Insulin-like Growth Factor-binding Protein-3 Plays an Important Role in Regulating Pharyngeal Skeleton and Inner Ear Formation and Differentiation*

Insulin-like growth factor-binding protein (IGFBP)-3 is the major insulin-like growth factor (IGF) carrier protein in the bloodstream. IGFBP-3 prolongs the half-life of circulating IGFs and prevents their potential hypoglycemic effect. IGFBP-3 is also expressed in many pe-ripheral tissues in fetal and adult stages. In vitro , IG-FBP-3 can inhibit or potentiate IGF actions and even possesses IGF-independent activities, suggesting that local IGFBP-3 may also have paracrine/autocrine function(s). The in vivo function of IGFBP-3, however, is unclear. In this study, we elucidate the developmental role of IGFBP-3 using the zebrafish model. IGFBP-3 mRNA expression is first detected in the migrating cranial neural crest cells and subsequently in pharyngeal arches in zebrafish embryos. IGFBP-3 mRNA is also persistently expressed in the developing inner ears. To determine the role of IGFBP-3 in these tissues, we ablated the IGFBP-3 gene product using morpholino-modified antisense oligonucleotides (MOs). The IGFBP-3 knocked down embryos had delayed pharyngeal skeleton mor-phogenesis and greatly reduced pharyngeal cartilage differentiation. Knockdown of also decreased inner ear size and cell differentiation and reintroduction of a of IG-FBP-3 the suggest that in regulating pharyngeal cartilage and inner ear development and growth in zebrafish.

Insulin-like growth factors (IGFs) 1 are evolutionarily conserved polypeptides that are essential for fetal and prenatal growth and survival (1,2). In extracellular fluids, IGFs are complexed with an IGF-binding protein (IGFBP). Six distinct IGFBPs, designated as IGFBP-1 to -6, have been isolated and cloned from human and other vertebrate species, and each represents an individual gene product (3)(4)(5). IGFBP-3 is the most abundant IGFBP in the blood. The majority of the IGFs circulate as a 150-kDa ternary complex that consists of IGF, IGFBP-3 (45 kDa), and acid-labile subunit (88 kDa) (4). When associated with the 150-kDa complex, the IGFs do not readily leave the vascular compartment, and their half-lives are prolonged to 12-15 h in comparison with the 10-min half-life of free IGFs (6). Therefore, the 150-kDa complex serves as a carrier protein of IGFs, protects IGFs from degradation, and prevents the possible hypoglycemic effects of high concentrations of IGFs in the blood. When needed by the organism, IGFs are released from the ternary complex by proteolysis of IG-FBP-3 and move across the vascular wall (7).
In addition to circulating IGFBP-3 originating from the liver, IGFBP-3 is also synthesized and secreted locally in a variety of fetal and adult tissues (8 -15), suggesting that it may have paracrine and/or autocrine functions. In vitro, IGFBP-3 significantly inhibited IGF-1-stimulated DNA synthesis in human skin fibroblasts when added together with IGF-1 but potentiated IGF-1 actions when it was preincubated with those cells (4). It has been proposed that cell surface association of IG-FBP-3 is required for potentiation of the bioactivity of IGFs (16). The cell surface-associated IGFBP-3 has a 10-fold lower affinity for IGF compared with IGFBP-3 in solution (17,18). This switch of binding affinity may result in an equilibrium favoring binding of IGF-I to its receptor and account for the potentiating action of IGFBP-3. Partial proteolysis of IGFBP-3 may also be associated with the potentiation effect because the proteolytic fragments have a much lower affinity for IGF-I (19). There is also evidence that IGFBP-3 possesses biological activities that are ligand-independent. IGFBP-3 inhibited cell growth and accelerated apoptosis in the absence of IGFs and in IGF-1R-null cells (20,21). An IGFBP-3 mutant that does not bind to IGFs possesses that same proapoptotic activity (22). The molecular mechanisms underlying the proapoptotic actions of IGFBP-3 are not well understood, but IGFBP-3 has been reported to bind a membrane protein (23,24) and to stimulate Smad2 and Smad3 phosphorylation (25). This membrane protein/receptor has recently been identified to be low density lipoprotein receptor-related receptor-1 (26). Intriguingly, several studies have demonstrated that IGFBP-3 is localized in the nucleus (27)(28)(29)(30) and can interact with retinoid X receptor ␣ (31).
Despite a large amount of in vitro data on various possible cellular actions of IGFBP-3 in a wide variety of cultured mammalian cells, the in vivo physiological functions of IGFBP-3 remain unclear. IGFBP-3 knock-out mice were normal in size (32), leaving in question the functional role of IGFBP-3. Several properties of the mouse model have complicated the interpre-tation of this result. First, the great redundancy in expression patterns among IGFBPs in mouse may mask some developmental role of this IGFBP. Second, altered expression of other IGFBPs may compensate for the loss of IGFBP-3. In support of this view, serum levels of IGFBP-1, -3, and -4 were increased in IGFBP-2 knock-out mice (33). Finally, there exists the possibility of maternal compensation in gene knock-out progeny, via placental circulation.
The goal of this study is to determine the in vivo functions of IGFBP-3 in early development using the zebrafish model through both loss-and gain-of-function approaches. The zebrafish has emerged as an informative vertebrate model organism for study of the IGF signaling pathway in early development (5,34,35). This genetically tractable vertebrate develops externally, offering an excellent alternative model vertebrate organism in which to study gene and protein function in the absence of maternal compensation. The transparent zebrafish embryos also permit the visualization of tissue and organ formation in real time. Furthermore, major components of the zebrafish IGF signaling pathway, including IGF ligands, IGF receptors, and the intracellular signal transduction network, have been characterized, and they are highly conserved (5, 34 -36). The IGFBP-1, -2, and -3 genes are conserved in zebrafish as well (37)(38)(39)(40). Our results suggest that IGFBP-3 is expressed in the pharyngeal arch and inner ear and plays an important role in regulating the differentiation and growth of these tissues.

EXPERIMENTAL PROCEDURES
Materials-All chemicals and reagents were purchased from Fisher unless otherwise noted. Human IGF-I was purchased from GroPep (Adelaide, Australia). RNA polymerase and RNase-free DNase were purchased from Promega (Madison, WI). Restriction endonucleases were purchased from New England BioLabs (Beverly, MA). Oligonucleotide primers for PCR were purchased from Invitrogen. Morpholino-modified oligonucleotides were purchased from Gene Tools, LLC (Corvalis, OR).
Experimental Animals-Wild-type zebrafish (Danio rerio) were maintained on a 14-h light/10-h dark cycle at 28°C and fed twice daily. Embryos were obtained by natural crossing. Fertilized eggs were raised in embryo medium (41) at 28.5°C and staged according to Kimmel et al. (42). To inhibit embryo pigmentation, embryo medium was supplemented with 0.003% (w/v) 2-phenylthiourea (41). The embryos were fixed in 4% paraformaldehyde in 1ϫ PBS and stored at Ϫ20°C in 100% methanol or frozen in liquid nitrogen and stored at Ϫ80°C for further analysis.
Whole Mount in Situ Hybridization-Whole mount in situ hybridization was performed as reported previously (38). Images were taken with a Nikon DC50NN camera mounted to a Nikon Eclipse E600 microscope (Melville, NY).
Western Blot and Immunocytochemistry-For Western immunoblot and ligand blot analyses, embryos were homogenized in homogenization buffer (100 mM Tris, pH 6.8, 0.1% Tween 20, 10% glycerol, 1 g/ml aprotinin, 0.3 mM phenylmethylsulfonyl fluoride). Protein levels of each sample were quantified using a commercial protein assay kit (Pierce). Equal amounts of protein were subjected to SDS-PAGE (12.5%) and transferred to Immoblin-P membrane (Millipore Corp.). Western immunoblot analysis was performed as described previously (37) using a GFP antibody at a 1:1000 dilution (Torrey Pines Biolabs, Inc., Houston, TX). Western ligand blot was carried out as described by Shimizu et al. (43). The whole mount antibody staining was performed as previously reported (36) using a monoclonal antiacetylated tubulin (Sigma) at 1:1000 dilutions.
Microinjection-MO, plasmid DNA, or mRNA was microinjected into embryos at the 1-2-cell stage using drawn glass microcapillary pipettes attached to a micromanipulator. Injection was driven by compressed N 2 , under the control of a PV830 Pneumatic PicoPump (World Precision Instruments, Sarasota, FL). Microinjection volumes were estimated at 1 nl per embryo. In preliminary experiments, control and antisense MOs were injected at 1, 2, 4, 5, 6, and 8 ng/embryo. A nominal concentration of 2 ng of MO8A and 6 ng of MO1A per embryo yielded consistent and reproducible effects, whereas equivalent amounts of control MO-injected embryos were indistinguishable from wild-type embryos. After injection, embryos were placed in embryo rearing medium (41) and kept at 28.5°C.
Confirmation of MO Specificity-A 201-bp DNA fragment corresponding to the 5Ј-untranslated region (100 bp containing the MO target regions) and partial coding sequence (111 bp, which encodes the signal peptide and the first 16 residues of the predicted mature protein) of the zebrafish IGFBP-3 gene was generated by PCR (forward primer, 5Ј-tagaattCTTCGGCTCCATGACGGGACTG-3Ј; reverse primer, 5Ј-at-accggtAGCGGACAAGTCAAGCAGCAC-3Ј; the linker sequences are in lowercase type). The amplified DNA was subcloned in frame into pEGFP-N1 (Clontech) using the EcoRI and AgeI sites to fuse the IG-FBP-3 partial sequence N-terminally of EGFP. Subsequently, a 934-bp DNA fragment containing the 201-bp IGFBP-3 sequence and the open reading frame of EGFP was released from pEGFP-N1-BP3UTR by digestion with EcoRI and NotI. After the NotI 5Ј overhang was filled, the blunt-ended IGFBP-3:EGFP DNA was subcloned into the pCS2ϩ vector (a gift from Dr. Victoria Prince, University of Chicago) using EcoRI and StuI sites to produce the pCS2ϩBP3:EGFP reporter construct. To generate a control EGFP reporter construct lacking the MO target sequence, the IGFBP-3 sequence was removed from the pCS2ϩBP3:EGFP by digestion with EcoRI and AgeI. The resultant pCS2ϩEGFP DNA was blunt-ended and self-ligated. The orientation and accuracy of sequence were verified by DNA sequencing.
Overexpression of IGFBP-3-Two full-length IGFBP-3 expression constructs were engineered for the rescuing experiments. First, a MOresistant native IGFBP-3 construct was generated by subcloning the entire open reading frame of zebrafish IGFBP-3 (in which 5 bp in the MO targeted sequence was mutated) into the pCS2ϩ vector. In addition, a DNA fragment containing the entire open reading frame of zebrafish IGFBP-3, in which 5 bp in the MO targeted sequence and stop codon were mutated, was generated by PCR (5Ј-GAATTCGCCACCAT-GACAGGTCTGTGTGCGCTC-3Ј/5Ј-CGACCGGTGCCTTTGTCTCCAT-GT-3Ј) and subcloned into the pCS2ϩEGFP construct using the EcoRI and AgeI sites, resulting in the pCS2ϩIGFBP-3-EGFP construct. All constructs were verified by DNA sequencing. For mRNA synthesis, the plasmids were linearized. Capped mRNA synthesis was carried out using a commercial kit (Megascript kit; Ambion, Inc., Austin, TX) following the manufacturer's instruction. The capped IGFBP-3 or EGFP mRNAs (35 pg/embryo) were introduced into one-cell embryos by microinjection as described above.
Cartilage Staining-Alcian blue was used for cartilage staining in fixed larvae following a published procedure (44). Briefly, the fixed fish were rinsed with acid-alcohol solution (0.37% HCl, 70% ethanol) and stained for 1 h with Alcian blue (0.1% in acid-alcohol). After several changes of acid-alcohol, the fish were put in 50% glycerol and 0.25% KOH solution overnight and then softened by 1% H 2 O 2 .
Statistics-Values are means Ϯ S.E. Differences among groups were analyzed by one-or two-way analysis of variance followed by Fisher's least significant difference using Stat View software (SAS Institute, Cary, NC). Significance was accepted at p Ͻ 0.05.

Temporal and Spatial Expression Patterns of IGFBP-3 mRNA in Development-
Since functional analysis of a gene often relies on its specific temporal and spatial expression pattern to form a testable hypothesis, we first mapped the IGFBP-3 mRNA expression patterns by whole mount in situ hybridization. As shown in Fig. 1, IGFBP-3 mRNA was not detected until 14 h postfertilization (hpf) (A and B), at which point IGFBP-3 mRNA was found in the cranial neural crest cells (C and D, indicated by an arrow). These cells are fated to migrate to the pharyngeal arch region to form each arch segment. At ϳ18 hpf, IGFBP-3 mRNA became detectable in the newly formed otic capsule (Fig. 1, G and H, indicated by the white arrowhead) in addition to its prominent expression in cranial neural crest cells (arrow). At 24 hpf, IGFBP-3 was detected in the cranial neural crest cells (arrow) and otic vesicles (inner ears) (white arrowhead) ( Fig. 1, I-K). A closer view of the otic vesicle indicates that IGFBP-3 mRNA was specifically expressed by the epithelial cells located in the anterior lower portion (Fig. 1L). This expression pattern persisted until ϳ30 hpf. Between 30 and 48 hpf, IGFBP-3 mRNA expression in the otic vesicle was restricted to the two patches of epithelial cells that form the anterior and posterior maculae ( Fig. 1, M-P). At 48 -60 hpf, IGFBP-3 mRNA was detected in the differentiating pharyngeal arches and in the pectoral fin buds (Fig. 1, Q-S). IGFBP-3 mRNA was no longer detected in these tissues at 72 hpf, but it remained highly expressed in the inner ears (Fig. 1T). The spatial and temporal expression pattern suggests that IGFBP-3 may play a regulatory role in the formation and/or differentiation of pharyngeal arches and inner ears.
Effects of IGFBP-3 Knockdown on Pharyngeal Arch and Otic Vesicle Formation and Differentiation-To test whether IG-FBP-3 is involved in the formation and/or differentiation of the pharyngeal arches and inner ears (otic vesicles), a MO-based loss-of-function approach was taken. For this, two antisense MOs were designed to target the IGFBP-3 gene. Their efficacy and specificity in knocking down the IGFBP-3 gene product were determined using a reporter construct (pCS2ϩBP3: EGFP). Microinjection of pCS2ϩBP3:EGFP DNA into zebrafish embryos yielded mosaic EGFP expression, clearly visible at 24 hpf by fluorescence microscopy ( Fig. 2A). Co-injection of BP3:EGFP DNA with an IGFBP-3-targeting MO abolished BP3:EGFP expression ( Fig. 2A). Similar results were obtained with MO1A. Likewise, EGFP signal was detected in BP3: EGFP-injected but not in wild-type embryos by immunoblot analysis (Fig. 2B, lanes 1 and 2). Co-injection of MO8A abolished the GFP expression, whereas co-injection of a standard control MO or an antisense MO designed against IGFBP-2, a related member in the IGFBP gene family, had no such effect (Fig. 2B, lanes 3-5). The MO knockdown effect is specific, because MO8A did not affect the expression of a control EGFP plasmid that lacks the targeting sequence (Fig. 2B, lanes 6 and  7). Similarly, embryos co-injected with IGFBP-3 MO and an expression plasmid encoding zebrafish IGFBP-2:EGFP fusion protein 2 produced embryos with mosaic GFP expression, indicating that the MOs targeted against IGFBP-3 did not suppress translation of a structurally related gene.
We next investigated the effects of IGFBP-3 knockdown on the formation and/or differentiation of pharyngeal arches. Control morpholino-injected embryos were morphologically indistinguishable from their wild-type siblings. In comparison, em-2 Y. Li and C. Duan, unpublished data. bryos injected with IGFBP-3-targeting MO8A ("morphant" embryos) exhibited underdeveloped otic vesicles and delayed pharyngeal arch formation, most evident at 48 hpf (Fig. 3,  A-D). Compared with the control embryos, morphant embryos had greatly reduced lower jaws (Fig. 3D, arrow). The morphants also had edema in the hindbrain region (Fig. 3D, arrowhead). MO1A injection resulted in similar phenotypic changes (data not shown), albeit at a higher dose (6 ng). MO8A was therefore used for subsequent experiments.
Pharyngeal cartilages are derived from migrating cranial neural crest cells (45). The appearance of pharyngeal arches is a characteristic morphogenic event during the pharyngula period, and it can be monitored by dlx2 mRNA expression (44). At 24 hpf, normal embryos form four pairs of arches: the mandibular, the hyoid, the first, and the second branchial arches. As shown in Fig. 3E, dlx2 mRNA was detected in all four postmigratory crest cells/arches. In comparison, only three postmigratory arch crests were found in the IGFBP-3 knockdown group. Likewise, the knockdown group only had four pairs of crests at 30 hpf, whereas the control embryos had already developed five postmigratory crests at this time (Fig. 3, G and H). We further compared the cartilage pattern in morphant larvae with their sibling controls using Alcian blue staining. Cartilage in neurocranium appeared relatively normal, albeit smaller, in the morphants (Fig. 3, I-L). In contrast, Meckel's and ceratohyal cartilage were absent in the morphants at 72 hpf (Fig. 3J). In addition to neurocranium and the Meckel's and ceratohyal cartilage, there are five branchial arches in the controls at 96 hpf (Fig. 3K). In the morphant group, all five branchial arches were absent, and the size of Meckel's and ceratohyal cartilage was greatly reduced (Fig. 3L). These data suggest that the loss of IGFBP-3 causes a marked delay in pharyngeal arch formation and a reduction in pharyngeal cartilage differentiation.
In addition to the pharyngeal skeletal defects, IGFBP-3 morphant embryos also exhibited defects in otic vesicles. Development of the zebrafish otic vesicle begins with the thickening of the epidermal layer into a two-cell-thick placode at 13.5 hpf (the 9-somite stage). The palcode invaginates and forms a lumen (otic vesicle) at about 18 hpf (42). Cells in the otic vesicle can be labeled by the expression of claudin-a, which is a major transmembrane protein in tight junctions and is specifically expressed in otic vesicle (46). As shown in Fig. 4, A and B, knockdown of IGFBP-3 had no effect on lumen formation. The otic vesicle contained two otoliths as in the controls. The size of the otic vesicle in the morphants, however, was markedly reduced. claudin-a expression analysis indicated that the reduced otic vesicle size was primarily due to reduced cell number (Fig. 4, C and D; also see Fig. 7B). The hair cells, grouped in two small 2-8-cell patches, begin to appear at 24 hpf. These hair cells can be visualized by antiacetylated tubulin antibody Note the reduced lower jaw (arrow) and brain edema (arrowhead) in the morphant at 48 hpf (D). E and F, whole mount in situ hybridization analysis of 24 hpf embryos for dlx2, a pharyngeal maker. At this stage, control embryos have formed four pairs of pharyngeal arches, the mandibular (labeled as 1), hyoid (2), and the first (3) and second branchial arches (4). Note the lack of the second branchial arch (4) in the IGFBP-3 morphant. G and H, whole mount in situ hybridization analysis of 30 hpf embryos for dlx2. At this stage, control embryos (G) have formed five pairs of pharyngeal arches, the mandibular (1), the hyoid (2), and the first (3), second (4), and third branchial arches (5). Note the lack of the last branchial arch (5)  staining (47; Fig. 4, E and G). In IGFBP-3-knocked down embryos, these hair cells were absent (Fig. 4, F and H), probably due to a lack of hair cell differentiation. From 42 hpf onward, columns of tissue grow out from the walls of the otic vesicle, fusing in the center of the space to generate semicircular canals (45). Sensory epithelia called cristae develop in the semicircu-lar canals, which primarily sense angular acceleration. In the control embryo, three semicircle canal protrusions are observed (Fig. 4I). These structures, however, were not seen in any morphant embryos, whereas the two otoliths are present in both groups (Fig. 4J). Whole mount in situ hybridization analysis of control-injected (Fig. 4K), and MO8A-injected (Fig. 4L) embryos at 48 hpf using a claudin-a riboprobe also revealed that the otic vesicle in the morphants remained an empty sac. To quantify the effect of IGFBP-3 knockdown on inner ear growth, we measured the length of otic vesicle in the control and morphant embryos at 48 hpf. The value (mean Ϯ S.E.) was 144.0 Ϯ 1.7 m (n ϭ 18) in the control group. The vesicle size of the morphant group was significantly smaller (104.4 Ϯ 3.6 m, n ϭ 18, p Ͻ 0.05). These results suggest that IGFBP-3 is required for the proper development and growth of inner ears.
To determine whether IGFBP-3 knockdown effects are specific to craniofacial skeletal tissues and inner ears, we examined brain, heart, and vascular patterning using specific molecular markers. eng3 is one of the engrailed transcription factors that are important regulators of midbrain-hindbrain boundary development in zebrafish (48). krox20 is a transcription factor and is expressed in rhombomeres 3 and 5 of the hindbrain (49). As shown in Fig. 5, A-D, knockdown of IG-FBP-3 did not affect their expression patterns. Likewise, there was no detectable difference in the mRNA expression pattern of nkx2.5, a cardiac marker, or fli-1, a vascular marker (50), indicating that knockdown of IGFBP-3 did not affect these tissues. Disruptions in the expression of one IGFBP gene in the mouse model has previously been shown to influence the expression of other IGFBP gene family members (33). We therefore carried out in situ hybridization analysis for IGFBP-2 mRNA, a related gene in the IGFBP gene family. As shown in Fig. 5B, IGFBP-2 mRNA expression was detected in brain boundary regions, particularly in cells along the forebrainmidbrain and midbrain-hindbrain boundaries. Knockdown of IGFBP-3 had no detectable effect on IGFBP-2 mRNA expres-  sion, suggesting that there is no compensatory increase in IGFBP-2 gene expression due to the loss of IGFBP-3.

Rescuing IGFBP-3 Knockdown-induced Defects by Overexpressing a MO-resistant IGFBP-3-
To prove that the defects in the IGFBP-3 morphant embryos were indeed due to the loss of IGFBP-3, pCS2ϩIGFBP-3-EGFP, a MO-resistant IGFBP-3 construct, was engineered. To confirm that the IGFBP-3-EGFP is a functional IGFBP, IGFBP-3-EGFP capped mRNA was injected into embryos. Western immunoblot analysis of the injected embryos using a GFP antibody detected a major band at ϳ60 kDa, corresponding to the IGFBP-3-EGFP fusion protein (Fig. 6A, left). This protein was not detected in wild-type embryos. Western ligand blot analysis revealed that IGFBP-3-EGFP bound to IGF-I (Fig. 6A, right). To show that the encoded IGFBP-3 is indeed resistant to MOs, we introduced the pCS2ϩIGFBP-3-EGFP DNA into zebrafish embryos by microinjection. This yielded mosaic GFP expression (Fig. 6B, b). Co-injection with 2 ng of IGFBP-3 targeting antisense MO (MO8A) had no effect on the IGFBP3-EGFP expression (Fig.  6B, c). At this dose, MO8A completely abolished the BP3:EGFP reporter gene expression (Fig. 2).
Next, the rescue effect of the MO-resistant intact IGFBP-3 was examined. As shown in Fig. 7A, control morpholino-injected embryos were morphologically and biochemically indistinguishable from their wild-type siblings (Fig. 7A, a and b; e and f; i and j; and m and n). Compared with the control embryos, morphant embryos had smaller and underdeveloped otic vesicles (Fig. 7A, c, k, and o). Whereas the wild-type and control MO-injected embryos had four pairs of pharyngeal arches (e and f), the IGFBP-3-knocked down embryos had only three pairs (g). Overexpression of the MO-resistant IGFBP-3 "rescued" this phenotype (Fig. 7A, h). In fact, the dlx2 mRNA expression patterns of embryos in the IGFBP-3 mRNA plus MO8A group were indistinguishable from that of the control and wild-type groups. The body size of the IGFBP-3-injected embryos, however, was somewhat smaller. Similar "rescuing effects" were obtained with otic vesicle-associated phenotypes. As shown in Fig. 7A, l, co-injection of IGFBP-3 mRNA restored the IGFBP-3 knockdown-induced reduction in otic vesicle size. To further quantify these effects, we determined the number of epithelial cells in the otic vesicles. As shown in Fig. 7B, in the IGFBP-3-knocked down embryos, there were 23.84 Ϯ 0.37 cells per optical section (n ϭ 24). This value is significantly lower than that of the wild-type (31.11 Ϯ 0.34, n ϭ 27) and control MO-injected groups (30.94 Ϯ 0.54, n ϭ 16), respectively (p Ͻ 0.05). Overexpression of IGFBP-3 restored the epithelia cell number back to the control levels (29.52 Ϯ 0.52, n ϭ 21). Overexpression of the MO-resistant IGFBP-3 also restored the hair cell differentiation (Fig. 7A, p). At 24 hpf, 100% of wildtype embryos (n ϭ 39) and control embryos (n ϭ 7) had already formed two clusters of differentiated hair cells. In comparison, 29 of 48 (60.42%) morphant embryos examined completely lacked these two clusters of cells, and the remaining 19 morphant embryos (39.58%) showed markedly reduced signal. In the IGFBP-3 mRNA and MO8A co-injected embryos, 37 of 46 (80.43%) of the embryos had the two spots of hair cells. These results confirm that exogenous IGFBP-3 is sufficient to rescue the changes of pharyngeal arches and otic vesicle caused by MO injection, verifying that these developmental defects are caused by the loss of IGFBP-3 function.

DISCUSSION
In this study, we show that IGFBP-3 mRNA exhibits spatially and temporally distinct patterns of expression during zebrafish embryogenesis. Specifically, IGFBP-3 mRNA is predominantly expressed in the developing inner ears and the pharyngeal arches. Using a MO-based targeted gene knockdown approach, we demonstrate that loss of IGFBP-3 resulted in developmental defects in the pharyngeal arches and inner ears. Furthermore, reintroduction of a MO-resistant IGFBP-3 to an IGFBP-3 knockdown background rescued these defects. These data suggest that IGFBP-3 plays a critical role in regulating pharyngeal cartilage and otic vesicle development and growth in zebrafish.
Previous studies have shown that IGFBP-3 is produced in many mammalian fetal tissues (8 -15). These studies focus on a single or a few fetal tissues for a limited period (almost all in advanced fetal stages after the completion of organogenesis), and few studies surveyed the majority of tissues during the entire embryogenesis using a standardized methodology. The developmental expression of IGFBP-3 has not been reported in a nonmammalian vertebrate until this study. Taking advantage of the free living, transparent, and fast developing zebrafish embryos, we have determined the spatial and temporal expression pattern of IGFBP-3 throughout embryogenesis. Our results reveal that IGFBP-3 mRNA expression begins at the end of the gastrula stage and continues throughout embryogenesis in zebrafish. Spatially, IGFBP-3 mRNA is highly expressed in the developing otic vesicles and pharyngeal arches and their precursor cells. The temporal patterns in these two expression domains are somewhat different. Whereas IGFBP-3 mRNA is persistently expressed in the epithelial cells in the developing inner ears, its expression in the cranial skeletal tissues is more dynamic. IGFBP-3 mRNA is initially expressed in the cranial neural crest cells and subsequently in pharyngeal arches. Its expression in these tissues ceases after 72 hpf. The spatial expression pattern of IGFBP-3 mRNA in zebrafish embryos varies somewhat from what has been reported in mammals. In mammals, fetal liver is one of the primary expression sites (10,15,51). We were unable to detect IGFBP-3 mRNA in the liver in zebrafish embryos and larvae. In contrast, IGFBP-1 mRNA is easily detected in the liver of zebrafish embryos (38). Determination of whether the zebrafish IGFBP-3 gene is expressed in the liver in juvenile and/or adults awaits future studies.
A major finding in this study is that IGFBP-3 plays an important role in regulating pharyngeal cartilage and otic vesicle development and growth in vivo. This conclusion is sup- ported by several lines of evidence. First, IGFBP-3 mRNA is highly expressed in the developing inner ear and the pharyngeal arches. Second, targeted knockdown of IGFBP-3 resulted in delayed pharyngeal skeleton morphogenesis, reduced pha-ryngeal cartilage differentiation, reduced vesicle size, and defects in hair cell and semicircular canal differentiation. In comparison, knockdown of IGFBP-3 did not affect the formation and differentiation of tissues/organs that have no or low IGFBP-3 mRNA, such as midbrain, hindbrain, heart, and vasculature. Furthermore, overexpression of a MO-resistant form of IGFBP-3 in the IGFBP-3 knocked down embryos rescued these IGFBP-3 MO-caused defects.
Although IGFBP-3 knock-out mice have been generated (32), these animals appeared normal in size and shape. There are several possible explanations for the difference between the mouse study and our study using the zebrafish model. First, there appears to be greater redundancy in the spatial and temporal expressions of different members of the IGFBP gene family in rodents. Previous studies have shown that most rodent fetal tissues/cell types produce multiple IGFBPs (8 -15). In comparison, there is little overlap in the expression domains of different IGFBPs during zebrafish embryogenesis (5,33,37) (this study). Second, disruption to the expression of one component of the IGF system often influences the expression of other IGF family members in the mouse model (33,52). For instance, it was reported that serum levels of IGFBP-1, -3, and -4 were increased in IGFBP-2 knock-out mice (33). The compensatory adjustments in the expression of other IGFBPs may have minimized deleterious effects of the loss of one IGFBP gene. In contrast, knockdown of the zebrafish IGFBP-3 gene did not alter the expression of IGFBP-2 in zebrafish embryos. Third, there exists the possibility of maternal compensation in gene knock-out progeny, via placental circulation, in the mouse model. In comparison, zebrafish embryos develop externally in the absence of placental compensation. These findings underscore the utility of the zebrafish model in studying IGF/IGFBP biology, especially during early development.
The mode of action of IGFBP-3 in regulating pharyngeal arch and otic vesicle growth and development is not clear at present. In vitro studies using a variety of cultured mammalian cells have shown that IGFBP-3 is capable of inhibiting or potentiating IGF actions and even possesses IGF-independent actions (3,4). Recently, we have shown that MO-based knockdown of the zebrafish IGF-I receptors also resulted in defects in pharyngeal arch formation, reduced vesicle size, and defective hair cell and semicircular canal differentiation along with other phenotypes (53). Therefore, it is possible that IGFBP-3 may regulate pharyngeal arch, and otic vesicle growth and development is through its interaction with IGF-I and/or IGF-II. Future studies are needed to elucidate the mechanisms of IG-FBP-3 action and how IGFBP-3 works together with IGFs and/or other molecules in vivo.
In summary, we have shown that IGFBP-3 mRNA is expressed in temporally and spatially specific patterns during zebrafish embryonic development and that targeted knockdown of IGFBP-3 results in delayed pharyngeal skeleton formation and reduced cartilage and otic vesicle differentiation, and reintroduction of IGFBP-3 to the knocked down embryos rescued these developmental defects. The defects in skeletal and inner ear development are localized in regions associated with high levels of IGFBP-3 mRNA expression, suggesting an important paracrine function for IGFBP-3. To our knowledge, this study is the first to provide in vivo evidence for the requirement of IGFBP-3 in vertebrate pharyngeal skeletal and inner ear development.