A zebrafish model of lethal congenital contracture syndrome 1 reveals Gle1 function in spinal neural precursor survival and motor axon arborization

Arthrogryposis multiplex congenita (AMC) is a heterogeneous group of human disorders with an estimated worldwide incidence of 1 in 3000 births (Hall, 1985). It is, however, more common in genetically isolated groups such as the Finnish and Israeli Bedouin populations (Lebenthal et al., 1970; Pakkasjarvi et al., 2006). To date, over 150 AMC conditions varying in etiology, pathogenesis and other organ anomalies have been described with multiple deformities of the joints (contractures) at birth being the predominant shared symptom (Hall, 1997). Among these AMC conditions, lethal congenital contracture syndrome 1 (LCCS1) is a distinct entity and one of the most severe forms manifested in utero, with prenatal death of all affected fetuses by week 32 of gestation (Herva et al., 1985; Makela-Bengs et al., 1998). LCCS1 is an autosomal recessive disorder that was originally reported in ten families from northeastern Finland (Herva et al., 1985). Affected LCCS1 fetuses show joint contractures, early loss of voluntary muscle movement, small jaws (micrognathia), incomplete lung development and muscular atrophy (Herva et al., 1985). Of note, LCCS1 fetuses exhibit a distinct neuropathology with marked loss of the ventral spinal cord, both in the anterior-horn motoneurons and the ventral and lateral funiculi (with less alteration of the sensory nuclei and the dorsal funiculus) (Herva et al., 1988). These observations have led to speculation that the LCCS1 pathology is due to defects in motoneuron development and maturation.

Recently, a genetic study linked GLE1 mutations to LCCS1 (Nousiainen et al., 2008). The major mutation, Fin_Major, is a single-nucleotide substitution that generates an illegitimate splice acceptor site in intron 3, producing a three amino acid insertion in the human GLE1 protein. Gle1 was first identified as an essential mRNA export factor in both S. cerevisiae and human cells (Kendirgi et al., 2003; Murphy and Wente, 1996; Watkins et al., 1998). In addition to mRNA export, we recently discovered a novel role for Gle1 in mediating efficient translation initiation and termination (Bolger et al., 2008). This work, together with substantial biochemical and cell biological analysis of Gle1 function (reviewed by Folkmann et al., 2011), reveals that Gle1 is a key regulator of multiple post-transcriptional gene expression steps. To date, no studies of the role of Gle1 in vertebrate development have been reported. Moreover, a specific role for Gle1 in motoneuron development has not been documented.

Here we investigate the link between Gle1 function and LCCS1 pathology using zebrafish (Danio rerio) as a model system. Using a gle1 insertional mutant line and an antisense morpholino knockdown strategy, we show that disrupting Gle1 function leads to morphological anomalies that parallel key phenotypes of LCCS1 human fetuses. Gle1 depletion also results in reduced spinal cord motoneurons, and Gle1 activity is required extrinsically in the environment for motor axon arborization. Surprisingly, motoneuron reduction is caused by apoptosis of neural precursors, rather than acute motoneuron death. This directly contrasts with the previously proposed LCCS1 pathophysiological mechanism (Nousiainen et al., 2008). Based on this first functional characterization of Gle1 in vertebrate development, we propose that rapidly dividing cells, including the organ precursors in both neuronal and non-neuronal tissues, have a high demand for Gle1 activity as they undergo proliferation. As a result of Gle1 deficiency, the death of proliferative organ precursors during early embryonic development is likely to be the common underlying cellular mechanism that leads to the pleiotropic abnormalities in affected LCCS1 fetuses.

Wild-type AB and gle1^hi4161a zebrafish (Danio rerio) strains were bred and maintained using standard procedures (Westerfield, 2000). Embryos were obtained by natural spawning and staged according to Kimmel et al. (Kimmel et al., 1995). We used transgenic line Tg(gfap:GFP) (Bernardos and Raymond, 2006) and generated lines Tg(mnx1:TagRFP-T) and Tg(mnx1:H2A-TagRFP-T) (see below). In some cases, 0.003% (w/v) 1-phenyl-2-thiourea (PTU) was added to the medium to block trunk pigmentation. Wild-type, heterozygous and homozygous gle1^hi4161a embryos were genotyped using a three-primer PCR strategy (Wang et al., 2007) (see supplementary material Table S1 for primers).

To clone zebrafish gle1, total RNA was isolated from wild-type AB embryos (shield stage) using TRIzol (Invitrogen, Carlsbad, CA, USA). cDNA was made from 2 μg total RNA by reverse transcription using an oligo(dT)₁₆ primer (Roche Diagnostics, Indianapolis, IN, USA) and SuperScript III reverse transcriptase (Invitrogen). The coding sequences for zebrafish gle1a and gle1b were amplified from cDNA using Phusion high-fidelity DNA polymerase (New England Biolabs, Ipswich, MA, USA) with specific primers (supplementary material Table S1) for insertion into pDONR221 using Gateway BP Clonase II enzyme mix (Invitrogen) to generate pME-drgle1a and pME-drgle1b.

Zebrafish mnx1 upstream elements were PCR amplified from purified wild-type zebrafish genomic DNA using Phusion high-fidelity DNA polymerase with specific primers (supplementary material Table S1) and inserted into pDONRP4-P1R using Gateway BP Clonase II enzyme mix to generate p5E-mnx1. p5E-mnx1 contains the 5136 bp upstream elements of mnx1, including the non-coding sequence of the first exon.

The plasmid pCMV/SP6-hsGLE1b with human GLE1b driven by the CMV/SP6 promoter was generated by MultiSite Gateway cloning (Kwan et al., 2007) using the PCR-amplified human GLE1b sequence (Kendirgi et al., 2005). For the Fin_Major mutation, we performed standard PCR-based site-directed mutagenesis (Braman et al., 1996) using plasmid pCMV/SP6-hsGLE1b to insert CCTTTTCAG between positions 432 and 433 of the coding DNA.

In situ hybridizations were performed as described (Thisse and Thisse, 2008). Digoxygenin-labeled gle1 antisense RNA probes (Roche Diagnostics) were prepared from DraI-linearized zebrafish gle1 plasmid using SP6 RNA polymerase. Embryos were cleared in serial incubations of glycerol (25, 50 and 75%) before imaging.

Whole-mount immunohistochemistry was performed as described (Sarmah et al., 2005). Indirect immunofluorescence was performed using the following antisera: anti-Islet1/2 [39.4D5, Developmental Studies Hybridoma Bank (DSHB), University of Iowa; 1:200]; anti-zn-8 (DSHB; 1:200); anti-Elavl (16A11, Invitrogen; 1:250); anti-cleaved caspase 3 (Abcam, Cambridge, MA, USA; 1:500); and anti-Sox2 (Abcam; 1:500). Alexa Fluor-conjugated secondary antibodies (Invitrogen) were used.

For cryosections, fixed embryos were embedded in 1.5% agarose/5% sucrose, equilibrated in 30% sucrose and frozen in a 2-methyl-butane/liquid nitrogen bath. Blocks were sectioned (10 μm) using a Leica CM3050 cryostat. Antibody staining of cryosections was performed as for the whole-mount procedure with no detergent added. Stained sections were mounted in mounting medium (Vector Laboratories, Burlingame, CA, USA) before imaging. Sometimes, nuclei were counterstained with 1 μM TO-PRO-3 dye (Invitrogen).

For histology sections, embryos were fixed in Bouin’s solution (Sigma-Aldrich, St Louis, MO, USA) and embedded in JB-4 plastic resin polymer (Polysciences, Warrington, PA, USA). Blocks were sectioned (7 μm) using a Leica RM2265 microtome. Sections were dried on a heating plate and stained with Hematoxylin and Eosin (Sigma-Aldrich).

Larvae at 5 dpf were fixed in 4% paraformaldehyde in PBS at 4°C overnight, washed in PBS, and bleached in 10% H₂O₂/0.5% KOH for 1 hour at room temperature (RT). Bleached embryos were washed twice in PBS and incubated in Alcian Blue solution (0.1% Alcian Blue, 70% ethanol, 1% HCl) overnight at RT. Embryos were then washed once in acidic ethanol (70% ethanol, 5% HCl) and cleared in acidic ethanol overnight. Embryos were dehydrated through increasing ethanol concentrations (85 to 100%), transferred to 80% glycerol and stored at 4°C.

TUNEL assays were performed using the In Situ Cell Death Detection Kit (Roche Diagnostics) following the manufacturer’s instructions with the modification that the TUNEL reaction (1 hour at 37°C) was stopped by incubation in 300 mM sodium chloride/30 mM sodium citrate for 15 minutes at RT. Acridine Orange staining was performed by incubating live embryos in 2 μg/ml Acridine Orange (Sigma-Aldrich) in E3 medium (5 mM NaCl, 0.17 mM KCl, 0.4 mM CaCl₂, 0.16 mM MgSO₄) for 30 minutes at RT. Embryos were washed with E3 medium twice for 10 minutes each before imaging.

One-cell stage embryos from gle1^hi4161a/+ intercrosses were injected with a mix of two non-overlapping morpholino antisense oligonucleotides (MOs) designed to block the translation of gle1: drgle1ATG1, 5′-TCAGAAGGCATTTTGAGGGCATTAA-3′; and drgle1UTR1, 5′-ACACCTTTAGCAGCCCAAACAAGCC-3′; 1.5 ng each MO per embryo (Gene Tools). Five-nucleotide-mismatched MOs were used as controls. Pipettes were pulled on a micropipette puller (Model P-2000, Sutter Instruments, Novato, CA, USA). Injections were performed with an air-injection apparatus (Pneumatic PicoPump PV820, World Precision Instruments, Sarasota, FL, USA). The injected volume (typically ∼1 nl) was measured with a microruler.

In rescue assays, capped human GLE1b mRNAs were prepared using the mMessage mMachine RNA Synthesis Kit (Ambion, Carlsbad, CA, USA) following the manufacturer’s instructions and injected into one-cell stage embryos, which were first injected with gle1 MOs. For each MO knockdown and rescue experiment, embryos from the same clutch were used as experimental subjects and controls. Approximately 140 pg of capped RNA was injected per embryo unless stated otherwise.

To label motoneurons, we used Tol2-mediated transgenesis (Kawakami, 2004; Kwan et al., 2007) to generate lines in which TagRFP-T or a histone variant H2AFV-fused TagRFP-T (H2A-TagRFP-T) was expressed from zebrafish mnx1 upstream elements. Mixtures of 25 pg of the transgenesis plasmid and 25 pg transposase mRNA were injected into wild-type one-cell stage embryos. Injected founders were raised and progeny were screened for the respective transgenesis markers to establish Tg(mnx1:TagRFP-T) and Tg(mnx1:H2A-TagRFP-T) lines.

Wild-type and gle1 MO-injected embryos from gle1^hi4161a/+ intercrosses were used. Donor and host embryos contain mnx1:GFP and mnx1:TagRFP-T transgenes, respectively, to differentially label motoneurons in the resulting chimeras. At blastula stage, ∼40 cells were transplanted from donor embryos to host embryos with a pulled needle connected to an oil-filled syringe (CellTram Vario Microinjector, Eppendorf, Hauppauge, NY, USA). After transplantation, donor-host pairs were raised individually in 48-well plates. Hosts with GFP⁺ and TagRFP-T⁺ motoneurons were imaged at 3 dpf.

Embryos subjected to whole-mount in situ hybridization or Alcian Blue staining were mounted in a bridged slide containing one drop of 80% glycerol, coverslipped and imaged using a Zeiss Axio Imager Z1 microscope equipped with a Zeiss AxioCam MRm digital camera. JB-4 plastic sections were mounted in Cytoseal XYL mounting medium (Richard-Allan Scientific, Kalamazoo, MI, USA) and imaged as above. For live imaging, embryos were anesthetized with 0.03% Tricaine (Sigma-Aldrich) and mounted in 0.8% low-melting-point agarose (SeaPlaque GTG agarose, Lonza, Basel, Switzerland) on glass-bottomed 35-mm dishes (MatTek, Ashland, MA, USA). Live embryos and immunostained cryosections were imaged using 20× (0.7 NA) and 63× (1.4 NA) oil-immersion objectives, respectively, using a Leica TCS SP5 laser-scanning confocal microscope. Confocal images were processed using LAS AF (Leica) and ImageJ (NIH, USA) software. Single optical sections were exported as TIFF files. Brightness and contrast were adjusted (if necessary) using Adobe Photoshop CS5 in accordance with journal policy. Figures were composed with Adobe Illustrator CS5. Statistical analyses (e.g. unpaired t-test) were performed and graphed using Prism4 (GraphPad, La Jolla, CA, USA) software. Statistical significance was set at P<0.005.

Through Ensembl database searches, we identified gle1 in the zebrafish (Danio rerio) genome as the sole ortholog of human GLE1. By reverse transcriptase PCR (RT-PCR), we isolated coding sequences corresponding to both zebrafish Gle1 isoforms from the total RNA of wild-type embryos (supplementary material Fig. S1). Zebrafish gle1 and human GLE1 (Kendirgi et al., 2003) have an identical exon-intron organization and both are predicted to encode two protein isoforms from alternatively spliced transcripts. In addition, the predicted translated gene products of zebrafish gle1 share high degrees of similarity with the corresponding human GLE1 proteins (50% identity and 69% similarity). RT-PCR and in situ hybridization experiments showed that gle1 transcripts are maternally inherited and widely expressed from the one-cell stage through 5 days post-fertilization (dpf) (Fig. 1; data not shown). From 48 hours post-fertilization (hpf), gle1 was prominent in the CNS, eyes, pharyngeal arches, fin buds and digestive system (Fig. 1J,K).

To analyze the consequences of Gle1 loss of function, we used an insertional mutant line, gle1^hi4161a (Amsterdam et al., 2004), in which a provirus was integrated 34 bp downstream of the gle1 translation start codon, resulting in a non-functional gle1 allele (supplementary material Fig. S1). Homozygous gle1^{hi4161a/hi4161a} embryos (designated gle1^–/–) were initially indistinguishable from their wild-type or heterozygous siblings. However, a spectrum of morphological anomalies was observed at ∼48 hpf, including smaller head and eyes (Fig. 2B,D, arrowheads), underdeveloped pectoral fins (Fig. 2D, arrow) and cell death in the head and spinal cord (Fig. 2F, arrow and inset). Some also displayed upward or downward body axis curvatures, which were maintained or exacerbated with time (Fig. 2J). By 5 dpf, gle1^–/– larvae developed pronounced pleiotropic phenotypes, including small eyes (Fig. 2J; supplementary material Fig. S2A), pericardial and subcutaneous edema (Fig. 2J,L, asterisks), unfolded intestinal epithelia (Fig. 2L, arrowhead) and a distinct craniofacial defect. For the latter, the ventral viscerocranium, which consists of a series of seven pharyngeal arch cartilages (Schilling, 1997), was almost absent, whereas the dorsal neurocranium was still largely intact (compare Fig. 2H with 2G). This phenotype corresponds to the small jaw defect (micrognathia) of human LCCS1 patients.

In addition, gle1^–/– mutants also became increasingly immotile with time, displayed little spontaneous swimming and were insensitive to touch (data not shown). By measuring the size of spinal transverse sections, we further found that the spinal cord of 5-dpf gle1^–/– larvae was ∼73% of that of the wild type (Fig. 2K,L; supplementary material Fig. S2B). This spinal atrophy phenotype highly resembled the degeneration of spinal cord reported in LCCS1 (Herva et al., 1988).

Taken together, the variety of morphological phenotypes observed in zebrafish gle1^–/– embryos and early larvae, which include immobility, micrognathia-like craniofacial defects and diminished spinal cords, paralleled characteristic pathologies observed in LCCS1 human fetuses. Thus, we concluded that the zebrafish gle1^–/– mutant is an excellent model system for studying LCCS1.

Because loss of anterior-horn motoneurons and their descending tracts in the spinal cord is a hallmark of LCCS1 pathology (Herva et al., 1988), we examined motoneuron development in gle1^–/– embryos. To directly visualize zebrafish motoneurons, we crossed the gle1^hi4161a/+ line with Tg(mnx1:GFP) (formerly hb9:GFP) (Flanagan-Steet et al., 2005) or Tg(mnx1:TagRFP-T) (this study) fish, in which spinal motoneurons and their processes were robustly labeled with GFP or the red fluorescent protein TagRFP-T (Shaner et al., 2008), respectively. In these transgenic lines, motoneuron development was compared between wild-type and gle1^–/– backgrounds. At 1 dpf, motoneuron development in gle1^–/– embryos was indistinguishable from that in wild-type embryos. However, at 2 dpf, and similar to the timing of the onset of morphological defects, gle1^–/– embryos started to show motoneuron defects. Notably, gle1^–/– embryos had fewer spinal motoneurons than their wild-type siblings, as revealed by fewer GFP-labeled or TagRFP-T-labeled cells in the spinal cord (Fig. 3B, bracket). We also quantified the population size of motoneurons and compared it with that of Rohon-Beard sensory neurons by anti-Islet1/2 staining (Korzh et al., 1993). In agreement with the observation from mnx1:GFP/TagRFP-T-labeled gle1^–/– embryos, anti-Islet1/2 staining revealed that gle1^–/– embryos had ∼25% fewer motoneurons than their wild-type siblings (P<0.0001, unpaired t-test) at 2 dpf. However, there was no significant difference in Rohon-Beard sensory neuron number between gle1^–/– and wild-type embryos (P>0.09, unpaired t-test) (Table 1; supplementary material Fig. S3). These results indicate that at 2 dpf the population size of spinal motoneurons is selectively affected by reduced Gle1 activity.

In addition to fewer motoneurons, gle1^–/– embryos also exhibited defects in motor axon arborization at 2 dpf. Compared with wild type, within the trunk muscle the motor axon arbors in gle1^–/– embryos consistently lacked rostrally projecting nerves around the horizontal myoseptum (Fig. 3B,D). This phenotype was not due to developmental delay because the lack of rostral nerves never recovered as the embryos developed further. Moreover, the horizontal myoseptum appeared intact, suggesting that there was no indirect perturbation (supplementary material Fig. S4). Ventrally projecting nerves in gle1^–/– embryos also appeared disorganized, often showing excessive and ectopic branching (Fig. 3B,D). The motor nerve bundles were also thinner overall in gle1^–/– embryos.

Collectively, starting at 2 dpf, at least two motoneuron defects were observed in gle1^–/– embryos: fewer spinal motoneurons and disorganized motor axon arborization.

Based on the maternally deposited gle1 transcripts (Fig. 1A), we speculated that the relatively late onset of phenotypes in gle1^–/– embryos (at 2 dpf, Figs 2, 3) could be due to a maternal Gle1 contribution. To test this, we injected one-cell stage embryos with a mixture of two non-overlapping morpholino antisense oligonucleotides (MOs) designed to block gle1 translation. We determined the maximal MO concentration that had a negligible phenotypic effect on wild-type embryos. This suboptimal dose was then used with a mixed population of embryos derived from gle1^+/– intercrosses. Following phenotypic analysis, we genotyped individual embryos and found that gle1 knockdown in all three possible genotypic backgrounds (+/+, +/– and –/–) had similar effects on morphology and motoneuron number. The gle1 MO-injected embryos (gle1 morphants) showed widespread cell death in the CNS as early as 1 dpf (supplementary material Fig. S5D), a full day earlier than gle1^–/– embryos (Fig. 2F). At 1 dpf, gle1 morphants also developed fewer motoneurons, but their motor axon trajectories were similar to those of controls (supplementary material compare Fig. S5E with S5F). However, after 2 dpf, gle1 morphants displayed pronounced defects in motor axon arborization, including the lack of rostrally (Fig. 4B,E) and dorsally (Fig. 4E) projecting nerves and disorganized motor axon trajectories (Fig. 4B,E). Overall, the motoneuron phenotypes in gle1 morphants were very similar to, but more severe than, those observed in zygotic gle1^–/– mutants (Fig. 3B,D).

Importantly, several control experiments validated that the exacerbated CNS cell death and motoneuron defects observed in gle1 morphants were specific to the knockdown of Gle1 activity (maternal and zygotic) and not due to nonspecific off-target effects. First, injecting five-nucleotide-mismatched gle1 MOs at the same dose as the gle1 MOs did not cause any perturbations (data not shown; supplementary material Fig. S5A,B,E). Second, the gle1 morphant phenotypes could be rescued by expressing human GLE1 (see below). Overall, our Gle1 loss-of-function studies indicate that Gle1 activity is required for motoneuron survival and/or formation, as well as for motor axon arborization. Reducing maternal Gle1 activity allows the otherwise milder and later onset phenotypes of zygotic gle1^–/– mutants to be revealed.

Zebrafish have two populations of spinal motoneurons: primary motoneurons (PMNs) and secondary motoneurons (SMNs). By 24 hpf, on each side of each segmentally arranged myotomal muscle of the zebrafish trunk, there are three PMNs that are individually classified as CaP, MiP and RoP by the stereotyped positions of their cell bodies and distinct innervation patterns of their axons (Eisen et al., 1986; Myers et al., 1986; Westerfield et al., 1986). Zebrafish SMNs arise later during segmentation and are more numerous than PMNs (Myers et al., 1986; Westerfield et al., 1986). Starting at ∼25 hpf, SMNs extend their axons along the paths pioneered by primary axons (Myers, 1985; Pike et al., 1992). Axons of both PMNs and SMNs eventually form the dorsal and ventral motor nerves (Beattie, 2000; Pike et al., 1992).

Based on the phenotypes of 1-dpf gle1 morphants (supplementary material Fig. S5), we speculated that PMN formation and axogenesis would initiate normally in gle1 morphants, whereas SMN formation would be compromised. To test whether SMN formation was indeed affected upon the loss of Gle1 activity, we used a zn-8 (Alcama) antibody, which transiently recognizes the somata and fasciculated axons of SMNs but not of PMNs (Fashena and Westerfield, 1999). At 2 dpf in the control embryos, the somata and the fasciculated segments of the rostrally projecting nerves of SMNs were robustly co-labeled by zn-8 antibody and GFP expressed from the mnx1:GFP transgene (which labels both PMNs and SMNs) (Fig. 5A-C). By contrast, in gle1 morphants, the detection of zn-8-positive SMNs was severely reduced (Fig. 5E), whereas the GFP-labeled PMNs were still present (Fig. 5D). These results indicate that attenuation of Gle1 activity severely affects the formation of SMNs but not of PMNs. As such, the SMN perturbation would account for the overall reduction of motoneurons, the thinner axon bundles and the lack of rostrally projecting nerves in the gle1 mutants and morphants (Figs 3, 4).

Given that CNS cell death was one of the earliest phenotypes observed in gle1 mutants and morphants (Fig. 2F; supplementary material Fig. S5D), we next investigated whether Gle1 depletion causes motoneuron death, thus leading to the observed overall reduction of SMNs. To detect motoneuron death we performed TUNEL (terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling) in gle1 mutants and morphants, where the motoneuron nuclei were fluorescently labeled by the fusion protein H2A-TagRFP-T expressed from the mnx1:H2A-TagRFP-T transgene (see Materials and methods). Surprisingly, the TUNEL assay rarely detected motoneuron death: only 1 out of 108 spinal transverse sections from gle1 mutants and morphants (n=12) showed TUNEL⁺ H2A-TagRFP-T⁺ doubly labeled motoneurons (supplementary material Fig. S6, arrows), although TUNEL⁺ signals were evident in those sections. These results indicate that cells other than motoneurons in the spinal cord die upon Gle1 depletion.

To identify the dying cells in the Gle1-depleted spinal cords, we examined the positions of apoptotic cells relative to other cell types with specific markers. We labeled all postmitotic neurons with anti-Elavl antibody (previously called anti-HuC/D) (Marusich et al., 1994) and apoptotic cells with anti-cleaved caspase 3 antibody. Consistent with the results from the TUNEL assay, anti-cleaved caspase 3 staining revealed pronounced apoptotic cell death in the spinal cord of gle1 morphants. These apoptotic cells were distributed widely along both the dorsoventral and apical-basal axes of the spinal cord (supplementary material Fig. S7J). Notably, cleaved caspase 3⁺ cells were often located close to the central canal or immediately adjacent to the basally located Elavl⁺ cells (examples shown in supplementary material Fig. S7A-I). Strikingly, after extensive examination of samples at different developmental stages (more than 200 spinal transverse sections, n=26, at 24, 36 and 52 hpf), we did not find cleaved caspase 3⁺ Elavl⁺ doubly labeled cells, indicating that the differentiated neurons did not undergo apoptosis upon Gle1 depletion.

During neurogenesis, neural precursor cells are patterned along the dorsoventral axis of the neural tube, proliferate and then migrate laterally to differentiate into defined neuronal classes (Bylund et al., 2003; Kintner, 2002). The exclusive non-overlapping distribution of cleaved caspase 3⁺ cells and basally located Elavl⁺ neurons in gle1 morphants suggested that at least some of the apoptotic cells might have been neural precursors. Many neural precursors adopt features of radial glia, which give rise to specific classes of neurons in the brain (Anthony et al., 2004; Malatesta et al., 2003; Miyata et al., 2001; Noctor et al., 2002). In the zebrafish spinal cord, a subset of proliferative glial fibrillary acidic protein (Gfap)-positive radial glia reside along the central canal and serve as the neural and glial precursors that develop into motoneurons and oligodendrocytes (Kim et al., 2008). Using the Tg(gfap:GFP) line to mark radial glia (Bernardos and Raymond, 2006), we found that more than 70% of cleaved caspase 3⁺ cells in the spinal cord of gle1 morphants (n=23) were also Gfap⁺; many of those cells were located along the central canal (examples shown in Fig. 6A-L, arrows). We also observed some Elavl⁺ Gfap⁺ cells as previously described (Kim et al., 2008). GFP is stable and its fluorescence can persist after the transgene is no longer transcribed. Thus, the newly differentiated neurons (Elavl⁺) derived from Gfap⁺ radial glia might be GFP⁺. Furthermore, we found that almost all Gfap⁺ radial cells in the spinal cord expressed Sox2, a known neural stem cell marker (Bylund et al., 2003; Graham et al., 2003) (supplementary material Fig. S8A-D). Sox2⁺ cells were close to the apical surface and were mutually exclusive with motoneurons and Elavl⁺ neurons (supplementary material Fig. S8E-P; data not shown). Thus, the distribution of Sox2⁺ neural precursors highly resembles that of the apoptotic cells in Gle1-depleted spinal cords, suggesting that these two populations extensively overlap. Indeed, TUNEL and anti-Sox2 double labeling revealed TUNEL⁺ Sox2⁺ doubly labeled cells in the spinal cord upon Gle1 depletion (Fig. 6Q-T, arrowheads), confirming that Sox2⁺ neural precursors die upon Gle1 depletion. Taken together, we conclude that neural precursor cells, rather than differentiated neurons, undergo apoptosis upon Gle1 depletion.

To address whether defects in cell division underlie the death of Gle1-depleted neural precursors, we examined cell cycle progression of spinal neural precursors in the control and gle1 morphant embryos using a ‘percent labeled mitoses’ paradigm (Locker et al., 2006) (supplementary material Fig. S9). The control and Gle1-depleted neural precursors progressed through the cell cycle at a similar rate between 26 and 33 hpf (∼3.5 hours per cycle). After 33 hpf, the majority of neural precursors in the control embryos exited the cell cycle and transitioned into differentiation. However, ∼80% of Gle1-depleted precursors proceeded through a full cell cycle between 32 and 36 hpf (compared with fewer than 10% of controls) (supplementary material Fig. S9D). Thus, Gle1-depleted neural precursors failed to exit the cell cycle and/or transition into differentiation. Such a failure might trigger apoptosis of Gle1-deficient neural precursors.

In addition to impaired neurogenesis, Gle1 depletion also led to aberrant motor axon outgrowth. Given that an active crosstalk between neurons and their environment is required for navigating axons to form correct arborization patterns (for reviews, see Beattie, 2000; Chisholm and Tessier-Lavigne, 1999), Gle1 could be required intrinsically in motoneurons and/or extrinsically in the environment for proper axon arborization. To distinguish between these possibilities, we performed mosaic analyses of chimeras generated by cell transplantation. To increase the robustness of obtaining desired chimeras, we used gle1 morphants as donors or hosts. We further employed donor and host embryos with motoneurons labeled by mnx1:GFP and mnx1:TagRFP-T transgenes, respectively, to enable observation of donor and host motor axon behaviors in the same chimeras simultaneously.

Transplantation of wild-type donor cells (mnx1:GFP⁺) into gle1 morphant hosts (mnx1:TagRFP-T⁺) revealed that GFP⁺ wild-type motoneurons primarily followed the aberrant motor axon trajectories of TagRFP-T⁺ motoneurons in the gle1 morphant hosts (Fig. 7B; n=10 chimeras). These results indicate that Gle1 activity is required in the environment for normal motor axon arborization; wild-type intrinsic Gle1 activity in the motoneurons is not sufficient. Conversely, when the donor cells from gle1 morphants were transplanted into wild-type hosts, the resulting gle1 morphant-derived motoneurons consistently exhibited the normal axon arborization patterns (GFP⁺) of their neighboring wild-type motor axons (TagRFP-T⁺) (Fig. 7C; n=17 chimeras). Notably, the rostrally projecting axons now developed from the cells of gle1 morphant donors (Fig. 7C, asterisks). This was in stark contrast to the phenotypes observed in gle1 mutants or morphants, in which the rostrally projecting axons failed to form (Figs 3, 4). These results further indicated that Gle1 acts in a non-cell-autonomous manner during motor axon arborization.

To directly test the use of zebrafish as a model for studying LCCS1, we investigated the effects of expressing human GLE1 in the gle1 morphants. The CNS cell death and aberrant axonal trajectories were rescued by co-injecting gle1 MOs with in vitro transcribed human GLE1 mRNA (encoding GLE1 that is either untagged or EGFP tagged at the N-terminus) (Fig. 4C,F; supplementary material Fig. S10B,D). To compare the wild-type and Fin_Major mutant forms of human GLE1, we assessed the overall consequence of gle1 MO-induced head cell death – categorized as severe, mild or wild-type-like – after injecting the corresponding human GLE1 mRNAs. When gle1 morphants were injected with buffer alone, as expected no rescue of head cell death was observed, with ∼75% of gle1 morphants showing severe head cell death (Fig. 8). Injecting wild-type human GLE1 mRNA, however, efficiently rescued head cell death, with ∼75% of human GLE1 mRNA-injected gle1 morphants showing wild-type-like head morphology. In striking contrast, injecting human GLE1 Fin_Major mRNA exhibited very limited, if any, rescuing effect on head cell death (Fig. 8). Western blotting experiments confirmed that the wild-type and Fin_Major GLE1 proteins were comparably expressed in the embryos (supplementary material Fig. S11). This cross-species rescue demonstrates that Gle1 function is evolutionarily conserved between human and zebrafish. In addition, this directly supports the use of zebrafish as a model with which to determine the pathophysiological consequences of human LCCS1.

Using zebrafish to investigate the function of the gle1 gene mutated in LCCS1, we have established the first vertebrate model with which to study this human lethal congenital disease. We find that the zebrafish gle1 mutant and morphant phenotypes parallel several key LCCS1 syndromes. Moreover, expressing human GLE1 rescues zebrafish Gle1 deficiency phenotypes, whereas expressing the human GLE1 Fin_Major mutant does not. Further analysis of neuronal development shows that lack of Gle1 function directly impacts the survival of neural precursors and motor axon arborization. Surprisingly, differentiated neurons (e.g. those born before maternal Gle1 is depleted) rarely die upon Gle1 depletion. These observations are in sharp contrast to conclusions from prior work that proposed that LCCS1 pathology is linked to perturbations of motoneuron development and maturation (Nousiainen et al., 2008). Our studies raise the possibility that susceptibility of highly proliferative cells to Gle1 deficiency, including organ precursors, constitutes the cellular basis of pleiotropic phenotypes in LCCS1.

This new interpretation of the LCCS1 pathophysiology also uncovers important links between critical developmental events and specific steps in the molecular regulation of gene expression. Gle1 is essential for mRNA export (Kendirgi et al., 2003; Murphy and Wente, 1996; Watkins et al., 1998) and also regulates translation (Bolger et al., 2008). In highly proliferative cells, Gle1 roles in mRNA export and translation could be crucial to support a high rate of protein synthesis. During mRNA export at nuclear pore complexes (NPCs), Gle1 binds inositol hexakisphosphate (IP₆) and the Gle1-IP₆ complex facilitates ATP loading and activation of Dbp5 (Alcazar-Roman et al., 2006; Noble et al., 2011; Weirich et al., 2006). Dbp5 is an essential RNA-dependent ATPase of the DEAD-box protein family (Schmitt et al., 1999; Snay-Hodge et al., 1998; Tseng et al., 1998). Members of this DEAD-box protein family serve as RNA helicases and/or RNA-protein complex remodelers during many essential aspects of RNA metabolism (Cordin et al., 2006; Jankowsky, 2011; Jankowsky and Bowers, 2006; Linder, 2006; Rocak and Linder, 2004). For Dbp5, conversion to the ADP-bound conformation triggers the release of proteins from the mRNA during export and potentially during translation termination (Gross et al., 2007; Noble et al., 2011; Tran et al., 2007). The Dbp5 cycles of ATP hydrolysis at the NPC are controlled by both Gle1-IP₆ and an NPC-localized ADP release factor, Nup159 (Hodge et al., 2011; Noble et al., 2011). In addition to working with IP₆ and Dbp5 during mRNA export and translation termination, Gle1 also mediates translation initiation in an IP₆- and Dbp5-independent manner (Bolger et al., 2008; Bolger and Wente, 2011). With Gle1 being a key regulator during multiple phases of mRNA metabolism, we speculate that loss of Dbp5 function or IP₆ production might have similar developmental and disease impacts to Gle1 disruption. Analysis of zebrafish null mutants for either dbp5 (ddx19 – Zebrafish Information Network) or ipk1 (ippk – Zebrafish Information Network; the kinase responsible for IP₆ generation) has not been reported. We have previously characterized a zebrafish ipk1 morphant with early developmental defects linked to ciliary function (Sarmah et al., 2005; Sarmah et al., 2007); however, given the high binding affinity of Gle1 for IP₆ (Alcazar-Roman et al., 2010), a zebrafish ipk1 null is needed to test whether the complete loss of IP₆ production phenocopies the gle1^–/– mutant. Unfortunately, a mouse null for Ipk1 is embryonic lethal prior to day 8.5 (Verbsky et al., 2005), making it difficult to pinpoint specific developmental defects.

Although Gle1 is an essential regulator of several distinct steps of the mRNA life cycle, it remains unclear whether Gle1 is required at the same levels in different cell types and at different developmental stages. Our zebrafish studies reveal ubiquitous gle1 expression through early developmental stages, supporting a general requirement of Gle1. However, after 1 dpf, gle1 expression appears more restricted to certain tissues. Of note, these tissues with prominent gle1 expression are the same as those that show the most pronounced phenotypes in gle1 mutants at later stages. Furthermore, these gle1-enriched tissues are those that maintain populations of highly proliferative cells at early larval stages (e.g. 3 dpf) (de Jong-Curtain et al., 2009). Therefore, we propose that although Gle1 is likely to be required in all cells, highly proliferative cells demand a higher level of Gle1 to support the robust metabolic activity required during proliferation. Thus, these cells/tissues are the most vulnerable to disruption of Gle1 function and consequently exhibit the most severe phenotypes.

In addition to neuronal deficiency, Gle1-depleted embryos exhibit impaired motor axon arborization in the trunk. There are a plethora of receptor-ligand interactions between neurons and the environment that are required for outgrowing axons to find their correct synaptic targets and to form stereotyped axon arbors (Chisholm and Tessier-Lavigne, 1999). It is therefore surprising that Gle1 activity in the environment appears sufficient to restore normal axon arbors to Gle1-depleted motoneurons. These data further support the idea that Gle1 is required differentially in different tissues. An exciting future challenge will be to identify the cues provided by the wild-type environment and understand how they help organize motor axon arborization. Alternatively, the establishment of normal trajectories of Gle1-depleted motoneurons in the wild-type hosts could be explained by axon fasciculation. The trajectories are laid down by PMNs and are followed by SMN axons. Gle1-depleted motor axons could retain the ability to follow the normal PMN axon trajectories, resulting in normal motor axon arbors.

Multiple insights have been gained from our molecular and pathological analyses of gle1 in zebrafish. In terms of the molecular mechanism, the major mutation in LCCS1 patients, Fin_Major, is a three amino acid insertion in the putative coiled-coil domain of human GLE1 (Nousiainen et al., 2008; Watkins et al., 1998). To date, several Gle1 interaction partners have been identified. Gle1 associates with the NPC through interaction with nucleoporin 42 (Nup42) in yeast (Murphy and Wente, 1996; Strahm et al., 1999) or with hCG1 (the human ortholog of Nup42) and NUP155 in human cells (Kendirgi et al., 2005; Rayala et al., 2004). Gle1 also directly binds IP₆ (Alcazar-Roman et al., 2010; Alcazar-Roman et al., 2006; Montpetit et al., 2011; Weirich et al., 2006). However, the domain corresponding to the Fin_Major insertion is not required for interaction with IP₆, Nup42/hCG1 or NUP155 (Alcazar-Roman et al., 2010; Kendirgi et al., 2005; Rayala et al., 2004). Thus, the perturbation could impact a novel interaction partner involved in mRNA export or translation. Alternatively, the Fin_Major mutation might instead be a functionally null allele, as we find no significant rescue of the head cell death phenotype when human Fin_Major is expressed in zebrafish gle1 morphants. Additional studies are needed to determine whether the Fin_Major mutation disrupts only a subset of Gle1 functions (mRNA export and/or translation). The zebrafish rescue assay developed here provides a valuable in vivo system with which to further test gle1 disease alleles.

We propose that the wide spectrum of LCCS1 syndromes is of both neurogenic and non-neurogenic origin. Furthermore, the manifestation of phenotypes could be directly correlated with the high demand for Gle1 in highly proliferative cells. For the neurogenic aspect of the disease, apoptosis of neural precursors, not of differentiated neurons, might be the underlying cause of spinal cord atrophy. This in turn would lead to fetal immobility and, ultimately, multiple joint contractures, phenotypes that are often secondary to immobility (Moessinger, 1983). As for phenotypes such as edema, micrognathia and underdeveloped organs, we speculate that they are of non-neurogenic origin. However, as in the case of neuronal tissues, these non-neuronal defects could be caused by death of corresponding organ precursors. In support of the latter, we find that proliferative domains of these tissues in zebrafish, including the ciliary marginal zone of the eyes (supplementary material Fig. S2A), intestinal epithelia (Fig. 2L) and pharyngeal arches (data not shown), all show signs of apoptosis upon the loss of Gle1 activity.

In summary, we have provided the first insight into the functional consequences of gle1 mutations in vertebrate development and LCCS1 pathology. The fact that the etiology of motoneuron defects in LCCS1 is developmental and not neurodegenerative in origin, sets the stage for considering strategies that might rescue such defects.

We gratefully acknowledge Watson Folk for technical assistance; Wenbiao Chen, Josh Gamse, Ela Knapik, Lila Solnica-Krezel and Wen-Der Wang for sharing reagents and advice; the Zebrafish International Resource Center (supported by NIH P40 RR012546); Koichi Kawakami for Tol2 transposon system; Chi-Bin Chien for Gateway plasmids; Dirk Meyer for mnx1:GFP line and constructs; and Wente laboratory members for discussions. Monoclonal antibodies 39.4D5 and zn-8 (developed by T. M. Jessell, S. Brenner-Morton and B. Trevarrow) were obtained from the Developmental Studies Hybridoma Bank (NICHD supported and maintained by The University of Iowa, Department of Biology, Iowa City, IA 52242, USA).