SH2-domain-containing inositol phosphatase 2 (SHIP2) belongs to a small family of phosphoinositide 5-phosphatases that help terminate intracellular signaling initiated by activated receptor tyrosine kinases. Mammalian SHIP2 is viewed primarily as an attenuator of insulin signaling and has become a prominent candidate target for therapeutic agents that are designed to augment insulin signaling. Despite this view, no signaling pathway has yet been demonstrated as being affected directly by SHIP2 function in vivo, and in vitro studies indicate that the protein may function in multiple signaling pathways. Here, we analyze the role of a SHIP2 family member in the early zebrafish embryo where developmental and gene expression defects can be used to assay specific signaling pathways. The zebrafish ship2a transcript is maternally supplied, and inhibiting the expression of its protein product results in the expansion of dorsal tissue fates at the expense of ventral ones. We show that the developmental defects are the result of perturbation of fibroblast growth factor (FGF) signaling in the early embryo. Loss of Ship2a leads to an increased and expanded expression of outputs of FGF-mediated signaling, including FGF-dependent gene expression and activated mitogen-activated protein kinase (MAPK) signaling. Our findings demonstrate that Ship2a attenuates the FGF signaling pathway in vivo and functions in the establishment of normal tissue patterning in the early embryo. We suggest that modulation of FGF signaling may be a principal function of SHIP2 in mammals.
Negative feedback loops and constitutive dampening of the function of signal transduction components play important roles in both intracellular and intercellular signaling pathways. Attenuation of signaling has multiple functions, including resetting responsiveness, limiting the duration of signaling, or modulating the strength of a signal. Negative feedback loops are intrinsic to the mechanisms that maintain homeostasis, and the disruption of feedback mechanisms is a prominent causative component of the pathophysiology of metabolic disorders such as diabetes and growth control disorders such as cancer (Leslie and Downes, 2004; Engelman et al., 2006; Lazar and Saltiel, 2006; Muoio and Newgard, 2008). Negative feedback of signaling events also plays important roles in the patterning of tissue precursors in the embryo, contributing to the shaping of morphogen gradients and the establishment of positional information (Manahan et al., 2004; Affolter and Basler, 2007). The finding that signaling events are attenuated at multiple levels, including the sequestration or destruction of extracellular ligands and intracellular activators; dephosphorylation of members of kinase cascades; and competition for target promoter sites, indicates that modulation of signaling strength is a significant aspect of the ability of a cell to sense and respond to its environment (De Robertis et al., 2000; Logan and Nusse, 2004; Manahan et al., 2004; Tsang and Dawid, 2004; Affolter and Basler, 2007).
Phosphoinositides are both immediate effector components of receptor tyrosine kinase (RTK)-initiated signaling and major targets for the negative feedback modulation of RTK signaling. Activation of RTKs results in the recruitment and activation of phosphoinositide 3-kinase (PI3K) resulting in the local production of phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P3], which serves as a focus for the assembly and activation of signal transduction machinery (Lazar and Saltiel, 2006; Ooms et al., 2009). Multiple inositol phosphatases are recruited to remove PtdIns(3,4,5)P3; their activities generate derivative PtdIns, which may serve to terminate signaling or promote distinct signaling cascades. For example, the inositide 3-phosphatase PTEN (phosphatase and tensin homologue on chromosome 10) is a primary attenuator of signaling in many contexts and consequently functions as a tumor suppressor (Leslie and Downes, 2004).
The src homology 2 (SH2)-domain-containing inositol 5-phosphatases (SHIPs) comprise a small family of proteins whose primary role appears to be downregulation of RTK signaling cascades (Backers et al., 2003). SHIPs are poised to interact with signaling cascades. They are expressed constitutively, facilitating their rapid activity-dependent recruitment to RTKs at the plasma membrane (Backers et al., 2003; Wang et al., 2004). SHIPs are multi-domain proteins that are capable of attenuating signaling in at least two ways: by virtue of their SH2 domain, which interacts directly with RTKs or with the adaptor molecules Shc and Grb2 and, thereby, presumably disrupts signaling interactions; or owing to their phosphatase activity, which catalyzes the dephosphorylation of PtdIns(3,4,5)P3 to phosphatidylinositol (3,4)-bisphosphate [PtdIns(3,4)P2] and antagonizes the PI3K-stimulated pathway (Pesesse et al., 1998; Wisniewski et al., 1999; Pesesse et al., 2001). As PtdIns(3,4)P2 also serves to support signaling through Akt (protein kinase B) and PDK1 (phosphoinositide-dependent kinase 1), SHIP2-dependent dephosphorylation may have modulatory functions in addition to attenuation of a signal (Ooms et al., 2009).
SHIPs attenuate signaling by mechanisms that are likely to affect multiple RTK pathways. Indeed, in vitro studies indicate that SHIPs can negatively regulate signaling events mediated by insulin, nerve growth factor (NGF), epidermal growth factor (EGF), c-Met, macrophage colony-stimulating factor (M-CSF) or platelet-derived growth factor receptor (PDGFR) (Pesesse et al., 2001; Wang et al., 2004; Koch et al., 2005; Aoki et al., 2007; Artemenko et al., 2009). Nevertheless, the true in vivo targets of SHIPs are unknown. In mammals, SHIP1 (also known as INPP5D) is expressed in a tissue-restricted manner, being present at high levels in cells of the hematopoietic lineage. Loss of Ship1 function in mice produces a mild lymphoproliferative phenotype in which both macrophages and B lymphocytes are hypersensitive to stimuli (Helgason et al., 1998; Liu et al., 1999). In contrast, SHIP2 (also known as INPPL1) is expressed broadly, and its immediate functions in vivo have yet to be defined (Muraille et al., 1999; Decker and Saltiel, 2005).
Whereas in vitro studies have implicated SHIP2 in multiple biological processes, many analyses have highlighted its role in insulin signaling (Clement et al., 2001; Sasaoka et al., 2001; Wada et al., 2001; Lazar and Saltiel, 2006). Overexpression of SHIP2 in humans or mice has been linked with susceptibility to type 2 diabetes and insulin resistance (Marion et al., 2002; Fukui et al., 2005; Kagawa et al., 2008). Consistent with these observations, the loss of function characterization of the Ship2/Inppl1 gene in mice has demonstrated a fundamental role of SHIP2 in the regulation of energy metabolism (Clement et al., 2001; Sleeman et al., 2005). Whereas one mutant allele affecting the Ship2 locus was associated with increased sensitivity to insulin (Clement et al., 2001), mice that were homozygous for a simple null allele of Ship2 were viable with mild skeletal defects; normal glucose levels and insulin tolerance; and resistance to high-fat diet-induced obesity (Sleeman et al., 2005). Although the cell types that mediate the physiological functions of SHIP2 and the molecular mechanism of SHIP2 action remain unresolved, these findings have been interpreted as indicating that both neural and non-neural functions of SHIP2 underlie the metabolism phenotype (Birnbaum, 2005; Decker and Saltiel, 2005; Lazar and Saltiel, 2006). As SHIP2 is viewed as a potential target for the development of new therapeutic agents to ameliorate the effects of diabetes (Lazar and Saltiel, 2006; Sasaoka et al., 2006), it is important to identify specific cellular signaling events that SHIP2 modulates in vivo.
As many tissue induction and patterning events in the vertebrate embryo have been linked to specific signaling pathways, we reasoned that an analysis of the earliest developmental defects associated with loss of SHIP2 function in vivo might indicate a signal transduction pathway that directly requires SHIP2 activity. We analyzed the consequences of loss of function of the zebrafish ship2a gene on the development of the zebrafish embryo and traced the origins of these defects to perturbation of a single signaling pathway. We show here that maternal Ship2a protein is needed to establish proper dorsoventral (DV) patterning in the zebrafish embryo. As interactions between the WNT, bone morphogenic protein (BMP) and fibroblast growth factor (FGF) signaling pathways establish and maintain the position-dependent identities of tissue precursors in the gastrula stage vertebrate embryo (De Robertis et al., 2000; Hammerschmidt and Mullins, 2002; Furthauer et al., 2004), we examined gene expression outputs dependent on each of these pathways in embryos deficient in Ship2a. We demonstrate that maternal Ship2a protein is required to establish proper DV patterning by virtue of its function as a negative regulator of FGF signaling in the pre-gastrula zebrafish embryo.
Identification and expression of zebrafish Ship2a
The human SHIP2 protein sequence was used in a BLAST homology search to identify expressed and genomic sequences that encoded the presumed zebrafish ortholog. A full-length zebrafish cDNA clone with a predicted open-reading frame encoding 1266 amino acids was generated by reverse transcription (RT)-PCR of mRNA isolated from 24-hour post-fertilization (hpf) embryos. The zebrafish protein, designated here as Ship2a, has the signature domains and the motif organization that unambiguously identify it as a SHIP2 family member, including an N-terminal SH2 domain, a central inositol polyphosphate 5-phosphatase catalytic (IPPc) domain, an NPXY motif, a proline-rich domain (PRD) and a C-terminal sterile alpha-motif (SAM), a specific motif that distinguishes SHIP2 from SHIP1 proteins (Fig. 1A) (see also the complete sequence and domain identification in supplementary material Fig. S1) (Backers et al., 2003). Ship2a is, overall, 65% identical to human SHIP2 and shares even greater identity with the functional domains of the human protein (Fig. 1A). The zebrafish genome also contains a second ship2 gene, ship2b (chromosome 14, zv7 scaffold contig BX088596.8), and a more distantly related gene, ship1 (chromosome 6, zv7 scaffold contig CR847938.10). Although the Ship2b protein sequence exhibits the typical SHIP2 domain structure, it is only approximately 44% identical to either human SHIP2 or zebrafish Ship2a. Comparative sequence analyses indicate that zebrafish Ship2a is the zebrafish protein that is most closely related to human and mouse SHIP2 (Fig. 1B; supplementary material Fig. S1). Moreover, zebrafish ship2a maps to the distal portion of chromosome 15 (zv7 scaffold 1537.2-1538.5), immediately adjacent to phox2a, which is a syntenic relation that is present in the mouse and human genomes, further supporting its identity as the true ortholog of mammalian SHIP2.
Zebrafish ship2a transcripts are broadly distributed throughout the embryo from the one-cell stage to 24 hpf (Fig. 1C–E and data not shown). Because ship2a is highly expressed in the one-cell zygote and in all blastomeres through the cleavage stages, we conclude that ship2a is a maternally supplied transcript that is ubiquitously expressed in the early embryo.
Maternal Ship2a protein is required for DV patterning
To identify an in vivo role of ship2a, we depleted the maternal Ship2a product from the embryo with antisense morpholinos (MOs) and examined how loss of Ship2 affected embryonic development. We verified that the sequence of the maternal transcript was identical to the one cloned from 24-hpf embryos and generated two translation-blocking MOs (ship2a MO1 and ship2a MO2), each complementary to a different non-overlapping sequence near the AUG start codon (see Methods). MOs were injected into the yolk of one-cell stage embryos shortly after fertilization, under conditions that enabled each MO to completely block the translation of a co-injected reporter mRNA, consisting of 5′ ship2a sequences fused to green fluorescent protein (GFP)-coding sequences (supplementary material Fig. S2). Furthermore, the effects of the MOs were highly target specific in that injection of either translation-blocking MO produced an identical phenotype that was uniformly expressed by treated gastrula-stage embryos, and distinct from that produced by effective splice blocking or other control MOs (Fig. 2J,L; supplementary material Fig. S3; and additional data not shown).
Embryos depleted of maternal Ship2a exhibited a characteristic dorsalization phenotype at the 5–6-somite stage, affecting multiple tissues and germ layers (Nguyen et al., 1998). Lateral expansion of somites (paraxial mesoderm) was readily visible (n=22/23 embryos), and lateral expansion of the hindbrain (neuroectoderm) was revealed by krox-20 expression in rhombomeres 3 and 5 (n=19/21 embryos) (Fig. 2A–D). In addition, as is often observed in dorsalized embryos, neural crest progenitors, which are present at the border between the neural and non-neural ectoderm and are indicated by foxd3 expression, were absent or generated in reduced numbers (n=24/27 embryos) (Fig. 2E,F). Most ship2 morphants died at segmentation stages; those that did survive to 24 hpf, which presumably represented weak versions of the prevailing phenotype, exhibited morphology consistent with a dorsalized patterning defect (Fig. 2G,H) (Stachel et al., 1993).
As ventrolateral expansion of dorsal tissues can be caused by perturbation of patterning in the gastrula-stage embryo (Hammerschmidt and Mullins, 2002), we analyzed patterning in terms of gene expression domains in Ship2a-depleted gastrulae. Analysis of foxb1.2 and eve1 expression at the sphere stage (4 hpf) revealed that gene expression domains marking prospective DV identities were already shifted in the early gastrula of ship2a-morphant embryos. foxb1.2 encodes a forkhead domain transcription factor that is normally expressed in the dorsolateral region; eve1 is a homeobox gene expressed in the ventroposterior region of the pre-gastrula embryo (Joly et al., 1993; Odenthal and Nusslein-Volhard, 1998). In Ship2a-depleted embryos, the foxb1.2 domain expanded ventrally so that it was expressed circumferentially (n=62/68 embryos) (Fig. 2I–L), and eve1 expression was severely reduced (n=18/19 embryos) (Fig. 2M,N and supplementary material Fig. S3E,F). Although identities were shifted along the DV axis, no alteration in the organizer was detected as assayed by goosecoid (gsc) expression (n=21/21 embryos) (Fig. 2O,P).
DV patterning of the gastrula specifically required Ship2a product derived from maternally supplied mRNA. Injection of 12–15 ng of a ship2a splice-blocking MO completely inhibited production of mature zygotic ship2a mRNA as assayed by RT-PCR, but failed to produce a dorsalized phenotype (data not shown).
Loss of maternal Ship2a affects the BMP but not WNT signaling pathway
Whereas maternal Ship2a did not seem to contribute to specification of organizer tissue, the altered expression of the foxb1.2 and eve1 transcription factor genes indicated that Ship2a was required before the onset of gastrulation to establish proper DV patterning of gene expression domains in the embryo. The BMP and WNT signaling pathways both contribute directly to assigning developmental fates along the DV axis of the early vertebrate embryo (Christian and Moon, 1993; Imai et al., 2001; Hammerschmidt and Mullins, 2002), and hence expression of components of these pathways was examined in embryos that were depleted of maternal Ship2a protein. Loss of maternal Ship2a resulted in substantially decreased expression of bmp7 in pre-gastrula and early gastrula-stage embryos. Normally, bmp7 is widely and abundantly expressed soon after the onset of zygotic transcription (30–40% epiboly) (Fig. 3A), and becomes restricted ventrally by the onset of gastrulation (50% epiboly) (Fig. 3C). However, depletion of maternal Ship2a caused a severe reduction in bmp7 expression at these stages (n=35/38 embryos) (Fig. 3B,D). Similarly, bmp4 expression was reduced in morphants (supplementary material Fig. S3A,B). By contrast, analysis of WNT signaling targets indicated that the WNT pathway was unaltered by loss of Ship2a. vox and vent are ventrally expressed genes that are directly regulated by WNT8 and encode transcriptional repressors that function to repress ventral expression of organizer genes (Imai et al., 2001; Ramel and Lekven, 2004). Embryos depleted of maternal Ship2a exhibited normal expression of vox (n=17/19 embryos) or vent (n=20/22 embryos) at 40% epiboly (Fig. 3E–H). Similarly, expression of bozozok (boz), one of the earliest zygotically expressed genes and a target of the dorsal WNT signaling pathway (Fekany et al., 1999; Koos and Ho, 1999), was not altered at the oblong-sphere stage (3.7–4 hpf) in ship2a-morphant embryos (n=27/28 embryos) (Fig. 3I,J). Together, these results indicate that, prior to the onset of gastrulation, BMP expression but not WNT signaling is dependent on maternal Ship2a.
FGF-dependent gene expression is elevated upon loss of maternal Ship2a
Two phenotypes observed in Ship2a-depleted embryos, dorsalization and diminished bmp gene expression, can be caused by the expansion of FGF signaling in the pre-gastrula embryo (Furthauer et al., 2001; Furthauer et al., 2004; Tsang et al., 2004). Dorsal-restricted expression of FGF in the blastula limits the expression of BMPs to the ventral side of the embryo (Furthauer et al., 2004). As FGF can function in a dose-dependent manner, precise regulation of spatial, temporal and quantitative aspects of FGF signaling is required for the correct DV patterning of gene expression and prospective tissue fates. DV patterning is so sensitive to FGF signaling that its perturbation has been used successfully to detect modifiers of the early embryonic FGF signaling cascade (for a review, see Tsang and Dawid, 2004; Thisse and Thisse, 2005).
The mitogen-activated protein kinase (MAPK) signaling pathway is required to mediate the effects of FGF in the early zebrafish embryo (Furthauer et al., 2001; Furthauer et al., 2004; Tsang et al., 2004). Signaling through the FGF receptor leads to increased expression of genes that encode both pathway mediators, including the Ets transcription factors Pea3 and Erm, and pathway inhibitors, including Mkp3, Sef and Sproutys (Raible and Brand, 2001; Roehl and Nusslein-Volhard, 2001; Tsang and Dawid, 2004; Tsang et al., 2004; Thisse and Thisse, 2005). As a measure of FGF signaling in maternal Ship2a-depleted embryos, we first analyzed the expression of mkp3, whose product MAP kinase phosphatase 3 (Dusp6) acts as a feedback attenuator of FGF signaling by dephosphorylating the active form of MAPK (Tsang et al., 2004). Shortly after the onset of zygotic gene expression, at the oblong-sphere stage (3.7–4 hpf), mkp3 expression is localized under the control of FGF signaling to the future dorsal side of wild-type (WT) embryos (Fig. 4A,C). Depletion of maternal Ship2a resulted in a dramatic expansion of mkp3 expression so that transcripts were distributed throughout the entire blastoderm in an apparently graded fashion at the oblong-sphere stage (n=14/17 embryos) (Fig. 4B,D). Over the following hour of development, between the dome stage and 50% epiboly (4.3–5.3 hpf), the normal mkp3 expression domain broadens circumferentially, and transcripts become distributed around the entire outer margin of the blastoderm, the site of mesoderm induction and a region of high FGF signaling (Fig. 4E; supplementary material Fig. S3C). In Ship2a-depleted embryos, ventral expansion of mkp3 expression occurred precociously and extended away from its normal domain at the edge of the blastoderm margin and deep into the internal portion of the embryo (n=18/19 embryos) (Fig. 4F; supplementary material Fig. S3D). Similarly, depletion of maternal Ship2a resulted in increased expression of erm, a target upregulated by FGF signaling (n=30/34 embryos) (Fig. 4G,H). The erm and mkp3 expression domains were not identical, indicating that the two genes may be responsive to different levels of FGF signaling, or that factors in addition to FGF regulate their transcription. In either case, the results indicated that loss of maternal Ship2a caused augmented and expanded expression of genes whose transcription responds to FGF signaling. These findings are consistent with the interpretation that FGF signaling was hyperactivated in embryos with reduced Ship2a.
Maternal Ship2a is required to attenuate FGF signaling
To measure directly the contribution of Ship2a to FGF signaling, we assessed the state of activation of the FGFR-Ras-MAPK pathway with an antibody specific to the active, diphosphorylated form of MAPK (pMAPK) (Shinya et al., 2001). In control embryos, cells with elevated levels of pMAPK are concentrated at the future dorsal region of the embryo at the oblong-sphere stage (3.7–4 hpf) (n=42/42 embryos) (Fig. 5A,B). In embryos that are depleted of maternal Ship2a, cells with high levels of pMAPK were distributed throughout the blastoderm (n=26/29 embryos, ship2a MO1) (Fig. 5C,D). Similar to the pattern of the expanded expression domains of mkp3 and erm, it appeared that cells with elevated FGF signaling were not randomly distributed following the loss of Ship2a, but rather that signaling extended aberrantly beyond its normal focus emanating from one portion of the blastoderm margin.
The results were consistent with the hypothesis that maternal Ship2a functions specifically to attenuate FGF signaling in the early embryo. In this model, the consequences of loss of Ship2 could be observed only in the context of effective intercellular FGF signaling. If maternal Ship2 indeed acts downstream of receptor activation, then inhibition of FGF receptor function should inhibit the activation of pMAPK regardless of whether or not maternal Ship2 is present. To test the hypothesis, FGF signaling in the embryo was blocked using SU5402, a small molecule that is a specific inhibitor of FGF receptors (Mohammadi et al., 1997). Incubation of embryos in 60 μM SU5402, beginning at the blastula stage (128-cell stage or ∼2.25 hpf), completely inhibited the expression of elevated pMAPK as measured at the oblong-sphere stage (n=24/24 embryos) (Fig. 5E). When embryos that were depleted of maternal Ship2 were also treated with SU5402, cells with elevated pMAPK were not observed (n=18/18 embryos) (Fig. 5F). Thus, loss of maternal Ship2a does not directly stimulate MAPK phosphorylation. Rather, these data indicate that Ship2a modulates the amount of pMAPK that is generated and/or maintained following activation of the FGF pathway.
Dorsal-restricted FGF signaling contributes to DV patterning of the embryo in part by repressing BMP expression and signaling on the future dorsal side of the embryo (Furthauer et al., 2004). Thus, the enhanced activity of FGF-dependent signaling that is observed in ship2a-depleted embryos may be responsible for the downregulation of bmp gene expression observed in ship2a MO-treated embryos, leading to their dorsalization (Figs 2 and 3). If Ship2a interacts specifically with the FGF pathway in the early embryo, and only indirectly affects BMP-dependent ventral identity development, then ventral identities might be restored to ship2a MO-treated embryos following ectopic induction of BMP signaling. We tested whether the tissue patterning defects caused by loss of maternal Ship2a function could be rescued by overexpression of bmp2b. Ventral identity was measured by eve1 expression, which is normally limited to the future ventral side in WT embryos at 50% epiboly (Fig. 6A). Injection of bmp2b mRNA at the one-cell stage caused a dorsalward expansion of the eve1 expression domain, indicated by circumferential expression of eve1 (Fig. 6B). Although eve1 expression was lost when maternal Ship2a function was blocked by MO1 or MO2 (Fig. 6C,E), co-injection of bmp2b mRNA with the MOs strongly enhanced eve1 expression, often restoring its WT expression pattern (Fig. 6D,F). These results indicate that the role of Ship2a in attenuating FGF signaling in the early embryo is sufficient to explain how loss of maternal Ship2a causes subsequent dorsalization of development.
An in vivo function of the SHIP2 protein
Defining the in vivo roles of, and requirements for, SHIP2 activity is important, as SHIP2 has been viewed as a regulator of glucose homeostasis and a favored potential molecular target in the development of therapies for type 2 diabetes (Decker and Saltiel, 2005; Lazar and Saltiel, 2006; Sasaoka et al., 2006). SHIP2 has been shown to be capable of modulating many pathways in vitro, and recent screens to identify components of the FGFR1 and FGFR2 signaling cascades found that the SHIP2 protein population responded dramatically to ligand presentation, becoming highly tyrosine-phosphorylated immediately upon FGF stimulation of mammalian cultured cells (Hinsby et al., 2004; Luo et al., 2009). Here we demonstrate that SHIP2 attenuates the response to FGF stimulation in vivo with the significant effect that it limits the spread of FGF signaling in the zebrafish embryo. We note that SHIP2 may have a similar role in mouse development, as Ship2 null mutant mice exhibit runting and skeletofacial anomalies (Sleeman et al., 2005); these defects may have arisen from aberrant regulation of FGF signaling (Muenke et al., 1994; Kan et al., 2002; Cohen, 2004). Conversely, the finding that Ship2a functions in FGF signaling in the pre-gastrula embryo does not preclude the possibility that it and/or Ship2b may function in additional signaling pathways in zebrafish.
FGF signaling is utilized in many different biological contexts, including regulation of cell proliferation, cell differentiation, cell migration, developmental fate patterning and metabolic homeostasis (Ornitz, 2000; Thisse and Thisse, 2005; Beenken and Mohammadi, 2009). Abnormal levels of FGF signaling play fundamental roles in the origins of birth defects and cancers, and can modulate metabolic rates and obesity (Turner and Grose, 2010; Beenken and Mohammadi, 2009). The work here shows that Ship2a functions to diminish FGF signaling. During early embryonic development a major role of FGF signaling is in assigning tissue fates, and negative regulation of the signal is key to this process (Tsang and Dawid, 2004; Thisse and Thisse, 2005). Following Ship2a depletion, the highest levels of FGF signaling-dependent gene expression and activated pMAPK were observed at their normal locations, indicating that sources of FGF in the embryo were probably not altered. Rather, in the absence of maternal Ship2a, the range of FGF signaling was extended in the pre-gastrula zebrafish embryo. Furthermore, overexpression of bmp2b mRNA in maternal Ship2a-depleted embryos could overcome the tissue patterning defects resulting from expanded FGF signaling. The ability of forced bmp gene expression to bypass the requirement for Ship2a in DV patterning indicates that the main function of Ship2a in the early embryo is probably restricted to its role in attenuating FGF signaling. In all, several co-expressed factors are dedicated to the regulation of FGF signal transduction intensity in the early embryo, although loss of Ship2a results in a far stronger dorsalization phenotype than is seen following depletion of either Sef or Sprouty4 (Furthauer et al., 2004). As the absence of any one component appears to have a readily detectable effect in the zebrafish embryo, it may be that there is considerable cross-dependence among the modulators.
SHIP2, FGFs and metabolism
Mice that are homozygous for a null mutation in Ship2/Inppl1 displayed resistance to obesity and an elevated basal metabolic rate when placed on a high-fat diet, but did not appear to exhibit altered glucose homeostasis or insulin sensitivity (Sleeman et al., 2005). In addition, the mice displayed runting and distinctive craniofacial skeletal anomalies. As a complex interplay of genes and molecular mechanisms affect regulation of obesity, lipid metabolism and insulin resistance, these phenotypic characteristics do not unambiguously implicate a single molecular pathway (Doria et al., 2008; Lusis et al., 2008; Muoio and Newgard, 2008); however, they are consistent with the effects resulting from augmentation of FGF signaling.
Although the majority of FGF proteins function as paracrine factors during development, a subfamily of FGFs (19, 21 and 23) act as endocrine factors regulating cholesterol, lipid, glucose, bile acid and phosphate metabolism (Beenken and Mohammadi, 2009). A link between FGF function and both regulation of obesity and insulin sensitivity comes from studies in which FGF19 or 21 was systemically administered or overexpressed in transgenic mice (Tomlinson et al., 2002; Kharitonenkov et al., 2005; Coskun et al., 2008; Xu et al., 2009). Transgenic mice expressing either FGF19 or 21 were resistant to dietary induced obesity, and systemic administration of FGF19 or 21 in ob/ob or db/db mice ameliorated obesity and hyperglycemia. As Ship2 null mice have normal insulin and glucose levels and tolerances, it is possible that the primary role of mammalian SHIP2 in vivo may be to affect metabolism through modulation of endocrine FGF signaling. This role need not exclude the possibility that SHIP2 also functions directly in the insulin-signaling pathway. Analysis of zygotic zebrafish Ship2a function may provide insight into its possible role in both FGF and insulin signaling in vivo.
Finally, the work presented here does not indicate the mechanism by which Ship2a downregulates FGF signaling in the embryo. It is somewhat unexpected to find that a protein that is best recognized for its phosphoinositide phosphatase activity clearly regulates the levels of activated MAPK present in cells responding to FGF. Nevertheless, it is consistent with previous studies that have shown that SHIP2 can modulate levels of activated MAPK in Chinese hamster ovary (CHO) cells stimulated with insulin (Blero et al., 2001; Kagawa et al., 2005). It is possible that Ship2a might function through its SH2 domain to compete with, and displace, molecules such as Shc and Grb2 from the activated FGF receptor and thereby reduce the number of active signaling complexes (Sasaoka et al., 2003). Consistent with this idea, SHIP1 or SHIP2 mutants lacking phosphatase activity are capable of inhibiting MAPK activity in 3T3-L1 cells stimulated with insulin or platelet-derived growth factor (PDGF) (Sharma et al., 2005; Artemenko et al., 2009). It is also possible that Ship2a attenuates the FGF pathway by functioning as a negative regulator of the PI3K signaling pathway (Wisniewski et al., 1999). Inhibition of PI3K activity in the early zebrafish embryo did not affect early dorosventral patterning even though it produced gastrulation defects (Montero et al., 2003). Thus, it is unlikely that loss of Ship2a-mediated modulation of PI3K signaling would be sufficient to account for the defects seen in Ship2a-depleted embryos. The ability of various forms of SHIP2 (i.e. IPPc mutants or SH2 mutants) to rescue any of the morphant phenotype will address the relative contributions of maternal Ship2a to different signal transduction pathways.
Zebrafish were maintained under standard conditions (Westerfield, 2000). Wild-type animals were of the AB strain. All experiments were performed in accordance with, and under the supervision of, the University of Utah’s Institutional Animal Care and Use Committee (IACUC).
Cloning and analysis of zebrafish ship2 sequences
Human SHIP2 (NP_001558) was used to BLAST search the zebrafish genome (www.ensembl.org/Danio_rerio/), and the putative translational start site and 3′ UTR of zebrafish ship2a were identified. The primers ship2F, 5′-CCGCGATGGCCGCTGTGGGTC-3′ and ship2R, 5′-GGAATTCCTCAGGTCAGATTCACCAGTGC-3′ were used to amplify the full coding sequence from WT AB strain cDNA, which was then subcloned into pGEM-T (Promega). The zebrafish ship2a and related ship2b sequences were deposited in GenBank (Accession numbers DQ272661 and DQ272662, respectively). Sequence alignment and phylogenetic analysis were performed using The Biology Workbench at the San Diego Supercomputer Center (http://workbench.sdsc.edu). Human SHIP2 (NP_001558), mouse SHIP2 (NP_034697), human SHIP1 (NP_001017915) and zebrafish Ship1 (XP_001923007) sequences were used in the comparison with the two zebrafish Ship2 proteins.
Morpholino oligonucleotide injections
Two non-overlapping antisense translation-blocking MOs were synthesized (Gene Tools): ship2a MO1 (−20 to +5 with respect to the translation initiation site), 5′-GCCATCGCGGAGACTCACACCGGAG-3′ (ship2a translational start site underlined) and ship2a MO2 (−75 to −50) 5′-CGACGCTCACCGCTGATCCTGCTTT-3′. Approximately 1 nl of morpholino was injected into the yolk of one-cell stage WT embryos.
Analysis of MO effectiveness
The 5′ UTR of ship2a (−109 to +6 with respect to the translation start site) was amplified from cDNA generated from 16–32-cell AB embryos and cloned into a CS2+ vector containing the gfp cDNA. 5′ UTR GFP mRNA was synthesized in vitro (Message Maker Kit, Epicentre Biotechnology) and either 400 pg chimeric mRNA or 400 pg chimeric mRNA plus ship2a MO1 or ship2a MO2 were injected into fertilized eggs. Embryos were collected at 4–5 hpf and processed for immunoblot analysis, or fixed in 4% paraformaldehyde and processed for immunohistochemistry (IHC).
Groups of 30 WT, 5′ UTR GFP mRNA-injected, 5′ UTR GFP mRNA plus ship2a MO1-injected, or 5′ UTR GFP mRNA plus ship2 MO2-injected embryos were harvested at 4–5 hpf and partially de-yolked with forceps. Embryos were then placed in 1 ml of cold PBSEE (PBS, 10 mM EDTA, 10 mM EGTA) and cells were dissociated with vigorous pipetting. Cells were centrifuged (1000 rpm, 5 minutes at 4°C), the supernatant was discarded, and the pellet was washed four times with cold PBSEE. Cells were finally centrifuged (3000 rpm, 1 minute at 4°C) and the pellet was resuspended and homogenized in lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.1% Tween-20, 0.05% NP-40) with protease inhibitors (complete protease inhibitor cocktail tablets, Roche Applied Science). The homogenized sample was centrifuged (10,000 rpm, 5 minutes), and the supernatant was stored at −80°C. 25 μg of total lysate (approximately 12 embryos) was resolved by electrophoresis through 4–20% polyacrylamide gradient gels (NuSep Inc.). Blots of the gels were probed with anti-GFP (1:4000, Molecular Probes) and anti-Paf1 (1:6500, Bethyl Laboratories, Inc.), and the signal was developed following incubation with enhanced chemoluminescence (ECL Plus, GE Healthcare).
Embryos for IHC were fixed with fresh 4% paraformaldehyle in 1× PBS buffer overnight at 4°C (Jurynec et al., 2008). Anti-GFP (1:1000; Molecular Probes) or anti-MAPK, activated (diphospho-ERK1 and 2) (pMAPK) antibody (1:10,000; Sigma M 8159) (Shinya et al., 2001) were diluted in DMSO-PBT-SS (PBS supplemented with 1% DSMO, 0.2% BSA, 0.1% Triton X-100 and 10% sheep serum). GFP immunoreactivity was detected by anti-rabbit Alexa 488 (1:200; Molecular Probes). Embryos were mounted in 75% glycerol and photographed under epifluorescence and bright field illumination. pMAPK immunoreactivity was detected by Alexa 488 tyramide amplification (Molecular Probes) following incubation with a goat anti-mouse IgG-HRP antibody (1:200, Molecular Probes). Embryos were mounted in 75% glycerol and images were obtained using a Leica DMRXE laser scanning confocal microscope.
SU5042 was dissolved in DMSO and diluted in embryo-rearing medium (Westerfield, 2000). Embryos were treated continuously, beginning at the 128-cell stage. For the lot of SU5402 used here, 60 μM was the lowest concentration that was capable of inhibiting pMAPK expression completely, as judged by anti-pMAPK IHC at the oblong-sphere stages. Embryos were injected shortly after fertilization with MO and allowed to develop until the 128-cell stage, at which time MO-injected and control embryos were treated with either 60 μM SU5402 or DMSO. Embryos were processed for anti-pMAPK IHC at the oblong-sphere stages (3.7–4 hpf).
We thank Ken Poss for his kind gift of SU5042, Nathan Bahary for helping to isolate a YAC clone containing ship2a, and all of our Utah colleagues for constructive criticism. M.J.J. was supported by a Ruth L. Kirschstein NRSA predoctoral fellowship (F31 NS044665) and NIH postdoctoral fellowships (5T32NS07493 and 5T32HL079874). These studies were supported by NIH grants HD37572 and HD048886 (D.J.G.). Deposited in PMC for release after 12 months.
M.J.J. and D.J.G. conceived the study, analyzed data and wrote the manuscript. M.J.J. designed and performed the experiments with assistance from D.J.G.
The authors declare no competing financial interests.