Distinct roles for the cell adhesion molecule Contactin2 in the development and function of neural circuits in zebrafish

Abstract

Contactin2 (Cntn2)/Transient Axonal Glycoprotein 1 (Tag1), a neural cell adhesion molecule, has established roles in neuronal migration and axon fasciculation in chick and mouse. In zebrafish, antisense morpholino-based studies have indicated roles for cntn2 in the migration of facial branchiomotor (FBM) neurons, the guidance of the axons of the nucleus of the medial longitudinal fascicle (nucMLF), and the outgrowth of Rohon-Beard (RB) central axons. To study functions of Cntn2 in later stages of neuronal development, we generated cntn2 mutant zebrafish using CRISPR-Cas9. Using a null mutant allele, we detected genetic interactions between cntn2 and the planar cell polarity gene vangl2, as shown previously with cntn2 morphants, demonstrating a function for cntn2 during FBM neuron migration in a sensitized background of reduced planar cell polarity signaling. In addition, maternal-zygotic (MZ) cntn2 mutant larvae exhibited aberrant touch responses and swimming, suggestive of defects in sensorimotor circuits, consistent with studies in mice. However, the nucMLF axon convergence FBM neuron migration, and RB outgrowth defects seen in morphants were not seen in the mutants, and we show here that they are likely off-target effects of morpholinos. However, MLF axons exhibited local defasciculation in MZcntn2 mutants, consistent with a role for Cntn2 in axon fasciculation. These data demonstrate distinct roles for zebrafish cntn2 in neuronal migration and axon fasciculation, and in the function of sensorimotor circuits.

INTRODUCTION

Neuronal migration and axon guidance are critical developmental processes that are essential for establishing functional neural circuits underlying complex cognitive and motor functions. During development of the central nervous system, many newborn neurons travel long distances away from the germinal zone to their final locations, whereupon they extend axons and dendrites to assemble functional neural networks. Precise neuronal migration and axon guidance are dependent upon cell-cell and cell-substrate interactions (Chao et al., 2009; Marin et al., 2010), which are mediated by a number of membrane-associated molecules. Membrane proteins playing important roles in neurodevelopment include the cell adhesion molecules of the immunoglobulin superfamily (IgSF-CAMs), which are anchored to the membrane by a glycosylphosphatidyl inositol (GPI) linkage or a single transmembrane domain, have varying numbers of Ig-like extracellular domains, and are widely expressed during vertebrate neural development (Gennarini et al., 2016; Stoeckli, 2004; Stoeckli and Landmesser, 1995).

Contactin2 (Cntn2)/Transient Axonal Glycoprotein 1 (Tag1), a GPI-linked IgSF-CAM, has six Ig-like domains and four fibronectin domains (Furley et al., 1990), and can bind homophilically (Kunz et al., 1998) as well as heterophilically with other IgSF CAMs such as L1, NgCAM, and NrCAM (Buchstaller et al., 1996; Fitzli et al., 2000; Kuhn et al., 1991; Kunz et al., 1998; Pavlou et al., 2002; Suter et al., 1995). Cntn2 can also interact with contactin-associated protein-like 2 (Caspr2) at the juxtaparanode region in neurons (Tzimourakas et al., 2007). Cntn2 is broadly expressed by various cortical, cranial nerve, and spinal neurons and axons (Dodd et al., 1988; Yamamoto et al., 1986). It has been demonstrated to play a role in the fasciculation and guidance of dorsal root ganglion axons in mice (Kunz et al., 1998; Law et al., 2008) and spinal commissural axons in chick (Stoeckli and Landmesser, 1995), in the migration of mitral cells in the mouse olfactory bulb (Bastakis et al., 2015), and in sensorimotor gating and coordination in mice (Savvaki et al., 2008).

In zebrafish, cntn2 is expressed in specific neuronal types during development (Liu and Halloran, 2005; Sittaramane et al., 2009; Warren et al., 1999; Wolman et al., 2008) (Fig. S1). Morpholino (MO)-mediated knockdown studies indicated that cntn2 is required for the organization of the midbrain nucleus of the medial longitudinal fascicle (nucMLF) and the convergence of nucMLF axons into the fascicle (Wolman et al., 2008), the growth and fasciculation of Rohon-Beard (RB) central axons in the spinal cord (Liu and Halloran, 2005), and the caudal migration of facial branchiomotor (FBM) neurons in the hindbrain (Sittaramane et al., 2009). Interestingly, cntn2 genetically interacts with the planar cell polarity gene vangl2, which encodes a four-pass transmembrane protein (Jessen et al., 2002), during FBM neuron migration (Sittaramane et al., 2009), suggesting that Cntn2 and Vangl2 participate in a common mechanism to regulate this process. While these MO-based data are consistent with Cntn2-associated functions in mouse and chick, potential later functions for zebrafish cntn2 in sensorimotor circuits and behavior could not be examined because of reduced efficacy of MO-mediated knockdown after 1–2 days of development (Bill et al., 2009). Moreover, although appropriate controls (Eisen and Smith, 2008) were performed, a caveat of MO-mediated knockdown experiments is the possibility of off-target effects.

Therefore, to investigate roles for zebrafish cntn2 at later developmental stages and to address potential concerns with the MO knockdown studies, we generated loss-of-function mutations in cntn2 using CRISPR/Cas9. We generated two cntn2 mutant lines (cntn2^zou20 and cntn2^zou22) representing null alleles. Using the null mutants, we detected genetic interactions between cntn2 and vangl2, as shown previously with morphants, definitively demonstrating a function for cntn2 in FBM neuron migration. In addition, maternal-zygotic (MZ) cntn2 mutant larvae exhibited aberrant touch responses and swimming, suggestive of defects in sensorimotor circuits, consistent with studies in mice. However, the nucMLF axon convergence, FBM neuron migration defects, and RB outgrowth defects seen in cntn2 morphants were not seen in MZcntn2 mutants, and our data indicate that the morphant phenotypes are likely off-target effects of the morpholino. Interestingly, MLF axons exhibited local defasciculation in MZcntn2 mutants, consistent with a role for Cntn2 in axon fasciculation. These data demonstrate distinct developmental roles for zebrafish cntn2 in neuronal migration and axon fasciculation, and in the function of sensorimotor circuits.

MATERIALS AND METHODS

Animals

Zebrafish (Danio rerio) were maintained following standard protocols and University of Missouri ACUC guidelines as described previously (Sittaramane et al., 2013; Westerfield, 1995). Embryos were grown in E3 medium at 28.5 °C, staged by hours post fertilization (hpf) (Kimmel et al., 1995), and embryo age was verified at somitogenesis stages. Tg(isl1:gfp) (Higashijima et al., 2000) and Tg(pitx2c:gfp) (Wolman et al., 2008) fish were used to analyze FBM neuron migration, and nucMLF neurons and axons, respectively. The following two mutant lines were used: trilobite (tri^tc240a, (Hammerschmidt et al., 1996)) and sleepy (sly^ti263a, (Odenthal et al., 1996); kindly provided by Dr. Clarissa Henry, University of Maine, Orono) fish carrying mutations in the vangl2 and lamc1 genes, respectively. CRISPR-Cas9 cntn2 mutant lines were generated in the Tg(isl1:gfp) background.

Generation of cntn2 null alleles using CRISPR-Cas9

The sgRNA design tool at http://crispr.mit.edu was used to identify possible target sites. Two complementary oligonucleotides (5’ -TAGGCTAACAATGGTGCCGCAT-3’; 5’–AAACATGCGGCACCATTGTTAG-3’) corresponding to the sgRNA target site in exon 4 (Fig. 1A) were used. sgRNA and Cas9 RNA were synthesized as described (Jao et al., 2013). The following primers were used for genotyping: Fwd: 5’ -GACATATCATTTGCTGAAGAGTCG- 3’; Rev: 5’ -CTCGCGTTAGCAAGATACAAGTTA -3’. PCR products spanning the target site were digested overnight with BanI (NEB), and successful mutagenesis was indicated by the loss of the BanI site and the presence of an undigested PCR product (Fig. 1C). Injected embryos were grown to adulthood to generate founder fish, which were identified by outcrossing to wildtype fish and genotyping the F1 progeny for mutations in cntn2.

Figure 1. Generation and validation of CRISPR-generated cntn2 mutant.

(A) Genomic structure of cntn2 containing 19 exons, with the CRISPR target site (sgRNA) in the 4^th exon. The target site is highlighted in blue while the 7bp insertion and the 11bp deletion in cntn2^zou20 and cntn2^zou22 alleles are highlighted in green and indicated by dash marks, respectively. (B) Domain structure of Cntn2 containing six immunoglobulin (Ig) domains, four fibronectin (FN) domains, and a glycosylphosphatidylinositol (GPI)-anchor linked to the plasma membrane. Predicted amino acid sequences of wildtype (WT) cntn2, and cntn2^zou20 and cntn^zou22 alleles containing multiple stop codons (*). The highlighted AAs (blue) correspond to the CRISPR target site in the gene. (C) A PCR product (528bp, zou20; 510bp, zou22) spanning the target site digested with BanI differentiate between three genotypes: Wildtype (2 cut bands), heterozygote (1 uncut and 2 cut bands) and homozygote (1 uncut band). (D) Dorsal views of wildtype (WT) and mutant (zou20) hindbrains processed for cntn2 in situ hybridization (ISH) (upper panels) and anti-Cntn2 immunohistochemistry (IHC) (lower panels). Arrowheads indicate migrated FBM neurons and asterisks mark sensory ganglia. In cntn2 (zou20) mutants, cntn2 expression is greatly reduced, and Cntn2 protein is not detectable. Scale bar = 50 μm. (E) Western blot analysis of Cntn2 in cntn2^zou20 and cntn^zou22 embryos at 48 hpf. Cntn2 protein is not detectable in cntn2 mutants. Loading control is α-tubulin.

Morpholino injections

Antisense cntn2 morpholino (5’ -CCACACCCAGACCAGACACTTATTT- 3’;(Liu and Halloran, 2005; Sittaramane et al., 2009) and standard control morpholino (5’ -CCTCTTACCTCAGTTACAATTTATA- 3’) were obtained from Gene Tools (Corvallis, OR). The cntn2 MO used in this study was previously shown to knockdown Cntn2 protein levels (Liu and Halloran, 2005), which we validated (Fig. S2).

Immunohistochemistry and in situ hybridization

Immunohistochemistry and in situ hybridization were performed using standard protocols (Sittaramane et al., 2009; Vanderlaan et al., 2005). The following primary antibodies were used: rabbit anti-GFP (Invitrogen, 1:2000 dilution), zn5/8 (Developmental Studies Hybridoma Bank (DSHB), 1:10 dilution), rabbit anti-Cntn2 (kindly provided by Dr. Claudia Stuermer, University of Konstanz (Lang et al., 2001); 1:500 dilution), zn-12 (ZIRC, 1:250 dilution). The following secondary antibodies were used: chicken anti-rabbit Alexa Fluor 488 (Invitrogen; 1:500 dilution), goat anti-mouse Alexa Flour 568 (Life technologies, 1:500 dilution), donkey anti-rabbit Alexa Flour 568 (Invitrogen, 1:500 dilution). Images were taken with an Olympus BX60 or Leica TCP SP8 MP confocal microscope, processed for brightness and contrast, and composed into figures using Photoshop or Powerpoint software.

Western blot

Total protein was isolated from the heads of 2 dpf larvae as previously described (Hoffman et al., 2016). Western blotting was performed using standard protocols (Pan et al., 2014). The following primary antibodies were used: rabbit anti-Cntn2 ((Lang et al., 2001); 1:500 dilution), anti-acetylated α-tubulin (Sigma, 1:2000 dilution). The following secondary antibodies were used: goat anti-rabbit IgG, HRP conjugate (Santa Cruz Biotechnology, 1:5000 dilution) and HRP-goat anti-mouse IgG (H+L) (Invitrogen, 1:2500 dilution).

Quantification of FBM neurons in the hindbrain

Z-stack images (0.5 μm thickness, 150–200 slices) were taken of dorsally-mounted stained embryos with Leica TCP SP8 MP confocal microscope at 40X magnification. Leica Application Suite X (LAS X) software was used to generate a 3D image of selected stacks of confocal images. FBM neurons at all dorsoventral levels were counted in the 3D image by examining each level.

Genotyping immunostained embryos

The head fragments of antibody-stained embryos were mounted for quantification of FBM neuron migration, and the corresponding tail fragments were genotyped. Genomic region spanning the point mutation in the trilobite (tri^tc240a) allele (Jessen et al., 2002) was PCR amplified using following primers: Fwd: 5’-GCCTGGATGGTCACAGATTT-3’; Rev: 5’ CCGGAAGTTTATCAGTATGGGAAACAC-3’. PCR product was purified using a PCR purification kit (Qiagen) according to the manufacturer’s instructions and sequenced to detect the point mutation in each sample.

DiI injections and photoconversion of DiI fluorescence

Embryos were fixed overnight at 4°C in 4% paraformaldehyde, washed in PBS, and embedded in 0.6% agarose (in Danieu buffer) on a glass slide. A 2 mg/ml solution of the lipophilic dye, 1,1’-Dioctadecyl-3,3,3’,3’-Tetramethylindocarbocyanine Perchlorate (DiI, ThermoFisher Scientific) in dimethylformamide, was pressure injected into the nasal retina, targeting the retinal ganglion cell layer. Following overnight incubation at room temperature to allow for transport of DiI, the fluorescent signal was converted into a brown precipitate as previously described (Chandrasekhar et al., 1997). Deyolked embryos were incubated in 3,3’-diaminobenzidine (DAB, Sigma; 0.5 mg/ml in PO₄ buffer) for 15 min, mounted in DAB on a slide, and the fluorescence was photoconverted by exposing the labeled axons to epifluorescence for 20–30 min. Photoconverted embryos were post-fixed, and mounted in 70% glycerol.

Quantification of outgrowth of central axons of Rohon-Beard (RB) neurons

Rohon-Beard neurons were sparsely labeled by injecting 12.5 pg of ngn1:GFP-caax DNA (Andersen et al., 2011) and 25 pg of Tol2 transposase RNA per embryo at the one-cell stage. Injected embryos were fixed at 18 hpf for 4 hours at room temperature, and immunostained for GFP. RB neurons were binned into three groups based on the position of the cell body along the anterior-posterior (A-P) axis (1–5 somites, 6–10 somites, 11–16 somites). As a proxy to measuring the lengths of the central axons, the somite numbers corresponding to the tips of the anteriorly- and posteriorly-extending axons (ascending and descending axons, respectively) were recorded.

Analysis of larval behaviors

Larvae at 2 dpf were placed in petri dishes in E3 medium, and aligned to the center of field of view (20 mm in diameter). The tactile stimulus was a single touch to the head or trunk of the larva with a dissection needle. Each larva was tested only once. The larval responses were binned into three categories (Fig. 6): No response (no movement after touch), Weak Response (muted movement with larva remaining in the field of view), and Strong Response (rapid and vigorous movement with larva swimming out of the field of view).

Figure 6. cntn2 mutants exhibit defective touch responses.

(A, B) Distribution of touch-evoked escape responses of 2 dpf cntn2+/+ and MZcntn2−/− embryos following a head touch (A) or a trunk touch (B). The larval responses were binned into three categories: No response (no movement after touch), Weak Response (muted movement with larva remaining in the field of view), and Strong Response (rapid and vigorous movement with larva swimming out of the field of view). MZcntn2−/− mutants responded similarly to control cntn2+/+ larvae when touched on the head. However, they exhibited much weaker escape responses compared to cntn2+/+ larvae when touched in the trunk. Data pooled from 2 experiments (number of embryos in parenthesis). *Chi-square test at p<0.05; NS: not significant.

For analysis of swimming behavior, 7 dpf larvae were placed singly in individual wells of a 24-well plate. Swimming activity was monitored at room temperature (21–22°C) using a DanioVision system and EthoVision XT 8.0 locomotion tracking software (Noldus). Larvae were allowed to acclimate in the DanioVision system for 30 minutes in the dark. Swimming activity was then monitored for an hour each in light and dark phases. Distance moved and moving duration parameters were analyzed using EthoVision. Larvae were processed for genotyping the cntn2 locus, and tracking data were pooled for analysis according to genotype.

Statistical analysis

We performed the chi-square tests using online software (www.quantpsy.org/calc.htm) to detect any significant differences in the distribution of: 1) FBM neuron migration phenotypes among groups used for genetic interaction studies, 2) touch-evoked escape responses for cntn2+/+ and MZcntn2 mutant embryos following a head touch or a trunk touch, and 3) MLF defasciculation phenotypes in cntn2+/+ and MZcntn2 mutants. We performed unpaired t-tests using GraphPad Prism 7 software to detect any significant differences in: 1) distance moved and moving duration between cntn2+/− and MZcntn2 mutant embryos, 2) RGC axon fascicle thickness in cntn2+/+, cntn2+/− and MZcntn2 mutant embryos, and 3) RB central axon length in cntn2+/+ and MZcntn2 mutant embryos.

RESULTS

CRISPR-generated cntn2 alleles are null

To examine roles for cntn2 in larval zebrafish, and to test the validity of the cntn2 morphant phenotypes, we generated loss-of-function mutations in cntn2 using CRISPR-Cas9 technology (Jao et al., 2013). We injected a single guide RNA (sgRNA) targeting exon 4, which encodes the second Ig domain of Cntn2 (Fig. 1A, B). We identified three independent alleles (cntn2^zou20, cntn2^zou21, and cntn2^zou22), and established lines for the cntn2^zou20 and cntn2^zou22 alleles. The cntn2^zou20 allele has a 7 base pair (bp) insertion, and the cntn2^zou22 allele has an 11 bp deletion (Fig. 1A, C), which generate frameshifts and premature stop codons (Fig. 1B). Whole-mount in situ hybridization revealed that cntn2 expression is reduced in cntn2^zou20 (Fig. 1D) and cntn2^zou22 (data not shown) homozygous mutant embryos, consistent with nonsense-mediated decay (Chang et al., 2007). Importantly, a polyclonal antibody raised against the full-length Cntn2 protein (Lang et al., 2001) failed to detect any protein following immunohistochemistry (Fig. 1D) or western blotting (Fig. 1E) in both zygotic cntn2^zou20 and cntn2^zou22 homozygous mutants. These results strongly suggest that the CRISPR-generated cntn2 alleles are null. Interestingly, zygotic as well as maternal-zygotic (MZ) cntn2 homozygous mutants are viable and fertile, and appear morphologically normal.

FBM neurons migrate normally in cntn2 mutants

In zebrafish, facial branchiomotor (FBM) neurons are born in rhombomere 4 (r4) and migrate posteriorly into r6 and r7 (Chandrasekhar et al., 1997; Higashijima et al., 2000) (Fig. 2A). cntn2 is expressed in FBM neurons in zebrafish and mouse (Garel et al., 2000; Sittaramane et al., 2009; Warren et al., 1999) (Fig. S1B). In cntn2 morpholino (MO)-injected embryos (morphants), FBM neurons exhibit migration defects, frequently failing to migrate out of r4 (Sittaramane et al., 2009). In cntn2 morphants (4–6 ng MO dose/embryo), FBM neurons migrated normally in ~50% of the embryos, and either migrated poorly (Partial block) or largely failed to migrate (Severe block) in the remaining embryos (Fig. 2B, E). “Partial block” indicates reduced FBM neuron migration out of r4 on one or both sides, with FBM neurons found throughout the migratory pathway from r4 to r7, whereas “Severe block” indicates that a large majority of FBM neurons (estimated to be substantially greater than 50%) remained in r4 on both sides, in the characteristic pattern seen previously in genetic mutants like trilobite (Bingham et al., 2002). To our surprise, FBM neurons migrated normally into r6 and r7 in 100% of zygotic cntn2 homozygotes (Fig. 2C, E) of both alleles. The discrepancy in FBM migration phenotypes between cntn2 morphants and mutants may be explained by the presence of maternally contributed mRNA/protein in mutants but not in morphants. To test this possibility, we examined the migration of FBM neurons in maternal-zygotic (MZ) cntn2 mutants. FBM neurons migrated normally in 95% of MZcntn2−/− embryos (Fig. 2D, E), strongly suggesting that cntn2 does not have an essential function during FBM neuron migration, and may play a subtle role at best.

Figure 2. FBM neuron migration is affected in cntn2 morphants but not in cntn2 mutants.

Panels A-D show dorsal views of the hindbrain with anterior to the left. Tg(isl1:gfp) embryos were fixed at 48 hpf, and processed for immunohistochemistry with zn5 antibody (red) to label hindbrain commissural neurons and axons at rhombomere boundaries, and anti-GFP antibody (green) to label FBM neurons (arrowheads). (A) FBM neurons (arrowheads) migrate normally into r6 and r7 in an uninjected embryo. (B) FBM neurons largely fail to migrate out of r4 in a cntn2 MO-injected embryo. (C, D) FBM neurons migrate normally in zygotic mutant (Zcntn2−/−) (C), and maternal-zygotic mutant (MZcntn2−/−) (D) embryos. Scale bar in D, 50 μm for A-D. (E) Quantification of FBM neuron migration defects. Number in parenthesis denotes number of embryos. Data are from 3 to 4 experiments.

Genetic interaction between cntn2 and vangl2 for FBM neuron migration is preserved in mutants

Vangl2, a four-pass transmembrane protein and a core component of Wnt/planar cell polarity pathway (Gray et al., 2011), is expressed ubiquitously in the hindbrain during the period of FBM neuron migration (Sittaramane et al., 2013; Sittaramane et al., 2009). We showed previously that cntn2 and vangl2 genetically interact during FBM neuron migration since injection of a suboptimal dose of cntn2 MO enhances the weak FBM neuron migration phenotype of vangl2 heterozygotes (Sittaramane et al., 2009). To test whether cntn2 mutants support this conclusion, we examined the offspring of MZcntn2−/− homozygous females and vangl2−/− homozygous mutant males, which are 100% double heterozygous embryos. FBM neuron migration defects were more severe and frequent in cntn2+/−; vangl2+/− double heterozygotes (Fig. 3D, E), compared to cntn2+/− or vangl2+/− embryos alone (Fig. 3B, C, E). FBM neurons migrated normally in 100% of cntn2+/− embryos (Fig. 3B, E), and failed to migrate out of r4 in only 5% of vangl2+/− embryos (Fig. 3C, E). By contrast, FBM neurons failed to migrate out of r4 in 17% of cntn2+/−; vangl2+/− embryos (Fig. 3D, E). To further strengthen this subtle genetic interaction phenotype, we quantified FBM neuron migration in embryos obtained from a cross between vangl2+/− heterozygotes and cntn2−/− homozygous mutants, generating 50% cntn2+/−; vangl2+/+ and 50% double heterozygous (cntn2+/−; vangl2+/−) embryos. Since the genotypes of the embryos was unknown at the time of neuronal quantification, we selected ten immunostained embryos from each of three phenotypic groups corresponding to normal migration, partial and severe migration block (Fig. 3F–H), counted the number and examined the distribution of motor neurons in r4, r5, and (r6+r7) (Fig. 3I). Nearly 70% of FBM neurons failed to migrate out of r4 in “severe block” embryos, while 87% and 65% of neurons migrated out of r4 in “normal migration” and “partial block” embryos, respectively, providing high confidence in our phenotypic classes (Fig. 3E). Importantly, all embryos in the partial (n=10) and severe block (n=10) categories were subsequently genotyped as double heterozygotes, while most of the “normal migration” embryos (8/10) were genotyped as cntn2+/−; vangl2+/+. Taken together, the distribution of FBM migration phenotypes generated in two different genetic crosses supports the likelihood of genetic interactions between cntn2 and vangl2. This genetic requirement for cntn2, defined using a null allele, is revealed only in a sensitized genetic background of reduced planar cell polarity signaling, suggesting that cntn2 plays a definitive but minor role during FBM migration.

Figure 3. cntn2 interacts genetically with vangl2 but not with lamc1.

Panels A-D and F-H show dorsal views of the hindbrain with anterior to the left. Tg(isl1:gfp) embryos were fixed at 48 hpf, and processed for immunohistochemistry with zn5 antibody (red) to label hindbrain commissural neurons and axons at rhombomere boundaries, and anti-GFP antibody (green) to label FBM neurons (arrowheads). (A) FBM neurons migrate normally in a control embryo. (B) FBM neurons migrate normally in a cntn2 heterozygous (cntn2+/−) embryo. (C) FBM neurons migrate poorly in a vangl2 heterozygous (vangl2+/−) embryo, with neurons located along the entire migratory pathway from r4 to r6. (D) FBM neurons fail to migrate out of r4 in a cntn2; vangl2 double heterozygote (cntn2+/−; vangl2+/−). Scale bar in D, 50 μm for A-D. (E) Quantification of genetic interaction data. Number in parenthesis denotes number of embryos. **Chi-square test at p<0.001; NS: not significant. Data are from 2 to 4 experiments. (F-H) Offsprings of vangl2+/− heterozygous and cntn2−/− homozygous mutants exhibit normal, partial block, and severe block phenotypes for FBM neuron migration. Embryos exhibiting partial block (10/10) and severe block (10/10) were all identified as cntn2; vangl2 double heterozygote (cntn2+/−; vangl2+/−) and a majority of embryos (8/10) exhibiting normal migration were identified as cntn2+/−; vangl2+/+ by genotyping. (I) Quantification of non-migrated FBM neurons in r4, partially migrated FBM neurons in r5 and fully migrated FBM neurons in r6, and r7. Number in parenthesis denotes number of cells. Scale bar in F, 50 μm for F-H.

To further explore a role for cntn2 in FBM neuron migration, we examined another sensitized genetic background known to affect FBM neuron migration. We showed previously using suboptimal cntn2 MO doses and the bashful mutant that cntn2 genetically interacts with laminina1 (lama1), which encodes a broadly expressed Laminin subunit Laminina1 (Sittaramane et al., 2009). Since FBM neurons exhibit similar migration defects in bashful (lama1−/−) and sleepy (lamc1−/−) mutants (Chandrasekhar, 2004), which affect different Laminin subunits, we tested for genetic interactions between cntn2 and lamc1. We examined the offspring of MZcntn2−/− homozygous females and lamc1+/− heterozygous males, 50% of which are double heterozygous embryos. We found no embryos with severe FBM neuron migration defects, and the fraction of embryos with partial defects was comparable to that seen in cntn2+/− and lamc1+/− populations (data not shown). These data indicate that cntn2 and lamc1 do not interact genetically for FBM neuron migration, and suggest that some morphant-associated phenotypes may not reflect true functions of cntn2.

cntn2 mutants lack nucMLF axon convergence defects seen in morphants but exhibit MLF defasciculation

Given the discrepancy between cntn2 morphant and mutant FBM neuron migration phenotypes, we examined other phenotypes previously seen in morphants, such as the nucMLF axon convergence defects (Wolman et al., 2008). First, we confirmed the morphant phenotype by injecting cntn2 MO into Tg(pitx2c:gfp) embryos, which express GFP in nucMLF cell bodies and axons (Wolman et al., 2008). In control MO-injected embryos, the nucMLF was found as bilateral groups of tightly clustered cells, which extended tight axon fascicles posteriorly (Fig. 4A, E). However, in cntn2 MO-injected embryos, the nucMLF cells were loosely packed, their axons were defasciculated and they failed to converge normally (Fig. 4B, E). In contrast to the morphants, we did not find any nucMLF axon convergence defects in either zygotic (Zcntn2−/−) or maternal-zygotic (MZcntn2−/−) embryos (Fig. 4C–E), consistent with the difference between morphant- and mutant-associated FBM neuron phenotypes (Fig. 2).

Figure 4. cntn2 mutants show MLF defasciculation but lacks nucMLF defects seen in morphants.

Panels A-D show ventral views of the midbrain with anterior to the left. (A-D) Confocal projections of Tg(pitx2c:gfp) embryos labeled with anti-GFP antibody at 24 hpf. Panels F-I show ventral views of the midbrain and anterior hindbrain region with anterior to the left. (A) In a control MO-injected embryo, the nucMLF is found as bilateral groups of tightly clustered cells (delineated by dashed outline). Their axons form tight fascicles (arrow) immediately posterior to the neuron clusters. (B) In a cntn2 MO-injected embryo, the nucMLF neurons are loosely packed, and their axons are defasciculated. (C, D) The nucMLF neurons and axons converge normally in zygotic (Zcntn2−/−) and maternal-zygotic (MZcntn2−/−) mutants. (E) Quantification of nucMLF defects. Number in parenthesis denotes number of embryos. Data are from 2 to 4 experiments. (F-I) Zn-12 antibody labeling of the MLF axons in 24 hpf embryos. MLF axons in a cntn2+/+ form a tight fascicle (F); however, MLF axons are defasciculated (arrowheads) in MZcntn2−/− embryos (G-I). Black dotted line in F shows the cut-off point (for scoring) where the trigeminal sensory axons enter the hindbrain in r2. (J) Quantification of MLF defasciculation defects. Number in parenthesis denotes number of embryos. Data are from 2 experiments. Scale bar in D, 50 μm for A-D; Scale bar in F, 50 μm for F-I.

We also examined the ability of MLF axons to maintain fasciculation in the absence of Cntn2. MLF axons remained tightly fasciculated as they extended through the midbrain and anterior hindbrain (Fig. 4F). Interestingly, MZcntn2−/− embryos showed defasciculation of MLF axons in these regions (Fig. 4G–I). Because the MLF axons are often loosely fasciculated posterior to r2, we restricted our analysis to the area anterior to r2 (delineated by dotted line in Fig. 4F, corresponding to the hindbrain entry point for trigeminal sensory axons). We categorized embryos in blinded fashion into three groups based on the severity of the defasciculation phenotype: None (no defasciculation), Weak (small splits in the fascicle or very short stretches of defasciculated axons on one or both sides (Fig. 4G)), and Strong (extensive stretches of split fascicles or axon defasciculation on one or both sides (Fig. 4H, I)). About ~85% of MZcntn2 mutant embryos (n=32) exhibited strong defasciculation, and <5% showed no defects (Fig. 4J). In marked contrast, only ~28% of cntn2+/+ embryos (n=32) exhibited strong defasciculation, and ~50% showed no defasciculation (Fig. 4J). These data suggest strongly that Cntn2 is involved in maintaining adhesions between MLF axons in the zebrafish midbrain and anterior hindbrain.

FBM neuron and nucMLF defects in cntn2 morphants are likely off-target effects

The difference between cntn2 morphant and mutant phenotypes can be explained either by off-target effects of the morpholino (MO) (Eisen and Smith, 2008; Kok et al., 2015) or by genetic compensation from related genes in mutants but not in morphants (Rossi et al., 2015). To distinguish between these alternatives, we examined the effects of injecting cntn2 MO into cntn2+/+ and MZcntn2−/− embryos. If cntn2 MO injection causes defects in cntn2+/+ embryos but not in MZcntn2−/− embryos, then it would suggest genetic compensation (Fig. 5A). However, defects in both cntn2+/+ and MZcntn2−/− embryos following cntn2 MO injection would imply an off-target effect of the MO (Fig. 5A). As expected, FBM neuron migration and the development of nucMLF neurons and axons occurred normally in control MO-injected cntn2+/+ and MZcntn2−/− embryos (Fig. 5B–E, J, K). Consistent with previous studies (Sittaramane et al., 2009), migration of FBM neurons was strongly affected in cntn2+/+ embryos injected with cntn2 MO (Fig. 5F, J). Similarly, the organization and axonogenesis of nucMLF neurons was greatly perturbed in cntn2+/+ embryos injected with cntn2 MO (Fig. 5G, K), as shown previously (Wolman et al., 2008). Surprisingly, MZcntn2−/− embryos, which lack functional cntn2 mRNA and contain no detectable Cntn2 protein, also exhibited FBM neuron migration and nucMLF convergence phenotypes following cntn2 MO injection (Fig. 5H–K). These data demonstrate that the FBM and nucMLF neuronal defects seen in cntn2 morphants are likely due to off-target effects of the MO.

Figure 5. Some neuronal defects in cntn2 morphants are likely to be off-target effects.

(A) Experiments to distinguish between genetic compensation in MZcntn2−/− mutants, and off-target effects of the cntn2 MO. Normal development of FBM and nucMLF neuron in MZcntn2−/− mutants injected with cntn2 MO would suggest compensation. However, defective development of both cell types in these embryos would suggest off-target effects. (B, D, F, H) Dorsal views of the hindbrain with anterior to the left. Tg(isl1:gfp) embryos were fixed at 48 hpf, and processed for immunohistochemistry with zn5 antibody (red) to label hindbrain commissural neurons and axons at rhombomere boundaries, and anti-GFP antibody (green) to label FBM neurons (arrowheads). (B, D) FBM neurons migrate normally in control MO-injected cntn2+/+ (B) and in MZcntn2−/− (D) embryos. (F, H) Migration of FBM neurons is greatly reduced in cntn2 MO-injected cntn2+/+ (F) and MZcntn2−/− (H) embryos. (C, E, G, I) Ventral views of the midbrain, with anterior to the left, of Tg(pitx2c:gfp) embryos labeled with anti-GFP antibody. (C, E) Normal nucMLF development in control MO-injected cntn2+/+ (C) and MZcntn2−/− (E) embryos. (G, I) Defective nucMLF development in cntn2 MO-injected cntn2+/+ (G) and MZcntn2−/− (I) embryos. (J) Quantification of data presented in B, D, F and H. Number in parenthesis denotes number of embryos. (K) Quantification of data presented in C, E, G and I. Number in parenthesis denotes number of embryos. Scale bar in B, 50 μm for B, D, F, and H; Scale bar in F, 50 μm for C, E, G, and I.

cntn2 mutants exhibit a defective escape response and have swimming deficits

While some of the cntn2 morphant phenotypes appear to be off-target effects, the definitive phenotypes seen in MZcntn2 mutants (cntn2-vangl2 genetic interaction for FBM neuron migration and MLF defasciculation) motivated us to look for phenotypes in other cntn2-expressing cell types. In cntn2 morphant embryos, Rohon-Beard (RB) central axons grew more slowly and showed some defasciculation compared to control embryos (Liu and Halloran, 2005). To test whether the growth of RB central axons was affected in cntn2 mutants, we compared the lengths of RB central axons in equivalent-stage cntn2+/+ and MZcntn2−/− embryos. RB neurons were sparsely labeled using a ngn1:GFP-caax Tol2 construct (Andersen et al., 2011). There was no reduction in RB central axon lengths in MZcntn2 mutants (Fig. S3), suggesting that the loss of Cntn2 does not affect RB central axon growth rates as seen in morphants. In addition, we did not observe obvious defasciculation of RB central axons in zn-12 antibody-stained MZcntn2−/− embryos at 24 hpf (data not shown). However, because the RB central axon defasciculation effect was variable in the morphants (Liu and Halloran, 2005), we reasoned that a behavioral test such as the touch-evoked escape response (Granato et al., 1996) would be a more sensitive assay for putative functional defects associated with these neurons. The escape responses of 2 dpf larvae to head or trunk touch were binned (blinded to genotype) into three categories: No response, Weak response, and Strong response (see Materials and Methods). Larvae of both genotypes (cntn2+/+ and MZcntn2−/−) behaved similarly when touched on the head (Fig. 6A). However, larger numbers of MZcntn2−/− mutant larvae than cntn2+/+ larvae exhibited weak responses when touched in the trunk (Fig. 6B). Touch-evoked escape response defects in the trunk but not in the head suggest that MZcntn2 mutants may have a defect associated specifically with the RB neuron-mediated spinal sensorimotor circuit. Importantly, outgrowth of spinal motor axons was not affected in mutants (Fig. S4) consistent with a role for cntn2 in assembling the sensory component of the escape response circuit. While there was no clear defect in RB central axon outgrowth in mutants, we cannot rule out defects in other aspects of RB morphology or differentiation.

To further examine the effects of Cntn2 protein loss on the function of sensorimotor circuits, we analyzed swimming activity using the DanioVision system. Since Cntn2 is expressed in a subset of retinal ganglion cells (RGCs) and their axons (Fig. S1E–F), we examined light-evoked behaviors by measuring the swimming activity of 7 dpf cntn2+/− and MZcntn2−/− larvae during “Lights off” and “Lights on” phases (60 minutes each) (Fig. 7A). Larval genotypes were determined after swimming activity measurements (See Materials and Methods). Total swimming distance and swimming duration were significantly lower in MZcntn2−/− mutants compared to cntn2+/− siblings during both phases (Fig. 7B, C). Importantly, total distance moved and activity during the “Lights off” phase were significantly lower than during the “Lights on” phase in both cntn2+/− and MZcntn2−/− larvae (Fig. 7B, C), indicating that light-dependent movement is conserved, and suggesting that the swimming deficits of MZcntn2 mutants result from circuit defects downstream of the visual pathway. We tested directly for visual pathway defects by examining the morphology of RGC axons following anterograde labeling with lipophilic dye DiI and photoconversion (see Materials and Methods). Since RGC axon fascicles were thinner in MZcntn2 mutants when compared to cntn2+/+ cousins, but not when compared to cntn2+/− siblings (Fig. S5), potential defects in the visual system cannot be ruled out. Taken together, the escape response and swimming activity phenotypes suggest that MZcntn2 mutants have deficits in sensorimotor circuitry, consistent with cntn2 expression in cell types (RB and RGC neurons) associated with these circuits.

Figure 7. cntn2 mutants exhibit swimming deficits.

(A) Swimming assay and analysis. (B, C) Distance moved and moving duration are compared between cntn2+/− heterozygote and MZcntn2−/− mutant larvae during Lights off (B) and Lights on (C) phases. There were significant differences between cntn2+/− (n=43) and MZcntn2−/− (n=27). MZcntn2−/− larvae moved less than cntn2+/− siblings during both lights off (unpaired t-test, *p<0.05) and lights on (unpaired t-test with Welch’s correction, **p<0.001) phases (B, C). MZcntn2−/− larvae also moved for shorter duration compared to cntn2+/− siblings during both lights off (unpaired t-test, *p<0.05) and lights on (unpaired t-test, **p<0.001) phases (B, C). Error bars show Mean ± SD.

DISCUSSION

A large number of membrane-associated molecules have been implicated in the guidance of migrating neurons and projecting axons during nervous system development (Maness and Schachner, 2007). With the advent and ease of application of CRISPR/Cas9 technology, it is now possible to readily test the biological role of any gene in zebrafish (Hwang et al., 2013). Here, we used CRISPR/Cas9 to generate a null mutant for a neural cell adhesion molecule Cntn2. Our data demonstrate that while some of the previously-described morphant phenotypes may result from off-target effects, cntn2 also plays definitive roles in assembling neural circuits in zebrafish.

Comparison of cntn2 morphant and mutant phenotypes

Morpholino-mediated knockdown experiments previously suggested roles for cntn2 in the caudal migration of FBM neurons (Sittaramane et al., 2009), the guidance of axons from the nucMLF (Wolman et al., 2008), and the growth of RB central axons (Liu and Halloran, 2005). However, MZcntn2 mutants did not phenocopy most of these morphant phenotypes. There are several explanations for the differences between the morphant and mutant phenotypes. First, the mutant might make partially functional Cntn2 protein due to translation in-frame from start codons downstream of the mutation in exon 4. Second, the use of alternative, cryptic splice sites may skip exon 4 and yet generate truncated but functional Cntn2 protein. However, these outcomes appear unlikely since no Cntn2 protein is detected in MZ mutants (both alleles) using a polyclonal antibody generated using full-length Cntn2 protein (Lang et al., 2001). Such an antibody would likely recognize epitopes distributed throughout the length of the protein, and therefore detect truncated protein generated by alternative splicing or translational start sites. Exhaustive analysis of the genome assembly (GRCz11, 2017) did not identify any closely related genes (>75% similarity in Immunoglobulin and Fibronectin domains), suggesting that the absence of a strong mutant phenotype is not due to genetic redundancy.

Although a few studies have shown that morphants and mutants for some genes have similar or overlapping phenotypes (Bill et al., 2009; Phillips et al., 2011), there are other cases where the two phenotypes do not match (Lebedeva et al., 2017; Moore et al., 2016). An analysis of several genes affecting development and organogenesis showed that 80% of the morphant phenotypes were not seen in mutants, and concluded that morpholinos produce a high false-positive rate (Kok et al., 2015). Two models can explain the discrepancy in phenotypes between morphants and mutants. In the first model, MOs can generate off-target effects by specifically knocking down the expression of another unidentified gene, but they can also be toxic and generate non-specific phenotypes. In the second model, genetic compensation from related genes may occur in mutants, but not in morphants, resulting in the absence of a loss-of-function phenotype (Rossi et al., 2015). For instance, in egfl7 mutants, vascular defects generated in egfl7 morphants were not seen. However, several members of the emilin gene family were upregulated in mutants, potentially compensating for the loss of egfl7 (Rossi et al., 2015). It is possible that genomic disturbances caused by deleterious mutations activate compensatory pathways in mutants, while MOs, which only block the translation of mRNA, fail to do so.

In light of these studies, one explanation for the difference between the cntn2 morphant and mutant phenotypes is that other Cntn-like or other Ig superfamily CAM genes may be overexpressed and compensate for loss of cntn2 function in mutants but not morphants. The genetic compensation and off-target effect models were distinguished by examining the effect of injecting MO into the mutants (Fig. 5A; (Stainier et al., 2017)). Since the cntn2 morphant phenotypes for FBM and nucMLF neurons were readily seen in null mutants injected with cntn2 MO, it is unlikely that these phenotypes are due to loss of Cntn2 function, but rather are due to off-target effects. Nevertheless, we have also documented cntn2 mutant-specific phenotypes associated with the FBM and nucMLF neurons, indicating that cntn2 plays definitive roles in the development of these neuronal populations.

Role for Cntn2 in FBM neuron migration

During nervous system development, migrating neurons respond to several cues in the environment, and interact with neighboring cells (Marin et al., 2010). Vangl2, a core component of Wnt/Planar Cell Polarity (PCP) pathway (Gray et al., 2011), is expressed ubiquitously in the hindbrain, and functions both within the FBM neurons and in the surrounding neuroepithelial cells for FBM neuron migration (Davey et al., 2016; Jessen et al., 2002; Sittaramane et al., 2013; Sittaramane et al., 2009). cntn2 is expressed in branchiomotor neurons, but not in the surrounding neuroepithelial cells, during FBM neuron migration (Sittaramane et al., 2009; Warren et al., 1999), which indicates that cntn2 likely functions in a cell-autonomous fashion. Laminina1 (lama1), which encodes a subunit of the extracellular matrix protein Laminin1, is expressed broadly (Sittaramane et al., 2009) and plays a role in FBM neuron migration (Paulus and Halloran, 2006). Using a combination of mutants and MOs, we previously showed genetic interactions between cntn2 and vangl2, and between cntn2 and lama1 (Sittaramane et al., 2009). Here, using only mutants, we again observed genetic interactions between cntn2 and vangl2, but not between cntn2 and lamc1, which encodes a different obligatory subunit of Laminin1. These data indicate that a role for cntn2 in FBM neuron migration can be discerned in a sensitized vangl2+/− background but not in a sensitized lamc1+/− background. In addition to vangl2 and lamc1, we tested for genetic interactions between cntn2 and other PCP (fzd3a, scrb1) and non-PCP genes (cdh2, gpr125) that have been implicated in FBM neuron migration (Li et al., 2013; Rebman et al., 2016; Stockinger et al., 2011; Wada et al., 2005; Wada et al., 2006). We detected genetic interactions for cntn2 only with vangl2, and not with any of the other genes tested (data not shown). Indeed, vangl2 is rather unique in exhibiting genetic interactions with several PCP and non-PCP genes (vangl2; fzd3a and vangl2; cdh2, for example), while many of the other combinations (fzd3a; scrb1 or fzd3a; cdh2, for example) also do not exhibit genetic interactions (Gurung, S., Hummel, D., Chandrasekhar, A., in preparation). Given the central role of Vangl2 in Wnt/PCP signaling, only the vangl2+/− background may be sensitive enough to detect a subtle role for Cntn2 in FBM neuron migration.

Role for Cntn2 in nucMLF development

The nucMLF neurons extend dendrites toward retino-recipient areas and project axons to the spinal cord (Gahtan et al., 2005; Kimmel et al., 1982), and have been implicated in modulating locomotor speed, optomotor response, escape swimming and prey capture (Sankrithi and O’Malley, 2010; Severi et al., 2014). Morpholino-mediated knockdown experiments suggested roles for cntn2 in the initial oriented growth of nucMLF axons, and their convergence into a tight fascicle (Wolman et al., 2008). We show here that MZcntn2 mutants do not recapitulate this nucMLF morphant phenotype. However, MLF axons in cntn2 mutants do exhibit defasciculation in the midbrain and anterior hindbrain (rhombomeres 1–2). These data suggest that Cntn2 facilitates adhesions between MLF axons, which is consistent with its ability to interact homophilically, as well as heterophilically with other cell adhesion molecules (Brummendorf and Rathjen, 1996).

FBM neurons migrate in close proximity to MLF axons along the entire pathway from r4 to r7 (Bingham et al., 2005; Wada et al., 2006; Wanner and Prince, 2013). The migration of FBM neurons from r5 to r6 depends upon interactions between FBM neurons and MLF axons mediated by the cell adhesion molecule Cadherin2 (Cdh2) (Wanner and Prince, 2013). Since MLF axons are frequently defasciculated in cntn2 mutants, it is possible that the interactions between FBM neurons and MLF axons are also partially disrupted in cntn2 mutants. It will be interesting to investigate potential roles for cntn2 and cdh2 in regulating FBM neuron migration indirectly through MLF axon-FBM neuron interactions.

Role for Cntn2 in sensorimotor circuits

In the mouse Cntn2 mutant, retinal ganglion cell (RGC) sensory axons, which normally express Cntn2, lose their preference to extend on Cntn2-coated substrates, suggesting that RGC axon fasciculation requires homophilic Cntn2 interactions (Chatzopoulou et al., 2008). In zebrafish, Cntn2 is expressed on RGC neurons and axons (Fig. S1E–F). Although no obvious defects in RGC fasciculation or pathfinding were evident in MZcntn2 mutants (Fig. S5), the RGC axon fascicle was thinner in some cases, suggesting possible effects on outgrowth. We tested for a potential role in the visual system by measuring swimming activity in light and dark phases. While swimming duration and distance were significantly decreased in MZ mutants compared to controls in both light and dark conditions, both measures of swimming were similarly affected in control and MZ mutants when switching from light to dark phase, indicating that the regulation of motor activity by the visual system was not affected in mutants. It is more likely that the swimming deficits are generated by deficits in circuits downstream of the visual system for two reasons. First, given the nucMLF fasciculation defects in MZcntn2 mutants, their swimming deficits are consistent with the role of the nucMLF in modulating locomotion in larval zebrafish (Severi et al., 2014). Second, some of the swimming deficits in MZ mutants may be attributed to a role for cntn2 in the outgrowth or function of sensory neurons in the spinal motor circuits. While we did not find any growth defects in Rohon-Beard (RB) central axons in MZcntn2 mutants (Fig. S3), as seen previously in cntn2 morphants (Liu and Halloran, 2005), other aspects of RB morphology or differentiation could potentially be affected. Therefore, our finding that cntn2 mutants exhibit weaker escape responses to trunk touch compared to wildtype controls supports a role for cntn2 in the function of the RB sensorimotor circuit. Since there are no defects in motor and RB axon outgrowth in cntn2 mutants, defects in the connectivity and/or function of sensorimotor components seem more likely. Indeed, some of the motor deficits seen in the mouse Cntn2 mutant (Savvaki et al., 2008) may result from improper development or function of dorsal root ganglion sensory axons (Kunz et al., 1998; Law et al., 2008).

In conclusion, our characterization of the cntn2 CRISPR mutant has revealed definitive roles for the Cntn2 cell adhesion molecule in the zebrafish nervous system. It will be of interest to identify other membrane proteins that function in concert with Cntn2 to regulate FBM neuron migration and MLF axon fasciculation in zebrafish.

Highlights.

• cntn2 has a definitive but subtle role during FBM neuron migration

• Most phenotypes seen in cntn2 morphants are likely due to off-target effects

• cntn2 mutants exhibit axon fasciculation defects

• cntn2 mutants exhibit swimming and touch response defects