The Rac-GAP alpha2-Chimaerin Signals via CRMP2 and Stathmins in the Development of the Ocular Motor System

Abstract

A precise sequence of axon guidance events is required for the development of the ocular motor system. Three cranial nerves grow toward, and connect with, six extraocular muscles in a stereotyped pattern, to control eye movements. The signaling protein alpha2-chimaerin (α2-CHN) plays a pivotal role in the formation of the ocular motor system; mutations in CHN1, encoding α2-CHN, cause the human eye movement disorder Duane Retraction Syndrome (DRS). Our research has demonstrated that the manipulation of α2-chn signaling in the zebrafish embryo leads to ocular motor axon wiring defects, although the signaling cascades regulated by α2-chn remain poorly understood. Here, we demonstrate that several cytoskeletal regulatory proteins—collapsin response mediator protein 2 (CRMP2; encoded by the gene dpysl2), stathmin1, and stathmin 2—bind to α2-CHN. dpysl2, stathmin1, and especially stathmin2 are expressed by ocular motor neurons. We find that the manipulation of dpysl2 and of stathmins in zebrafish larvae leads to defects in both the axon wiring of the ocular motor system and the optokinetic reflex, impairing horizontal eye movements. Knockdowns of these molecules in zebrafish larvae of either sex caused axon guidance phenotypes that included defasciculation and ectopic branching; in some cases, these phenotypes were reminiscent of DRS. chn1 knock-down phenotypes were rescued by the overexpression of CRMP2 and STMN1, suggesting that these proteins act in the same signaling pathway. These findings suggest that CRMP2 and stathmins signal downstream of α2-CHN to orchestrate ocular motor axon guidance and to control eye movements.

SIGNIFICANCE STATEMENT The precise control of eye movements is crucial for the life of vertebrate animals, including humans. In humans, this control depends on the arrangement of nerve wiring of the ocular motor system, composed of three nerves and six muscles, a system that is conserved across vertebrate phyla. Mutations in the protein alpha2-chimaerin have previously been shown to cause eye movement disorders (squint) and axon wiring defects in humans. Our recent work has unraveled how alpha2-chimaerin coordinates axon guidance of the ocular motor system in animal models. In this article, we demonstrate key roles for the proteins CRMP2 and stathmin 1/2 in the signaling pathway orchestrated by alpha2-chimaerin, potentially giving insight into the etiology of eye movement disorders in humans.

Introduction

The ocular motor system (OMS) provides an excellent model in which to study the mechanisms that control the axon guidance of motor neurons. Defects in the development of this system, composed of three nerves and six muscles, cause eye movement disorders (squint) and even partial blindness. Our research, and that of other groups, has revealed that the signaling protein alpha2-chimaerin (α2-CHN) is a key regulator of axon guidance in the ocular motor system (Miyake et al., 2008; Chilton and Guthrie, 2017).

We found that mutations in the chimaerin-1 gene (CHN1), encoding α2-CHN, cause the eye movement disorder Duane Retraction Syndrome (DRS; Miyake et al., 2008). DRS arises because of miswiring of the ocular motor system, as revealed by neuroimaging in patients (Demer et al., 2007a). Manipulations of α2-CHN signaling in animal models leads to striking disruptions of ocular motor connectivity, akin to DRS phenotypes in humans (Ferrario et al., 2012; Clark et al., 2013; Nugent et al., 2017). We have demonstrated that global downregulation of α2-chn signaling in the larval zebrafish causes axon guidance defects that include axon defasciculation, ectopic branching, and axon overshooting of muscle targets, as well as impairing eye movements. Conversely, overexpressing wild-type (WT) or gain-of-function (GOF) α2-CHN isoforms (G228S and L20F GOF mutants found in DRS patients) in single oculomotor axons leads to axon stalling and a loss of directionality in axon growth (Clark et al., 2013).

We have proposed that the role of α2-CHN is to integrate chemoattractive and chemorepulsive guidance information, leading to cytoskeletal remodeling in motor neurons. Chemoattractants present in the ocular motor system include CXCL12 and hepatocyte growth factor (HGF), whereas Semaphorin3A (Sema3A) ligand acts via Plexin A receptors as a chemorepellent (Lerner et al., 2010; Ferrario et al., 2012). Despite these insights into the role of α2-CHN and its upstream regulators, the signaling pathways that lie downstream of α2-CHN to regulate the cytoskeleton and steer growth cones are largely unknown, though changes in microtubule (MT) dynamics are a likely factor (Chilton and Guthrie, 2017; Whitman and Engle, 2017).

α2-CHN is a Rac-GAP protein that resides in the cytosol, but moves to the membrane in response to signaling, adopting an open conformation to engage with binding partners (Yang and Kazanietz, 2007; Colón-González et al., 2008). Among the known α2-CHN interactors are Rac1, collapsin-response-mediator protein 2 (CRMP2; encoded by the dpysl2 gene) and cyclin-dependent kinase 5 (Cdk5), which are cytoskeletal regulators (Brown et al., 2004). Rac1 is known to be a key effector of α2-CHN signaling, whereas CRMP2 plays a role in the α2-CHN-dependent migration of cortical neurons and in the Sema3A collapse response (Brown et al., 2004; Ip et al., 2012). It is unknown whether any of these candidates function in ocular motor system development.

We have used a proteomics-based approach to identify candidate α2-CHN-binding partners, and here we report that CRMP2 and stathmins (STMN1 and STMN2) are α2-CHN interacting proteins. Our experiments show that these proteins bind differentially to wild-type and mutant isoforms of α2-CHN, suggesting they may play a role in the etiology of DRS.

We have performed functional assays in the zebrafish to evaluate the role of these proteins in ocular motor wiring. Knocking down chn1, dpysl2, or stathmin 1 (stmn1)/stathmin 2 (stmn2) caused axon guidance defects in zebrafish larvae in vivo and an impairment of horizontal eye moments in the optokinetic reflex (OKR) behavioral assay. Overexpression of dpysl2 or stathmin 1/2 rescued defects caused by chn1 knockdown (KD), suggesting that these molecules participate in a signaling pathway. These data suggest that CRMP2 and stathmins are part of the cellular machinery that contributes to α2-CHN signaling during ocular motor development.

Materials and Methods

Fish breeding and maintenance.

Zebrafish (Danio rerio) were maintained at 28.5°C with a 10/14 h light/dark cycle; embryos of either sex were collected after natural spawning and were staged by the number of hours postfertilization, as previously described (Kimmel et al., 1995). Wild-type (AB background), Tg(isl1BAC:Gal4ff) (RRID:ZFIN_ZDB-ALT-121204–4), and Tg(UAS:GFP) (RRID:ZFIN_ZDB-ALT-080528–1) zebrafish strains were used in the present study (Tessadori et al., 2012; Asakawa et al., 2013).

RNA microinjection.

Zebrafish embryos of either sex were microinjected at the one-cell stage with glass micropipettes containing the corresponding 3 ng of morpholino (MO) or 50 pg of mRNA (or a combination of both) in 1× Danieau solution: 58 mm NaCl, 0.7 mm KCl, 0.4 mm MgSO₄, 0.6 mm Ca(NO₃)2, and 5 mm HEPES, pH 7.6 (Kimmel et al., 1995). For the whole mount experiments, the injected embryos were treated with 200 mm phenylthiourea at 24 h postfertilization (hpf) and were maintained until 72 hpf to ensure transparency for imaging.

Morpholino design and microinjection.

The MOs used to downregulate chn1 and dpysl2 expression were a previously published chn1 MO (Clark et al., 2013) targeting the exon 1/intron 1–2 boundary and a dpysl2 MO directed to the intron 1–2/exon 2 boundary (O'Brien et al., 2009). MOs designed against the two stmn1 and stmn2 zebrafish paralogs were used to block the translation of zebrafish stmn1a, stmn1b, stmn 2a, and stmn2b by targeting the ATG site (GeneTools). All of the following morpholinos and a nontargeting standard oligo were used as a control (GeneTools): chn1: 5′-GAGGACTCACCGAACACATGGATGG-3′; dpysl2: 5′-CACTCTGGAAACACAGATAAACACA-3′; stmn1a: 5′-GAATGTCACTTGTAGCAGCCATTGT-3′; stmn1b: 5′- CTCCAGAGGACGCCATTGCACTTTC-3′; stmn2a: 5′-CATGGTGTCGTTTCTTGAATCCGTA-3′; stmn2b: 5′-ACTGGTGCGTCTCAGCTTTA-3′; and Standard MO: 5′-CCTCTTACCTCAGTTACAATTTATA-3′.

Three nanograms of each of these morpholinos were injected into one-cell stage embryos.

RNAscope in situ hybridization.

Analysis of the expression patterns of stmn1a, stmn1b, stmn2a, stmn2b, and dpysl2b in larval zebrafish was performed using RNAscope (RNAscope Multiplex Fluorescent Reagent Kit V2, ACDBio; Gross-Thebing et al., 2014). Both a positive and negative control probe were used (designed by ACDBio). The negative control probe targets dapB of Bacillus subtilis, and the positive probe targets the zebrafish gad1b. The probes against stmn1a, stmn1b, stmn2a, stmn2b, and dpysl2b were also designed by ACDBio and covered the following mRNA sequences (GenBank accession numbers are in parentheses): 2–812 of stmn1a (NM_001040375.1); 525–1648 of stmn1b (NM_001017850.1); 774–2123 of stmn2a (NM_001005923.2); 2–1468 of stmn2b (NM_001024222.1); and 12–1378 of dpysl2b (XM_005155631.4).

In situ hybridization was performed according to the manufacturer's instructions. Briefly, Tg(isl1BAC:Gal4ff;UAS:GFP) zebrafish embryos of either sex were fixed at 48 hpf in 4% paraformaldehyde (Thermo Fisher Scientific) for 24 h. Larvae were dehydrated in methanol and processed, and the relevant probes were added for 2 h at 40°C. The samples were treated with Opal 650 Reagent (catalog #FP1496001KT, PerkinElmer) for 30 min at 40°C to visualize a fluorescent signal. The GFP signal was counterstained using an anti-GFP antibody (catalog #ab290, Abcam; RRID:AB_303395) overnight at 4°C. The ocular motor nuclei were then imaged with the larvae mounted dorsal side up, using an SP8 confocal microscope, and the images were processed using ImageJ (RRID:SCR_003070).

DNA constructs.

YFP constructs for mammalian expression were generated by introducing a previously cloned 2.6 kb human α2-CHN cDNA (Miyake et al., 2008), fused to YFP at the amino terminus, into a pCAG-IRES-GFP vector in which the IRES-GFP region has been substituted with the YFP-α2-CHN fusion protein coding sequence. In addition to WT α2-CHN (pCAG-YFP-α2-CHN-WT), two mutant α2-CHN isoforms harboring either the p.G228S substitution (pCAG-YFP-α2-CHN-G228S) or the p.L20F substitution (pCAG-YFP-α2-CHN-L20F; Miyake et al., 2008) were also generated by the same subcloning strategy using the UAS constructs generated previously (Clark et al., 2013). A pCAG-YFP construct was used as a control.

The stmn1 gene was cloned into the pUC57 vector by GenScript and subsequently cloned into the pCAG-YFP-α2-CHN construct in place of α2-CHN by PCR using the following primers: forward 5′-GCTAGCGCaTCTTCTGatatccagg-3′ and reverse 5′-ctcgagttagtcagcttcagtctcg-3′. The stmn2 gene was amplified from RNA extracted from adult brain tissue (kind gift from M. Hafezparast, University of Sussex, Brighton, UK) and cloned into the same pCAG-YFP vector using the following primers: forward 5′-GCTAGCATGGCTAAAACAGCAATGG-3′ and reverse 5′-CTCGAGTCAGCCAGACAGTTCAAC-3′. To generate the constructs used for coimmunoprecipitation, the YFP of the pCAG-YFP-STMN1 and pCAG-YFP-STMN2 vectors was replaced with mRFP, amplified from an EB1-mRFP vector (kind gift of M. Hafezparast, University of Sussex) using the following primers: forward, 5′-GAATTCatggcAtcAtccgaggacgtc-3′ and reverse 5′-GCTAGCTgcTccggtggagtg-3′.

For the RNA injections, a pCS2-GFP plasmid was used for the generation of GFP mRNA as a control. For the alpha2-chimaerin rescue experiments, an α2-CHN mRNA generated from a pCS2-α2-CHN plasmid coding for human α2-CHN was used. For the CRMP2 rescue experiments, a dpysl2 mRNA generated from a pCS2-5myc-dpysl2 plasmid coding for an MO-resistant version of the dpysl2 mRNA was used (Tanaka et al., 2012). For the stathmin rescue experiments, YFP-fused STMN1 was cloned into the pCS2-5myc-dpysl2 construct in place of 5myc-dpysl2. mRNA was produced in each case using Invitrogen mMessenger mMachine SP6 (Thermo Fisher Scientific) after linearization of the pCS2 plasmids, and the mRNA was purified using the RNA Clean and Concentrator kit (Zymo Research) according to the manufacturer instructions.

Primary tissue culture and transfection of rat cortical neurons.

Wistar rats (RRID:RGD_13508588) pregnant with embryonic day 18 embryos were supplied by Charles River UK. Embryos were dissected as previously described (Baskaran et al., 2018). For the transfections, 6 × 10⁶ dissociated cortical neurons per transfection were centrifuged at 1800 rpm for 5 min. The neurons were transfected with an Amaxa nucleofector 2b machine using rat neuron nucleofector solution (Lonza). The cell pellet was resuspended in 100 µl of nucleofector solution with 3 µg of DNA, transferred into the Amaxa cuvette and transfected using the O-003 program. 500 µl of pre-equilibrated culture medium was added, and cells were plated at a density of 2–3 × 10⁶ cells in 2 ml of culture medium per well. Transfection efficiency was ∼70%.

Neurons were cultured in six-well plates (Nunc) that had previously been coated with poly-lysine and laminin (Sigma-Aldrich). The culture medium was Neurobasal medium supplemented with 2% B-27, 1% glutamax, and 1% penicillin-streptomycin (all from Thermo Fisher Scientific) with 2% horse serum (Sigma-Aldrich). Cultures were incubated at 37°C with 5% CO₂ for 72 h (Baskaran et al., 2018). Cells were collected in Lysis buffer for immunoprecipitation (ChromoTek).

HEK 293T cells (kind gift of L. Lagnado, University of Sussex) were maintained in DMEM supplemented with 10% fetal bovine serum, glutamine, and penicillin/streptomycin (all from Thermo Fisher Scientific). Cells were transfected using the calcium phosphate method and cultured for 48 h before lysis (Lysis buffer, ChromoTek).

Immunoprecipitation and Western blot.

The GFP-TRAP_A (catalog #gta-20, ChromoTek; RRID:AB_2631357) system was used for the immunoprecipitation of YFP-α2-CHN-WT or mutant isoforms transfected into rat cortical neurons or HEK293T cells, and 1 mg of protein extract were used in each case. In each case, 20 µg of protein per lane was loaded for nonimmunoprecipitated extracts, and 25 µl of the total 100 µl eluted was used for the immunoprecipitated extracts. Western blots of zebrafish extracts were performed by lysis in RIPA buffer of 20 larvae at 3 d postfertilization (dpf).

To visualize the results of the immunoprecipitation, 25 µl of the precipitated extract was loaded onto a 10% acrylamide gel and analyzed by Western blot. For the Western blot, the following antibodies were used at the indicated concentrations: anti-chimaerin (1:10,000; catalog #ab156869, Abcam; RRID:AB_2732883); anti-GAPDH (1:5000; catalog #MAB374, Millipore; RRID:AB_2107445); anti-GFP (1:3000; catalog #ab5450, Abcam; RRID:AB_304897); and anti-DsRed (1:2000; catalog #632496, Takara Bio; RRID:AB_10013483). The following secondary antibodies were used: HRP anti-rabbit (1:5000; catalog #31460, Thermo Fisher Scientific; RRID:AB_228341); HRP anti-mouse (1:5000; catalog #A9044, Sigma-Aldrich; RRID:AB_258431); and HRP anti-goat (1:5000; catalog #A15999, Thermo Fisher Scientific; RRID:AB_2534673).

Proteomic experimental methods.

In sample preparation, rat cortical neurons were transfected either with a YFP plasmid as a nonspecific binding control or with one of the three plasmids expressing YFP-α2-CHN wild-type or one of the mutant isoforms. A total of 500 µg of protein extracts from each transfection were obtained and immunoprecipitated as described previously. Each transfection and immunoprecipitation was performed in triplicate and analyzed by mass spectrometry to identify the proteins that co-immunoprecipitated with α2-CHN.

Replicates of immunoprecipitated samples of rat cortical neurons were loaded on a 10% acrylamide gel. 1D gel bands were transferred into a 96-well PCR plate. Five bands per lane were cut into 1 mm² pieces, destained, reduced (DTT), and alkylated (iodoacetamide), and subjected to enzymatic digestion with chymotrypsin overnight at 37°C. After digestion, the supernatant was pipetted into a sample vial and loaded onto an autosampler for automated liquid chromatography (LC)-tandem mass spectrometry (MS/MS) analysis.

All LC-MS/MS experiments were performed using a Dionex Ultimate 3000 RSLC nanoUPLC system (Thermo Fisher Scientific) and a Q Exactive Orbitrap Mass Spectrometer (Thermo Fisher Scientific). The separation of peptides was performed by reverse-phase chromatography at a flow rate of 300 nl/min and a reverse-phase nano Easy-Spray Column (PepMap C18; particle size, 2 µm; pore size, 100 Å; inner diameter, 75 μm; length, 50 cm; Thermo Fisher Scientific). Peptides were loaded onto a precolumn (PepMap 100 C18; particle size, 5 µm; pore size, 100 Å; inner diameter, 300 µm; length, 5 mm; Thermo Fisher Scientific) from the Ultimate 3000 Autosampler with 0.1% formic acid for 3 min at a flow rate of 10 µl/min. After this period, the column valve was switched to allow elution of peptides from the precolumn onto the analytical column. Solvent A was water + 0.1% formic acid and solvent B was 80% acetonitrile, 20% water, plus 0.1% formic acid. The linear gradient used was 2–40% B in 30 min.

The LC eluent was sprayed into the mass spectrometer by means of an Easy-Spray source (Thermo Fisher Scientific). All mass/charge ratio (m/z) values of eluting ions were measured in an Orbitrap mass analyzer, set at a resolution of 70,000 and was scanned between m/z of 380–1500. Data-dependent scans (Top 20) were used to automatically isolate and generate fragment ions by higher-energy collisional dissociation (HCD; normalized collisional energy, 25%) in the HCD collision cell, and measurement of the resulting fragment ions was performed in the Orbitrap analyzer, set at a resolution of 17,500. Singly charged ions and ions with unassigned charge states were excluded from being selected for MS/MS, and a dynamic exclusion window of 20 s was used.

Database searching.

After the run, the data were processed using Protein Discoverer (version 2.1.; Thermo Fisher Scientific). Briefly, all MS/MS data were converted to mgf files, and the files were then submitted to the Mascot search algorithm (Matrix Science) and searched against the UniProt Rattus norvegicus database (April 2017) and common contaminant sequences. Variable modifications of oxidation (M), deamidation (NQ), and carbamidomethyl were applied. The peptide and fragment mass tolerances were set to 5 ppm and 0.1 Da, respectively. A significance threshold value of p < 0.01 and a peptide cutoff score of 20 were also applied.

Immunocytochemistry and proximity ligation assay.

Immunocytochemistry and proximity ligation assay (PLA) were performed on cultured primary rat cortical neurons. The neurons were transfected with YFP as a control, or with wild-type or mutant forms of YFP-tagged α2-CHN. Cells were coimmunostained with anti-GFP primary antibody (1:3000; Abcam catalog #ab5450, RRID:AB_304897), which recognizes the YFP-tag on transfected α2-CHN, along with either anti-CRMP2 (1:250; Sigma-Aldrich catalog #SAB4500669, RRID:AB_10743989), anti-stathmin 1 (1:400; Abcam catalog #ab52630) or anti-stathmin 2 (1:3000; Abcam, catalog #ab185956, RRID:AB_2773045) primary antibodies. Fluorescent secondary antibodies anti-rabbit Alexa Fluor 568 (1:1000; Thermo Fisher Scientific catalog #A10042, RRID:AB_2534017) and anti-goat Alexa Fluor 488 (1:1000, Thermo Fisher Scientific catalog #A-11 055, RRID:AB_2534102), were then added and DAPI (1:1000; Sigma-Aldrich D9542) was used to visualize the nucleus of the cells. The resultant signals were imaged using an SP8 confocal microscope.

For the PLA, a kit (Sigma-Aldrich) was used according to the manufacturer instructions. The same primary antibodies were used as in immunocytochemistry and PLA probes were added that recognize the goat and rabbit primary antibodies. The probes are tagged with complementary DNA oligos that hybridize if they come in close proximity. The hybridized DNA was then amplified and visualized with fluorescently labeled probes using the Duolink In Situ Red Starter Kit Goat/Rabbit (catalog #DUO92105, Sigma-Aldrich). The cells were then counterstained with the anti-goat Alexa Fluor 488 to visualize transfected cells (Alam, 2018). The resultant signals were imaged using an SP8 confocal microscope, and the images were analyzed in ImageJ by measuring the percentage area of the cells containing the PLA signal.

Phenotypic analysis of axon pathways.

To analyze the ocular motor system wiring and axon guidance, whole-mount double immunostaining of 3 dpf larvae of either sex was performed as previously described (Clark et al., 2013). The following antibodies were used: anti-acetylated tubulin (1:1000; catalog #5335S, Cell Signaling Technology; RRID:AB_10544694) and anti-myosin heavy chain (1:250; catalog #A4.1025, Developmental Studies Hybridoma Bank; RRID:AB_528356).

Using the microscopy techniques described below, we analyzed the presence of ectopic branches and defasciculation in each of the four branches of the oculomotor nerve (OMN), and the main bundle before the branching point, as well as in the trochlear and abducens nerves. We also measured the total length of each of the OMN branches from the branching point to the point where it innervates the muscles. In addition to this quantification of the phenotype of each individual branch and nerve, we quantified the defasciculation, defined as axons that left the main nerve bundle but ultimately followed the same path and innervated the correct target. Instances of ectopic branching were scored as the axons that left the main bundle and failed to innervate the correct target, either wandering or innervating a different muscle. We measured the severity of the phenotype in the ocular motor system as the percentage of defasciculated nerves/nerves presenting ectopic branches across the OMN main bundle before branching, the four OMN branches, and the abducens and trochlear nerves.

Microscopy and image analysis.

After dechorionation, larvae of either sex were embedded in 1% low-melting point agarose (Sigma-Aldrich) in Danieau's solution and mounted on a SuperGlue (Henkel)-lined glass slide. Imaging was performed using a multiphoton microscope A1R-MP+ (Nikon) system equipped with a Coherent Chameleon Vision II laser, tunable from 690 to 1050 nm coupled to an Eclipse Ni-E FN Upright Microscope. Samples were imaged under an excitation wavelength of 780 nm, with 40× [numerical aperture (NA), 0.8] water-immersion objectives. dpysl2 and stmn1/2 morpholino-injected larvae and rescue experiments were imaged using a custom-built two-photon microscope equipped with a mode-locked titanium sapphire laser (Chameleon, Coherent) tuned to 770 nm and an Olympus LUMPlanFI 403 water-immersion objective (NA, 0.8).

Fluorescence emission was captured by both the objective and a substage oil condenser (Olympus), filtered through GFP emission filters (catalog #HQ 535/50, Chroma Technology) before detection with photomultiplier tubes (Hamamatsu). Scanning and image acquisition were controlled under ScanImage version 3.6 software (Pologruto et al., 2003). Image sequences were typically acquired at 10 Hz (256 3 100 pixels/frame, 1 ms/line). When acquiring z-stacks, optical sections were spaced by 1 µm. Image processing was performed using Fiji (RRID:SCR_002285). Single neurons were extracted from z-stacks using the Simple Neurite Tracer Plugin (Longair et al., 2011).

Optokinetic response assays.

At 5 dpf, wild-type or morpholino-injected larvae of either sex were placed dorsal side up in a 35 mm Petri dish filled with 2% low-melting point agarose prepared in Danieau's solution, and their OKR was assayed as previously described (Neuhauss, 2003; Brockerhoff, 2006). For the OKR recordings, we used a custom setup with a projector aligned to a cold mirror (Thorlabs) that projected vertical stripes to a cylindrical screen with the fish positioned at the center. Underneath the screen, a 54 LED infrared lamp illuminates the fish through the dichroic mirror and an infrared filter above the screen (52 mm; catalog #R72, HOYA) allowed only the infrared light to pass through.

The resulting infrared light was captured by a camera (Thorlabs) positioned vertically on the top of the setup and connected to the computer where the eye movements of the zebrafish were recorded using custom-written software in LabView (RRID:SCR_014325) and then analyzed using custom software written in MATLAB (RRID:SCR_001622). The eye movements were recorded when following stripes moving in both clockwise and anticlockwise directions at different speeds and with different spatial resolution and contrast levels. The range of movement (from the resting position of the eye toward rostral or caudal positions in degrees), saccadic velocity (how fast each eye reset to the resting position) and the gain (eye velocity response under varying angular velocity of the stimulus) were measured individually for each eye and each direction of movement; adduction (toward the midline) and abduction (away from the midline) (Neuhauss, 2003; Huang et al., 2011). For the analysis of the OKR, we used the maximum eye range and saccadic velocity, and we measured the gain of each eye for each of five different speeds.

Experimental design and statistical analyses.

Analysis of the ocular motor system anatomy phenotypes was performed on GraphPad Prism (GraphPad Software). MO knockdown of chn1, dpysl2, or stmn1/2 were compared with standard MOs using the Mann–Whitney test. For rescue experiments, larvae injected with chn1 MO and dpysl2 or stmn1 RNA were compared both to larvae injected with chn1 and standard MO separately using the Mann–Whitney test. For OKR experiments, chn1, dpysl2, or stmn1/2 MO knockdown was compared with standard MO, eye range and saccadic velocity were analyzed using t test, and gain was analyzed using Kruskal–Wallis one-way ANOVA.

Results

Mutations in CHN1 encoding the α2-CHN protein cause the eye movement disorder DRS because of miswiring of the ocular motor system, as revealed by neuroimaging in patients (Demer et al., 2007b; Dumars et al., 2008). The signaling cascades to the cytoskeleton by which α2-CHN regulates axon guidance are largely unknown. We initially aimed to elucidate which proteins are involved downstream of α2-CHN in the development of the ocular motor system, and to identify whether the GOF mutations linked to DRS cause an alteration in the interactome of the protein, thus leading to the axonal defects observed in patients.

Mass spectrometry identification of α2-chimaerin interacting partners

To identify which proteins could participate in the α2-CHN signaling module, we designed a mass spectrometry screening of α2-CHN-interacting proteins using rat cortical neurons transfected with either the wild-type protein (YFP-α2-CHN-WT) or one of two mutant isoforms (YFP-α2-CHN-L20F or YFP-α2-CHN-G228S). These mutant GOF isoforms were chosen as they have been well characterized, and we previously showed that their expression in chick and zebrafish systems cause striking defects in axon guidance (Miyake et al., 2008; Ferrario et al., 2012; Clark et al., 2013). YFP was used as a negative control for binding. We used three replicates of each of the α2-CHN-transfected immunoprecipitated samples and 2 replicates of the YFP-transfected samples with mass spectrometry analysis of the immunoprecipitated extracts to identify which components of each sample could be effectors of α2-CHN.

The dataset obtained in this analysis contained 2155 proteins from 11 samples. By considering only proteins present in at least two-thirds of replicates with a high confidence of identification, we found 627 proteins across all four conditions. After removing those proteins present in the YFP-negative control, 165 proteins were found to coprecipitate in the samples transfected with YFP-α2-CHN isoforms (Fig. 1A). In this way, we identified 102 proteins that coprecipitated in the YFP-α2-CHN-WT samples, while 109 and 63 coprecipitated in the YFP-α2-CHN-G228S and YFP-α2-CHN-L20F samples, respectively (Fig. 1A). We identified 39 proteins with a specific interaction with α2-CHN-WT but not with the GOF mutants, while 12 proteins were found in the GOF but not the wild-type samples (Fig. 1A). These data suggest that DRS-related mutations in the α2-CHN molecule lead to changes in the repertoire of protein partners, which would be likely to affect downstream signaling.

Figure 1.

α2-Chimaerin coprecipitates with cytoskeletal regulatory proteins and mutations alter the pattern of these interactions. A, Venn diagram of the proteins identified from immunoprecipitation of the fusion proteins YFP-α2-CHN-WT (WT), YFP-α2-CHN-L20F (L20F), or YFP-α2-CHN-G228S (G228S) and subsequent mass spectrometry. Proteins shown were identified with high confidence in at least two of three samples of the triplicate. The size of each circle corresponds with the total number of proteins identified in each sample. B–D, String (https://string-db.org/) interactome analysis of the proteins included in A, showing the interactions described between the proteins present in the sample and based on data mining; the thickness of the connecting lines represents the strength of the supporting evidence for the interaction. Color coding as in E. E, Pathway ID and functional enrichment analysis of each protein sample, showing the false discovery rate for and the pathway ID of the three biological processes shown in the data.

A STRING interaction analysis and a GO pathway identification of each subset of proteins (Fig. 1B–D) show that among those coprecipitating in the YFP-α2-CHN-transfected cells, there is a clear enrichment of proteins involved in neuron projection development, cytoskeleton organization, and microtubule-based processes, with a false discovery rate <0.01 for all three categories in every sample (Fig. 1E). These data are consistent with some of the known roles of α2-CHN as a regulator of axon growth and a modulator of microtubule and cytoskeletal dynamics. Consistent with these findings, two of the proteins that coprecipitated with α2-CHN in these experiments were the microtubule regulators STMN1 and STMN2. The collapsin response mediator protein (CRMP2) was found in the YFP-α2-CHN-transfected samples, but also in small amounts in the YFP-only transfected samples. Previous mass spectrometry experiments from our group using tandem mass tag label quantification corroborated an increased presence of CRMP2 in those samples transfected with YFP-α2-CHN when compared with the levels of CRMP2 present in YFP-only transfected cells (data not shown). Since CRMP2 has been shown to directly bind to the SH2 domain and act downstream of α2-CHN in other processes (Brown et al., 2004; Ip et al., 2012), we decided to study its role as a downstream effector of α2-CHN in OMS development.

STMN1/2 interact with α2-chimaerin

Stathmins 1 and 2 belong to a family of phosphoproteins that bind tubulin dimers, preventing them from being incorporated into MTs. They are also believed to destabilize MTs (Chauvin and Sobel, 2015). STMN1 activity is negatively regulated by phosphorylation at several phosphorylation sites by specific kinases, including Cdk5 and JNK (Amayed et al., 2002). Stathmins play key roles in processes such as nerve cell differentiation, axonogenesis, axon growth, dendritic growth, and branch formation, and they are promising candidates to mediate α2-CHN signaling because of their modulation of microtubule dynamics.

Whereas interaction of CRMP2 and α2-CHN has been shown previously (Brown et al., 2004), this has not previously been tested for the stathmins. To verify the interaction of STMN1 and STMN2 with α2-CHN observed in the mass spectrometry, we performed an analysis by Western blot and immunoprecipitation. HEK293T cells were cotransfected with YFP-α2-CHN constructs and RFP-STMN1 or RFP-STMN2. The cells were lysed after 2 d in culture, and the proteins were immunoprecipitated as before. We could detect an interaction between both STMN1 and STMN2 and α2-CHN isoforms (Fig. 2A). Interestingly, these interactions were specific to certain α2-CHN isoforms. STMN1 did not show a detectable interaction with wild-type α2-CHN but interacted weakly with the L20F isoform and most strongly with the G228S isoform. By contrast, STMN2 interacted with all three α2-CHN isoforms, interacting weakly with wild-type α2-CHN and more strongly with both mutant isoforms. These data confirm the direct binding of two novel α2-CHN-interacting proteins identified from mass spectrometry data and suggest that α2-CHN wild-type and mutant isoforms show differential levels of binding with these partners.

Figure 2.

α2-Chimaerin interacts with STMN1/2. Input and coimmunoprecipitation of STMN1 and STMN2 in HEK293T cells cotransfected with YFP, or the fusion proteins YFP-α2-CHN-WT (WT), YFP-α2-CHN-L20F (L20F), or YFP-α2-CHN-G228S (G228S) with RFP-tagged STMN1 or STMN2. Immunoprecipitation was performed using anti-GFP antibodies. Membranes were blotted using anti-DsRed, anti-GFP, and anti-GAPDH antibodies. STMN1 coprecipitates with the G228S isoform of α2-CHN, while STMN2 coprecipitates with both isoforms. RFP-tagged STMN1 is ∼45 kDa (asterisk), and STMN2 is 48 kDa (dash). The sizes are indicated on the blot.

Proximity ligation assay between α2-CHN and CRMP2 and STMN1/2

To further confirm the interactions of α2-CHN with CRMP2 and STMN1/2 as observed by mass spectrometry analysis and immunoprecipitation, we used a Proximity Ligation Assay (PLA) to establish whether both interacting partners of α2-CHN colocalize in rat embryonic cortical neurons transfected with YFP-α2-CHN-WT. As shown in Figure 3, A and A′, both YFP-α2-CHN and endogenous CRMP2 are present along the neural projections and within the cytoplasm, confirming the expression of both proteins within the axon and the cell body. Using the same set of antibodies, we tested whether these proteins are not only expressed in the same regions, but if they colocalize using PLA, which generates a fluorescent signal only when both proteins are closer than 40 nm. Signal could not be detected in the negative control, in which the anti-CRMP2 antibody was absent to eliminate a possible cross-detection between secondary antibodies (Fig. 3B,B′). However, as can be observed by the strong signal in Figure 3, C and C′, this assay demonstrated the positive colocalization of α2-CHN with CRMP2. Experiments using a negative control (transfection of YFP alone) failed to show any signal (data not shown).

Figure 3.

α2-Chimaerin colocalizes with CRMP2 and stathmin 1/2 in rat cortical neurons. Immunocytochemistry and imaging of YFP-α2-CHN-transfected cells. A–C, Detection of CRMP2 in transfected YFP-positive cells. DAPI signal from cell body is shown in yellow, YFP is shown in cyan, and CRMP2 is shown in magenta. A, A′, Double immunocytochemistry detection of YFP-α2-CHN and CRMP2. B, B′, Negative control of PLA in the absence of CRMP2 primary antibody. C, C′, Proximity ligation assay of CRMP2 in a YFP-positive cell. D, D′, Similar panels for STMN1 immunocytochemistry with STMN1 in magenta. E, F′, Negative control (E, E′) and PLA (F, F′). G, G′, Similar panels for STMN2 immunocytochemistry with STMN2 magenta. H–I′. Negative control (H, H′) and PLA (I, I′). White arrowheads in C′, F′, and I′ denote colocalization of the proteins in cell bodies, and open arrowheads in the same panels denote colocalization in axons. Scale bars, 50 µm.

The experiment was repeated to detect the expression of both STMN1 and STMN2. Endogenous STMN1 in rat embryonic cortical neurons can be observed uniformly distributed throughout the cell (Fig. 3D,D′), whereas STMN2 is localized predominantly at the Golgi, within vesicles in the cytoplasm and in a punctate distribution along the neural projections (Fig. 3G,G′). As before, we were unable to detect a positive PLA signal in the absence of either anti-STMN1 or STMN2 antibody (Fig. 3E,E′,H,H′), while we detected a strong colocalization signal with α2-CHN, showing fluorescent puncta the cell soma and the neuronal projections, for both STMN1 and STMN2 (Fig. 3F,F′,I,I′).

These results suggest that CRMP2 and STMN1/2 colocalize to similar cellular regions with α2-CHN, permitting their interaction as signaling partners, in particular in growing axons.

dpysl2 and stathmins are expressed in the ocular motor system in zebrafish

The OMS is well conserved across species, including zebrafish, and is composed of six extraocular muscles (EOMs) and three cranial motor nerves (CNs). The neuronal nuclei of these nerves, those of the OMN (CN III), trochlear nerve (CN IV), and abducens nerve (CN VI), are located in the midbrain, rhombomere 1, and rhombomeres 5/6 of the hindbrain, respectively. The development of the OMS can be easily visualized in vivo using the Tg(Isl1BAC:Gal4ff;UAS:GFP) zebrafish line, in which cranial motor neurons express GFP, and the expression of novel genes implicated in OMS development can be characterized.

Following their identification as putative interacting partners of α2-CHN, we sought to determine the expression patterns of CRMP2, encoded by dpysl2, and stathmins 1 and 2, both of which have been duplicated within the zebrafish genome, and are encoded by stmn1a, stmn1b, stmn2a, and stmn2b. We performed an RNAscope in situ hybridization in Tg(Isl1BAC:Gal4ff;UAS:GFP) larval zebrafish at 48 hpf, followed by dorsal imaging of the midbrain/hindbrain region (Fig. 4B–D: CN III, top panel, CN IV, bottom panel), schematized in Figure 4A. A negative control probe was used to establish a background level of nonspecific, fluorescent signaling (Fig. 4B). Probes targeting dpysl2 demonstrated a substantial expression level throughout the nervous system, including the midbrain and hindbrain. Here we observed general expression in the neuroepithelium, including the CN III and CN IV nuclei, as shown by the GFP fluorescence (Fig. 4C). Analysis of the expression patterns of both paralogs of stmn1 and stmn2 revealed differences in each case. stmn1a was expressed at a very low level throughout the neuroepithelium, with very little signal present in either nucleus (Fig. 4D). In contrast, stmn1b was expressed at a much higher level, especially in the hindbrain and including the region of the trochlear nucleus (Fig. 4E). Both stmn2a and stmn2b were expressed at similar levels throughout the midbrain and hindbrain with stmn2a present within the CN III and CN IV nuclei (Fig. 4F), whereas stmn2b appeared to be more highly expressed in the CN IV nucleus (Fig. 4G). These data indicate that the temporal and spatial expression patterns of the putative interacting partners of α2-CHN highlighted by the mass spectrometry data support their potential role in the development of the ocular motor system. In the case of the oculomotor nucleus, the highest expression was seen for dpsyl2 and stmn2a.

Figure 4.

Expression of dpysl2 and stmn1/2 in the ocular motor system of the larval zebrafish. A, Schematic representation of the dorsal view of CN III (oculomotor) and CN IV (trochlear) nuclei at 48 hpf. Midline and midbrain–hindbrain boundary are shown, with anterior (A) and lateral (L) axes. B, Fluorescent signal corresponding to the negative control probe following RNAscope in situ hybridization in Tg(isl1BAC:Gal4;UAS:GFP) larvae. Anti-GFP immunohistochemistry (cyan) was used to counterstain and highlight the cranial motor neurons of the oculomotor nerve (CN III) and trochlear nerve (CN IV), shown by dotted lines. C–G, Fluorescent signal following RNAscope in situ hybridization is shown in magenta using probes against dpysl2 (C), stmn1a (D), stmn1b (E), stmn2a (F), and stmn2b (G). Scale bars, 50 µm.

Analysis of the normal anatomy of the ocular motor system in zebrafish

The OMS is additionally composed of the following six EOMs: the superior rectus (SR), medial rectus (MR), inferior rectus (IR), inferior oblique (IO), superior oblique (SO), and lateral rectus (LR). The oculomotor nerve (CN III) innervates the SR, MR, IR, and IO muscles through four branches, while the SO and LR muscles are innervated by the trochlear (CN IV) and abducens (CN VI) nerves, respectively (Fig. 5A). Our previous work has demonstrated that morpholino knockdown of α2-CHN during OMS development in the zebrafish leads to a range of phenotypes, including defasciculation and ectopic branching of the cranial nerves (Clark et al., 2013).

Figure 5.

α2-Chimaerin knockdown produces defects in the development of the OMS. A, Schematic representation of a lateral view of the ocular motor system in zebrafish, with anterior facing to the right. The dotted line indicates the edge of the brain. Oculomotor (III), trochlear (IV), and abducens (VI) nerves are shown in cyan, and extraocular muscles are shown in magenta. B, Schematic representation of the typical defects observed in OMS development as a result of manipulations presented here. Yellow arrows indicate ectopic branching, and white arrows indicate defasciculation. Ci, Neurite tracer extracted image following whole mount staining of the OMS of a 3 dpf wild-type zebrafish larva using anti-tubulin (cyan) and anti-myosin heavy chain (magenta). Scale bars, 100 µm. Cii, Ciii, Insets, Representative defasciculation and ectopic branching events at a higher magnification (Cii shows oculomotor main branch; Ciii shows abducens). Dii, Diii, Eii, Eiii, Whole mount staining of 3 dpf zebrafish larvae injected with 3 ng of Standard MO (Dii, Diii, Eii, Eiii) and 3 ng of chn1 MO (Eii, Eiii). In each case, Dii, Diii, Eii, and Eiii are as in C. F, Quantification of the average defasciculation and ectopic branching across the OMS, shown as a percentage. G, Quantification of the defasciculation events (percentage) analyzed in the main bundle of the OMN, and each of the four branches of the OMN, trochlear nerve (Tro), and abducens nerve (Abd). H, Quantification of the ectopic branching events (percentage) analyzed in each branch. I, Western blot detection of α2-chn in embryos injected with 3 ng of Standard MO or chn1 MO. A 65% reduction in the expression levels of α2-chn in 3 dpf zebrafish larvae (n = 3 Western blots, n = 20 injected embryos each). For phenotypic analysis: noninjected, n = 25 embryos; standard MO, n = 17 embryos; chn1 MO, n = 23 embryos. Mann–Whitney was used test for comparison of phenotypes: **p < 0.005, ***p < 0.0005. Scale bars: Ci, Di, Ei, 100 µm; Cii, Ciii, Dii, Diii, Eii, Eiii, 20 µm.

We therefore sought to analyze in more detail the variability patterns of axon growth in wild-type embryos to act as a baseline for manipulations. To visualize the developing OMS in vivo, 3 dpf wild-type zebrafish larvae were double immunostained with antibodies against tubulin and myosin heavy chain to identify the cranial nerves and EOMs, respectively. The larvae were imaged from the lateral view, and the ocular motor nerve trajectories were extracted as described previously (Clark et al., 2013). We then quantitated some of the variability in axon pathways, defining defasciculation as axons that leave the main nerve bundle but follow the same path and innervate the correct muscle target (Fig. 5B, white arrows). Ectopic branching was defined as axons that left the main bundle and failed to innervate the correct target (Fig. 5B, yellow arrows).

Whereas the general structure of axon pathways remained conserved, we observed some variations in the pattern of innervation in the larvae (Fig. 5C), which affected in particular the main OMN nerve bundle where it split into branches to the SR, IR, and MR muscles (Fig. 5Cii) and the behavior of specific nerve branches (Fig. 5Ciii). We observed occasional ectopic branches targeting the EOMs or defasciculation events in which one of the branches split into two or more bundles before reaching its target muscle. Given this variability, we decided to analyze and quantify the range of phenotypes present in normal larvae before examining defects produced by experimental intervention. We performed a quantification of the defasciculation and ectopic branching prevalence in the OMS as a whole, taking into account the main bundle of the OMN, all four branches to muscles, the trochlear nerve, and the abducens nerve, and found that defasciculation had an average frequency of 28%, while ectopic branching occurred in 6% of all the branches/nerves analyzed (Fig. 5F). A detailed analysis of these results according to individual nerve branch is shown in Table 1. These results demonstrate that a small level of intrinsic variability is present in noninjected larvae. Slight defasciculation defects were present in at least 10% of the cases across all branches and were most frequent in the IO branch (63%; Fig. 5G). The presence of ectopic branches was quite rare, being the most frequent in the IR (17%; Fig. 5H).

Table 1.

Quantification of observed neuroanatomical phenotypes in the ocular motor system following morpholino knockdown

	Oculomotor nerve										Trochlear nerve		Abducens nerve		Total		N
	Main bundle		Superior rectus		Inferior rectus		Medial rectus		Inferior oblique
	%	p value	%	p value	%	p value	%	p value	%	p value	%	p value	%	p value	%	p value
Defasciculation
Standard MO	41%		47%		18%		12%		35%		6%		18%		25%		17
Noninjected	38%	1.0000	29%	0.3281	25%	0.7113	13%	1.0000	63%	0.1180	16%	0.6052	10%	0.6396	28%	1.0000	25
chn1 MO	58%	0.3580	42%	1.0000	52%	0.0504	44%	0.0414	77%	0.0105	23%	0.2056	32%	0.4644	48%	0.0034	23
dpysl2 MO	84%	0.0070	70%	0.1991	46%	0.0958	64%	0.0025	72%	0.0274	7%	1.000	20%	1.000	55%	0.0009	20
stmn1a MO	91%	0.0161	82%	0.1150	45%	0.1998	64%	0.0104	89%	0.0145	36%	0.0618	18%	1.000	60%	<0.0001	11
stmn1a + 1b MO	74%	0.0895	79%	0.0819	53%	0.0411	63%	0.0022	53%	0.3351	56%	0.0027	74%	0.0011	64%	<0.0001	19
stmn2a MO	67%	0.2635	100%	0.0033	83%	0.0007	67%	0.0045	90%	0.0140	33%	0.1296	75%	0.0032	73%	<0.0001	12
stmn2a + 2b MO	100%	0.0008	89%	0.0116	61%	0.0189	53%	0.0275	89%	0.0054	53%	0.0024	60%	0.0324	73%	<0.0001	15
stmn1a + 1b + 2a + 2b MO	88%	0.0075	79%	0.1014	57%	0.0411	43%	0.0365	78%	0.0166	50%	0.0072	54%	0.0176	64%	<0.0001	24
Ectopic branching
Standard MO	12%		0%		0%		12%		0%		0%		0%		3%		17
Noninjected	7%	1.0000	0%	1.0000	17%	0.1284	14%	1.0000	4%	1.0000	0%		0%		6%	0.2660	25
chn1 MO	65%	0.0004	38%	0.0072	28%	0.0300	24%	0.4388	4%	1.0000	0%		0%		24%	0.0002	23
dpysl2 MO	24%	0.4388	30%	0.0295	29%	0.0295	27%	0.4260	88%	0.0002	0%		0%		31%	<0.0001	20
stmn1a MO	36%	0.1741	36%	0.0161	0%	1.0000	0%	0.5053	22%	0.1108	0%		0%		13%	0.0047	11
stmn1a + 1b MO	58%	0.0058	5%	1.0000	11%	0.4873	5%	0.5929	16%	0.2310	0%		0%		14%	0.0008	19
stmn2a MO	17%	1.0000	67%	0.0001	8%	0.4138	17%	1.0000	20%	0.1282	0%		0%		18%	<0.0001	12
stmn2a + 2b MO	83%	<0.0001	17%	0.2286	17%	0.2310	0%	0.2286	22%	0.1062	0%		0%		21%	<0.0001	15
stmn1a + 1b + 2a + 2b MO	50%	0.0140	38%	0.0094	13%	0.4873	9%	0.5843	30%	0.0473	0%		0%		20%	<0.0001	24

Quantification is given for the total ocular motor system (averaged across scores for each component) as well as the main bundle and all four branches of the oculomotor nerve, the trochlear nerve, and the abducens nerve individually. Each of the chn1, dpysl2, stmn1, and stmn2 morpholino injections was compared with the standard MO (shown in italic) using the Mann–Whitney test. Exact p values for each comparison are given. P values that are significant are denoted in italics, and higher levels of significance are shown in bold.

As a control, we injected a nontargeting MO (Standard MO) to establish whether any alteration in the pattern of innervation could be observed. We injected one-cell stage zebrafish embryos with 3 ng of Standard MO and studied the anatomy as before at 3 dpf (Fig. 5D). In each case, the levels of defasciculation and ectopic branching observed were very similar to those previously observed in the non-injected larvae and neither manipulation affected the length of the OMN branches (data not shown). The incidences of defasciculation and ectopic branching in the OMS as a whole were 25% and 3%, respectively (Fig. 5F), values not significantly different from those observed in the non-injected larvae (p > 0.05 in each case). Defasciculation was most frequently observed in the SR (47%; Fig. 5G), and, while ectopic branching was again rare, it was most often present in the MR (12%; Fig. 5H). Neither non-injected nor standard MO-injected embryos showed significant defasciculation or ectopic branching of the trochlear or abducens nerves (for trochlear, compare Fig. 5Ci, Di, area outlined by dotted line in SO muscle; Fig. 5Ciii, Diii, abducens). The similarities in the overall effects of this control injection when compared with the non-injected larvae suggest that the standard morpholino did not negatively impact normal OMS development.

chn1 knockdown increases defasciculation and ectopic branching in the OMS

To analyze the effects of manipulating CRMP2 or stathmin signaling in the development of the OMS, we initially used morpholino knock-down KD of chn1 to determine a baseline for alterations in the ocular motor system. This approach allowed us to compare the effects of knocking down putative downstream effectors of α2-CHN on the development of the ocular motor system using the standard morpholino as a negative control.

First, to establish the efficiency of the chn1 knockdown induced by the MO, we analyzed by Western blot the expression levels of α2-CHN using protein extracts of 3 dpf zebrafish injected with 3 ng of either the standard or chn1 MO. The results confirmed the effectiveness of the MO knockdown, reducing α2-CHN expression levels by 65% (Fig. 5I). We therefore proceeded to study the subsequent effect on the development of the ocular motor system, comparing the frequency of defects observed with the standard MO control.

In chn1 MO-injected larvae, the axons of the ocular motor system projected toward their muscle targets but showed a significant increase in defasciculation when averaged across all nerves and their branches (48%; Fig. 5E,F, Table 1). However, the trochlear and abducens nerves did not a show a significant increase, and the defects were focused on the oculomotor nerve. The main bundle of the OMN presented a defasciculated phenotype compared with controls (58%; Fig. 5Eii,G). The branch to the IO muscle showed the highest degree of defasciculation (77%; Table 1), which was significantly increased compared with the control. The chn1 KD also produced a significant increase in the incidence of ectopic branches overall in the OMS (24%; Fig. 5Eiii,F). This phenomenon was frequent in the OMN main bundle and the SR branch (65% and 38%, respectively; Fig. 5H). We observed some ectopic branches that did not appear to target any muscle. For example, in some cases sprouts from the OMN main bundle projected dorsally toward the LR muscle (Fig. 5Eii, yellow arrows) reminiscent of phenotypes seen in DRS in humans. The trochlear and abducens nerves (to the SO and the LR, respectively), did not show increased ectopic branching, and although defasciculation was slightly increased in the trochlear nerve following chn1 KD, this was not significant (Fig. 5Ei). In summary, for the oculomotor nerve, there was a strong effect of chn1 knockdown in increasing nerve defasciculation, especially of the IO branch, and ectopic branching of the OMN main bundle and the SR branch.

CRMP2 and stmn1/2 are required for normal ocular motor system development

To evaluate whether CRMP2 and STMN1/2 are components of the α2-CHN signaling pathway, we used morpholino-mediated KD and anatomic analysis as above (Fig. 6A). Upon knockdown of CRMP2, encoded by dpysl2, the overall level of defasciculation was similar to that with chn1 MO and was significantly higher than the control (55%; Fig. 6B,E). We detected a notable increase in defasciculation across all branches of the oculomotor nerve, particularly in the main bundle (84%) and in the MR and IO branches (64% and 72%) relative to controls (Fig. 6F). Additionally, the KD of dpysl2 produced a statistically significant increase in the frequency of ectopic branches (31%; Fig. 6E). These data show that, overall, both chn1 and dpsyl2 KDs produced an increase in defasciculation and ectopic branching in the OMS (compare Figs. 5F, 6E), with the IO branch (the longest and most distal of the branches) showing increased defasciculation for both manipulations. However, there were some specific differences. For ectopic branching, chn1 KD had the greatest effect on the OMN main bundle and SR branch, whereas dpsyl2 KD affected the IO branch most. Together, these results suggest that both α2-CHN and CRMP2 (encoded by dpysl2) participate in oculomotor development in the zebrafish and may form part of a signaling cascade for axon guidance.

Figure 6.

Knockdown of dpysl2 and stmn1/2 leads to defects in OMS development. Neurite tracer-extracted image following whole-mount staining of the OMS of 3 dpf wild-type zebrafish larvae using anti-tubulin (cyan) and anti-myosin heavy chain (magenta). Abbreviations are as in Figure 5. Ai–Diii, Representative images of the OMS of larvae following injection of standard (A), dpysl2 (B), stmn1a + stmn1b (C), and stmn2a + stmn2b morpholinos (D). Yellow arrows indicate ectopic branching, and white arrows indicate defasciculation. E, Quantification of the average defasciculation and ectopic branching of the OMS shown as a percentage. F, Quantification of the defasciculation events (percentage) analyzed in the main bundle of the OMN, each of the four branches of the OMN, and trochlear and abducens nerves. G, Quantification of the ectopic branching events (percentage) analyzed in each branch. Standard MO, n = 17; dpysl2 MO, n = 20; stmn1a + stmn1b, n = 19; stmn2a + stmn2b, n = 15. Mann–Whitney test for comparison of phenotypes; **p < 0,005, ***p < 0.0005. Scale bars: Ai, Bi, Ci, Di, 100 µm; Aii, Aiii, Bii, Biii, Cii, Ciii, Dii, Diii, 20 µm.

In zebrafish, stmn1 and stmn2 are duplicated and all four stathmin isoforms are expressed in the CNS (Thisse et al., 2004; Burzynski et al., 2009). We have additionally demonstrated that they are expressed at varying levels in the nuclei of the OMS especially stmn2a (Fig. 4). To examine whether STMN1/2 are involved in the development of the ocular motor system, we used translation-blocking MOs to target stmn1a, stmn1b, stmn2a, or stmn2b, or various combinations. Effects on the anatomy of the oculomotor system were analyzed as above.

Larvae injected with stmn1a MO showed a general increase in defasciculation, which was statistically significant when compared with the standard morpholino (60%; Table 1), coupled with a modest increase in ectopic branching (13%). The defasciculation particularly affected the main bundle and the IO/MR branch of the OMN (91%, 89%, and 64% respectively; all showed a significant increase; Table 1). Ectopic branching affected the SR (36%) and main bundle (36%) most. Neither the abducens nor the trochlear nerves showed significant ectopic branching phenotypes (Table 1). Similar results were observed in embryos injected with stmn2a MO, with an increase in defasciculation to 73% overall (Table 1). All four branches of the OMN were affected, with a frequency between 67% (MR) and 100% (SR), whereas the frequency of defasciculation increased significantly for the abducens nerve (75%; Table 1). Therefore, stmn2a knockdown had a larger effect on defasciculation overall than stmn1a knockdown. stmn2a knockdown also produced an increase in ectopic branching to 18% overall, with the SR branch of the OMN most affected (67%), as for stmn1a MO (Table 1) but with an enhanced effect. Again, the increase in ectopic branching did not affect the trochlear and abducens nerves. Overall, the stronger effects of stmn2a knockdown in the oculomotor nerve compared with that of stmn1a are consistent with the higher expression of stmn2a observed in the oculomotor nucleus (Fig. 4F).

We next sought to determine the impact of the combined knockdown of both stmn1 or stmn2 paralogs. The injection of morpholinos targeting stmn1a and stmn1b together increased overall defasciculation to 64%, and ectopic branching to 14% (Fig. 6C,E, Table 1). This corresponds with a moderate increase in the effect on defasciculation when compared with the single-MO injections. Notably, the double knockdown led to an increase in ectopic branching of the main bundle of the OMN to 58% (Fig. 6G). There was also a significant increase in defasciculation of both the abducens and the trochlear nerves (74% and 56%; Fig. 6F), which was not seen for stmn1a KD alone, and which may correspond with the high level of stmn1b expression observed in the trochlear nucleus (Fig. 4E).

Similarly, KD of stmn2a and stmn2b led to high levels of defasciculation and ectopic branching, both of which were significantly different from controls (73% and 21%, respectively; Fig. 6D,E). Defasciculation of the main bundle of the OMN could be observed in 100% of samples analyzed, and was also present in the SR and IO branches in 89% of cases as well as at significantly increased levels across all other branches of the OMN (Fig. 6F, Table 1). Increased defasciculation was observed in both the trochlear and abducens nerves. Ectopic branching affected the main bundle of the OMN, which in some cases projected toward the LR muscle (83%), mirroring the DRS-like phenotype, which was also seen on knockdown of chn1 and dpysl2 (Fig. 6G). Together, the combined knockdown of stmn1 or stmn2 paralog pairs significantly elevated the frequency of phenotypes across nerve branches relative to the single knockdowns and affected all three ocular motor nerves. stmn2a + stmn2b KD in particular showed the strongest effect.

Finally, we studied the effect of the combined injection of MOs directed against stmn1a, stmn1b, stmn2a, and stmn2b. This resulted in a similar overall level of defasciculation and ectopic branching compared with double-knockdown approaches (64% and 20%, respectively; Table 1). The pattern of defects observed across the nerves and branches analyzed was not noticeably different from that elicited by stmn2a/2b KD, which may reflect a more important role for stmn2a/2b in OMS development, or a partial redundancy of stmn1 and stmn2 in the system.

In summary, the MO-mediated knockdown of chn1 mainly resulted in an increase in defasciculation and ectopic branches, some of which overshot toward the wrong targets. Similar defects, with higher frequency, could be observed on knockdown of dpysl2, with the strongest effect overall on ectopic branching. MO-mediated knockdown of stmn1a and/or stmn1b produced similar defasciculation defects but with a moderate effect on the appearance of ectopic branches. The largest effects observed occurred as a result of stmn2a and stmn2b knockdown, which did not increase further when stmn1a/1b/stmn2a/2b were simultaneously knocked down. These results indicate that these proteins form part of an α2-CHN signaling module and play important roles in normal OMS development. Overall, it appears than stmn2a/2b play an especially important role in OMS development. There may be partial redundancy between the role of these stathmins in the ocular motor system.

EOMs and CN nuclei develop normally in chn1, dpysl2, and stmn1/2 knock-down conditions

Axon guidance is a complex process requiring the integration of both attractive and repulsive external cues. The defects observed in the larval zebrafish OMS following the disruption of α2-CHN signaling, such as the presence of ectopic branching, indicate a perturbation of axon guidance events. Previous work by our group and others has shown that Semaphorin and CXCR4/CXCL12 are important guidance molecules that are expressed by the extraocular muscles (Lerner et al., 2010; Ferrario et al., 2012; Whitman et al., 2018). To determine whether the observed OMS defects could be because of an altered extracellular environment resulting from disruptions to extraocular muscle development, we immunostained 3 dpf larvae for a muscle-specific antigen. The normal muscle organization, schematized in Figure 7A, could be observed in both the noninjected and standard MO-injected controls, as well as in larvae injected with chn1, dpysl2, or stmn1/2 MOs (Fig. 7B–F). These results suggest that the extracellular environment of the OMS is unaffected in morphants, and that the observed axon guidance defects occur because of perturbations in the integration of external cues.

Figure 7.

Extraocular muscles and cranial nuclei of the OMS develop normally in morphants. A, Schematic representation of the position and organization of the six extraocular muscles (magenta). Abbreviations are as in Figure 5. B–F, Representative images following whole-mount immunostaining of the EOMs, using an anti-myosin heavy chain antibody (magenta) at 72 hpf: wild-type, noninjected larvae (B), standard (C), chn1 (D), dpysl2 (E), and stmn1a + stmn1b + stmn2a + stmn2b (F) morpholino-injected larvae. G, Schematic representation of the position and organization of the nuclei of the oculomotor (CN III) and trochlear (CN IV) cranial nerves of the OMS as visualized in Tg(isl1BAC:Gal4;UAS:GFP) larvae. H–L, Representative dorsal view images of the GFP signal at 3 dpf shown in grayscale; dotted lines highlight CN III and CN IV: wild-type, noninjected larvae (H); standard (I), chn1 (J), dpysl2 (K), and stmn1a + stmn1b + stmn2a + stmn2b (L) morpholino-injected larvae. Although there is some variability in the level of GFP expression, the overall layout is unaffected. Scale bars, 100 µm.

Similarly, axon guidance defects in the ocular motor system could result from disruptions to the development of the neurons of the three cranial nerves. We repeated the MO injections in Tg(isl1BAC:Gal4;UAS:GFP) embryos in which the cranial motor neurons express GFP (Tessadori et al., 2012). These larvae were imaged dorsally at 3 dpf to visualize the oculomotor and trochlear nuclei (CN III and CN IV, respectively), schematized in Figure 7G. Both the noninjected and standard MO-injected larvae demonstrated a normal arrangement and number of neurons in CN III and CN IV (Fig. 7H,I). Larvae injected with chn1, dpysl2, or stmn1/2 MOs also showed normal development of the oculomotor and trochlear nuclei (Fig. 7J,K,L). Together, these results indicate that the cranial nerves and extraocular muscles of the OMS develop normally in morphants, and suggest that an α2-CHN signaling module is crucial for axon guidance of ocular motor neurons.

CRMP2 and STMN1/2 overexpression rescues chn1 knockdown phenotype

The similarities in phenotype observed after injection of morpholinos against chn1, dpysl2, stmn1, or stmn2 suggest an involvement of these proteins in the same pathway. To analyze whether this is the case in the OMS of zebrafish, we coinjected one-cell stage embryos with chn1 MO and mRNA encoding dpysl2 (Tanaka et al., 2012), STMN1, or CHN1 as a control. Neuroanatomical phenotypes were compared after whole-mount staining analysis of 3 dpf injected larvae, as previously with both standard and chn1 MO-injected larvae (Fig. 8A,B).

Figure 8.

Overexpression of CHN1, CRMP2, and STMN1 rescues chn1 knock-down phenotypes. Representative neurite tracer-extracted images following whole-mount staining of the OMS of 3 dpf wild-type zebrafish larvae using anti-tubulin (cyan) and anti-myosin heavy chain (magenta). A–C, E, G, Representative images of larvae following the injection of standard MO (A), chn1 MO (B), chn1 MO with CHN1 RNA (C), chn1 MO with dpysl2 RNA (E), and chn1 MO with YFP-STMN1 RNA (G). A–C, E, G, Insets in i show higher magnification of observed defasciculation (white arrows) of specific nerve branches in ii and iii. D, Quantification of the average defasciculation and ectopic branching across the OMS shown as a percentage, comparing larvae injected with standard MO with GFP RNA, chn1 MO with GFP mRNA, and chn1 MO with CHN1 RNA. Standard MO, n = 36; chn1 MO, n = 19; and chn1 MO with CHN1 RNA, n = 40. F, Quantification of the average defasciculation and ectopic branching across the OMS shown as a percentage, comparing standard MO, chn1 MO, and chn1 MO with dpysl2 RNA-injected larvae. Standard MO, n = 17; chn1 MO, n = 23; and chn1 MO with dpysl2 RNA, n = 13. H, Quantification of the average defasciculation and ectopic branching across the OMS shown as a percentage, comparing standard MO-injected embryos with embryos injected with GFP RNA, chn1 MO with GFP RNA, and chn1 MO with YFP-STMN1 RNA. Standard MO + GFP RNA, n = 35; chn1 MO + GFP RNA, n = 15; chn1 MO + STMN1 RNA, n = 17. Mann–Whitney test for comparison of phenotypes: *p < 0.05, **p < 0.005, ***p < 0.0005. Scale bars: Ai, Ci, Ei, Gi, 100 µm; Aii, Aiii, Cii, Ciii, Eii, Eiii, Gii, Giii, 20 µm.

We first compared embryos injected with standard MO and GFP mRNA with chn1 MO and GFP mRNA (with and without CHN1 RNA to effect a rescue). GFP RNA was coinjected as a control for the RNA rescue. Larvae coinjected with GFP mRNA and the standard MO presented similar overall levels of axon guidance defects compared with previous standard MO-injected larvae (20%; Fig. 8D, Table 2; p > 0.05). Coinjection of GFP RNA with chn1 morpholino resulted in significant increases in the overall frequency of defasciculation and ectopic branching compared with standard MO + GFP (Fig. 8D, Table 2; p < 0.001 and p < 0.0001, respectively). On the other hand, the effect of knockdown by chn1MO + GFP mRNA could be rescued by the injection of CHN1 RNA (Fig. 8C,D, Table 2; p < 0.01 and p < 0.0001), relative to chn1 MO + GFP for defasciculation and ectopic branching, respectively, and showed no significant difference from standard MO + GFP RNA (Table 2; p > 0.05 in each case).

Table 2.

Quantification of observed neuroanatomical phenotypes in the ocular motor system following morpholino knockdown and mRNA overexpression

Defasciculation				Ectopic branches
chn1 rescue				chn1 rescue
Standard MO + GFP RNA		20%	N = 36	Standard MO + GFP RNA		3%	N = 36
chn1 MO + GFP RNA		33%	N = 19	chn1 MO + GFP RNA		20%	N = 19
chn1 MO + chn1 RNA		22%	N = 40	chn1 MO + CHN1 RNA		3%	N = 40
			p value				p value
Standard MO + GFP RNA	vs	chn1 MO + GFP RNA	0.0004	Standard MO + GFP RNA	vs	chn1 MO + GFP RNA	<0.0001
Standard MO + GFP RNA	vs	chn1 MO + CHN1 RNA	0.3321	Standard MO + GFP RNA	vs	chn1 MO + CHN1 RNA	1.0000
chn1 MO + GFP RNA	vs	chn1 MO + CHN1 RNA	0.0035	chn1 MO + GFP RNA	vs	chn1 MO + CHN1 RNA	<0.0001
Standard MO		25%	N = 17	Standard MO		3%	N = 17
chn1 MO		48%	N = 23	chn1 MO		24%	N = 23
dpysl2 rescue				dpysl2 rescue
chn1 MO + dpysl2 RNA		27%	N = 13	chn1 MO + dpysl2 RNA		9%	N = 13
			p value				p value
Standard MO	vs	chn1 MO	0.0034	Standard MO	vs	chn1 MO	0.0002
Standard MO	vs	chn1 MO + dpysl2 RNA	0.5760	Standard MO	vs	chn1 MO + dpysl2 RNA	0.2170
chn1 MO	vs	chn1 MO + dpysl2 RNA	0.0101	chn1 MO	vs	chn1 MO + dpysl2 RNA	0.0331
stmn1 rescue				stmn1 rescue
Standard MO + GFP RNA		29%	N = 35	Standard MO + GFP RNA		4%	N = 35
chn1 MO + GFP RNA		49%	N = 15	chn1 MO + GFP RNA		17%	N = 15
chn1 MO + STMN1 RNA		35%	N = 17	chn1 MO + STMN1 RNA		8%	N = 17
			p value				p value
Standard MO + GFP RNA	vs	chn1 MO + GFP RNA	0.0003	Standard MO + GFP RNA	vs	chn1 MO + GFP RNA	0.0019
Standard MO + GFP RNA	vs	chn1 MO + STMN1 RNA	0.0982	Standard MO + GFP RNA	vs	chn1 MO + STMN1 RNA	0.4270
chn1 MO + GFP RNA	vs	chn1 MO + STMN1 RNA	0.0367	chn1 MO + GFP RNA	vs	chn1 MO + STMN1 RNA	0.0358

Quantification given for the whole ocular motor system, averaged across scores for each component. Coinjection of standard MO and GFP mRNA, chn1 MO and GFP mRNA, and chn1 MO and CHN1 mRNA are each compared to assess the success of the rescue. chn1 MO coinjected with dpysl2 mRNA was compared with standard MO or chn1 MO on their own. Coinjection of chn1 MO and STMN1 mRNA was compared with coinjections of standard MO and GFP mRNA, and chn1 MO and GFP mRNA. Comparisons were performed using Mann–Whitney test; p values are given. P values that are significant are denoted in italics, and higher levels of significance are shown in bold.

We next injected with dpysl2 mRNA as a control and found no phenotype and similar levels of defasciculation and ectopic branching relative to noninjected controls (data not shown). However, when dpysl2 RNA was coinjected with chn1 MO, there was reduced defasciculation and ectopic branching compared with phenotypes resulting from chn1 MO alone (Fig. 8E,F, Table 2). Defasciculation was reduced from 48% to 27% (p = 0.01) and ectopic branching from 24% to 9% (p < 0.05). The levels of defasciculation and ectopic branching observed after rescue were similar to those observed in larvae injected with the standard MO (p > 0.05 in each case). This reflects a dpysl2-mediated rescue of the phenotype resulting from the knockdown of chn1, indicating that the CRMP2 protein is able to compensate for decreased α2-chn levels and supporting the idea that both proteins are involved in the same pathway.

Similarly, we assessed the ability of STMN1 to rescue defects resulting from morpholino-mediated chn1 knockdown, by coinjecting mRNA for YFP-tagged STMN1 with chn1 MOs (Fig. 8G,H, Table 2) and comparing with the results of injection of chn1 MO + GFP RNA. Interestingly, coinjection of YFP-STMN1 with chn1 morpholino partially reduced these defects in OMS development. The overall frequency of defasciculation decreased from 49% to 35%, while ectopic branching was reduced from 17% to 8%, statistically significantly different from GFP coinjected with standard MO (p < 0.05 in each case). As observed with the controls, YFP-STMN1 expression alongside chn1 knockdown resulted in the persistence of some defects (e.g., defasciculation and ectopic branching in the main bundle of the OMN; Fig. 8G). However, these defects were not observed at levels that were significantly higher than in controls (p > 0.05 in each case). These data demonstrate that the overexpression of STMN1 protein is capable of rescuing defects resulting from the dysregulation of α2-CHN. Together, these results indicate that CRMP2 and STMN1 can compensate at least partially for the defects in the development of the OMS that occur due to the loss of α2-CHN signaling. In addition to the observation that each protein interacts directly with α2-CHN, our data suggest that CRMP2, STMN1, and STMN2 form part of a downstream signaling cascade and play crucial roles in normal OMS development.

chn1 knockdown perturbs the optokinetic reflex

To identify whether the perturbation of ocular motor anatomy described here would lead to defects in eye movement, we used the OKR assay to assess changes in horizontal eye movements. We quantitated the OKR using a video camera to film eye movements of immobilized 5 dpf zebrafish viewing projected stripes with varying parameters of speed, contrast, and thickness. We tracked the movement of each individual eye and analyzed the resulting traces using custom-built software to derive values for eye gain and eye range (Fig. 9A). The range of movement is indicated as nasal movement (adduction) for positive (clockwise) range of the left eye and negative (anticlockwise) range for the right eye. Temporal eye movement (abduction) is denoted as a negative range of movement in the left eye and a positive range of movement in the right eye (Fig. 9B). For the analysis of these tests, we identified the dominant eye (defined as the eye performing better with respect to range), and the weak eye and completed all statistical analyses to compare both groups of eyes. In larvae injected with chn1 MO, we observed an impairment in adduction of both the dominant and weak eyes when compared with the noninjected controls. In this case, we can see a larger range of values for chn1 MO-injected embryos, with much lower eye ranges in some larvae, especially for the weak eye (Fig. 9C).

Figure 9.

chn1 MO-injected zebrafish larvae show reduced optokinetic reflex. A, Schematic representation of the OKR setup. B, Diagram of zebrafish larval eye movements showing abduction and adduction movements for left and right eye. C, Effect of the chn1 MO injection on the abduction and adduction range of movements separated into dominant and weak eyes. In all cases, injected fish showed higher variability and a reduction in adduction. D, Similar results observed in the case of saccadic velocity. E, F, Gain curves for abduction (E) and adduction (F) movements. chn1 MO-injected fish show lower gain levels, especially when comparing weak eyes. Range and saccadic velocity compared with t test; gain was compared using two-factor ANOVA with replication: **p < 0.01. Noninjected, n = 13; chn1 MO, n = 13.

Similarly, the saccadic velocity was reduced in chn1 MO-injected larvae, which was significant for the adduction of the dominant eye (Fig. 9D). We also measured the OKR gain (sensitivity) in relation to the changing velocity of the stimulus for dominant/weak eyes and abduction/adduction responses (Fig. 9E,F). For both directions of movements, chn1 MO-injected larvae performed worse than noninjected controls, having a notable reduction in adduction and abduction for the weak eye, (p < 0.001 and p < 0.01, respectively). Whereas the dominant eyes of injected larvae performed similarly to noninjected larvae in adduction movements, they showed a gain reduction at higher speeds of abduction movements (p < 0.002; Fig. 9E). These results demonstrate that the neuroanatomical defects observed on morpholino-mediated knockdown of chn1 correlate with functional defects in larval eye movements.

dpysl2 and stmn1/2 knockdown results in defects in the optokinetic reflex

In a similar fashion, we next sought to establish whether zebrafish larvae injected with dpysl2 or stmn1/stmn2 MOs also showed an impairment of the OKR. Morpholino-mediated knockdown of dpysl2 reduced the eye range for adduction movements (Fig. 10A). The saccadic velocity was also reduced in these larvae, particularly affecting abduction of dominant and weak eyes (Fig. 10B). Finally, the OKR gain was also reduced in these MO-injected fish, again affecting both types of movements and the dominant and weak eyes (p < 0.05 in all cases; Fig. 10C,D). The effect of the dpysl2 knockdown on the OKR response was more severe than that caused by the chn1 MO, affecting both eyes instead of mainly one in the case of the chn1 knockdown.

Figure 10.

dpysl2 and stmn1/2 MO-injected zebrafish larvae showed reduced optokinetic reflex. A, Effect of the dpysl2 MO injection on the abduction and adduction range of movements separated into dominant and weak eyes. In all cases, injected fish showed higher variability and values in the lowest range. B, Similar results observed in the case of saccadic velocity. C, D, Gain curves for abduction (C) and adduction (D) movements. dpysl2 MO-injected fish show lower gain levels, for both dominant and weak eyes, and especially affecting adduction. Noninjected, n = 16; dpysl2 MO, n = 14. Values were compared using two-factor ANOVA with replication: **p < 0.01. E, Effect of the stmn1a + stmn1b + stmn2a + stmn2b MO injection on the abduction and adduction range separated into dominant and weak eyes. F, Saccadic velocity. G, H, Gain curves for abduction (G) and adduction (H) movements. Standard MO, n = 41; stmn1/2 MO, n = 39. Range and saccadic velocity were compared using the t test; gain compared using two-factor ANOVA analysis with replication: *p < 0.05, **p < 0.005, ***p < 0.0005. ns, Nonsignificant.

Finally, we analyzed the OKR of larvae injected with a combination of stmn1 and stmn2 morpholinos. This knockdown was performed with all four stmn MOs to reduce the possibility of genetic compensation. We did not observe a statistically significant difference in OMS defects between individual or combined stmn MOs. Neither the eye range nor saccadic velocity were statistically significantly different from the standard MO-injected controls in either the dominant or weak eyes (Fig. 10E,F). However, a small decrease in the gain in abduction movements could be observed in the dominant eye, while there was no difference in adduction movements (Fig. 10G,H).

These results show that the anatomic defects observed previously with the MO injections produce perturbations of horizontal eye movements for larvae with chn1 or dpysl2 knockdowns. The more severe defects observed with the dpysl2 MO correlated with a lower OKR fitness compared with chn1 KD. However, this correlation was not borne out for the strong axon guidance phenotypes seen in stmn KD embryos, where significant defects of the OKR were not observed (see Discussion). Impediments to eye movement might be caused by the decreased number of axons reaching their muscle targets because of being lost or incorrectly targeting other muscles (ectopic branching phenotype). In this case, the most significant defects for the alteration of the OKR response are those particularly affecting the MR and LR muscles, as the horizontal eye movements are controlled by contraction of the MR (adduction) and the LR (abduction).

Discussion

In this study, we have analyzed the interactome of α2-CHN and discovered novel interacting partners, including many proteins that are involved in axon outgrowth, cytoskeletal reorganization, and microtubule dynamics. Loss of function approaches in the zebrafish, with neuroanatomical analysis and behavioral studies, showed that CRMP2 and STMN1/2 play key roles in the development of the ocular motor system. Expression analysis confirmed the expression of these genes in the cranial motor neurons of the OMS. Knockdown of these components of this α2-CHN “signaling module” produced striking defects in ocular motor axon guidance, including an increase in defasciculation and ectopic branching; larvae with CRMP2 KD also showed an impairment of eye movements. These findings are consistent with previous data suggesting that the role of α2-CHN is mediated via regulation of the cytoskeleton (Clark et al., 2013). We now show that CRMP2 and STMN1/2 are likely to be key effectors of this regulation, as their overexpression could rescue axon guidance defects resulting from α2-CHN knockdown. These results suggest that CRMP2 and STMN1/2 are candidates to act downstream of α2-CHN in the etiology of cranial motor axon wiring disorders such as DRS (Demer et al., 2007b; Miyake et al., 2008; Clark et al., 2013; Nugent et al., 2017).

Significance of the phenotypes

The main phenotypes observed in response to downregulation of components of the α2-CHN signaling module consisted of axon-debundling defects (defasciculation) and ectopic branching. For chn1 KD, these defects are consistent with those described in our previous work using siRNAs in the chick, and morpholinos in the zebrafish (Ferrario et al., 2012; Clark et al., 2013), as well as those observed in chn1 knock-out mice (Nugent et al., 2017). These phenotypes contrast with those resulting from the introduction of α2-CHN gain-of-function mutations, which lead to stalling phenotypes of the OMN (Ferrario et al., 2012; Clark et al., 2013), of the abducens nerve, and in some cases misinnervation of the LR muscle by the OMN—a DRS-like phenotype (Nugent et al., 2017). Misdirected sprouts (ectopic branching) of the OMN toward the LR muscle were also seen to result from loss-of-function (LOF) of α2-CHN in a previous study (Ferrario et al., 2012) as well as in the current study in LOF of α2-CHN, CRMP2, and STMNs.

Phenotypes observed following CRMP2 and STMN KD were similar to those observed with chn1 MO, with differences in the pattern of defects whereby more severe phenotypes involved more nerve branches. For defasciculation defects, the severity followed a sequence of STMN2a/2b > CRMP2 > α2-CHN. The most distal branch of the oculomotor nerve to the IO muscle showed a significant increase in defasciculation in KDs for α2-CHN, CRMP2, and STMN2a/b. The oculomotor main bundle and MR branch manifest defects in CRMP2 and STMN2a/b KDs (though not in α2-CHN KD). For the STMN2a/2b KD, defects in the SR and IR branches were added to this repertoire. Therefore, it appears that there is a gradient of sensitivity to perturbation of the CHN signaling module, starting with the IO branch (which navigates the longest pathway), then MR, SR, and IR. Only STMN manipulations led to significant defects in the trochlear and abducens nerves, suggesting that STMNs may play a more important role in cytoskeletal regulation of the trochlear (and abducens) nerves than CRMP2. Enhanced severity of defects in STMN KDs, for example, may suggest that signaling pathways other than those mediated by α2-CHN might converge on these downstream targets. Other molecules that could play a role in axonal fasciculation could be Semaphorin3D signaling (Wolman et al., 2007), various cell adhesion molecules (Pollerberg et al., 2013), and Cadherin 7 and Cadherin 6b, which have also been linked to growth, branching, and guidance defects of cranial motor axons (Barnes et al., 2010).

Behavioral phenotypes

Using OKR analysis, we showed that the defects caused by chn1 and dpysl2 knockdown in the anatomy of the OMS of zebrafish larvae are linked to defective eye movements. These predominantly affected adduction movements (toward the midline) that would be mediated by the MR muscle. Eye range, saccadic velocity, and eye gain were affected to differing degrees. It is notable that the effects of KD manipulations were usually asymmetric, mirroring the effects of DRS in humans (Whitman and Engle, 2017). It is perhaps surprising that OKR defects were not recorded following KDs of stathmins, given that these anatomic defects were the most striking. One possibility to explain this is compensatory regulation of other family members: STMN3 and STMN4 have been shown in response to STMN KD, for example (Liedtke et al., 2002; Yoshie et al., 2006). Where defasciculation defects are linked to the perturbation of eye movements, maintenance of a tightly fasciculated nerve bundle may be crucial for the correct function of the axons and/or the formation of neuromuscular junctions (NMJs; Serra et al., 2012). It is notable that there are some features of extraocular muscles such as low fatigue resistance that may lead to even mild NMJ formation defects causing visual symptoms, whereas similar defects in trunk skeletal muscle may be asymptomatic (Serra et al., 2012). Recent studies in zebrafish expressing a mutant TDP-43 protein have shown that changes in fasciculation and branching of motor axons can cause alterations in the formation of NMJ and motor defects (Patten et al., 2017). Further studies will be needed to determine whether alterations of the α2-CHN signaling module are linked to defects in the formation of NMJs in the zebrafish ocular motor system.

Role of the α2-CHN signaling module

We propose that the α2-CHN signaling module is an important integrator of signaling, relaying axon guidance information to the cytoskeleton. DRS may arise because of the incorrect regulation of several signaling pathways that converge on the cytoskeleton (Fig. 11). The full spectrum of guidance cues that function in ocular motor system development has yet to be identified. We have previously demonstrated that Semaphorin A/Plexin A signaling, is required for OMS development and is mediated via α2-CHN in the chick (Ferrario et al., 2012). Work in mice has shown that α2-CHN also acts downstream of EphA4 in the OMS (Nugent et al., 2017), as well as during repulsion at the spinal cord midline (Katori et al., 2017).

Figure 11.

α2-CHN is an integrator of multiple signaling pathways. Proposed model of the role of α2-CHN as an integrator of extracellular signals. α2-CHN serves as a link between extracellular signaling cues acting in the OMS (based on the published literature) and the modulation of cytoskeletal dynamics. We propose that α2-CHN acts as a scaffold for the microtubule-interacting proteins CRMP2 and STMN1/2, and potentially for the engagement of their kinases, allowing for changes in phosphorylation (i.e., its function). This in turn would cause changes in microtubule stability and affect axonal growth.

The present study shows for the first time the involvement of CRMP2 in the development of the OMS. This is consistent with published data showing that CRMP2 is a component of the growth cone collapse/inhibition response triggered by Sema3A and Ephrin A5-EphA4 repulsion (Brown et al., 2004). Phenotypes resulting from the attenuation of CRMP2 activity may therefore correspond with a loss of axon repulsion from areas surrounding the ocular motor nerves, and be dependent on Sema3A or ephrins. A nonexclusive possibility is that α2-CHN/CRMP2 lies downstream of other ligand–receptor systems. In cortical neuronal migration in mice, α2-CHN acts via CRMP2 activity to regulate microtubule stability in a mechanism that is independent of EphA4 activation but downstream of neuropilin/TrkB signaling (Ip et al., 2012). We lack evidence of which guidance cues may regulate STMNs in the OMS, but STMN2 has been shown to lie downstream of EphBs in growth cones (Suh et al., 2004). Other cues that may lie upstream include the CXCR4/CXCL12 signaling pathway, which has been shown to be involved in OMS development in chicks and mice (Lerner et al., 2010; Whitman et al., 2018), as well as HGF (Lerner et al., 2010; Ferrario et al., 2012).

Regulation of CRMP2 and STMNs

We propose that α2-CHN regulates the activity of both CRMP2 and STMN1/2 (Fig. 11), with an alteration in the balance of activity leading to an alteration in microtubule dynamic instability, leading to axon growth defects. Mechanisms of regulation of microtubule dynamics include a hyperstabilization of microtubules by stathmin phosphorylation (Tararuk et al., 2006) or microtubule destabilization by CRMP2 phosphorylation (Fukata et al., 2002; Uchida et al., 2005). The ability of CRMP2 to regulate axon outgrowth is negatively regulated by phosphorylation, which reduces its affinity for binding tubulin dimers and destabilizes microtubules. Hyperphosphorylation of CRMP2 could be one of the mechanisms by which α2-CHN mutations lead to DRS. STMN1/2 bind to tubulin heterodimers producing MT stabilization; STMN is negatively regulated by phosphorylation, which reduces its ability to bind tubulin (Marklund et al., 1996). While STMN2 (SCG10) has been linked to neurite outgrowth in hippocampal primary cultured neurons (Morii et al., 2006), our work is the first to show a link between its role and the function of α2-CHN.

We speculate that as the phosphorylation of stathmin by kinases such as CDK5 and JNK is needed for the regulation of its function, α2-CHN might play a key role as a scaffolding protein bringing together CDK5 and JNK for the phosphorylation of stathmin and the regulation of its activity (Brown et al., 2004; Tararuk et al., 2006; Devaux et al., 2012; Yip et al., 2014). Given that α- and β-chimaerins play wide biological roles, including neuronal migration and axon and dendrite remodeling (Yang and Kazanietz, 2007), there is potential for these signaling pathways to be involved in a multiplicity of aspects of neural development.