Glycerophospholipid regulation of modality-specific sensory axon guidance in the spinal cord

See allHide authors and affiliations

Science  28 Aug 2015:
Vol. 349, Issue 6251, pp. 974-977
DOI: 10.1126/science.aab3516

Axon paths in developing spinal cords

Sensory neurons entering the spinal cord take different paths as inputs for pain and proprioception diverge. Working with chick and mouse embryos, Guy et al. found that glycerophospholipids produced by radial glial cells guide these neural fibers, or axons, in the developing spinal cord. A soluble glycerophospholipid released by the cells provided an inhibitory signal to the pain-sensitive axons, keeping them on their own unique pathway.

Science, this issue p. 974


Glycerophospholipids, the structural components of cell membranes, have not been considered to be spatial cues for intercellular signaling because of their ubiquitous distribution. We identified lyso-phosphatidyl-β-d-glucoside (LysoPtdGlc), a hydrophilic glycerophospholipid, and demonstrated its role in modality-specific repulsive guidance of spinal cord sensory axons. LysoPtdGlc is locally synthesized and released by radial glia in a patterned spatial distribution to regulate the targeting of nociceptive but not proprioceptive central axon projections. Library screening identified the G protein–coupled receptor GPR55 as a high-affinity receptor for LysoPtdGlc, and GPR55 deletion or LysoPtdGlc loss of function in vivo caused the misallocation of nociceptive axons into proprioceptive zones. These findings show that LysoPtdGlc/GPR55 is a lipid-based signaling system in glia-neuron communication for neural development.

During nervous system development, axon-guiding proteins play major roles in network formation, but lipids may also be a family of intercellular guidance cues because of their abundance and diversity in nervous tissue (1, 2). Phosphatidyl-β-d-glucoside {PtdGlc [2-O-arachidoyl-1-O-(1-stearoyl)-sn-glycer-3-yl]-β-d-glucopyranosyl phosphate; fig. S1A} is a membrane glycerophospholipid (3) that is localized to radial glia and nascent astrocytes (4, 5). We showed that PtdGlc and its hydrolytic derivative lyso-phosphatidyl-β-d-glucoside (LysoPtdGlc; fig. S1A) occur in a patterned distribution in the developing spinal cord and mediate glia-neuron communication to guide nociceptive afferent axons.

Nanoliquid chromatography/tandem mass spectrometry detected PtdGlc in proliferative glia isolated from spinal cords of stage (st.) 34 chick embryos (fig. S1, B and C). Their conditioned medium contained water-soluble LysoPtdGlc (fig. S1, D and E) but not PtdGlc, which suggests that PtdGlc can be hydrolyzed in glial membranes and released as LysoPtdGlc. We double-stained glial cultures with DIM21, an antibody to PtdGlc (6, 7), and EAP3, an antibody to transitin, a marker of radial glia (8). 94.1% of cultured glia were labeled positive for PtdGlc, and of these cells, 91.1% coexpressed PtdGlc and transitin (fig. S1, F and G). In spinal cord sections, PtdGlc is distinct from cells labeled with the neuronal marker NeuN, whereas it colocalizes with transitin (fig. S2). Because these data suggested that spinal cord radial glia produce PtdGlc and can release LysoPtdGlc into the extracellular environment, we examined a role for these glycerophospholipids in spinal cord development.

Dorsal root ganglion (DRG) axons of different sensory modalities elongate through a common tract, the dorsal root, but segregate as they elongate in the spinal cord (fig. S3). Then these axons bifurcate and extend in the white matter both rostrally and caudally (9), with nociceptive afferents positioned laterally and proprioceptive afferents occupying a more dorsomedial position, the primordial dorsal funiculus (10) (fig. S4A). To examine this initial segregation process, we observed the spatiotemporal relationship between PtdGlc-positive radial glia and DRG afferents. We labeled transverse sections with DIM21 and antibody to either TrkA or TrkC to stain nerve growth factor (NGF)–responsive nociceptive or neurotrophin-3 (NT-3)–responsive proprioceptive axons (Fig. 1). At st. 28, DIM21 labels PtdGlc in cell bodies lining the dorsal midline as well as the primordial dorsal funiculus that receives the pial processes of PtdGlc-positive cells (Fig. 1A). TrkA-expressing axon terminals are localized to the primordial dorsal root entry zone (DREZ), whereas those expressing TrkC extend into the PtdGlc-positive primordial dorsal funiculus. At st. 34, PtdGlc is abundant in the dorsomedial spinal cord, colocalizing with TrkC but not TrkA (Fig. 1B). We used mass spectrometry to quantify LysoPtdGlc in spinal cord subregions (fig. S5). At st. 34, LysoPtdGlc was undetectable in the dorsolateral spinal cord, whereas approximately 7.4 μM LysoPtdGlc was present in the dorsomedial cord (see supplementary methods for the protocol), indicating that LysoPtdGlc codistributes with PtdGlc in this area. The demarcation of the TrkA and PtdGlc domains compared to the colocalization of TrkC with PtdGlc at all stages observed (fig. S4) led us to hypothesize that PtdGlc or its derivative LysoPtdGlc has a repulsive effect that is specific to nociceptive afferents.

Fig. 1 Intraspinal localization of PtdGlc and chemorepulsive activity toward DRG afferents.

(A and B) Transverse sections of st. 28 (A) and st. 34 (B) chick lumbar spinal cord with hematoxylin and eosin (H & E) stain (scale bar, 100 μm), double immunofluorescence of PtdGlc and either TrkA or TrkC, and grayscale images showing the unilateral dorsal cord from green/magenta merges at higher magnification (scale bar, 50 μm). The arrow and arrowhead indicate the dorsal root entry zone (DREZ) and the primordial dorsal funiculus, respectively. (C) Collagen gel explant assay. Dorsomedial (DM) or dorsolateral (DL) st. 34 spinal cord explants were cocultured with a DRG explant enriched for either TrkA or TrkC neurons by adding NGF or NT-3 to the medium, respectively. DRG axons were visualized with antibody to β-tubulin. Images are composites of three or four photomicrographs. Scale bar, 500 μm. (D) Explants were scored using a scale of 0 to 10, with 0 representing the greatest chemorepulsion (see the supplementary methods for the scoring protocol). Bars represent mean chemorepulsion score ± SEM, and numbers in boxes represent cultures per group. ***P < 0.001; Kruskal-Wallis test.

To assess the repulsive activity of the dorsomedial cord, we cocultured one DRG explant and one spinal cord subregion explant dissected from st. 34 embryos in a collagen matrix, with either NGF or NT-3 added to the culture medium (Fig. 1, C and D) to enrich explants with NGF- or NT-3–responsive neurons (11). Explants of dorsomedial, but not dorsolateral, cord repelled NGF-responsive axons. NT-3–responsive axons did not display a chemorepulsive response to spinal cord explants of either subregion.

To test whether LysoPtdGlc alone is sufficient for the chemorepulsion, we prepared dissociated cultures of st. 28 or 36 DRG neurons and quantified growth cone turning responses to a concentration gradient of LysoPtdGlc in the culture medium (Fig. 2). DRG neurons that had been maintained with bath-applied NGF or NT-3 showed attractive responses to a gradient of NGF or NT-3, respectively, but not to a gradient of the opposite neurotrophin (Fig. 2B), indicating that we could selectively subject NGF- or NT-3–responsive neurons to this assay. PtdGlc purified from embryonic brain tissue (4) was hydrolyzed in vitro to produce LysoPtdGlc, and a gradient of LysoPtdGlc chemorepelled NGF-responsive but not NT-3–responsive axons (Fig. 2, A and B). In this assay, the LysoPtdGlc concentration near the growth cone was approximately 0.1% of in-pipette concentration (12), or 1 nM. The same concentration of lysophosphatidic acid (LPA) did not repel NGF-responsive axons (low LPA in Fig. 2B), excluding the possibility that LysoPtdGlc was chemorepulsive after being hydrolyzed to LPA in the medium. We synthesized LysoPtdGlc (fig. S6 and supplementary text) to avoid the possibility of contaminants in brain-derived preparations and found that synthetic LysoPtdGlc also repelled NGF-responsive but not NT-3–responsive axons (fig. S7). Lysophosphatidylcholine, a structurally related lysophospholipid, did not induce axon turning (Fig. 2B).

Fig. 2 LysoPtdGlc repels NGF-responsive but not NT-3–responsive DRG axons.

(A) Dissociated st. 36 chick DRG neurons were cultured with either NGF or NT-3 and exposed to a concentration gradient of purified LysoPtdGlc. The number in each panel indicates minutes after initiation, and the arrowhead shows the direction of LysoPtdGlc gradient; scale bar, 10 μm. (B) Quantification of axon turning assay. Bars represent mean turning angle ± SEM with positive and negative values indicating attraction and repulsion, respectively. Numbers in parentheses are growth cones tested per group. The y-axis labels indicate the lipid or neurotrophin gradient, and in parentheses the neurotrophin added to the culture medium. Abbreviations and in-pipette concentrations used are as follows: LysoPtdGlc (purified; 1 μM); LysoPtdCho, lysophosphatidylcholine (1 μM); low LPA (1 μM); LPA (20 μM); NGF (50 μg/ml); NT-3 (50 μg/ml). ***P < 0.001, **P < 0.01, *P < 0.05; Kruskal-Wallis test. (C) Antibody to LysoPtdGlc blocked the chemorepulsive turning of NGF-responsive st. 28 DRG axons induced by LysoPtdGlc. Bars represent mean turning angle ± SEM. Numbers in parentheses are growth cones tested per group. In-pipette concentrations used are as follows: LysoPtdGlc (synthetic; 10 μM); LPA (20 μM); Sema3A (10 μg/ml). ***P < 0.001; Kruskal-Wallis test.

Using the Autonomously Diversifying Library System (13) we raised a function-blocking antibody to LysoPtdGlc that exhibited negligible binding to PtdGlc or other lipids tested (fig. S8). This antibody abolished the repulsive turning of NGF-responsive axons induced by LysoPtdGlc, but not that by LPA or semaphorin 3A (Sema3A) (Fig. 2C), and attenuated the chemorepulsive activity of dorsomedial spinal cord explants toward NGF-responsive DRG axons (fig. S9).

We injected anti-LysoPtdGlc into the spinal cord in ovo, either alone or as a cocktail with anti–neuropilin-1 (NRP-1). Antibody-injected embryos were fixed at st. 28 and phenotyped by labeling nociceptive afferents by 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI) injection into the dorsomedial DRG (14) (fig. S10). Whereas DiI-labeled nociceptive afferents were confined to the DREZ in control embryos, these afferents showed dorsomedial projection into the primordial dorsal funiculus after injection of anti-LysoPtdGlc alone or a cocktail of the two antibodies (Fig. 3, A to C). This guidance error of nociceptive afferents was also confirmed by TrkA immunofluorescence (Fig. 3, D and E) and quantitative pixel analysis (fig. S11). Anti-LysoPtdGlc injection alone induced guidance errors (Fig. 3, C and E), most likely because Sema3A is absent from the lateral spinal cord (15) and may not participate in nociceptive axon guidance in the oval bundle of His. Antibody injection did not affect PtdGlc or TrkC domains (figs. S11 and S12). These results indicate that LysoPtdGlc-mediated chemorepulsion prevents TrkA-positive nociceptive axons from entering the primordial dorsal funiculus.

Fig. 3 LysoPtdGlc regulates nociceptive afferent projections in developing spinal cord.

Injection of function-blocking antibody into the spinal cord in ovo attenuated LysoPtdGlc signaling. (A) Transverse lumbar spinal cord sections showing unilateral DiI-labeled putative nociceptive afferents in embryos injected with control antibody, anti-LysoPtdGlc, or a cocktail of anti-LysoPtdGlc and anti-NRP1. Scale bar, 75 μm. (B) Method of evaluating normal and dorsomedial projection phenotypes. The dorsolateral white matter known as the oval bundle of His is composed of the DREZ and the primordial dorsal funiculus (PDF). For quantification, the oval bundle of His was divided into dorsal and ventral halves as indicated by the broken line. Sections showing DiI- or TrkA antibody–labeled afferents only in the ventral half were considered normal, whereas those showing positive labeling in the dorsal half were classed as dorsomedial projections. (C) Quantification of dorsomedial projection of DiI-labeled afferents after antibody injection. (D) TrkA immunofluorescence in control antibody or anti-LysoPtdGlc–injected spinal cords. Scale bar, 50 μm. (E) Quantification of TrkA+ dorsomedial projection after antibody injection. [(C) and (E)] Bars represent mean percentage ± SEM of sections showing dorsomedial projection of all embryos in each test group. Numbers within each bar indicate the number of embryos examined per group. ***P < 0.001, **P < 0.01, *P < 0.05; Kruskal-Wallis test.

We next searched for a receptor and downstream signals that mediate axon repulsive responses to LysoPtdGlc. We found that CT04, an inhibitor of Rho guanosine triphosphatases (16), or Y-27632, an inhibitor of Rho-associated protein kinase (17), abolished LysoPtdGlc-induced repulsion in NGF-responsive neurons (fig. S13A). Because Rho can be activated downstream of Gα12/13 (18), we used targeted inhibitory Gα12/13 C-terminal peptides (19) and found that Gα13 mediates LysoPtdGlc-induced chemorepulsion (fig. S13B).

We conducted screening to identify a G protein–coupled receptor (GPCR) mediating LysoPtdGlc, using the transforming growth factor-α (TGF-α) shedding assay (20). Of 115 GPCRs tested, only GPR55 responded to LysoPtdGlc (fig. S14). The half-maximal effective concentration (EC50) of LysoPtdGlc was 16 nM (fig. S15). Although GPR55 has been reported as a receptor for lysophosphatidylinositol (LysoPtdIns) (21), the EC50 of LysoPtdIns was 110 nM (fig. S15B), indicating that LysoPtdGlc is the more potent ligand. Gα12 and Gα13 knockdown abolished the LysoPtdGlc-induced TGF-α response, which could be rescued by a chimeric Gα protein that consisted of a Gαq backbone with the Gα13 C-terminal sequence (fig. S16), indicating that LysoPtdGlc signals through Gα13-coupled GPR55.

After confirming GPR55 expression in DRG neurons by polymerase chain reaction (fig. S17), we used a Gpr55 knockout mouse (22) to test whether GPR55 mediates repulsive axon responses to LysoPtdGlc. In vitro, LysoPtdGlc did not affect the directional growth of NT-3–responsive axons but chemorepelled NGF-responsive axons in a GPR55-dependent manner (Fig. 4, A and B). Although it has been reported that LysoPtdIns induces neurite retraction through GPR55 (23), LysoPtdIns chemorepelled Gpr55−/− axons (Fig. 4B), indicating that LysoPtdIns can signal through GPR55-independent pathways as reported (24). To examine Gpr55 function in vivo, we labeled nociceptive afferents in the spinal cord of embryonic day 14.0 wild-type or Gpr55−/− mice with TrkA antibody (Fig. 4C) or DiI (fig. S18). Nociceptive afferents were localized mostly to the PtdGlc-sparse DREZ in the wild type, but displayed dorsomedial projection in Gpr55−/− mice (Fig. 4, C to E) to where PtdGlc is concentrated and normally TrkC+ afferents predominate (fig. S20). These data indicate that Gpr55 deletion in the mouse causes nociceptive afferent projection errors analogous to those induced by LysoPtdGlc loss of function in the chick. Thus, we demonstrated a mechanism for a diffusible glycerophospholipid that mediates glia-neuron intercellular communication to guide nociceptive axons in the central nervous system. This lipid-mediated guidance may underlie the role of GPR55 in nociception in vivo (25).

Fig. 4 GPR55 is a receptor for LysoPtdGlc-mediated axon guidance.

(A) LysoPtdGlc chemorepels NGF-responsive DRG axons of wild-type (WT) but not Gpr55−/− mice. The number in the panels indicates minutes after initiation, and the arrowheads show the direction of the gradient; scale bar, 10 μm. (B) Quantified turning responses of WT or Gpr55−/− axons to gradients of LysoPtdGlc, LPA, LysoPtdIns, or Sema3A. NGF or NT-3 was added to the culture medium as indicated in parentheses. Bars represent mean turning angle SEM; numbers in parentheses indicate growth cones tested. **P < 0.01; one-way analysis of variance with Tukey’s post-test. (C) TrkA immunofluorescence in WT or Gpr55−/− mouse lumbar spinal cord. Solid circles indicate the midline; scale bar, 200 μm. (D) Method of evaluating TrkA immunofluorescence. Gray-shaded area represents the TrkA+ domain, and the broken line indicates the midline. If the TrkA+ domain in the medial quarter (MQ) of the spinal cord was equal to or greater in thickness than the TrkA+ domain in the DREZ (arrowheads), the section was classified as a dorsomedial projection. (E) Quantification of TrkA+ dorsomedial afferent projection in WT and Gpr55−/− mice. Bars represent mean percentage SEM of sections showing dorsomedial projection of all embryos in each test group. Numbers within each bar indicate the number of embryos examined per group. *P < 0.05; t test.

Supplementary Materials

Materials and Methods

Additional Author Notes

Supplementary Text

Figs. S1 to S28

References (2649)


  1. ACKNOWLEDGMENTS: We thank Y. Shinoda, Y. Osaka, A. Shuto, H. Akiyama, and E. Stoeckli for technical assistance; S. Higashiyama, J. Miyazaki, and F. Lefcort for reagents; and C. Yokoyama for critical reading of the manuscript. Funding for this research was provided by Japan Society for the Promotion of Science Foreign Postdoctoral Fellowships (A.T.G. and P.G.), AMED-CREST (J.A.), PRESTO (A.I.), the Mizutani Foundation for Glycoscience (P.G.), a Naito Foundation Subsidy for Promotion of Specific Research Projects (Y.H.), the Platform for Dynamic Approaches to Living System (K.O. and A.M.), and a Ministry of Education, Culture, Sports, Science and Technology (MEXT) Grant-in-Aid for Scientific Research (22116001 to J.A.) and for Scientific Research on Innovative Areas (23114003 to K.O. and A.M., 24110519 to Y.H., and 21200009 to H.K.). All reagents in this study are available from H.K. or Y.H. under a materials transfer agreement with RIKEN. The authors declare the following financial interests: Y.H., H.K., K.O., A.M., and RIKEN have a patent issued (20130142813) relating to the production of anti-LysoPtdGlc, and K.O. is an executive director of Chiome Bioscience Inc. The supplemental material contains additional data.
View Abstract

Navigate This Article