hnRNP I is required to generate the Ca²⁺ signal that causes egg activation in zebrafish

Egg activation ensues immediately after fertilization or egg laying,initiating a variety of cellular processes within the egg. One early event in egg activation, known as cortical granule exocytosis (CGE), modifies the vitelline membrane (also known as the zona pellucida or chorion) of the egg,and this is crucial to block polyspermy and to create a barrier to protect the developing embryo. During CGE, cortical granules (CGs), which are stimulus-dependent secretory vesicles located at the egg cortex, exocytose their contents, including enzymes and structural proteins into the perivitelline space, resulting in a permanently modified vitelline membrane(Tsaadon et al., 2006; Wassarman et al., 2001; Wessel et al., 2001).

Egg activation is initiated by a rise in cytosolic Ca²⁺. The release of Ca²⁺ from the endoplasmic reticulum (ER) is typically triggered by fertilization and requires inositol 1,4,5-trisphosphate(IP₃) (Whitaker,2006; Runft et al.,2002; Stricker,1999). IP₃ is generated by activation of the phosphoinositide pathway, whereby enzymes of the phospholipase C (PLC) family cleave the lipid phosphatidylinositol 4,5-bisphosphate (PIP₂),generating IP₃. IP₃ subsequently serves as a Ca²⁺-release messenger at egg activation to stimulate the release of Ca²⁺ from the ER (Tsaadon et al., 2006).

In zebrafish, a similar rise of cytosolic Ca²⁺ is observed upon egg activation, although egg activation in zebrafish does not require fertilization, but instead depends on a hypotonic environment, which arises when eggs are laid in normal water conditions(Hart and Fluck, 1995; Hart and Yu, 1980; Iwamatsu et al., 1997; Lee et al., 1999). As in other species, CGE in zebrafish occurs as a first response to elevated cytoplasmic Ca²⁺ levels during egg activation. CGE is required to modify the initially soft chorion, expanding and hardening it, which then serves as an important protective structure for the developing fish embryo.

We performed a maternal-effect mutant screen in the zebrafish to identify novel regulatory components of the egg activation process(Dosch et al., 2004; Wagner et al., 2004). We report here the characterization of a zebrafish maternal-effect mutant, brom bones, which is defective in egg activation. We show that the eggs of brom bones mutant females display multiple defects during egg activation, including a failure in CGE, cytoplasmic segregation, and a reduction in or absence of the Ca²⁺ wave that initiates egg activation. Providing Ca²⁺ or the Ca²⁺-release messenger, IP₃, into brom bones mutant eggs rescues the egg activation defects, indicating that brom bones is a regulator of IP₃-mediated Ca²⁺ release. Interestingly, when the egg activation defects are incompletely penetrant, embryos from brom bones mutant females develop to 1 day post fertilization (dpf) and display a ventralization of the embryonic axis. Dorsal axis formation requires a cortical microtubule array that translocates a vegetally localized dorsal determinant asymmetrically to establish the dorsoventral axis. We show that in brom bones mutant embryos this cortical microtubule array is disorganized due to persisting, non-exocytosed CGs from earlier egg activation defects. We positionally cloned the brom bones maternal-effect mutant gene and found that it encodes hnRNP I, an RNA-binding protein belonging to the heterogeneous nuclear ribonucleoprotein (hnRNP) family. Importantly, our results demonstrate a new function for hnRNP I in vertebrate oogenesis, i.e. regulating the IP₃-mediated Ca²⁺ rise that activates the egg for zygotic development.

Whole-mount in situ hybridization and in situ hybridization on ovary sections were carried out essentially as described(Schulte-Merker et al., 1992; Shimamura et al., 1994). In situ probes were boz (Yamanaka et al., 1998), squint(Erter et al., 1998; Rebagliati et al., 1998), chordin (Miller-Bertoglio et al.,1997), gata2 (Detrich et al., 1995), brl(Suzuki et al., 2000), Vg1 (Helde and Grunwald,1993), mago nashi(Pozzoli et al., 2004), dazl (Maegawa et al.,1999) and brom bones (see brom bonescloning).

Whole-mount antibody staining was done essentially as described(Kelly et al., 2000), using aβ-catenin antibody (Schneider et al.,1996) with the ABC Kit (Vector Laboratories). To detect F-actin,rhodamine-phalloidin (Invitrogen) was used, as described(Becker and Hart, 1999). Microtubules were detected with anti-β-tubulin monoclonal antibody(Developmental Studies Hybridoma Bank)(Topczewski and Solnica-Krezel,1999).

Cortical granules were visualized by labeling embryos with 50 μg/ml FITC-conjugated Maclura pomifera agglutinin (EY Laboratories, San Mateo, CA, USA). Images were taken from an MZ12.5 stereomicroscope (Leica)with a ColorSNAP-cf digital camera (Photometrics) or a Zeiss LSM510 confocal microscope, and processed using Adobe Photoshop.

Fixation of the embryos for histology, embedding in JB-4 plastic(Polysciences) and sectioning were performed as described previously(Pack et al., 1996). The sections were processed for Toluidine Blue staining(Selman et al., 1993). For electron microscopy, dissected ovaries and squeezed eggs from wild-type and brom bones mutant females, selected based on strong egg activation defects among their progeny, were fixed and processed according to standard protocols (Hayat, 1986). Images were captured at various magnifications using a Hamamatsu CCD camera and AMT 12-HR software.

Western blotting was performed according to standard protocols. Antibodies used were anti-β-catenin (polyclonal, 1:1000), anti-β-tubulin(1:1000) and HPR-conjugated secondary antibodies (1:4000, Amersham). Blots were developed with the ECL Kit (Amersham).

Mutant or wild-type sibling adult female DNA was pooled for bulk segregant analysis. Simple sequence length polymorphism (SSLP) markers were identified from the zebrafish genetic map (Shimoda et al., 1999) or generated from simple-sequence repeats found within BAC DKEY-57I11 [(CA)n1 and (CA)n2]. Open reading frames (ORFs) of candidates in BACs DKEY-57I11, DKEY-119N4 and CH211-253D24 were predicted using Genscan(http://genes.mit.edu/GENSCAN.html)and then confirmed by BLAST against the NCBI zebrafish EST database. The ORFs of the candidates were analyzed by complete sequencing of cDNAs isolated from adult mutant and wild-type ovary or brain tissue (prg2 was only isolated from brain, as it was not detectable in the ovary). The point mutation found in the hnrnp I cDNA was substantiated by sequencing PCR products from genomic DNA of mutant and wild-type females, as well as the original ENU-mutagenized male from which this mutant chromosome was derived. For genotyping, genomic DNA was PCR-amplified with primers 5′-TAACATTAAACAGTCTTTAGATCGA-3′ and 5′-GCTGCCATTGTCTGTGCATCAAAGG-3′, and digested with ClaI to detect a restriction fragment length polymorphism generated by the mutation. The full-length sequence of the brom bones transcript was further confirmed by 5′ rapid amplification of cDNA ends using the GeneRacer Kit (Invitrogen) and by sequencing ESTs that encode the hnrnp I transcript, CD015134 and CA470908. Full-length hnrnp I coding region was amplified and cloned into the CS2-MT vector to obtain pCS2-hnRNP I,to make brom bones antisense RNA probe.

Embryos obtained from wild-type and brom bones mutant females were split into two groups immediately after fertilization. One subset of embryos was placed at 20°C for 2 hours and then shifted back to 28.5°C. Non-treated control groups were placed at 28.5°C. The phenotype was scored at 1 dpf.

Unactivated eggs were collected by squeezing anesthetized females and these were then injected with aequorin (supplied by Dr Osamu Shimomura) or AquaLite(Invitrogen), as described (Lee et al.,1999). Data were collected using a photon imaging microscope (PIM)as previously reported (Lee et al.,1999; Webb et al.,1997).

Unactivated eggs were collected and kept in Hank's buffered saline containing 0.5% BSA pH 7.0 to delay egg activation(Sakai et al., 1997). For Ca²⁺ injection, 2 nl of 50 mM CaCl₂ was injected into the center of unactivated eggs. Control eggs were injected with 50 mM KCl solution. For IP₃ injection, 2 nl of IP₃ (10 μg/μl in 125 mM KCl; BIOMOL Research Laboratories, CA-430) was injected into the center of unactivated eggs. Control eggs were injected with 125 mM KCl solution only. Following injection, eggs were kept in Hank's buffered saline containing 0.5% BSA for 15 minutes and then transferred into E3 medium for activation. Chorion elevation and cytoplasmic segregation in the injected eggs were monitored 1 hour after egg activation.

Embryos from brom bones maternal-effect mutant females (henceforth called brom bones mutant embryos) display two major phenotypes: a variably penetrant failure in egg activation and, in those eggs that successfully activate, a subsequent ventralization of the embryonic axis at 1 dpf (Fig. 1)(Wagner et al., 2004). A summary of the relative proportion of these two phenotypes within representative single clutches is shown in Table 1.

Egg activation in zebrafish is marked by elevation of the chorion and by segregation of the cytoplasm, which is initially integrated within the yolk,to the animal pole to form the single-cell blastodisc(Fernandez et al., 2006; Hart and Fluck, 1995). Chorion elevation is a consequence of CGE and is completed within minutes of egg activation (Becker and Hart,1999). The chorions of brom bones mutant eggs did not fully elevate upon egg activation (Fig. 1A,B, arrows), and remained soft rather than becoming rigid. Cytoplasmic segregation was also blocked in brom bones mutant eggs. Normally, cytoplasm begins to segregate from the yolk by ∼10 minutes post egg activation (mpa) and is completed during early cleavage stages(Fig. 1A)(Hart and Fluck, 1995). The most strongly affected brom bones mutant eggs lacked obvious cytoplasmic segregation to the animal pole(Fig. 1B). Consequently, most of the cytoplasm remained in the yolk, causing the eggs to be mildly opaque,rather than becoming translucent like wild-type eggs, following cytoplasmic segregation. These brom bones mutant eggs had very small blastodiscs(Fig. 1B, asterisk) and showed no cell cleavage. Moderately affected eggs showed reduced and delayed cytoplasmic segregation, generating a moderately sized blastodisc at the 1-cell stage. During cleavage stages, these mutant embryos exhibited cleavage defects (Fig. 1D, arrow),displayed large cytoplasmic inclusions in the yolk(Fig. 1D, arrowheads) and died before 1 dpf. Mildly affected embryos exhibited slightly delayed and less effective cytoplasmic streaming to the blastodisc. Small cytoplasmic inclusions were apparent in the yolk of mutant embryos even during gastrula stages (Fig. 1F, arrowheads). Cellular cleavage and gastrulation appeared normal in these mildly affected embryos (Fig. 1F).

brom bones mutant embryos that were mildly affected in egg activation developed a variably penetrant ventralized phenotype at 1 dpf, as previously reported (Wagner et al.,2004). The most strongly affected embryos appeared radially ventralized, with little or no trunk development and only a few tail somites(Fig. 1H). Intermediate strength embryos lacked a notochord and displayed strongly reduced or absent head structures (Fig. 1I,J). Mildly ventralized mutant embryos exhibited a small head and eyes(Fig. 1K). The fraction of ventralized embryos from different homozygous mutant females varied(Table 1). Unaffected mutant embryos survived to adulthood (data not shown).

We first examined the egg activation defects in brom bonesmutants. As in other animals, CGE in zebrafish is among the first cellular responses of egg activation. CGE consists of a cascade of steps that normally complete within 5 minutes after an egg is activated. These steps include the movement of granules to the plasma membrane (PM), fusion of the CG membranes with the PM, release of the granule contents, and conclude with membrane retrieval through compensatory endocytosis(Becker and Hart, 1999; Donovan and Hart, 1986). Dynamic changes in the actin cytoskeleton are crucial to allow the CGs to fuse with the PM, and subsequently retrieve the vacant CG membrane following exocytosis (Becker and Hart,1996; Becker and Hart,1999; Sokac et al.,2003). Since the CGE-driven processes of chorion elevation and hardening are impaired in brom bones mutant eggs, we investigated whether CGE was defective by examining actin cytoskeletal dynamics.

Zebrafish eggs contain a cortical, sub-PM F-actin network that acts as a barrier to CGE prior to egg activation(Becker and Hart, 1996; Becker and Hart, 1999). Using rhodamine-phalloidin (RhPh) to label F-actin, we found that brom bones eggs displayed a relatively normal distribution of cortical F-actin prior to egg activation (Fig. 2A,B). Following egg activation, cortical F-actin reorganizes to allow exocytosis of CGs and then acts in compensatory endocytosis to retrieve the emptied CG membrane (CG crypt) from the egg surface. In wild-type eggs,large-scale CGE occurs within 2 mpa, which is evident by numerous patches of intense RhPh staining at the egg surface(Fig. 2C). These F-actin patches are cortical crypts at the egg surface undergoing F-actin-mediated compensatory endocytosis to retrieve the vacant granule membrane. In striking contrast, mutant eggs at 2 mpa displayed only a few patches of intense F-actin at the PM (Fig. 2D). A small number of dark circular profiles were seen on the egg surface that lacked or showed reduced RhPh fluorescence, resembling the sites where CGs are in fusion with the PM (Becker and Hart,1999; Sokac et al.,2003). The F-actin distribution on the mutant egg surface at 2 mpa indicates that large-scale CGE is compromised in brom bones mutant eggs.

To determine if CGE is simply delayed or slower in brom bonesmutant eggs, we examined F-actin organization at 5 and 30 mpa. Wild-type eggs at 5 mpa exhibited no F-actin patches and instead F-actin was distributed evenly at the egg cortex (Fig. 2E). This speckled appearance of F-actin indicates that CGE had been completed and all CG crypt membranes were retrieved by endocytosis,leaving the egg surface membrane completely remodeled(Becker and Hart, 1999; Sokac et al., 2003). In mutant eggs at 5 mpa, numerous intense circular profiles of F-actin were present on the egg surface (Fig. 2F,arrows). Surprisingly, these cortical circular profiles of F-actin persisted in mutant eggs up to 30 mpa (Fig. 2H). These circular profiles might represent non-exocytosed CGs.

To determine if CGE is blocked in brom bones mutants and if the circular profiles represent non-exocytosed CGs, eggs at 30 mpa were double-labeled with RhPh and FITC-conjugated Maclura pomiferaagglutinin (MPA), which specifically labels CG contents(Becker and Hart, 1999). Indeed, the majority of circular actin profiles observed in brom bones mutant eggs at 30 mpa contained MPA-labeled granule contents(Fig. 2G-L). These non-exocytosed granules were located either beneath(Fig. 2M, arrowheads) or within(Fig. 2M, arrow) a cortical F-actin meshwork. These results show that in strongly affected brom bones mutant eggs, CGE is largely blocked, thus probably causing the failure in chorion elevation and hardening(Fig. 1A,B).

CGE is initiated by a wave of elevated cytoplasmic Ca²⁺(Runft et al., 2002), which is triggered by interaction between IP₃ and its receptors on the ER,the intracellular Ca²⁺ storage sites. The increased cytoplasmic Ca²⁺ initiates CGE at least in part by leading to the reorganization of the cortical actin cytoskeleton barrier(Tsaadon et al., 2006). In zebrafish, a single wave of cytosolic Ca²⁺ traverses the egg during egg activation (Lee et al.,1999). CGE is accelerated when actin is depolymerized, whereas CGE is blocked when actin depolymerization is inhibited(Wolenski and Hart, 1988; Becker and Hart, 1999),suggesting that CGE is accomplished through Ca²⁺-mediated actin reorganization. To determine if an altered Ca²⁺ wave could be the cause of impaired CGE and chorion elevation in brom bones mutant eggs, we examined the Ca²⁺ wave triggered at egg activation in brom bones eggs.

To visualize Ca²⁺ in vivo, a recombinant Ca²⁺-sensitive bioluminescent protein was injected into the center of unactivated eggs and allowed to diffuse evenly throughout the egg for 20 minutes. The eggs were activated and bioluminescence was recorded to measure Ca²⁺ levels, followed by the acquisition of bright-field photos to examine the extent of chorion elevation(Lee et al., 1999). As shown in Fig. 3A, brom bonesmutant eggs exhibited a defect in the increase in Ca²⁺ levels,which correlated with the extent of chorion elevation. Mutants with no detectable increase in Ca²⁺ upon egg activation exhibited no obvious chorion elevation (n=6 eggs), whereas mutant eggs with a small increase in Ca²⁺ showed intermediate chorion elevation(n=7). Mutant eggs with relatively normal Ca²⁺ levels displayed normal chorion elevation (n=15). These results suggest that the egg activation defects in brom bones mutant eggs are caused by the defective increase in Ca²⁺ levels.

To test if the defective increase in Ca²⁺ levels is the cause of the egg activation defects, we provided Ca²⁺ to brom bonesmutant eggs and determined whether it could rescue the egg activation defects. We found that brom bones mutant eggs could be activated by Ca²⁺ injection, in contrast to control KCl-injected mutant eggs. In three individual experiments the percentage of activated eggs, as indicated by chorion elevation and blastodisc formation, was significantly increased in the Ca²⁺-injected groups when compared with control groups(Fig. 3B; the bright-field live egg images show the I to V scoring series). These results indicate that defective egg activation in brom bones mutant eggs is caused by an impaired increase in Ca²⁺.

We next examined whether IP₃, the major signal that releases Ca²⁺, can rescue the brom bones egg activation defects. We found that brom bones mutant eggs could be activated by IP₃ injection, in contrast to control buffer-injected mutant eggs(Fig. 3C,D). The percentage of mutant eggs that showed both normal chorion elevation and normal blastodisc formation (data not shown) was significantly increased in IP₃-injected groups when compared with control groups (47% versus 0% in experiment 1 and 81% versus 15% in experiment 2)(Fig. 3E).

Taken together, we found that the egg activation defects in brom bones mutant eggs are correlated with a failure in Ca²⁺elevation and that these defects can be rescued by providing Ca²⁺or IP₃ to mutant eggs prior to egg activation. These results are consistent with a model whereby insufficient IP₃ is generated in strongly affected brom bones mutant eggs following entry into spawning medium, causing reduced or no Ca²⁺ elevation, which leads to little or no CGE and thus a lack of chorion elevation and cytoplasmic segregation. Thus, all of the machinery downstream of IP₃ necessary for egg activation is present and functional in brom bones mutant eggs, but fails to be activated owing to an apparent insufficiency of IP₃.

We examined histologically and by electron microscopy the ovaries of brom bones mutant females to investigate whether a morphological defect in oogenesis could account for the egg activation defects. In brom bones mutant oocytes, CG distribution and morphology looks similar to wild-type oocytes (Fig. 4A,B,arrows). At stage IV of oogenesis, the oocyte cortex, including CGs and yolk globules, appeared similar in brom bones and wild-type oocytes(Fig. 4C,D). These results suggest that a gross morphological defect in oogenesis in brom bonesmutant females does not account for the egg activation defect.

We next investigated the dorsoventral axis defect in brom bonesmutant embryos. In zebrafish and Xenopus embryos, one of the earliest molecular steps in the establishment of the dorsoventral axis is the translocation of β-catenin, a key mediator of canonical Wnt signaling,into the nuclei of dorsal marginal blastomeres at the midblastula stage(Heasman et al., 1994; Schneider et al., 1996; Fagotto et al., 1996). Nuclear-localized β-catenin in turn acts as a transcriptional effector to activate dorsal-specific genes such as bozozok (boz; dharma - Zebrafish Information Network), squint(ndr1 - Zebrafish Information Network) and chordin (reviewed by Schier and Talbot, 2005). These dorsal genes and their targets block the influence of ventralizing signals such as BMPs and promote dorsal cell fate specification(Little and Mullins,2006).

We examined the expression patterns of dorsally and ventrally restricted genes at the beginning of gastrulation. We found that the normally dorsally localized expression of chordin was reduced or absent in brom bones mutant embryos, whereas the ventrally restricted expression of gata2 (gata2a - Zebrafish Information Network) was reciprocally expanded dorsolaterally and frequently circumferentially(Fig. 5A-D′; data not shown). Further characterization showed that the dorsal defects were already apparent at sphere stage (midblastula stage) in brom bones mutant embryos, as indicated by the strongly reduced or absent expression of squint and boz (Fig. 5E,F, arrows; data not shown).

We then examined the nuclear localization of β-catenin in the dorsal blastomeres. We found that in brom bones mutant embryos,β-catenin nuclear accumulation was variably reduced and correlated with the range of ventralized phenotypes observed in mutant embryos(Fig. 5G,H; data not shown). These results indicate that dorsal axis formation in brom bonesmutants is impaired upstream of β-catenin nuclear localization.β-catenin localization at the cell membrane of mutant embryos was as intense as in wild-type embryos, suggesting that the absence of β-catenin in dorsal blastomere nuclei of brom bones mutants is not due to reduced maternal β-catenin protein abundance (compare G and H in Fig. 5). We tested if normal levels of β-catenin protein were present in brom bones mutants and found by western blot that they were similar in 1-cell-stage mutant and wild-type embryos (Fig. 5I). Thus, we conclude that brom bones mutants are ventralized due to a failure in β-catenin nuclear localization in dorsal blastomeres at midblastula stages.

In zebrafish and Xenopus, β-catenin nuclear translocation in dorsal blastomeres requires the asymmetric transport of a dorsal determinant at early cleavage stages via microtubules(Suprenant and Marsh, 1987; Jesuthasan and Strahle, 1997; Rowning et al. 1997; Miller et al., 1999). In zebrafish, parallel arrays of microtubules in the vegetal yolk cytoplasmic layer at the 1-cell stage and in the yolk cytoplasmic layer at early cleavage stages are required for this movement (Fig. 6A,E) (Jesuthasan and Strahle,1997). Severe cold treatment that disrupts the cortical microtubule network prior to the 32-cell stage results in a failure ofβ-catenin nuclear translocation into dorsal blastomeres(Jesuthasan and Strahle,1997). We found that the expressivity of the ventralized phenotype in brom bones mutants was cold sensitive. After a moderate cold treatment that does not affect wild-type embryos (0-2 hours post fertilization at 20°C), brom bones mutant embryos displayed a significant exacerbation in ventralization (Table 2).

Since low temperature is known to destabilize microtubules(Tilney and Porter, 1967), we asked if the microtubule networks required for dorsal determinant translocation are intact in brom bones mutant embryos. To address this question, whole-mount β-tubulin immunostaining was performed. Parallel arrays of microtubules were observed at the vegetal cortex of wild-type embryos at 20 minutes post fertilization (mpf)(Fig. 6B). By contrast, mutant embryos at the same stage showed variably disorganized microtubules at the vegetal pole. In the most severely affected mutant embryos, no parallel arrays of microtubules at the vegetal cortex were visible. Instead, these mutant embryos formed a short and irregular microtubule meshwork around numerous circular profiles (Fig. 6C). In intermediately affected mutant embryos, microtubule arrays appeared short at the vegetal cortex, often interrupted by the presence of circular profiles similar to those found in severely affected embryos but less abundant(Fig. 6D, arrows). Mildly affected mutant embryos showed virtually normal microtubule organization (data not shown).

We further examined microtubule organization in the yolk cytoplasmic layer(YCL) at the 16-cell stage. Long microtubules formed cortical arrays in wild-type embryos that extended beyond the equator(Fig. 6F). By contrast, in mutant embryos, microtubules in the YCL were sparse and failed to form arrays,due to interruption by circular profiles similar to those observed at earlier stages (Fig. 6G,H, arrowheads). The extent of microtubule disorganization in brom bones mutants correlated with the variation in dorsoventral patterning defects. In general, brom bones homozygous mutant females that produced embryos with the strongest ventralized phenotype also displayed the most severe microtubule organization defects.

We hypothesized that the circular profiles disrupting the microtubule arrays are persistent CGs that failed to exocytose owing to the earlier egg activation defect. Indeed, persistent CGs were evident histologically in early cleavage-stage mutant embryos (Fig. 6I,J). Thus we examined the relationship between the microtubule arrays and the persisting CGs in double-labeling experiments. We found that the circular profiles that block the cortical microtubule arrays were indeed the sites of persistent CGs (Fig. 6C-D″,G-H″). These results suggest that completion of CGE is required for the proper organization of cortical microtubules acting in dorsal determinant translocation at early cleavage stages.

To better understand how brom bones functions, we positionally cloned the brom bones gene. We scanned the genome for linkage of the brom bones mutation to ∼170 SSLP markers and found that the mutation was located between SSLP markers z9361 and z11023, close to the centromere on chromosome 2. Further analysis of 984 meioses placed the brom bones gene within a genetic interval of 0.3 centimorgans (cM),0.2 cM centromeric to z8951 (two recombinants in 984 meioses) and 0.1 cM telomeric to (CA)n2 (one recombinant in 984 meioses)(Fig. 7A). This genetic interval is fully covered by three sequenced and overlapping BAC clones, which correspond to a physical interval of 248 kb between markers z8951 and (CA)n2. Using the Genscan program and the Sanger Ensembl database of zebrafish genomic sequence(www.ensembl.org),we found four transcription units within this physical interval: glutamate receptor ionotropic NMDA 3B (grin3b), plasticity-related gene 2 (prg2; fj45h08 - Zebrafish Information Network), paralemmin and heterogeneous nuclear ribonucleoprotein I(hnrnp I). The open reading frames of these four genes were sequenced and only a single point mutation was identified in the hnrnp I gene. hnrnp I is also known as ptb4 (polypyrimidine tract-binding protein 4) (Ghetti et al., 1992; Wagner and Garcia-Blanco, 2001) and ptbp1a. This point mutation causes a premature stop codon in the hnrnp I gene(Fig. 7B). We sequenced genomic DNA from the original G0 mutagenized male fish, from which the brom bones mutant chromosome was derived following ENU mutagenesis(Wagner et al., 2004). We found that this parental fish did not contain the brom bonesmutation, which excludes the possibility that this base change existed as a background polymorphism. Instead, this base change represents an ENU-induced mutation. We conclude that the brom bones mutant is defective in the hnrnp I gene.

Structurally, hnRNP I is composed of four RNA-recognition motifs (RRMs)with interdomain linker regions. Its amino-terminal extension contains both nuclear localization and export signals(Spellman et al., 2005). Zebrafish hnRNP I has the same domains as hnRNP I in other species, except that its amino-terminal extension contains 18 additional amino acids(Fig. 7C,D). The predicted protein sequence of hnRNP I shows high similarity with hnRNP I in human(75.3%), rat (73.9%), Xenopus (72.2%) and Drosophila (46.8%)(Fig. 7D). The brom bones point mutation is a single nucleotide C to T transversion,converting Gln 254 to a stop codon (Fig. 7B). The mutant transcript would yield a truncated protein only containing the amino-terminal extension, the first RRM and part of the second RRM (Fig. 7C), and is therefore likely to be a strong loss-of-function mutation.

We examined the distribution of the hnrnp I transcript during oogenesis. hnrnp I is expressed throughout oogenesis. It was found throughout stage I and II oocytes, whereas in stage III oocytes it was enriched in cortical regions (Fig. 7E, arrowheads). By contrast, hnrnp I transcripts were not detectable in stage IV and V oocytes (data not shown).

Studies in Xenopus oocytes show that hnRNP I is directly involved in localizing the Vg1 transcript to the vegetal pole(Cote et al., 1999; Kress et al., 2004). To determine if hnRNP I plays a similar role in mRNA localization during zebrafish oogenesis, we examined Vg1 (dvr1 - Zebrafish Information Network) localization in brom bones mutant oocytes by in situ hybridization on ovary sections. We selected ovaries from females that consistently showed strong egg activation defects for this analysis. We found that Vg1 transcripts were localized normally in brom bonesmutant oocytes (data not shown). However, in zebrafish, unlike in Xenopus oocytes, Vg1 RNA localizes to the animal pole rather than the vegetal pole of oocytes (Helde and Grunwald, 1993; Bally-Cuif et al., 1998). Therefore, it is possible that although hnRNP I was not required to localize animal pole transcripts, it could play a conserved role in the localization of transcripts to the vegetal pole. To test this possibility we examined three transcripts that localize to the vegetal pole, mago nashi, dazl and bruno-like (brl; cugbp1 - Zebrafish Information Network)(Maegawa et al., 1999; Suzuki et al., 2000; Howley and Ho, 2000; Pozzoli et al., 2004; Marlow and Mullins, 2008). We found that these transcripts also localized normally in brom bonesmutant oocytes (Fig. 8; data not shown).

The difference in hnRNP I function in transcript localization along the animal-vegetal pole in Xenopus versus zebrafish might be due to subfunctionalization or redundancy between multiple hnRNP I genes in the zebrafish. We recently found a second hnRNP I gene in the zebrafish genome. The second hnRNP I gene is located on chromosome 11, as a genome duplication of a small region of chromosome 2 where hnrnp I is located. The second hnRNP I is highly similar to hnrnp I (84% protein sequence identity). Since this hnRNP I gene is also expressed during zebrafish oogenesis (W.M., unpublished), it could play a role during oogenesis. Whether this second hnRNP I protein directs mRNA localization along the oocyte animal-vegetal axis will require further study.

Here we report the characterization of a zebrafish maternal-effect mutant, brom bones, which disrupts the hnrnp I gene. We show that loss of hnRNP I function impairs the Ca²⁺ wave that normally traverses the egg upon entry into spawning medium, causing a defect in egg activation, which includes impaired CGE, chorion elevation and cytoplasmic segregation. This function of hnRNP I is distinct from its previously described functions in other animals, which include roles in spermatogenesis,wing development, and RNA localization and translational regulation during oogenesis (Dansereau et al.,2002; Cote et al.,1999; Robida and Singh,2003; Kress et al.,2004; Czaplinski and Mattaj,2006; Lewis et al.,2008; Besse et al.,2009). Our studies reveal a new crucial role for hnRNP I in the initiation of egg activation in zebrafish.

A crucial factor in initiating egg activation at fertilization is a rise in free Ca²⁺ in the egg cytosol, although the mechanism by which Ca²⁺ evokes egg activation remains unknown. A Ca²⁺ rise in the egg cytosol upon sperm-egg fusion has been observed in virtually all animal and plant species studied, and is mediated by activation of the phosphoinositide pathway (Stricker,1999; Runft et al.,2002; Tsaadon et al.,2006). The activated phosphoinositide pathway produces IP₃, which interacts with its receptor (IP₃R) on the ER to mediate the release of Ca²⁺. In zebrafish, when spawning medium contacts the egg, a single Ca²⁺ wave initiates at the animal pole and traverses the egg to the vegetal pole(Lee et al., 1999). brom bones mutant eggs display a defect in the Ca²⁺ wave, which correlates with the strength of the egg activation defect. This suggests that failure of egg activation in brom bones mutants is caused by an insufficient rise in cytosolic Ca²⁺. In support of this view,providing Ca²⁺ or IP₃ to brom bones mutant eggs rescues the egg activation defects. These results strongly suggest that hnRNP I plays an important role in the control of egg activation through the regulation of IP₃-mediated cytosolic Ca²⁺ release.

The IP₃ rescue data also indicate that the machinery downstream of IP₃ required for egg activation, for example the machinery required for intracellular Ca²⁺ storage and IP₃activation of its receptor IP₃R, is intact in brom bonesmutants. In support of this view, electron microscopy studies on unactivated brom bones mutant eggs suggest that the distribution and density of the ER, the major intracellular Ca²⁺ storage site in the egg cortex, are normal in brom bones mutant eggs (data not shown). How hnRNP I regulates IP₃-mediated cytosolic Ca²⁺ elevation then becomes an intriguing question. Since IP₃ is generated by hydrolysis of phosphatidylinositol 4,5-bisphosphate (PIP₂), hnRNP I might be required for the production of upstream activators of IP₃at fertilization. These might include enzymes of the phospholipase C (PLC)family, as well as cytoplasmic tyrosine kinases of the Src family, which act upstream of PLC (Runft et al.,2002; Tsaadon et al.,2006). An increase of Src family activity and PLC activity upon fertilization has been observed in many organisms, including zebrafish(Wu and Kinsey, 2002; Runft et al., 2002; Kinsey et al., 2003; Sharma and Kinsey, 2006; Tsaadon et al., 2006).

hnRNP I probably regulates its target mRNA during oogenesis. Since egg activation is completed within minutes following egg laying in zebrafish, the factors involved in eliciting the Ca²⁺ wave are expected to be present as proteins in the egg, and therefore do not require mRNA translation or post-transcriptional regulation at this time. Thus, hnRNP I is unlikely to regulate its target(s) in the egg. We expect that it regulates mRNA target(s)during oogenesis, the protein product(s) of which function either directly or indirectly in activating the egg to initiate zygotic development. A possible time window for hnRNP I function is during oocyte maturation, when oocytes in most species acquire their competence to undergo egg activation-associated events, such as CGE.

brom bones mutants also display defects in dorsoventral embryonic axis formation and organization of the cortical microtubules, which are implicated in establishing the dorsoventral axis. Interestingly, the severity of microtubule organization defects correlates with the severity of the ventralization phenotype, suggesting that the former defect might be the cause of the latter. Failure to assemble contiguous microtubule arrays in the yolk cytoplasmic layer in brom bones mutant embryos is associated with incomplete CGE owing to the earlier egg activation defect. Incomplete CGE results in a large number of CGs remaining at the cortex of cleavage stage brom bones mutant embryos. These persisting CGs block the proper assembly of the cortical microtubules in eggs/embryos, which we postulate disrupt dorsal determinant translocation from the vegetal pole to the future dorsal side, and thus ventralize these embryos. However, it is also possible that hnrnp I regulates dorsal axis formation through an independent process. For example, it could control the IP₃-mediated cytosolic Ca²⁺ elevation during the blastula stage that is implicated in dorsal axis formation (Lyman Gingerich et al., 2005; Slusarski et al.,1997a; Slusarski et al.,1997b; Westfall et al.,2003) (reviewed by Slusarski and Pelegri, 2007). Whether hnrnp I regulates IP₃-mediated cytosolic Ca²⁺ elevation in different processes during zebrafish development will require further study.