Receptor-Ligand Interaction between Vitellogenin Receptor (VtgR) and Vitellogenin (Vtg), Implications on Low Density Lipoprotein Receptor and Apolipoprotein B/E

The vitellogenin receptor (VtgR) belongs to the low density lipoprotein receptor (LDLR) gene family. It mediates the uptake of vitellogenin (Vtg) in oocyte development of oviparous animals. In this study, we cloned and characterized two forms ofOreochromis aureus VtgR. Northern analysis showed that VtgR was specifically expressed in ovarian tissues. However, reverse transcription-PCR indicates that either there are trace levels of expression of VtgR or a homolog of LDLR exists in nonovarian tissues. The VtgR is highly homologous to the very low density lipoprotein receptor. To better understand the mechanism by which similar structural modules in the ligand-binding domain bind different ligands, we used the yeast two-hybrid system to screen for the minimal interaction motifs in Vtg and VtgR. The amino-terminal region of the lipovitellin I domain of Vtg interacts with the ligand-binding domain of VtgR. The first three ligand-binding repeats of the receptor were found to be essential for ligand binding. Computational analysis of the binding sequence indicates that Vtg has a similar receptor-binding region to apolipoprotein (apo) E and apoB. Site-directed mutagenesis of this region indicates electrostatic interaction between Vtg and its receptor. Sequence analysis suggests the coevolution of receptor-ligand pairs for the LDLR/apo superfamily and suggests that the mode of binding of LDLR/very low density lipoprotein receptor to apoB and apoE is inherited from the electrostatic attraction of VtgR and Vtg.

ovary via blood circulation. This major yolk precursor protein binds to vitellogenin receptor (VtgR) on the surface of oocytes and is taken up by receptor-mediated endocytosis (1). A large amount of Vtg accumulates in the oocytes within a relatively short time. Once in the oocytes, Vtg is cleaved into yolk proteins, namely, lipovitellin (LV) and phosvitin, which are stored as nutrients for the developing embryo (2). Sequence analysis showed that the amino-terminal 700 amino acids of Vtg and apolipoprotein (apo) B-100 are homologous, although the similarity is limited (3). Coincidentally, Vtg also binds lipids and transports them into the oocytes. The sequence and functional relationship of these two proteins support the idea that they have a common ancestor.
VtgR belongs to the low density lipoprotein receptor (LDLR) family (4). The members of this family bind to various ligands and are involved in lipid metabolism in both vertebrates and invertebrates. These receptors have common structural features (5,6) including (i) cysteine-rich ligand-binding repeats (LBRs), (ii) cysteine-rich epidermal growth factor precursor (EGFP)-like repeats spaced by cysteine-poor spacer regions, (iii) a single transmembrane domain, and (iv) a short carboxylterminal cytoplasmic tail. In addition, a short region highly enriched in serine and threonine residues may exist in some receptors. The number of LBRs varies among different receptors. LDLR contains seven LBRs, whereas very low density lipoprotein receptor (VLDLR) and VtgR in vertebrates have eight LBRs. Larger receptors such as LDLR-related protein and megalin have more than 30 LBRs in several clusters (7,8).
Each LBR consists of about 40 amino acids including 6 cysteine residues, participating in the formation of three disulfide bonds, which are crucial for its proper folding (9). At the carboxyl terminus of each LBR, there is a consensus acidic tripeptide, Ser-Asp-Glu (SDE). Recent structural study by NMR and x-ray diffraction analysis of LBRs 1, 2, 5, and 6 from LDLR have revealed that the side chains of many of the aspartate and glutamate residues in the consensus peptides are involved in coordinating the calcium ion into a folded calcium cage (10 -15).
The binding sites of Vtg for VtgR were presumed to be located on the lipovitellin I domain, LV1 (16). Residue modification studies showed that lysine and arginine residues were important for binding with the acidic clusters in LBR of VtgR through ionic interactions (17). However, new structural studies of LBR (10 -15) indicate that those acidic residues might not be accessible to Vtg. This necessitates a reassessment of cur-rent models for the binding of VtgR to Vtg. Because the sequences of LBRs in different receptors are highly homologous, their backbone structures are very likely to be identical (9). The distinct affinity to different ligands may result from differential participation of individual LBRs, for example, repeat 5 is essential for binding of apoE, and repeats 2-7 cooperatively bind apoB (18). Thus, Vtg binding may require the involvement of different LBRs of VtgR.
Given the relationship among VtgR, LDLR, and VLDLR and the relationship between Vtg and apoB, the understanding of Vtg-VtgR recognition would contribute insights into the mechanism of interaction between LDLR gene family members and their ligands. Many Vtg and VtgR genes have been cloned and characterized in recent years (19). However, knowledge on Vtg-VtgR interaction remains limited. In our study, we have cloned full-length and different domains of Vtg and VtgR from tilapia and examined their interactions. Using yeast two-hybrid and GST pull-down assays, we found that the minimal binding domain of VtgR was the first three LBRs and that there might be more than one binding site. The receptor recognizes the 84-amino acid fragment in the amino-terminal of Vtg. Site-directed mutagenesis study indicates that lysine 185 in Vtg is particularly important for electrostatic interaction with VtgR.
Northern Analysis-The total RNAs from ovary, liver, muscle, brain, spleen, and intestine were purified using Trizol reagent (Invitrogen). Aliquots of 20 g of total RNAs were resolved in 1% agarose formaldehyde gel. The separated RNAs were transferred onto a nylon membrane (GeneScreenPlus; PerkinElmer Life Sciences) using capillary transfer. The membrane was UV-cross-linked, prehybridized in DIG Easy Hyb (Roche Molecular Biochemicals) for 4 h at 50°C, and hybridized overnight at 50°C in Easy Hyb with DIG-labeled PCR fragment B as probe. After hybridization, the membrane was washed twice in 2ϫ SSC, 0.1% SDS for 15 min each at room temperature and twice in 0.5ϫ SSC, 0.1% SDS for 15 min each at 65°C. The membrane was incubated for 1 h in blocking solution followed by a 30-min incubation in anti-DIG-AP (Roche Molecular Biochemicals) diluted 1:10,000 in blocking solution. After incubation, the membrane was washed twice in washing buffer (100 mM maleic acid, 150 mM NaCl, 0.3% Tween 20, pH 7.5), and the hybridized band was detected by chemiluminescence (CSPD reagent; Roche Molecular Biochemicals). The ␤-actin cDNA from zebrafish was also labeled with DIG and used as a normalization control.
RT-PCR Analysis-Approximately 500 ng each of mRNA from ovary, liver, muscle, brain, spleen, and intestine, isolated using the Oligotex Direct mRNA kit, was reverse transcribed with oligo(dT) and Thermoscript reverse transcriptase. The cDNAs were amplified by PCR using 0.2 M concentrations of C1 (5Ј-GTGCTCCAGTCTTCAGAG-3Ј) and G (5Ј-ATCGCGGGGTACGTGTG-3Ј) primers that are specific for VtgR, in a final volume of 50 l, containing 2 units of platinum Taq DNA polymerase, 20 mM Tris-HCl (pH 8.4), 50 mM KCl, and 1.7 mM MgCl 2 . The PCR reaction was carried out by incubating the samples at 94°C for 1 min, followed by 35 cycles of 94°C for 30 s, 52°C for 30 s, and 72°C for 1 min.
Yeast Two-hybrid Constructs-Unless otherwise stated, the Vtg and VtgR insert cDNAs were generated by linker PCR using platinum Pfx polymerase. The Vtg and VtgR cDNAs were then cloned in-frame into pGADT7 and pGBKT7 yeast two-hybrid vectors (Clontech), respectively. The template cDNAs used for VtgR constructs were reverse transcription products of ovary mRNA. The cDNAs for Vtg constructs were derived from pOAVtg1 (20). The full-length cDNA of VtgR flanked by NdeI-XmaI was cloned into pGBKT7 to obtain pGBK-VtgR. All sequential LBR deletion constructs were PCR-amplified and cloned into pGBKT7 to obtain the respective pGBK-LBR constructs. The VtgRs lacking the LBR domain were PCR-amplified and inserted into pG-BKT7 NdeI-XmaI sites to obtain pGBK-NLBR. The full-length Vtg insert was PCR-amplified and cloned into NdeI sites of pGADT7. The LV1 of Vtg was cloned into NdeI-XmaI sites of pGADT7. The deletion constructs of Vtg were obtained by digesting pGADVtg with the corresponding enzymes and religating the plasmids. All constructs were sequenced to confirm the inserts.
␤-Galactosidase Reporter Assay-The yeast strain Y187 (Clontech) was cotransformed by the lithium acetate method with pGAD and pGBK plasmids encoding the DNA binding domain and activator fusion protein, respectively. As controls, the pGAD constructs and pGBK constructs were also cotransformed with either pGBKT7 or pGADT7 vector. The transformed yeast cells were plated on SD minimal medium (Clontech) agar plates supplemented with -Trp-Leu DO supplement (Clontech). The cultures were grown for 4 days at 30°C, until colonies reached 1-2 mm in diameter. Three individual colonies were inoculated into SD liquid cultures for ␤-galactosidase assays. The ␤-galactosidase activities were quantified using ␤-galactosidase assay kits (Pierce). The absorbance of the cultures at 600 nm was recorded. Aliquots of 350 l of each mid-log phase liquid culture were mixed with 350 l of Y-PER reagent (Pierce), followed by 300 l of a 1 M Na 2 CO 3 stop solution when the yellow color appeared. As a control, 350 l of SD medium was assayed as blank. The absorbance of each sample at 420 nm was recorded against the blank.
In Vitro Pull-Down Assays-The ligand-binding domain of VtgR was amplified by PCR and cloned into the Escherichia coli expression vector, pGEX4T-3 (Amersham Biosciences), in-frame with GST. The cDNA fragment corresponding to amino acids 162-246 of Vtg was amplified using PCR and cloned in-frame into the NdeI-EcoRI site of expression vector pET22-b (Novagen). To facilitate cloning, these restriction sites were added to the primers in both cases. The recombinant plasmids were transformed into E. coli strain BL21. The bacterial culture was grown at 25°C to A 600 nm ϭ 0.6 -0.8 and then induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h. Cells harvested from 10 ml of culture were resuspended in 0.5 ml of phosphate-buffered saline and sonicated. Triton X-100 was added to 1%, and the cell debris was removed by centrifugation at 13,000 ϫ g for 30 min. Glutathione-Sepharose 4B (Amersham Biosciences) was incubated overnight at 4°C with GST or GST-LBR lysate, either with or without 2 mM DDT. The beads were washed three times with 1ϫ phosphate-buffered saline and 1% Triton X-100 and then incubated with the second lysate of pET22-b or pET22b-Vtg construct. The beads were washed three times with 1ϫ phosphate-buffered saline and 1% Triton X-100, and the eluates were analyzed on 15% SDS-PAGE. Before using cell lysates in binding assays, aliquots of each protein sample were analyzed on SDS-PAGE gels and stained with Coomassie Blue to estimate protein concentrations, and sample volumes were adjusted to equalize the amounts of fusion proteins used in each assay.
Site-directed Mutagenesis-The QuikChange XL mutagenesis kit (Stratagene) was used to mutate 3 residues in Vtg and 1 residue in VtgR. These are sites that were anticipated to contribute crucially to their interaction. The primers were designed according to the manual. The mutation reaction was carried out by incubating samples at 95°C for 1 min, followed by 18 cycles of 94°C for 50 s, 60°C for 50 s, and 68°C for 28 min. After PCR, the sample was incubated with 1 l of DpnI at 37°C for 1 h to digest the template DNA. Five l of reaction was used to transform the XL10-Gold competent cells. The mutation was confirmed by sequencing. The mutated constructs were used to transform the yeast, and the resulting interaction was analyzed by ␤-galactosidase assay.

RESULTS
Two Forms of VtgR cDNA Were Identified-To clone VtgR cDNA, RT-PCR was performed based on primers designed to obtain a 688-bp fragment A in the EGFP-like domain (Fig. 1).
The sequence information was used to design another two pairs of primers. By PCR, two fragments (fragments B and C) overlapping fragment A were cloned and sequenced. Fragment B was 734 bp, encompassing LBRs 5-8. Two forms of fragment C, of 580 and 520 bp, were observed in the PCR product. The corresponding amino acid sequence shows that the 580-bp cDNA has the O-linked sugar domain, which is a threonineand serine-rich region. The 520-bp fragment lacks this region. The primers for 5Ј-and 3Ј-RACE were used to obtain the full-length cDNA sequence (GenBank TM accession number AF514281). The analysis of cDNA sequences revealed two open reading frames of 2500 and 2560 bp, encoding 800 and 820 residues, respectively. The alignment of the amino acid sequences of VtgR to the VLDLR and VtgR from other species showed high homology.
Differential Expression of VtgR in Various Tissues-To examine the transcription of the VtgR gene, RT-PCR was carried out with C1 and G primers flanking the O-linked sugar domain. All the tissues consistently exhibited two forms of VtgR mRNA of 420 and 360 nucleotides ( Fig. 2A). The size difference between these two forms was probably attributable to the differential splicing of a short region, as observed in the comparison of the sequences of these two PCR products. By aligning with VtgR and VLDLR genes of other species, this region was found to be located in the O-linked sugar domain. Interestingly, when Northern analysis was performed using fragment B (containing LBRs 5-8) as the probe, only the ovary exhibited one transcript of VtgR mRNA of 3.3-kb nucleotides (Fig. 2B). These data indicate that the VtgR gene was transcribed in both ovarian and nonovarian tissues. However, only the ovarian VtgR mRNA is sufficiently abundant to be detectable by Northern analysis. In our study, all the RT-PCR products from different tissues share the same sequence. This result excludes the possibility that the RT-PCR amplified the LDLR that is homologous to VtgR. When fragment C was obtained from ovarian mRNA by RT-PCR, the dominant product was the one lacking the O-linked sugar domain. Additionally, in 3Ј-RACE, the same product was obtained (data not shown). These data indicate that the major form of VtgR in the ovary lacks the O-linked sugar domain. In chicken, the VtgR was reported to function as VLDLR/VtgR in different tissues (4). The existence of VtgR mRNA in nonovarian tissues suggests that this receptor may function as VLDLR or VtgR in different piscine tissues.
The LBR of VtgR Is Sufficient for Vtg Binding-In an effort to locate the important interactive domains between VtgR and Vtg, we first examined the interactions between full-length Vtg and different domains of VtgR using yeast two-hybrid assays to confirm that the yeast two-hybrid assay is suitable for the study of interaction between Vtg and VtgR. Although two forms of VtgR were identified in the ovary, the results indicate that the major form lacks the O-linked sugar domain. Therefore, the VtgR constructs used in the interaction studies were from the VtgR lacking O-linked sugar domain. Three constructs of VtgR, full-length VtgR, the extracellular part of VtgR, and the LBR of VtgR, were tested for binding to Vtg. The quantitative ␤-galactosidase assay confirms the binding of the ligand-binding domain of VtgR or extracellular part of VtgR to full-length Vtg in vivo (Fig. 3). However, full-length VtgR did not show binding with full-length Vtg. This may be due to hindrance by the transmembrane domain on the transportation of VtgR fusion protein into the nucleus, where the interaction-dependent activation of reporter gene transcription occurs in the yeast twohybrid assay. Although some earlier studies have indicated that the EGFP-like domain might be required for ligand binding (21,22), our results clearly confirm that deletion of other parts of VtgR did not disrupt the binding between LBRs and Vtg. The controls showed that the interactions were not within the fusion proteins and the activation domain or DNA-binding domain encoded by the pGADT7 or pGBKT7 vector. These data suggest that the LBR itself is sufficient for binding Vtg. The Two different products, C big (656 bp) and C small (596 bp), were amplified. Primer D was used to obtain the 5Ј end fragment D (872 bp) by 5Ј-RACE, and primer C1 was used to derive the 3Ј end fragment E (1792 bp) by 3Ј-RACE. Primer G was used to study the expression of two forms of VtgR by RT-PCR analysis.

FIG. 2. Tissue-specific expression of VtgR.
A, detection of VtgR mRNA in different tissues by RT-PCR. mRNA at 200 ng each from ovary, liver, spleen, muscle, intestine, brain, and heart was used as template for reverse transcription, followed by amplification with two primers, C1 and G flanking O-linked sugar domain. Aliquots of 10 l from each reaction were loaded onto the agarose gel together with 100-bp DNA marker (M). B, Northern analysis of VtgR mRNA in different tissues. mRNA at 200 ng each from liver, spleen, muscle, intestine, brain, heart, and ovary was loaded onto the agarose gel, and RNA was transferred to the membrane. The membrane was probed with VtgR fragment B and normalized with zebrafish ␤-actin.
EGFP homology domain is not necessary for the receptor-ligand interaction. Therefore, further studies on the receptor focused on the LBD. LBRs 1-3 of VtgR Are Involved in the Interaction with Vtg-To determine which subdomains of the receptor are important to Vtg binding, 25 different fragments of the VtgR ligand-binding domain were cloned into pGBKT7. The interaction between the pGBKT7-VtgR deletion constructs and pGAD-Vtg was tested by the yeast two-hybrid assay. All the conditions used in the transformation of yeast and the enzymatic assays for all the constructs were the same. This made the yeast two-hybrid assay semiquantitative. The difference in the ␤-galactosidase activities most probably arose from the difference in the interaction.
Five carboxyl-terminal deletion constructs of LBD, including LBR1-7, LBR1-6, LBR1-5, LBR1-4, and LBR1-3, did not disrupt the binding to Vtg. The two carboxyl-terminal deletion constructs of LBD, namely, LBR1-2 and LBR1, lost binding to Vtg (Fig. 4). When the first LBR was deleted in constructs LBR2-8, LBR 2-7, LBR 2-6, and LBR2-3, dramatic drops in the ␤-galactosidase activities were observed. LBR2 alone did not show significant binding to Vtg. When both the first and second LBRs were deleted, all 11 constructs lost binding of Vtg. It is therefore clear that the first three LBRs are most important for binding Vtg. Without LBR3, the construct LBR1-2 exhibited weaker interaction with Vtg. On the other hand, constructs containing only LBR3 did not bind Vtg. Similarly, the deletion of LBR1 from various combinations of LBD (viz., For each construct, the result was the mean of three individual transformed colonies assayed. The A 420 nm of the ␤-galactosidase assay reactions and A 600 nm of the culture were recorded. The relative ␤-galactosidase activities were obtained by calculating the ratio of A 420 nm to A 600 nm . Full-length Vtg was fused with the activation domain (AD), and the different constructs of VtgR were fused with the DNA-binding domain (BD). To rule out nonspecific binding between the Vtg/VtgR and AD/BD backbones, the constructs containing Vtg or VtgR domains were also assayed for potential interactions with the vector backbones, pGADT7 or pGBKT7.

FIG. 4. The interactions between LBD of VtgR and different deletion constructs of Vtg.
Seven carboxyl-terminal deletion constructs of Vtg and fulllength Vtg were fused with activation domain (AD) and assayed for their interaction with the LBD of VtgR. The first three LBRs are most critical for VtgR-Vtg interaction. Individually, LBR1, LBR2, or LBR3 is nonfunctional, thus suggesting the global effects of LBRs 1-3 on binding Vtg. constructs LBR2-8, LBR2-7, LBR2-6, LBR2-4, and LBR2-3) attenuated but did not abolish the interaction with Vtg. These data indicate that the interaction between LBD and Vtg utilizes more than one LBR in the LBD. There may be individual binding sites in the first three LBRs. However, LBRs 1-3 are not mutually exclusive in their interaction with Vtg. It is possible that one single binding site is insufficient to stabilize the interaction. Thus, the minimum ligand-binding region is a combination between either LBR1 and LBR2 or LBR2 and LBR3 or LBR1 and LBR3. Thus, the region required for maximal binding of Vtg is LBRs 1, 2, and 3.
The LBR of VtgR Interacts with the Amino-terminal Fragment of Vtg between Ala 162 and Ile 246 -To define the actual binding sites in Vtg for VtgR, seven 3Ј deletion constructs of Vtg together with full-length Vtg were tested for interaction with the VtgR LBRs (Fig. 5). The LBRs interacted with fulllength Vtg and interacted to different extents with the following deletion constructs: pGADVtg-ClaI (1584 amino acids), pGADVtg-XmaI (1505 amino acids), pGADLV1 (1089 amino acids), pGADVtg-BamHI (286 amino acids), and pGADVtg-EcoRI (246 amino acids). No interaction was observed with pGADVtg-SacI (162 amino acids) and pGADVtg-XhoI (52 amino acids). Deletion of Vtg upstream of 246 amino acids completely abolished its binding to VtgR. Thus, the Vtg binding site is either between amino acids 162 and 246 or between amino acids 1 and 162. The deletion from 162 to the carboxyl terminus may cause a drastic change in the conformation of the upstream region that is necessary for binding with VtgR. Therefore, to delineate the actual binding site, the aminoterminal 84-amino acid fragment (flanked by SacI and EcoRI), VtgSE, containing amino acids 162-246 was subcloned into the pGAD vector to examine its interaction with LBR. The ␤-galactosidase assay shows that this VtgSE fragment has a binding capacity similar to that of full-length Vtg (Fig. 6A), thus indicating that this region contains the crucial binding site for VtgR and that deletion of other parts of Vtg does not seem to affect its binding to receptor.
Direct Binding of LBR and Vtg in Vitro-To test whether the interaction between LBD and the short fragment VtgSE is direct and to confirm the results of the yeast two-hybrid assay, the LBD and VtgSE were tested by in vitro pull-down assay. LBR and VtgSE were expressed as GST fusion protein and His fusion protein, respectively, in E. coli. The GST and GST-LBR proteins were immobilized on the glutathione-Sepharose beads and incubated with VtgSE cell lysate. After extensive washing, the elution of GST-LBR contained VtgSE, whereas the elution of GST did not (Fig. 6B). Thus, it is clear that the LBR and VtgSE interact directly in vitro. In the presence of DDT, the GST-LBR did not bind VtgSE. This suggests that the interaction was dependent not only on the charged residues in LBRs but also on the integrity of the disulfide bonds, which are crucial for the three-dimensional structure to confer the functional LBD.
Site-directed Mutagenesis of Vtg Reveals the Importance of Lys 185 for Interaction with VtgR-By aligning the sequence of the binding site in Vtg with the apoB major binding site and the apoE binding site for LDLR (23,24), a short motif in Vtg with basic residues was observed (Fig. 7A). This suggests that vitellogenin may utilize the same mechanism as apoB and apoE to bind the VtgR. To test this hypothesis, we mutated the positively charged amino acids at His 182 , Lys 185 , and Lys 187 in VtgSE into This result indicates that the amino-terminal Vtg region upstream of EcoRI contains the binding site for VtgR.

FIG. 6. Direct binding between VtgSE and VtgR is confirmed by GST pull-down assay.
A, yeast two-hybrid assay of the interaction between VtgSE and the LBD of VtgR. VtgSE was fused with the activation domain (AD) and tested for interaction with the LBD of VtgR. As the positive control, the interaction between full-length Vtg and the LBD of VtgR was also assayed. B, in vitro direct binding between VtgSE and LBD shown by GST pull-down. GST-LBD was expressed in E. coli and coupled to glutathione beads with or without 2 mM dithiothreitol. After washing, the expressed VtgSE was incubated with the beads. The unbound proteins were washed away, and the bound proteins were analyzed on SDS-PAGE. The protein bands were visualized by Coomassie Blue staining. Interaction between GST (without LBD) and pET22-b vector backbone was used as a negative control.
alanine. These residues are correspondingly important in apoB and apoE for binding to their respective receptors. We examined the effect on binding of VtgR by yeast two-hybrid assay. The mutation constructs VtgSE(H182A) and VtgSE(K187A) showed similar binding of LBD and LBR1-3. However, the binding between the mutation construct VtgSE(K185A) and either LBD or LBR1-3 is attenuated (Fig. 8). This result highlights the importance of the residue Lys 185 in the interaction between Vtg and VtgR.
When the sequences of different VtgR LBRs were aligned (Fig. 7C), the signature sequences of the LDLR superfamily emerged as well conserved. In LBR3, the three conserved acidic residues are EDE, whereas in all other LBRs, the sequence is SDE. Bajari et al. (25) have searched for the minimal binding site in chicken VtgR for receptor-associated protein (RAP) and proposed that the EDE region in LBR3 might be important for ligand binding because it has the highest negative charge density. Consistently, our study has empirically confirmed that LBR3 is critical to the interaction with Vtg. To demonstrate a potential relationship between this subtle difference in the acidic residues in LBR3 and its affinity for Vtg, we mutated the E144S in both the LBD and LBR1-3 and tested their interac-tions with Vtg and VtgSE. However, this mutation did not affect the binding between LBD/LBR1-3 and Vtg/VtgSE. Furthermore, the mutation of SDE to EDE in LBR6 of LBR4 -7 did not show any gain of function for binding Vtg (data not shown). Hence, contrary to chicken VtgR, the increased negative charge density in EDE of piscine VtgR is not related to ligand binding. Instead, we propose that the EDE region in LBR3 is more likely to play a role in the formation of the calcium cage, which is also found in other LBRs of LDLR (9).

DISCUSSION
In this study, using RT-PCR and 5Ј-and 3Ј-RACE, we cloned a piscine VtgR, a member of the LDLR family. The VtgR contains eight low density lipoprotein complement type A ligand-binding repeats. Northern analysis indicates that this receptor is expressed mainly in ovarian tissues. Using the yeast two-hybrid system, we show that the ligand-binding domain is sufficient to bind Vtg. The deletions of different LBRs provide evidence that LBRs 1, 2, and 3 constitute the major ligandbinding subdomain in VtgR for binding of Vtg. The binding sites within Vtg are localized to 84 amino acids at the aminoterminal of LV1, between amino acids 162 and 248. Sequence analysis of the binding sites in Vtg and VtgR together with other members of LDLR and apolipoprotein suggests that VtgR may interact with Vtg via electrostatic attraction.
The piscine VtgR is highly homologous to the VtgR of amphibians, birds, and the LDLR/VLDLR of mammals. Both VtgR and VLDLR contain eight type A repeats in the ligand-binding domain. It was reported that the receptor in chicken functions as both VtgR and VLDLR in different tissues (26). In mammals, members of the LDLR family function not only in receptor-mediated endocytosis but also in transducing signals that are important during embryonic development (27). Given the fact that VLDLR/VtgR plays diverse roles in different tissues, it was proposed that they could be uniformly named LR8, according to the number of repeats in ligand-binding domain (28). The existence of two forms of VtgR, one with and one without O-linked sugar domain, was also reported in other species and in other tissues (28 -30). The function of this Olinked sugar domain is still unclear. It may be responsible for controlling receptor recycling and degradation (31,32). In the present study, Northern analysis showed an apparent single transcript. This may be attributable to the difference between the two forms of mRNA, which is subtle and very short (60 bp), compared with full-length VtgR mRNA (ϳ3.5 kb). Previous studies indicated that lack of this 60-bp region would not affect the binding of the receptors to their ligands. The presence of VtgR transcript in the nonovarian tissues suggests that there are piscine versions of VLDLR/LDLR, which have distinct functions from VtgR in the ovarian tissues. However, the mRNA levels of both forms of VtgR, with or without O-linked sugar domain, in nonovarian tissues are very low. It is still not clear whether this trace amount of transcription has any distinct biological significance.
The ligand-binding domains in members of the LDLR family have been studied extensively. It is interesting that the highly conserved ligand-binding domains can form promiscuous interactions with various ligands. Analysis of the naturally occurring and engineered mutations in LDLRs has shown that LBRs 2-6 are required for the binding of apoB, whereas the binding of apoE depends critically on LBR5. Previous efforts to search for the minimal binding domain in chicken VtgR indicated that LBR3 is important for binding of RAP (25). It was reported that the minimal binding unit in LDLR for RAP binding was any two adjacent repeats (33). There is, however, no report on the participation of LBRs in the interaction with Vtg. Using the yeast two-hybrid system, we show that binding to Vtg requires LBRs 1, 2, and 3. Savonen et al. (34) reported that RAP interacts equally well with repeats 1-3 and 1-5, but not with repeats 6 -8 in VLDLR. An analysis by Mikhailenko et al. (35) suggested that the RAP binding site of VLDLR is located within the amino-terminal four repeats and also suggested that the first three repeats are especially important for RAP binding. Consistent with these reports, we also observed the importance of the first three LBRs to binding Vtg. LBR3 is the only repeat that contains EDE instead of the consensus sequence, SDE. In chicken VtgR/VLDLR, this acidic region has been proposed to bind the basic residues on RAP because it has the highest negative charge density (25). However, according to the known structures of LBRs 1, 2, 5, and 6, many residues in these regions are involved in the formation of calcium cage and are most likely not accessible to the ligand. In our study, sitedirected mutagenesis in LBR3 affirmed that the change of EDE sequence to SDE did not affect the binding of Vtg. In addition, mutagenesis of SDE to EDE in LBR6 failed to gain the function of binding Vtg. This also indicates that EDE sequence alone is insufficient for Vtg binding. The structural determinants in different LBRs, viz., three disulfide bonds and calcium cageforming amino acids, are highly conserved. The backbone structures of those LBRs are expected to be similar. Thus the specificity of the LBRs to different ligands may be attributable to the nonconserved acidic groups, which are still available for interaction. The GST pull-down assay in the presence of dithiothreitol confirms the importance of disulfide bonds toward an appropriate architecture of LBRs for effective binding of Vtg (Fig. 6B). This clearly demonstrates that the primary sequence of LBD is not sufficient for binding Vtg. The three-dimensional structure of LBD must constitute the correct surface patch, which can recognize Vtg.
Both Vtg and apolipoprotein belong to the large lipid transfer protein, and they were found to be evolutionarily related not only in function but also in sequence (36). In our study, we found that the binding region, VtgSE, showed a pattern similar to the binding sites of apoB and apoE. The site-directed mutagenesis of the basic residues in VtgSE further confirmed that the positively charged residue, Lys 185 , plays a crucial role in receptor binding. Lys 185 is highly conserved in Vtgs. In the single deviant case in chicken Vtg, the lysine residue is also substituted with a basic residue, arginine (Fig. 7B). These observations strongly suggest that electrostatic interaction is involved. The existence of common basic residues in the receptor-binding region of Vtg, apoB and apoE on one hand, and the inaccessibility of the conserved acidic triplets in LBRs on the other, is apparently contradictory to expectations for electrostatic interactions. Nevertheless, it is still possible that electrostatic interactions between receptors and ligands exist via other negatively charged residues present within LBRs. This assumption is supported by reports that all the LBRs in LDLR have negatively charged surface patch (37).
The crystal structure of lamprey Vtg is already known (38,39). By alignment of the amino acid sequence, the receptorbinding region, VtgSE, was located to the LV1n part of lamprey Vtg. LV1n contains 4 ␣ helices and 12 ␤ strands, 11 of which form a barrel-like conformation. It is not clear whether this VtgSE receptor site is in the ␤ strands or the ␣ helices. Using the secondary structure prediction program, we predict that VtgSE is located in the ␣ helices (data not shown). In both apoB and apoE, the receptor-binding region is in the ␣ helices conformation, thus suggesting that they may utilize a similar mechanism for binding.
In nature, Vtg exists as a dimer containing symmetric binding sites (40). In our study, the receptor-binding site was found to be located in the VtgSE region. Comparable activities in the yeast two-hybrid assays observed with VtgSE and full-length Vtg indicate that the VtgSE region might be the only binding site in Vtg. The early study indicated that Vtg dimer and VtgR interact in a 1:1 stoichiometry (41). Therefore, to bind the Vtg dimer, VtgR must contain two Vtg-binding sites. This was supported by the results of our studies in deletion constructs of VtgR. LBRs 1-3 may contain more than one binding site for Vtg. Fig. 9 illustrates our proposed model of Vtg-VtgR interaction. Two molecules of Vtg dimerize through the dimerization domain in LV1 (39). The symmetric receptor binding sites in Vtg bind to two sites in the VtgR LBRs 1-3. The carboxylterminal of Vtg will form the lipid-binding cavity to transport lipid into the oocytes. Thus transportation of 2 Vtgs/VtgR molecule into the oocytes presents an efficient mechanism to meet FIG. 9. The model of Vtg-VtgR interaction. The LBRs are labeled 1-8. The EGFP-like repeats A, B, and C are annotated in VtgR. In a stoichiometric ratio of interaction between 2Vtgs:1VtgR, the Vtg dimer facilitates the binding of the LBRs 1-3 via the SE fragment (amino acids 162-248) in each amino-terminal of the Vtgs. There is more than one ligand-binding site within LBRs 1-3, and binding of VtgR to Vtg requires the combination of these three LBRs, which make electrostatic contacts with the VtgSE regions of the Vtg dimer. the temporal demands of oogenesis.
The receptor-ligand pairs of VtgR-Vtg and LDLR/VLDLRapolipoprotein have existed together in fish, amphibians, reptiles, and birds for millions of years (42,43). In the lower species, including insects and nematode, VtgR was the predominant form of receptor. The existence of VtgR in more ancient species such as Caenorhabditis elegans indicates that the LDLR and VLDLR might have evolved from VtgR by mutation and gene shuffling, and Vtg is the ancestor of apolipoprotein. It is accepted that apoB and Vtg share a common ancestor. However, the amino-terminal location of the receptor-binding site in Vtg is in contrast to apoB, which contains receptor-binding site in the carboxyl-terminal. This difference may arise from the structural change for better adaptation to the function. The mode of binding of LDLR/VLDLR to apolipoprotein is inherited from the electrostatic attraction of VtgR-Vtg. However, the sequence changes in apoB, especially in the region for lipid binding, probably facilitate the specific function of lipid transportation (39,44). Responsively, the mutation in LDLR/VLDLR in different LBRs accumulates for the improved specific binding to apoB, thus LDLR and VLDLR utilize different LBRs for apoB binding. This coevolution of receptor-ligand pairs probably creates the current functionally and structurally distinct receptors and ligands. The receptor-binding sites in many ligands of LDLR family members are still unknown. We predict that they also contain domains rich in basic residues for binding electrostatically with their cognate receptors.