|Home | About | Journals | Submit | Contact Us | Français|
The chemokine SDF1 activates a cAMP mediated signaling pathway which antagonizes retinal responses to the midline repellent slit. Using male and female zebrafish (Danio rerio), we show that knocking down the calmodulin activated adenylate cyclase ADCY8 makes retinal axons insensitive to the SDF1. Experiments in vivo confirm a mutual antagonism between slit signaling and ADCY8 mediated signaling. Unexpectedly, knockdown of ADCY8 or another calmodulin activated cyclase, ADCY1, induces ipsilateral mis-projections of retinal axons that would normally cross the ventral midline. We demonstrate a cell autonomous requirement for ADCY8 in retinal neurons for normal midline crossing. These findings are the first to show that ADCY8 is required for axonal pathfinding before axons reach their targets. They support a model in which ADCY8 is an essential component of a signaling pathway that opposes repellent signaling. Finally, they demonstrate that ADCY8 helps regulate retinal sensitivity to midline guidance cues.
Axons navigate through the developing nervous system by responding to differentially localized guidance cues in their immediate environment. Many diverse families of secreted and cell surface signaling molecules have been found to guide axons through their attractive and repellent effects on growth cones (Tessier-Lavigne and Goodman, 1996; Dickson, 2002). Growth cones routinely encounter multiple cues simultaneously and must integrate their activities into unitary guidance decisions. Understanding how axons navigate through their native environment will require an understanding of how multiple cues interact with one another to affect growth cone behavior.
One such interaction is the antagonistic relationship between the chemokine SDF1 and repellent guidance cues. SDF1 has been shown to make a variety of axons less sensitive to multiple axonal repellents including semaphorin3A, semaphorin3C, and slit2 in vitro (Chalasani et al., 2003). It has also been shown to antagonize slit/robo signaling in vivo (Chalasani et al., 2007). Pharmacological studies in cultured chick sensory and retinal axons showed that SDF1's anti-repellent effect is mediated by the seven transmembrane receptor CXCR4, a pertusis toxin-sensitive G-protein coupled pathway, the stimulation of calcium-calmodulin, and the activation of PKA. An analogue of cAMP stimulates the anti-repellent pathway while a cAMP antagonist blocks the anti-repellent effects of SDF1 (Chalasani et al., 2003). These results suggest that SDF1 signals through a calmodulin activated intermediary to elevate cAMP. Altering levels of cyclic nucleotides has been shown to dramatically alter growth cone responses to both attractants and repellents in vitro (Song and Poo, 1999; Piper et al., 2007), but there is very little understanding of how these levels are controlled in vivo. There are ten recognized adenylate cyclases that produce cAMP in higher vertebrates (Willoughby and Cooper, 2007). Two of them, ADCY1 and ADCY8, are known to be activated when they bind calcium activated calmodulin (Ferguson and Storm, 2004). ADCY1 is maximally activated by a combination of activated Gαs and calmodulin, while ADCY8 is activated by calmodulin alone (Wayman et al., 1994; Nielsen et al., 1996). Since SDF1 initiates its anti-repellent effects through a pertussis toxin sensitive Gαi or Gαo, and since it is not blocked by interfering with Gαs-mediated signaling (Twery et. al., unpublished), we hypothesized that SDF1's anti-repellent effects are mediated by calmodulin induced ADCY8 generation of cAMP. We therefore tested whether knockdown of ADCY8 interferes with SDF1-mediated anti-repellent signaling in cultured zebrafish retinal axons and during retinal pathfinding in the embryonic zebrafish.
Consistent with the hypothesis that calmodulin-activated cyclases participate in the SDF1 activated anti-repellent pathway, knockdown of ADCY8 makes zebrafish retinal axons more responsive to slits in the presence of SDF1 in vitro. Either ADCY8 or SDF1 knockdown tends to rescue specific retinal misprojections in mutant embryos in which retinal sensitivity to midline repellents is reduced. Surprisingly, knockdown of zADCY8 in the embryo induces abnormal ipsilateral retinal projections, and these mis-projections can be rescued by the simultaneous reduction or elimination of slit/robo repellent signaling. These findings demonstrate a requirement for the calmodulin-activated adenylate cyclase ADCY8 in retinal axon pathfinding in vivo and they are consistent with a model in which ADCY8 is part of a pathway that antagonizes repellent slits expressed at the ventral midline.
Changes in cAMP levels were monitored in retinal ganglion cells by Fluorescence Resonance Energy Transfer (FRET) using the cAMP sensor ICUE3 (DiPilato and Zhang, 2009). E5 chicken retinae were dissected and dissociated after treatement with 0.25% trypsin for 12 min at 37°C. Dissociated retinal neurons were electroporated with ICUE3 plasmid using the rat neuron Nucleofector kit and plated onto the glass bottoms of MatTak dish coated with laminin. Neurons were allowed to attach after plating for 6 hours in F12 medium supplemented with 6 mg/ml glucose, 2mM glutamine, 100 µM putrescine, 200 µM progesterone, 5 µg/ml insulin, 20 µg/ml NGF, 100 U/ml penicillin, 100 µg/ml streptomycin and 500 µl bovine pituitary extract (BPE) per 100 ml. The medium was then exchanged with F12 medium supplemented with all of the above components except BPE and cultured overnight at 37°C and 5% CO2. On the second day, the F12 medium was replaced with L15 medium supplemented with all of the above components except BPE and the temperature was maintained at 37°C on a heating stage at atmospheric CO2. Retinal ganglion cells were easily recognized as cells with long processes. Time-lapse imaging of the FRET indicator was performed on a Leica TSP2 confocal microscope. The indicator was excited at 458nm and dual emission images were collected at 465–490nm (cyan) and 535–560nm (yellow) every 30 seconds. Background was subtracted from both cyan and yellow channel images and the ratios of cyan-to-yellow emissions were then calculated for each time point and normalized to the average emission ratio just before treatment.
Wildtype Tü strain zebrafish were raised and maintained according to Mullins et al (Mullins et al., 1994). Zebrafish embryos were staged by time after fertilization and/or morphology (Kimmel et al., 1995). Heterozygous transgenic Brn3c:GAP43-GFP (Xiao et al., 2005) or Isl2b:GFP fish (Pittman et al., 2008) were used to produce embryos for cell transplantation experiments or in situ hybridization studies. Homozygous zebrafish robo2 astrayte284 or astrayti272z mutant fish (Fricke et al., 2001) were mated to produce homozygous mutant embryos for morpholino injections. All fish lines were kindly made available by the Chien laboratory (Utah). The astray mutations were identified by PCR as described previously (Chalasani et al., 2007). Fish were raised in 0.006% phenylthiourea to prevent pigmentation in embryonic and larval fish.
The full-length protein sequences and calmodulin-binding sites of human ADCY1 and ADCY8 were used to BLAST zebrafish genomic and EST databases at the Sanger Centre and NCBI. Two close relatives of ADCY1 and one of ADCY8 were found. Total RNA was extracted from 48hpf zebrafish embryos and reverse-transcribed into cDNA. Primers for RT-PCR were designed according to the zebrafish genomic and EST sequences. The sequences were completed by 5’- and 3’-RACE (Rapid Amplification of cDNA Ends, BD Biosciences, K1802-1). According to the seventh version of the Ensemble zebrafish database (Zv7, release 50; http://www.ensembl.org/Danio_rerio/Info/Index), zADCY1a is located on chromosome 20 while both zADCY1b and zADCY8 are located on chromosome 2.
The ICUE3 construct used as a cAMP sensor for FRET imaging was a kind gift from Dr. Zhang (DiPilato and Zhang, 2009). The pHuC:Gal4-VP16 and pUAS:EGFP plasmids were a kind gift from Dr. Jesuthasan (Hendricks and Jesuthasan, 2007). The pUAS:GAP43-mGFP plasmid was made to label axons by replacing the EGFP coding region from pUAS:EGFP with a coding sequence for GAP43-mGFP (Xiao et al., 2005). The zebrafish ADCY8 expressing plasmid pUAS:ADCY8 was made by replacing the EGFP coding region from pUAS:EGFP with the zADCY8 coding region.
In situ hybridization was performed as described (Chalasani et al., 2007) with minor modifications. Briefly, digoxigenin-labeled anti-sense cRNA probes were incubated with embryos to detect the expression pattern of various transcripts. anti-DIG-POD (Roche, Cat#11 207 733 910) was applied and the signal was amplified with a Cyanine 3-coupled tyramide system method (PerkinElmer, NEL 744). Immunostaining was performed simultaneously where needed to detect the co-localization of GFP labeled axons.
Immunostaining was performed as previously described (Chalasani et al., 2007). Briefly, zebrafish larvae were fixed overnight in 4%PFA and treated with 0.2% collagenase for 2.5 to 3 hours to facilitate antibody access into the tissue. Goat anti-GFP (1:500, Rockland Immunochemicals Inc., Cat# 600-101–215) and anti-goat IgG Alexa Fluor 488 (1:500; Invitrogen) were used to visualize green fluorescent protein (GFP)-positive cells and axons. Immunostained larvae were imaged with a Leica (Wetzlar, Germany) TSP2 confocal microscope and all Z-series stacks are shown as two dimensional projections.
Anti-sense morpholinos were synthesized by Gene-Tools, LLC (Philomath, OR). Two were designed to block pre-mRNA splicing and targeted exons within the first guanylate cyclase domain of zADCY8. Both cause shifts in the reading frame and almost immediate truncation of the translated protein. The morpholino sequences were: (E3) 5’-AAG ACA GAA ATT ACC TCA CGT TCT C-3’ (underlined nucleotides correspond to exon sequence) and (E4) 5’-AAG TGT GTT TAC TTA CGT GTG CCA G-3’. The sequence of the standard control morpholino from Gene-Tools is 5'-CCT CTT ACC TCA GTT ACA ATT TAT A-3’. Morpholinos were micro-injected into the yolks of one to two-cell stage embryos immediately after fertilization at the following concentrations: E3: 1.8ng/embryo; E4: 3.2ng/embryo, and E3+E4: 1.0ng E3 + 1.5ng E4/embryo. To characterize the effects of morpholinos on splicing, RT-PCR was performed using cDNA templates prepared from the RNA of 48hpf injected embryos. The PCR products were sequenced to confirm deletion of the expected exons.
Zebrafish larvae at 5dpf were fixed in 4% paraformaldehyde (PFA) overnight and mounted in 1.3% agarose on glass slides. Retinal axons from the left eye were anterogradely labeled with the lipophilic tracer DiI (D282; Invitrogen) and those from the right eye with DiD (D7757; Invitrogen), both dissolved at 2.5 mg/ml in dimethylformamide. After injection, the dyes were allowed to diffuse overnight at room temperature. The dye-labeled retinotectal axons were scanned on a Leica TSP2 confocal microscope and all images are presented as projections of Z series.
Zebrafish retinal explants were prepared and cultured as previously described (Chalasani et al., 2007). Control or zADCY8 morpholinos were injected into 1-cell stage embryos. At 2dpf, the embryos were anesthetized and the eyes were removed. Retinal explants from zADCY8 and from control morpholino treated fish larvae (approximately 10 retinas per condition) were cultured on opposite halves of poly-lysine (200 µg/ml) and laminin (80 µg/ml) coated glass bottomed MatTek dishes. After 24h in culture, retinal axons were allowed to grow for a 75 minute control period and then for an additional 75 minute treatment period after the addition of supernatants containing: (1) mock transfected, (2) human slit2-transfected, (3) zebrafish SDF1a-transfected, or (4) human slit2-transfected plus zebrafish SDF1a. Images were taken with a CCD camera (Spot; Phase 3 Imaging, Glen Mills, PA) at the beginning of the control and treatment periods and the end of treatment period. The migration distances of retinal axons during control and experimental periods were analyzed using Image-Pro software.
Donor embryos from Brn3c:GAP43-GFP transgenic parents were injected at the 1-cell stage with a mixture of rhodamine-conjugated Dextran (Molecular Probes, D3308) and zADCY8 morpholinos. Host embryos were uninjected wild type and stage-matched with donor embryos. At the sphere stage, embryos were dechorionated and about 20–50 donor embryo blastocysts were transplanted near the animal pole of the host, the region fated to become retina (Kozlowski et al., 1997). The donor embryos were phenotyped at 5dpf and the host embryos were fixed at 4dpf for immunostaining analysis.
Electroporation of the eye was performed as described with slight modifications (Hendricks and Jesuthasan, 2007). Embryos were injected at the 1-cell stage with zADCY8 morpholinos. At about 24hpf, TE buffer containing the plasmids HuC:Gal4-VP16, UAS:GAP43-GFP, with or without pUAS:zADCY8 were injected into the left eye and subjected to five 13V or 14V pulses 1 millisecond in duration and 500ms apart using an ECM 830 square-wave electroporation system (BTX, Division of Genetronics, Inc., San Diego, CA). The HuC promoter drives gene expression in early differentiated neurons (Park et al., 2000). At 4dpf, the electroporated fish larvae were fixed for immunostaining and analysis.
Previous studies have defined the outlines of a signaling pathway through which the chemokine SDF1 reduces axonal responses to repellent guidance cues in cultured embryonic primary neurons (Chalasani et al., 2003). This pathway is unusual as it appears to require an increase in cAMP that is induced by a pertussis toxin sensitive Gαi/o-protein intermediary. This presumed increase has not been measured directly and contrasts with measured reductions of cAMP levels in SDF1 treated astrocytes and epithelial cells (Dwinell et al., 2004; Warrington et al., 2007). We used the FRET-based cyclic cAMP sensor ICUE3 (Indicator of cAMP using Epac) (DiPilato and Zhang, 2009) to directly detect SDF1 induced changes in cAMP levels in cultured embryonic retinal ganglion cells. An expression vector encoding ICUE3 was transfected into cultured primary chick retinal ganglion cells. The ratio of cyan to yellow fluorescence was monitored to track changes in cAMP levels. This ratio increased in SDF1 treated cells over the course of more than 10 minutes, indicating that cAMP levels rose in response to SDF1 treatment (Figure 1A,B,C). No increase in cAMP was detected without the application of SDF1, and even greater increases were observed in response to application of the adenylate cyclase stimulator forskolin (Figure1D). The SDF1-induced increase in cAMP, but not the forskolin induced increase in cAMP, was abolished by the calmodulin antagonist Calmidazolium (CMZ) (Figure 1E). These findings demonstrate that in primary embryonic retinal ganglion cells, SDF1 can induce an increase in cAMP that is dependent on calmodulin activation. They are consistent with previous findings in embryonic neurons and support the hypothesis that SDF1's effects are mediated through calmodulin activated adenylate cyclases.
BLAST searches identified three adenylate cyclases (ADCYs) in the Ensembl zebrafish genome database that contain sequences matching calmodulin binding sites in higher vertebrate ADCY1 and ADCY8 (Vorherr et al., 1993; Gu and Cooper, 1999). The full coding sequences of three zebrafish ADCYs were reconstructed by 5'- and 3'-RACE using 48 hpf zebrafish cDNA as a template. Two of them are most closely related to human ADCY1 and the third to ADCY8, so we have named them zADCY1a, zADCY1b, and zADCY8 respectively (GeneBank accession numbers: ADCY1a, GU169394; ADCY1b, GU169395; ADCY8, FJ472834). The organization of human as compared to zebrafish exons and introns, guanylate cyclase domains, and calmodulin binding sites are all highly conserved (Figure 2A,B and Figure S1A,B,C). The amino acid sequences of the zebrafish guanylate cyclase domains are approximately 90% identical to those in the corresponding human ADCYs.
This study focuses on the functional characterization of zADCY8 during retinal axon outgrowth. The predicted gene structure of zADCY8 is strikingly similar to that of hADCY8. Both have 18 exons. The numbers of nucleotides in each exon are similar between fish and humans except for the first and last exons, which are both shorter in zebrafish. We have tentatively numbered the exons in zADCY8 according to their human counterparts. There are two calmodulin-binding domains, one near the N-terminus and the other near the C-terminus. Both are required for calmodulin regulation of enzymatic activity (Gu and Cooper, 1999; Simpson et al., 2006). The N-terminal most calmodulin-binding site is thought to recruit calmodulin to ADCY8 while the interaction of calmodulin with the C-terminal most site is thought to stimulate ADCY8 cyclase activity (Simpson et al., 2006). Both calmodulin binding sites are highly conserved between zebrafish and higher vertebrates (Figure 2A). There are only a small number of conservative substitutions which would not be expected to interfere with calmodulin-binding (Rhoads and Friedberg, 1997).
In situs for zADCY8 in 36hpf zebrafish embryos demonstrate expression in the CNS as a whole and in the retinal ganglion cell layer of the eye (Figure 2C,D). To confirm that zADCY8 is expressed in retinal ganglion cells, in situs were performed in Isl2b:GFP transgenic zebrafish in which all differentiated retinal ganglion cells (RGCs) express GFP (Pittman et al., 2008). zADCY8 transcripts were detected within individual RGC cells in 36hpf embryos (Figure 2E.F). This is a relatively early time when the first RGCs have been born and their axons are just beginning to cross the ventral midline (Burrill and Easter, 1995). Two non-overlapping RNA probes detected the same expression pattern for zADCY8. zADCY1b, but not zADCY1a, is also expressed at comparable levels and times in retinal ganglion cells (Figure S1E-F). These results demonstrate that the calmodulin-stimulated adenylate cyclases zADCY8 and ADCY1b are expressed in RGCs at a time consistent with their playing a role in retinal axon guidance.
ADCYs require both guanylate cyclase domains for catalytic activity (Willoughby and Cooper, 2007). We designed morpholinos that were predicted to cause disruption of the first guanylate-cyclase domain of zADCY8 by causing early exons to be skipped during the splicing of pre-mRNAs (Morcos, 2007). Skipping these exons should cause a shift in reading frame and premature termination of translation. The efficacy of each morpholino was tested by performing RT-PCR for targeted sequences using cDNA prepared from morpholino injected as compared to control 48 hpf embryos. Morpholinos targeting either exon 3 or 4 of zADCY8 were tested singly and in combination (Figure 2B). Agarose gel electrophoresis and sequencing of the RT-PCR products indicated that each mopholino induces truncated PCR products of the expected size (Figure 2B, arrow heads), loss of either exon 3 or exon 4 as expected, and generation of an early termination codon in the truncated sequences. Combining the two morpholinos induces a mixture of PCR products missing exon 3, exon 4, or both exons (Figure 2B, arrow heads). Most subsequent knockdown experiments were performed with the two morpholinos together, but where noted, phenotypes were confirmed with single morpholinos. Comparable splice blocking morpholinos against exons 5 or 7 of ADCY1a (Figure S1B) and exons 4 and 7 of ADCY1b (Figure S1C) were made and tested.
We hypothesized that SDF1 induced calmodulin activation stimulates zADCY8 and that the resulting elevation of cAMP causes retinal axons to be relatively insensitive to repellent signals. We therefore tested whether zADCY8 knockdown makes zebrafish retinal ganglion cell axons less responsive to the anti-repellent effects of SDF1a. Retinae were harvested from zADCY8 morpholino or control morpholino injected larvae at 2dpf, cut into small pieces, and cultured on laminin coated glass. Each coverslip had control explants on one side and zADCY8 morpholino containing explants on the other (see methods). Many individual growth cones were photographed at the beginning and end of a 75 minute non-treatment period, the bathing medium was changed, and the same growth cones were photographed again after a second 75 minute treatment period (Figure 3A1–C3). These photographs were used to measure the forward progress of individual growth cones during the two periods. During the treatment period either the repellent slit2, the chemokine SDF1a, or both were added to the cultures.
Knockdown of zADCY8 does not affect the rate of retinal axon outgrowth (Figure 3D-F, compare the average migration distances between control or zADCY8 morpholino treated axons during the non-treatment periods). Retinal axons containing control or zADCY8 morpholino respond equally well to the repellent slit2 (Figure 3D). Retinal axons containing control or zADCY8 morpholinos advance at the same rate in the presence of zSDF1a (Figure 3E). Consistent with our previous findings4, retinal axons containing control morpholino are significantly less responsive to slit2 in the presence of zSDF1a as compared to its absence (compare empty columns in Figure 3D and 3F). In contrast, retinal axons containing zADCY8 morpholino are equally sensitive to slit2 in the presence of zSDF1a as compared to its absence (compare filled columns in Figure 3D and 3F). The difference is most apparent in experiments where both slit2 and zSDF1 are presented together during the experimental period (Figure 3F). Control morpholino containing retinal axons continue to advance, while zADCY8 morpholino containing axons retract. These findings demonstrate that zADCY8 knockdown makes zebrafish retinal ganglion cell axons insensitive to SDF1a. They are consistent with zADCY8 serving as an essential step in the SDF1a mediated anti-repellent pathway. Importantly, they demonstrate a cell autonomous change in retinal ganglion cell responsiveness to a repellent guidance cue when zADCY8 is knocked down.
We examined the consequences of zADCY8 knockdown on axonal pathfinding in the zebrafish retinal projection. After exiting the eye, retinal axons extend through the ventral diencephalon towards the ventral surface of the brain where they cross the midline at the optic chiasm. As they continue to extend, they grow dorsally and posteriorly to the tectum. In contrast to binocular animals, all retinal axons cross the midline in the zebrafish. We used anterograde dye tracing in 5 dpf zebrafish larvae to compare retinal axon trajectories in zADCY8 morphants. All retinal axons crossed the midline normally and projected to the contralateral tectum in control morpholino-treated larvae (Figure 4A). However, in a significant number of zADCY8 morphants, some or all retinal axons failed to cross the midline and mis-projected instead to the ipsilateral tectum (Figure 4B–D). Abnormal ipsilateral projections were detected in 62% of E3 morpholino treated eyes and in 12% of E4 morpholino treated eyes (Figure 4E). Combining the E3 and E4 morpholinos, using reduced amounts of each that individually produced no ipsilateral mis-projections, produced robust ipsilateral mis-projections. Synergy between the E3 and E4 morpholinos is consistent with their inducing their effects by acting on a common target. This combination of morpholinos is less likely to induce off-target effects than higher concentrations of single morpholinos and was therefore used in most subsequent experiments. Although mis-projecting retinal axons did not cross the midline, they joined the normal ipsilateral optic tract and projected with grossly normal retinotopy into the ipsilateral tectum (Figure S2).
Knocking down zADCY8 caused a marked reduction in the overall size of the embryo (Figure S4E). This effect could in principle be caused by the knockdown of zADCY8 itself since it is widely expressed, especially in the central nervous system, or by a toxic reaction to morpholinos that induces p53 mediated apoptotic cell death (Robu et al., 2007). We co-injected a p53 morpholino and zADCY8 morpholinos together in order to reduce potential p53 mediated apototic cell death (Robu et al., 2007) and found that the eyes and head were still small in morphant as compared to wildtype embryos. Ipsilateral retinal mis-projections were observed at comparable frequencies in embryos containing ADCY8 or concurrent ADCY8 and p53 morpholinos (data not shown). Knockdown of zADCY8 also induces a developmental delay in the eye. The initial formation of the optic chiasm, which occurs around 36hpf in normal embryos, did not occur until 18hrs later in zADCY8 morphants. The delay appeared to be greater in the eye as compared to the trunk as judged by the expression pattern of cxcr4b in the lateral line primordium (David et al., 2002) (data not shown). These results suggest that ADCY8 knockdown affects both the size and rate of development of the eye and other central nervous tissues. Subsequent experiments were therefore aimed at determining whether the retinal axon misguidance phenotype is induced directly by the loss of zADCY8 in retinal ganglion cells or as an indirect consequence of developmental perturbations along the retinal pathway.
Knockdown of zADCY8 might alter retinal responses to midline guidance cues, or alternatively, alter the distributions of those cues at the midline. Since there is a significant developmental delay in zADCY8 morphants, we precisely stage-matched zADCY8 morphants with wildtype embryos according to key morphological markers (Kimmel et al., 1995) and the maturity of retinal axon projections as visualized in Isl2b:GFP transgenics (Pittman et al., 2008). We compared a number of midline markers and/or potential guidance cues in normal and morphant larvae at the time when the first retinal axons cross the midline equivalent to 36 hpf in untreated embryos. These included the midline morphogen and potential guidance cue sonic hedgehog (Trousse et al., 2001), nkx2.2 (Marcus et al., 1999), sema3d (Sakai and Halloran, 2006), and the likely repellent cues slit1a, slit1b, slit2, and slit3 (Hutson and Chien, 2002; Hutson et al., 2003). We found that each of these signaling molecules is expressed at the midline in similar patterns in both wildtype and zADCY8 morphant embryos (Figure S3, A–C' and G–J'). We also examined the expression patterns of two receptors expressed in retinal ganglion cells, robo2 and cxcr4b, each of which is required for normal retinal axon guidance (Fricke et al., 2001; Li et al., 2005). Both were detected in the RGC layer and their expression patterns are comparable between wildtype and zADCY8 morphants (Figure S3, E–F’). Another potential retinal axon guidance receptor, neuropilin1a (Liu et al., 2004), also maintains a normal expression pattern in zADCY8 morphants (Figure S3, D-D’). All of these key cues and receptors are expressed in their appropriate patterns at the actual time that retinal axons first cross the ventral midline. What is more, their normal expression suggests that midline tissues differentiate normally. The errors we see in retinal pathfinding are therefore difficult to explain by the disruption of midline tissues, or from the loss of known midline guidance cues or their receptors.
We took two independent approaches to determine whether knockdown of zADCY8 alters retinal axon responses to midline guidance cues. In the first, we tested whether retinal ganglion cells harvested from zADCY8 morphant embryos and transplanted into untreated wildtype embryos extend axons normally across the ventral midline or whether they mis-project ipsilaterally. Multipotent progenitors were harvested from Brn3c:GAP43-GFP transgenic fish in which a fluorescent axonal tracer is expressed in approximately 50% of cells that differentiate into retinal ganglion cells (Xiao et al., 2005). These donor embryos were injected at the one cell stage with zADCY8 morpholinos and dextran red. Blastocyst cells were harvested from them at 4 hpf and transplanted into stage-matched wild type host embryos near the animal pole (Figure 5A). Chimeric embryos were allowed to grow to 4 dpf. Embryos with GFP labeled retinal ganglion cells in only one eye were selected using a dissecting fluorescent microscope for analysis. GFP staining was enhanced by immunofluorescence and examined with a confocal fluorescent microscope. Transplanted zADCY8 morpholino containing retinal ganglion cell axons with ipsilateral mis-projections were detected in 7 of 39 embryos (Figure 5C,D). Since only a very small proportion of retinal neurons derive from the transplant, and since most axons project normally even when all retinal neurons contain zADCY8 morpholino, this approach would be expected to underestimate the rate at which mis-projections occur when morpholino containing retinal axons extend in normal tissues. Transplanted retinal axon trajectories that did not contain zADCY8 morpholino projected normally across the midline in 32 of 32 embryos (Figure 5B). These results are significant to the p<0.05 level (Fisher's exact test) and are consistent with a retinal ganglion cell autonomous requirement for zADCY8 for normal midline crossing. Further, they strongly suggest that ipsilateral mis-projections in morpholino treated embryos do not arise from a general delay in embryonic development, since ipsilateral projections occured in chimeric embryos in which the overall rate of development was normal. However, since a small number of scattered dextran-labeled transplanted cells were generally present near the ventral midline in chimeric embryos, these results do not absolutely rule out the possibility that morpholino treated midline cells contribute to the formation of ipsilateral misprojections. We were unsuccessful in attempts to transplant wildtype blastocyst cells into zADCY8 morpholino treated embryos, as the recipient embryos were too fragile to survive the transplantation procedure.
We used a second approach to confirm that zADCY8 is required within retinal ganglion cells for normal pathfinding. Full length zADCY8 was transfected into retinal neurons of morpholino treated embryos to determine whether its expression rescues midline guidance errors. zADCY8 morpholino treated 24 hpf embryos were injected in one eye with a combination of expression plasmids containing either HuC:Gal4-VP16 and UAS:GAP43-GFP; or with HuC:Gal4-VP16, UAS:GAP43-GFP, and UAS:zADCY8 plasmids. The HuC promoter drives expression in neurons just before they extend axonal processes (Park et al., 2000). The injected eye was immediately electroporated to transfect the DNA into retinal neurons (Figure 6A). Expression of electropporated constructs was detected in only a small number of retinal ganglion cells. In one experiment in which we quantified expressing cells, only an average of 2.2 ganglion cells expressed the control GFP construct and 2.2 cells expressed ADCY8. Ipsilaterally mis-projecting retinal axons were detected in 14% (8 of 56) of zADCY8 morphant embryos expressing only the control GFP construct (Figure 6B,D). This low frequency of ipsilateral mis-projections can be explained by the low proportion of retinal neurons that were labeled, since most retinal axons project contralaterally even in morphant embryos. Ipsilaterally mis-projecting axons were detected in only 4% (3 of 75) of morphant embryos in which full length zADCY8 was transfected into the retina (Figure 6C,D). Expression of full length zADCY8 in non-morphant embryos did not affect retinal midline crossing. Thus, expression of full length zADCY8 rescued ipsilateral mis-projections in morphant embryos. These results argue strongly that ipsilateral mis-projections are not caused by a general delay of development in morphant embryos since rescue was successfully accomplished without reversing the overall developmental delay. They support the idea that zADCY8 activity within retinal ganglion cells influences the behavior of their axons at the midline.
We tested whether the calmodulin activated cyclases zADCY1a or zADCY1b contribute to normal guidance of retinal axons at the midline. Multiple splice-altering morpholinos were confirmed to induce premature stop codons within the first of their two cyclase domains (Figure S1B,C). As expected from the absence of zADCY1a expression in retinal ganglion cells, zADCY1a morpholinos, even at excessive concentrations that caused approximately 50% mortality and induced reduced eye and brain size in the surviving embryos (Figure S4C), produced almost no ipsilateral misprojecting retinal axons (Figure S4H,K). The same was true for a morpholino targeting sema3a1 (Figure S4B,G,K). In contrast, morpholinos directed against ADCY1b did induce ipsilateral retinal misprojections in 22% of morphants (Figure S4I,K). Since zADCY1b and zADCY8 morphants phenocopied one another, we next tested whether these two cyclases act cooperatively to promote midline crossing. Combining morpholinos to each at doses too low to induce ipsilateral misprojections on their own, induced ipsilateral misprojections in 28% of morphants. These findings suggest that the calmodulin activated cyclases zADCY8 and ADCY1b work together to promote retinal axon crossing of the ventral midline.
Slit is a repellent for retinal ganglion cell axons whose expression near the ventral midline helps determine where the optic chiasm forms (Erskine et al., 2000; Niclou et al., 2000; Hutson and Chien, 2002; Plump et al., 2002; Chalasani et al., 2003; Kreibich et al., 2004; Chalasani et al., 2007). Robo2 is the only slit receptor expressed in retinal ganglion cells as their axons cross the ventral midline in zebrafish (Lee et al., 2001). Zebrafish that have a partial loss of robo2 function display a number of retinal guidance errors, some of which can be rescued by interrupting SDF1a signaling (Chalasani et al., 2007). If zADCY8 is an essential component within the SDF1 signaling pathway, we predicted that its knockdown should tend to rescue the partial loss of robo2 phenotype. There are several retinal axon pathfinding errors in partial loss of robo2 astrayte284 mutants, including extra midline crossings, anterior and posterior mis-projections, and some ipsilateral mis-projections (Figure 7A) (Fricke et al., 2001). Knocking down SDF1a or its receptor CXCR4b partially rescues the anterior mis-projections in astrayte284 mutants (Chalasani et al., 2007). We focused on this specific class of pathfinding errors when we assessed the effect of zADCY8 knockdown on astray fish. We found that the injection of zADCY8 morpholinos into astrayte284 mutant embryos induces significant rescue of anterior mis-projections (Figure 7B and C).
Neither SDF1a or CXCR4b knockdown rescues anterior mis-projections of retinal axons in zebrafish that lack functional robo2 altogether, presumably because there is no residual slit/robo signaling that can be strengthened by reducing SDF1 signaling (Chalasani et al., 2007). If zADCY8 is a key step in the SDF1 signaling pathway, then zADCY8 knockdown would not be expected to rescue anterior mis-projections in mutants with a complete loss of robo2. In robo2 null mutants, such as astrayti272, retinal axon pathfinding errors are more severe than in astrayte284 mutant embryos (Figure 7D) (Fricke et al., 2001). zADCY8 knockdown has no significant effect on anterior mis-projections of retinal axons in astrayti272 mutant embryos (Figure 7E and F). RT-PCR was performed to confirm that morpholinos knocked down zADCY8 in both astrayte284 and astrayti272 mutants (data not shown).
Since ADCY8 is part of a signaling pathway that reduces axonal responses to repellents, its knockdown would be expected to make retinal axons more sensitive to repellent cues. Ipsilateral misprojections in zADCY8 morphant embryos could potentially be explained by retinal axon hypersensitivity to repellents at the midline. Slits are potent repellents which are expressed at the midline and play an important role guiding retinal axons (Erskine et al., 2000; Niclou et al., 2000; Hutson and Chien, 2002; Kreibich et al., 2004; Chalasani et al., 2007). We therefore tested whether reducing or eliminating slit signaling in robo2 mutants reduces ipsilateral misprojections in zADCY8 morphants. We limited our analysis to strong ipsilateral misprojections in morphant embryos, defined as>50% of retinal axons projecting ipsilaterally, since smaller ipsilateral misprojections sometimes occur in robo2 mutants. Morpholino-induced strong ipsilateral mis-projections are less prevalent in either astrayte284 mutants that have a partial loss of robo2 or in astrayti272 mutants that have a complete loss of robo2 as compared to wildtype embryos (Figure 8D). These observations are consistent with the idea that zADCY8 is part of a signaling pathway that antagonizes slit/robo mediated repellent signaling.
Our studies reveal an essential role for calmodulin-activated adenylate cyclases in early retinal pathfinding in the zebrafish embryo. Morpholino knockdown of zADCY8 induces a significant number of retinal ganglion cell axons to be deflected away from the ventral midline instead of following their normal trajectory to the contralateral tectum (Figure 4). Deflected axons generally join the ipsilateral optic tract and travel to the ipsilateral tectum, but in some instances, they travel to entirely inappropriate anterior locations. A second calmodulin activated cyclase, ADCY1b, acts cooperatively with zADCY8 to facilitate midline crossing of retinal axons (Figure S4). The pathfinding errors observed in ADCY8 knockdown embryos arise, at least in part, by a cell autonomous requirement for ADCY8 within retinal ganglion cells. This was demonstrated in three ways. First, ipsilateral misprojections arise from axons of retinal neurons containing zADCY8 morpholino that are transplanted into untreated wildtype embryos (Figure 5). This indicates that knockdown axons make errors even when interacting with normal host tissues. Second, errors in retinal projections in morpholino treated embryos were rescued by the re-expression of morpholino insensitive zADCY8 in retinal neurons (Figure 6). This demonstrates that zADCY8 expressing axons are more likely to navigate correctly when confronted with knockdown tissues than zADCY8 knockdown axons. Third, experiments in vitro demonstrated that knocking down zADCY8 makes retinal ganglion cells relatively insensitive to SDF1a. This shows that zADCY8 is required within retinal ganglion cells for SDF1 signaling. These findings do not preclude a non-autonomous contribution of zADCY8 to retinal axon pathfinding in vitro, but they provide decisive support for a retinal ganglion cell autonomous contribution to normal guidance.
Previous studies have suggested that ADCY1 and ADCY3 play important roles in determining where axons terminate within their targets, but have not reported pathfinding errors en route to their targets. Retinal projections have been shown to be perturbed in ADCY1 mutant mice (Welker et al., 1996; Abdel-Majid et al., 1998). Both the orderly retinotopic mapping of retinal projections and the normal separation of ipsi and contralateral axons are significantly disrupted in the lateral geniculate of the thalamus (Ravary et al., 2003; Nicol et al., 2006). These alterations in sensory projections have been interpreted as resulting from ACDY1's presumptive role in activity dependent remodeling of sensory connectivity (Ravary et al., 2003; Nicol et al., 2006). Our results suggest the alternative possibility that guidance errors contribute to some of the mapping errors observed in ADCY1 knockout mice. Abnormal ipsilateral trajectories in the visual system, like those in zADCY8 morphant fish, could lead to the appearance of abnormal segregation of ipsi and contralateral projections in the geniculate.
An adenylate cyclase that is indirectly regulated by calmodulin (Wei et al., 1996), ADCY3, has recently been shown to be essential for the normal convergence of olfactory sensory cell axons into specific glomeruli in the main olfactory bulb. ADCY3 is expressed in sensory neurons within the olfactory epithelium (Wong et al., 2000). The axons of many olfactory sensory neurons mis-project to multiple abnormal locations in the bulb in ADCY3−/− mice, and single glomeruli form aberrantly from sensory axons expressing dissimilar olfactory receptors (Chesler et al., 2007; Col et al., 2007; Zou et al., 2007). Guidance of olfactory axon projections in ADCY3 mutant mice may be disrupted by perturbations in cAMP levels that affect how axons respond to guidance cues (Song and Poo, 1999; Chalasani et al., 2003). This could occur if cAMP levels control the expression levels of guidance cue receptors on the surfaces of olfactory axons (Imai et al., 2006). These previously reported roles of ADCY1 and ADCY3 demonstrate their role in determining where axons terminate within their targets. Our results are the first to demonstrate that a calmodulin activated ADCY is involved in primary axon guidance as axons extend towards their target.
What accounts for the abnormal deflection of retinal axons away from the ventral midline in zADCY8 knockdown embryos? Our findings argue that ADCY8 is required in retinal ganglion cells, and errors arise from either a diminution of retinal sensitivity to midline attractants, the potentiation of retinal responsiveness to midline repellents, or both. Previous work has shown that the chemokine SDF1 makes chick retinal axons less sensitive to multiple repellents in vitro. This effect is mediated by a pertussis toxin and calmodulin-sensitive elevation of cAMP and the activation of PKA (Chalasani et al., 2003). We hypothesized that calmodulin-activated ADCYs are key components of this anti-repellent signaling pathway. If true, then the knockdown of zADCY8 should make zebrafish retinal axons less responsive to SDF1 and more sensitive to slits. Consistent with this prediction, retinal axons harvested from zADCY8 morphants retract in response to a mixture of SDF1 and slit2 much more strongly than retinal axons injected with a control morpholino (Figure 3). These results show that retinal axons in zADCY8 morphants are insensitive to the anti-repellent effects of SDF1, and that this change is cell autonomous.
zADCY8 knockdown makes retinal axons less responsive to SDF1, and consequently more sensitive to slit, in vitro. Can antagonism between SDF1 and slit signaling be observed in vivo? In a previous study we showed that retinal axons extend ectopically towards an anterior region of slit expression in mutant fish in which slit/robo signaling is reduced. Further, we found that knocking down SDF1/CXCR4 signaling tended to prevent abnormal anterior extension in mutants with reduced robo2, but not in mutants lacking robo2 altogether (Chalasani et al., 2007). Rescue of the anterior mis-projection therefore requires the presence of some robo2 activity. These results suggest that SDF1/CXCR4 signaling antagonizes slit/robo2 signaling and that residual slit/robo2 signaling is boosted by reducing SDF1/CXCR4 signaling. If zADCY8 is a key signaling step in an SDF1 triggered anti-repellent pathway, then zADCY8 knockdown should also tend to rescue anterior mis-projections in mutants that have a partial, but not a total, loss of robo2. This prediction was confirmed by experiment in this study. zADCY8 knockdown rescues anterior retinal misprojections in partial but not complete loss of robo2 function mutants (Figure 7). Both our in vitro and our in vivo findings therefore support a model in which zADCY8 constitutes a key intermediate in an SDF1 activated anti-repellent signaling pathway.
The generalization of this same argument as applied to other potential antagonists of slit signaling at the midline could explain why zADCY8 knockdown induces midline deflection of retinal axons. Slits expressed near the midline play an important role in determining where the optic chiasm forms (Erskine et al., 2000; Niclou et al., 2000; Hutson and Chien, 2002; Plump et al., 2002; Chalasani et al., 2007). Disruptions of slit1 and slit2 in mice induce retinal defasciculation and ectopic crossings at the chiasm (Plump et al., 2002), while loss of the slit receptor robo2 induces widening of the chiasm and ectopic projections of retinal axons to inappropriate targets (Plachez et al., 2008). Retinal axons cross the ventral midline in unusual locations and project ectopically in zebrafish mutants with reduced or missing robo2 (Fricke et al., 2001; Hutson and Chien, 2002). The chiasm normally forms where slit expression is low (Erskine et al., 2000; Niclou et al., 2000; Hutson and Chien, 2002; Plump et al., 2002), but since slits are secreted and are expressed in nearby midline tissues, it is likely that low levels of slit proteins are present even at the chiasm where retinal axons cross. One plausible explanation for ipsilateral misrouting of retinal projections in zADCY8 morphants is that a loss of anti-repellent signaling makes retinal axons hypersensitive to slits in the chiasm.
If zADCY8 knockdown induces abnormal ipsilateral projections by making retinal axons overly sensitive to midline slits, then a reduction in slit/robo signaling would be predicted to rescue these errors. Supporting this idea is the finding that zADCY8 morpholino induced ipsilateral projections are much less frequent in mutant embryos in which robo2 is either reduced or absent. Although our findings are consistent with the idea that ADCY8 knockdown makes retinal axons hypersensitive to midline repellents, we cannot exclude the alternative possibility that loss of zADCY8 interferes with an attractive mechanism that helps retinal axons cross the midline. Whether zADCY8 is required for an anti-repellent or an attractive response to a midline cue, that cue is unlikely to be SDF1. SDF1 is not expressed at the midline and ipsilateral misprojections are not observed in either SDF 1 or cxcr4b mutants or morphants (Li et al., 2005). Several G-protein coupled signaling pathways have been proposed to affect retinal axon guidance. The adenosine receptor A2b is thought to have a positive influence on cAMP levels in retinal axons that modulates their response to netrin (Shewan et al., 2002), while sonic hedgehog is thought to mediate a repellent response by reducing cAMP levels in retinal growth cones (Trousse et al., 2001). Activation of the metabotropic gluatamate receptor mGluR1 reduces the repellent activity of slit2 on retinal axons in a cAMP-dependent manner (Kreibich et al., 2004). It is not known whether any of these signals influence ADCY1 or ADCY8 activity, nor have any of these GPCRs been reported to affect axonal crossing at the midline.
Studies in vitro have demonstrated that modulating cAMP levels within the growth cone can have a profound effect on axon guidance, but the role of cAMP in axonal pathfinding in vivo has been less well studied. Our findings demonstrate an essential and previously unappreciated role for calmodulin activated adendylate cyclases ADCY1b and ADCY8 in retinal axon pathfinding in vivo. Further, they show that ADCY8 is part of a signaling pathway that antagonizes repellent signaling. They support the idea that retinal axon sensitivity to midline guidance cues is regulated by cAMP levels. Future studies will focus on the identification of additional cues, besides the chemokine SDF1, that influence axonal pathfinding through the activation of ADCY8.
The authors thank Drs Greg Bashaw, Michael Granato, and Mathew Dalva for reading the manuscript and their comments. We thank Dr. Jin Zhang for the ICUE3 construct and Dr. Suresh Jesuthasan for the HuC:Gal4-VP16 and UAS:EGFP constructs. The Brn3c:GAP43-GFP transgenic fish line was a gift from Dr. Herwig Baier. We thank Dr. Chi-Bin Chien for the astray mutants and the Isl2b:GFP transgenic fish. Radioactive in situs were performed by the Histology and Gene Expression Core of Molecular Cardiology Research Center at UPENN with the help of Arun Padmana from Jonathan Epstein's laboratory. We thank Drs. Sreekanth Chalasani, Mark Lush, Naomi Twery, Angela Sabol, Vanisha Lakhina and other members of the Raper laboratory for many thoughtful and stimulating discussions. We also thank everyone in the Granato-Mullins fish facility for their help. This work was supported by NIH 9RO1-DA025407.