Subcellular Localization of a 2-Arachidonoyl Glycerol Signaling Cassette in Retinal Ganglion Cell Axonal Growth In Vitro

Abstract

Purpose: To investigate whether the subcellular distribution of endocannabinoid (eCB) system (ECS) components in growing RGC axons is consistent with the formation of eCB-enriched “hotspots” and the role of the ECS in RGC axonal growth.

Methods: We used immunocytochemistry and image analysis to quantify axonal expression of the ECS components diacylglycerol lipase alpha (DGLα), monoacylglycerol lipase (MGL), and cannabinoid receptor type 1 (CB1R) in a mouse retinal explant model. We tested whether pharmacologic antagonists of CB1R and inhibitors of eCB degradation modulate ECS component expression and axonal growth.

Results: DGLα expression was higher in the distal RGC axon than in the growth cone central domain (GCCD) (95% confidence interval [CI], 106.5%–122.4% at 15 μm proximal to the GCCD), whereas MGL expression in the same region was not significantly different (95% CI, 88.8%–102.1%). In more proximal axon segments, DGLα and MGL expression were both lower than in the GCCD, whereas CB1R expression was 2.5-fold higher in this region (95% CI, 220.3%–278.4% at 50 μm proximal to the GCCD). The presence of CB1R antagonist O-2050 disrupted profiles of ECS component expression and increased axonal growth (95% CI for the difference of median axon lengths 26.6–55.6 μM).

Conclusions: Our results demonstrate an ECS topology in RGC axons that is consistent with formation of eCB-enriched hotspots and suggest that the ECS has a role in CB1R-dependent inhibition of RGC axonal growth in vitro.

Visual signals sent from the retina to the brain must pass through axons of the retinal ganglion cells. The death of these cells in optic neuropathies such as glaucoma is a major cause of blindness. There is significant interest in identifying molecular signaling pathways that modulate the regeneration of adult retinal ganglion cell (RGC) axons by surviving ganglion cell somata^1,2 and promote the establishment of new neural connections between transplanted stem cell–derived RGCs and native visual circuitry.^3,4 Although some mechanisms that govern the successful regeneration of damaged neural connections in the adult central nervous system may be unique, fundamental aspects of embryonic neuritogenesis, polarization, and axonal growth are likely to remain important.^1,5–8 Models of embryonic RGC axonal growth may help identify important signaling pathways in regeneration-, replacement-, or rescue-based therapeutic approaches to optic neuropathies. 

Endocannabinoids (eCBs) are lipid messengers that modulate several types of developmental axonal growth by activating cannabinoid receptor type 1 (CB1R) and G protein-coupled receptor 55 (GPR55).^9,10 Studies of the eCB system (ECS) in long-range glutamatergic projections have focused on the development of reciprocal connections between the neocortex and thalamus.^9,11–13 In this central nervous system region, finely tuned, complementary subcellular distributions of ECS components are poised to promote efficient growth as corticothalamic and thalamocortical axons with opposing trajectories cross paths (the corticothalamic-thalamocortical “handshake”).^9,11 “Hotspots” of eCB 2-arachidonoyl glycerol (2-AG) enrichment may be formed by the exclusion of 2-AG-degrading monoacylglycerol lipase (MGL) from corticofugal growth cones (GCs), where 2-AG-synthesizing diacylglycerol lipase alpha (DGLα) is enriched and coincident with CB1R.^9,12,14 Axonal MGL forms an intracellular barrier to GC-derived 2-AG in a significant proportion of corticothalamic axons (CTAs), whereas developing thalamocortical axons form an extracellular barrier against 2-AG by expressing MGL, but not CB1R.¹¹ This topology limits CB1R signaling competence to the CTA GC, and autocrine activation of CB1R by 2-AG at the CTA GC plasmalemma promotes normal fasciculation and may be sufficient for stimulating directional growth.^9,11–13,15 Functionally significant ECS components are expressed in RGC axons during development,^12,16–19 but no reciprocal connections between thalamic or extrageniculate targets and retina are known to exist. There is no clear analog to the CTA-thalamocortical axons handshake during retinofugal projection growth, and this opens the issue of whether developing RGC axons express a distinct arrangement of ECS components. 

We used an embryonic mouse retinal explant model to examine the topology and function of the ECS in RGC axons and growth cones. We investigated whether the RGC ECS exhibits a subcellular arrangement that is consistent with formation of eCB hotspots and whether CB1R promotes or inhibits functional readouts associated with RGC axonal growth in vitro. 

Methods

Materials

Follicular unit extraction micropunches were from Mediquip Surgical (San Francisco, CA, USA). C57 black 6 (C57BL/6) mice were from Charles River Laboratories. Rabbit polyclonal anti-CB1R (used at 1:200 final dilution), anti-MGL (1:50), and anti-DGLα (1:100) antibodies and corresponding antigen peptides were purchased from Frontier Institute Co. (Ishikari, Japan). All reagents for retinal ganglion cell growth medium (RGC-GM) and growth substrate preparation were sourced as previously described.²⁰ O-2050 (100 nM), JZL184 (100 nM), and AM251 (100 nM) were from Tocris Bioscience (Bristol, UK). Cyanine (Cy2)-donkey anti-mouse immunoglobulin G (IgG) antibody (1:400 final dilution), Indodicarbocyanine (Cy5)-donkey anti-rabbit IgG antibody (1:400), normal donkey serum, normal rabbit serum, and bovine serum albumin (BSA; IgG and protease free) were from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). Molecular Probes Alexa Fluor 546 phalloidin was from ThermoFisher Scientific (Grand Island, NY, USA). Mouse monoclonal anti-βIII-tubulin antibody (1:1000) was from Promega (Madison, WI, USA). Glycerol gelatin aqueous mounting medium was from Sigma-Aldrich (St. Louis, MO, USA). 

Embryonic Retinal Explant Culture and Drug Treatments

All animal procedures were performed in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Retinal explants (2 to 4 per eye) were taken from embryonic day 15 (E15) C57BL/6 mice with a 300 μm follicular unit extraction punch and plated on poly-D-lysine and laminin coated²⁰ glass coverslips in RGC-GM. RGC-GM was composed of Neurobasal medium/Dulbecco's modified Eagle's medium (1:1) plus insulin (5 μg/ml), sodium pyruvate (110 μg/ml), thyroxine (40 ng/ml), L-glutamine (292 μg/ml), neuronal supplement 21 (NS21) (1X), N-acetylcysteine (5 μg/ml), BSA (100 μg/ml), transferrin (100 μg/ml), putrescine (160 μg/ml), progesterone (60 ng/ml), sodium selenite (40 ng/ml), forskolin (4.2 μg/ml), brain-derived neurotrophic factor (50 ng/ml), ciliary neurotrophic factor (50 ng/ml), and penicillin-streptomycin.^20,21 Drugs or dimethyl sulfoxide vehicle were added immediately after plating, and stimulation time was 16 hours. The authors were masked to treatment conditions until all image acquisition, postacquisition processing, and statistical analyses were completed. 

Immunocytochemistry

Cultures were maintained for 16 hours in the presence or absence of the indicated drugs and then fixed in PBS at pH 7.4 with 4% paraformaldehyde and 4% sucrose for 10 minutes at room temperature (RT). Cultures were then rinsed with PBS and permeabilized in PBS plus 0.25% Triton X-100 and 10 μg/ml Hoechst 33258 for 10 minutes at RT. Cultures were rinsed with PBS plus 2% BSA (2%BSA-PBS) and blocked with 10% normal donkey serum in 5%BSA-PBS for 2 hours at RT with gentle agitation. Cultures were then incubated with selected combinations of primary antibodies in 2%BSA-PBS with or without corresponding antigen peptides (50 μg/ml) for 24 hours at 4°C with gentle agitation. The specificity of the ECS antibodies was previously demonstrated with Western blot, immunohistochemistry, knockout mice, and RNA in situ hybridization.^22–24 The cultures were rinsed with 2%BSA-PBS and incubated with Alexa Fluor 546 phalloidin and secondary antibodies in 2%BSA-PBS for 2 hours at RT with gentle agitation. Alexa Fluor 546 phalloidin was prepared according to the vendor's instructions and used at a final dilution of 1:50. Final dilutions of primary and secondary antibodies were as noted in the Materials section. The coverslips were rinsed again in 2%BSA-PBS and mounted on glass slides with glycerol gelatin aqueous mounting medium. The rabbit anti-MGL antibody used here was previously noted to have significant cross-reactivity with other primary antibodies.²⁴ We therefore adapted the two-step method developed by Uchigashima et al.²⁴ for multiple immunofluorescence involving MGL. In these cases, after blocking with 10% normal donkey serum in 5%BSA-PBS, cultures were incubated with rabbit polyclonal anti-MGL antibody only (1:50) in PBS + 0.1% Tween 20 (PBST) for 24 hours at 4°C with gentle agitation. Cultures were then rinsed with PBST and incubated for 2 hours with Cy5-donkey anti-rabbit IgG in PBST at RT with gentle agitation. The cultures were rinsed again with PBST and blocked with 10% normal rabbit serum in PBS for 2 hours at RT to mask the MGL antibody. Finally, the cultures were incubated with a second selection of primary antibodies and were processed and mounted as described previously. 

Confocal Microscopy, Image Analysis, and Statistical Analysis

Images were obtained with a FLUOVIEW FV1000 confocal microscope (Olympus, Center Valley, PA, USA) at eight-bit depth and 10X or 100X primary magnification. For high magnification imaging of GCs and axons, laser power and gain were set such that only a miniscule proportion of pixels in the brightest fields registered detector saturation. Two or three 0.5-μm optical sections acquired in a Z-stack was usually sufficient to image a growth cone and up to 100 μm of distal axon. Each explant was systematically circled starting from the 12 o'clock position, and images of as many axons as possible were acquired under identical microscope, laser, and software settings for any groups that would subsequently be compared. For low magnification imaging of whole explants for the assessment of axonal growth, two or three approximately 10-μm optical sections acquired in a Z-stack was usually sufficient to image the entire axonal growth area surrounding the explant body. Data were exported from FLUOVIEW software as raw eight-bit gray values, and maximum value Z-stack projections were created with ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). It was occasionally necessary to create mosaics of two to four low power fields to image all axons from explants displaying exuberant axonal growth. 

For high magnification images of GCs and axons, F-actin and βIII-tubulin signals were used to create regions of interest (ROIs) that defined GC central domains (GCCDs), peripheral domains (GCPDs), and consecutive 5-μm segments of axon.⁵ Raw eight-bit gray values were recorded and used to quantify ECS component expression within each ROI. For the analysis of ECS component axonal microgradients, gray values were normalized to the GCCD. For low magnification images of whole explants, a binary mask of the nuclear stain was used to define the region occupied by the explant body, and the explant body centroid coordinates were recorded. The area of the roughly circular explant body was used to calculate an explant body radius, and a masked observer (DTS) tagged all individually identifiable terminations of RGC axons or axon fascicles and recorded corresponding coordinates. Axon length was then calculated by determining the distance between each tagged axon termination and its respective explant body centroid and subtracting the explant body radius. 

Data were analyzed in Minitab software (Minitab Inc., State College, PA, USA) with Student's t-test, Mann-Whitney U test, or piecewise linear least squares regression as indicated in the Table. P values less than 0.05 and the nonoverlap of 95% confidence intervals (CIs) were considered statistically significant. 

Table

 

Statistical Table

 

Results

Subcellular Arrangement of ECS Components in Developing RGC Axons

We quantified the subcellular expression of ECS components, including MGL, DGLα, and CB1R in GCCDs, GCPDs, and 5 μm segments of stabilized axon from RGCs in cultured E15.5 mouse retina explants, to determine whether the configuration is consistent with the formation of eCB hotspots. All three of these ECS elements were present in the GCs and axons of RGCs (Fig. 1). GCPDs stained lightly for all three elements, whereas GCCDs contained abundant MGL and DGLα (Figs. 1C, 1G, 1K). MGL expression gradually decreased from the GCCD toward the cell body, but was not significantly different from the GCCD in the distal axon (95% CI, 88.8%–102.1% at 15 μm proximal to the GCCD; Figs. 1C, 1D). In contrast, DGLα staining in the distal axon was more intense than in GCCDs (95% CI, 106.5%–122.4% at 15 μm proximal to the GCCD; Figs. 1G, 1H), and expression tapered beginning around the 25-μm axon segment (Figs. 1G, 1H). CB1R staining intensity increased more than twofold over the same regions (Figs. 1K, 1L). In more proximal axon segments, DGLα and MGL expression were both lower than in the GCCD (95% CI, 74.5%–99.2% and 67.8%–91.2%, respectively, at 50 μm proximal to the GCCD; Figs. 1C, 1D, 1G, 1H). In contrast, CB1R expression was approximately 2.5-fold higher in this region (95% CI, 220.3%–278.4% at 50 μm proximal to the GCCD; Figs. 1K, 1L). Immunopositivity of all three ECS components was reduced by the presence of the corresponding antigen peptide (Fig. 2). 

Figure 1

 

Localization of a 2-AG signaling cassette in growing RGC axons is consistent with formation of eCB hotspots. Representative images demonstrate subcellular localization of MGL (C), DGLα (G), and CB1R (K). ROIs for GCPDs, GCCDs, and consecutive 5-μm segments of axon were defined as indicated in blue using βIII-tubulin immunopositivity (A, E, I) and F-actin (phalloidin) staining (B, F, J). Mean gray values were recorded for each ROI, normalized to that of the GCCD, and summarized in D, H, and L as 95% confidence intervals. Arrows indicate that the 25-μm segment was used to divide the axonal domain into two intervals for piecewise linear regressions. MGL: n = 43 axons. DGLα: n = 56 axons. CB1R: n = 35 axons.

Figure 2

 

Immunopositivity for ECS components is reduced by the presence of a specific antigen peptide. Representative images demonstrate that MGL (C), DGLα (F), and CB1R (I) immunopositivity is strongly reduced in the presence of a specific antigen peptide (AP; L, O, R).

Effective 2-AG Tone at CB1R and RGC Axonal Growth

We used pharmacologic approaches to assess whether CB1R activity impacts the growth of RGC axons in culture. We first tested whether effective 2-AG tone promotes or inhibits axonal growth with the potent and selective MGL inhibitor JZL184 and the in vitro CB1R neutral antagonist O-2050.^25–27 JZL184 enhances the endogenous 2-AG tone, whereas O-2050 behaves as a neutral, or silent, cannabinoid receptor antagonist with high affinity for CB1R, at least in vitro.²⁶ O-2050 produced significantly increased median axon length that was reversed by MGL inhibition with JZL184 (Fig. 3E). 

Figure 3

 

CB1R neutral antagonism causes increases in axon growth that are sensitive to MGL inhibition. Representative images demonstrate axon growth from explant bodies as revealed by βIII-tubulin immunopositivity (A–D). Colored inset in A demonstrates how Hoechst 33258 nuclear stain was used to segment the explant body (yellow) from the axonal outgrowth area (blue). Hoechst staining was also used to locate the explant body centroid. Small blue dots in A–D identify terminations of axons tagged by a masked observer. E, Box-and-whisker plots for axon length in the presence and absence of the indicated pharmacologic agents. Axon length was calculated by subtracting the explant body diameter from the tag-to-centroid distance. Median axon length was significantly increased by O-2050, and this was reversed by JZL184 (P values as indicated; n = 370 to 514 tags per treatment across 9–10 individual explants per treatment).

We next tested whether CB1R-selective²⁸ antagonist/inverse agonist²⁹ AM251 modulates axonal growth. We found that median axon length was increased by the presence of AM251 (Fig. 4C). The length (Fig. 3E), but not the mean number, of axons per explant (Fig. 4D) was increased by O-2050, whereas AM251 increased both the median axon length (Fig. 4C) and the mean number of tagged axon tips per explant (Fig. 4E). 

Figure 4

 

CB1R antagonism/inverse agonism causes marked increases in axon growth. Representative images demonstrate axon growth in the presence or absence of CB1R inverse agonist AM251 (A, B). C, Median axon length was significantly increased by the presence of AM251 (P values as indicated; AM251 treatment, n = 610 total tags across 8 explants; vehicle, n = 476 total tags across seven explants). D, E, The mean number of axon tips tagged per explant was significantly increased by the presence of AM251 (E), but not O-2050 (D) (P values as indicated; n = 7–10 explants per treatment).

Effective 2-AG Tone at CB1R and ECS Component Subcellular Gradients

If CB1R activity regulates ECS component expression or localization, then the pharmacologic treatments described previously might be expected to alter the physiologic microgradient profiles we observed. We tested this by quantifying MGL, DGLα, and CB1R expression following MGL inhibition and CB1R blockade (Fig. 5). We selected a common ROI, the GCCD, to compare the signal intensities of the three ECS components across all treatment conditions (Figs. 5A–C). MGL intensity was increased by the presence of CB1R antagonist O-2050 and reduced below baseline by the presence of MGL inhibitor JZL184 (Fig. 5A). An increase in CB1R expression in the GCCD also accompanied CB1R antagonism (Fig. 5B), whereas DGLα expression remained flat (Fig. 5C). Although absolute MGL expression was modulated by an effective 2-AG tone, the GCCD-normalized gradient profile of MGL from the GCCD to the proximal axon remained unchanged (Fig. 5D) by the CB1R antagonism. In contrast, the normally opposed gradients of DGLα and CB1R were flattened by O-2050; DGLα was no longer relatively enriched in the distal axon and GCCD (Fig. 5E), and CB1R expression increased less dramatically when moving from the GCCD toward the soma (Fig. 5F). 

Figure 5

 

Altered effective 2-AG tone at CB1R disrupts physiologic gradients of ECS components. Neutral CB1R antagonist O-2050 increases MGL (A) and CB1R (C), but not DGLα (B) expression. Basal MGL expression is decreased by the presence of selective MGL inhibition with JZL184 (A). A–C, P values as indicated; n = 48 to 103 axons across at least two explants. D–F, CB1R neutral antagonism (blue) disrupts physiologic subcellular gradients of DGLα (E) and CB1R (F), but not MGL (D). Yellow regions indicate nonoverlapping 95% confidence intervals.

Discussion

Most knowledge of ECS-modulated glutamatergic projection growth comes from studies of developing reciprocal connections between the neocortex and thalamus.^9,11–13 The subcellular distribution of ECS components is arranged to promote efficient growth as corticothalamic and thalamocortical axons with opposing-trajectory cross paths. No such “handshake” occurs during the growth of glutamatergic retinofugal projections, and the arrangement of ECS components in regenerating RGC axons is unknown. 

We cultured embryonic retinal explants and quantified subcellular distributions of MGL, DGLα, and CB1R in growing RGC axons and growth cones with immunocytochemistry, confocal microscopy, and image analysis. Figure 6 compares the subcellular arrangement of the 2-AG signaling cassette previously observed in CTAs^9,11,12,14 with the topology we found in RGC axons (Figs. 1, 2). In CTAs, MGL is restricted from the growth cone, where DGLα is enriched and coincident with CB1R, and thalamocortical axons express MGL, but not CB1R.^9,11,12,14 The coincidence of DGLα and CB1R, but not MGL, in the CTA growth cone raises the possibility of spatially restricted, cell autonomous 2-AG signaling at that location. MGL and DGLα are abundant in the GCCD of growing RGC axons, but differential gradients of enzyme expression predict an enhanced 2-AG tone in the distal axon and GCCD (Figs. 1, 2). 2-AG-synthesizing DGLα expression is enriched in this region, whereas 2-AG-degrading MGL is more homogenously expressed throughout the axon. Interestingly, CB1R is present within GCCDs and GCPDs, including the filopodia, but most of the CB1R signal is consistently located in axon segments proximal to the putative 2-AG hotspots, which suggests the presence of opposing intracellular ligand-receptor gradients. This arrangement could be an alternative basis for the spatial restriction of CB1R signaling competence in long-range projections that do not participate in a “handshake” of reciprocal neural connections. The RGC ECS appears favorably positioned to promote central domain consolidation⁵ by modulation of the actomyosin contractility³⁰ and stabilization of the microtubules in nascent axon segments, for example, by CB1R-dependent phosphorylation-degradation of microtubule destabilizing proteins.^6,31 The abrupt transition to DGLα-poor axon segments that occurs beginning at approximately 25 μm from the GCCD is reminiscent of the abrupt transition from tyrosinated tubulin-rich to tyrosinated tubulin-poor microtubule domains that occurs in the same region.^6,32–34 This raises the intriguing possibility that CB1R activity restricted to the distal RGC axon and GCCD controls microtubule half-life and dynamic instability by influencing the balance of microtubule posttranslational modifications.^6,32–34 

Figure 6

 

Retinofugal projections exhibit a distinct arrangement of ECS components that is consistent with the formation of 2-AG hotspots. The region of the RGC axon predicted to have highest 2-AG tone is the distal axon and GCCD. This region leads most CB1R expression with respect to the direction of axonal growth, which suggests opposing intracellular ligand-receptor gradients. Pharmacologic manipulations of the ECS system altered RGC axon growth, ECS component expression levels, and profiles of ECS component subcellular microgradients.

Although some progress toward understanding the functions of the ECS in RGC axonal growth has been made,^16–18 the area can benefit from independent confirmation and extension. Attempts to determine whether CB1R activation stimulates or inhibits neuritogenesis and axonal growth in a variety of primary neuron culture models have produced mixed and apparently contradictory results.¹⁰ For example, the earliest study that identified a link between ECS signaling and RGC axonal growth implicated diacylglycerol lipase activity as a necessary component of signal transduction initiated by attractive guidance cues,³⁵ whereas recent research with newer pharmacologic and genetic approaches suggests that DGLα-derived 2-AG acting at CB1R inhibits axonal growth from RGCs.¹⁸ Similar to CTAs, RGC axons form long-range projections that target thalamic nuclei, but influential studies of cultured mid-gestation pyramidal neurons, which form CTAs in vivo, conclude that CB1R activation is sufficient to promote axonal growth.^12,13 Because previous studies to determine whether CB1R activity promotes or inhibits axonal growth have yielded contradictory results,^{10,13,14,18,35,36} we next tested whether effective 2-AG tone impacts RGC axonal growth in vitro by combining pharmacologic MGL inhibition with CB1R antagonists (Figs. 3, 4). We found that CB1R antagonists increase axon length, and this can be reversed when endogenous 2-AG is elevated by JZL184. This suggests that ligand-dependent activation of cannabinoid receptors inhibits axonal growth, but one of the agents we used, O-2050, also has high affinity for type 2 cannabinoid receptor (CB2R).²⁶ In the context of central nervous system development, CB2R expression is commonly regarded as predominant in neural progenitor cell populations, whereas CB1R is upregulated at the expense of CB2R upon commitment to a neuronal lineage.⁹ Nonetheless, evidence that CB2R activation modulates RGC axon guidance has been reported,¹⁷ so we tested whether the CB1R-selective antagonist/inverse agonist AM251 modulates RGC axonal growth. O-2050 is expected to prevent binding of endogenous ligand to CB1R without biasing the normal equilibrium between active and inactive receptor conformation.^10,26,36,37 The magnitude of AM251's effect could be caused by inhibition of both ligand-dependent and constitutive CB1R activity by this inverse agonist. Overall, these results are consistent with the view that stimulation of the classical cannabinoid receptor CB1R (and possibly CB2R) tends to inhibit developmental RGC axonal growth in vitro.^10,17,18 

Synaptogenesis is associated with rearrangements of axonal ECS components that are predicted to terminate hotspot formation.^9,12,14,19 MGL restriction from the CTA growth cone is lifted after target contact.¹² Axonal MGL expression subsequently terminates and is redistributed to the perikaryon,¹² whereas DGLα is redistributed from the axons to the dendritic fields.¹⁹ The mechanisms that control this reorganization are not fully understood, but activation of the growth cone CB1R by target-derived 2-AG could play a role. If this is correct, then pharmacologic manipulation of the ECS could disrupt normal ECS gradient profiles. We found that opposing DGLα and CB1R gradients were abolished by CB1R antagonism (Fig. 5). Moreover, if target-derived 2-AG ultimately promotes the downregulation of axonal MGL, this could explain why JZL184 accelerated the loss of axonal MGL expression. Previous studies showed that CB1R knockout mice did not significantly alter DGLα or MGL expression, but the authors were unable to exclude subtle alterations to CB1R-dependent control of these ECS components with this in vivo model.¹² Our results support the hypothesis that CB1R can exert upstream control of ECS component levels and microgradient profiles in retinofugal projections. 

Our experimental approach has a number of limitations. For example, one of the agents we used, CB1R-selective antagonist/inverse agonist AM251, is also a potent GPR55 agonist.³⁸ GPR55 activity was recently shown to promote RGC axonal growth in a similar model system,¹⁶ and we cannot rule out that the effect of AM251 is partly a result of GPR55 agonism. Similarly, we cannot rule out a contribution by CB2R, as noted earlier. Experiments with complementary molecular biology approaches are in progress to confirm and extend our findings with pharmacologic tools. Because 2-AG can be synthesized and degraded via several pathways,^39,40 our evidence supporting formation of 2-AG hotspots in RGC axons is circumstantial. In the adult brain, DGLα synthesizes most 2-AG for ECS-mediated signaling events,^39,41 but the closely related DGLβ isoform⁴² is present in certain developing axons¹³ and sometimes exhibits a subcellular localization that is distinct from DGLα.^13,43 DGLβ is expressed in developing retinofugal projections in vivo,¹⁹ but its detailed subcellular localization in RGC axons is not established. MGL is responsible for the bulk of 2-AG hydrolyzing activity in the adult brain,⁴⁴ but other bona fide ECS components (e.g., ABHD6)⁴⁵ make substantial contributions.⁴⁴ Detailed subcellular localization of other 2-AG hydrolyzing enzymes in developing axons is also not established. 

The ECS participates in diverse functions of the adult visual system ranging from corneal endothelial cell chemotaxis to retinal processing of visual information and the regulation of intraocular pressure.^46–49 The ECS also modulates visual system development by way of retinofugal projection growth and target selection.^16–18 In this context, there are several outstanding questions about the ECS with significant therapeutic implications that should be addressed by future studies. At embryonic day 15 in mice, a significant proportion of RGCs has established a nascent axon, and it is likely that our explant procedure causes axotomy of at least some of these. However, we did not explicitly examine this phenomenon, and successful regeneration of axotomized processes at this early developmental stage is expected as a matter of course. Whether elements of the ECS are expressed in successfully regenerating adult RGC axons,^1,2,50 which normally have a limited capacity to form new growth cones after injury,⁷ remains unknown. Whether the ECS has important functions in morphogenesis of adult stem cell-derived neurons has also not been addressed. Broadly, it is important to note that the classical cannabinoid receptors CB1R and CB2R are only two members of phylogenetically related subfamilies²⁵ of rhodopsin-like (class A) GPCRs, and several members share significant sequence homology and functional overlap with the cannabinoid receptors. Lysophosphatidic acid, sphingosine-1-phosphate, lysophosphoinositol, and other lipid mediators are potent ligands for a large variety of cognate rhodopsin-like GPCRs that have been shown to control axonal growth.^16,51–57 Given the large number of possible pathways for 2-AG and other eCB synthesis and degradation, a more ideal approach would be direct observation of axonal lipids. Advances in technologies such as matrix-assisted laser desorption/ionization (MALDI) imaging mass spectrometry might eventually make this feasible.^58–60 

We have presented evidence that the RGC ECS exhibits a subcellular arrangement that is consistent with the formation of eCB hotspots but distinct from other glutamatergic projections. We also showed that CB1R inhibits functional readouts associated with RGC axonal growth in vitro, and we found that effective 2-AG tone at CB1R can bidirectionally modulate the expression and topology of axonal ECS components. Overall, these results suggest that the ECS has a role in the CB1R-dependent inhibition of RGC axonal growth in vitro. 

Acknowledgments

The authors thank Motokazu Uchigashima for assistance with development of immunocytochemistry protocols and Siyang “Charlie” Chaili for assistance with image analysis. The authors acknowledge Natik Piri for helpful discussions. Supported by Research to Prevent Blindness and the UCLA EyeSTAR Program. DTS is a recipient of the Frank Stein and Paul S. May Grant for Innovative Glaucoma Research (Glaucoma Research Foundation). 

Disclosure: D.T. Stark, None; J. Caprioli, None