The Midline Axon Crossing Decision Is Regulated through an Activity-Dependent Mechanism by the NMDA Receptor

Abstract Axon guidance in vertebrates is controlled by genetic cascades as well as by intrinsic activity-dependent refinement of connections. Midline axon crossing is one of the best studied pathfinding models and is fundamental to the establishment of bilaterally symmetric nervous systems. However, it is not known whether crossing requires intrinsic activity in axons, and what controls that activity. Further, a mechanism linking neuronal activity and gene expression has not been identified for axon pathfinding. Using embryonic zebrafish, we found that the NMDA receptor (NMDAR) NR1.1 subunit (grin1a) is expressed in commissural axons. Pharmacological inhibition of grin1a, hypoxia exposure reduction of grin1a expression, or CRISPR knock-down of grin1a leads to defects in midline crossing. Inhibition of neuronal activity phenocopies the effects of grin1a loss on midline crossing. By combining pharmacological inhibition of the NMDAR with optogenetic stimulation to precisely restore neuronal activity, we observed rescue of midline crossing. This suggests that the NMDAR controls pathfinding by an activity-dependent mechanism. We further show that the NMDAR may act, via modulating activity, on the transcription factor arxa (mammalian Arx), a known regulator of midline pathfinding. These findings uncover a novel role for the NMDAR in controlling activity to regulate commissural pathfinding and identify arxa as a key link between the genetic and activity-dependent regulation of midline axon guidance.


Introduction
A key common feature in the CNS of invertebrates and vertebrates is the establishment of a bilaterally symmetric nervous system, which in turn requires forming connections between the two sides. Most axons cross the midline once and do not cross again, while a smaller percentage remain ipsilateral and never cross (Garbe and Bashaw, 2004). Genetic regulation of the midline crossing decision has been intensely studied since the 1990s, leading to the identification of canonical axon guidance gene families, including the Robos, Slits, Netrins, Dccs, Uncs, Derailed, Frizzleds, and Wnts (Dickson, 2002). Abnormalities of commissural axon tracts, caused by genetic disorders or premature birth, are associated with a variety of neurodevelopmental disabilities ranging from isolated intellectual impairment to autism spectrum disorder (Hardan et al., 2009;Izzi and Charron, 2011;. The standard model of pathfinding in vertebrates consists of two parts: an early, "hard-wired" genetically controlled phase; and later, neuronal activity refinement of connections, particularly of synapses. A defined genetic cascade, culminating particularly in the expression of cellsurface receptors and ligands, guides early pathfinding; whereas intrinsic neuronal activity appears to be regulated separately for later aspects of connectivity development (Bixby and Harris, 1991;Katz and Shatz, 1996;Tessier-Lavigne and Goodman, 1996;Yamamoto and López-Bendito, 2012;Kutsarova et al., 2017). However, recent studies have shown an early role for neuronal activity in the development of normal connectivity (Hanson and Landmesser, 2004;Spitzer, 2006). For example, the terminal projection determination of retinofugal, retinotectal, and commissural axons require neuronal activity (Schmidt et al., 2000;Mizuno et al., 2007, Dhande et al., 2011, and interhemispheric coordinated activity is necessary for final targeting of axon projections once they have crossed the midline (Suárez et al., 2014). Whether there is a mechanism linking genetic regulation and neuronal activity, or whether the two act independently, is uncertain, with conflicting evidence for both possibilities (Hanson and Landmesser, 2004;Nicol et al., 2007;Cang et al., 2008;Benjumeda et al., 2013).
Thus, key unresolved questions remain regarding the role of activity and commissural axon projections. First, it is not known whether the choice for midline crossing requires intrinsic activity in the axons. If activity is required for the midline crossing choice, what controls their activity? Second, in early development, during the period classically considered to be primarily hard-wired and genetically regulated, is there a mechanism linking neuronal activity and gene expression?
To address these questions, we tested the role of activity in the midline crossing choice using embryonic zebrafish (Danio rerio) to visualize guidance decisions. The embryonic zebrafish is well suited for analysis of axon guidance: it is transparent, facilitating visualization of pathfinding; and experimentally accessible to pharmacological, optogenetic, and genetic manipulation. To explore the requirement for activity, we examined the role of the NMDAR for three reasons: first, because of its known role in axon refinement and branch maturation (Ben Fredj et al., 2010); second, because the NMDAR is widely expressed in the early zebrafish embryo CNS (Cox et al., 2005); and third, prior work has shown that NMDARs are present in axon growth cones Gill et al., 2015), but their involvement in axon guidance has not been studied and no evaluation of their functional role has been performed.
Here, we report that the NMDAR is required for midline axon crossing. We found that the NR1.1 subunit of the NMDAR (grin1a) is expressed in a population of commissural axons. Pharmacological inhibition or CRISPR knockdown of grin1 leads to defects in midline crossing. Inhibition of neuronal activity phenocopies the effects of grin1 loss. Finally, we use optogenetic stimulation to precisely rescue neuronal activity in pharmacologically treated embryos, demonstrating that the NMDAR's role in pathfinding is mediated by neuronal activity. We find that the NMDAR may act, via modulating activity, on the transcription factor arxa (mammalian Arx), a known regulator of midline pathfinding. Thus, our study demonstrates a novel role for the NMDAR in early commissural pathfinding and demonstrates an early requirement for neuronal activity in midline axon guidance.

Ethics statement
Zebrafish experiments were performed with supervision and in accordance of guidelines from the University of Utah Institutional Animal Care and Use Committee, regulated by federal law (the Animal Welfare Act and Public Health Services Regulation Act) by the United States Department of Agriculture and the Office of Laboratory Animal Welfare at the National Institutes of Health, and accredited by the Association for Assessment and Accreditation of Laboratory Care International.

Fish stocks and husbandry
Adult fish were bred according to standard methods and staged by time and morphology (Kimmel et al., 1995). Embryos were raised at 28.5°C in E3 embryo medium with methylene blue and embryos beyond 24 h postfertilization (hpf) were treated with phenylthiourea (PTU) to prevent pigmentation.

Hypoxia conditions
To induce hypoxia, embryonic zebrafish were placed in a sealed Plexiglas chamber connected via a controller that monitored and adjusted nitrogen gas flow to a desired pO 2 set point (Biospherix, Inc.;Stevenson et al., 2012). Embryos were incubated in 1% O 2 from 24 to 36 hpf. Morphologic staging was used to determine age at fixation for analysis.
Scoring tract of the commissure of the posterior tuberculum (TCPTc) C/L ratios Evaluation of the TCPTc axon midline crossing defects was determined by the following approach. Briefly, a confocal z-stack was taken of the region using identical confocal settings (20ϫ objective, imaging speed: 8 s/ pixel, laser power: 30%, gain: 1.25%, offset: 5%). We used ImageJ to generate a maximum intensity projection of 11 slices (step size: 2 m); then subtracted the background fluorescence by using a rolling ball radius of 50 pixels. We measured the total fluorescence intensity of a rectangular area (10 ϫ 30 m) placed over the commissure of TCPT pathway, or of the longitudinal tract before its decussation into the commissure. A ratio of commissural versus longitudinal axon intensity (C/L ratio) was calculated (Stevenson et al., 2012;Xing et al., 2015). Since experimental variation was noted, results were only compared between experiments performed on the same day. At least 10 embryos were imaged for each treatment.
Immunohistochemistry was performed as described previously . Following fixation and dehydration in methanol, embryos were rehydrated, permeabilized using proteinase K [10 g/ml in PBS with 0.1% Tween 20 (PBST)] at 28°C for 60 minutes (min) (8 min for 24 hpf, 20 min for 36 hpf, and 30 min for 48 hpf) without rocking, washed twice in PBST for 5 min then re-fixed for 15 min. Embryos were then washed in PBST, blocked in PBST/1% DMSO/2% BSA/5% normal goat serum (NGS), and then incubated overnight in a primary antibody solution diluted in PBST/1% DMSO/2% BSA/2% NGS at 4°C. The next day embryos were washed in PBST/1% DMSO/1% NGS for a minimum of 6 h, followed by incubation overnight with secondary antibodies, and washed the following day.
Whole-mount in situ hybridization for NR1.1 was performed using an anti-sense probe (Cox et al., 2005) following our previous protocol (Bonkowsky and Chien, 2005).
For in situ hybridization combined with immunohistochemistry, after in situ hybridization was completed, embryos were transferred sequentially through 5%, 15%, and 30% sucrose in PBS. Embryos were then embedded in optimal cutting temperature compound (OCT; Tissue-TeK) using a dry ice-ethanol bath and stored at Ϫ80°C overnight. Fourteen-micrometer sections were cut for immunostaining. Sections were boiled for 20 min with citrate buffer (10 mM sodium citrate and 0.05% Tween 20; pH 6.0) for antigen retrieval, and immunostaining was performed as described above.

RNA extraction, reverse transcription, and quantitative real-time polymerase chain reaction (qRT-PCR)
The 24 hpf whole embryos or the head tissue of 48 hpf embryos were homogenized in TRIzol reagent (Ambion) and total RNA was extracted and purified with the RNeasy mini kit (QIAGEN). An aliquot of each extract was used for spectrophotometry to determine RNA quality and concentration. cDNA was synthesized from total RNA (2 g; 20-l final reaction volume) with oligo(dT) priming using SuperScript III reverse transcriptase (ThermoFisher) according to the manufacturer's instructions. The resulting cDNA was diluted 4 -20 times for the real time quantitative PCR. Samples were prepared with three independent biological replicates. RT-PCR was performed on an ABI prism 7900 HT instrument (Applied System) with a SYBR green fluorescence label. qRT-PCR for grin1a and grin1b was performed on 24 and 48 hpf embryos. Reactions were run with the following conditions: 95°C for 10 min and 40 cycles of 95°C 20 s (s)/60°C 20 s/72°C 40 s. Each reaction was performed in triplicate and the mean of replicates was calculated; results were normalized to the mRNA level of each gene in normoxia embryos at 24 and 48 hpf with elongation factor 1␣ (elf1a) transcript levels as a control using the 2 Ϫ⌬⌬CT method (Livak et al., 2001). For statistical analysis, Student's t test was used to compare between normoxia and hypoxia groups. Primers for grin1a were forward 5'-AAGCCAACTACGCTGGAAGG-3' and reverse 5'-CAATTGGGCGTCCTGGGAT-3'; for grin1b were forward 5'-TTGACAACAAGCGAGGACCC-3' and reverse 5'-CGTCTTCACTTGCAGACAGGA-3'; and for elf1a were forward 5'-CTTCTCAGGCTGACTGTGC-3' and reverse 5'-CCGCTAGCATTACCCTCC-3' (McCurley et al., 2008).

Grin1a and grin1b CRISPR/Cas9
One-cell stage embryos were injected with 400 pg of cas9 nuclease 3NLS (Integrated DNA Technologies) and 200 pg each of grin1a and grin1b sgRNA. sgRNAs were targeted to the following sequences in grin1a: GCAG-GAGCAGGAGAACAACG and in grin1b: GATGGCAC TCTCCGTGTGCG. sgRNA synthesis was followed the protocol as described previously. Embryos were collected at 24 hpf for mutation analysis using PCR and high-resolution melt analysis (HRMA). Genomic DNA was prepared from 24 hpf embryos and PCR was performed using the following primers: grin1a forward 5'-ATGCGT CTGCTTCTGCTGGC-3' and reverse 5'-CTTCTGGCTC AGGACAGCCC-3', and grin1b forward 5'-GTCACCCATA AGCCAAACGCC-3' and reverse 5'-GACCTGAACAGAC TGTTTTACCTGG-3'. HRMA was performed on a Light-Scanner-96 instrument (Idaho Technology) from 65°C to 95°C with a temperature transition rate of 0.1°C/s. For sequencing the targeted regions in F0 founders, PCR amplicons were cloned into pCR4-TOPO TA vector (Invitrogen). Plasmids were prepared from single colonies and Sanger sequenced to identify the mutation(s).

Microscopy and image analysis
Image acquisition and analysis were performed as described previously (Stevenson et al., 2012). Immunostained embryos were transferred sequentially into 20%, 50%, and 80% glycerol in PBST, mounted in a glass slide with a #0 coverslip placed over a well made using electrical tape, and imaged on a confocal microscope (Olympus FV-1000). Confocal stacks were projected in ImageJ and composed with Adobe Illustrator.
To compare fluorescence intensity of neuron cell bodies in control versus APV-treated embryos, we took confocal images with identical settings. Following image acquisition, we generated maximal intensity z-stack images and subtracted background using the rolling ball (100) setting in ImageJ. For each hemi-telencephalon we measured the fluorescence intensity using a square of set size which included all the cell bodies. We calculated total fluorescence intensity ϭ area‫ء‬mean intensity, and then averaged the hemi-telencephalon values for each single embryo.

Drug treatment
The drugs used in this study were 500 M D-APV (Tocris), 500 M MK-801 (Tocris) and 500 M NMDA (Tocris). Embryos were manually dechorionated at 24 hpf before drug exposure. PTU-containing E3 media were mixed with the respective drugs unless otherwise noted. Embryos were incubated with the drugs from 24 to 72 hpf and were fixed at 72 hpf.
For NMDA experiments, magnesium-free E3 medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl 2 , and 0.00002% methyl blue) were used. The embryos were incubated in PTU-containing magnesium-free E3 medium mixed with NMDA in normoxic or hypoxic conditions.

Statistical analysis
Analysis was done blinded to experimental condition and/or genotype. Experiments were performed before assignment of sex in zebrafish, during embryonic and larval stages. C/L ratios and Western blotting intensities were tested for significant differences using Student's t test. qRT-PCR p values were adjusted using Benjamini-Hochberg false discovery rate. For all analyses, p Ͻ 0.05 was considered significant. Analyses were conducted with Prism 6 (GraphPad).
We examined expression of the NMDAR subunit GluN1 (grin1a). We found that grin1a is expressed in TCPTc axons at 24 hpf, as they extend and prepare to cross the midline (Fig. 1A-B''). Next, we evaluated whether there is a source of glutamate that could serve as a ligand for the NMDAR. We observed that TCPTc axons pass near regions expressing Vglut1 (slc17a7a), the vesicular glutamate transporter expressed on neurons that release glutamate (Takamori et al., 2001; Fig. 1C-D'''). These findings show that the NMDAR is expressed in TCPT commissural axons, and that there is a glutamate source for the TCPT axons as they prepare to cross the midline.

Inhibition of NMDAR decreases midline crossing
We used pharmacological inhibition of the NMDAR to test its necessity for midline crossing. To quantify the extent of the crossing defect, we measured the ratio of commissural to longitudinal axon crossing (the C/L ratio) in the TCPT commissural axons of Tg(foxP2-enhancerA.   New Research 5-phosphonopentanoic acid), a competitive NMDAR antagonist that blocks the glutamate binding site; or MK-801 (dizocilpine), an uncompetitive NMDAR antagonist that is use-and voltage-dependent; we found a significant decrease in crossing axons at 72 hpf (one-way ANOVA; control, APV, MK-801; n ϭ 32, n ϭ 19, n ϭ 15; mean C/L 1.0, 0.61, 0.59; SD 0.43, 0.30, 0.23; F (2,63) ϭ 10.3, p Ͻ 0.001; post hoc Tukey's HSD control versus APV, p Ͻ 0.01; control vs MK-801, p Ͻ 0.01; Fig. 1E,F,I; Extended Data Figs. 1-1, 1-2). Importantly, the decrease in crossing axons was not simply due to a delay in development; the C/L ratio was also still significantly different at 96 hpf, Ͼ24 h after the end of drug exposure (n ϭ 15 each, control and APV, mean C/L 1.0, 0.76; SD 0.20, 0.30; p Ͻ 0.05, Student's t test; Fig. 1I). Also, the effect of APV was not due to an overall change in axon outgrowth. Visualization with the pan-axonal anti-acetylated tubulin antibody showed no difference between control and APV-treated embryos in terms of overall axon staining (Fig. 1G,H).
Treatment with APV or MK-801 did not adversely affect embryo morphology or development. Further, the treatment did not change properties of the reporting transgene: we found that APV did not change fluorescence of the foxP2.A.2:egfp-caax transgene line. We measured intensities of the neuron cell bodies and found that in control compared to APV-treated, average total fluorescence pixel intensity was 1.11 ϫ 10 6 versus 1.18 ϫ 10 6 (n ϭ 8 each, both hemi-telencephalon cell groups counted, SD 5.6 ϫ 10 4 vs 3.3 ϫ 10 5 , Student's t test, p ϭ 0.62). These results suggest that NMDAR function is necessary for normal midline crossing.

NMDAR interacts genetically with known modulators of midline crossing
Previous work has shown that the midline crossing decision can be modulated by exogenous signaling due to potential environmental factors, including hypoxia or exposure to serotonergic compounds (Pocock and Hobert, 2008;Bonnin et al., 2011;Stevenson et al., 2012;Xing et al., 2015). Since we had observed that pharmacological inhibition of NMDAR caused a similar midline axon phenotype, this suggested a potential shared or interacting mechanism. To explore this, we tested whether the NMDAR interacted with hypoxia and with serotonin signaling. It has been shown that hypoxia reduces TCPTc crossing (Stevenson et al., 2012), and hypoxia is known to affect NMDAR activity and expression (Kobayashi and Millhorn, 2000;Mishra et al., 2001). We assessed grin1a and grin1b, the paralogs of the GluN1 subunit of the NMDAR. First, we found that hypoxia reduced expression of grin1a and grin1b ( Fig. 2A,B). Addition of exogenous NMDA rescued the effects of hypoxia on TCPTc crossing ( Fig. 2C; one-way ANOVA; normoxia, normoxia with NMDA, hypoxia, hypoxia with NMDA; n ϭ 18 each; mean C/L 1.0, 0.93, 0.42, 0.81; SD 0.29, 0.39, 0.43, 0.42; F (3,73) ϭ 8,7, p Ͻ 0.001; post hoc Tukey's HSD normoxia vs hypoxia, p Ͻ 0.01; hypoxia vs hypoxia with APV, p Ͻ 0.05; normoxia with APV vs hypoxia, p Ͻ 0.05).
In contrast, blockade with APV did not worsen the phenotype caused by hypoxia ( Fig. 2D; one-way ANOVA; We also explored the interaction between serotonergic signaling and NMDAR's effects, because it has been shown that the serotonin receptor htr2a is necessary for TCPT crossing and is a mediator of hypoxia's effect (Xing et al., 2015). We found that increasing available serotonin by using fluoxetine, a serotonin re-uptake inhibitor, restored a normal ratio of midline crossing in APVtreated embryos ( Fig. 2E; one-way ANOVA; control, APV, fluoxetine, fluoxetine with APV; n ϭ 16 each; mean C/L 1.0, 0.56, 0.96, 1.0; SD 0.16, 0.21, 0.25, 0.24; F (3,66) ϭ 4.6, p Ͻ 0.01; post hoc Tukey's HSD, control vs APV, p Ͻ 0.05; APV vs fluoxetine, p Ͻ 0.05; APV vs APV with fluoxetine, p Ͻ 0.01). These results suggest that the role of the NMDAR in midline crossing interacts with previously described mechanisms regulating midline axon crossing. Further, this suggests that hypoxia's effects on midline crossing are caused in part by a reduction in NMDAR expression.

NMDAR CRISPR knock-down and pathfinding
Next, we evaluated whether knock-down of grin1a and grin1b using CRISPR led to reduced TCPTc crossing. After CRISPR knock-down of grin1a and grin1b, PCR amplification followed by HRMA showed loss of the wildtype alleles for grin1a and grin1b; sequencing of 16 PCR amplicons from the grin1a and grin1b locus showed a Ͼ80% mutagenesis rate at the target sites (Fig. 3A). Examples of the mutations at the grin1b locus show out of frame deletions or insertions in 14/16 clones (Fig. 3A); 2/16 clones were larger deletions or rearrangements in which the 5' end could not be identified. With CRISPR knock-down we observed a Ͼ50% reduction in GluN1 expression on a Western blotting (Fig. 3B,C). CRISPR grin1a/b knock-down led to visibly and quantifiably less crossing ( Fig. 3D-F). These results show that grin1a and grin1b mediate the function of midline crossing for the NMDAR.

NMDAR regulation of the midline crossing decision is neuronal activity dependent
The NMDAR can affect cellular processes through several mechanisms, including second messenger systems and activity dependent mechanisms (Papadia and Hardingham, 2007;Kilb et al., 2011;Zhang-Hooks et al., 2016). We had found previously that midline axon crossing of the TCPTc requires down-regulation of ephrinB2a (Stevenson et al., 2012;Xing et al., 2015), and eph/ephrins have been shown to have functional effects on NMDAR function (Dalva et al., 2000;Takasu et al., 2002;Nolt et al., 2011). We tested whether NMDAR regulated ephrinB2a levels. Following APV treatment, we did not find any changes in ephrinB2a protein levels in the telencephalon (Fig. 4A), using Western blotting with four separate experiments of protein extracts derived from embryo heads. This argues against ephrinB2a mediating NMDAR's effect on pathfinding. However, inhibition of neuronal activity, using the sodium channel blocker tricaine (MS-222) (Ramlochan-  4B). Further, similar to the observed interaction between APV blockade and rescue with fluoxetine, we found that fluoxetine could rescue the TCPT crossing defect caused by tricaine treatment (Fig. 4C). These results indirectly suggest that the NMDAR could act to control the midline pathfinding decision through an activity-dependent mechanism.
To demonstrate that the NMDAR was regulating activity levels; and to test the cell-type specificity of NMDAR action, we conducted a series of optogenetic experiments (Fig. 4D,E). Embryos were dechorionated and exposed to APV and 470-nm light exposure from 1 to 3 dpf. All embryos were double-transgenic for Tg(foxP2.A.2:egfpcaax); Tg(foxP2.A.2:Gal4-VP16). Some embryos carried Channelrhodopsin2 under the inducible control of a UAS promoter, Tg(UAS:ChR2-YFP), which was determined by post hoc genotyping. If neuronal activity could rescue pathfinding, only triple-transgenic embryos would have normal midline pathfinding in the presence of APV. We found that triple-transgenic, but not double-transgenic embryos, had normal midline TCPT crossing. in situ expression analysis c-fos of experimental embryos following 470-nm light exposure showed that embryos carrying ChR2-YFP had increased neuronal activity, indicated by increased c-fos expression (red arrow in region of foxP2.A.2 neurons; Fig. 4D). These results show that the NMDAR is acting by regulating neuronal activity; and that the activity is required specifically in the foxP2.A.2 neurons.

Arx is targeted by NMDAR
To investigate how the NMDAR could affect pathfinding, we explored downstream effectors of axon pathfinding. As mentioned above we did not find differences in EphrinB2 expression with NMDAR blockade. We tested effects on the transcription factor arxa (the zebrafish ortholog of mammalian Arx), which has been shown by mouse knock-outs and in human genetic conditions to be necessary for normal midline commissure formation (Shoubridge et al., 2010;Simonet et al., 2015). We found coexpression of arxa in foxP2.A.2 neurons during early development (Fig. 5A). To determine if arxa might be responsive to changes in NMDAR-modulated neuronal activity, we measured arxa expression after pharmacological blockade or genetic knock-down of the NMDAR. We performed qRT-PCR at 48 hpf in experimental triplicates, with n Ͼ 10 embryos in each group (Fig. 5B). The relative log2 fold change were calculated relative to wild type, using 28S rRNA standard, SD for each in parenthesis, with p values adjusted using Benjamini-Hochberg false discovery rate. We found significant decreases (p Ͻ 0.01) in arxa expression with APV or tricaine exposure, or with grin1a/b knock-down: respective log2 fold changes were 0.028 (0.023), 0.068 (0.037), and 0.54 (0.15). These results suggest that arxa might act downstream of the NMDAR to affect changes in midline crossing.

Discussion
Midline axon pathfinding is a critical process in the development of CNS circuitry, and our results indicate an early activity-dependent step regulated through the  NMDAR. To our knowledge this is the first demonstration that the NMDAR is involved in control of axon pathfinding. Interestingly, our results show that it is by regulating neuronal activity that the NMDAR affects the midline crossing decision, and potentially by targeting the transcription factor gene arxa.
NMDARs are known to play other roles in neuronal development including neuronal survival, dendritic and axonal arborization, and synapse formation (Ewald and Cline, 2009;Gambrill and Barria, 2011;Chakraborty et al., 2017). Also, NMDAR expression has been shown at axon growth cones Gill et al., 2015) but no functional role had been shown previously.
The classical model of vertebrate axon connectivity development consisted of a relatively discrete two-step process (Katz and Shatz, 1996). This was defined by an initial stage that was genetically deterministic, leading to the development of a scaffold of axon tracts. Second, an activity-dependent stage then refined axon connections and synapse development. In invertebrates such as Caenorhabditis elegans and Drosophila, the early genetic regulation of midline commissure formation appears relatively hard-wired, with stereotypy of each neuron and its projection(s) (Sánchez-Soriano et al., 2007;Chisholm et al., 2016).
However, recent data have revealed limitations of the classical model for understanding vertebrate axon path-finding. First, the significant variability in vertebrate axon tracts is not consistent with genetic hard-wiring (Lu et al., 2009;Mueller et al., 2013;Finn et al., 2015;. Second, it is now apparent that transcriptional, translational, and post-translational mechanisms all contribute to regulation of pathfinding choices (Stoeckli, 2017). Third, multiple studies have shown roles for spontaneous activity in guidance decisions earlier than had been realized (for reviewed, see Kirkby et al., 2013). This includes data showing that pathfinding decisions require rhythmic bursting neuronal activity (Hanson and Landmesser, 2004;Mizuno et al., 2007;Menelaou et al., 2008;Wang et al., 2009;Dhande et al., 2011;Plazas et al., 2013;Burbridge et al., 2014); and that corpus callosum pathfinding across the midline requires balanced hemispheric activity (Suárez et al., 2014). The spontaneous activity was shown to regulate axon guidance receptor expression, including that of EphA4, EphB1, and PlexinA3 (Hanson and Landmesser, 2004;Kastanenka and Landmesser, 2010;Plazas et al., 2013). Whether the activity directly or indirectly regulates the axon guidance receptors was not examined. Finally, environmental, nongenetic factors affect final axon trajectories and overall connectivity (Brun et al., 2009).
Our data point to an early role for neuronal activity in midline axon crossing mediated by the NMDAR. We found early expression of the NMDAR on midline crossing ax- Red boxed inset shows sense probe control for arxa. B, arxa expression is reduced following blockade (APV) or knock-down (CRISPR) of NMDAR or inhibiting activity (tricaine). qRT-PCR at 48 hpf, experimental triplicates; n Ͼ 10 embryos each group. Relative log2 fold change calculated relative to wild-type, elf1a standard, SD for each in parenthesis, p values adjusted using Benjamini-Hochberg false discovery rate; ‫‪p‬ءء‬ Ͻ 0.01; error bars, SEM.
ons, and inhibition (pharmacologic) or reduction (hypoxia, CRISPR) of the NMDAR disrupted crossing. Using precise optogenetic stimulation, we could rescue NMDAR blockade, suggesting that the NMDAR-dependent pathfinding was regulated by activity. We showed that the transcription factor arxa, a known regulator of downstream guidance receptors and a gene implicated in human midline crossing defects (Shoubridge et al., 2010;Friocourt and Parnavelas, 2011;Simonet et al., 2015), was down-regulated by NMDAR inhibition or knock-down. Interestingly, we observed that both APV and tricaine cause a greater decrease in arxa levels than grin1a/b CRISPR knock-down. A possible reason for this difference in effect is that the CRISPR knock-down targets exon 1 in the grin1a and grin1b subunits, possibly there is residual NMDAR activity by incomplete knockdown or alternative splicing (Bai and Hoffman, 2009), the former is consistent with our Western blottings showing that there is still protein expression. In contrast, APV or tricaine block any signaling or activity from the NMDAR.
In summary, these experiments provide evidence for an early activity-dependent step in midline axon pathfinding and provide a possible mechanistic insight into the regulation of activity by the NMDAR and a downstream effector via arxa. Understanding a potential link between activity regulated through the NMDAR, and axon pathfinding, may reveal unexpected insights into a range of neurodevelopmental connectivity disorders. For example, premature birth, which exposes the developing nervous system to chronic hypoxia (Martin et al., 2011), is associated with a range of disturbances in connectivity development even in the absence of destructive lesions (Kwon et al., 2016;Thomason et al., 2017).