Abstract
Oligodendrocyte progenitor cells (OPCs) are specified from discrete precursor populations during gliogenesis and migrate extensively from their origins, ultimately distributing throughout the brain and spinal cord during early development. Subsequently, a subset of OPCs differentiates into mature oligodendrocytes, which myelinate axons. This process is necessary for efficient neuronal signaling and organism survival. Previous studies have identified several factors that influence OPC development, including excitatory glutamatergic synapses that form between neurons and OPCs during myelination. However, little is known about how glutamate signaling affects OPC migration before myelination. In this study, we use in vivo, time-lapse imaging in zebrafish in conjunction with genetic and pharmacological perturbation to investigate OPC migration and myelination when the GluR4A ionotropic glutamate receptor subunit is disrupted. In our studies, we observed that gria4a mutant embryos and larvae displayed abnormal OPC migration and altered dorsoventral distribution in the spinal cord. Genetic mosaic analysis confirmed that these effects were cell-autonomous, and we identified that voltage-gated calcium channels were downstream of glutamate receptor signaling in OPCs and could rescue the migration and myelination defects we observed when glutamate signaling was perturbed. These results offer new insights into the complex system of neuron-OPC interactions and reveal a cell-autonomous role for glutamatergic signaling in OPCs during neural development.
SIGNIFICANCE STATEMENT The migration of oligodendrocyte progenitor cells (OPCs) is an essential process during development that leads to uniform oligodendrocyte distribution and sufficient myelination for central nervous system function. Here, we demonstrate that the AMPA receptor (AMPAR) subunit GluR4A is an important driver of OPC migration and myelination in vivo and that activated voltage-gated calcium channels are downstream of glutamate receptor signaling in mediating this migration.
- glutamate signaling
- myelin
- oligodendrocyte
- oligodendrocyte progenitor cell
- zebrafish
Introduction
During development, oligodendrocyte progenitor cells (OPCs) migrate from their ventral spinal cord origin and distribute throughout the CNS, and a subset differentiates into mature oligodendrocytes that myelinate axonal segments (Barres and Raff, 1999; Kucenas et al., 2008; Emery, 2010; de Faria et al., 2019; Mitew et al., 2014; Nave and Werner, 2014; Marisca et al., 2020). During this process, OPCs interact extensively with each other and their environment, and are influenced by a variety of signals, including growth factors, chemokines, neurotransmitters, and extracellular matrix components (Kirby et al., 2006; Frost et al., 2009; Hughes et al., 2013; Sanchez-Rodriguez et al., 2018; Suzuki et al., 2019). Previously, we identified adenosine signaling via neuronal A2A receptors as an upstream regulator of OPC migration (Fontenas et al., 2019). In this study, we sought to extend those findings and identify an axonally-derived cue that modulates OPC migration during development.
Previous studies demonstrate that neuronal signaling influences OPC development via various mechanisms, including OPC-expressed ionotropic AMPA receptors (AMPARs; Bergles et al., 2000; Lin and Bergles, 2004). The activation of OPC AMPARs can affect cellular processes by facilitating membrane depolarization and subsequent activation of voltage-gated calcium channels, resulting in Ca2+ influx (Gudz et al., 2006; Paez et al., 2010; Giesen et al., 2020). In vivo, neuronal activity affects oligodendrocyte axon selection for myelination and stabilization of myelin sheath formation (Hines et al., 2015; Koudelka et al., 2016), and neuronal signaling is also implicated in OPC proliferation (Barres and Raff, 1999; Gibson et al., 2014) and survival (Kougioumtzidou et al., 2017). Several studies have reported that AMPAR-mediated Ca2+ influx activates OPC migration in vitro (Gudz et al., 2006; Paez et al., 2010; Harlow et al., 2015), but this effect has not been described in vivo (Harlow et al., 2015; Kougioumtzidou et al., 2017) and currently, there is no consensus on how AMPARs affect OPC migration in vivo.
In addition to AMPAR-mediated activation of voltage-gated calcium channels, extracellular Ca2+ influx can occur via calcium-permeable AMPARs. There are eight AMPAR subunit genes encoded in the zebrafish genome, compared with four in mammals, because of a whole-genome duplication event in the evolutionary history of teleost fish (Meyer and Schartl, 1999; Sobolevsky et al., 2003). Of these eight zebrafish AMPAR subunit genes, gria2a and gria2b are calcium impermeable (Kung et al., 2001; Sobolevsky et al., 2003). The remaining six AMPAR subunit genes, including gria4a, which encodes GluR4A, are calcium-permeable (Kung et al., 2001). The calcium permeability of an AMPAR is determined by the permeability of each of its subunits (Burnashev, 1998), and calcium-permeable AMPARs are a major source of Ca2+ influx in OPCs (Ge et al., 2006). This mechanism of Ca2+ influx via calcium-permeable AMAPRs has been studied extensively in the context of excitotoxicity (Feldmeyer et al., 1999; Higuchi et al., 2000; Lee et al., 2001; Isaac et al., 2007; Evonuk et al., 2020) and extracellular Ca2+ influx can cause a wide array of cellular effects in OPCs, including changes in proliferation and differentiation (Chen et al., 2018), cell death and survival (Danesi et al., 2019), myelination (Krasnow et al., 2018), and cell morphology and migration (Gudz et al., 2006; Paez et al., 2010; Harlow et al., 2015). In primary rat cultures, OPCs and immature oligodendrocytes highly express GluR4, while GluR4 expression and Ca2+ influx was decreased in mature, myelin basic protein (MBP)-expressing oligodendrocytes, indicating that AMPAR subunit composition and downstream Ca2+ influx may control the onset of myelination (Itoh et al., 2002). However, the relationship between AMPAR subunit composition and downstream Ca2+ influx with regard to OPC migration in vivo, is not fully understood (Harlow et al., 2015).
Here, using in vivo, time-lapse imaging coupled with a new gria4a CRISPR mutant, we demonstrate that the migration and myelination patterns of spinal cord OPCs and oligodendrocytes are disrupted in the presence of perturbed GluR4A. Our results highlight, in vivo, a role for AMPAR activation downstream of neuronal activity and identify the GluR4A subunit as a regulator of OPC migration before myelination during vertebrate development.
Materials and Methods
Zebrafish
The use of all animals in this study was approved by the University of Virginia Institutional Animal Care and Use Committee. The zebrafish lines used were: AB* (wild type), Tg(olig2:dsred)vu19, Tg(olig2:egfp)vu12 (Shin et al., 2003), Tg(mbp:egfp-CAAX) (Almeida et al., 2011), Tg(sox10:megfp)sl3 (Smith et al., 2014), Gt(foxd3:mCherry)ct110R (Hochgreb-Hägele and Bronner, 2013), Tg(Xla.Tubb:dsred) (Peri and Nüsslein-Volhard, 2008), Tg(sox10:SF-iGluSnFR)uva56, and gria4auva43 (Table 1). To visualize individual oligodendrocyte lineage cells (OLCs), one-cell gria4a+/+, gria4a+/uva43 and gria4auva43/uva43 embryos were transiently injected with 2 nl of 20 ng/µl mbp:egfp-CAAX (Almeida et al., 2011). Embryos were incubated at 28.5°C in egg water, and phenylthiourea (PTU; 0.004%) was used to reduce pigment formation for imaging (Karlsson et al., 2001). Embryos of either sex were used in all experiments (Kimmel et al., 1995), and ages are listed in hours postfertilization (hpf) or days postfertilization (dpf).
Generation of gria4auva43/uva43 mutants and the Tg(sox10:SF-iGluSnFR) line
Guide RNA targeting gria4a (TGAGAGGTTCATGCACGGCG) was designed using CHOPCHOP (Labun et al., 2018) and the protocol described by Gagnon et al. (2014). The 3′ constant and 5′ gria4a-specific oligonucleotides were combined as described by Nakayama et al. (2014), and the sgRNA transcribed using the Ambion Megascript T7 kit. An injection of 2 nl of the sgRNA (200–400 ng/μl) with Cas9 protein (500 ng/μl) was performed into one-cell stage AB* embryos. Once the injected fish reached adulthood, they were bred to AB* wild-type adults and DNA isolated from a pool of 8 embryos was amplified with PCR using the primers TTAGTGTGTCAGCATTTGGGAC and TAGTTTGATCACCATCGGTTTG, then sequenced to identify potential founders. Tg(olig2:dsred) and Tg(mbp:egfp-CAAX) were crossed with the injected adult fish (F0) with a germline mutation in gria4a to produce stable heterozygote mutants. These fish (F1) were raised to adulthood, and fin clipping and sequencing was used to determine each mutation. We identified a 13-bp insertion resulting in a premature stop codon, and fish bearing this mutation were crossed to produce all subsequent generations. This 13-bp insertion produces a restriction site for MslI not found in the wild-type gria4a gene, and restriction digest of the amplified fragment followed by agarose gel electrophoresis on a 1.5% agarose gel was used for genotyping. This digest produces two bands at 149 and 75 bp for the mutant gene and does not affect the 224-bp wild-type fragment. All three bands are visible following digestion of heterozygous DNA. Whenever possible, genotyping was performed after analyzing results to allow for blind analysis of all studies.
Tg(sox10:SF-iGluSnFR) fish were generated via the Tol2kit Gateway-based cloning system (Kwan et al., 2007). Vectors used for making the Gateway expression constructs were p5E-sox10(−4.9) (Carney et al., 2006), pME-SF-iGluSnFR.A184V (Marvin et al., 2018), p3E-poly A (Kwan et al., 2007), and the destination vector pDesTol2CG2 (Kwan et al., 2007). All three expression vectors were inserted into the pDesTol2CG2 destination vector via LR reactions (Kwan et al., 2007). Resulting LR plasmids were amplified and sequenced to confirm appropriate insertion of each expression construct. After sequence confirmation, the plasmid was injected into one-cell stage AB* embryos at a concentration of 25 ng/μl with an equal concentration of Tol2 transposase mRNA. Fish expressing the pDesTol2CG2 heart reporter and SF-iGluSnFR fluorescence were raised to adulthood and were crossed to AB* adults to screen for founders (Kawakami, 2004).
Cell transplantation
Donor embryos were injected with ThermoFisher Dextran Cascade Blue dye (3000 MW) immediately following fertilization at the one-cell stage. All embryos were manually dechorionated at 4 hpf in 2% agarose-lined dishes and flamed pipettes were used to transfer dechorionated embryos. Transplantation was performed using the Eppendorf CellTram vario transplantation rig, with ∼30 cells transferred from each donor embryo and inserted into the expected spinal cord region of the 4- to 5-hpf host embryo according to fate mapping (Kimmel et al., 1995). Transplantation was performed in 1× Danieau with 1× penicillin-streptomycin, and hosts were incubated at 28.5°C in 0.3× Danieau with 1× penicillin-streptomycin following the procedure. The hosts were later imaged as described below and both donor and host embryos were kept and genotyped as described above.
Chemical treatments
Drug treatments [NBQX, (±)-Bay K 8644] were performed by adding a working solution of each drug, dissolved in 1% dimethylsulfoxide (DMSO), to 96-well plates containing one embryo per well. (±)-Bay K 8644 was applied at a treatment concentration of 5 μm as previously described (Koleilat et al., 2019). A treatment of 1% DMSO was used as a control for each drug.
Confocal imaging
All confocal images were taken with a 40× water objective (NA = 1.1) mounted on a motorized Zeiss AxioObserver Z1 microscope equipped with a Quorum WaveFX-XI (Quorum Technologies Inc.) or Andor CSU-W (Andor Oxford Instruments) spinning disk confocal system. Embryos used for imaging were anesthetized with 0.01% 3-aminobenzoic acid ester (Tricaine) and mounted in 0.8% low-melting-point agarose in 35 mm glass-bottomed Petri dishes, then immersed in egg water with PTU (0.004%). For drug treatment experiments, the immersion during imaging contained the same dosage administered in the 96-well plate, described above. In time-lapse experiments, images were taken every 10 min unless otherwise noted. Images were processed with MetaMorph and ImageJ softwares, and adjustments were limited to levels and contrast. Cell migration tracking was performed on time-lapses with images every 20 min, for consistency, using the ImageJ Manual Tracking plugin, as described below. Myelin quantification was performed through the z-stack of each image using the measurement feature of ImageJ.
Cell tracking
The ImageJ plugin Manual Tracking was used to determine speed and distance traveled of migrating cells. For each cell, a track was created by clicking the center of the cell body at each time point in a time-lapse. Distance migrated was calculated by summing the “distance” output, and migration speed was calculated by averaging the “velocity” output, both scaled to µm. Further documentation can be found on https://imagej.nih.gov/ij/plugins/track/track.html.
Myelin measurement
The “measure” feature of ImageJ was used to quantify the length of myelin internodes. To do this, the line tool was used to mark myelin sheaths appearing in either individual z-planes or z-stacks of images of Tg(mbp:egfp-CAAX) embryos or embryos transiently injected with the mbp:egfp-CAAX construct (see figure legends). The results were compiled and scaled to µm.
Photo-uncaging glutamate
We bathed embryos and larvae in 1 µm MNI-glutamate (Tocris catalogue number 1490; 25 mm stock solution in DMSO) 1 h before imaging. An equal amount of DMSO (0.4%) was used as a vehicle control. During treatment, larvae were kept in the dark as much as possible and we mounted larvae for live imaging as described above. We performed focal uncaging of MNI-glutamate with a 404-nm laser focused to a circular region of interest (50-µm2 surface, ∼20–40 µm from sox10+ OPC membranes) in the center of the acquired z-stack, followed by time-lapse imaging for 1 min, imaging every 32 ms, or for 2 h, imaging every minute.
In situ hybridization
Embryos and larvae used were fixed in 4% paraformaldehyde (PFA) overnight at 4°C and stored in 100% methanol. In situ RNA hybridization was used to localize the expression of gria4a in the zebrafish spinal cord. The primers used to produce the probes were as follows: forward primer, ACATCTCTCGCAGAGGCAAT; reverse primer, GGGCTTGGAGAAATCAATCA. Probes were prepared with DIG labeling, as previously described (Fontenas et al., 2019). After in situ hybridization, embryos and larvae were embedded in 1.5% agarose and saturated with 30% sucrose in distilled water, then frozen in 2-methylbutane cooled by immersion in liquid nitrogen. We produced 25 µm cross-sections of the embryos/larvae with a cryostat microtome, then immersed in 75% glycerol and imaged the slides with the 40× water objective (NA = 1.1) on a Zeiss AxioObserver microscope. Images were imported to Adobe photoshop, and adjustments were limited to levels, contrast, and cropping.
Immunohistochemistry
Embryos and larvae were fixed in 4% PFA for 1 h, then embedded in 1.5% agarose and saturated with 30% sucrose in distilled water. Then, they were frozen in 2-methylbutane cooled by immersion in liquid nitrogen. We produced 25 µm cross-sections of the embryos/larvae with a cryostat microtome. Antibodies used were as following: rabbit anti-Sox10 (1:1000; Binari et al., 2013) and chicken anti-GFP (Aves Lab; 1:500). Fluorescent detection of antibody labeling was performed using Alexa Fluor 555 donkey anti-rabbit (1:500) and Alexa Fluor 488 donkey anti-chicken (1:500). Slides were imaged using a 40× oil objective (NA = 1.1) on a Zeiss AxioObserver ZI microscope. Images were imported to ImageJ, and adjustments were limited to levels, contrast, and cropping. For immunohistochemistry following in situ hybridization, whole mount in situ hybridization was performed as described above and then embryos or larvae were embedded in agarose, cryosectioned, and then immunohistochemistry was performed as described.
Cell dissociation for flow cytometry
For cell dissociation, 72 hpf larvae with transgenes Tg(olig2:dsred);Tg(sox10:megfp) for OPCs, Tg(sox10:megfp);Gt(foxd3:mCherry) for Schwann cells, and Tg(nbt:dsred) for neurons, were chilled in egg water on ice, then the heads were removed using a scalpel. The cells were dissociated according to a previously published method (Zhu et al., 2019). Briefly, the trunks were immersed in calcium-free Ringer's solution with 2.5 mm EDTA and rocked for 15 min at 4°C, then washed three times with chilled Dulbecco's PBS (D-PBS) and transferred with 100 µl D-PBS into microcentrifuge tubes. The trunk pieces were broken up using a pellet pestle followed by the addition of 30 µl Liberase TM. The samples were incubated for 15 min at 28.5°C, 1 ml Trypsin with 5% EDTA was added, then the samples were incubated again for 15 min at 28.5°C. The samples were pipetted into 5-ml D-PBS with 1% BSA, then filtered by passing through a 40 µm cell strainer, and a syringe plunger was used to gently mash the samples into a Petri dish. The sample was passed through a new 40 µm cell strainer and transferred to a microcentrifuge tube, then washed twice with D-PBS with 1% BSA. Three samples of cells (1750 ± 750) of each type were used for RNA-sequencing. Total RNA was extracted using the RNeasy Micro kit (QIAGEN), then cDNA was prepared using Smart-Seq v4 Ultra Low Input RNA kit for sequencing (Takara). Library preparation was performed using the NEB Next Ultra II DNA Library Prep kit for Illumina (NEB).
RNA-seq analysis
RNA-seq analysis was performed following previously published methods (Zhu et al., 2019). Briefly, abundance of transcripts from RNA-seq datasets were imported into the DESeq2 pipeline using the R package tximport (Love et al., 2014; Soneson et al., 2015). R package AnnotationDbi was used to acquire ENTREZ IDs and Gene Symbols (Pagès et al., 2020). Normalized FPKM counts were generated using the fpkm() function in the DESeq2 package (Love et al., 2014). Differentially expressed genes were analyzed using the DESeq2 package (abs(log2FoldChange)>1&padj < 0.1; Love et al., 2014). Lists of differentially expressed genes were then used to perform functional analysis using the clusterProfiler package (Yu et al., 2012). Genes with padj = = NA were removed to exclude genes with low counts. Raw sequencing data are available on GEO: [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE174486]. Transcript abundance files are available on Mendeley Data: [https://data.mendeley.com/datasets/cpfwssf2ns/2].
RT-PCR
mRNA was extracted and cDNA synthesized as described previously (Peterson and Freeman, 2009) with the use of RNAeasy Mini kit (QIAGEN) and High-capacity cDNA Reverse Transcription kit (AppliedBiosystems) according to manufacturer instructions. Equal amounts of mRNA were used for cDNA synthesis, and RT-PCR was performed using GoTaq green mastermix (Promega). Primers used were: gria4a: forward, CAAGGACTCGGGAAGTAAGGAC and reverse, CCGAGAAGGTGAGCTTCAGTT; gria4b: forward, ACAAGGCCTGTTGGACAAATTGAAA and reverse, CACGGTAAGTCGGATGACACC; gria2b: forward, CTGAACGAGCAGGGGCTTC and reverse, AGCATCTGTGAACGTCATCCC; and ef1α: forward, GAGACTGGTGTCCTCAAGCC and reverse, CCAACGTTGTCACCAGGAGT. Expression was quantified using ImageJ and compared with relative expression in wild-type larvae.
Statistical analyses
Statistics for all quantitative experiments were conducted using GraphPad Prism. To compare between controls and treatment groups, a two-sided unpaired t test was used. To compare across all genotypes or to compare multiple treatment groups with control, a one-way ANOVA with Tukey's multiple comparison test was used. For quantification of the difference in fluorescence intensity on cell membranes relative to the background, a one-sample t test with a hypothetical value of 1 was used. Categorical data were analyzed using GraphPad Prism's one-tailed Fisher's exact test. The significance level was set at α = 0.05 (ps < 0.05 were considered significant).
Results
The AMPAR subunit gene gria4a is expressed by spinal cord OPCs
While the AMPAR subunit genes gria1a, gria1b, gria2b, and gria4a are expressed in the developing zebrafish spinal cord, it was unclear whether any were expressed by OPCs (Hoppmann et al., 2008). To determine which, if any, of these genes are selectively expressed in spinal cord OPCs, we performed RNA sequencing of OLCs (olig2+/sox10+ cells collected from the trunks of Tg(olig2:dsred);Tg(sox10:megfp) larvae at 72 hpf), Schwann cells (foxd3+/sox10+ cells collected from the trunks of Gt(foxd3:mCherry);Tg(sox10:megfp) larvae at 72 hpf), and neurons (nbt+ cells collected from the trunks of Tg(nbt:dsred) larvae at 72 hpf) and compared the expression of each AMPAR subunit gene across these cell types. RNA sequencing revealed that at 72 hpf, both gria2b and gria4a were highly expressed in OLCs, but gria4a was the subunit with unique expression in OLCs and was not highly expressed by either Schwann cells or neurons (Fig. 1A). To validate these results, we performed in situ hybridization with a mRNA probe specific to gria4a at 48 and 72 hpf and found expression in cells closely associated with the spinal cord white matter (Fig. 1B). To confirm that gria4a was expressed by OLCs, we performed in situ hybridization for gria4a and immunohistochemistry with an antibody specific to Sox10, which labels OLCs, on 80 hpf Tg(olig2:egfp) larvae, and observed Sox10+/olig2+/gria4a+ OLCs in the dorsal spinal cord (Fig. 1C). OLCs in the ventral spinal cord were typically Sox10+/olig2+/gria4a- (Fig. 1C). Together, these results indicate that the AMPAR subunit gene gria4a, which encodes the AMPAR subunit GluR4A, is expressed by a subset of spinal cord OLCs during development.
Because the AMPAR subunit gene gria4a is highly expressed by OLCs and not neurons at 72 hpf (Fig. 1A), we hypothesized that this gene was a good candidate to mediate OPC migration before myelination. To test this hypothesis, we used CRISPR/Cas9 gene editing to create mutants that perturbed the GluR4A subunit. We injected a gria4a gRNA that targeted the sixth exon of the gria4a gene (Fig. 1D), into one-cell embryos and grew these putative mutant founders to adulthood. Upon crossing them to identify germline founders carrying a mutation in gria4a, we identified one allele that caused a 13-bp insertion and subsequent frameshift mutation early in the gene after amino acid 289 (Fig. 1D). This frameshift mutation resulted in a premature stop codon, which terminates the GluR4A polypeptide before translation of the critical glutamate-binding and transmembrane domains (Fig. 1E). These mutant embryos and larvae were morphologically indistinguishable from their clutchmates both at 2 and 5 dpf (Fig. 1F), and adults were homozygous viable.
To investigate whether gria4a expression was reduced in gria4a mutant embryos and larvae, we performed in situ hybridization and this revealed that there was gria4a transcript in the spinal cord of gria4auva43/uva43 mutants, both at 48 and 72 hpf, although at lower levels than seen in wild-type controls (compare Fig. 1G and B). To confirm this reduction in transcript level, we performed RT-PCR and observed that in 48 hpf gria4auva43/uva43 embryos, gria4a mRNA expression was decreased by ∼20% when compared with wild-type larvae (mutant expression of 81.1 ± 1.3% of wild-type level; Fig. 1H), possibly caused by nonsense-mediated decay. Because loss of gria4a may result in genetic composition by other AMPAR subunits and an altered composition of subunits may change Ca2+ influx into the cells, we also analyzed expression of gria4b and gria2b. We observed that while expression of gria4b was low in wild-type larvae, expression increased both in heterozygous (146.1 ± 34.7% increase compared with the wild-type level) and homozygous larvae (142.0 ± 36.5% increase compared with the wild-type level). Additionally, gria2b expression also increased in heterozygous (123.6 ± 14.5% increase compared with the wild-type level) and homozygous larvae (124.8 ± 17.8% increase compared with the wild-type level; Fig. 1H). Thus, in gria4auva43/uva43 and gria4a+/uva43 larvae there are reduced levels of GluR4A, as well as increased expression of GluR4B and GluR2B, indicating there may be some compensation for the loss of a functional copy of the gria4a gene.
GluR4A is required for OPC migration and distribution in the spinal cord
We hypothesized that mutation of GluR4A would likely affect OPC sensing of the neurotransmitter glutamate as we recently identified neuronal activity as important for OPC migration (Fontenas et al., 2019). Previous studies demonstrate that glutamate can be released at sites along CNS axons outside of the synapse (Kukley et al., 2007; Ziskin et al., 2007; Wake et al., 2015; Kula et al., 2019). Therefore, we hypothesized that gria4auva43/uva43 mutants would exhibit abnormal OPC migration because they would be unable to sense this axonally-derived glutamate. To investigate this hypothesis, we performed in vivo, time-lapse imaging using Tg(olig2:dsred);Tg(mbp:egfp-CAAX) transgenic embryos and larvae, where olig2 and mbp regulatory sequences drive expression of fluorescent reporters in OPCs and motor neurons for olig2, and in myelin-producing cells for mbp, and imaged the dorsal migration of spinal cord OPCs (Shin et al., 2003; Almeida et al., 2011). In time-lapse movies from 56 to 72 hpf, we observed similar numbers of dorsal OPCs across genotypes at 56 hpf, which is just as these cells are migrating from their ventral spinal cord origin (Fig. 2A). However, by 72 hpf, there were significantly fewer dorsal OPCs in gria4auva43/uva43 larvae compared with wild-type clutchmates (p = 0.0041; Fig. 2B) and a similar trend was observed compared with heterozygous clutchmates (p = 0.0659; Fig. 2B). When we compared the difference between dorsal OPCs at 72 hpf and 56 hpf, OPCs in gria4auva43/uva43 larvae exhibited a significantly reduced net dorsal migration when compared with wild-type (p = 0.0083) and heterozygous (p = 0.0013) larvae (Fig. 2C). These results demonstrate that gria4a is necessary for dorsal OPC migration by 72 hpf.
To better understand what causes the reduced dorsal migration, we analyzed our time-lapse videos using the Manual Tracking ImageJ plugin for cell tracking (Fig. 2D–G). Quantification of distance migrated revealed that OPCs in gria4auva43/uva43 larvae migrated a shorter distance (p = 0.0092; Fig. 2F) and at a slower speed (p = 0.0309; Fig. 2G) than their wild-type counterparts between 56 and 72 hpf. Interestingly, although gria4a+/uva43 OPCs migrated a significantly longer distance than gria4auva43/uva43 OPCs (p = 0.0105; Fig. 2F), their migration speed, although not significantly different from either wild-type or mutant OPCs, was intermediate between the two (Fig. 2G). To determine whether the difference in migration between wild-type and mutant OPCs was the result of a developmental delay, we repeated the experiment at a later stage and imaged the larvae from 80 to 96 hpf. The migration during this period revealed that there were fewer dorsal olig2+ cells in gria4auva43/uva43 larvae compared with gria4a+/+ and gria4a+/uva43 larvae at both 80 (p = 0.0030 gria4auva43/uva43 vs gria4a+/+; p = 0.0010 gria4a+/uva43 vs gria4auva43/uva43; Fig. 2H) and 96 hpf (p = 0.0008 gria4auva43/uva43 vs gria4a+/+; p = 0.0513 gria4a+/uva43 vs gria4auva43/uva43; Fig. 2I). Additionally, there was minimal net dorsal OLC migration with no significant difference across the genotypes from 80 to 96 hpf (Fig. 2J), which demonstrates that the variation in OLC migration observed from 56 to 72 hpf was not the result of a developmental delay. Finally, we wanted to determine whether there were any changes to OPC proliferation or survival in gria4a mutant larvae, which has previously been described (Kougioumtzidou et al., 2017). To do this, we quantified the number of cell divisions and cell death events from in vivo, time-lapse imaging data taken from 56 to 72 hpf. These studies revealed that OPC proliferation and survival were unchanged in the gria4auva43/uva43 larvae when compared with heterozygous and wild-type siblings (Fig. 2K,L). Together, these findings demonstrate that mutation of GluR4A results in reduced dorsal OPC migration and altered migration dynamics of spinal cord OPCs from 56 to 72 hpf.
To independently confirm that perturbed AMPAR signaling was the cause of the dorsal OPC migration defect we observed in gria4auva43/uva43 mutants, we used the drug NBQX, an ionotropic glutamate receptor inhibitor (Deng et al., 2003; Evonuk et al., 2020) to investigate the role of AMPAR signaling in OPC migration. To do this, we treated Tg(olig2:dsred);Tg(nkx2.2a:megfp) embryos from 30 to 72 hpf with either 40 μm NBQX dissolved in 1% DMSO or with 1% DMSO as a control. We then performed in vivo, time-lapse imaging from 56 to 72 hpf and compared dorsal OPC migration between the two groups. Similar to gria4a mutant embryos and larvae, wild-type embryos treated with NBQX and DMSO larvae looked healthy (Fig. 3A). When we assayed OPC migration between 56 and 72 hpf, we observed a significant reduction in the number of nkx2.2a+/olig2+ OPCs that migrated dorsally between 56 and 72 hpf in NBQX-treated larvae compared with DMSO-treated larvae (p = 0.0305; Fig. 3B,C). Similar to what we observed in gria4a mutant larvae, OPC proliferation and cell death were not affected in wild-type larvae treated with NBQX (proliferation: DMSO 0.40 ± 0.51 cells per somite vs NBQX 0.38 ± 0.51 cells per somite; cell death: DMSO 0.20 ± 0.41 cells per somite vs NBQX 0.46 ± 0.66 cells per somite; data not shown).
To reveal any compensatory mechanisms that may be taking place in gria4a mutant larvae, we treated both gria4a+/uva43 and gria4auva43/uva43 embryos with NBQX from 30 to 72 hpf, and imaged OPC migration from 56 to 72 hpf. As OPCs already migrate significantly less in mutants, we did not expect to observe any effect of NBQX on dorsal OPC migration in mutant larvae. However, to our surprise, we observed a trend of increased dorsally-migrated OPCs at 72 hpf in NBQX-treated gria4a heterozygous larvae (p = 0.0663; Fig. 3D), and a significant increase in dorsal OPCs at 72 hpf in gria4a mutant larvae when compared with DMSO-treated mutant larvae (p = 0.0427; Fig. 3B,E). We hypothesize that this observed difference in the effect of NBQX on migration between wild-type and mutant larvae may be related to the altered composition of the AMPAR subunits in gria4a mutant larvae, thereby changing the calcium permeability of the AMPARs, and we will discuss this further in the discussion. Similar to what we observed in wild-type larvae, OPC proliferation and cell death were unchanged in heterozygous and mutant larvae treated with NBQX (gria4a+/uva43 proliferation: DMSO 0.50 ± 0.90 cells per somite vs NBQX 0.18 ± 0.40 cells per somite; cell death: DMSO 0.17 ± 0.39 cells per somite vs NBQX 0.55 ± 0.69 cells per somite; gria4auva43/uva43 proliferation: DMSO 0.18 ± 0.40 cells per somite vs NBQX 0.25 ± 0.25 cells per somite; cell death: DMSO 0.27 ± 0.47 cells per somite vs NBQX 0.17 ± 0.39 cells per somite; data not shown). Based on the effects of NBQX on wild-type and mutant OPC migration (Fig. 3C,E), we conclude that AMPAR signaling plays a key role in promoting dorsal OPC migration, and this finding provides validation for the OPC migration defects observed in gria4a mutants.
To confirm that the reduction in OPC migration was caused by an inability of the gria4auva43/uva43 OPCs to detect glutamate, we locally increased the availability of glutamate by photo-uncaging MNI-glutamate in the dorsal the spinal cord at 56 hpf (Hughes and Appel, 2020). To verify that the glutamate-uncaging method we used was functional, we used a fluorescent glutamate biosensor (SF-iGluSnFR; Marvin et al., 2018), which is a cell surface biosensor that detects extracellular glutamate, expressed under the sox10 promoter. To do this, we incubated Tg(sox10:SF-iGluSnFR) larvae with 1 μm MNI-glutamate in 0.4% DMSO for 1 h before imaging. We then mounted the larvae for in vivo imaging and focally uncaged the MNI-glutamate using a 404-nm pulsed nitrogen dye laser in a region of interest just dorsal to iGluSnFR+ cells (∼20–40 µm from measured regions on OPC membranes, in line with previous studies used for uncaging MNI-glutamate; Fig. 4A, red circle; Li et al., 2012; Hughes and Appel, 2020). We then imaged every 32 ms for 1 min before and immediately after uncaging of the MNI-glutamate and compared the t-stacks of the two time-lapses. We then measured the change in fluorescence intensity across SF-iGluSnFR+ OPC membranes and size-matched control regions immediately next to iGluSnFR+ cells (Fig. 4A–C, yellow regions mark cell membranes and cyan regions mark neighboring background). After uncaging, we observed larger changes in iGluSnFR fluorescence intensity over time on sox10+ cell membranes compared with size-matched regions in the background (Fig. 4B, representative data). Quantification of the change in fluorescence of the t-stacks revealed that the uncaging of MNI-glutamate resulted in significant changes in SF-iGluSnFR fluorescence on the cell-membranes of sox10+ cells compared with size-matched regions of background (p < 0.0001; Fig. 4C). From these results, we conclude that OPCs can detect glutamate release in the spinal cord.
Once we were confident that OPCs can respond to glutamate in vivo, we performed in vivo, time-lapse imaging from 56 to 58 hpf in Tg(olig2:dsred);gria4+/+, gria4a+/uva43, and gria4auva43/uva43 larvae after focal uncaging of MNI-glutamate in the dorsal spinal cord (uncaging occurred ∼20–40 µm from SF-iGluSnFR+ OPCs; Fig. 4C). After uncaging, we observed that gria4a+/+ OPCs increased their migration speed (p = 0.0048; Fig. 4D) and average migration distance (p = 0.0094; Fig. 4E) in response to the increased local availability of glutamate. In contrast, both gria4a+/uva43 (p = 0.0018) and gria4auva43/uva43 OPCs (p < 0.0001) migrated significantly slower (Fig. 4D) and for shorter distances (gria4a+/uva43 p = 0.0012; gria4auva43/uva43 OPCs p < 0.0001) when compared with gria4a+/+ OPCs in response to the local glutamate burst. In fact, gria4auva43/uva43 OPCs treated with DMSO were indistinguishable from gria4auva43/uva43 OPCs exposed to uncaged MNI-glutamate, demonstrating that mutation of gria4a affects OPC sensing of, and migration toward, glutamate.
Next, we sought to determine whether the altered OPC migration we observed in gria4auva43/uva43 larvae was maintained past developmental myelination stages, at 6 dpf. To investigate this, we performed immunohistochemistry on transverse trunk sections of mutant and control larvae to investigate the number and distribution of Sox10+ spinal cord OPCs and oligodendrocytes at 3 and 6 dpf. For each spinal cord section, we counted and recorded the XY-coordinates of each Sox10+ cell, then compiled these coordinates to generate heat maps for each developmental stage and genotype (Fig. 5A,B). We stratified the positional data into dorsal (upper 2/3 of Y area) and ventral (lower 1/3 of Y area) categories, roughly divided by the upper boundary of the primary motor neuron (pMN) domain. Based on these categories, we quantified dorsal and ventral Sox10+ cells in each genotype. At both stages, there were significantly fewer dorsal Sox10+ cells in the gria4auva43/uva43 spinal cord compared with the gria4a+/+ spinal cord, and there were more ventral Sox10+ cells in the gria4auva43/uva43 spinal cord compared with the gria4a+/+ spinal cord at both 3 (p = 0.0080; Fig. 5A,C) and 6 dpf (p = 0.0159; Fig. 5B,E), based on the proportion of dorsal cells in each section (n = 20 for all groups at 3 dpf, n= 10 for all groups at 6 dpf). Interestingly, at 3 dpf, there were slightly more dorsal OPCs in gria4a+/uva43 larvae compared with in gria4a+/+ larvae (p = 0.0478; Fig. 5A,C). However, at both timepoints, the total number of Sox10+ cells (both dorsal and ventral) across genotypes was similar (Fig. 5F), which demonstrates that cell proliferation and survival were not affected by the gria4a mutation, and this matches the proliferation and cell death counts we did previously (Fig. 2K,L). These results reveal that the OPC migration defect in gria4auva43/uva43 larvae persists beyond early development and that GluR4A is necessary for normal distribution of OPCs throughout the spinal cord.
Mutation of gria4a alters OPC migration in a cell-autonomous manner
Our gria4a CRIPSR mutant line is a global mutant, so we sought to determine whether the observed OPC migration differences were cell-autonomous, or alternatively, the consequence of aberrant GluR4A in other cell types, including neurons. To investigate this, we created genetic mosaics (Gansner et al., 2017). To do so, we transplanted blastula cells from dextran-injected gria4auva43/uva43;Tg(mbp:egfp-CAAX) embryos into gria4a+/+ embryos of the same stage, then used in vivo, time-lapse imaging to track the migration and dorsal/ventral location of transplanted cells from 56 to 72 hpf (Fig. 6A). At 56 hpf, 66.7% of transplanted gria4a+/+ mbp+/dextran+ cells were in the dorsal spinal cord, whereas only 37% of transplanted gria4auva43/uva43 mbp+/dextran+ cells were located dorsally (p = 0.0417; Fig. 7B). At 72 hpf, this difference was even more pronounced, and we observed 83.3% of the transplanted gria4a+/+ mbp+/dextran+ cells in the dorsal spinal cord, compared with 28% dorsally-located, transplanted gria4auva43/uva43 mbp+/dextran+ cells (p = 0.0003; Fig. 6C).
To further characterize the cell-autonomy of gria4auva43/uva43 in OLCs, we analyzed the migration dynamics of the transplanted cells and found that gria4auva43/uva43 cells transplanted into gria4a+/+ host embryos migrated at a slower speed (p < 0.0001; Fig. 6D) and for a shorter total distance (p < 0.0001; Fig. 6E) from 56 to 72 hpf, as compared with transplanted gria4a+/+ cells, similar to what we observed in gria4auva43/uva43 larvae. Interestingly, we observed the same trend with gria4a+/uva43 OLCs with respect to a slower migration speed, although it was not significantly different from wild-type OLCs (p = 0.0783; Fig. 6D). However, gria4a+/uva43 OLCs did migrate significantly less distance than OLCs in gria4a+/+ larvae (p = 0.0011; Fig. 6E), which we hypothesize is evidence of an intermediate phenotype in gria4a+/uva43 OLCs. Combined, these results indicate that the spinal cord OPC migration and regional distribution phenotypes we observed in gria4auva43/uva43 embryos and larvae are cell-autonomous, and mutation of gria4a solely in OPCs is sufficient to disrupt their migration.
To determine whether there were any non-cell-autonomous effects of gria4a on OLC migration, we transplanted blastula cells from dextran-injected gria4a+/+;Tg(mbp:egfp-CAAX) embryos into gria4auva43/uva43;Tg(mbp:egfp-CAAX) host embryos at the same stage and used in vivo, time-lapse imaging to analyze the migration dynamics of transplanted OLCs between 56 and 72 hpf. As a control, we also transplanted cells from dextran-injected gria4auva43/uva43;Tg(mbp:egfp-CAAX) embryos into gria4auva43/uva43;Tg(mbp:egfp-CAAX) host embryos to control for transplantation effects. At 56 hpf, 66.7% of the transplanted gria4a+/+ mbp+/dextran+ cells were located in the ventral spinal cord, and the remaining 33.3% were in the dorsal spinal cord, while all of the transplanted gria4auva43/uva43 mbp+/dextran+ cells were in the ventral spinal cord (p = 0.0191; Fig. 6F). At 72 hpf, 53.3% of the transplanted gria4a+/+ mbp+/dextran+ cells were located in the ventral spinal cord, and the other 46.7% were located in the dorsal spinal cord, while 94.1% of the transplanted gria4auva43/uva43 mbp+/dextran+ cells were located ventrally and only 5.9% were located dorsally (p = 0.0133; Fig. 6G). The greater number of transplanted gria4a+/+ OLCs in the dorsal spinal cord compared with the transplanted gria4auva43/uva43 OLCs is consistent with the migration effects we observed both with gria4auva43/uva43 OLCs transplanted into gria4a+/+ embryos and global gria4auva43/uva43 embryos, which provides evidence for this altered distribution pattern as a cell-autonomous phenomenon. However, we did observe a greater proportion of dorsal OLCs when cells were transplanted into gria4a+/+ embryos compared with cells transplanted into gria4auva43/uva43 embryos, which indicates that there may be a minor non-cell-autonomous effect also influencing the distribution of OLCs in the developing spinal cord.
To further explore this non-cell-autonomous effect of loss of GluR4A on OLC migration, we analyzed the migration dynamics of OLCs transplanted into gria4auva43/uva43 mutant host embryos from 56 to 72 hpf using the Manual Tracking plugin in ImageJ. Wild-type OLCs transplanted from gria4a+/+ embryos into mutant hosts migrated faster (p = 0.0005; Fig. 6H) and for a greater distance (p = 0.0215; Fig. 6I) than mutant OLCs transplanted from gria4auva43/uva43 embryos into mutant host embryos between 56 and 72 hpf (Fig. 6H,I). These differences in migration dynamics were similar to those observed in both the global gria4auva43/uva43 migration analysis and the transplant into wild-type hosts. To determine the contributions of cell-autonomous versus non-cell-autonomous effects on these migration dynamics, we compared the results from OLCs transplanted into either wild-type or mutant host embryos. The migration speed of gria4auva43/uva43 OLCs transplanted into gria4auva43/uva43 embryos was similar to that of gria4auva43/uva43 OLCs transplanted into gria4a+/+ embryos (p = 0.7859; Fig. 6J), and the speed of gria4a+/+ OLCs transplanted into gria4auva43/uva43 embryos was similar to that of gria4a+/+ OLCs transplanted into gria4a+/+ embryos (p = 0.4745; Fig. 6J). Furthermore, the distance traveled by gria4auva43/uva43 OLCs transplanted into gria4auva43/uva43 embryos was similar to that of gria4auva43/uva43 OLCs transplanted into gria4a+/+ embryos (p = 0.2511; Fig. 6K), and the distance migrated by gria4a+/+ OLCs transplanted into gria4auva43/uva43 embryos was similar to that of gria4a+/+ OLCs transplanted into gria4a+/+ embryos (p = 0.6566; Fig. 6K). Together, these results show that gria4a cell-autonomously drives OLC migration speed and distance, with some possible minor contributions of non-cell-autonomous effects to the dorsoventral distribution of OLCs.
Mutation of gria4a reduces myelination in the dorsal spinal cord
Several recent studies demonstrate that oligodendrocyte myelination of CNS axons can be activity dependent (Hines et al., 2015; Koudelka et al., 2016; de Faria et al., 2019; Noori et al., 2020). Therefore, we hypothesized that the combination of reduced dorsal OPC migration and reduced sensitivity of OLCs to neuronal activity because of mutation of gria4a would lead to a reduction in dorsal spinal cord myelination in mutant larvae. To investigate this hypothesis, we used the Tg(mbp:egfp-CAAX) transgenic line to perform in vivo imaging at 3 dpf when myelination is occurring in the developing spinal cord. From the z-stacks of these images, we quantified the number of myelin internodes in the dorsal spinal cord and measured the internode length in both mutant and wild-type larvae (Fig. 7A–C). From these studies, we observed fewer myelin internodes in both gria4a+/uva43 and gria4auva43/uva43 larvae compared with gria4a+/+ larvae at 3 dpf (p = 0.0016 gria4a+/+ vs gria4a+/uva43; p = 0.0012 gria4a+/+ va gria4auva43/uva43; Fig. 7A,B), although the average internode length was similar across genotypes (p = 0.3008; Fig. 7A,C). Notably, the slightly larger number of dorsal OPCs in heterozygous larvae at 3 dpf (Fig. 5A,C) may indicate some compensation in the heterozygous larvae for the decrease in myelination we observed.
Because we observed an interesting migration result when treating gria4a mutant larvae with NBQX, we hypothesized that it may also affect myelination in mutant larvae. To investigate this, we treated Tg(olig2:dsred);Tg(nkx2.2a:megfp) embryos from 55 to 80 hpf, which is after most OLCs have migrated out of the ventral spinal cord, with either 40 μm NBQX dissolved in 1% DMSO or with 1% DMSO as a control. We then performed in vivo imaging and compared dorsal myelination between the different genotypes. Blocking AMPAR signaling with NBQX from 55 to 80 hpf did not significantly change the number of myelin internodes per somite, nor affected the average length per internode, in wild-type and heterozygous larvae (Fig. 7D,E). However, treatment with NBQX decreased the internode length as well as number of dorsal myelin internodes in gria4a mutant larvae (number of internodes p = 0.0347, internode length p = 0.0497; Fig. 7D,E). Therefore, while NBQX blocks OPC migration in wild-type larvae (Fig. 3), it did not significantly alter myelination. On the other hand, NBQX increased OPC migration in gria4a mutant larvae (Fig. 3), while it significantly decreased myelination.
We also imaged myelin at 5 dpf in the Tg(mbp:egfp-CAAX) transgenic line, to assess whether the decreased myelination was caused by a delay in development. Because of increased myelination at this stage, our measurements of myelin internode number and length were conducted in a single z-plane for all groups. At 5 dpf, the number of myelin internodes remained significantly reduced in gria4auva43/uva43 larvae compared with gria4a+/+ larvae (p < 0.0001 gria4auva43/uva43; Fig. 8A,B), while the number of internodes in gria4a+/uva43 was now significantly larger than in gria4auva43/uva43 larvae (p = 0.0027; Fig. 8A,B), although still not completely at the same level as observed in gria4a+/+ larvae. The average internode length was not significantly different between the groups (p = 0.8968; Fig. 8C). Therefore, we conclude that both the decrease in dorsal OPC migration and myelination in gria4auva43/uva43 larvae were not caused by a developmental delay. This is in line with our observation that mutant larvae appear healthy and indistinguishable from wild-type and heterozygous larvae with regard to body length and eye size both at 2 and 5 dpf (Fig. 1F). Interestingly, at both 3 and 5 dpf, the gria4a+/uva43 myelin phenotypes were either intermediate between the gria4a+/+ and gria4auva43/uva43 phenotypes, or comparable to the gria4auva43/uva43 phenotype. However, the intermediate myelin phenotype in gria4a+/uva43 larvae was independent of a significant change in dorsal OPC distribution compared with gria4a+/+ larvae. This leads us to hypothesize that GluR4A may affect migration and myelination via distinct mechanisms, which we will address in more detail in the discussion.
In the previous set of studies, we used the stable Tg(mbp:egfp-CAAX) line, which makes it is difficult to distinguish internodes from individual oligodendrocytes. Therefore, to assess myelin on an individual cell basis, we injected the mbp:egfp-CAAX construct into gria4a+/+, gria4a+/uva43, and gria4auva43/uva43 larvae and imaged individual oligodendrocytes at 5 dpf. We observed that although the average length per internode of dorsal oligodendrocytes was comparable between the different genotypes (Fig. 8D,E), gria4auva43/uva43 oligodendrocytes made more internodes per cell compared with gria4a+/+ oligodendrocytes (p = 0.0188; Fig. 8D,F). Interestingly, there was no difference in the number of internodes or average internode length per cell in ventral oligodendrocytes across the different genotypes (Fig. 8D,G,H).
Glutamate signaling via AMPARs regulates OPC migration and myelination by modulating voltage-gated calcium channels
Several proposed mechanisms for AMPAR-mediated glutamate signaling as a regulator of OPC development involve downstream Ca2+ influx via voltage-gated calcium channels (Gudz et al., 2006; Paez et al., 2010; Harlow et al., 2015). To determine whether voltage-gated calcium channel activation plays a role in directing OPC migration, we treated gria4a mutant, heterozygous, and wild-type Tg(olig2:dsred);Tg(mbp:egfp-CAAX) embryos with either 5 μm the L-type voltage-gated calcium channel agonist (±)-Bay K 8644 in 1% DMSO or 1% DMSO alone from 55 to 72 hpf and performed in vivo, time-lapse imaging. Both wild-type and mutant larvae treated with (±)-Bay K 8644 looked healthy and indistinguishable from DMSO-treated larvae (Fig. 9A). At 56 hpf, we observed a similar number of dorsal OPCs across all genotypes and conditions (p = 0.4721; Fig. 9B). In contrast, at 72 hpf, we observed fewer dorsal OPCs in gria4auva43/uva43 embryos treated with DMSO compared with gria4a+/+ embryos treated with DMSO (p = 0.0008; Fig. 9C). However, in gria4auva43/uva43 embryos treated with (±)-Bay K 8644, there were more dorsal OPCs at 72 hpf compared with the gria4auva43/uva43 DMSO control group (p < 0.0001; Fig. 9C), and there was a similar number of dorsal OPCs in both the gria4a+/+ (±)-Bay K 8644-treated group (p = 0.5983) and the gria4a+/+ DMSO control group (p = 0.9578). These results indicate Ca2+ signaling induces OPC migration downstream of AMPAR activation (Fig. 9D), and that stimulation of voltage-gated calcium channels with (±)-Bay K 8644 effectively rescues the OPC migration phenotype in gria4auva43/uva43 embryos. Notably, OPC migration in wild-type larvae did not increase with (±)-Bay K 8644 treatment, whereas migration of OPCs in gria4auva43/uva43 larvae did. We hypothesize this is because wild-type OPCs have already reached a maximum threshold of Ca+ influx that drives OPC migration under physiological conditions.
Similar to its effect on OPC migration, we hypothesized that voltage-gated calcium channel activation with (±)-Bay K 8644 would rescue the myelin phenotype in gria4a mutant larvae. To assess this, we treated gria4a wild-type, heterozygous, and mutant Tg(mbp:egfp-CAAX) embryos with either 5 μm (±)-Bay K 8644 in 1% DMSO or 1% DMSO alone from 55 to 72 hpf or from 55 to 80 hpf, and performed in vivo imaging of myelin at 80 hpf. In gria4a+/+ larvae, treatment with (±)-Bay K 8644 did not alter the number or the length of myelin internodes (Fig. 10A–C). In gria4a+/uva43 larvae, (±)-Bay K 8644 also did not significantly change the number of myelin internodes (Fig. 10A,D). However, the average internode length was slightly decreased on a 24-h treatment (p = 0.0371; Fig. 10A,E). Interestingly, in gria4auva43/uva43 larvae treated for 24 h, but not 16 h, with (±)-Bay K 8644, we observed a significant increase in the number of dorsal myelin internodes (p = 0.0192; Fig. 10A,F), while the internode length was not altered (Fig. 10A,G). Therefore, while OPC migration is restored after 16 h of voltage-gated calcium channel activation with (±)-Bay K 8644 (Fig. 9B), voltage-gated calcium channel activation is also required for efficient myelination, as only a 24-h treatment with (±)-Bay K 8644 rescued the myelin phenotype in gria4a mutant larvae.
Discussion
During development, spinal cord OPCs must migrate extensively from their origin in the ventral spinal cord to distribute evenly throughout the CNS and myelinate axons. Several studies have examined how OPC AMPAR activation may stimulate this migration (Gudz et al., 2006; Paez et al., 2010), but in vivo experiments have yielded conflicting results (Harlow et al., 2015; Kougioumtzidou et al., 2017). One study demonstrated in vitro that in wild-type mouse OPCs, AMPA accelerates OPC migration. Notably, AMPA treatment resulted in internalization of the calcium-impermeable GluR2 subunit and increased Ca2+ influx. However, in ex vivo cerebellar slices from GluR2-null mouse, OPCs did not accelerate in response to AMPAR stimulation and resulted in reduced baseline Ca2+ influx in OPCs, possibly because of reduced neuronal activity (Harlow et al., 2015). The authors describe this effect on migration as the result of complex formation between PLP, αv-integrin, and the GluR2 subunit, which does not occur in the absence of GluR2 (Harlow et al., 2015). Thus, GluR2 was required for complex formation and initiation of migration followed by internalization of GluR2 and an increase in Ca2+ influx. Although not discussed in detail by the authors, GluR4 was also bound by to PLP complex. Our results reveal similar effects on OPC migration following GluR4A disruption, suggesting that these effects are not specific to GluR2 and that they may both act via AMPAR-mediated Ca2+ influx, not solely through PLP-integrin complex formation. We conclude from our studies that GluR4A is an important AMPAR subunit for OPC migration and provide evidence for neuron-OPC AMPAR signaling that directly affects OPC migration before myelination in vivo.
AMPAR subunit composition modulates calcium signaling in OPCs
Several studies have examined the activation of voltage-gated calcium channels as downstream mediators of AMPAR-mediated membrane depolarization in OPCs as a possible driver of developmental processes including membrane extension and myelination (Yuan et al., 1998; Butt, 2006; Gallo and Armstrong, 2008; Giesen et al., 2020). Following voltage-gated calcium channel-mediated Ca2+ influx into OPCs, numerous pathways involved in migration are activated, including PKC-dependent process extension, TrK activation, and PKA activation (Paez et al., 2010). In mice, the expression of voltage-gated calcium channels in OLCs varies with developmental stage; OPCs express these channels at a higher level, and have correspondingly higher Ca2+ currents than later stages of OLCs (Blankenfeld Gv et al., 1992). This developmental variation suggests roles for voltage-gated calcium channel-mediated Ca2+ influx at specific points in OPC development, in line with our findings that Ca2+ influx stimulates OPC migration.
The types of AMPAR subunits expressed in each cell affect its Ca2+ influx. At 72 hpf, OLCs in the zebrafish spinal cord predominantly express two AMPAR subunits: gria2b and gria4a (Fig. 1A; Hoppmann et al., 2008). A posttranscriptional Q/R modification which makes mammalian GluR2 calcium impermeable is genetically encoded in the gria2b gene in zebrafish (Kung et al., 2001; Sobolevsky et al., 2003). Like other AMPAR subunits, gria4a is calcium-permeable, and the calcium permeability of each heterotetrameric AMPAR is dependent on its constituent subunits (Burnashev, 1998). In embryos harboring a mutation in gria4a, OPCs would likely have increased levels of AMPARs with a higher degree of calcium impermeability because of the relative increase in calcium-impermeable GluR2B subunits. In line with this, the loss of GluR4 in mature mouse oligodendrocytes leads to a decrease of Ca2+ influx (Evonuk et al., 2020). While in zebrafish some genetic compensation of other calcium-permeable subunits may occur, it is likely that gria4auva43/uva43 OPCs would have reduced Ca2+ influx following AMPAR stimulation compared with their wild-type counterparts. Notably, both in gria4a mutant and heterozygous larvae, we detected increased expression of the calcium impermeable gria2b subunit gene. Therefore, gria4a+/uva43 OPCs would likely also have reduced Ca2+ influx following AMPAR stimulation compared with their wild-type counterparts, although more than in gria4auva43/uva43 OPCs as they express one wild-type copy of gria4a. This is in line with our observation that gria4a+/uva43 OPCs exhibit an intermediate phenotype with regard to their migration speed and distance traveled.
GluR4A cell autonomously drives OPC migration through glutamate sensing
In our studies, transplanted gria4auva43/uva43 OPCs in wild-type larvae exhibited altered migration dynamics similar to those observed in global gria4auva43/uva43 larvae, indicating that OPC specific perturbation to GluR4A is sufficient to alter OPC migration in a cell-autonomous manner. Interestingly, gria4a+/uva43 OPCs that were transplanted into wild-type larvae also migrated slightly slower than when in global heterozygous larvae (Fig. 6). However, in gria4a+/uva43 larvae, we observed a trend of reduced OPC migration speed, although it was not significant (Fig. 2). Therefore, it is not completely unexpected that we would see an intermediate phenotype when heterozygous OPCs were transplanted into wild-type hosts and were surrounded by a wild-type environment, which may cause the migration perturbations to heterozygous OPCs to become more obvious. Additionally, we found that wild-type OPCs transplanted into gria4auva43/uva43 embryos demonstrated the presence of mild, non-cell-autonomous effects of the loss of GluR4A on OPC migration, which may be a result of reduced neuronal signaling and subsequent reduced neurotransmitter sensing by OPCs. With the uncaging of the MNI-glutamate, we directly assessed the ability of OPCs to sense glutamate. With this method, we created a local burst of available glutamate. In control conditions, there appeared to be two migratory gria4a+/+ OPC populations: a slow-migrating and fast-migrating population. However, after a focal increase in glutamate availability after MNI-glutamate uncaging, we observed a shift in migration speed in these populations, with the previously slow-migrating gria4a+/+ OPCs increasing their speed to be comparable to the fast-migrating OPCs. Almost all gria4a+/uva43 and gria4auva43/uva43 OPCs, however, were slow migrating and did not significantly increase migration speed, nor distance, in response to glutamate uncaging. We hypothesize that this lack of response to uncaged glutamate is because of the presence of the mutated GluR4A receptor. Therefore, the speed and distance of OPC migration appears to be regulated by glutamate signaling and not the inherent ability of OPCs to migrate.
Regulation of myelination by AMPAR signaling
While myelination can occur in an activity-independent manner (Barres and Raff, 1999; Bechler et al., 2018; de Faria et al., 2019), it can also be highly regulated by neuronal signaling (Barres and Raff, 1993; Hughes et al., 2013; Gautier et al., 2015; Hines et al., 2015; Mensch et al., 2015; de Faria et al., 2019). Activity-dependent myelination can be mediated by multiple mechanisms, including those involving vesicle release (Gautier et al., 2015) and NMDA receptor signaling (Lundgaard et al., 2013), and Ca2+ transients are known to stabilize myelin sheath formation and promote sheath elongation (Hines et al., 2015; Baraban et al., 2018; Krasnow et al., 2018). Because numerous interrelated pathways influence myelination, it is not surprising that we observed that gria4a+/uva43 oligodendrocytes produced less dorsal myelin than wild-type controls, as GluR4A may affect migration and myelination through distinct mechanisms. This also became apparent in our experiments with NBQX.
Previous studies have demonstrated that in rodents, AMPAR signaling regulates OPC differentiation (Gautier et al., 2015; Chen et al., 2018) and remyelination (Gautier et al., 2015). Interestingly, in vitro studies demonstrate that before the expression of MBP, there is a change in the AMPAR subunits expressed in OLCs (Itoh et al., 2002). Rat OLCs express GluR4 at an early stage, but in mature oligodendrocytes, the expression of GluR4 is decreased, hence resulting in relatively more calcium impermeable AMPAR containing GluR2.
Based on this, we hypothesize that calcium permeable AMPAR signaling enhances migration and/or blocks differentiation into mature oligodendrocytes, while calcium impermeable AMPAR signaling inhibits migration and/or stimulates differentiation. We were especially fascinated by the results obtained after treatment with NBQX, which has an opposite effect on gria4a+/+ and gria4auva43/uva43 OPC migration and myelination. In gria4a mutant larvae, where there are more calcium impermeable AMPARs because of a mutation in gria4a and increased expression of gria2b, OPC migration, differentiation, and myelination are regulated differently than in wild-type larvae. Because NBQX is a generic AMPAR antagonist, it can inhibit all AMPA subunits. In wild-type larvae, NBQX treatment from 30 to 72 hpf would block GluR4A function, thus blocking OPC migration. However, in our gria4auva43/uva43 larvae, NBQX treatment from 30 to 72 hpf would also have a significant block of GluR2B function because of the increased expression levels of this subunit, which could therefore, result in increased migration and also affect myelination. These data demonstrate that fine-tuned, temporal control of expression of distinct AMPAR subunits is required for both efficient dorsal OPC migration and myelination, and that this balance between calcium-permeable and calcium-impermeable subunits determines when OPCs migrate and differentiate.
Downstream of glutamate sensing, Ca2+ influx drives OPC migration and initiation of differentiation, as evidenced by the rescued OPC migration and myelination in gria4auva43/uva43 embryos treated with L-type voltage-gated Ca2+ channel agonist (±)-Bay K 8644. This is in line with a previous studies that revealed that voltage-operated Ca2+ channels affect OPC migration and that voltage-gated Ca2+ influx in oligodendroglial cells is critical for normal myelination (Cheli et al., 2016) and Ca2+ transients are known to stabilize myelin sheath formation and promote sheath elongation (Hines et al., 2015; Baraban et al., 2018; Krasnow et al., 2018). In another study using primary cultures of mouse OPCs, L-type voltage gated channel signaling increased OPC morphologic differentiation as well as the expression of mature oligodendrocyte markers (Cheli et al., 2015). In our work, we do not observe any increase in the number of myelin internodes or the average internode length in response to (±)-Bay K 8644 in gria4a+/+ larvae, and the average myelin internode length in gria4a+/uva43 larvae even slightly decreased. Interestingly, a recent study in zebrafish demonstrated that some OPCs participate extensively in signaling involving Ca2+ transients, but preferentially proliferate rather than differentiate into myelin-producing cells (Marisca et al., 2020). However, in our study we did observe a positive effect of (±)-Bay K 8644 on myelination in gria4auva43/uva43 larvae. Notably, while OPC migration in gria4auva43/uva43 larvae is restored after 16 h of voltage-gated calcium channel activation with (±)-Bay K 8644, there is only a rescue of myelination on a 24-h treatment with (±)-Bay K 8644. Therefore, is seems that Ca2+ influx via voltage-gated calcium channels is especially important for initiation of myelination.
AMPAR signaling affects multiple processes in OLCs
The effects of inhibiting AMPAR signaling in OLCs are context-specific and diverse. One recent study reported that an induced, mature oligodendrocyte-specific knock-out of gria4 in mice ameliorated EAE symptoms by reducing oligodendrocyte vulnerability to excitotoxicity (Evonuk et al., 2020). In multiple sclerosis (MS), insufficient migration of OPCs to lesions results in poor remyelination (Boyd et al., 2013). NBQX inhibits the early phases of remyelination in vivo in toxin-induced demyelinated lesions in the rat brain, which was attributed to a direct effect on OPCs (Gautier et al., 2015). Inhibition of neuronal activity increased the numbers of OPCs in these demyelinated lesions, while OPC differentiation was reduced (Gautier et al., 2015). Additionally, the loss of GluR4 in rat brain inhibits OPC survival (Kougioumtzidou et al., 2017). Our findings show that GluR4A AMPAR activation drives OPC migration, which raises the question of whether GluR4A-deficient zebrafish would have enhanced remyelination capacity because of reduced vulnerability to excitotoxicity, or would fail to remyelinate MS-like lesions because of a defect in OPC migration, as seen in early myelination of the dorsal spinal cord. The expression of AMPAR subunits in OLCs is tightly regulated; changing one subunit can control migration, proliferation, differentiation, and/or myelination depending on the spatiotemporal environment, which makes OPC AMPAR signaling an exciting area for future research.
Footnotes
- Received October 2, 2020.
- Revision received April 28, 2021.
- Accepted May 3, 2021.
*M.P. and I.L.W. are co-first authors.
This work was supported by the Arnold and Mabel Beckman Foundation (M.P.), The Owens Family Foundation (S.K.), and National Institutes of Health Grants NS072212 (to S.K.) and NS107525 (to S.K.). We thank members of the Kucenas Lab, past and present, for valuable discussions, Dr. Alexandria Hughes and Dr. Laura Fontenas for technical help with glutamate uncaging in vivo, and Lori Tocke for zebrafish care.
The authors declare no competing financial interests.
- Correspondence should be addressed to Sarah Kucenas at sk4ub{at}virginia.edu
- Copyright © 2021 the authors