Zebrafish mutations in gart and paics identify crucial roles for de novo purine synthesis in vertebrate pigmentation and ocular development

Purines can be generated de novo or they can be recycled from sources within the body through a `salvage pathway'(Zalkin and Dixon, 1992; Zollner, 1982). De novo synthesis starts with the precursor 5-phospho-D-ribosyl-1-pyrophosphate (PRPP)and this is converted to inosine monophosphate (IMP) through ten enzymatic steps (Fig. 1). IMP then feeds into pathways leading to AMP and GMP formation, and these serve as precursors for their high energy di- or tri-phosphate forms (the `ATP pathway' and the`GTP pathway').

Although much work has been performed to identify the biochemical mechanisms underlying de novo purine biosynthesis in prokaryotic and eukaryotic systems (Zalkin and Dixon,1992), the role that purines and purinergic signaling play during embryonic development in animals has not been well characterized. Perhaps the most well-studied model for the function of the de novo purine synthesis pathway during development comes from work in Drosophila, in which mutations in prat (ppat)(Clark, 1994), ade2(pfas) (Henikoff et al.,1986b; Johnstone et al.,1985), ade3 (gart)(Henikoff et al., 1986a; Johnstone et al., 1985) and ade5 (paics) (O'Donnell et al., 2000) have been identified, as well as mutations in the GTP pathway components IMPDH (raspberry) and GMP synthase (burgundy)(Johnstone et al., 1985; Long et al., 2006). Each of these de novo pathway mutants shows a subset of developmental defects that have collectively been termed the `purine syndrome' phenotype(Tiong et al., 1989). Purine syndrome defects include a reduction in the red pteridine-derived eye pigments, bristle and leg malformations, small wings, and pupal lethality(Tiong et al., 1989). Additionally, IMPDH and GMP synthase mutants possess defects in R-cell axon pathfinding (Long et al.,2006).

Developmental roles for the de novo purine synthesis pathway in vertebrates have been elusive, however, purinergic signaling is known to have several functions during vertebrate embryonic development. Recent studies have highlighted a crucial role for retinal pigmented epithelium (RPE)-released ATP in regulating retinoblast proliferation(Martins and Pearson, 2008; Pearson et al., 2002; Pearson et al., 2005). Moreover, E-NTPDase2, an enzyme that converts ATP to ADP, has been shown to function in eye formation upstream of the eye field transcription factor network (Massé et al.,2007). Additionally, GTP is thought to serve as the precursor for the formation of each of the pigment types in the zebrafish embryo(Ziegler, 2003). Although these studies highlight important developmental roles for purine nucleotides or their products, how the de novo pathway itself functions during vertebrate embryonic development remains unclear.

With an interest in this pathway, we have characterized two recessive zebrafish mutations that affect de novo purine synthesis, gart and paics (Amsterdam et al.,2004; Gross et al.,2005). gart encodes a trifunctional enzyme that catalyzes steps 2, 3 and 5 of IMP synthesis, and paics encodes a bifunctional enzyme that catalyzes steps 6 and 7 of this process(Fig. 1). gart and paics mutants are microphthalmic and they possess pigmentation defects that affect each of the pigment cell types of the embryo: the melanin-containing melanosomes in skin melanocytes and the RPE; the guanine stack-containing iridosomes in iridophores; and the pteridine-containing pterinosomes in xanthophores. Utilizing gart and paicsmutants, and morpholino knockdown of components unique to the ATP or GTP pathways, we present data supporting a model in which an ATP pathway is required for eye growth, and a GTP pathway is required for pigmentation.

Zebrafish (Danio rerio) adults and embryos were maintained at 28.5°C on a 14-hour light/10-hour dark cycle. Animals were treated in accordance with the University of Texas at Austin provisions governing animal use and care. Mutant alleles used in this study were paics^hi2688, gart^hi3526b and adssl^hi3081.

Embryos (10-20) were homogenized in TRIzol Reagent (Invitrogen) using a 25-gauge needle. Total RNA was purified by chloroform extraction and isopropanol precipitation. cDNA was synthesized from 500 ng of total RNA using an iScript cDNA Synthesis Kit (BioRad). PCR was performed using 1.25 μL of cDNA. Primer sequences are available upon request.

Histology was performed as described in Nuckels and Gross(Nuckels and Gross, 2007).

Embryos were processed for histology and transverse 70-90 nm sections were cut through the central retina and mounted onto grids. Sections were stained with lead citrate and uranyl acetate. Images were obtained on a Philips EM 208 transmission electron microscope (80 kV) via an AMT Advantage HR 1MB digital camera. Images were processed using Adobe Photoshop CS2.

Hybridizations were performed essentially as described by Jowett and Lettice (Jowett and Lettice,1994). cDNA clones encoding PNPase, gch and dctwere provided by David Parichy (University of Washington, Seattle, WA, USA)and those encoding ccnd1 and ccnb1 were purchased from ZIRC(Eugene, OR, USA).

Melanin quantification was performed as described in Maldonado et al.(Maldonado et al., 2006). At least 20 embryos were analyzed per trial, and 2-6 trials were performed using different clutches of embryos. The data were graphed and statistical significance was calculated using a Student's t-test (GraphPad Prism).

Immunohistochemistry was performed as described in Uribe and Gross(Uribe and Gross, 2007). The following antibodies and dilutions were used: zpr1 (1:200); zn8 (1:100); zpr2(1:200); and goat anti-mouse Cy3 secondary (1:500). Imaging was performed on a Zeiss LSM5 Pascal laser scanning confocal microscope.

BrdU incorporation assays were performed as described in Lee et al.(Lee et al., 2008). Cell counts were performed on sections through the central retina. BrdU⁺nuclei were counted from sections obtained from three to five different embryos, derived from at least two different mating pairs. Averages of BrdU⁺ cells were compared with wild-type controls using a Student's t-test for statistical significance (GraphPad Prism).

TUNEL assays were performed on cryosections using a TMR red in situ cell death detection kit (Roche) and were imaged by confocal microscopy.

adss, gmps and standard control morpholinos (MOs) were purchased from Gene Tools (Philomath, OR, USA): adss-MO,5′-TCCACCCTGCACAAACACTGACGTT-3′; gmps-MO,5′-CACCTACTGACAGTGCTCACCTGAA-3′; and were injected into Oregon AB embryos. Splicing alterations were verified by RT-PCR and DNA sequencing.

FACS analysis was performed as in Bessa et al.(Bessa et al., 2008). DNA content was analyzed on a BD FACSCalibur flow cytometer using at least 25,000 cells/condition. Data analysis was performed using FlowJo software and statistical analysis was performed using a two parametric, unpaired t-test (GraphPad Prism).

Homozygous gart and paics mutant embryos possess pigmentation defects in which nearly all xanthophore-derived yellow pigmentation and iridophore-derived silver/reflective pigmentation is absent(Fig. 2). Melanophore/RPE-derived black pigmentation is present but embryos appeared lighter than their wild-type siblings. This lighter appearance could result from decreases in overall melanin levels, or from defects in melanosome distribution within mutant pigment cells. To differentiate between these possibilities we utilized a biochemical purification method to quantify melanin levels at 3.5 dpf and 5 dpf(Maldonado et al., 2006). At both times, gart and paics mutants possessed significantly less melanin (see Fig. S1 in the supplementary material).

In addition to pigmentation defects, gart and paicsmutants were also microphthalmic (Fig. 2; Table 1). paics mutants were first phenotypically identifiable at 48 hpf. Histologically, the paics mutant eye was dramatically smaller than that of a wild-type sibling, and retinal lamination and neuronal differentiation were delayed (Fig. 3A,B). gart mutants were not identifiable until ∼54 hpf and, like paics mutants, gart mutants also showed delayed retinal lamination and neuronal differentiation(Fig. 3C,D). Both gartand paics mutants remained microphthalmic at 3 dpf, although retinal lamination and neuronal differentiation appeared to have recovered(Fig. 3E-G). By 5 dpf, the mutant eye had grown considerably but was still smaller than that of wild-type siblings (Fig. 3H-J, Table 1). Lamination was normal and, histologically, all cell types appeared to be present. To confirm this latter point, immunohistochemical analyses were performed: retinal ganglion cells (RGCs) and red/green double cones were present in the mutant retinas(see Fig. S2 in the supplementary material), as well as rods, bipolar cells and amacrine cells (data not shown). RPE differentiation was also unaffected,although in some mutants clumps of morphologically abnormal RPE were occasionally detected (see Fig. S2I in the supplementary material). Examination of RGC axon growth and organization within the optic nerve and optic tract revealed no obvious defects in gart and paicsmutants (data not shown).

Finally, whereas most zygotic gart and paics mutants were embryonic lethal, we were able to rear ∼5-10% of homozygous mutants to adulthood. In these mutants, eye size was normal and pigmentation levels recovered over 1-2 months, at which point the mutants resembled wild-type and heterozygous siblings in their levels of pigmentation (see Fig. S3 in the supplementary material). No additional defects were observed in these fish.

Because paics and gart mutants were generated through retroviral insertional mutagenesis(Amsterdam et al., 2004; Golling et al., 2002; Gross et al., 2005), we first wanted to determine what effect proviral insertion had on the expression of the locus. In Fig. 4A and Fig. 5A, the approximate position of each proviral insertion has been mapped onto a schematic of the mutated gene. RT-PCR was performed on total RNA samples isolated from 3 dpf wild-type AB, wild-type/heterozygous siblings and/or homozygous mutant embryos from each line, and transcript levels were assayed using primer sets flanking and/or downstream of the proviral insert. In gart^hi3526b,the proviral insertion is located in the predicted 5′ UTR, which is located roughly 5 kb from the predicted first coding exon(Fig. 4A). Amplification of the regions between the UTR and exon 4 (Fig. 4B), exons 1 and 4, or exons 13 and 14(Fig. 4C) all indicated that the overall levels of gart were unaffected in the homozygous mutant embryos. However, amplification of the region spanning the 3′ end of the proviral insert and exon 4 revealed that in homozygous mutants, a portion of the proviral insert is transcribed and retained in the resulting mRNA(Fig. 4B), probably impairing translation and synthesis of Gart. Polyclonal antiserum derived against human GART did not recognize the zebrafish protein, so we were unable to confirm this experimentally.

Analysis of paics revealed the presence of at least two different transcripts in wild-type embryos (Fig. 5A,B), the first of which, paics^LONG, is encoded by at least 15 exons (Fig. 5A). All of the sequences encoding the conserved enzymatic domains of Paics are found in exons 7-15 of paics^LONG, and exon 7 contains a putative translational start site that would be predicted to generate a full-length Paics protein similar in size to the human ortholog. The second paics transcript, paics^SHORT(Fig. 5B), is composed of the latter nine exons of paics^LONG, in addition to a tenth noncoding first exon. The putative translational start site for paics^SHORT is located in exon 2. RT-PCR using primer sets that amplify the region between exon 4 and 9 of paics^LONG(primers 3F/4R in Fig. 5), and exon 1 and 4 of paics^SHORT (UTR/4R) indicates that both paics transcripts are expressed in wild-type embryos(Fig. 5C). Moreover, sequencing of these PCR products confirms this predicted intron/exon organization (data not shown).

The proviral insert in paics^hi2688 mutants is located in exon 7 of paics^LONG and exon 2 of paics^SHORT. RT-PCR using primer sets that assay for the presence of the proviral insert in each of the transcripts (UTR/V) revealed that the viral insert is spliced into paics^SHORT(Fig. 5D) but not paics^LONG (data not shown). Conversely, using a primer set that amplifies the region surrounding the site of proviral insertion (3F/4R)in paics^LONG, the proviral insert leads to an alteration in mRNA splicing whereby exon 7 is removed from paics^LONG(Fig. 5C), a result confirmed by DNA sequencing (data not shown). The finding that exon 7 is absent from paics^LONG enabled the utilization of primer sets within the corresponding exon 2 of paics^SHORT, but downstream of the proviral insertion (7F/4R and 7-8F/4R), to demonstrate that the presence of viral DNA in paics^SHORT does not lead to detectable changes in transcript levels of this splice form(Fig. 5D). We have not determined how much viral sequence is maintained in paics^SHORT; however, RT-PCR utilizing a primer set that amplifies exon 1 to exon 4 of paics^SHORT was not successful in homozygous mutant samples, even when PCR extension times were increased to over four minutes, suggesting that much of the virus is retained in the transcript (Fig. 5C). These data indicate that the proviral insertion disrupts paics^LONG by causing exon 7 to be spliced out, either preventing translation of paics^LONG, if exon 7 contains the translational start site, or resulting in a translational frameshift, if exon 7 is located within the endogenous Paics^LONG protein. For paics^SHORT, data indicate that the proviral insertion does not affect transcript levels; rather, it probably prevents translation of paics^SHORT. Polyclonal antiserum derived against human PAICS did not recognize the zebrafish protein, so we were unable to confirm this experimentally.

To begin to understand the developmental defects in gart and paics mutants, we first wanted to determine their expression profiles(Fig. 6A,B). Maternal gart expression was detected by RT-PCR at the 2-cell stage and maintained expression was detected at 1, 3 and 6 dpf(Fig. 6A). To assay paics expression, and to differentiate between the two paicssplice forms, we utilized a primer set that amplifies the region between exons 6 and 12 (6F/12R), which will only amplify paics^LONG, and a primer set that amplifies the region from the noncoding exon 1 (intron 6 in paics^LONG) to exon 7 (UTR/12R) that will only amplify paics^SHORT. Interestingly, only paics^SHORT was provided maternally, whereas maintained zygotic expression of both splice forms was detected at 1, 3 and 6 dpf(Fig. 6B). gart and paics are highly expressed within the eye, central nervous system and pigment cells throughout development(Thisse and Thisse, 2004)(data not shown).

Given the maternal expression of both gart and paics, we reasoned that more severe developmental phenotypes were probably masked in zygotic mutants due to the presence of maternal transcripts. To test this hypothesis, we crossed homozygous gart and paics females to homozygous or heterozygous males to generate maternal-zygotic mutants(gart^MZ and paics^MZ), and maternal-effect mutants (gart^ME and paics^ME), which lacked both maternal mRNA and protein and a functional maternal copy of each gene. These latter mutants enabled us to assay whether the zygotic paternal genome alone could direct normal embryonic development. gart^ME mutants were drastically delayed in development when compared with wild-type embryos derived from heterozygous incrosses (Fig. 6C,D). Most gart^ME mutants lived for ∼36-48 hpf and then began to degenerate, dying by 3 dpf. Despite this high degree of lethality, we were able to rear a small number of gart^ME mutants to adulthood. These fish were runted over their first 2-3 months but eventually grew to near wild-type size, showing normal pigmentation and no overt defects(data not shown). All gart^MZ mutants arrested in development at ∼16-18 hpf, displaying severe defects in anterior development (Fig. 6E). Surprisingly, paics^ME mutants developed relatively normally throughout the first 2-3 days of embryonic development(Fig. 6F), but after this point they began to degenerate and most died 5-7 dpf. paics^MEmutants resembled mutants derived from heterozygous mothers, indicating that zygotic expression of the paternal copy of paics was unable to rescue pigment formation or eye growth. Beyond these defects however, no additional phenotypes were observed prior to the onset of degeneration in paics^ME mutants. Like gart^ME mutants,a small number of paics^ME fish were viable and although they were runted during the first few months, as adults they were also normal(data not shown). paics^MZ mutants survived longer than gart^MZ mutants, developing to ∼60-72 hpf before dying. paics^MZ mutants were small, nearly albino and possessed cardiac edemas and widespread CNS degeneration(Fig. 6G). RT-PCR using total RNA isolated from 2-cell-stage paics^MZ embryos revealed that, as expected, paics^MZ mutants lacked expression of maternally provided paics^SHORT transcripts(Fig. 6B).

We next wanted to determine the underlying cellular mechanisms of the pigmentation phenotypes in gart and paics mutants. Two possible models to explain these defects are: (1) pigment cells are not properly specified or are not able to properly migrate and generate the larval pigment pattern; or (2) pigment cell specification and migration are normal,but these cells are impaired in their biochemical ability to synthesize pigments. Quantification of melanin levels in gart and paicsmutants supports the latter of these models (see Fig. S1 in the supplementary material). Histological and TEM imaging of the pigmented iris and RPE enabled a rapid analysis of each pigment cell type(Fig. 7A,C). Additionally,examination of the posterior of the eye enabled a secondary analysis of melanosomes in the RPE and of iridophores within the sclera(Fig. 7B). In wild-type eyes, a dense layer of melanosome-containing RPE is obvious and within the iris one can observe numerous round pterinosomes and several iridosomes, which appear reflective due to the highly ordered guanine stacks(Fig. 7A). TEM analysis highlights the presence of these organelles: melanosomes in the RPE, and pterinosomes and guanine stacks in iridosomes within the iris(Fig. 7C). At the posterior of the eye, the RPE contains melanosomes and the sclera contains interspersed iridosomes (Fig. 7B). gart and paics mutants possess each of these pigment cell types, supporting the hypothesis that their specification and differentiation are unaffected in the mutants. However, the number of pigment-containing pterinosomes and iridosomes appears to be reduced in the iris(Fig. 7D,F,G,I) and to a lesser extent in the RPE and sclera (Fig. 7E,H) of the mutants, suggesting that in the absence of sufficient pigment, the formation or integrity of these pigment-containing organelles might be compromised.

To further analyze pigmentation defects in gart and paicsmutants, we utilized in situ hybridization markers for pigment cell types to determine if their overall number or distribution was compromised. dopachrome tautomerase (dct) encodes a protein that converts dopachrome to 5,6-dihydroxyindole-2-carboxylic acid (DHICA) and 5,6-dihydroxyindole (DHI) in melanosomes(Fig. 11). Using dctas a marker for melanosomes, no obvious differences in the number or distribution of melanocytes within the dermis, or in the intensity of expression in the RPE, were noted in gart or paics mutants(Fig. 7J,K). Similarly, using GTP cyclohydrolase I (gch, gch2 - ZFIN), which encodes a protein that converts GTP to 6-pyruvoyl-H₄ pterin synthase, as a marker for xanthophores (Fig. 7L,M), and purine nucleotide phosphorylase (PNPase,xdh - ZFIN), which encodes a protein that converts inosine to hypoxanthine, as a marker for iridophores(Fig. 7N,O), revealed no obvious differences in the number or distribution of these cell types in the mutants.

ATP levels influence cell proliferation in a number of contexts, and recent studies have highlighted a crucial role for RPE-released ATP in regulating retinoblast proliferation in the developing eye(Martins and Pearson, 2008; Pearson et al., 2002; Pearson et al., 2005). The findings that the proteins encoded by gart and paics provide IMP, the precursor for ATP (Fig. 1), and gart and paics mutants are microphthalmic (Figs 2,3),suggested that these proteins might play a role in regulating eye growth. We envisioned two possible mechanisms that could underlie the microphthalmia in the mutants: (1) defects in retinoblast proliferation, whereby the cell cycle in retinoblasts is prolonged or cell cycle exit is compromised, could lead to a smaller eye; and/or (2) increases in apoptosis could lead to a smaller eye. To test the first of these hypotheses we utilized BrdU incorporation assays to compare the percentage and location of proliferative cells between wild-type and mutant eyes (Fig. 8). BrdU incorporation from 48 to 50 hpf indicated that 20.3% of cells in the wild-type retina were in S phase (Fig. 8H), and most were located at the retinal periphery(Fig. 8A). By comparison, 40.9%of cells in the paics retina were in S phase(Fig. 8H), and these cells were ectopically located throughout the central retina(Fig. 8B). BrdU incorporation from 54 to 56 hpf indicated that 13.9% of cells in the wild-type retina were in S phase (Fig. 8H), and nearly all of these were located at the ciliary marginal zones (CMZs; Fig. 8C). In the gartretina, 25.5% of cells were in S phase(Fig. 8H) and, like paics, they were ectopically located within the central retina(Fig. 8D). BrdU incorporation from 96 to 98 hpf indicated that 2.9% of cells in the wild-type retina were in S phase (Fig. 8H), and again,all of these were located at the retinal periphery(Fig. 8E). By comparison, 5.5%of the paics retina and 6.6% of the gart retina was BrdU-positive (Fig. 8H);however, all of these S-phase cells were confined to the retinal periphery(Fig. 8F,G).

Histological analysis did not identify pyknotic nuclei in gart and paics mutant eyes at the time at which microphthalmia was first identifiable (Fig. 3A-D). To further test the latter hypothesis for microphthalmia, we utilized TUNEL staining to identify apoptotic nuclei. As reported by others(Cole and Ross, 2001; Biehlmaier et al., 2001),apoptotic nuclei in the wild-type retina were rare(Fig. 8I,K). No differences in the number of TUNEL⁺ cells were detected between wild-type siblings and paics mutants at 48 hpf (Fig. 8I,J) or gart mutants at 54 hpf(Fig. 8K,L), nor were differences detected at later time points (data not shown). Thus, apoptosis is not likely to contribute to the microphthalmia in gart and paics mutants. An absence of increased levels of cell death, combined with an increase in the proportion of the mutant retina that is BrdU⁺, supports the hypothesis that defects in cell cycle progression lead to microphthalmia in gart and paicsmutants.

IMP serves as a precursor for both ATP and GTP biosynthesis(Fig. 11). ATP can modulate retinoblast proliferation and cell cycle duration within the developing vertebrate eye (Pearson et al.,2002; Pearson et al.,2005), whereas components downstream of GTP are involved in zebrafish pigmentation (Le Guyader et al.,2005; Ziegler,2003), suggesting independent functions for the two purines during development. Since Gart and Paics function in IMP synthesis upstream of the ATP and GTP synthesis pathways, it is not surprising that gart and paics mutants display both microphthalmia and pigmentation phenotypes. We wanted to test this model predicting that the ATP and GTP pathways were separable into distinct developmental roles, and we hypothesized that if one perturbed only the ATP synthesis pathway, microphthalmia with normal pigmentation should result. Conversely, if one perturbed only the GTP synthesis pathway, pigmentation defects should result, but eye size would be unaffected.

To test the hypothesis that the ATP pathway is involved in regulating eye growth, we designed a morpholino antisense oligo targeting the intron 1/exon 2 splice junction of adenylosuccinate synthase (adss; adss-MO; Fig. 9C). adss encodes the protein that catalyzes the first reaction in the ATP pathway, converting IMP to adenylosuccinate (Fig. 11). Injection of 3.33 ng of adss-MO resulted in an absence of adss transcripts in adss morphants', probably through nonsense-mediated mRNA decay (see Fig. S4A in the supplementary material). As predicted, adss morphants were microphthalmic, with eye size reduced by almost 30% (Fig. 9D,E; Table 2). Importantly, however,pigmentation was unaffected; xanthophore, iridophore and melanophore/RPE pigmentation all appeared similar to that observed in control MO-injected siblings. Additionally, an insertional mutation affecting adss has been reported and, phenotypically, adss^hi3081 mutants have microphthalmia but normal pigmentation(Amsterdam et al., 2004) (data not shown).

To test the hypothesis that GTP feeds into a pathway required for pigmentation and not eye growth, we utilized a morpholino targeting the exon 2/intron 2 splice junction of GMP synthase (gmps; gmps-MO; Fig. 9F), which encodes the enzyme that converts XMP to GMP (Fig. 11). Injection of 3.9 ng of gmps-MO disrupted splicing of exon 2 of gmps, as verified by RT-PCR (see Fig. S4B in the supplementary material) and DNA sequencing (see Fig. S4C in the supplementary material). All gmps morphants possessed obvious pigmentation defects that resembled zygotic gart and paics mutants. Melanin pigmentation was reduced at 36 hpf (data not shown), and there was a near absence of xanthophore and iridophore pigmentation at 3 dpf(Fig. 9G,H). Eye size was largely unaffected, although the eyes of gmps morphants were 4%smaller than those in control-injected embryos(Table 2).

We next wanted to determine the cellular mechanism through which the ATP pathway regulated eye growth. Like gart and paics mutants, adss morphants possessed a significantly higher proportion of BrdU⁺ cells in their retinas when compared with control MO-injected embryos (Fig. 10A), and many of these were ectopically located within the central retina (data not shown). These results suggest at least two possible roles for the ATP pathway in regulating the cell cycle in retinoblasts: it could be required for normal S-phase progression; and/or it could be required for S-phase exit and G2/M progression. To test these models, we utilized flow cytometry to quantify DNA content and compare the proportions of retinal cells in each phase of the cell cycle between control MO-injected and adss morphant retinas. Quantification at 48 hpf and 72 hpf revealed significant changes in the proportion of cells in the G1 and S phases of the cell cycle in the adss morphant retina, when compared with controls(Fig. 10B,C). At both times,the proportion of cells in the S phase was increased in the adssmorphant retina, whereas the proportion of cells in the G1 phase was decreased. The proportion of cells in the G2/M phases was also decreased in the adss morphant retina, but these differences were not statistically significant at either time point (P<0.05). cyclin d1 (ccnd1) serves as a marker of the G1/S transition,and, generally, of proliferative cells in the retina(Levine and Green, 2004). In control MO-injected embryos, ccnd1 is localized to the zones of proliferative cells at the retinal periphery at 48 hpf(Fig. 10D). In adssmorphants, ccnd1 remained expressed within the central retina(Fig. 10D), indicating that these cells had not exited the cell cycle. cyclin b1(ccnb1), a marker of the G2/M transition(Nasmyth, 1996), was distributed normally in adss morphants, however(Fig. 10D). An expansion of ccnd1 into the central retina of adss morphants correlates with the increased number of BrdU⁺ cells in the central retina of gart and paics mutants, and adss morphants, and it correlates with the increased proportion of S-phase cells found in the adss morphant retina. When combined, these data support a model in which the ATP pathway is required in retinoblasts for their timely progression through S phase. When the pathway is impaired, S phase is prolonged, the rate of cell division decreases and microphthalmia results.

Using zebrafish mutations and/or morpholino-induced loss of function in gart, paics, adss and gmps, our results support a model in which de novo purine synthesis is required for two diverse developmental functions: an ATP pathway mediates eye growth by regulating retinoblast proliferation and cell cycle progression, and a GTP pathway mediates embryonic pigmentation, probably serving as a precursor for the biosynthesis of each of the pigment molecules (Fig. 11). Zygotic loss of function for gart and paicsaffects de novo purine biosynthesis upstream of the bifurcation of the ATP and GTP pathways, and therefore gart and paics mutants show defects in both eye growth and pigmentation. adss mutants and morphants are only affected in the ATP pathway, and this results in eye growth defects resembling those in paics and gart, but with normal pigmentation. Conversely, gmps morphants are only affected in the GTP pathway and this results in paics and gart-like pigmentation defects but relatively normal eye size. Our data do not exclude the possibility that the GTP pathway also contributes to eye growth because eye size was mildly smaller in gmps morphants; however, this contribution is likely to be far less significant than that of the ATP portion given the substantial decrease in eye size observed in adss mutants and morphants. Finally, we demonstrate a crucial role for maternal gartand paics function as both maternal-zygotic and maternal-effect mutants display more severe developmental defects than their zygotic counterparts. These results highlight a crucial role for de novo purine synthesis during vertebrate development and they further subdivide these developmental roles into those that depend on ATP- or GTP-specific pathways.

Our results, when coupled with those from several previous reports, provide possible mechanistic insights into the phenotypes observed in gart, paics,adss and gmps mutants and morphants and into how de novo purine biosynthesis mediates such diverse developmental functions. With regard to retinoblast proliferation and cell cycle progression, it has been demonstrated that ATP, released from the RPE, can nonautonomously modulate the rate of the retinoblast cell cycle and thereby regulate overall rates of eye growth(Martins and Pearson, 2008; Pearson et al., 2002; Pearson et al., 2005). Using chicken retinal explants and pharmacological manipulations, ATP was shown to mechanistically act by modulating the rate of mitosis(Pearson et al., 2002) and/or the rate of DNA synthesis (Sugioka et al.,1999; Pearson et al.,2005). Given the key enzymatic roles that the proteins encoded by the gart, paics and adss loci play in the biosynthesis of ATP, the most parsimonious explanation for the cellular mechanism underlying the microphthalmia observed in gart, paics and adsszebrafish mutants and/or morphants is that, in the absence of sufficient ATP,DNA synthesis during S phase is delayed. This leads to a net lengthening of the cell cycle in each retinoblast, as well as a higher overall proportion of retinoblasts within the developing eye remaining in S phase. Over time, each round of cell division takes longer, leading to fewer total cells in the retina and, thus, to microphthalmia. Homozygous gart and paics mutants eventually recover and, as adults, eye size is normal,which is likely to reflect the activity of the nucleotide salvage pathway.

With respect to the pigmentation defects in gart and paics mutants and gmps morphants, our results support a model whereby Gart- and Paics-dependent IMP formation feeds into a Gmps-dependent GTP pathway that impinges on the synthesis of the three pigment types (Fig. 11). GTP is thought to serve as the precursor for synthesis of each of these pigment molecules (Ziegler, 2003), and in the absence of sufficient IMP levels (gart and paics) or the ability to convert XMP to GMP (gmps morphants) pigment molecules are absent or deficient. The finding that pigmentation recovers in homozygous adult gart and paics mutants again probably reflects the activity of the salvage pathway.

Interestingly, developmental phenotypes in gart and paicsmutants are different than those resulting from deficiencies in de novo pyrimidine biosynthesis (Link et al.,2001; Willer et al.,2005). In the pyrimidine biosynthesis mutant perplexed,the gene encoding Cad, an enzyme that catalyzes the first three steps of de novo pyrimidine synthesis, is disrupted(Willer et al., 2005). perplexed mutants have defects in both proliferation and neuronal differentiation within the retina, and they possess defects in jaw and pectoral fin formation. Pigmentation in these mutants is normal, however. Additionally, de novo pyrimidine mutants in Drosophila often possess defects in wing formation (Rawls and Fristrom, 1975; Falk,1976), whereas mutants in de novo purine biosynthesis have defects in eye color (Tiong et al.,1989). These studies, when combined with our results, support a model in which de novo purine biosynthesis and de novo pyrimidine biosynthesis play important but tissue-specific roles during animal development.

Two human congenital disorders have been described that affect the de novo purine synthesis pathway. The first, adenylosuccinate lyase deficiency,disrupts adenylosuccinate lyase (ADSL), whose protein product catalyzes two stages of de novo purine synthesis: (1) the eighth step in IMP synthesis (Fig. 1); and(2) the conversion of adenylosuccinate to AMP in the ATP pathway(Fig. 11)(Marie et al., 1999; Spiegel et al., 2006; Stone et al., 1992). Patients with ADSL mutations have severe mental retardation, epilepsy and autistic features, although no ocular phenotypes have been reported. The second disorder, AICA-ribosiduria, results from a deficiency in ATIC,whose protein product catalyzes the ninth and tenth reactions of IMP synthesis(Fig. 1). The only reported patient with this disorder had dysmorphic features of the face and body,mental retardation, epilepsy and congenital blindness(Marie et al., 2004). The molecular underpinnings of these disorders are not well understood, however,it has been demonstrated that the substrates for each enzyme accumulate in afflicted patients, suggesting that these intermediates might be deleterious for neuronal development and function. Zebrafish models of de novo purine synthesis deficiencies might thus be useful in elucidating how defects in this pathway can lead to defects in human embryonic development and CNS function.