|Home | About | Journals | Submit | Contact Us | Français|
S.K. produced all of the data except for the electron microscopy images shown in Figure 5c,d, which were produced by E.W. N.T. created the Tg(sox10(7.2):mrfp) transgenic line and H.-C.P. created the Tg(olig2:dsred2) transgenic line. B.A. supervised the experiments and wrote the manuscript with S.K. and K.B.
Motor function requires that motor axons extend from the spinal cord at regular intervals and that they are myelinated by Schwann cells. Little attention has been given to another cellular structure, the perineurium, which ensheaths the motor nerve, forming a flexible, protective barrier. Consequently, the origin of perineurial cells and their roles in motor nerve formation are poorly understood. Using time-lapse imaging in zebrafish, we show that perineurial cells are born in the CNS, arising as ventral spinal-cord glia before migrating into the periphery. In embryos lacking perineurial glia, motor neurons inappropriately migrated outside of the spinal cord and had aberrant axonal projections, indicating that perineurial glia carry out barrier and guidance functions at motor axon exit points. Additionally, reciprocal signaling between perineurial glia and Schwann cells was necessary for motor nerve ensheathment by both cell types. These insights reveal a new class of CNS-born glia that critically contributes to motor nerve development.
The formation of spinal motor nerves requires coordinated interactions between several types of cells. Motor neurons extend axons into developing muscle fields from the spinal cord through segmentally positioned motor exit points (MEPs). These axons encounter neural crest–derived boundary cap cells clustered at MEPs, which permit the axons, but not the cell bodies, to exit from the spinal cord1. As motor axons approach their targets, they are sequentially wrapped and then myelinated by Schwann cells, glial cells that also develop from neural crest.
The myelinated motor nerve is surrounded by a flexible cellular sheath called the perineurium, first described in 1841 by Henle and later named by Key and Retzius2. The perineurium consists of uninterrupted, concentric rings of flattened cells that are connected by tight junctions and encase motor nerves from the MEP to the neuromuscular junction (NMJ)2–7. The perineurial sheath serves as a barrier, protecting axons from ionic flux, toxins and infection4,8–10. Therefore, formation of the perineurium is essential for peripheral nerve function.
Previous studies have suggested that, during development, the perineurium appears to form by a series of steps in which nearby mesenchymal cells first assemble as a loosely organized tube around the nerve and then mature to create a multilayered barrier11. The maturation step requires the signaling molecule Desert hedgehog (Dhh), which is expressed by Schwann cells. In the absence of Dhh, the perineurium is disorganized and is permeable to macromolecules and inflammatory cells12. Although these studies revealed a critical feature of perineurial cell differentiation, the origin of perineurial cells and how they initially associate with motor nerves remain unknown.
In Drosophila, motor axon–ensheathing glia are born in the lateral edges of the CNS and then migrate out in a chain-like fashion along the motor nerve13–17. The peripheral ensheathing glia of flies and peri-neurial cells of vertebrates have similar functional properties, raising the possibility that they have similar developmental origins. We tested this hypothesis by investigating the origin and development of perineurial cells in zebrafish. Using transgenic reporter genes and time-lapse imaging, we directly determined that, as in Drosophila, glial cells born in the CNS migrate into the periphery to ensheath motor nerves. Specifically, we observed that ventral spinal-cord glia emerged from MEPs and migrated along the entire length of motor axons, ensheathing both axons and Schwann cells and forming the perineurium. In the absence of these perineurial cells, we saw that motor axons exited the spinal cord ectopically and Schwann cells failed to wrap motor nerves, indicating that these cells help direct development of spinal ventral nerve roots. Finally, we observed that perineurial glia failed to ensheath motor nerves in mutant embryos with defective Schwann cell development, suggesting that the wrapping behavior of perineurial glia is instructed by signals derived from Schwann cells. These studies reveal intimate, orchestrated interactions between motor neurons, Schwann cells and perineurial glia in the formation of peripheral motor nerves.
In Drosophila, glial cells that wrap motor nerves originate in the CNS, close to motor neurons13,16. In zebrafish, lateral floor plate cells are immediately ventral to motor neurons18 and, consequently, in a position to similarly interact with pathfinding motor axons. We therefore hypothesized that zebrafish perineurial cells arise from the lateral floor plate. To test this idea, we used in situ RNA hybridization to examine the expression of genes that have been previously associated with the lateral floor plate. Ventral spinal-cord cells expressed nkx2.2a, nkx2.2b, foxa (fkd4), foxa2 (fkd1), nkx6.2 and nkx2.9 (data not shown). Among these markers, only nkx2.2a was expressed outside the CNS in a pattern that is consistent with the distribution of ventral motor nerves.
We investigated nkx2.2a expression in detail by observing the distribution of nkx2.2a RNA and enhanced green fluorescent protein (EGFP) encoded by a transgene (Fig. 1). At 24 and 36 h postfertilization (hpf), during the time of motor nerve formation19, nkx2.2a RNA was restricted to the lateral floor plate (Fig. 1a,c). However, by 48 hpf, nkx2.2a expression was evident in cells that were outside of the ventral spinal cord, lying close to the notochord (Fig. 1e). By 72 hpf, bilateral stripes of nkx2.2a expression extended ventrally from the spinal cord on either side of the notochord (Fig. 1g). From a lateral view, it was apparent that peripheral nkx2.2a expression was segmentally iterated, similar to the pattern of motor nerves in the periphery (data not shown).
We next examined transverse sections of Tg(nkx2.2a:megfp) embryos, in which membrane-tethered EGFP is expressed under the control of nkx2.2a regulatory sequences20,21. Similar to nkx2.2a expression, at 24–36 hpf, EGFP+ cells were restricted to the ventral spinal cord (Fig. 1b,d). However, at 48 and 72 hpf, EGFP+ cells formed sheath-like structures in the periphery that were continuous with the ventral spinal cord and ran ventrally down the sides of the notochord (Fig. 1f,h). Peripheral EGFP+ cells often had multiple fine membrane extensions, which was suggestive of migratory activity (Fig. 1h).
The segmental distribution and sheath-like morphology of peripheral nkx2.2a+ cells suggested that these cells wrap around motor axons. To test this hypothesis, we investigated the relationship between nkx2.2a+ cells and motor axons that were marked by DsRed2 fluorescent protein expressed under the control of olig2 regulatory DNA22. Transverse sections of Tg(nkx2.2a:megfp);Tg(olig2:dsred2) embryos at 48 hpf showed that EGFP+ peripheral cells enveloped DsRed2+ motor axons from the MEPs to the horizontal myoseptum(Fig. 2a). At 72 hpf, EGFP+ cells occupied nearly the full length of the primary branches of the motor roots, but were not associated with secondary axon branches (Fig. 2b). By 96 hpf, EGFP+ cells also had wrapped the secondary motor-axon branches and appeared to extend to the nerve terminals (Fig. 2c and data not shown). Distal ventral nerve roots have both motor and sensory processes. Labeling with an antibody to acetylated tubulin revealed that EGFP+ cells ensheathed mixed nerves at 96 hpf, but did not extend along the sensory branch to the dorsal root ganglia (data not shown). Sections obtained from an 8-month-old Tg(nkx2.2a:megfp) fish revealed that motor nerve–associated EGFP+ cells persisted into adulthood (Fig. 2d). These observations indicate that peripheral nkx2.2a+ cells progressively ensheath proximal motor nerves and distal mixed motor-sensory nerves during development, forming a structure that persists for the lifetime of the animal.
We next investigated the association of nkx2.2a+ peripheral cells with Schwann cells (Fig. 3). We crossed Tg(nkx2.2a:megfp) fish with Tg(sox10(7.2):mrfp) fish, which express membrane-tethered red fluorescence protein (RFP) in Schwann cells under the control of sox10 regulatory DNA20,23, to examine the spatial relationship of the two cell types. At 24 hpf, Schwann cells, but not EGFP+ cells, were present at segmentally iterated positions along the trunk (Fig. 3a). By 48 hpf, however, EGFP+ cells were now loosely associated with Schwann cells close to the spinal cord (Fig. 3b). By 72 hpf, EGFP+ cells appeared to be tightly wrapped around Schwann cells along the motor nerves, but not around Schwann cells of the lateral line nerve (Fig. 3c). To follow up this association between nkx2.2a+ cells and Schwann cells, we labeled Tg(nkx2.2a:megfp) fish with an antibody specific to myelin basic protein (MBP)24. This similarly revealed that EGFP+ cells surround MBP+ Schwann cells (Fig. 3d). Thus, peripheral nkx2.2a+ cells ensheath both motor axons and Schwann cells.
One interpretation of the data described above is that the peripheral nkx2.2a+ cells that wrap motor axons and Schwann cells originate in the ventral spinal cord and migrate out to join the peripheral nervous system (PNS). However, an alternative interpretation of the data is that peripherally located cells, such as mesenchyme, progressively express nkx2.2a and wrap around motor axons and Schwann cells in a proximal to distal sequence, beginning near the MEPs. To discriminate between these possibilities, we carried out in vivo time-lapse imaging of Tg(nkx2.2a:megfp) embryos.
Beginning at approximately 45 hpf, long nkx2.2a+ filopodia began to emerge from the ventral spinal cord at segmentally repeated positions (Fig. 4a and Supplementary Video 1 online). During the next several hours, nkx2.2a+ cell bodies followed their processes, migrating from the spinal cord into the periphery (Fig. 4a and Supplementary Video 1). These cell bodies sometimes migrated back into the spinal cord, but always returned to the periphery where they remained (Supplementary Video 1). The nkx2.2a+ cells often divided as they moved further into the adjacent muscle territories (Supplementary Video 2 online).
Most peripheral nkx2.2a+ cell bodies migrated from the spinal cord by 48 hpf, approximately 30 h after the first motor axons exited. To better understand the relationship between motor axons and nkx2.2a+ cells, we used Tg(nkx2.2a:megfp);Tg(olig2:dsred2) embryos for time-lapse imaging. This analysis confirmed that nkx2.2a+ cells emerged from the spinal cord at the MEPs (Fig. 4b and Supplementary Video 3 online). Initially, EGFP+ cells appeared to be broad and flat and not wrapped around motor axons. However, nkx2.2a+ processes eventually formed sheaths that extended along and around the axons (Fig. 4b and Supplementary Video 3). These time-lapse analyses provide direct evidence that peripheral nkx2.2a+ cells originate in the CNS, exit the spinal cord at MEPs and ensheath motor axons.
Because nkx2.2a+ cells that originate in the CNS wrap around both Schwann cells and axons, it seemed likely that they form the perineurium. To test this hypothesis, we assayed several defining characteristics of the perineurium, as described in other animals. First, we labeled Tg(nkx2.2a:megfp) fish with an antibody specific to zona occludins 1 (ZO-1), which labels tight junctions, a hallmark of the perineurium2,4,5. At 60 d postfertilization (dpf), EGFP fluorescence highlighted membranes that were associated with motor nerves (Fig. 5a). ZO-1 was concentrated at the same membranes, indicating the presence of tight junctions (Fig. 5a). A second feature of the perineurium is that it extends the full length of the motor nerve to the NMJ4. We therefore labeled Tg(nkx2.2a:megfp) fish with an SV2 antibody to visualize the synaptic vesicle pools concentrated at NMJs25. In sections of dorsal somatic muscle, SV2 labeling was always associated with EGFP+ processes (Fig. 5b), indicating that peripheral nkx2.2a+ cells extended to the NMJ. Finally, ultrastructural examination of the perineurium of rodents and birds by electron microscopy has shown concentric layers of flat cells linked by electron-dense tight junctions6,11. Similarly, electron microscopy studies of zebrafish showed that motor nerves were surrounded by concentric layers of flat cells, and that these cells were linked by characteristic electron-dense tight junctions (Fig. 5c,d). Each cell layer was associated with basal lamina on both sides, another characteristic feature of perineurial cells (Fig. 5d). On the basis of these criteria, we conclude that the zebrafish perineurium is structurally similar to the perineurium of other vertebrate species and arises from nkx2.2a+ cells that migrate from the CNS, instead of arising from mesenchymal cells in the vicinity of motor nerves. Hereafter, we will refer to these cells as perineurial glia to reflect their neural origin.
In the Drosophila CNS, premigratory peripheral glia are necessary for the stereotyped projection of motor axons into the periphery17. These cells therefore appear to act as guideposts for motor axons in the CNS. We hypothesized that the nkx2.2a+ lateral floor-plate cells that give rise to perineurial glia similarly influence exit of pioneering neurites from the spinal cord in zebrafish. Previous work established that nkx2.2a is required for lateral floor-plate development in zebrafish18.We therefore interfered with nkx2.2afunction using antisense morpholino oligonucleotides, and assayed the consequences on motor projections.
To test the efficacy of nkx2.2a morpholino oligonucleotides, we first assessed expression of genes that mark ventral spinal-cord cells. In wild-type embryos, foxa2 expression marks both medial and lateral floor-plate cells26 (Supplementary Fig. 1 online). In morpholino oligonucleotide–injected embryos, foxa2 expression was reduced to a band of 1 to 2 medial floor-plate cells (Supplementary Fig. 1). nkx2.2b expression normally marks lateral floor plate and interspersed V3 interneurons, but in the absence of nkx2.2a function, nkx2.2b was only expressed in presumptive V3 interneurons18. Consistent with this, fewer ventral spinal-cord cells expressed nkx2.2b in nkx2.2a morpholino oligonucleotide–injected embryos compared with controls (Supplementary Fig. 1). Thus, nkx2.2a morpholino oligonucleotide injections effectively reduce gene expression that marks the lateral floor plate.
We designed nkx2.2a MO1 (morpholino oligonucleotide 1) to block translation of the endogenous nkx2.2a gene, but not translation of egfp mRNA expressed by the transgene (Methods). This allowed us to use EGFP fluorescence as a marker of cells that expressed nkx2.2a in MO1-injected embryos (Fig. 6). We injected nkx2.2a MO1 into Tg(nkx2.2a:megfp);Tg(olig2:dsred2) embryos. Similar to wild-type embryos, ventral spinal-cord cells of injected embryos expressed high levels of EGFP (Fig. 6a–f). However, EGFP+ cells failed to migrate from the ventral spinal cord (Fig. 6d,f and Supplementary Table 1 online). Although motor axons of MO1-injected larvae were associated with NMJs, perineurial cell tight junctions were absent (Supplementary Fig. 2 online). Taken together with data showing that nkx2.2a is necessary for expression of genes that mark lateral floor plate18 (Supplementary Fig. 1) and that nkx2.2a+ lateral floor-plate cells give rise to perineurial glia (Fig. 1–Fig. 5), these results show that formation of the perineurium requires nkx2.2a function.
In control embryos, motor nerve projections were highly regular and tightly fasciculated (Fig. 6a,b and Supplementary Tables 1 and 2 online). In contrast, many motor axons in all morpholino oligonucleotide–injected embryos exited the spinal cord at irregular positions, causing a severe disruption of the normal segmental distribution (Fig. 6d,i and Supplementary Table 2). The motor axons in morpholino oligonucleotide–injected embryos were often stunted relative to control embryos, and the nerves appeared thinner and had more branches, suggesting that the motor axons did not form tight fascicles (Fig. 6a–f,j and Supplementary Tables 1 and 2). Additionally, a subset of morpholino oligonucleotide–injected embryos had DsRed2+ cell bodies outside of the spinal cord, showing that, in the absence of lateral floor plate and perineurial glia, motor neuron soma followed their axons into the periphery (Fig. 6e,f and Supplementary Table 1). These data indicate that nkx2.2a+ cells are necessary for exit of bundled motor axons at regular positions in the ventral spinal cord and to prevent motor neuron migration from the CNS.
As an independent method to eliminate lateral floor plate and perineurial glia, we interfered with Hedgehog (Hh) signaling. Zebrafish embryos homozygous for mutations of sonic hedgehog a (shha), which encodes one of five known Hh proteins, lack lateral floor plate, but have motor neurons27. Consequently, we speculated that low concentrations of cyclopamine, a pharmacological inhibitor of the Hh signaling pathway28,29, could similarly block lateral floor-plate formation with-out disrupting motor neuron development. We first determined that treatment with 25 µM cyclopamine at 9–10 hpf abolished nkx2.2a expression (Supplementary Fig. 3 online), but did not affect motor neuron number (n = 30 embryos) (Fig. 6g,h). We next exposed Tg(olig2:dsred2) embryos to 25 µM cyclopamine at 9 hpf (n = 60 embryos). Similar to nkx2.2a morpholino oligonucleotide–injected embryos, motor axons exited the spinal cord at irregular positions and appeared to be poorly fasciculated (Fig. 6g–j and Supplementary Table 2). Thus, two independent, but complementary, approaches support the conclusion that nkx2.2a+ lateral floor-plate cells and their descendent perineurial glia guide motor axon exit from the CNS and facilitate axonal fasciculation.
To investigate the effect of the absence of perineurial glia on Schwann cell development, we injected Tg(nkx2.2a:megfp);Tg(sox10(7.2):mrfp) embryos with nkx2.2a morpholino oligonucleotides. At 24 hpf, Schwann cells in morpholino oligonucleotide–injected embryos remained close to the ventral spinal cord, instead of extending along motor axons as they do during the normal developmental progression (Fig. 7a,b). By 48 hpf, Schwann cells in wild-type embryos had begun to form contiguous sheaths around motor axons (Fig. 7c). In contrast, Schwann cells in morpholino oligonucleotide–injected embryos appeared to be isolated, as individual, irregularly shaped cells at ectopic positions (Fig. 7d and Supplementary Table 1). Consistent with this, MBP was absent from motor nerves in morpholino oligonucleotide– injected larvae (Supplementary Fig. 2). Thus, in the absence of perineurial glia, Schwann cells failed to wrap around and myelinate motor nerves normally.
Previous studies demonstrated that Schwann cell–derived Dhh is necessary for development of the perineurial sheath, leading to the proposal that Dhh promotes a mesenchymal-to-epithelial transformation of nearby fibroblasts12. However, the above data provide compelling evidence that the perineurium forms from glial cells that migrate out of the ventral spinal cord, changing this interpretation. To investigate whether perineurial glia differentiation requires signals from Schwann cells, we examined colorless (cls) mutant fish30, which are deficient for Sox10, a transcription factor that is necessary for Schwann cell development31–33.
To image Schwann cells in the absence of Sox10, we produced cls mutant embryos carrying the Tg(sox10(7.2):mrfp) transgene. sox10+ cells were present in mutant embryos, but their morphologies were markedly different from wild type (Supplementary Fig. 4 online). Specifically, sox10+ cells often appeared as individuals, rather than as contiguous columns of cells associated with the motor nerves, and many did not wrap around axons. Thus, in the absence of Sox10 function, Schwann cells failed to coordinately wrap around motor nerves.
Next, we created Tg(nkx2.2a:megfp) embryos homozygous for two different cls alleles, clsm241 and clstw11 (Fig. 8). Time-lapse imaging in wild-type embryos showed that perineurial glia extend into the periphery as contiguous sheaths around motor nerves (Supplementary Video 3). In contrast, nkx2.2a+ cells in cls mutant embryos migrated away from the ventral spinal cord as individuals and never wrapped around motor nerves (Fig. 8a,b and Supplementary Videos 4 and 5 online). In mutant embryos, nkx2.2a+ cells had abnormally long and excessively active filopodia, suggesting that signals from Schwann cells are necessary to appropriately guide migration (Fig. 8a,b and Supplementary Videos 4 and 5). Thus, we conclude that Schwann cell differentiation is necessary for the coordinated migration and motor nerve–ensheathing activities of perineurial glia.
Finally, we asked whether the failure of perineurial cell differentiation in sox10 mutant embryos affects motor nerve formation. We labeled cls mutant embryos with antibody to znp-1, which labels primary motor axons, the first motor axons to project from the spinal cord19. By 48 hpf, primary motor axons in both clsm241 and clstw11 mutant embryos exited the spinal cord, but were more highly branched than those in wild type, suggesting that they were poorly fasciculated (Fig. 8c–e,g and Supplementary Table 2). However, in contrast to nkx2.2a morpholino oligonucleotide–injected embryos, all motor axons projected from the spinal cord at their normal, segmental positions (Fig. 8c–f and Supplementary Table 2). We interpret this to mean that, in cls mutant embryos, perineurial glia can act at MEPs to help guide motor axon exit, but that the subsequent failure of motor nerve wrapping by Schwann cells and perineurial glia permits abnormal motor axon branching.
The CNS is shielded from infection, mechanical insults and depolarizing ionic conditions by the meninges, a system of membranes composed of dura mater, arachnoid mater and pia mater. Pia mater, the innermost layer that directly contacts neurons, consists of multiple layers of flat, squamous epithelial cells. Peripheral nerves are similarly enclosed by the flattened cells of the perineurium. The pia mater is continuous with the perineurium at the interfaces of the CNS and PNS, and together they serve to safely partition the nervous system from the rest of the body. The origins and developmental roles of these ensheathing cells have received considerably less attention than the neurons and associated myelinating glia that they protect. Consequently, a question as simple as, “Where do perineurial cells come from?” has remained unanswered since they were reported more than 150 years ago.
One possibility is that perineurial cells arise from mesoderm. Perineurial cells have been described as mesenchyme or fibroblasts because, similar to fibroblasts, they have collagen fibers4,11,34 and it seemed that the perineurium was organized during development from cells that were loosely associated with nerves11,12. In support of this idea, fibroblasts cultured in the presence of neurons and Schwann cells form perineurial-like structures35. However, other characteristics, such as association of perineurial cells with basal lamina, are quite different from fibroblasts4. Another possibility is that perineurial cells arise from neural crest. However, in culture, neural crest–derived Schwann cells do not form perineurium35 and in vivo cell-lineage experiments have shown that neural crest contributes to nerve-associated endoneurium, but not to perineurium36. A third possible source of perineurial cells is the CNS. Examination of neural crest–ablated chick embryos and xenografted amphibian embryos revealed an apparent contribution of neural tube cells to ventral nerves37,38. Similarly, labeled chick neural-tube grafts gave rise to cells that were associated with ventral nerves that were described as Schwann sheath cells39. However, the identities of neural tube–derived cells in these experiments were not determined.
More recent cell-labeling strategies indicate that CNS cells migrate along cranial nerves, giving rise to a large variety of derivates40. However, the cells were not directly observed as they underwent their implicit migrations, and the conclusions remain controversial because other investigators, after failing to achieve similar results, have suggested that problems in the cell-labeling strategies account for the apparent origin of peripheral cells in the neural tube41,42. The evidence for CNS-derived cells that associate with peripheral nerves is much stronger in Drosophila. Clonal analyses show that many peripheral glia arise from CNS neuroblasts13, implying that peripheral glia migrate from the ventral nerve cord. Furthermore, transgenic reporter-gene expression patterns indicate that glial cells born near nerve exit and entry points migrate out of the CNS along axons and subsequently ensheath them16. These CNS-derived peripheral glia contribute to the blood-brain barrier15,43 and are therefore functionally similar to the perineurial cells of vertebrates.
In the present study, we took advantage of the ability to directly observe cells in developing zebrafish embryos to investigate the question of perineurial cell origin. Because many neural developmental mechanisms are conserved among invertebrates and vertebrates, we reasoned that, similar to Drosophila, glial cells that arise close to motor neurons in the CNS might migrate out along motor axons to form the perineurium. In static images of zebrafish embryos at different stages, we observed that nkx2.2a is expressed first in ventral spinal cord cells and later in peripheral columns that are continuous with the neural tube. Direct evidence came from time-lapse imaging showing that nkx2.2a+ ventral spinal cord cells emerge through the MEPs, wrap around both motor axons and Schwann cells and extend as sheaths to the motor nerve terminals. The ensheathing nkx2.2a+ cells form numerous tight junctions and do not express MBP, indicating that they form the perineurium. Thus, our work provides evidence that motor nerve–associated perineurial cells originate in the CNS. Because these peripheral cells arise from ventral spinal cord, mutations that alter neural tube dorsoventral patterning may have previously unanticipated consequences for PNS development.
A notable feature of embryos without perineurial glia is the fact that motor axons exit the spinal cord at ectopic positions. Moreover, these axons are abnormally branched and have defasciculated trajectories. To our surprise, motor-neuron cell bodies often escape the spinal cord, a phenotype previously associated with reduced levels of Hh signaling44. These observations raise the possibility that perineurial glia have two distinct functions before migration. First, the ectopic motor-axon exit implies that, similar to CNS glia of flies17, zebrafish perineurial glia help determine the position of MEPs. Second, the aberrant migration of motor neuron soma from the spinal cord suggests that perineurial glia contribute to the interface between the CNS and PNS. Neural crest–derived boundary cap cells, which transiently assemble at MEPs on the outside of the spinal cord, have a similar role in confining motor-neuron cell bodies to the neural tube in rodents and birds1. Future experiments will need to address whether this function of perineurial glia is unique to zebrafish or if perineurial glia and boundary cap cells coordinately act at MEPs in all vertebrates.
After migration into the periphery, perineurial glia direct the formation of tightly fasciculated, myelinated motor nerves. In the absence of perineurial glia, Schwann cells fail to target and wrap around motor axons. This could be an indirect consequence of ectopic axon exit. However, many axons still project from their normal positions in nkx2.2a morpholino oligonucleotide–injected embryos, raising the possibility that cues derived from perineurial glia attract Schwann cells or promote their differentiation. Conversely, Schwann cells also direct perineurial cell behavior. In cls/sox10 mutant embryos, in which Schwann cells do not differentiate, nkx2.2a+ cells migrate from the spinal cord, but do not wrap around motor nerves, implying the action of a Schwann cell–to–perineurial cell signal. In fact, one such signal has already been identified. Mouse Schwann cells express Dhh, and in its absence perineurial cells fail to create tight barriers around motor nerves. These findings led to the proposal that Dhh causes nearby fibroblasts to undergo a mesenchymal-to-epithelial transition and form the perineurium12. Our work suggests that reciprocal signaling between two classes of glia coordinates their differentiation and orchestrates development of fully myelinated and ensheathed motor nerves.
In summary, our work shows that the perineurium develops from glial cells that originate in the CNS and that these are actively involved in the formation of motor nerve roots through interactions with motor axons and Schwann cells. Our data should prompt a reexamination of the identity and origin of perineurial cells in mammals and further investigation of the molecules that promote their differentiation. Defects in perineurial glia might contribute to diseases disrupting nerve root development or maintenance. For example, Charcot-Marie-Tooth disease, which comprises a group of peripheral neuropathies, often results from defects of axons or Schwann cells, but the primary defects underlying some Charcot-Marie-Tooth disease variants are not well understood45–47. Additionally, many peripheral nerve–sheath tumors are described as perineurial48,49, raising the possibility that defects in perineurial glia proliferation contribute to cancer. A more complete understanding of the role of perineurial glia in neural development should help to bring new insights into these diseases of the peripheral nervous system.
All animal studies were approved by the Vanderbilt University Institutional Animal Care and Use Committee. Zebrafish strains used in this study included AB, Tg(nkx2.2a:megfp)vu17 (ref. 20), Tg(olig2:dsred2)vu19, Tg(sox10 (7.2):mrfp)vu234, clstw11 (refs. 30,50) and clsm241. Embryos were produced by pair-wise matings, raised at 28.5 °C in egg water or embryo medium and staged according to hpf or dpf. Embryos used for in situ hybridization, immunocytochemistry and microscopy were treated with 0.003% phenylthiourea in egg water to reduce pigmentation.
At 24 hpf, all embryos used for live imaging were manually dechorionated and transferred to egg water containing phenylthiourea. At specified stages, embryos were anesthetized using 3-aminobenzoic acid ethyl ester (Tricaine), immersed in 0.8% low–melting temperature agarose and mounted on their sides in glass-bottomed 35-mm Petri dishes (World Precision Instruments). All images were captured using a 40× oil-immersion (NA = 1.3) objective mounted on a motorized Zeiss Axiovert 200 microscope equipped with a PerkinElmer ERS spinning-disk confocal system. During time-lapse experiments, a heated stage chamber was used to maintain embryos at 28.5 °C. Z image stacks were collected every 5–10 min, and three-dimensional datasets were compiled using Sorenson 3 video compression (Sorenson Media) and exported to QuickTime (Apple) to create movies.
Embryos and larvae were fixed in 4% paraformaldehyde for 24 h, stored in 100% methanol at −20 °C and processed for in situ RNA hybridization. Plasmids were linearized with appropriate restriction enzymes and cRNA preparation was carried out using Roche DIG-labeling reagents and T3, T7 or SP6 RNA polymerases (NEB). After the in situ hybridization, embryos were embedded in 1.5% agarose/30% sucrose and frozen in 2-methyl butane chilled by immersion in liquid nitrogen. We collected 10-µm transverse sections on microscope slides using a cryostat microtome and covered with 75% glycerol. Images were obtained using a Retiga Exi–cooled CCD camera (Qimaging) mounted on an Olympus AX70 microscope equipped with Openlab software (Improvision). All images were imported into Adobe Photoshop. Adjustments were limited to levels, contrast, color matching settings and cropping.
Embryos and larvae were fixed in AB Fix (4% paraformaldehyde, 8% sucrose, 1× PBS) for 3 h at 23 °C or overnight at 4 °C and embedded as described above. We collected 10–50-µm transverse sections using a cryostat microtome. Sections were rehydrated in 1× PBS for 60 min at 23 °C and preblocked in 2% sheep serum/BSA–1× PBS for 30 min. Sections were incubated with primary antibody overnight at 4 °C. The primary antibodies used included mouse antibody to ZO-1 (Zymed, 1:200), mouse antibody to SV2 (Developmental Studies Hybridoma Bank, 1:500) and rabbit antibody to MBP24 (gift from Will Talbot, Stanford University, 1:1,000). Sections were washed extensively with 1× PBS, incubated for 3 h at 23 °C with either Alexa Fluor 568 goat antibody to rabbit or Alexa Fluor 568 goat antibody to mouse (Molecular Probes) as secondary antibodies for detection of primary antibodies, and washed with 1× PBS for 30 min. Sections were mounted in Vectashield (Vector Laboratories) and imaged using the confocal microscope described above. Image adjustments were limited to contrast enhancement and levels settings using Volocity software (Improvision) and Adobe Photoshop.
60-dpf adults were euthanized with Tricaine and the trunk was cut into thin sections (2–3 mm in width) to allow for good penetration of fixative. Samples were fixed in 2% glutaraldehyde in 0.05 M PBS for 1 h and placed into fresh 2% glutaraldehyde overnight. Samples were washed in PBS for 30 min, transferred to 1% OsO4 in PBS with 1.5% K3Fe(CN)63H2O276 for 3 h and washed three times in dH2O. Preparations were stained en bloc in 1% aqueous uranyl acetate for 1 h and washed three times in dH2O. Samples were then subjected to ethanol dehydration: 30% ethanol for 10 min twice, 50% ethanol for 10 min twice, 50% ethanol for 15 min twice, 70% ethanol for 30 min twice, 100% ethanol for 20 min three times and 100% ethanol for 60 min twice, and passed through two 20-min incubations with propylene oxide as a transition. Preparations were transferred to a 1:1 araldite:propylene-oxide mixture for 1 h and 1:3 araldite:propyleneoxide for 3 h, which was followed by complete infiltration of pure araldite facilitated with the help of a vacuum oven for 1 h. Samples were placed into fresh araldite and left to cure overnight in a 60 °C oven. Ultra-thin sections (60–70 nm) were obtained on a Leica UCT Ultracut microtome, transferred to formvar-coated grids and examined on a Phillips CM10 TEM equipped with an AMT 2 mega-pixel camera.
Embryos were incubated beginning at 9 hpf in embryo medium containing 25 µM cyclopamine (Toronto Research Chemicals), diluted from a 10 mM stock in ethanol, at 28.5 °C. To stop the treatment, embryos were rinsed three times in fresh embryo medium.
Morpholino antisense oligonucleotides were purchased from Gene Tools. nkx2.2a MO1 (5′-CCG TCT TTG TGT TGG TCA ACG ACA T-3′) was complementary to a sequence spanning just before and including the translation start codon18. nkx2.2a MO2 (5′-AAG TTG CTG CAC CAG TTT GAC AAT C -3′) was complementary to a sequence in the 5′ UTR. Morpholino oligonucleotides were dissolved in water to create a stock solution of 3 mM and diluted in 2× injection buffer (5 mg ml−1 Phenol red, 40 mM HEPES and 240 mM KCl) to create a working injection concentration of 0.5 mM for MO1 and 0.75 mM for MO2. We injected 2–4 nl into the yolk just below the single cell of fertilized embryos. All morpholino oligonucleotide injected embryos were raised in embryo medium at 28.5 °C.
To measure motor nerve–root periodicity in morpholino oligonucleotide–injected and wild-type embryos, composite Z image stacks were compiled using Volocity software. The distance between each motor nerve root was determined by drawing a line from the center of each exiting motor nerve to the next. Root periodicity was determined between somites 8 and 12. The number of axon branches to the horizontal myoseptum was determined by following axon bundles using the XYZ function of Volocity. Individual Z images were sequentially observed and branches counted. For morpholino oligonucleotide–injected embryos in which the horizontal myoseptum was not clear, axon branches were counted over a distance of 52 µm from the spinal cord, which is the average distance to the horizontal myoseptum in control embryos. Axon branching was also determined between somites 8 and 12. All graphically presented data represents the mean of the analyzed data. Statistical analyses were performed with Prism software. The level of significance was determined by either paired or unpaired t-tests using a confidence interval of 95%.
Note: Supplementary information is available on the Nature Neuroscience website.
We thank T. Piotrowski, J. Shin, W. Talbot and R. Karlstrom for reagents and fish, M. Bhat, V. Auld and members of the Appel lab for valuable discussions, and J. Weston for comments on the manuscript. Reagents also were provided by the Zebrafish International Resource Center, supported by grant P40 RR012546 from the US National Institutes of Health (NIH) National Center for Research Resources. This work was supported by the Post Doctoral Training Program in Neurogenomics-MH65215-03 (S.K.), NIH grant R01 NS046668 (B.A.), NIH grant R01 GM054544 (K.B.) and a zebrafish initiative funded by the Vanderbilt University Academic Venture Capital Fund.