bp deletion, V44C_fs*63), (st153: 7 bp deletion, P40T_fs*68). The larger deletions st151 (nrg1 type I) and st152 (nrg1 type II) were readily genotyped using primers (Table S1) that generate a shorter product from the mutant allele that is distinguishable from the wildtype allele on an agarose gel. st150 disrupts a BsaJI restriction enzyme site; PCR primers oAMS_618/619 amplify a 430 bp fragment from the wildtype allele that BsaJ1 cleaves into three fragments (208, 158 and 64 bp), while the mutant product is cleaved into two fragments (218, 208 bp). st153 disrupts a BslI site; primers CRD_L2/R2 amplify a 488 bp fragment from the mutant allele that is cleaved by BslI into three fragments (304, 99, 77 bp), whereas the wildtype product is cleaved into four fragments (235, 99, 77 and 77 bp). Genotyping primers are described in Table S1. DNA sequences have been deposited in GenBank.

To examine nrg1 mRNA splicing in these mutants, we isolated RNA from whole individual fish at 3.5 dpf (RNeasy Plus Micro, Qiagen), amplified cDNA (SuperScript IV, Invitrogen), and used primers to amplify across all exon-boundaries from the 5′ isoform-specific exons to the first EGF exon. We detected expected splicing of all isoforms in each isoform-specific mutant and the EGF mutant lines, and we detected no evidence of cryptic splice sites activated by the lesions. DNA sequences have been deposited in GenBank.

In situ hybridization and RNAScope

In situ hybridization for mbp mRNA in whole-mount larvae was performed using standard methods.^26,74 To probe for nrg1 type II and GFP mRNA, we followed the RNAScope protocol of Gross-Thebing et al.,^75,76 using a custom-generated nrg1 type II probe set directed at the nrg1 type II-specific first exon, in combination with probes for EGFP (ACD). 22–36hpf animals were anesthetized with Tricaine and fixed in 4% PFA in PBS for 40 min, washed and stored in methanol overnight at −2°C. Animals were rehydrated from 100% to 0% MeOH in 25% steps using PBS-0.01% Tween (PBSTw). Animals were treated with Protease III for 20 min, washed 3X with PBSTw, and probed overnight in horizontal tubes in a 40°C water bath. Subsequent washing steps were performed with 0.2X SSC, 0.01% Tween at room temperature while fluorescent detection reagent steps were performed in horizontal tubes in a 4°C water bath. Animals were mounted in agarose and imaged using a Zeiss LSM700.

Oligodendrocyte sheath analysis

The stable transgenic line claudink:Gal4-UAS:GFP-CAAX³⁵ was crossed to nrg1 mutant lines and used to visualize the myelin sheaths of individual oligodendrocytes at 3.5 dpf. At this stage, a number of oligodendrocytes have migrated to the dorsal region of the spinal cord and many have completed the process of myelin sheath initiation and maturation.³⁶ In this dorsal region, all the sheaths generated by an oligodendrocyte can be visualized and analyzed discretely from neighboring oligodendrocytes.

The extensive myelination in the spinal cord at 5 dpf precludes analysis of individual oligodendrocytes with the cldnk:GFP-CAAX transgenic line. For sheath analysis at 5 dpf, we injected mbp:GFP-CAAX³⁷ to scatter-label individual oligodendrocytes in transiently transgenic animals, as done previously.³⁶ At both 3.5 and 5 dpf, we anesthetized and embedded live larvae in agarose and imaged using a Zeiss LSM700 or Leica SP5 confocal microscope, blind to genotype. We imaged oligodendrocytes in the dorsal spinal cord in the region of somites 4–12. After imaging, fish were genotyped. Sheaths belonging to an individual oligodendrocyte were quantitated in Fiji/ImageJ⁷³ by drawing a segmented line along the length of each sheath and measuring length; a script in Rstudio collated sheath parameters per cell.

Antifreeze protein 3′UTR construct cloning

p3E-afp_UTR (pAMS448) was created and used as a 3′ Entry clone for LR-generated constructs in this study. Sequence encoding the ocean pout antifreeze (afp) protein 3′UTR⁷⁷ was ordered as a gBlock (IDT) and cloned into pDonP2R-P3 using Gibson cloning (NEB).

Myelin reporter analysis

The GFP-cntn1a and GFP-2A-tdTcntn1a myelin reporters have been previously used to visualize myelin sheaths on individual neurons.^7,71 We used a multiple-vector strategy in transient transgenic animals in order to visualize neuronal type and myelination status in our stable nrg1 mutant lines. We utilized Tol2kit vectors⁷⁰ that employ Gateway cloning to construct all Tol2 transgenic vectors in this paper. We created pTol2_6XUAS:GFP-cntn1a (pDEL73) using LR cloning, and received the kind gift of pTol2_10XUAS:GFP-2A-tdTcntn1a⁷¹ to express the contactin myelin reporter. To drive pan-neuronal expression of these vectors we created pTol2_NBT:Kal4 (pDEL72) and pTol2_HuC:Kal4 (pDEL102) using LR cloning (p5E-NBT,^45,78 p5E-HuC,⁶⁸ pME-KalTA4⁶⁹). To randomly label neurons we injected 0.0006 ng each of a Kal4 driver vector and a myelin reporter vector with Tol2 mRNA into embryos at the one-cell stage. Neuronal type and myelin sheaths were imaged at 5.5–6 dpf using a Zeiss LSM700 blind to genotype. After imaging, larvae were genotyped. Neuronal type, and myelination status were determined, and data were collated and analyzed using scripts in RStudio.

nrg1 type II-reporter transgene

Using zebrafish genomic DNA prepared using DNeasy Blood & Tissue Kits (Qiagen) as a template, we amplified two overlapping fragments covering approximately 3kb of DNA upstream from the nrg1 type II translational initiation codon using KAPA polymerase (Roche), and the following primer pairs: (T2_3055_G_F1; 1677_R) and (1399_F; T2_G_R8). Fragments were joined using overlapextension PCR and inserted into pDonrP4-P1R using Gibson cloning (NEBuilder HiFi) to create a p5E Gateway vector (pDEL61). Using LR cloning we created the Tol2 vector pDEL63, in which sequences from upstream of nrg1 type II drive expression of membrane bound GFP, with a cmlc2:GFP marker of transgenesis. Tol2 mRNA and pDEL63 were injected into embryos at the one-cell stage, and animals with GFP signal in the heart were raised to adulthood. Stably expressing lines were established from the F0 generation and screened for GFP expression outside of the heart.

Rohon-Beard-specific regulatory element LSEC cloning

A short section of the lateral stripe element (LSE) upstream of the coding region of the neurogenin1 gene previously shown⁵¹ to drive expression in developing Rohon-Beard neurons was amplified with the following primer pair: (ng1_LSE_C_F1; ng1_LSE_C_F1). This fragment was inserted into pDonrP4-P1R as above to create p5E_ngn1_LSE_C (pDEL98; hereafter referred to as LSEC, describing a minimal “critical” region expressed in RBs). pME_GFP-CAAX was generated using Infusion cloning (Takara) and combined with pDEL98 to create the expression vector pTol2_LSEC:GFP-CAAX, used to assess the strength and specificity of LSEC expression.

Assessing RB-specificity of zCREST, LSEC, and isl1(ss) regulatory elements

The specificity of expression of these gene regulatory elements was assessed in transient transgenic experiments as above. zCREST3:dsRedexpress⁷² (p826), zCREST3:mCherry⁷² (p456), LSEC:GFP (pDEL116), isl1(ss):GFP⁷² (p11) and isl1(ss):RFP⁷² (p12) were injected individually with Tol2 mRNA at the one-cell stage. Embryos were mounted and spinal cord neuron type and number were counted at 1, 2, and 3 days post fertilization using a Zeiss LSM700.

Cell-type-specific CRISPR

We used previously described Gateway-cloning compatible Tol2 constructs that express Cas9 using a tissue-specific promoter and gene-specific sgRNAs using the U6 promoter; cmlc2:GFP was included as a marker of transgenesis.^46,47 sgRNAs are described in Table S1; nrg1 sgRNAs are the same as those used to generate the stable lines, with the exception of nrg1 type III. A new sgRNA (coDEL_161) for nrg1 type III was designed to avoid a SNP in the previously targeted type III-specific sequence that might lower cutting efficiency. All sgRNAs used in cell-type specific CRISPR experiments caused extensive cleavage of their target sites in embryos injected with sgRNA and Cas9 protein, as evaluated at 1.5 dpf by observing a high degree of mixed sequence in Sanger sequencing of the amplified target regions (Figure S5). Four erbb2 sgRNAs were designed and evaluated for efficiency by injecting sgRNA and Cas9 protein (UC Berkeley) into wildtype embryos at the one-cell stage and assessing mbp expression along the peripheral lateral line nerve by in situ hybridization at 4.5 dpf; the most efficient sgRNA (coDEL_171) was used in subsequent experiments. Complementary oligos encoding the sgRNAs were annealed and cloned into the BseRI site of the cell-specific CRISPR destination vector using ligase (NEB). We used a vector with non-specific sgRNA sequence as control; this vector will allow cell-type specific Cas9 expression along with an sgRNA that is not targeted to any site in the zebrafish genome. We used the zCREST3^50,72 and LSEC⁵¹ element to drive expression of Cas9 in Rohon-Beard neurons, the NBT regulatory element to drive expression pan-neuronally,^45,78 and the claudink regulatory element to drive expression in oligodendrocytes.³⁵ All cell-type specific CRISPR Tol2 expression vectors (pDEL125-129; pDEL133-135; pDEL142-143; pDEL157-161) were created using LR cloning as above. To analyze the effect on myelin sheaths, we injected cell-specific CRISPR constructs and Tol2 mRNA early during the one-cell stage into cldnk:GFP stable transgenic embryos, selected fish with efficient construct expression (GFP in heart) at 3.5 dpf, and imaged myelin sheaths as above in these F0 animals. Reporter analysis indicates that zCREST3 drives expression specifically in RB neurons (Figure S4). Additional evidence that this approach inactivates nrg1 type II in a specific cell population derives from Sanger sequencing and SYBR Green melt curve analysis of embryos injected with the zCREST:Cas9 or NBT:Cas9 targeting constructs. These approaches detected no cleavage of nrg1 type II target sequences in whole animals injected with the cell type specific targeting constructs, presumably because RB neurons are a very small fraction of the total embryo; extensive cleavage was evident in positive control embryos injected with the same sgRNA and Cas9 protein (Figure S5).

SYBR GREEN melt curve analysis

SYBR Green melt curve analysis was performed using SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) on an Applied Biosystems QuantStudio5 PCR machine (ThermoFisher).

QUANTIFICATION AND STATISTICAL ANALYSIS

Graphs and statistical analysis were done in GraphPad Prism. Data were first tested for normality using D’Agostino & Pearson, Anderson-Darling, and Shapiro-Wilk normality tests. If normal, comparison of means was done with ANOVA and t test for pairwise comparisons, adjusting the p value in the case of multiple comparisons. Non-normal groups were compared using the Mann-Whitney test. For comparison of proportions of myelinated axons, Fisher’s exact test was used. Exact p values are reported. Error bars indicate mean ± standard deviation.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Critical commercial assays
RNAscope® Probe-EGFP	Advanced Cell Diagnostics	Cat No. 400281-C2
RNAscope® Probe-Dr-nrg1-O2-C1	Advanced Cell Diagnostics	Cat No. 1153491-C1; NM_001277084.1
RNAscope® Fluorescent Multiplex Kit v1	Advanced Cell Diagnostics	Cat No. 320851
Deposited data
nrg1 type I cDNA seq from wt	this paper	GenBank: OP680574
nrg1 type I cDNA seq from st150 mutant	this paper	GenBank: OP680575
nrg1 type I cDNA seq from st151 mutant	this paper	GenBank: OP680576
nrg1 type I cDNA seq from st152 mutant	this paper	GenBank: OP680577
nrg1 type I cDNA seq from st153 mutant	this paper	GenBank: OP680578
nrg1 type II cDNA seq from wt	this paper	GenBank: OP680569
nrg1 type II cDNA seq from st150 mutant	this paper	GenBank: OP680570
nrg1 type II cDNA seq from st151 mutant	this paper	GenBank: OP680571
nrg1 type II cDNA seq from st152 mutant	this paper	GenBank: OP680572
nrg1 type II cDNA seq from st153 mutant	this paper	GenBank: OP680573
nrg1 type III cDNA seq from wt	this paper	GenBank: OP680564
nrg1 type III cDNA seq from st150 mutant	this paper	GenBank: OP680565
nrg1 type III cDNA seq from st151 mutant	this paper	GenBank: OP680566
nrg1 type III cDNA seq from st152 mutant	this paper	GenBank: OP680567
nrg1 type III cDNA seq from st153 mutant	this paper	GenBank: OP680568
nrg1 genomic DNA seq from wt at st150 lesion	this paper	GenBank: OP690417
nrg1 genomic DNA seq from st150 mutant at st150 lesion	this paper	GenBank: OP690418
nrg1 genomic DNA seq from wt at st151 lesion	this paper	GenBank: OP690419
nrg1 genomic DNA seq from st151 mutant at st151 lesion	this paper	GenBank: OP690420
nrg1 genomic DNA seq from wt at st152 lesion	this paper	GenBank: OP690421
nrg1 genomic DNA seq from st152 mutant at st152 lesion	this paper	GenBank: OP690422
nrg1 genomic DNA seq from wt at st153 lesion	this paper	GenBank: OP690423
nrg1 genomic DNA seq from st153 mutant at st153 lesion	this paper	GenBank: OP690424
Experimental models: Organisms/strains
nrg1 EGF mutant	this paper	st150
nrg1 type I mutant	this paper	st151
nrg1 type II mutant	this paper	st152
nrg1 type III mutant	this paper	st153
Tg([-3kb]nrg1 typeII:GFP-CAAX)	this paper	st162
Tg(claudink:GAL4-UAS:GFP-CAAX)	(Münzel et al. 2012)35	N/A
Tg(NBT:dsRed)	(Peri and Nüsslein-Volhard 2008)45	zf148Tg
Oligonucleotides
For all oligonucleotides used, see Table S1	This paper	N/A
Recombinant DNA
pTol2_mbp:GFP-CAAX	(Almeida et al. 2011)37	N/A
p5E-NBT	(Peri and Nüsslein-Volhard 2008)45	N/A
p5E-HuC	(Shiau et al. 2013)68	N/A
pME-KalTA4	(Almeida and Lyons 2015)69	N/A
p3E-afp_UTR	this paper	pAMS448
Tol2kit	(Kwan et al. 2007)70	N/A
pTol2_NBT:Kal4	this paper	pDEL72
pTol2_HuC:Kal4	this paper	pDEL102
p5E-6XUAS	(Almeida and Lyons 2015)69	N/A
pCR8/GTW-GFP_cntn1a	(Koudelka et al. 2016)7	N/A
pTol2_6XUAS:GFP-cntn	this paper	pDEL73
pTol2_6XUAS:tdT-cntn	this paper	pDEL74
pTol2_10UAS-EGFP-2A-zftdTomato-cntn1a_crymCherry	(Almeida et al. 2021)71	N/A
p5E_(−3kb)nrg1 type II	this paper	pDEL61
pTol2_(−3kb)nrg1_typeII:GFP-CAAX	this paper	pDEL63
p5E_claudinK	(Münzel et al. 2012)35	N/A
p5E_ngn1_LSE_C	this paper	pDEL98
pME_GFP-CAAX	this paper	pELB01
pTol2_LSEC:GFP-CAAX	this paper	pDEL116
pME_memtagRFPt T2A cas9	(Marshall-Phelps et al. 2020)47	N/A
pME_cas9_2A_GFP	Addgene	#63155; RRID:Addgene_63155
pDest_U6_A2_CG2	Addgene	#63156; RRID:Addgene_63156
pDest_U6_A2_CG2_typeEGF_gRNA	this paper	pDEL110
pDest_U6_A2_CG2_T1_gRNA	this paper	pDEL111
pDest_U6_A2_CG2_T2_gRNA	this paper	pDEL112
pDest_U6_A2_CG2_T3_gRNA	this paper	pDEL113
pTol2_ngn1:redCas9; control_sgRNA (LSEC)	this paper	pDEL127
pTol2_ngn1:Cas9green; EGF_sgRNA (LSEC)	this paper	pDEL128
pTol2_ngn1:Cas9green; T1_sgRNA (LSEC)	this paper	pDEL126
pTol2_ngn1:Cas9green; T2_sgRNA (LSEC)	this paper	pDEL129
pTol2_ngn1:Cas9green; T3_sgRNA (LSEC)	this paper	pDEL125
p5E_zCREST3	(Palanca et al. 2013)72	p410
zCREST3:redCas9; control_gRNA	this paper	pDEL157
zCREST3:Cas9green; EGF_gRNA	this paper	pDEL158
zCREST3:Cas9green; T1_gRNA	this paper	pDEL159
zCREST3:Cas9green; T2_gRNA	this paper	pDEL160
zCREST3:Cas9green; T3_gRNA	this paper	pDEL161
pDest_U6_A2_CG2_erbb2_gRNA171	this paper	pDEL133
pTol2_claudink:Cas9green; control_gRNA	this paper	pDEL143
pTol2_NBT:Cas9green; erbb2_gRNA171	this paper	pDEL134
pTol2_LSEC:Cas9green; erbb2_gRNA171	this paper	pDEL135
pTol2_claudink:Cas9green; erbb2_gRNA171	this paper	pDEL142
pTol2_CREST3(ss):lexA; lexAOP:dsRedexpress	(Palanca et al. 2013)72	p826
pTol2_CREST3(ss):Gal4; UAS:mCh	(Palanca et al. 2013)72	p456
pTol2_isl1(ss):Gal4; UAS:GFP	(Palanca et al. 2013)72	p11
pTol2_isl1(ss):Gal4; UAS:RFP	(Palanca et al. 2013)72	p12
Software and algorithms
Fiji	(Schindelin et al. 2012)73	RRID: SCR_002285
GraphPad Prism	GraphPad Software	RRID:SCR_002798
Adobe Illustrator	Adobe	RRID: SCR_010279
Rstudio	RStudio	RRID:SCR_000432
R	R Project for Statistical Computing	RRID:SCR_001905
Snapgene	Snapgene	RRID:SCR_015052

Highlights.

• nrg1 type II is required for normal myelination of diverse neuronal classes

• nrg1 type II is expressed in unmyelinated sensory neurons

• Myelination of interneurons requires nrg1 type II function in unmyelinated neurons

• The Nrg1 receptor erbb2 is required in neurons for normal myelination