DNA Binding and Transactivation Characteristics of the Mosquito Ecdysone Receptor-Ultraspiracle Complex*

The steroid hormone 20-hydroxyecdysone is a key regulatory factor, controlling blood-meal triggered egg maturation in mosquitoes. To elucidate the ecdysone hierarchy governing this event, we cloned and characterized the ecdysone receptor (AaEcR) and the nuclear receptorUltraspiracle (AaUSP), a retinoid X receptor homologue, from the mosquito, Aedes aegypti, which form a functional complex capable of ligand and DNA binding. Here we analyzed the DNA-binding properties of the AaEcR·AaUSP heterodimer with respect to the effects of nucleotide sequence, orientation, and spacing between half-sites in natural Drosophila and synthetic ecdysone response element (EcREs). By using an electrophoretic gel mobility shift assay, we showed that AaEcR·AaUSP exhibits a broad binding specificity, forming complexes with inverted (IR) and direct (DR) repeats of the nuclear receptor response element half-site consensus sequence AGGTCA separated by spacers of variable length. A single nucleotide spacer was optimal for both imperfect (IRhsp-1) and perfect (IRper-1) inverted repeats; adding or removing 1 base pair in an IRhsp-1 spacer practically abolished binding. However, changing the half-site to the consensus sequence AGGTCA (IRper-1) increased binding of AaEcR·AaUSP 10-fold over IRhsp-1 and, at the same time, reduced the stringency of the spacer length requirement, with IRper-0 to IRper-5 showing detectable binding. Spacer length was less important in DRs of AGGTCA (DR-0 to DR-5); although 4 bp was optimal, DR-3 and DR-5 bound AaEcR·AaUSP almost as efficiently as DR-4. Furthermore, AaEcR·AaUSP also bound DRs separated by 11–13 nucleotide spacers. Competition experiments and direct estimation of binding affinity (K d ) indicated that, given identical consensus half-sites and an optimal spacer, the AaEcR·AaUSP heterodimer bound an IR with higher affinity than a DR. Co-transfection assays utilizing CV-1 cells demonstrated that the mosquito EcR·USP heterodimer is capable of transactivating reporter constructs containing either IR-1 or DR-4. The levels of transactivation are correlated with the respective binding affinities of the response elements (IRper-1 > DR-4 > IRhsp-1). Taken together, these analyses predict broad variability in the EcREs of mosquito ecdysone-responsive genes.

ceptor in Drosophila is a heterodimer of the ecdysone receptor (EcR) protein (19) and an RXR homologue, Ultraspiracle (USP) (20 -22). The first ecdysone response element (EcRE) was identified in the promoter of the Drosophila heat shock protein-27 gene. It is an imperfect palindrome with only a 1-bp spacer (IR hsp -1), rather than the 3-bp spacer typical of vertebrate steroid HREs (23). Several EcREs have been identified in the regulatory regions of the following four more Drosophila genes: Eip28/29 (24), Fbp-1 (25), Sgs-4 (26), and Lsp-2 (27), each containing imperfect inverted repeats with a 1-bp spacer (IR-1). These findings, together with DNA binding and in vitro transactivation studies, suggested that natural Drosophila EcREs are predominantly IR-1s. However, it was later found that Drosophila EcR⅐USP (DmEcR⅐DmUSP) can bind synthetic DRs of (A/G)GGTCA with spacers of 3-5 nucleotides and can activate reporter gene constructs containing these direct repeats in Drosophila Schneider-2 (S2) cells (28,29). Finally, the EcRE of the Drosophila nested gene (ng) is a direct repeat of AGGTCA with a 12-nucleotide spacer (30).
The maintenance and dispersal of mosquito-borne disease depends upon successful reproduction of the mosquito, and 20E plays a crucial role in regulation of vitellogenesis and oogenesis (15,16,31,32). The processes of egg maturation and disease transmission are intimately associated through the mutual requirement for blood. Therefore, elucidation of the role of ecdysone receptor in mosquito reproduction is of significant biological and epidemiological importance. Although several target genes for the 20E-mediated regulatory cascade have been identified (32)(33)(34), native EcREs in mosquitoes are still unknown. In the mosquito Aedes aegypti, cDNAs of one ecdysone receptor (AaEcR) (35) and two USP isoforms (36) have been cloned. Compared with vertebrate nuclear receptors, insect EcR and USP homologues show unexpectedly high levels of sequence diversity (36,37). It is, therefore, difficult to predict the DNA binding specificity of the mosquito EcR⅐USP heterodimer. Although DNA binding domain determinants of halfsite sequence specificity have been identified (38 -40), determinants of half-site spacing and orientation as well as flanking sequence preferences are less well understood. In order to address these questions, we have analyzed the DNA binding properties of the AaEcR⅐AaUSP heterodimer. We used electrophoretic gel mobility shift assays (EMSA) with synthetic oligonucleotides and in vitro synthesized AaEcR and AaUSP to investigate the effects of the sequence, orientation, and spacing of half-sites on the DNA binding properties of the mosquito EcR⅐USP heterodimer. Finally, we have used CV-1 cells in order to correlate the DNA binding properties of AaEcR⅐AaUSP with their ability to transactivate the reporter gene constructs containing EcREs.

MATERIALS AND METHODS
In Vitro Synthesis of Nuclear Receptor Proteins-The nuclear receptor proteins were synthesized by coupled in vitro transcription-translation (TNT) system (Promega). AaEcR and AaUSPb cDNAs containing full open reading frames were subcloned into pGEM3Z (Promega) as described previously (36). For comparison, the 2.1-kilobase pair EcoRI fragment from pZ7-1-DmUSP (41) and the 3.3-kilobase pair BamHI fragment from pACT-DmEcR-B1 (19) bearing the entire open reading frames of Drosophila USP and EcR, respectively, were subcloned into pGEM7Z (ϩ) (Promega). The in vitro transcription/translation reactions programmed by the circular plasmid DNAs utilized the SP6 promoter. To confirm the synthesis of proteins with expected sizes, control TNT reactions were performed in the presence of [ 35 S]methionine, and the resulting reactions were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and autoradiography.
Oligonucleotides and Probes-Oligonucleotides were purchased either from the Macromolecular Structure Facility of the Biochemistry Department at Michigan State University or Life Technologies, Inc. For DNA binding studies, a pair of sense and antisense oligonucleotides was annealed and resolved by 15 or 20% non-denaturing PAGE, and the appropriate bands of double-stranded oligonucleotides were electroeluted. Ten picomoles of double-stranded oligonucleotide were endlabeled with T4 DNA kinase (Life Technologies, Inc.) and 50 Ci of [␥ 32 P]ATP (NEN Life Science Products), and the unincorporated radioactivity was removed through a Sephadex G-25 (Amersham Pharmacia Biotech) spin column.
Electrophoretic Gel Mobility Shift Assay (EMSA)-One microliter of each TNT reaction was used alone or in combination as a protein source for EMSA. Proteins were incubated for 30 min at room temperature in 20 l of the electrophoretic mobility shift buffer, containing 10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 1 mM MgCl 2 , 0.5 mM dithiothreitol, 0.5 mM EDTA, 4% (v/v) glycerol, 0.05 mg/ml of poly(dI-dC)-poly(dI-dC), 0.3 mg/ml single-stranded DNA (5Ј-CTAACAAAGTTCGCCTGGACTA-GAACGGCC-3Ј), 0.5 M 20E, and for competition experiments, the indicated amounts of unlabeled competitor oligonucleotides. This was followed by the addition of 0.05 pmol of 32 P-labeled probe and incubation for another 30 min. The reaction mixture was resolved using a 6% non-denaturing PAGE at a constant voltage of 150 V for 90 min at room temperature. The gel was dried, and the distribution of radioactivity was visualized either by autoradiography or by PhosphorImaging for quantitative analysis using ImageQuant TM software (Molecular Dynamics).
Equilibrium Dissociation Constant (K d ) Estimation-K d values of AaEcR⅐AaUSP binding to potential EcREs were estimated according to Fawell et al. (42) using the EMSA procedure described above. Protein samples were first incubated in the electrophoretic mobility shift buffer containing 0.5 M 20E for 30 min and then with several different concentrations of labeled double-stranded oligonucleotides for another 30 min. Bound and free probes were separated by non-denaturing PAGE and quantified by the PhosphorImager. Saturation curves and Scatchard plots (43) were calculated for at least three independent experiments, and the mean value was taken as the K d .
Antibodies-The antiserum raised against AaEcR (anti-AaEcR) was prepared as follows. The HincII-EcoRI fragment of the AaEcR cDNA clone was subcloned into pMAL-c2 (New England Biolabs) and expressed in Escherichia coli TB1 strain. The AaEcR protein, which was fused to maltose-binding protein, was concentrated by amylose resin according to the manufacturer's instructions. The fusion protein was further purified by SDS-PAGE followed by electroelution. A New Zealand White rabbit was immunized by subcutaneous injection with 100 g of the purified fusion protein emulsified in TiterMax (CytRx) adjuvant. Blood was collected at 2-week intervals, and the titer of specific antibodies was estimated by Western blotting of the TNT reaction programmed by pGEM3Z-AaEcR.
Reporter and Expression Plasmids and Cell Transfection Assays-The BamHI-EcoRI fragment of AaEcR cDNA (35) and the EcoRI fragments of AaUSPb cDNA (36) were subcloned into the corresponding sites of pCDNA3.1/zeo (Invitrogen). Translatability of these constructs was checked by in vitro TNT-coupled transcription/translation (Promega), which was followed by EMSA to verify binding properties of the expressed receptors. The reporter plasmid ⌬MTV-5xIR hsp -1-CAT (chloramphenicol acetyltransferase), containing five copies of IR hsp -1 was used in initial transfection assays (22). To make other reporter plasmids, oligonucleotides IR hsp -1(agcttcaaGGGTTCaTGCACTtgtccatcg), DR-4 (agcttcaagTGACCTcctgTGACCTtgtccatcg), and IR per -1 (agcttcaa-gAGGTCAaTGACCTtgtccatcg) were ligated into the HindIII site of ⌬MTV-CAT (45). Constructs harboring a single copy of either IR hsp -1, IR per -1, or DR-4 were used for a comparative study of transactivation by EcR⅐USP. Three copies of DR-4 were placed before the CAT gene to make the reporter construct ⌬MTV-3xDR4-CAT. All reporter constructs were confirmed by sequencing. The expression plasmid CMV-␤-galactosidase was a kind gift from Dr. L. Karl Olson (Department of Physiology, Michigan State University).
The green African monkey kidney CV-1 cell line (American Tissue Culture Collection, Bethesda, MD) was maintained in Dulbecco's modified Eagle's medium containing 10% calf serum. 2 ϫ 10 5 cells were seeded in 6-well plates for 18 -24 h before transfection. Transfection was performed using LipofectAMINE (Life Technologies, Inc.) according to the manufacturer's instruction. In brief, 0.4 g each of AaEcR, AaUSP, and CMV-␤-galactosidase expression plasmid, and 1.2 g of the reporter plasmid were mixed with LipofectAMINE and transfected in OPTI-MEM (Life Technologies, Inc.) for 3-5 h. The transfection mixture was removed, and the cells were further incubated in OPTI-MEM supplemented by 5% charcoal-stripped calf serum for 36 -48 h in the presence of ethanol vehicle or 1 M Muristerone A (Sigma). Each well received 2.4 g of total DNA. pCDNA3.1/Zeo (ϩ) was used as a carrier for equalizing the amount of DNA allocated to each well. CAT assays were performed as described by Herbomel et al. (46) for 2 h and ␤-galactosidase assays for 60 -90 min. CAT activity was normalized with ␤-galactosidase activity.

Binding of the AaEcR⅐AaUSP Heterodimer to Inverted Repeats: the Effect of Spacer Length and Half-site Nucleotide
Sequence-By using EMSA with in vitro TNT-expressed mosquito EcR and USP isoforms, we previously demonstrated that the AaEcR⅐AaUSP complex bound a 30-base pair oligonucleotide corresponding to the Drosophila hsp27 EcRE (36). This element was designated as IR hsp -1 (an imperfect inverted repeat with a 1-bp spacer). In this study, we report in detail the DNA binding characteristics of mosquito EcR heterodimerized with the mosquito USPb isoform (designated hereafter as USP). Our analyses indicated, however, that the binding properties of the AaEcR⅐AaUSPa heterodimer are generally similar to those of AaEcR⅐AaUSPb (not shown).
First, we confirmed binding of the AaEcR⅐AaUSP heterodimer to IR hsp -1 by utilizing anti-AaEcR and anti-DmUSP antibodies: when EMSA were performed in the presence of either of the antibodies, AaEcR⅐AaUSP⅐IR hsp -1 complexes were supershifted (not shown). Next, we investigated the role of IR spacer nucleotide length in the binding of the AaEcR⅐AaUSP heterodimer. In the first series of experiments, we tested the ability of IR hsp s with various spacer lengths to compete against IR hsp -1 binding to the AaEcR⅐AaUSP complex ( Fig. 1). A 50-fold molar excess of the appropriate cold IR was added to each EMSA reaction with radiolabeled IR hsp -1, and the intensity of the resulting bands was measured. Self-competition with cold IR hsp -1 led to a 96% reduction in binding intensity (Fig. 1, A  and B). IR hsp -0 also was revealed to be an efficient competitor, displacing 82% of the bound IR hsp -1 probe. However, the ability of IR hsp s to compete with IR hsp -1 progressively declined as the number of spacer nucleotides increased from 2 to 5. In a second series of experiments, the direct binding of IR hsp -0 -5 to AaEcR⅐AaUSP was tested by EMSA: apart from IR hsp -1, only IR hsp -0 exhibited a detectable retarded band (not shown).
Next, we tested AaEcR⅐AaUSP binding to a perfect inverted repeat consisting of AGGTCA half-sites with a 1-bp spacer (IR per -1). The EMSA revealed a strong retarded band when both TNT-generated AaEcR and AaUSP proteins were included in the reaction ( Fig. 2A, lane 1). Binding of AaEcR⅐AaUSP to IR per -1 was confirmed by supershift assays incorporating anti-AaEcR and anti-DmUSP ( Fig. 2A, lanes 2 and 3). Interestingly, a faster migrating weak band, which was supershifted by anti-DmUSP, was detected when AaUSP alone was tested with IR per -1 (not shown).
Perfect inverted repeats with nucleotide spaces from 0 to 5 (IR per -0 -5) had much higher affinity to AaEcR⅐AaUSP than the imperfect repeats motifs (IR hsp -0 -5); the retarded bands were clearly recognizable in all reactions (Fig. 2B). The retarded bands with IR per -0 and IR per -1 were considerably more intense than those formed by IR per -2 to IR per -5. Together with the results of IR hsp s, this indicates preferential binding of AaEcR⅐AaUSP to IR motifs with 1 and 0 nucleotide spacers (Fig. 2B). This preference was also confirmed by competition EMSA (Fig. 3). AaEcR⅐AaUSP was incubated with a fixed amount of radiolabeled IR per -1 and competed by increasing amounts of unlabeled IR per s of various spacers. In the selfcompetition control, binding was almost completely competed away by inclusion of a 25-fold molar excess of cold IR per -1. IR per -0 also competed well, with a 25-fold molar excess competing away 93% of IR per -1 binding to AaEcR⅐AaUSP. Twenty five-fold molar excess of IR per -2, IR per -3, and IR per -5 displaced only 56, 11, and 32% of binding, respectively. IR per -5 consist-ently showed stronger competition than IR per -3. Efficiency of competition indicates that the DNA binding affinity of AaEcR⅐AaUSP toward IR per s follows the order IR-1 Ͼ IR-0 Ͼ IR-2 Ͼ IR-5 Ͼ IR-3.
Binding of the AaEcR⅐AaUSP Heterodimer to Direct Repeats of AGGTCA-We also measured the binding of mosquito EcR⅐USP to a set of synthetic elements containing direct re-

FIG. 1. Effect of spacer length in the imperfect inverted repeats (IR hsp s) on binding with AaEcR⅐AaUSP. A, EMSAs of in vitro
synthesized AaEcR⅐AaUSP incubated with 0.05 pmol of 32 P-labeled IR hsp -1 in the absence (lane 1) or presence of a 50-fold molar excess of unlabeled IR hsp -1 (lane 2) or its variants with varying length of spacer nucleotides between the two half-sites (IR hsp -0, IR hsp -2, IR hsp -3, IR hsp -5, lanes 3-6, see below for each sequence). The position of the AaEcR⅐AaUSP⅐ 32 P-IR hsp -1 complex is indicated by an arrowhead and the free probe by an asterisk. B, the radioactivity associated with the protein-DNA complexes in A was quantified by PhosphorImaging as described under "Materials and Methods." Each bar represents percentage binding relative to the control reaction without unlabeled competitors appearing in lane 1 of A. Oligonucleotides used in this experiment (only one strand of each probe is shown) are as follows: IR hsp -0, agaga-caagGGTTCATGCACTtgtccaa; IR hsp -1, agagacaagGGTTCAaTGCACTtgtccaat; IR hsp -2, agagacaagGGTTCAatTGCACTtgtccaa; IR hsp -3, aga-gacaagGGTTCAaatTGCACTtgtccaa; IR hsp -5, agagacaagGGTTCAaata-aTGCACTtgtccaa.
peats with spacers ranging from 0 to 5. The mosquito EcR⅐USP complex effectively bound direct repeats (DRs) of AGGTCA containing a 4-bp spacer (DR-4) (Fig. 4). The composition of the AaEcR⅐AaUSP complex was verified by supershift experiments using either anti-AaEcR or anti-DmUSP. The latter super-shifted the AaEcR⅐AaUSP⅐DR-4 complex as efficiently as it did the control complex DmEcR⅐DmUSP⅐DR-4 (not shown).
We investigated the possible effect of DR-4 flanking regions on AaEcR⅐AaUSP binding by testing the following three DR-4 response elements: DR-4/3C with flanking regions from Drosophila ng elements (30), DR-4/HS with flanking regions from Drosophila hsp27 EcRE (23), and DR-4/VT with flanking regions from the thyroid response element (47). Radiolabeled IR hsp -1 was displaced from AaEcR⅐AaUSP equally efficiently by 50-fold molar excess of cold DR-4s containing any of the three flanking regions (Fig. 4, lanes 1-5). Labeled DR-4s with different flanking regions strongly bound AaEcR⅐AaUSP, forming specific retardation bands of similar size and intensity (Fig. 4, lanes 6 -11). Thus, the flanking DNA sequences did not much affect specific binding of the heterodimer. Also of note is that incubation of AaEcR⅐AaUSP with DR-4/3C resulted in an additional band of higher mobility than the specific heterodimer band (Fig. 4, lane 6). This high mobility band was competed by an excess of the cold specific probe (Fig. 4, lane 7) and was specifically supershifted with anti-DmUSP, but not anti-AaEcR antibodies (not shown), suggesting that it might represent the binding of AaUSP alone.
The role of the spacer nucleotides in DRs was also investigated. In EMSA competition experiments, the AaEcR⅐AaUSP complex was incubated with labeled DR-4 in the absence or presence of 5-or 25-fold molar excess of cold DRs containing nucleotide spacers of various lengths (DR-0 to DR-5). Twenty five-fold molar excess of cold DR-4 was sufficient to displace the bound probe almost completely (Fig. 5). DR-3 and DR-5 were almost as efficient as DR-4 itself. DR-1 and DR-2 seemed to be slightly less efficient, but still displaced around 90% of the labeled probe. DR-0 was the weakest competitor, with only 2/3 of the bound probe displaced by a 25-fold molar excess of this response element (Fig. 5). Indeed, in direct EMSA binding experiments, DR-0 was the only DR sequence that did not exhibit detectable binding to AaEcR⅐AaUSP (not shown). Thus, spacer length in DRs appeared to be less critical for binding to EMSAs were carried out with AaEcR⅐AaUSP and 0.05 pmol each of radiolabeled IR per s with 0 to 5 spacer nucleotides. Position of the shifted AaEcR⅐AaUSP⅐DNA complex is indicated by an arrowhead and of the free probe by an asterisk. Oligonucleotide probes used are as follows: IR per -0, agagacaagAGGTCATGACCTtgtccaa; IR per -1, agagacaagAGGT-CAaTGACCTtgtccaat; IR per -2,agagacaagAGGTCAatTGACCTtgtccaa; IR per -3, agagacaagAGGTCAaatTGACCTtgtccaa; IR per -5, agagacaagAG-GTCAaataaTGACCTtgtccaa.

FIG. 3. Effect of spacer length of the perfect inverted repeats (IR per s) on binding with AaEcR⅐AaUSP.
AaEcR⅐AaUSP was incubated with 0.05 pmol of 32 P-labeled IR per -1 in the absence of unlabeled competitor or in the presence of different molar excess of unlabeled competitor oligonucleotides IR per 0 -5 (see Fig. 2 for sequences). Reactions were subjected to EMSA, and the radioactivity in the specific protein-DNA complexes was counted by PhosphorImaging. The radioactivity associated with the DNA-protein complex observed without competition was taken as the control and defined as 100%. Data are reported as a percentage of the control. the mosquito EcR⅐USP complexes than in IRs. Nevertheless, efficiency of competition (Fig. 5) indicates that the DNA binding affinity of AaEcR⅐AaUSP toward DRs follows the order DR-4 Ͼ DR-3 Ͼ DR-5 Ͼ DR-2 Ͼ DR-1 Ͼ DR-0.
We were also interested in determining whether the mosquito EcR⅐USP complex might be capable of recognizing more widely spaced direct repeats. It has been reported that the Drosophila EcR⅐USP complex recognizes a DR-12 sequence found within the Drosophila ng gene (the original element was called DR-11 because 7 bp were taken as the consensus halfsite) (30). The mosquito EcR⅐USP was capable of binding to the ng EcRE (Fig. 6); however, our analysis of this element identified more closely spaced cryptic direct repeats within the ng EcRE. Within the ng EcRE, there are consensus half-sites at either end and an imperfect half-site (AGGCCA) in the middle, such that it could form a DR-2 or DR-4 in combination with one of the terminal elements. Whereas binding of DmEcR⅐DmUSP was abolished when both terminal consensus half-sites in the DR-12 were mutated simultaneously (30), this does not rule out the possibility that the DR-2 or DR-4 might be the active element or might be a functionally significant part of a compound element.
To investigate the nature of AaEcR⅐AaUSP binding to the ng EcRE, we performed mutational analyses of all three half-sites in this sequence, in which a mutation was introduced independently into the 5Ј-proximal (DR-12/P), middle (DR-12/M), or 3Ј-distal (DR-12/D) half-sites. First, we conducted a competition assay with 32 P-labeled IR hsp -1 and 5-or 25-fold molar excess of cold DR-12, DR-12/P, DR-12/M, or DR-12/D (Fig. 7). Twenty five-fold molar excess of cold DR-4, used as a positive control, removed 98% of radioactive probe binding (Fig. 7, lane   3), and 25-fold molar excess of the ng EcRE eliminated about 92% (Fig. 7, lane 5). Mutating the proximal half-site (DR-12/P) cripples the DR-12, leaving only the imperfect DR-4 intact. DR-12/P was much a weaker competitor (Fig. 7, lanes 6 and 7). In contrast, DR-12/M and DR-12/D, in which only the imperfect DR-2 was preserved, retained most of the binding ability of the original ng EcRE, with DR-12/M competing more strongly than DR-12/D (Fig. 7, lanes 8 -11). The overall order of relative affinity of tested elements was DR-4 Ͼ ng EcRE ϭ DR12/M Ͼ DR12/D Ͼ Ͼ DR12/P. Importantly, DR-12/M exhibited the same level of competition as the original DR-12, suggesting that the latter indeed serves as a response element with a 12-nucleotide spacer.
In order to determine whether other widely spaced direct repeats might also function as EcREs, we tested binding to DR-11 and DR-13 elements (Fig. 6). These studies show that the EcRE⅐USP has considerable flexibility in its spacing requirements. Although binding to DR-12 was strongest of the three, specific binding was also clearly demonstrated for the DR-11 and DR-13 elements.
In addition, we tested binding of the AaEcR⅐AaUSP complex to Eip28/29, a composite element from the promoter of the Drosophila Eip28/29 gene, containing an imperfect IR-1 and an imperfect DR-3 (24). EMSA analyses showed that AaEcR⅐AaUSP bound this composite motif as a heterodimer (data not shown).
Binding Affinity of the AaEcR⅐AaUSP Heterodimer: Effect of Sequence and Orientation of Half-sites in the Response Element-We compared the effect of half-site orientation and sequence on the DNA binding affinity of the AaEcR⅐AaUSP heterodimer. For these analyses, we utilized only IR hsp -1, IR per -1, and DR-4, because these elements exhibited maximal binding in the respective categories. First, we performed competitive EMSA, in which 32 P-labeled IR hsp -1 was competed with 2.5-, 5.0-, and 10.0-fold molar excess of cold IR hsp -1, IR per -1, DR-4, or Eip-28/29 (Fig. 8). The results suggest that the binding affinity of IR per -1 to AaEcR⅐AaUSP was stronger than that of DR-4 or IR hsp -1, which varied insignificantly from each other. The Eip28/29 appeared to have the weakest binding affinity to AaEcR⅐AaUSP. Finally, to resolve quantitatively the differences in DNA binding affinity, we calculated the equilibrium dissociation constants (K d ) for IR hsp -1, IR per -1, and DR-4 binding to AaEcR⅐AaUSP. The EcR⅐USP heterodimer was incubated with increasing concentrations of radiolabeled probes (IR hsp -1, IR per -1, or DR-4) in the presence of 5 ϫ 10 Ϫ7 M 20E, the optimal concentration for EcR⅐USP DNA binding. 2 Saturation binding analyses and Scatchard analyses were used to estimate K d values for the binding of AaEcR⅐AaUSP to IR hsp -1 (Fig. 9), IR per -1, and DR-4 (not shown). The differences in K d values for IR hsp -1, IR per -1, and DR-4, which are in agreement with the results of the competition analyses (Fig. 8), indicate that AaEcR⅐AaUSP binds IR per -1 with an 8-fold higher affinity than to DR-4 and with 10-fold higher affinity than to IR hsp -1 ( Table I).
Transactivation of AaEcR⅐AaUSP: DNA Binding Affinity Corresponds to Transactivation Activity-Transactivation of AaEcR⅐AaUSP was studied using the CV-1 cell line. This mammalian cell line has no endogenous EcR and contains very low endogenous levels of RXR. It has been used to study transactivation of DmEcR⅐DmUSP (20 -22). The transactivation ability of the AaEcR⅐AaUSP heterodimer was assessed with the ⌬MTV-5xIR hsp -1-CAT reporter plasmid, which contains five tandem repeats of IR hsp -1. Transfection of CV-1 cells with the reporter plasmid alone resulted in a very low basal level of CAT activity (Fig. 10A). Co-transfection of the reporter plasmid with either AaEcR or AaUSP expression vector alone did not confer ecdysone responsiveness. However, strong induction (40-fold) of CAT activity was observed when the reporter plasmid was co-transfected with both AaEcR and AaUSP expression vectors and incubated with 1 M MurA, demonstrating that the AaEcR⅐AaUSP heterodimer activated reporter gene expression in a ligand-dependent manner.
Next, we tested whether DR-4 could function as an EcRE in CV-1 cells. We constructed a reporter plasmid (⌬MTV-3xDR-4-CAT) containing three copies of DR4. Co-transfection of this reporter construct with either the AaEcR or the AaUSP expression vector did not render CV-1 cells ecdysone-responsive (data not shown). However, co-transfecting the reporter construct with both AaEcR and AaUSP expression vectors rendered CV-1 cells highly responsive to ligand with 9-fold induction (Fig.  10B).
Finally, we elucidated whether the level of transactivation by AaEcR⅐AaUSP depends on the sequence and orientation of the half-sites in the EcRE and whether it is correlated with DNA binding affinity. We constructed ⌬MTV-CAT reporter plasmids containing a single copy of IR hsp -1, DR-4, or IR per -1. Co-transfection of these reporter plasmids with either AaEcR or AaUSP expression vector did not render CV-1 cells ecdysone-2 S. Wang and A. Raikhel, unpublished observations. Reactions were subjected to EMSA, and the radioactivity in the specific protein-DNA complexes was counted by PhosphorImaging. The radioactivity associated with protein-DNA complex observed without competition was taken as the control and defined as 100%. Data are reported as a percentage of the control.
responsive similar to results with the ⌬MTV-5xIR hsp -1-CAT reporter plasmid (data not shown). Co-transfection of both AaEcR and AaUSP with each of these reporter plasmids resulted in an increase in CAT activity in the presence of 1 M MurA (Fig. 11). The level of transactivation of the reporter plasmid containing only one copy of the EcRE is considerably lower than that with five copies or three copies (Fig. 10). The differences in levels of activation between tested EcREs were of lower magnitude than the differences between their binding affinities. However, the strength of binding directly corresponds to the level of transactivation for each class of EcRE, with IR per -1 Ͼ DR-4 Ͼ IR hsp -1 (Table I). A nonparametric STP (simultaneous test procedure) test (48) indicated that the differences in transactivation between the different elements are statistically significant (␣ ϭ 0.01).

DISCUSSION
In this paper, we provide further evidence that the AaEcR⅐AaUSP heterodimer is the functional mosquito ecdysone receptor and that it is capable of binding various DNA motifs oriented either as inverted or direct repeats. Data presented here parallels previous observations from several insect species that heterodimerization of EcR and USP is required for efficient binding of both the ligand and the response elements, as well as for gene transactivation (20 -22, 36, 49, 50). Analyses utilizing EMSA and anti-EcR and anti-USP antibodies clearly demonstrated that the AaEcR⅐AaUSP heterodimer exhibits specific binding to the various sequences of naturally occurring Drosophila EcREs as well as to the synthetic response elements tested in this study. This and a previous study (36) demonstrate that the AaEcR⅐AaUSP heterodimer is capable of binding to various DNA motifs with the consensus half-site sequence AGGTCA oriented as inverted repeats. Indeed, EcREs found in native Drosophila genes are predominantly through the indicated range of DNA probe concentrations. Radioactivity associated with free oligonucleotides and with protein-oligonucleotide complexes was determined separately, permitting the construction of a saturation curve (B) and a Scatchard plot (C). EMSA and quantification were repeated at least three times.   values) of IR hsp -1, IR per -1, and DR-4 to the AaEcR⅐AaUSP complex were measured by EMSA. In vitro translated AaEcR⅐AaUSP was incubated with increasing amounts of radiolabeled elements (0.4 -20 nM) and resolved by electrophoresis. Radioactivity associated with free oligonucleotides and with protein-oligonucleotide complexes was quantitated by phosphorImage analysis, permitting the construction of a saturation curve and a Scatchard plot (Fig. 9). EMSAs and quantifications were repeated at least three times, and the mean was taken as the K d value. For transfection assays, 0.4 g each of CMV-␤-gal actosidase, AaEcR, and AaUSP expression vectors were transiently co-transfected into CV-1 cells with 1.2 g of ⌬MTV-CAT reporter plasmid harboring one copy of IR hsp -1, DR-4 and IR per -1 (Fig.  11). After transfection, cells were incubated in the presence of vehicle ethanol or 1 M MurA for 36 or 48 h and harvested for CAT and ␤-gal actosidase assays. Fold induction was calculated from normalized CAT activity from two triplicate experiments. Standard errors (S.E.) were calculated by Microsoft Excel ™ . inverted imperfect palindromes; the binding affinities of Sgs-4, Lsp-2, Fbp-D, and Eip28/29 EcREs are weaker than the hsp27 EcRE (IR hsp -1), which is the most efficient natural EcRE identified for Drosophila to date (23,24,26,27,51). We obtained similar results when testing mosquito ecdysone receptor binding to the Eip28/29 and hsp27 EcREs (Fig. 8). Here, we have shown that the perfect palindrome IR per -1 binds 10 times more efficiently than IR hsp -1.
Our results suggest that spacer length likewise plays an important role in both imperfect (IR hsp ) and perfect (IR per ) inverted repeats, with a single nucleotide spacer being optimal for both. This finding is in agreement with conclusions drawn from studies performed with DmEcR⅐DmUSP (23,24,51). Moreover, we also found that whereas adding or removing one base pair from a spacer in IR hsp -1 practically abolishes binding, changing the half-site to the consensus sequence AGGTCA (IR per -1) reduces the stringency of the spacer length requirement, so that IR per -0 to IR per -5 exhibits detectable binding.
It has been demonstrated that Drosophila EcR⅐USP binds to direct repeats and that a 4-bp spacer is optimal (28,29). Our observations on the mosquito EcR⅐USP heterodimer suggest that this aspect of EcR⅐USP DNA binding specificity also displays a high degree of functional conservation. Direct binding and competition assays demonstrate that the nucleotide spacer length is less important in direct repeats of AGGTCA (DR-0 to DR-5) than in IRs. Although 4 bp is an optimal spacer length in the direct repeats, DR-3 and DR-5 bind AaEcR⅐AaUSP almost as efficiently as DR-4. The order of binding affinities of AaEcR⅐AaUSP to DRs (DR-4 Ͼ DR-3 Ͼ DR-5 Ͼ DR-2 Ͼ DR-1 Ͼ DR-0) corresponds closely to that recently reported for DmEcR⅐DmUSP (DR-4 Ͼ DR-5 Ͼ DR-3 Ͼ DR-1 Ͼ DR-2 Ͼ DR-0) (29).
Competition experiments and direct estimations of K d indicate that binding affinity depends on the sequence of the halfsite and is higher when the consensus is used for each half-site. However, given the same consensus half-site and an optimal spacer, the AaEcR⅐AaUSP heterodimer binds an inverted repeat with considerably more strength than a direct repeat. Our results significantly extend DNA binding studies on insect EcR⅐USP heterodimers by providing accurate measurements of dissociation constants for DR-4, IR hsp -1, and IR per -1.
Kato et al. (52) showed that in the chicken ovalbumin promoter region, there are multiple AGGTCA motifs arranged as direct repeats separated from each other by more than 100 bp. Despite such large spacers, they can act synergistically as a complex estrogen response element, indicating that widely spaced half-sites can cooperate to generate an efficient estrogen response element. Moreover, widely spaced direct repeats (10 -200 bp) can function as cis-acting response elements for retinoic acid and vitamin D receptors (10). In contrast to the specificity observed with shortly spaced DRs (DR-1 to DR-5), different receptors bind promiscuously to these widely spaced repeats to activate transcription in the presence of retinoic acid, vitamin D, or estrogen. Our tests have shown that although the AaEcR⅐AaUSP heterodimer exhibits relatively strong binding to DR-12, it is considerably weaker than to DR-3 or DR-4. Furthermore, AaEcR⅐AaUSP binds to other direct repeats separated by more than 10 nucleotides (DR-11 and DR-13) but with considerably lower affinity than to DR-12. Presently, it is not known whether insect EcR⅐USP heterodimers are capable of utilizing widely spaced half-sites (e.g. Ͼ100 bp) as response elements.
We observed that the ng EcRE, previously described as a DR-12, is a composite of three half-sites, suggesting the possibility that an internal 5Ј-proximal DR-2 and/or 3Ј-distal DR-4 might contribute to the functionality of this element. Mutating the 5Ј half-site in this element dramatically reduced its binding affinity, revealing its critical role in EcR⅐USP binding. In contrast, mutating the 3Ј half-site decreased binding affinity only slightly, whereas mutating the middle half-site did not have any obvious effect on the binding of the element. Thus, both competition and direct binding analyses of mutated ng EcRE suggest that in addition to functioning as a true DR-12, the ng element may also have a functional imperfect DR-2 located at its 5Ј end (Fig. 7).
Transactivation assays in CV-1 cells confirmed the finding of the DNA binding assays, demonstrating that the AaEcR⅐ AaUSP heterodimer is indeed the functional ecdysone receptor. Co-transfection of AaEcR and AaUSP expression vectors into CV-1 cells conferred 40-fold induction of the reporter plasmid ⌬MTV-5xIR hsp -1-CAT and 9-fold induction of ⌬MTV-3xDR4-CAT in response to 1 M MurA. We observed that the number of EcREs in a reporter construct is not directly proportional to the magnitude of reporter transactivation. A reporter plasmid containing a single copy of IR hsp -1 was induced only 2.5-fold compared with a 40-fold induction of the reporter containing five copies of the same EcRE. Therefore, in order to compare transactivation activities of IR and DR elements, we utilized the reporter plasmids containing only one copy of either IR hsp -1, DR-4, or IR per -1. Importantly, all three response elements were able to mediate ecdysone responsiveness of the reporter in CV-1 cells. The transactivation efficiencies of tested response element followed the order IR per -1 Ͼ DR-4 Ͼ IR hsp -1. Thus, despite the fact that the differences were not as dramatic as those for binding affinities measured for the same elements, the two sets of data are in agreement with one another (Table  I). By using transfection in Drosophila Schneider-3 cells and endogenous receptor pools, Martinez et al. (53) also showed that for DmEcR⅐DmUSP, IR per -1 was transactivated about twice as well as IR hsp -1 when placed before the thymidine kinase promoter. By using four copies of EcREs ahead of the hsp70 promoter, Vogtli et al. (54) reported that IR per -1 activated the reporter gene twice stronger than IR hsp -1 and DR-4 in the presence of endogenous DmEcR⅐DmUSP in S2 cells. These results agree with our observations for CV-1 cells.
DR-4, the optimal DR for binding to AaEcR⅐AaUSP (Fig. 5) and DmEcR⅐DmUSP (28,29), so far has not been identified as a natural EcRE in ecdysone-responsive genes in any organism. The functionality of DR-4 as EcRE is controversial in the literature. Antoniewski et al. (29) showed that various DRs, including DR-4, could act as functional EcREs for Drosophila EcR⅐USP in S2 cells. More recently, these results have been confirmed by Vogtli et al. (54). However, DR-4 failed to render Drosophila Kc cells ecdysone-responsive (37). Our data demonstrate that DR-4 also can act as a functional EcRE for AaEcR⅐AaUSP in mammalian CV-1 cells. Taken together, these observations suggest that direct repeats may serve as cell-specific EcREs. By placing two copies of either IR hsp -1 or DR-4 upstream of the thymidine kinase promoter for transfection assays in S2 cells, Antoniewski et al. (29) demonstrated that IR hsp -1 was a more potent EcRE than DR-4 for transactivation by exogenous DmEcR⅐DmUSP. In contrast, the situation was the reverse in our experiments in CV-1 cells. These differences could be due to the number of EcREs in the reporter constructs, to the type of cells used for the transactivation experiments, or they could also reflect true differences in the transactivation properties of Drosophila and mosquito EcR⅐USP heterodimers.
Taken together, our findings suggest that both IR and DR response elements can act as functional EcREs in the activation of mosquito genes. Our data predict wide variability among natural EcREs in mosquito genes. Moreover, because the level of gene transactivation by ecdysone depends on an EcRE sequence, differences between the various EcREs may be utilized as one of the mechanisms regulating the levels of ecdysone responsiveness. Recently, support for this hypothesis has been provided by discovery of EcREs in two mosquito yolk protein precursor genes, vitellogenin (Vg) and vitellogenic carboxypeptidase. Regulatory regions of both Vg and vitellogenic carboxypeptidase genes, expression of which is controlled by 20E (55,56), contain imperfect DR-1 and DR-2 elements, respectively. Several lines of analysis strongly suggest that both Vg DR-1 and vitellogenic carboxypeptidase DR-2 are functional EcREs. First, both of them specifically bind AaEcR⅐AaUSP in EMSA utilizing in vitro expressed receptors, as well as nuclear extracts from vitellogenic mosquito fat bodies. Second, a portion of the vitellogenic carboxypeptidase promoter containing the DR-2 EcRE confers ecdysone responsiveness in CV-1 cells. Finally, a reporter gene containing the regulatory region of either the Vg or vitellogenic carboxypeptidase gene with the respective EcRE was expressed at the correct stage when transformed into Drosophila. 3 Thus, the present study provides a solid foundation to search for native EcREs in mosquito genes. It will aid in the analysis of the regulatory mechanisms governing gene expression during the blood-meal activated events of reproduction and pathogen transmission in this critically important insect vector for both humans and animal diseases.