Abstract
Neuroepithelial cell (NEC) elongation is one of several key cell behaviors that mediate the tissue-level morphogenetic movements that shape the neural tube (NT), the precursor of the brain and spinal cord. However, the upstream signals that promote NEC elongation have been difficult to tease apart from those regulating apico-basal polarity and hingepoint formation, due to their confounding interdependence. The Repulsive Guidance Molecule a (Rgma)/Neogenin 1 (Neo1) signaling pathway plays a conserved role in NT formation (neurulation) and is reported to regulate both NEC elongation and apico-basal polarity, through signal transduction events that have not been identified. We examine here the role of Rgma/Neo1 signaling in zebrafish (sex unknown), an organism that does not use hingepoints to shape its hindbrain, thereby enabling a direct assessment of the role of this pathway in NEC elongation. We confirm that Rgma/Neo1 signaling is required for microtubule-mediated NEC elongation, and demonstrate via cell transplantation that Neo1 functions cell autonomously to promote elongation. However, in contrast to previous findings, our data do not support a role for this pathway in establishing apical junctional complexes. Last, we provide evidence that Rgma promotes Neo1 glycosylation and intramembrane proteolysis, resulting in the production of a transient, nuclear intracellular fragment (NeoICD). Partial rescue of Neo1a and Rgma knockdown embryos by overexpressing neoICD suggests that this proteolytic cleavage is essential for neurulation. Based on these observations, we propose that RGMA-induced NEO1 proteolysis orchestrates NT morphogenesis by promoting NEC elongation independently of the establishment of apical junctional complexes.
SIGNIFICANCE STATEMENT The neural tube, the CNS precursor, is shaped during neurulation. Neural tube defects occur frequently, yet underlying genetic risk factors are poorly understood. Neuroepithelial cell (NEC) elongation is essential for proper completion of neurulation. Thus, connecting NEC elongation with the molecular pathways that control this process is expected to reveal novel neural tube defect risk factors and increase our understanding of NT development. Effectors of cell elongation include microtubules and microtubule-associated proteins; however, upstream regulators remain controversial due to the confounding interdependence of cell elongation and establishment of apico-basal polarity. Here, we reveal that Rgma-Neo1 signaling controls NEC elongation independently of the establishment of apical junctional complexes and identify Rgma-induced Neo1 proteolytic cleavage as a key upstream signaling event.
Introduction
The CNS derives from the neural tube (NT) that is formed during neurulation. Hallmarks of primary neurulation include the thickening of the neural plate (NP), narrowing and lengthening of the NP, elevation of the lateral borders of the NP to form neural folds, fusion of the neural folds at the dorsal midline, and separation of the neural folds from the overlying non-neural ectoderm (Colas and Schoenwolf, 2001). Significant inroads have been made into understanding the cellular basis of these tissue-level changes; however, connecting specific cellular behaviors to the signaling pathways that control them has proven more challenging.
The zebrafish is ideally suited to study the cellular basis of morphogenesis because of the early accessibility and transparency of its embryo. Despite its mesenchymal-like appearance, the zebrafish NP is shaped into a tube by organized epithelial infolding, akin to primary neurulation (Papan and Campos-Ortega, 1994). The transient structure formed by infolding, termed the neural keel, becomes a solid neural rod by mid-somitogenesis and matures into a NT with a clearly defined midline and lumen following cavitation (Papan and Campos-Ortega, 1994).
During neurulation, cells undergo significant changes in shape, from cuboidal to columnar, as they elongate along their apico-basal axis (Schroeder, 1970; Burnside, 1971; Karfunkel, 1974; Schoenwolf and Franks, 1984). Cell elongation is required for thickening the NP of amniotes, resolving the bilayered NP of zebrafish and Xenopus embryos into a monolayered neuroepithelium via radial intercalation (Hong and Brewster, 2006; Kee et al., 2008) and elevating the neural folds, in absence of which the NT fails to close (Karfunkel, 1971, 1972; Suzuki et al., 2012). Apical constriction, another essential cell behavior, causes neuroepithelial cells (NECs) to adopt a wedge shape (Suzuki et al., 2012), resulting in the formation of one median and two dorsolateral hingepoints (Shum and Copp, 1996), around which the NP of amniotes bends and folds.
The upstream signaling events that control NEC elongation are not well understood. In a landmark study (Kee et al., 2008), Rgma and its receptor Neo1 were identified as upstream regulators of NEC elongation and establishment of apico-basal polarity. However, the interdependence of both of these processes in Xenopus (Suzuki et al., 2012) prevented a direct assessment of the role of Rgma-Neo1 signaling in controlling either process.
neo1 encodes a transmembrane receptor belonging to the immunoglobulin superfamily. Repulsive guidance molecules are a family of glycosylphosphatidylinositol-anchored proteins that function as membrane-bound, short-range guidance cues or as secreted, long-range signals (Tassew et al., 2012). Rgma-Neo1 signaling plays a conserved role in neurulation, as knockdown of the ligand and/or receptor prevents NT closure in mice (Niederkofler et al., 2004) and Xenopus (Kee et al., 2008). The signaling events activated downstream of Rgma-Neo1 interaction that promote cell elongation and NT closure are currently unknown; however, previous studies have identified several putative signaling mechanisms. Both Rgma and Neo1 are known to modulate bone morphogenetic protein signaling (Babitt et al., 2005; Zhou et al., 2010; Tian and Liu, 2013). Repulsive guidance molecules can also directly bind to Neo1, triggering structural changes to the actin cytoskeleton through the Rho family of small guanosine-5′-triphosphate-hydrolyzing GTPases (Hata et al., 2006; Conrad et al., 2007). Furthermore, Neo1 is sequentially cleaved by α- and γ-secretases in response to Rgma binding, leading to the release of a Neo1 intracellular domain (NeoICD) that regulates transcription (Goldschneider et al., 2008; van Erp et al., 2015). Neo1 proteolytic cleavage has been implicated in axonal pathfinding and neuronal migration (van Erp et al., 2015; Banerjee et al., 2016).
In zebrafish embryos, NEC elongation precedes the establishment of apical junctional complexes (Hong and Brewster, 2006), making it an ideal model for teasing apart the signals required for cell elongation specifically. This study aims to determine the role of Rgma-Neo1 in NEC elongation and to identify the signal transduction events triggered by ligand–receptor interaction that impinge on neurulation.
Materials and Methods
Husbandry, care, and use of zebrafish
WT zebrafish (Danio rerio) of the AB strain were reared and manipulated using protocols approved by the Institutional Animal Care and Use Committee at the University of Maryland Baltimore County. Fish were maintained in UV-irradiated, filtered running water and exposed to 14:10 light/dark cycle. Male and female fish were separated by a partition that was removed after first light to initiate spawning of fry (embryos). Embryos were collected and staged according to previously described methods (Kimmel et al., 1995). The sex of the embryos used is unknown.
Construction of dominant-negative neo1a plasmid
Custom-made neo1aδ1133-FLAG (GenScript) was modeled after a previously published dominant-negative neo1 construct used in Xenopus (Kee et al., 2008) and humans (Enns et al., 2012). To generate neo1aδ1133-FLAG, a partial zebrafish neo1a coding sequence (Accession no. Q8AY67) was subcloned from pUC57 into the pCS2+ vector. This insert encodes a mutant form of neo1a lacking the C terminal region and spans amino acids 1–1132. This sequence is immediately followed by one corresponding to a single C-terminal FLAG epitope tag (5′-GACTACAAAGACGATGACGACAAG-3′) (inserted after base pair 3397, corresponding to amino acid residue 1132), a translation termination sequence (TGA), and the full 3′ UTR and poly A tail (pA) of neo1a (Accession no. Q8AY67). For efficient in vitro transcription, the T7 promoter was introduced at the 5′ end of the insert followed by a 10-mer poly-thymine spacer. A SnaBI restriction site (TACGTA) was added 3′ to the neo1a pA just before the SV40 pA signal of the pCS2+ vector to provide a unique site for linearization of the vector before transcription.
Construction of mutant NeoICD plasmid
The nuclear localization sequence of NeoICD corresponds to the N terminal region of the ICD and has the following peptide sequence: RRTTTQQKKKR. To generate a peptide that fails to localize to the nucleus, the pCS2+ plasmid containing neoICD-6x myc (GenScript) was mutated to include two substitutions respective to full-length Neo1a, R1123A/R1124A, generating the mutant peptide sequence AATTTQQKKKR (designed based on Goldschneider et al., 2008). The resultant mutant construct was named neoICDR1123A/R1124A-6x myc (neoICDmut).
Morpholino (MO)-mediated Rgma and Neo1a knockdown
MOs.
MOs were commercially obtained (Gene Tools) and administered at 5 ng/embryo, with the exception of one Western blot assay as indicated in Results. Each MO was injected at the 1–4 cell stages into the yolk margin just underneath the embryo proper. The MO sequences are as follows: rgma 5′UTR MO1: 5′-ACACTTTTGGGTTGAGTTTCTTTTTGC-3′ (Gibert et al., 2011); rgma AUG MO2: 5′-CTCTCCTCTCCCTTTTTGAATGCAT-3′; neo1a AUG MO2: 5′-GGCTCCCCGCTCCGCCATCACTTTA-3′ (Mawdsley et al., 2004); and standard control MO (std MO): 5′-CCTCTTACCTCAGTTACAATTTATA-3′ (Schmajuk et al., 1999). neo1a MO2 is complementary to 25 and 19 consecutive bases of neo1a and neo1b, respectively.
Dominant-negative neo1a and neoICD.
Dominant-negative (neo1aδ1133-FLAG)-pCS2+ plasmid was linearized with SnaBI (NEB) and in vitro transcribed with T7 polymerase using the T7 mMESSAGE mMACHINE kit (Ambion, catalog #AM1344), according to the manufacturer's guidelines for GC-rich templates; 120 pg of column-purified neo1aδ1133-FLAG mRNA was injected into the cytoplasm of 1–4 cell stage embryos.
The coding sequence corresponding to the 981 bp intracellular domain of zebrafish neo1a (neoICD) was amplified from the full-length Neo1a subcloned into pCS2+ vector (GenScript) containing full-length neo1a coding sequence to produce a DNA template using the following primers: T7-neoICD forward: 5′-TAATACGACTCACTATAGggATGCTTTGTACACGTCGTACCACA-3′; and neoICD reverse: 5′-TCAGGCTGTTGTGATGGC-3′. PCR products were transcribed using T7 mMessage Machine kit (Ambion); 50 pg of column-purified neoICD mRNA was injected.
CRISPR/Cas9-induced rgma and neo1a indels
Target selection.
The online bioinformatics tool CHOPCHOP (Montague et al., 2014; Labun et al., 2016) was used to identify unique CRISPR targets for rgma (Ensembl reference number ENSDARG00000012248 from the zebrafish genome reference consortium assembly 9 [GRCz9]) and neo1a (ENSDARG00000102855 from the GRCz10 and GRCz11, NM_173218.2). Each guide RNA (gRNA) target sequence was chosen for its lack of homology to other sequences in the published zebrafish genome, predicted off-target effects, and self-complementation. The CRISPR target sequences are as follows: rgma exon 2: 5′-GGTGTAAATAGGGAGAGGAG-3′; rgma exon 3: 5′-GGTGGGACTGTGCGGGCACG-3′; neo1a exon 3: 5′-CCGCTGGGAACAGAATAAAG-3′; neo1a exon 3: 5′-CTAATGCCAGGCTCCAGAGG-3′; neo1b exon 3: 5′-GAGGGTGTTTCTGTTCGCAG-3′; and neo1b exon 8: 5′-AGAGACGCCACCACATCCCG-3′.
Generation of insertion-deletions (indels).
Embryos were mutagenized by active ribonucleoproteins consisting of the Alt-R CRISPR/Cas9 high-fidelity nuclear-targeted Cas9 protein (HF-nCas9) and exon-specific gRNAs according to the manufacturer's recommendations (IDT) (Kim et al., 2014; Burger et al., 2016; Anderson et al., 2017, 2018). Ribonucleoproteins were injected individually or multiplexed to target multiple exons. Early 1 cell stage embryos were injecting with 3 pmol of each gRNA and 0.5 ng of HF-nCas9.
High-resolution melt analysis.
Genomic DNA (gDNA) extractions from individual embryos were prepared by alkaline lysis (Meeker et al., 2007), separated from particulates via centrifugation, and diluted 1:100 to be used as PCR templates. SYBR Green qPCR was performed using Precision Melt Supermix (Bio-Rad) according to the manufacturer's suggestions. The following primer sets were used to amplify gDNA for 40 cycles as follows: rgma exon 2 forward: 5′-CACGCAGAGTACCTGACAAAAG-3′; rgma exon 2 reverse: 5′-AAGGAACTGGCAGACTTGTAGC-3′; rgma exon 3 forward: 5′-AGAAGAGGAGTTTTGCACTGCT-3′; rgma exon 3 reverse: 5′-TATGTGGTAGACTCCGCTCGTA-3′; neo1a exon 3 forward: 5′-CTTGCACAAAAAGTGGAACAAT-3′; neo1a exon 3 reverse: 5′-GTCTGTCTCTGTGGCGTTACTG-3′; neo1a exon 8 forward: 5′-CGTCACCCTACTGTCTTCACTG-3′; neo1a exon 8 reverse: 5′-TATTAATGTGACGTAGGTGCCC-3′; neo1b exon 3 forward: 5′-TGGTTTTCTGTAGTAACGCAGC-3′; neo1b exon 3 reverse: 5′-TCTATGGTGCAACGGTAGAGTC-3′; neo1b exon 8 forward: 5′-CTACTGACCATGTAGCGTCCAG-3′; and neo1b exon 8 reverse: 5′-TAGTGCCGTCCAGACTGTAGAA-3′.
Amplification and melt analysis at 0.2°C increments were performed on a CFX96 real-time system with C1000 cycler (Bio-Rad). Baseline normalization and difference curves were obtained by thresholding and processing the melt RFU data using uAnalyze (https://dna.utah.edu/uv/uanalyze.html). Melting point homogeneity and heterogeneity were expected among PCR products derived from WT and mosaic mutant embryos, respectively.
Microinjections of mRNA, gRNAs, MO, DNA, and protein
Microinjections were performed using a PCI 100 Microinjector (Harvard Apparatus). Embryos at the 1–4 cell stage were injected with 2 nl per embryo of appropriate concentration of mRNA, gRNA, MO, plasmid DNA, or protein. Cytoplasmic injections were performed for RNA injections while MOs and plasmid DNA were injected into the yolk-cytoplasm margin. The std MO was designed against the human β-globin intron that is not present in the zebrafish genome (Nasevicius and Ekker, 2000; Eisen and Smith, 2008; Bill et al., 2009; Moulton, 2017; Stainier et al., 2017).
Microinjection of cell surface markers mGFP and mRFP.
To achieve mosaic expression of the cell surface marker membrane-targeted green fluorescent protein (mGFP), plasmid (obtained from Richard Harland, University of California, Berkeley, CA) was prepared using a midiprep kit (Macherey-Nagel, catalog #740410.10) and injected (50–200 pg) into 1–4 cell stage embryos. Ubiquitous expression of the cell surface marker membrane-targeted red fluorescent protein (mRFP; plasmid provided by Steve Farber, Carnegie Institution for Science) (Megason and Fraser, 2003) was achieved by injecting mRNA encoding mRFP. Plasmid containing mRFP was linearized with NotI, transcribed using SP6 mMESSAGE mMACHINE kit (Ambion, catalog #AM1340) and injected (50 pg) at the 1–4 cell stage.
Microinjection of neoICD.
Rescue experiments were performed by injecting 1–4 cell stage embryos with 5 ng of neo1a MO2, 50 pg of neoICD/NeoICDmut, or both.
All other microinjection experiments using nucleic acid or protein were performed as described in previous sections.
Cell transplantation
Transplantation was performed as described previously (Kemp et al., 2009; Jayachandran et al., 2010). Briefly, 50–100 cells from donors, injected with mRFP RNA, were transplanted isochronically into the animal pole of host embryos at the sphere-dome stage injected with mGFP DNA.
Whole-mount in situ hybridization (WISH)
Embryos were fixed in 4% PFA at the desired stages and processed for in situ hybridization as previously described (Thisse and Thisse, 2008; Jayachandran et al., 2016), using riboprobes against zebrafish distal-less homeobox 3B (dlx3b), rgma, and neo1a. Plasmid containing dlx3b (obtained from Igor Dawid, National Institutes of Health) was linearized with NotI, and T7 polymerase was used to generate antisense probe. Sense and antisense riboprobes were synthesized from a pCRII-TOPO vector containing full-length zebrafish rgma (P.J. and R.B., unpublished) by linearizing/transcribing with NotI/SP6 and HindIII/T7, respectively. The neo1a EST cb333 (Zebrafish Information Network) was linearized with ScaI, and T7 or T3 polymerases were used to generate sense and antisense riboprobes, respectively.
Immunolabeling and staining
For microtubule (MT) analysis, embryos were fixed in MT assembly buffer containing 3.7% PFA, 0.25% glutaraldehyde, 0.2% Triton X-100, 1 mm MgCl2, 5 mm EGTA, and 0.5 μm paclitaxel as previously described (Topczewski and Solnica-Krezel, 1999; Tran et al., 2012). For cell membrane analysis, embryos injected with mGFP DNA or mRFP mRNA were fixed in 4% PFA and processed for immunofluorescence using 1:1000 rabbit anti-GFP antibodies (Invitrogen, catalog #A11122; RRID:AB_221569). mRFP fluorescence was detected directly by confocal microscopy (described below). For analysis of F-actin, PFA (4%)-fixed embryos were stained with 1:250 Phalloidin-Alexa-488 (Invitrogen, catalog #A12379) for 2 h according to the manufacturer's protocol. Nuclei were labeled with DAPI (Invitrogen, catalog #D1306) according to the manufacturer's instructions.
Imaging embryos
Bright field imaging.
Live, dechorionated embryos were immobilized using the sedative tricaine methanesulfonate at 0.001% w/v in embryo medium E3 (5 mm NaCl, 0.33 mm CaCl2, 0.17 mm KCl, 0.33 mm MgSO4). Embryos were mounted on depression slides atop droplets of 4% w/v methylcellulose in E3. Side views of 32 hpf larvae were captured at 50× total magnification on a Axioskop II compound microscope (Carl Zeiss) and AxioCam 503 CCD camera (Carl Zeiss).
For fixed preparations, embryos labeled using in situ hybridization were sectioned (40 μm) at the level of the hindbrain using a sectioning system (Vibratome, 1500), mounted in glycerol, and imaged at 50× magnification using an AxioSkopII microscope (Carl Zeiss) and CCD camera (Q-Capture).
Confocal microscopy.
Time-lapse microscopy was performed on dechorionated live embryos injected with fluorescent markers embedded in <1 mm holes that were bored in 1.2% low melting point agarose (Shelton Scientific-IBI) solidified on glass-bottom culture dishes with size 1.5 coverslips (Mattek). Embryos were imaged from a dorsal view using an inverted SP5 laser scanning confocal microscope (Leica Microsystems) with a 40× oil objective lens.
For fixed preparations, immunolabeled embryos were sectioned (40 μm) at the level of the hindbrain using a sectioning system (Vibratome, 1500) and mounted in 4% low melting point agarose for confocal imaging using an inverted confocal microscope (Leica Microsystems, SP5) with a 40× oil objective lens. Images were analyzed, processed, and annotated using ImageJ (National Institutes of Health), Photoshop (Adobe), and PowerPoint (Microsoft).
Western blot analysis
Embryos were mechanically lysed in precooled 1× RIPA buffer (20 mm Tris-HCl, pH 7.5, 150 mm NaCl, 1 mm Na2 EDTA, 1 mm EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mm sodium pyrophosphate, 1 mm β-glycerophosphate, 1 mm Na3VO4, and 1 μg/ml leupeptin) supplemented with protease inhibitors (1× final concentration) (Sigma-Aldrich, catalog #P8340). Bicinchoninic acid protein assay was used to determine the total protein concentration of embryonic lysates. Western blot analysis was performed by loading 7.5 μg total protein per lane into 7.5% Tris-glycine SDS-PAGE gels (Bio-Rad, catalog #4561026) that were transferred to methanol-activated PVDF membranes. Specific protein detection was performed using diluted primary antibodies as follows: rabbit anti-Neo1 (LSBIO, catalog #LS-B13131; RRID:AB_2620594) at 1:1000; custom-made rabbit anti-Rgma (GenScript; RRID:AB_2750931) at 0.64 ng/ml; mouse anti-FLAG (GeneTex, catalog #GTX18230; RRID:AB_423136) at 1:10,000; rabbit anti-GAPDH (GeneTex, catalog #GTX124503; RRID:AB_11165273) at 1:1000–1:5000 and goat anti-rabbit secondary antibodies (Cell Signaling Technology, catalog #7074; RRID:AB_2099233) at 1:10,000. Molecular weights were determined using prestained protein ladders (Thermo Fisher Scientific, catalog #26616; GE Healthcare, catalog #RPN800E). Protein abundance was detected using the SuperSignal West Femto enhanced chemiluminescence kit (Thermo Fisher Scientific, catalog #34095) and Blu-C high-contrast autoradiography film (Stellar Scientific, catalog #BLC-57-100).
Drug treatments
Inactivation of secretases.
To inactivate secretases, embryos were pretreated with 1% DMSO in embryo medium (E3) at standard temperature (28.5°C) for 1 h to permeabilize their chorions before treatment with inhibitors. The α-secretase/a disintegrin and metalloprotease domain 17 (ADAM17) inhibitor, TNFα protease inhibitor I (TAPI-1) and the γ-secretase inhibitor N-[N-(3, 5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester (DAPT) were prepared freshly in prewarmed 1% DMSO/E3 solution containing the proteasome and nuclear export inhibitors MG132 and leptomycin B (LMB), respectively, to suppress protein recycling and 0.002% methanol (MeOH) to enhance drug solubility. Embryos were treated with 1% DMSO, 100 μm TAPI-1, or 8 ng/μl DAPT starting and ending at the 6 somite and 20 somite stages (som), respectively, before mechanical lysis in 1× RIPA buffer.
Deglycosylation using PNGaseF.
A 20 μl lysate of 1 d-post-fertilization (dpf) embryos was split evenly between two tubes (10 μl each) on ice. Lysates were treated with 5 μl of PNGaseF buffer (1× final concentration) with or without PNGaseF enzyme using a kit (NEB, catalog #P0704). PNGaseF incubated on a thermocycler set at 37°C for 1 h. Lysates were subjected to Western blotting under reducing conditions.
Periodic acid Schiff assay
Proteins obtained from 1 dpf std MO- or rgma MO2-injected larval lysates were concentrated using a 10,000 molecular weight (MW) protein concentrator column (Pierce, catalog #88527) and separated using SDS-PAGE on 7.5% Tris-glycine polyacrylamide gels (Bio-Rad). Gels were processed for fuchsin labeling of periodic acid-oxidized glycans using the Glycoprotein Staining Kit (Thermo Fisher Scientific, catalog #24562), according to the manufacturer's recommendations. Replicate lanes were stained with Coomassie to assay total protein or transferred to PVDF membranes for Western blot analysis using Neo1 antibodies.
Subcellular fractionation
The 1 dpf larvae were mechanically lysed in precooled (4°C) 1× RIPA buffer (20 mm Tris-HCl, pH 7.5, 150 mm NaCl, 1 mm Na2 EDTA, 1 mm EGTA, 1% NP-40, 1% sodium deoxycholate, 2.5 mm sodium pyrophosphate, 1 mm β-glycerophosphate, 1 mm Na3VO4, and 1 μg/ml leupeptin) supplemented with protease inhibitors (1× final concentration). Lysates were centrifuged at 6000 revolutions per minute at 4°C for 6 min to pellet nuclei. Supernatants were transferred to fresh tubes and stored on ice. Pellets were washed twice in precooled (4°C) 1× RIPA with protease inhibitors and stored on ice; 6× reducing sample buffer (375 mm Tris-HCl, pH 6.8, 6% SDS, 9% β-mercaptoethanol, 0.03% bromophenol blue, 4.8% glycerol) was added to nuclear pellets and supernatants at a final concentration of 1× before boiling sample for Western blot analysis.
Measurements
All measurements of neural tissue were taken at the level of the hindbrain that was identified with several visual landmarks (shape of neural keel or NT, presence of otic vesicles).
NP width.
NP width was measured as the distance (in micrometers) between the lateral domains of dlx3b expression at the level of rhombomere 5, as confirmed by krox20 labeling.
NEC length-to-width ratio.
The NEC length-to-width ratio was obtained by measuring the long and short axes (in micrometers) of mGFP- or mRFP-expressing NECs in hindbrain sections of embryos at 4–5 som.
C division index, distribution, and orientation.
Measurements were taken from single-frame images of 8–10 som embryos labeled with anti-β-tubulin, counterstained with DAPI, and sectioned at the level of the hindbrain. ImageJ (National Institutes of Health) was used to quantify mitotic indices, the position of C divisions relative to the midline, and the angle of C divisions relative to the midline. Mitotic indices were quantified in the medial aspect of the neural tissue (region spanning 30 μm on either side of the midline) as the number of nuclei with condensed chromatin divided by the total number of nuclei multiplied by 100. To determine the angle of cell divisions, a line (radial distance, r) was drawn to connect the condensed chromatin of conjoined daughter cells at the telophase stage. The orientation of C divisions was scored as the elevation angle or azimuthal angle (θ) between r and the mediolateral (horizontal) axis of the neural tissue, such that a mediolaterally oriented C division has a value of 0°. For metaphase cells, r was inferred from the orientation of mitotic spindles stained by β-tubulin antibodies.
Experimental design and statistical analyses
Graphing and statistical analyses were performed using Prism 7 (GraphPad). Statistical significance was declared in circumstances where p ≤ 0.05. Experiments comparing three or more groups were analyzed by one-way ANOVA. When comparing 2 means within a dataset, p values were obtained by performing a post hoc Tukey test. Data were reported as the mean ± SEM unless otherwise indicated.
Rose plots (radial grids)
Image stacks from time-lapse microscopy were collapsed to a single frame for each time point using Volocity version 5.5 (PerkinElmer) and exported as Tif files. Only interphase cells that remained in-frame for the duration of the movie were analyzed. Using a plugin created for ImageJ (National Institutes of Health), individual cells were thresholded and outlined. Based on the centroid and the orientation relative to the midline, each cell was then divided into 8 radial segments. Finally, individual frames were superimposed, and any new membrane extensions of 3 pixels or more were counted for each segment and frame. The percentage of protrusions for each segment was then quantified and graphed in Mathematica version 9 (Wolfram).
Results
Tools for knocking down Rgma and Neo1
In contrast to other vertebrate models used to study neurulation, zebrafish undergo delayed establishment and maturation of apico-basal polarity, and hinge points have not been reported in this organism (Geldmacher-Voss et al., 2003; Lowery and Sive, 2004; Hong and Brewster, 2006), enabling the identification of molecular pathways controlling NEC elongation independently of those required for epithelialization. To address whether activation of Neo1 signaling is an early event that controls NEC elongation, we generated and validated several loss-of-function (LOF) tools for functional characterization of both Rgma and its receptor, Neo1.
Whereas rgma is encoded by a single gene in the zebrafish genome (accession #Q58EQ5), there are two neo1 paralogs, neo1a (accession #NM_173218) and neo1b (accession #E7F004), mapping to chromosomes 7 and 2, respectively. These paralogs share 65% overall amino acid homology and 99% homology in their C-terminal regions and may therefore be functionally redundant. LOF tools that target both neo1 paralogs in addition to rgma were therefore generated (Fig. 1), which include the following: CRISPR/Cas9-induced indels in the rgma (Fig. 1A–C,J), neo1a (Fig. 1A,D,E,L), and neo1b (Fig. 1A,F,G,L) loci using two gRNAs per gene, translation-blocking MOs (rgma MO1 and MO2) that target the 5′UTR (Fig. 1A, top) and the translational start site region (AUG) of rgma, respectively (Fig. 1A,K) as well as the neo1a AUG (neo1a MO2) (Fig. 1A,M) and a neo1a construct predicted to function in a dominant-negative manner to disrupt Rgma-Neo downstream signaling, neo1aδ1133-FLAG, which encodes a truncated protein lacking terminal amino acids 1133–1428 (Fig. 1I,N) (Wilson and Key, 2006).
rgma maternal-zygotic homozygous mutant lines generated using CRISPR/Cas9 genome editing (gRNAs targeting rgma exon 2 exclusively) appeared phenotypically WT (data not shown), a likely consequence of compensatory mechanisms that have been frequently reported, including upregulation of paralogs (Hartman et al., 2001; Gu et al., 2003; Conant and Wagner, 2004; Hanada et al., 2011; Sztal et al., 2018; To and Andrechek, 2018), nonhomologous genes (Peng et al., 2012; Li et al., 2014), and exon-skipping (Oda et al., 2016; Anderson et al., 2017; Moulton, 2017; Sharpe and Cooper, 2017). However, high-resolution melt analysis revealed that gRNAs targeting rgma exons 2 or 3 generated indels at high frequency (Fig. 1B,C), and the combined use of these gRNAs depleted Rgma protein in the first generation of injected embryos (F0s) (Fig. 1J). Likewise, we observed indels in embryos injected with Cas9 protein and gRNAs targeting neo1a exons 3 and 8 (Fig. 1D,E) or neo1b exons 3 and 8 (Fig. 1F,G) and a corresponding depletion of Neo1 protein levels in embryos injected with gRNAs targeting all 4 exons (Fig. 1L).
Western blot analysis further revealed that the MOs targeting rgma and neo1 were effective. rgma MO2 fully depleted Rgma protein (Fig. 1K). Using anti-Neo1, which is predicted to recognize the C terminus of both Neo1a and Neo1b, we observed that Neo1 was substantially, but not fully, depleted in neo1a MO2-injected embryos (Fig. 1M). This partial depletion suggests that neo1a MO targets the neo1a paralog only, which is consistent with the mismatched residues between the ATG region of neo1b and the neo1a MO2 sequence (Fig. 1H).
Last, we observed a 150 kD band in Western blots of protein extract from neo1aδ1133-FLAG-injected embryos using anti-FLAG, which corresponds to the truncated, FLAG-tagged protein. This band was not observed in control, uninjected embryos (Fig. 1I,N).
Morphological defects associated with Rgma and Neo1 LOF
To evaluate the efficacy of these LOF tools, embryos injected with rgma MO 1 or 2, neo1 MO2, neo1aδ1133-FLAG, or Cas9/gRNAs targeting rgma (rgmaex2+3), neo1a (neo1aex3+8), neo1b (neo1aex3+8), or neo1a and neo1b (neo1aex3+8/neo1bex3+8) were examined at 32 hpf for morphological defects (Fig. 2). In contrast to control embryos (uninjected and standard std MO-injected), perturbation of Rgma or Neo1 function caused graded phenotypes ranging from WT to severe (Fig. 2A–I). Phenotypes observed consisted of a shortened anterior–posterior axis, absence of brain morphological landmarks (forebrain, hindbrain, midbrain-hindbrain boundary), abnormally shaped somites, small eyes, and pigmentation. The mild, moderate, and severe phenotypic categories displayed one, two or three, or at least four of these traits, respectively (Fig. 2J). These observations indicate that rgma and neo1a/b are required to shape the early embryo, consistent with their maternal expression and distribution at later stages in neural tissue, mesoderm, and retina primordia (Thisse et al., 2001; Shen et al., 2002). Since neo1a and neo1b depletion causes severe morphological defects, they are at most partially redundant. Overall, these data support the model that Rgma and Neo1 function as a ligand-receptor pair during early development. Given the similarity between these phenotypes, we have used neo1a MO2 and rgma MO2 to perform some of the functional studies described below.
Rgma-Neo1a signaling promotes NEC elongation
To test the requirement of Rgma-Neo1 signaling in NEC elongation, we analyzed cell shapes in embryos co-injected with mGFP plasmid (for mosaic expression) and one of the following neo1a or rgma LOF tools: neo1a MO2, rgma MO2, neo1a gRNA exon 3 + 8 and HF-nCas9 or rgma gRNA exons 2 + 3 and and HF-nCas9, as previously described (Jayachandran et al., 2010) (Fig. 3). We observed that Neo1a- and Rgma-depleted NECs appeared rounder (Fig. 3B,B′,D,D′,C,C′,E,E′, respectively) than uninjected control cells (Fig. 3A,A′), consistent with impaired cell elongation. Measurement of length-to-width ratios (LWRs) confirmed a significant difference in cell shape between control cells and experimental groups (F(4,740) = 67.48, p < 0.0001, ANOVA) (Fig. 3F). Whereas the mean LWR for control cells was 3.58 ± 0.14 (n = 55 cells, 5 embryos, p < 0.0001, Tukey test), the LWRs of neo1a MO2 and neo1a gRNA exon 3 + 8 injected embryos was 1.54 ± 0.04 (n = 121 cells, 5 embryos, p < 0.0001, Tukey test) and 2.238 ± 0.07 (n = 188 cells,10 embryos, p < 0.0001, Tukey test), respectively, whereas the LWR of rgma MO2 and rgma gRNA exon 2 + 3 injected embryos was 1.55 ± 0.04 (n = 90 cells, 5 embryos, p < 0.0001, Tukey test) and 2.034 ± 0.05 (n = 291 cells, 10 embryos, p < 0.0001, Tukey test), respectively. To gain a better understanding of the dynamic cell shape changes that take place during infolding and the role of Neo1a signaling in this process, we performed time-lapse imaging of control and Neo1a-depleted embryos that mosaically express mRFP (Fig. 3G,H; Movies 1, 2). These movies, spanning 1–2 som to 5–6 som, revealed that control cells were initially rounded but gradually elongated along their prospective apico-basal polarity axis (Fig. 3G; n = 6 cells from 3 embryos). In contrast, NECs in neo1a MO2-injected embryos remained rounded throughout the duration of time-lapse imaging (87 min) (Fig. 3H; n = 6 cells from 3 embryos). To further analyze these cell behaviors, the outlines of interphase cells were traced, revealing a striking defect in cell elongation in Neo1a-depleted NECs compared with control cells (Fig. 3I and Fig. 3J, respectively). Furthermore, graphical representation of the distribution of the cell surface using rose plots confirmed that Neo1a-depleted NECs (Fig. 3L) failed to redistribute their plasma membrane along the mediolateral axis during infolding in contrast to control cells (Fig. 3K) (n = 6 cells/embryo; n = 3 embryos). All the cells analyzed displayed a similar behavior. Together, these data are consistent with the proposed role of Rgma-Neo1a signaling in promoting cell elongation during infolding.
rgma and neo1a are expressed in the developing NT
If rgma and neo1a directly control NEC elongation, both genes (or at least the receptor, neo1a) should be expressed in the developing NT. To test this, we analyzed their mRNA distribution in the presumptive hindbrain, at the NP (tailbud), neural keel (4–5 som), and neural rod (8–10 som) stages using WISH (Fig. 4). At the NP stage, cells in the deep neural layer express neo1a, consistent with the expression reported for its orthologue Xneog in Xenopus (Kee et al., 2008), and no labeling was detected in the superficial neural layer (Fig. 4A). neo1a expression was also observed in the lateral plate mesoderm at this and later stages (Fig. 4A–C). The spatial distribution of neo1a expands in older embryos, as weak expression is observed in the deep and superficial layers of the neural keel (Fig. 4B) and becomes more intense in the dorsal region of the neural rod (Fig. 4C). neo1b was expressed uniformly across the neural keel (data not shown), overlapping with the neo1a expression domain.
At the NP (Fig. 4G), neural keel (Fig. 4H), and neural rod stages (Fig. 4I), rgma is strongly expressed in the notochord and floor plate and to a lesser extent in the overlying neural tissue, where it is restricted to the ventral region of the neural keel and rod (Fig. 4H,I). This distribution is consistent with that of Xenopus Xrgma, which is robustly expressed in the deep and medial cells of NP and ventral NT, although no expression was detected in the notochord (Kee et al., 2008). The specificity of the labeling was confirmed using sense riboprobes for neo1a (Fig. 4D–F) and rgma (Fig. 4J–L). Thus, rgma and neo1a are expressed in a spatial and temporal manner that is consistent with their proposed function in regulating NEC elongation. However, the strong expression of both genes in the mesoderm raises the question of whether the NEC elongation defect we observe in Neo1a-depleted embryos may be indirectly attributed to defective convergent extension movements in the mesoderm (Mawdsley et al., 2004), especially given the severe anteroposterior axis shortening we observed in Neo1a knockdown embryos (Fig. 2E,F,H).
Neo1a functions cell autonomously in the neuroectoderm
To address whether Neo1a signaling is specifically required within the neural ectoderm to drive NEC elongation, we used cell transplantation to generate chimeric embryos consisting of WT and Neo1-depleted cells and measured NEC LWRs in transverse sections at the level of the hindbrain (F(3,349) = 85.9, p < 0.0001, ANOVA). mRFP and mGFP were used to differentially label cells from the donor and host, respectively. Donor embryos were injected at the 1 cell stage with mRNA encoding mRFP to uniformly label all cells. Hosts (recipients) were injected with DNA encoding mGFP, to mosaically express this reporter throughout the embryo (Fig. 5A–C′). When control (mRFP and Neo1a-expressing) cells were transplanted into host embryos, both donor and host NECs exhibited a comparable mean LWR (LWR of control donor cells = 3.54 ± 0.10; n = 100 cells from 18 embryos, LWR of control host cells = 3.70 ± 0.17; n = 67 cells from 14 embryos; p = 0.7348, Tukey test; Fig. 5D), which is consistent with previous measurements from uninjected embryos (Fig. 3F). However, when Neo1-depleted cells (from embryos injected with mRFP and neo1a MO2) were transplanted into host embryos, donor cells displayed lower LWRs compared with their neighboring host cells (LWR of host cells = 3.18 ± 0.10; n = 94 cells from 9 embryos, LWR of Neo1a-depleted cells = 1.54 ± 0.04; n = 92 cells from 10 embryos; p < 0.0001, Tukey test; Fig. 5D). Even though the reciprocal experiment, transplanting control (Neo1a-expressing) cells into neo1a MO2-injected hosts failed due to embryonic lethality, our transplantation data indicate that Neo1a functions in a cell-autonomous manner to regulate cell elongation in the neuroectoderm.
Rgma and Neo1a depletion impairs infolding
NEC elongation is a prerequisite for radial intercalation and establishment of transient contacts between contralateral NECs, cell behaviors that promote infolding of the NP (Papan and Campos-Ortega, 1994; Davidson and Keller, 1999; Marsden and DeSimone, 2001; Hong and Brewster, 2006). To test whether Rgma-Neo1a signaling is required for infolding, we compared the width of the NP in embryos injected with neo1a MO2 or rgma MO2 with that of control embryos that were either uninjected or injected with std MO (F(7,181) = 35.47, p < 0.0001, ANOVA). The 4–5 som embryos were processed for WISH using either dlx3b or krox 20 to image and measure the width of the NP at the level of rhombomere 5, which coincides with the position of the otic placode (Fig. 6A,B). We observed that the NP widths of embryos injected with rgma MO1 (248.4 ± 16.66 μm, n = 12 embryos, p = 0.0356, Tukey test), rgma MO2 (351.8 ± 9.254 μm, n = 18 embryos, p < 0.0001, Tukey test), neo1a MO2 (358.1 ± 12.22 μm, n = 7 embryos, p < 0.0001, Tukey test), neo1aδ1133-FLAG (340.0 ± 18.53 μm, n = 39 embryos, p < 0.0001, Tukey test), rgmaex2+3: (207.9, n = 22 embryos, p < 0.0001, Tukey test), and neo1aex3+8/neo1bex3+8 (213.3 ± 21.51 μm, n = 22 embryos, p = 0.0272, Tukey test) were significantly wider than those of uninjected (148.0 ± 1.413 μm, n = 60 embryos), and std MO-injected (145 ± 3.641 μm, n = 9 embryos, Fig. 6C). Together, these findings suggest that Rgma and Neo1a mediate NP infolding by regulating NEC elongation.
Rgma and Neo1a depletion causes duplication of the NT
In the zebrafish neural rod, daughter cells of NECs integrate into opposite sides of the NT, a mode of cell division known as midline-crossing division (C division). This mode of division is a mechanism used in this organism to establish apico-basal polarity, as the polarity protein Pard3 localizes to the cleavage furrow of mitotic progenitors cells and is mirror-symmetrically inherited by both daughter cells, marking their apical pole (which coincides with the apical midline of the NT) (Tawk et al., 2007). When convergence and/or infolding is delayed in planar cell polarity mutants, C divisions occur away from the midline of the widened neural tissue (neural rod stage), preventing midline-crossing and resulting in a duplication of the apical midline of the NT (Tawk et al., 2007; Clarke, 2009).
To test whether the widened neural keel in embryos with impaired Rgma-Neo1 signaling alters the frequency, distribution, or orientation of C divisions, we labeled 8–10 som embryos with the nuclear marker DAPI and anti-β-tubulin and examined mitotic progenitors in rgma MO2-injected embryos and uninjected controls, at the level of the hindbrain (Fig. 7A,B). We observed that Rgma knockdown did not significantly impact the mitotic index in the neural rod (control mean mitotic index: uninjected controls: 5.21 ± 0.99%, n = 4 embryos; rgma MO2-injected embryos: 2.91 ± 0.63%, n = 6 embryos, p = 0.0740, t = 2.054, df = 8, unpaired two-tailed t test). However, Rgma depletion biased the position of these divisions away from the midline of the neural rod compared with controls (mean percentage of medial cell divisions in control embryos: 100 ± 0 μm, n = 20 nuclei from 4 embryos; mean percentage of medial cell divisions in rgma MO2-injected embryos: 12.08 ± 9.8 μm, n = 26 nuclei from 6 embryos, p < 0.0001, t = 7.178, df = 8, unpaired two-tailed t test). Furthermore, the angle of cell division in Rgma-depleted embryos was increased relative to controls (mean angle of division of uninjected controls: 22.43 ± 4.77°, n = 20 nuclei from 4 embryos; mean angle of division of rgma MO2-injected embryos: 40.97 ± 4.44°, n = 26 nuclei from 6 embryos, p = 0.0071) (Fig. 7C). Interestingly, whereas the majority of cell divisions in uninjected control embryos occurred at ≤15° (52% of divisions in n = 20 nuclei from 4 control embryos), fewer cell divisions in Rgma-depleted embryos occurred at these angles (15.17% of divisions in n = 26 nuclei from 6 rgma MO2-injected embryos, p = 0.0091, F(5,40) = 3.578, two-way ANOVA) (Fig. 7C). These findings signify that C divisions occur at a normal frequency in Rgma-depleted embryos but are positioned away from the midline and their angle of division is oblique to the mediolateral axis of the neural tissue.
To assess whether abnormal C divisions in embryos defective for Rgma signaling contribute to NT defects (NTDs), we imaged the organization of the NT in hindbrain sections labeled with phalloidin, a marker for filamentous actin (F-actin) that is enriched at the cell cortex and apical pole of NECs (Fig. 8). The 1 dpf control embryos (uninjected) (Fig. 8A) and std MO-injected embryos (data not shown) had a normally shaped NT with a single lumen lined with phalloidin. In contrast, two apical midlines and lumens were observed in severely afflicted embryos depleted of Rgma (rgma MO2-injected and rgmaex2–3) (Fig. 8B,C) or Neo1a/b (embryos injected with neo1a MO2 or neo1aδ1133-FLAG) (Fig. 8D,E). These findings indicate that, as is observed with mutations in planar cell polarity genes, disruption of Rgma-Neo1a signaling can cause a duplication of the NT.
Rgma-Neo1a signaling is not required for establishing NEC apico-basal polarity
A previous study by Kee et al. (2008) revealed that depletion of Neo1 causes a dispersal of the tight junction marker zona occludens 1 (ZO1) in the dorsal aspect of the zebrafish neural rod, which led to the conclusion that Rgma-Neo signaling is required for the establishment of apico-basal polarity in NECs (Kee et al., 2008). However, in light of the duplication of the NT we observe in Rgma and Neo1a knockdown embryos (Fig. 8B–E), we re-examined this phenotype by analyzing ZO1 distribution in the neural rod (Fig. 8F,G) and NT (Fig. 8I) of rgma MO2-injected embryos and in the NT of neo1a-MO2 injected embryos (Fig. 8J). We observed that, while ZO1 puncta do indeed seem dispersed and disorganized in the dorsal neural rod of Rgma- and Neo1-depleted embryos, closer examination reveals that ZO1 label indeed forms a “V” shape in the neural rod (8–10 som) (Fig. 8G) that becomes more distinct by 1 dpf (Fig. 8I,J) and is indicative of a duplicated apical midline (Ciruna et al., 2006; Clarke, 2009; Compagnon and Heisenberg, 2013). Together, these observations indicate that Rgma-Neo1a signaling regulates NEC elongation independently of the establishment or maintenance of apical junction markers.
Neo1a and Rgma regulate MT organization during infolding
Previous studies from our laboratory and others have demonstrated that MTs and several MT-binding proteins are required for NEC elongation, neural fold elevation, and timely infolding of the NP (Karfunkel, 1971, 1972; Kee et al., 2008; Jayachandran et al., 2016). Furthermore, the Cooper laboratory showed that MTs are disorganized in Xenopus embryos depleted of Neo1 (Kee et al., 2008). To test whether Rgma and Neo1 are also required for proper MT organization in zebrafish, we labeled rgma MO2- and neo1a MO2-injected embryos with anti-β-tubulin at the neural keel stage (4–5 som) (Fig. 9). Interestingly, MTs in these embryos appear shortened and clustered around the nucleus (Fig. 9B–C′) compared with the linear arrays observed in control (uninjected) embryos (Fig. 9A,A′), suggesting that Rgma-Neo1 signaling promotes NEC elongation in an MT-dependent manner.
Full-length Neo1 is consecutively cleaved by α- and γ-secretase
To gain insight into the molecular events activated downstream of Rgma-Neo1a interaction that promote NEC elongation and infolding, we next investigated whether zebrafish Neo1 undergoes proteolytic cleavage. A growing list of single-pass (Type I) transmembrane receptors have been shown to undergo regulated intramembrane proteolysis. This process cleaves single-pass (Type I) transmembrane receptors twice: first at the cell surface, which releases the ectodomain into the extracellular space; and then within the plasma membrane, which solubilizes the ICD (Brown et al., 2000; Marambaud et al., 2002; Gudey et al., 2014; Almagro-Moreno et al., 2015). Consistent with RIP, α-secretase (ADAM17)-mediated ectodomain shedding of Neo1 was previously demonstrated in migrating neurons and regenerating axons (Goldschneider et al., 2008; Okamura et al., 2011; van Erp et al., 2015). Neo1 is also known to be cleaved by γ-secretase, an event that releases an intracellular Neo1 domain (NeoICD) (Goldschneider et al., 2008; van Erp et al., 2015). However, proteolysis of Neo1 during early embryonic development has not been reported.
To test whether zebrafish Neo1 is a sheddase substrate, we analyzed protein products of Neo1 in 1 dpf embryos (when Neo1 is abundant), in the presence of TAPI-1 (100 μm), an inhibitor of ADAM17-like metalloproteases, or vehicle. The proteasome inhibitor MG132 and the nuclear export inhibitor LMB were added to extracts to stabilize short-lived targets of proteasome degradation (Schulte et al., 2007) and retain ADAM17-like proteolytic products in the nucleus, respectively (Goldschneider et al., 2008; Maetzel et al., 2009; Bouillez et al., 2014). Drug treatments were administered from 6 som (neural keel) to 20 som, when protein extracts were harvested. Western blots were performed using anti-Neo1 that does not distinguish between Neo1a and Neo1b. In presence of vehicle only, we observed Neo1 fragments of three sizes, 225, 55, and 40 kDa (Fig. 10A,B). Treatment with TAPI-1 (α-secretase inhibitor) depleted the 55 kDa band and increased the amount of the 225 kDa band (Fig. 10B), suggesting that the 225 kDa protein is a substrate for ADAM17-like metalloproteases. Since ADAM17 proteolytic activity occurs at the juxtamembrane region of its substrates on the outer leaflet of the plasma membrane, we interpret the 55 kDa band to represent a membrane-embedded cleavage product of Neo1 and the 225 kDa fragment to be full-length Neo1. Our results are consistent with a model in which the Neo1 ectodomain is shed by an α-secretase at the expense of future ligand binding. The 40 kDa band was not altered by TAPI-1 treatment, possibly as a result of its embryonic production before drug treatment and sequestration in the nucleus away from proteases (in the presence of LMB).
We next tested whether the putative 55 kDa α-secretase/ADAM17-like cleavage product is a substrate for γ-secretase, an intramembrane protease complex. We performed Western blot analysis on lysates derived from 20 som embryos that were treated with DAPT, a γ-secretase inhibitor, from 6 som to 20 som, in the absence of LMB and MG132 (Fig. 10B) (Jiang et al., 2011; Kalantari et al., 2013). We observed that the relative abundance of the 55 kDa band in the DAPT-treated sample was higher than in the control. The 40 and 225 kDa bands of Neo1 were not resolved under these conditions. These results concur with others suggesting that the 55 kDa α-secretase cleavage product of Neo1 is further processed by γ-secretase-mediated proteolysis.
Neo1 is constitutively cleaved during early development
To address if and when Neo1 cleavage is developmentally regulated, we performed Western blot analysis using protein extract from embryos at stages relevant to the process of NP infolding, in the absence of MG132 and LMB. We observed an abundant 55 kDa band at the tailbud, 4–5 som, and 6 som stages (Fig. 10C), which coincide with dynamic morphogenetic movements in all germ layers. The low abundance of the 225 and 40 kDa fragments during early developmental stages suggests that the Neo1 ectodomain is constitutively shed during early development and that the γ-secretase cleavage product, NeoICD, is highly unstable (Goldschneider et al., 2008). In contrast, the 40 kDa cleavage product was highly abundant in 5 dpf embryos, even in the absence of LMB (Fig. 10C), suggesting the existence of endogenous mechanisms that stabilize this cleavage product.
Rgma stimulates Neo1 cleavage
The similarity of the Rgma and Neo1 LOF phenotypes (Figs. 2, 3, 6, 8) is consistent with these proteins functioning as a ligand-receptor pair during NT development, as reported in other contexts (Mawdsley et al., 2004; Wilson and Key, 2006; Kee et al., 2008). To test whether Neo1 undergoes RIP in response to Rgma binding, we examined Neo1 proteolytic cleavage in 1 dpf Rgma knockdown embryos using Western blot analysis (Fig. 10D–F). std MO-injected embryos (10 ng) produced the 55 and 40 kDa Neo1 fragments that are consistent with α- and γ-secretase processing, respectively (Fig. 10D, left lane). In contrast, depletion of Rgma with MO2 (5 and 10 ng) caused a concentration-dependent reduction in the amount of the 40 and 55 kDa fragments and a corresponding increase in the full-length product range (225 kDa) (Fig. 10D, middle and right lanes, respectively). A similar reduction in the 40 and 55 kDa Neo1 fragments is also observed in rgmaex2+3 mutants (Fig. 10E, right lane) compared with uninjected controls (Fig. 10E, left lane). These results suggest that Rgma dose-dependently regulates the bioavailability of all three specific Neo1 products, which is consistent with previous observations in cultured cells (Goldschneider et al., 2008; Banerjee et al., 2016). Because the 225 kDa Neo1 product is low-abundant at 1 dpf, we examined an overnight exposure of the blot in Figure 10D and observed this band in all lanes (Fig. 10F, top arrowhead); however, its abundance is much greater in rgma MO2-treated samples (Fig. 10F). Interestingly, the overexposed blot also revealed the accumulation of a novel Neo1 band of ∼200 kDa that intensified as endogenous Rgma levels decreased (Fig. 10F, bottom arrowhead). This size differential could potentially reflect an additional cleavage event occurring in the absence of Rgma that produces a Neo1 peptide lacking only the immunoglobulin repeats (Tassew et al., 2014); however, we favor a post-translational modification for the reasons mentioned below.
Neo1 is N-glycosylated during early development
Although zebrafish Neo1a and b paralogs are both estimated to be ∼150 kDa in mass based on amino acid sequence, they are often observed at a size >200 kDa in reducing and denaturing conditions, most likely due to post-translational modifications (PTMs). Bioinformatics approaches have predicted that Neo1 orthologs in mice, rats, and humans each have 4 disulfide bonds, 6 phosphoserine sites, 2 phosphothreonine sites, and 8 asparagine-linked glycosylation sites. Of these four PTMs, only N-glycosylation is expected to shift the molecular weight of zebrafish Neo1 sufficiently to account for the 225 and ∼200 kDa band sizes (Fig. 10F). This PTM has been linked to the regulation of protein stability, adhesion, and signaling in normal and diseased cells, making it an intriguing candidate for modulating Neo1 activity during morphogenesis (Eklund and Freeze, 2006; Freeze et al., 2012; Scott and Panin, 2014; Pinho and Reis, 2015; Stowell et al., 2015; Mallard and Tiralongo, 2017); however, in vivo experimental evidence of this Neo1 PTM is limited.
The tripeptide sequence recognized by N-glycosyltransferases, Asn-X-Ser/Th (where X is any residue except proline) is conserved in zebrafish Neo1 (Cooper et al., 2001). To test whether zebrafish Neo1 is N-glycosylated, we treated 1 dpf embryos with recombinant peptide-N-glycosidase F (PNGaseF), which removes oligosaccharides from asparagine (Asn) residues on glycoproteins (Tarentino et al., 1985) and examined Western blots of lysates. Buffer-treated lysates revealed a 225 kDa Neo1 peptide (Fig. 10G, left lane), whereas PNGaseF-treated lysates exhibited a downward shift in protein size (200 kDa, Fig. 10G, right lane), suggesting that the full-length peptide is indeed glycosylated and likely to at least partially account for the Neo1 225 kDa band size.
To directly test whether Rgma regulates Neo1 glycosylation, we performed an in-gel Fuchsin assay of periodic acid Schiff-oxidized glycoproteins using protein extracts from 1 dpf embryos (Fig. 10H) (Roth et al., 2012). We found that Fuchsin labeling of a ∼225 kDa protein was reduced in rgma MO2 (5 ng)-injected embryos compared with controls (Fig. 10H). We next probed replicate lanes with anti-Neo1 via Western blot analysis and observed Neo1 immunoreactivity that coincided with the mass of the Fuchsin-positive band. As expected, this ∼225 kDa Neo1 band was increased in the Rgma-depleted sample (Fig. 10H). These data suggest that Neo1 glycosylation is decreased in Rgma-depleted embryos.
NeoICD localizes to the nucleus
The 40 kDa NeoICD is known to solubilize in an Rgma-dependent manner (Goldschneider et al., 2008; van Erp et al., 2015; Banerjee et al., 2016), but its function during embryogenesis is unknown. The conserved N terminus of the NeoICD includes a nuclear localization signal (Goldschneider et al., 2008), which suggests that this fragment may translocate to the nucleus in the zebrafish embryo. To assess NeoICD distribution, we performed subcellular fractionation on 1 dpf larvae to isolate and purify the nuclear pellet. We found that the 55 kDa Neo1 fragment was excluded from the nuclear fraction (Fig. 10I, left lane) but present in the supernatant (Fig. 10I, right lane). Instead, the purified nuclear fraction was enriched for the 40 kDa NeoICD (Fig. 10I, left lane). Our results confirm those from studies using human cell culture and chick optic tectum explants, showing that γ-secretase-mediated Neo1 proteolysis produces a nuclear-localized soluble cytosolic peptide (Goldschneider et al., 2008; Banerjee et al., 2016). Although the NeoICD was observed at larval stages (1 and 5 dpf), it was undetectable in early embryos in the absence of LMB and MG132 (Fig. 10C), consistent with its previously reported rapid export from the nucleus and degradation (Goldschneider et al., 2008).
NeoICD partially rescues impaired infolding in Neo1a and Rgma-depleted embryos
To test the requirement of the NeoICD during infolding, we performed a rescue experiment by injecting 1 cell stage embryos with neo1a MO2 and rgma MO2 in the presence or absence of 50 pg of neo1a ICD mRNA (neoICD) or neoICD mutant construct (neoICDmut) (Fig. 10J) (F(7,77) = 64.27, p < 0.0001, ANOVA). Embryos injected with neo1a MO2 exhibited an abnormally widened NP (414.70 ± 28.81 μm, n = 8 embryos) compared with uninjected controls (123.50 ± 3.60 μm, n = 17 embryos, p < 0.0001, Tukey test) or neoICD-injected embryos (151.80 ± 7.89, n = 17 embryos, p < 0.0001, Tukey test). However, when neo1a MO2 was coinjected with neoICD, a partial rescue of the NP width (a decrease of 92.07 ± 20.62 μm, mean difference ± SEM difference) was observed in these embryos compared with those injected with neo1a MO2 alone (322.60 ± 15.68 μm, n = 9, p = 0.0003, Tukey test).
We next investigated whether NeoICD is sufficient to rescue Rgma LOF. If Rgma-dependent Neo cleavage is important to promote infolding, then NeoICD should rescue the morphological defects in embryos depleted of Rgma. Embryos coinjected with rgma MO2 and neoICD mRNA (50 pg) exhibited a narrower NP (274.60 ± 20.08 μm, n = 5, p = 0.0121, Tukey test) than embryos injected with rgma MO2 alone (368.90 ± 17.83 μm, n = 10, p = 0.0121, Tukey test), further supporting the model that Rgma-induced Neo cleavage is required for NT development.
To test whether the translocation of NeoICD into the nucleus is necessary for its function, we generated a zebrafish neoICD mutant construct, neoICDmut, carrying arginine-to-alanine substitutions in the nuclear localization sequence that prevent entry into the nucleus, as reported by Goldschneider et al. (2008). We further found that, in contrast to NeoICD, this mutant form failed to rescue Neo1-depleted embryos (390.90 ± 21.07 μm, n = 9, p < 0.0001, Tukey test) (Fig. 10J), suggesting that this peptide is required for transcriptional regulation. Although the NP of embryos injected with neoICDmut alone (231.40 ± 15.58 μm, n = 10, p = 0.0018, Tukey test) was significantly narrower than that of neo1a MO2-injected embryos, it was wider than that of uninjected embryos, potentially due to the accumulation of NeoICDmut in the cell causing toxicity. These results suggest a requirement for the NeoICD to promote the process of infolding, as this peptide partially overcomes the effects of Neo1 and Rgma depletion. The NeoICD has previously been shown to regulate gene transcription, and it is possible that it promotes the expression of genes implicated in cell elongation during NT morphogenesis.
Discussion
Rgma and Neo1 promote NEC elongation during infolding
Shaping of the NT is dependent on the ability of NECs to elongate; however, the molecular mechanisms underlying this cell behavior are poorly understood. In this study, we provide evidence that NEC elongation is mediated by Rgma-Neo1a signaling. Using MOs, CRISPR/Cas9 genome-editing tools and DN-Neo1 (neo1aδ1133−), we show that disruption of Rgma or Neo1 causes NECs to adopt an abnormally rounded shape. This defect is, however, transient as cells do eventually elongate (Fig. 8), suggesting the existence of compensatory mechanisms.
We further show that depletion of Rgma or Neo1a impairs infolding, resulting in an abnormally wide NP. Since our cellular analyses were performed at the neural rod stage (when infolding is complete in control embryos), we cannot rule out that the earlier stage of NP narrowing (convergence) is not defective in these embryos. Nevertheless, both processes are dependent on cell elongation and support a central role for Rgma/Neo1 signaling in promoting this cell behavior.
Consistent with their role in shaping the NT, both rgma and neo1 are expressed in the developing NT: rgma more strongly in the midline/floorplate (although also present in the medial aspect of the NT) and neo1 weakly throughout the neural tissue at the NP and neural keel stages. Although the expression of neo1a and neo1b overlap in the NP and keel, depletion of Neo1a alone causes NTDs, arguing against functional redundancy between these paralogs. Furthermore, while neo1a transcript is also highly expressed in the paraxial mesoderm (accounting for the axial elongation defects we observe), transplantation of Neo1a-depleted cells into WT hosts revealed a cell autonomous role for Neo1 in NEC elongation. Together, these data are consistent with a model whereby Rgma, secreted from the floorplate and/or notochord, binds to Neo1-expressing NECs, thereby promoting their elongation. However, it is also possible that Rgma-Neo1 interaction occurs in a cis configuration.
Rgma-Neo1 signaling promotes NEC elongation independently of the establishment of apical junctional complexes
Disruption of convergence and/or infolding of the NP has dramatic and well-documented consequences on later aspects of NT development in zebrafish that are primarily linked to ectopic positioning of C divisions. We also observe striking NTDs in Rgma- and Neo1-depleted embryos, as evidenced by ectopic (tissue-level) distribution of apical markers ZO1 and phalloidin. In a previous study, the broad distribution of ZO1 throughout the dorsal aspect of the NT was interpreted as a failure of Neo1-depleted cells to establish apico-basal polarity (Kee et al., 2008). However, we propose that ectopic positioning of the apical midline is a more likely explanation and that Rgma-Neo1 signaling controls NEC elongation independently from the establishment of apical junctional markers. The use of zebrafish in this study was instrumental in making this distinction, as cell elongation and epithelialization are temporally uncoupled in this organism (Hong and Brewster, 2006). It will be interesting in the future to determine whether Rgma-Neo1 signaling and the planar cell polarity pathway converge on common cellular targets to mediate cell elongation.
MT organization is disrupted downstream of Rgma and Neo1 signaling
In zebrafish, Xenopus, and mice, the MT network undergoes a rapid and dramatic change in organization, which drives cell elongation in the NP. Interphase MTs are initially broadly distributed throughout the cell and randomly oriented. However, they polymerize rapidly, assembling parallel to the long (apico-basal) axis of NECs (Karfunkel, 1971, 1972; Hong and Brewster, 2006; Lee et al., 2007; Cearns et al., 2016). Consistent with a requirement for MTs in NEC elongation, treatment with drugs that either destabilize (vinblastine, colchicine, and nocodazole) or hyperstabilize (paclitaxel) MTs prevents the induction or maintenance of this shape change (Karfunkel, 1971, 1972; Kee et al., 2008; Jayachandran et al., 2016). Morphogenetic movements that are MT-dependent are also perturbed in these drug-treated embryos, including narrowing of the NP and elevation of the neural folds (Karfunkel, 1971, 1972; Kee et al., 2008; Jayachandran et al., 2016). Thus, MT organization is important for both the establishment and maintenance of cell elongation and proper completion of NT morphogenesis. A possible mechanism via which MTs promote cell elongation may be the transport of cytoplasmic materials toward the extending ends of the cell (Burnside, 1971).
MT stability and organization are regulated by binding of MT-associated proteins (Hirokawa, 1994). Post-translational modifications of MT-associated proteins, such as phosphorylation, are thought to modulate their binding affinity to MTs and in turn alter MT stability (Díaz-Nido et al., 1988; Avila et al., 1994; Gordon-Weeks and Fischer, 2000). Deleted in colorectal cancer, a Netrin receptor closely related of Neo1 triggers phosphorylation of the neural-specific MT-binding protein Map1b upon binding to its ligand (Del Río et al., 2004), raising the interesting possibility that this MT-associated protein may also be activated or transcribed downstream of Neo1 signaling. In support of this model, we previously reported phenotypes in Map1b knockdown embryos that are very similar to those observed in Neo1a- and Rgma-depleted embryos, namely, a widened NP, fully or partially duplicated NTs, failure of NECs to elongate and shortened/disorganized MTs (Jayachandran et al., 2016).
Role of NeoICD and regulation by Rgma
Using Western blot analysis, we have demonstrated that Rgma stimulates the sequential proteolytic cleavage of zebrafish Neo1 by an ADAM17-like metalloprotease (van Erp et al., 2015) and γ-secretase to produce a soluble intracellular fragment, NeoICD. Given the low abundance of full-length Neo1 throughout early stages of development, we surmise that Rgma induces constitutive Neo1 ectodomain shedding. These observations are consistent with previous reports revealing that Rgma stimulates ADAM17-mediated Neo1 ectodomain shedding in differentiated neurons and in the mouse optic tectum (van Erp et al., 2015; Banerjee et al., 2016). Moreover, our subcellular fractionation assay revealed the presence of the zebrafish 40 kDa fragment in the nuclear fraction, similar to a previous report (Goldschneider et al., 2008). This soluble domain is unstable in early-stage embryos and best observed in the presence of proteasome and nuclear export inhibitors, suggesting that NeoICD activity and degradation are tightly regulated.
NeoICD is known to regulate the transcription of several promigratory genes, including cerebral cavernous malformations 2 (CCM2) and grainyhead-like 1 (GRHL1) in human embryonic kidney 293T (HEK293T) and Michigan cancer foundation 7 (MCF-7) human breast cancer cells (Goldschneider et al., 2008). Although it remains unclear whether NeoICD is implicated in the regulation of gene expression in the developing NT, the ability of exogenous NeoICD to partially rescue infolding defects in Neo1 and Rgma knockdown embryos lends support to the model that proteolytic cleavage of this cell surface receptor is important to promote NT morphogenesis. Of particular relevance is the finding that Rgma induces the release of the transcriptional coactivator LIM domain only 4 (LMO4) from the cytoplasmic tail of Neo1. After nuclear entry, LMO4 is proposed to modulate the transcriptional activity of mediators of neuronal outgrowth (Schaffar et al., 2008). It is currently unclear whether Rgma-Neo1-dependent transcriptional responses originate from NeoICD or LMO4, which rely on an endogenous nuclear localization sequence domain or bind to nuclear LIM interactor to translocate into the nucleus, respectively (Kenny et al., 1998). These observations provide a rationale for future work to identify targets downstream of the Rgma-Neo1 interaction that shape the NT, via transcriptional regulation or otherwise (Babitt et al., 2005; Hata et al., 2006; Conrad et al., 2007; Zhou et al., 2010; Tian and Liu, 2013).
N-glycosylation of NeoICD
The N-glycosylation events that we report for full-length Neo1 provide more evidence that the Neo1 biosynthetic pathway is conserved in zebrafish (Meyerhardt et al., 1997; Vielmetter et al., 1997; Kleene and Schachner, 2004; Liu et al., 2005; Wollscheid et al., 2009). Interestingly, Rgma depletion reduces the molecular mass of Neo1 from 225 to 200 kDa, which is consistent with Neo1 deglycosylation. These data suggest that Rgma promotes the retention of this PTM once Neo1 is inserted in the plasma membrane, although other models are possible. The functional significance of Neo1 glycosylation is unclear, but it is tempting to speculate that it could be linked to the regulation of its proteolytic cleavage.
Overall, this study reveals that Rgma-Neo1 signaling regulates cell elongation during neurulation independently of the establishment of apical junctional complexes and identifies Neo1 proteolysis as an early signaling event downstream of ligand–receptor interactions in the developing embryo. The proposed function of Rgma/Neo1 signaling is likely to be conserved in vertebrates, as rgma LOF in Xenopus and mouse embryos causes severe NTDs (Niederkofler et al., 2004; Kee et al., 2008). Future work identifying the transcriptional targets of the NeoICD is likely to provide novel insights into the molecular mechanisms that regulate NEC elongation during neurulation. It would also be of interest to determine whether the other Neo1 cleavage products, the shed ectodomain and the membrane-embedded 55 kDa fragment, play a regulatory role during neural development.
Footnotes
This work was supported by National Institutes of Health/National Institute of General Medical Sciences Grant GM085290-02S1 to V.O. and Grant GM085290 to R.B. S.B. was supported by a U.S. Department of Education GAANN Fellowship, Grant P200A120017, and National Institutes of Health/National Institute of General Medical Sciences Grant GM055036 (Meyerhoff Graduate Fellowship). M.N. was supported by National Institute of General Medical Sciences/National Institutes of Health T32 GM066706, National Science Foundation 1500511, and National Institutes of Health/National Institute of General Medical Sciences GM055036. E.V. was supported by Howard Hughes Medical Institute (Precollege and Undergraduate Science Education Program) Grant 52008090 to University of Maryland Baltimore County. The SP5 confocal microscope (Leica Microsystems) was purchased with funds from the National Science Foundation Grant DBI-0722569. We thank Robyn Goodman for cell transplantation experiments; and Neus Sanchez-Alberola, Julie Wolf, Yiannis Balanos, and Nilusha Jayasinghe for contributions to CRISPR/Cas9 tools.
The authors declare no competing financial interests.
- Correspondence should be addressed to Rachel Brewster at brewster{at}umbc.edu.