The Drosophila Juvenile Hormone Receptor Candidates Methoprene-tolerant (MET) and Germ Cell-expressed (GCE) Utilize a Conserved LIXXL Motif to Bind the FTZ-F1 Nuclear Receptor*

Background: JH signaling involves interactions between FTZ-F1 and candidate JH receptors MET and GCE. Results: Mutation of FTZ-F1 AF2 or LIXXL MET/GCE sequence disrupts interaction between the proteins. Conclusion: NR box-AF2 binding underlies FTZ-F1·MET and FTZ-F1·GCE heterodimer formation. Significance: Dissecting the interaction between FTZ-F1 and MET, GCE is critical to understanding the molecular basis of JH signaling. Juvenile hormone (JH) has been implicated in many developmental processes in holometabolous insects, but its mechanism of signaling remains controversial. We previously found that in Drosophila Schneider 2 cells, the nuclear receptor FTZ-F1 is required for activation of the E75A gene by JH. Here, we utilized insect two-hybrid assays to show that FTZ-F1 interacts with two JH receptor candidates, the bHLH-PAS paralogs MET and GCE, in a JH-dependent manner. These interactions are severely reduced when helix 12 of the FTZ-F1 activation function 2 (AF2) is removed, implicating AF2 as an interacting site. Through homology modeling, we found that MET and GCE possess a C-terminal α-helix featuring a conserved motif LIXXL that represents a novel nuclear receptor (NR) box. Docking simulations supported by two-hybrid experiments revealed that FTZ-F1·MET and FTZ-F1·GCE heterodimer formation involves a typical NR box-AF2 interaction but does not require the canonical charge clamp residues of FTZ-F1 and relies primarily on hydrophobic contacts, including a unique interaction with helix 4. Moreover, we identified paralog-specific features, including a secondary interaction site found only in MET. Our findings suggest that a novel NR box enables MET and GCE to interact JH-dependently with the AF2 of FTZ-F1.

larvae of holometabolous insects, high JH titer forces ecdysone to elicit molting, whereas a drop in JH during the final larval instar allows ecdysone to trigger entry into prepupal development (1). The ability of JH to delay metamorphosis has been observed directly in insects, such as the silkworm Bombyx mori, where removal of endogenous JH by ectopic JH esterase results in premature pupariation (2) and in the red flour beetle Tribolium castaneum, where exogenous JH prevents metamorphosis through the production of additional larval instars (3,4).
There is accumulating evidence that the antimetamorphic function of JH involves MET (methoprene-tolerant), a basichelix-loop-helix Per-AhR/Arnt-Sim (bHLH-PAS) protein and candidate JH receptor with homologs identified in all holometabola whose genomes have been sequenced. The Tribolium and Drosophila homologs of MET bind JH with nanomolar affinity when produced in vitro (5,6). In Tribolium MET is required to prevent premature pupariation in larvae (3,4) and acts upstream of the antimetamorphic gene Kr-h1, a target of JH activation (7). In Drosophila, the role of MET as a mediator of JH action is less clear. It is required in vivo for JH-dependent development of the adult eyes and CNS optic lobes (8) and for the lethal "methoprene syndrome" response to ectopic hormone (9). However, despite the fact that JH deficiency is lethal, Met null mutant flies are viable. In contrast to insects that depend on MET for survival, drosophilids possess a Met paralog encoded by the gce (germ cell-expressed) gene (10). GCE is also a bHLH-PAS protein and is capable of binding JH in vitro (5). Like Met, gce null mutants are viable and insensitive to ectopic JH (11). Although neither mutation is lethal, mutants missing both paralogs die as prepupae, suggesting that MET and GCE have some redundant functions in vivo. However, the paralogs are not completely redundant because their ability to substitute for one another in JH-dependent processes is tissue-specific (9). Recently, we found that JH activation of the nuclear receptor gene E75A in S2 cells requires GCE but not MET, indicating that GCE possesses paralog-specific regulatory functions (12). Despite the emerging evidence that both MET and GCE are involved in JH signaling, it is still not clear whether MET, GCE, or both proteins act as bona fide JH receptors in flies.
The structure of MET and GCE conforms with their proposed role in regulating transcription. The bHLH is a DNAbinding domain found in many transcription factors (13). PAS domains often bind ligands and can function as signal sensors for light, oxygen, redox potential, metabolites, and xenobiotic compounds (14). They can also mediate protein-protein interactions. For example, mosquito MET utilizes its PAS domains to interact with the bHLH-PAS partner FISC; together they bind a JH-responsive sequence and activate transcription of the early trypsin gene (15). The location of the bHLH and PAS domains in MET and GCE is typical of the bHLH-PAS family; the bHLH is at the N terminus and is followed by two PAS motifs, PAS A and PAS B, and a single PAS-associated C-terminal (PAC) motif. PAS motifs are a characteristic feature of PAS domains, and in MET and GCE they share considerable sequence identity with other PAS proteins (16,17). A complete PAS domain cannot always be identified by primary structure, however, because the corresponding sequences often have a limited degree of identity outside of the PAS motif (18). This appears to be the case for MET and GCE, where the tertiary structure of the PAS domains has not yet been determined. Another common feature of bHLH-PAS proteins is the presence of acidic, proline/serine (P/S), and glutamine-rich (Q R ) sequences in the C terminus that can serve as transactivation domains (19). However, the C-terminal regions of MET and GCE have not yet been functionally characterized.
Recently, we found that in addition to GCE, JH activation of E75A requires the orphan nuclear receptor FTZ-F1 (12). The ftz-f1 gene encodes two isoforms, ␣ftz-f1 and ␤ftz-f1, whose products share an identical DNA-binding (DBD) and ligandbinding (LBD) domains but have distinct N-terminal domains resulting from alternative promoter usage (Fig. 1A). The LBD of FTZ-F1 has a typical structure consisting of 12 ␣-helices and an anti-parallel ␤-sheet; it forms a coactivator interaction surface referred to as activation function 2 (AF2) from helices H3, H3Ј, H4, and H12 (20,21). In contrast to ligand-dependent AF2, the AF2 of FTZ-F1 is stabilized in an active conformation by H6, which serves as a pseudoligand and may allow FTZ-F1 to function as a ligand-independent activator (21). The presence of a constitutively active AF2 suggests that FTZ-F1 could act primarily through protein-protein interactions, and our finding that FTZ-F1 interacts with both MET and GCE (12) raises the possibility that FTZ-F1 mediates JH action indirectly through the formation of FTZ-F1⅐MET or FTZ-F1⅐GCE heterodimers. Previous examples of interactions between bHLH-PAS proteins and nuclear receptors have come from the steroid receptor coactivator (SRC) family of proteins. A centrally located domain in these proteins contains three copies of a motif, LXXLL, termed the nuclear receptor (NR) box, which adopts an ␣-helical conformation and binds to a hydrophobic groove formed by the AF2 (20,22). For the SRC coactivators, each NR box is critical to interaction with a different subset of nuclear receptors. The N-terminal bHLH and PAS domains of SRC proteins are dispensable to their interaction with nuclear receptors (23)(24)(25), whereas a Q R region in the C terminus is used as a second interaction site through a poorly understood mechanism (26 -29). In addition to the SRC family, NR boxes have been identified in a variety of other coregulators, including the homeodomain protein FTZ, which utilizes an NR box to interact with the AF2 of FTZ-F1 (21,30,31). No functional NR boxes have yet been identified in MET or GCE, so the mechanism of their interaction with FTZ-F1 remains unresolved.
Here, we define the secondary and tertiary structure of the MET and GCE proteins and explore the molecular basis of the FTZ-F1⅐MET and FTZ-F1⅐GCE interactions. We demonstrate that both MET and GCE interact JH-dependently with FTZ-F1 AF2, utilizing a novel NR box with a unique and paralog-distinct mechanism.

EXPERIMENTAL PROCEDURES
Cell Culture, Cell Transfection, and RNAi-Drosophila S2 cells were cultured in Schneider's medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS) at 25°C. JH III (a natural JH compound; Sigma) and methoprene (a synthetic JH analog) were dissolved in ethanol, and 1 ϫ 10 Ϫ1 M stock solutions were kept at Ϫ20°C. Hormones were added to a 1 ϫ 10 Ϫ6 M final concentration. Control cells were treated with an equal volume of ethanol. Transient transfection of expression plasmids and RNAi treatment of S2 cells were performed as described previously (12,32).
Plasmids and Mutagenesis-The ORFs for ␤FTZ-F1 and comFTZ-F1 (residues 175-803, ␤-isoform numbering) were amplified by PCR using ␤ftz-f1 cDNA acquired from the Drosophila Genomics Resource Center (EST clone RE02257) and cloned into the pMt/V5-His plasmid (Invitrogen) using primers with flanking restriction sites. For expression of ␣FTZ-F1, the ␣-specific coding region was amplified from S2 genomic DNA and cloned into the comFTZ-F1 expression plasmid, producing ␣FTZ-F1 ORF with a two-residue linker. Protein expression was confirmed by Western blot with anti-V5 antibodies (1:10,000 dilution). To produce double-stranded RNA (dsRNA) for RNAi knockdown of ␣ftz-f1, a DNA fragment containing ␣-specific protein coding sequences was amplified by PCR and cloned in both orientations into the pGEM-T Easy plasmid (Promega) using the following primers: 5Ј-TGC TCT AGA  ATG ACA CTA ATG GGC ACT GC-3Ј and 5Ј-TGC TCT  AGA TGC TAG TGT GGT TGC TGT TG-3Ј. Plasmids for the insect two-hybrid experiment (pIE2-(p65)AD, pIE2-(GAL4)DBD, and pUAS-Luc) were a gift from Dr. T. Kusakabe (Kyushu University). MET-DBD, GCE-DBD, and FTZ-F1-AD fusion constructs were generated using the pENTR3C plasmid and LR Clonase II kit (Invitrogen). The constitutive expression plasmid pAc5.1-LacZ-V5 (Invitrogen) was used in two-hybrid experiments to measure transfection efficiency. Point mutations were introduced using the QuikChange II site-directed mutagenesis kit (Stratagene) according to the manufacturer's instructions. All expression plasmids were sequenced (Genewiz) to confirm the ORFs. Primer sequences are available upon request.
RNA Extraction, Northern Hybridization, and Quantitative RT-PCR-Extraction of RNA, Northern blot hybridization, and quantitative RT-PCR were performed as described previously (12). Primer sequences are available upon request.
Luciferase and ␤-Galactosidase Reporter Assays-Assays were performed as described (12). Briefly, standardized protein extracts were used to measure luciferase using the Luciferase Assay System (Promega), and relative light units were averaged over 10 s for each sample. ␤-Galatosidase activity was measured with the ␤-Galactosidase Assay System (Promega), with absorbance at A 420 recorded every 60 s for 10 min. Luciferase activity was normalized to ␤-galactosidase activity for each sample.
Homology Modeling-Models of MET and GCE tertiary structure were built using Phyre2 (33). PSI-BLAST alignments of MET and GCE amino acid sequences were used to identify remote homologs; each identified 1000 high confidence (E Ͻ 0.001) matches primarily within the region containing the PAS motifs. The sequence identity to remote homologs was limited (ranging from 7 to 14%), indicative of highly informative alignments (33). Secondary structure was predicted using three independent algorithms (Psi-Pred, SSPro, and JNet), and disordered regions were predicted using Disopred. The structural predictions and PSI-BLAST profile were used to search empirical structures from the Structural Classification of Proteins database and the Protein Data Bank (PDB). Among the top scoring matches, nearly all of them were either helix-loop-helix or PAS folds. Models for MET and GCE were built using murine MyoD for the bHLH domains and PER (Drosophila Period) for the PAS domains.
To assess the quality of Phyre2 models, Z-scores were generated using QMEAN, which compares models to a reference set of empirical structures whose average Z-score is zero (34). Z-Scores for the bHLH domains of MET (Ϫ0.41) and GCE (Ϫ0.25) indicate models that are comparable with experimentally derived structures, whereas Z-scores for the PAS domains of MET (Ϫ1.91) and GCE (Ϫ2.85) indicate medium quality models. Local (i.e. per residue) QMEAN scoring of the PAS domain models found poor quality modeling in the regions directly adjacent to PAS loops and in the region linking the two PAS folds, whereas the ␤-strands and ␣-helices of the PAS folds are of higher quality. Final model illustrations were produced using PyMOL.
Identification of Conserved MET and GCE Residues-Multiple sequence alignment of MET and GCE homologs was performed using ClustalW (35). After alignment, the percentage of homologous (i.e. identical or similar) residues among all proteins was calculated for each conserved sequence block identified in the alignment. The complete set of conserved sequence blocks identified among distant MET and GCE homologs is shown in supplemental Fig. S2; for alignment of C-terminal MET and GCE sequences, see supplemental Fig. S3.
Simulated Docking of Protein Complexes-Docking models of FTZ-F1⅐MET and FTZ-F1⅐GCE complexes were generated with HADDOCK (36). The three-dimensional structure of the FTZ-F1 LBD is publicly available (PDB entry 2XHS). 9-residue NR box peptide structures for MET (LYLIENLQK, residues 552-560) and GCE (LRLIQNLQK, residues 416 -424) were generated de novo by manually mutating the side chains of the 9-residue FTZ NR box peptide from the FTZ-F1 crystal structure. Parallel docking was performed using the FTZ NR box (STLRALLTN, residues 107-115), whose binding to FTZ-F1 has been determined empirically (21). Docking by HADDOCK is driven by predictions of likely residues involved in complex formation (ambiguous interaction restraints (AIRs)); these may be active (interacting residue) or passive (solvent-accessible neighbor of interacting residue). AIRs for FTZ-F1 were generated using CPORT (37), and AIRs for MET and GCE were chosen based on experimental data (supplemental Table S1). For each docking experiment, 10,000 rigid body simulations were performed, followed by semiflexible simulated annealing, water refinement, and clustering of the top 400 simulations based on structural similarity; clusters were then assigned energy scores. The largest cluster of docking solutions in each experiment had the most favorable (i.e. lowest) energy score, and the scores were similar when either MET, GCE, or FTZ peptides were docked (supplemental Table S2). As expected, the structure of the docked FTZ peptide was similar to that of FTZ in the crystal structure (root mean square deviation of peptide from the lowest energy structure versus the crystal structure was 1.00 Å). Following docking simulations by HADDOCK, the lowest energy structures were refined using Rosetta FlexPepDock, which improves the accuracy of docking by allowing for flexibility in the backbone of short peptides (38). MolProbity (39) was used to identify hydrogen bonds and non-bonded contacts. Final model illustrations were produced in PyMOL.

Both Isoforms of FTZ-F1 Can Mediate JH Activation-ftz-f1
encodes two transcript isoforms, ␣ftz-f1 and ␤ftz-f1, which are differentially expressed during development (40,41). To determine which of these transcripts is present in S2 cells, probes targeting the isoform-specific 5Ј and common 3Ј regions ( Fig.  1A) were used to measure ftz-f1 gene expression by Northern blot hybridization. RNA samples were collected from S2 cells before and after treatment with ecdysone and from prepupal stage larvae. The ␣-probe detected specific transcripts in S2 cells (5.4 and 6.2 kb), whose expression was repressible by ecdysone, whereas the ␤-probe detected transcripts (4.8 and 5.6 kb) only in prepupal larvae (Fig. 1B). The common probe facilitated direct comparison of transcripts from S2 cells and prepupae; the longer transcripts found in S2 cells compared with prepupae are consistent with the longer 5Ј region of ␣ftz-f1. These results demonstrate that ␣ftz-f1 is the predominant transcript in S2 cells, indicating that ␣FTZ-F1 protein is responsible for mediating JH activation of E75A in this cell line.
To determine whether the function of FTZ-F1 in JH signaling is limited to ␣FTZ-F1, we studied JH inducibility of E75A in cells that ectopically express high levels of ␣FTZ-F1, ␤FTZ-F1, or an N-terminally truncated FTZ-F1 containing only the domains common to both isoforms (comFTZ-F1). Cells were transfected with empty plasmid or plasmid expressing one of the FTZ-F1 proteins and then treated with solvent or the JH analog methoprene for 1 h. In each case, the amount of protein expressed was comparable ( Fig. 2A). In the absence of ectopic protein, E75A is induced ϳ3-fold after 1 h of JH treatment (Fig.  2B). When either ␣FTZ-F1 or ␤FTZ-F1 is present, E75A transcription is modestly but significantly (p Ͻ 0.05) enhanced both in the absence and presence of JH, resulting in higher overall expression but a similar 3-fold level of JH induction (Fig. 2B). Removal of the isoform-specific region of FTZ-F1 that typically harbors ligand-independent activation function 1 (AF1) did not affect E75A transcription in the absence of JH but modestly enhanced transcription when hormone was present, resulting in 4-fold JH induction of E75A. These results indicate that regardless of whether the hormone is present or not, both ␣FTZ-F1 and ␤FTZ-F1 potentiate E75A transcription, but the domains shared between both isoforms are sufficient to potentiate JH-dependent induction.
To determine conclusively whether both isoforms are functionally equivalent in mediating JH signaling, we carried out an RNAi rescue experiment by utilizing the distinct 5Ј coding region of the ␣ftz-f1 transcript. First, ␣ftz-f1 expression was knocked down in S2 cells using dsRNA targeting the 5Ј region specific to the ␣ isoform. Cells were then transfected with empty plasmid or the ␤ftz-f1-encoding plasmid and treated for 1 h with methoprene. RNAi treatment reduced endogenous ␣ftz-f1 expression by about 70% (data not shown) and resulted in a strong and significant (p Ͻ 0.05) reduction of E75A JH-dependent transcription (Fig. 2C); this result confirms that the ␣ isoform is responsible for JH activation in this cell line. When ␤FTZ-F1 is expressed in an ␣ftz-f1 RNAi background, JH activation of E75A is restored. These results demonstrate that both ␣FTZ-F1 and ␤FTZ-F1 can mediate the JH-dependent activation of E75A.
The AF2 of FTZ-F1 Is Required for JH-dependent Interaction with MET and GCE-We previously found that FTZ-F1 interacts with the bHLH-PAS transcription factors MET and GCE in FIGURE 1. Isoform-specific expression of ftz-f1 in the S2 cell line. A, structural organization of the ftz-f1 gene. Black bars, exons; arrows, promoters for the ␣ and ␤ isoforms. Each isoform possesses a unique coding sequence at the 5Ј end. Rectangles above the exons indicate probes specific for the ␣ (white) and ␤ (gray) isoforms as well as a common probe (striped). Alternative polyadenylation sites are indicated by vertical arrows. Shown at the top is the size of ftz-f1 in kilobases. B, total RNA was isolated from S2 cells cultured in the presence of 1 ϫ 10 Ϫ6 M ecdysone for the time period indicated above each lane and from prepupae (PP). RNA samples were analyzed by Northern blot hybridization with radioactive probes for ␣-specific, ␤-specific, or common exons. Transcript sizes are indicated on the right. rp49 expression was used as a loading control. FIGURE 2. Both FTZ-F1 isoforms mediate JH-dependent activation of E75A. A, expression of ectopic FTZ-F1 proteins used in B and C was confirmed by Western blot hybridization (IB) using anti-V5 antibody, with tubulin as a control for loading. B, S2 cells were transfected with empty plasmid or expression plasmid encoding ␣FTZ-F1, ␤FTZ-F1, or common domains of FTZ-F1 as indicated (x axis) and then treated with ethanol solvent (light bars) or 1 ϫ 10 Ϫ6 M methoprene (dark bars) for 1 h. Total RNA was extracted, and E75A expression was measured as -fold abundance against rp49 by quantitative RT-PCR (y axis). Bars, mean Ϯ S.D. from three independent experiments. *, significant (p Ͻ 0.05) JH-dependent increase in E75A expression; **, significant JH-independent E75A activation. C, S2 cells were incubated for 48 h with doublestranded RNA targeting the 5Ј region of the ␣ftz-f1 transcript (␣ftz-f1 dsRNA), transfected with empty plasmid or ␤ftz-f1-encoding plasmid as indicated on the x axis, and treated with ethanol (light bars) or 1 ϫ 10 Ϫ6 M methoprene (dark bars) for 1 h. Expression of E75A relative to rp49 was measured as in B. Bars, mean Ϯ S.D. from three independent experiments. *, significant (p Ͻ 0.05) change in E75A expression.
GST pull-down assays (12). Because nuclear receptors typically utilize the AF2 function of their LBD to interact with coregulators (20), we reasoned that FTZ-F1 may use its AF2 to form heterodimers with MET and GCE. To test this possibility, we employed a two-hybrid system designed to quantitatively measure protein-protein interactions in cultured insect cells (42). GAL4-DBD fusions of MET or GCE were coexpressed in S2 cells with p65-AD fusions of ␣FTZ-F1, ␤FTZ-F1, or mutant FTZ-F1 with a 15-residue C-terminal deletion. The truncated protein (␤FTZ-F1 ⌬H12 ) lacks the helix H12, which is critical to the AF2 function of nuclear receptors (20). When MET or GCE are expressed alone, little or no reporter activity is seen when cells are treated with methoprene (Fig. 3). However, when either MET or GCE is coexpressed with ␣FTZ-F1, substantial JH-dependent reporter activity is evident (4.3-and 3.0-fold, respectively). Similar JH-dependent activity is observed when either MET or GCE is coexpressed with ␤FTZ-F1 (4.7-and 3.9-fold, respectively). These results demonstrate that in the two-hybrid system, MET and GCE interact with both isoforms of FTZ-F1, and these interactions are JH-dependent. When H12 is removed from FTZ-F1, the JH-dependent interaction of MET with the truncated FTZ-F1 is reduced significantly (p Ͻ 0.05), but the proteins still retain a notable level of interaction (Fig. 3A). For GCE, the interaction with truncated FTZ-F1 is nearly abolished (Fig. 3B). These results indicate that FTZ-F1 requires an intact AF2 to bind MET and GCE, with each paralog displaying distinct requirements for this motif.
Characterization of MET and GCE Structure Identifies Potential FTZ-F1 Interaction Sites-It has been well documented that the AF2 of nuclear receptors forms a groove that interacts with ␣-helical NR box peptides containing the sequence LXXLL (20). In addition to this motif, other ␣-helical structures resembling LXXLL (e.g. LXXIL, LLXXL, and FXXLL) have been identified as functional NR boxes (43)(44)(45). There is only one LXXLL motif, situated N-terminal to PAS A in MET, that potentially represents a canonical NR box, LMQLL. However, this motif is not found in GCE or any other insect homolog, and its deletion from MET has no impact on the interaction with FTZ-F1. In order to identify any ␣-helices in MET and GCE containing non-conventional NR box motifs, we characterized the structure of these proteins with Phyre2 (33). Phyre2 generates models of tertiary structure by homology to empirically derived templates, and when a template is not available, it provides information about secondary structures.
MET and GCE are multidomain proteins, containing both bHLH and PAS, and as a result Phyre2 identified three-dimensional models for each of these domains. Using the bHLH transcription factor MyoD (PDB entry 1MDY) as a template, models were built for MET and GCE (supplemental Fig. S1) whose bHLH corresponds to residues 38 -90 and 9 -58, respectively. The domain consists of a long ␣-helix containing the N-terminal basic region (MET 38 -63, GCE 9 -34), a loop (MET 64 -73, , and a second shorter ␣-helix (MET 74 -90, GCE 45-58). The structure of the MET and GCE bHLH is typical of this domain (46), and as expected, no putative NR boxes were identified in this region.
For modeling of the PAS region, Drosophila PER (PDB entry 1WA9) was used as a template. PER is a founding member of the PAS family and, similar to MET and GCE, possesses two tandem PAS motifs (47). MET and GCE both adopt two PAS folds (designated 1 and 2) that correspond to residues 134 -506 and 83-371, respectively (Fig. 4A). Each fold consists of a fivestrand antiparallel ␤-sheet flanked on one side by three or four ␣-helices (i.e. a canonical PAS fold) (18). PAS Fold 1 is composed of PAS A and three subsequent sequence blocks (I-III) (Fig. 4B). In this first fold, there are two flexible loops that vary in size between MET and GCE (Table 1), directly following PAS A and between blocks II and III. PAS Fold 2 is composed of PAS B and PAC, with no intervening loops. Inspection of individual secondary structures (Table 1) reveals that each ␣-helix in the PAS region is part of a PAS fold. Given that typical NR boxes, such as those found in the bHLH-PAS-containing SRC coactivators, are always isolated from other domains (22), the region of MET and GCE containing the PAS folds is unlikely to contain an NR box.
No template was found for the remaining C-terminal residues of MET (residues 507-716) and GCE (residues 373-689). Analysis of secondary structures in this region reveals that although most of the C terminus consists of paralog-specific features, there are two sequence blocks (IV and V) with similar ␣-helical structure and position in both proteins (Fig. 4B). Block IV (MET 507-524, GCE 373-390) is an 18-residue ␣-helix immediately following PAC. Block V (MET 547-558, GCE 411-422) is a 12-residue ␣-helix following block IV, flanked on either side by disordered regions. These helices, designated ␣IV and ␣V (Table 1), are distinct from other protein structures, and we considered them as candidate sites for an NR box.

Identification of Novel NR Box Motif in C Terminus of MET
and GCE-To assess the functional relevance of ␣IV and ␣V, a multiple sequence alignment was generated based on the assumption that functionally important residues are conserved. MET or GCE homologs have been identified in all holometabolous insects with sequenced genomes and in several cases have been implicated in JH signaling (3,8,12,15). We therefore aligned MET and GCE from Drosophila melanogaster to homologs from two insects of the order Diptera, a distant relative within the Drosophila genus (Drosophila grimshawi) and a mosquito species (Aedes aegypti). We also included in the alignment species from the distant orders Coleoptera (Tribolium castaneum) and Hymenoptera (Nasonia vitripennis). Together, these insects provide an evolutionary distance ranging from 60 to 300 million years (48,49). The alignment revealed that ␣IV and ␣V exhibit sequence homology (61 and 58%, respectively) similar to that displayed by other conserved domains (Table 1 and supplemental Fig. S2). These two ␣-helices (␣IV and ␣V) are thus well conserved, suggesting they provide some critical function in MET and GCE. The sequence alignments of ␣IV and ␣V were inspected for potential NR box motifs (Fig. 5A). Intriguingly, ␣V possesses a motif whose sequence, L(I/L)XXL, is reminiscent of the NR box. In contrast to the LXXLL sequence found only in MET (Fig. 4B, arrowhead), this LIXXL motif is present in both MET and GCE (Fig. 4B, vertical arrow) and is conserved in other insects (Fig. 5A). Because mutation of either of the outer leucine residues of an NR box is sufficient to completely abrogate its function (50), a conserved leucine at the ϩ1 position was mutated to alanine in MET and GCE (L554A and L418A, respectively), and their interaction with FTZ-F1 was measured in the two-hybrid system. For both MET L554A (Fig. 5B) and GCE L418A (Fig. 5C), we observed a significant (p Ͻ 0.05) reduction in the interaction with FTZ-F1 compared with the wild type proteins. By contrast, deletion of the ϩ4 and ϩ5 residues from the LXXLL motif of MET had no significant effect on the interaction with FTZ-F1 (Fig. 5B). The reduced ability of MET L554A and GCE L418A to interact with FTZ-F1 suggested that the conserved motif in ␣V is a novel NR box used to interact with AF2.
We noticed that mutation of the NR box did not fully abrogate the FTZ-F1⅐MET and FTZ-F1⅐GCE interactions. This was especially evident for MET, which retained roughly 50% of its interaction in the absence of the NR box (Fig. 5B), similar to the residual interaction of MET with H12-deficient FTZ-F1 (Fig.  3A). To identify other possible interacting sites, we inspected residues C-terminal to the NR box. Although this region is poorly conserved among distant homologs (Table 1) we found several regions conserved in the Drosophila genus, including a Q R region ( Fig. 4B and supplemental Fig. S3), which is used as a secondary NR interaction site by SRC family members (26 -29). Alignment of the MET and GCE Q R regions revealed a short block of common glutamine/glutamate residues (Fig. 5A). Mutation of amino acids within this Q R stretch by itself does not significantly disrupt the FTZ-F1⅐MET interaction, but when the NR box is also mutated, the interaction between FTZ-F1 and MET is completely abolished (Fig. 5B). A very dif-ferent result was seen for GCE, because mutation in the Q R region had no disruptive effect whatsoever on the interaction with FTZ-F1 (Fig. 5C). These results indicate that, in addition to the NR box, MET utilizes a secondary interaction site in the C-terminal Q R .
Unique NR Box-AF2 Complex Underlies FTZ-F1⅐MET and FTZ-F1⅐GCE Interactions-Because the NR box in MET and GCE consists of a novel sequence (LIXXL), we were interested in investigating the molecular basis of its interaction with FTZ-F1. We therefore modeled the interactions by simulated docking with HADDOCK (36) and Rosetta FlexPepDock (38) using the recently derived FTZ-F1 LBD crystal structure (PDB entry 2XHS) and 9-residue MET or GCE peptides generated de novo. Because the interaction between the FTZ-F1 AF2 and the FTZ NR box was characterized empirically through the crystal structure (21), the docking experiments allowed us to compare the interaction mechanism of the LIXXL motif with that of a canonical LXXLL motif.
A comparison of the lowest energy structures of MET and GCE with the empirically determined structure of FTZ is shown in Fig. 6A. The MET peptide (LYLIENLQK, residues 552-560) and GCE peptide (LRLIQNLQK, residues 416 -424) are positioned similarly to FTZ peptide along the AF2 groove formed by helices H3, H3Ј, H4, and H12 of FTZ-F1. The MET and GCE peptide backbones are shifted relative to FTZ (root mean square deviation of 2.74 and 2.00 Å, respectively), positioning them deeper in the hydrophobic groove and closer to H3 (Fig. 6A). The MET and GCE backbones also show distinct positioning from one another (root mean square deviation of 1.95 Å). The N-terminal half of GCE is shifted closer to the C-terminal end of H12 as compared with MET, whereas the C-terminal half of GCE lies deeper in the hydrophobic groove and closer to H4 (Fig. 6A).
The NR box peptides of MET and GCE are stabilized in the FTZ-F1 AF2 by contacts with a series of hydrophobic side chains (Fig. 6, B and C). Residues Leu-554/Leu-558 of MET (Fig. 6B) and Leu-418/Leu-422 of GCE (Fig. 6C), located at the  ϩ1 and ϩ5 positions of the NR box, are embedded deeply within the AF2 groove. Due to the different positioning of the peptides relative to FTZ-F1, the contacts made by these residues are distinct for MET and GCE. Leu-554 of MET interacts with Val-621, Leu-642, Gln-643, Leu-792, and Met-796 of FTZ-F1, and Leu-558 interacts with Met-639 and the aliphatic region of Arg-625 (Fig. 6B). Leu-418 of GCE interacts with Val-621, Met-639, Leu-642, Leu-792, and Met-796 of FTZ-F1, whereas Leu-422 interacts with Val-621, Arg-625, Val-635, Gln-638, and Leu-642 (Fig. 6C). A notable feature of both NR boxes is the positioning of the isoleucine at the ϩ2 position (Ile-555 in MET, Ile-419 in GCE). For canonical NR boxes, an inner hydrophobic residue at the ϩ4 position rests along the rim of the AF2 groove and interacts with side chains of H3 (20). By contrast, the ϩ2 isoleucine of the MET/GCE NR box rests along the opposing rim and interacts with Met-639 on H4 (Fig. 6, B and  C). This mode of interaction is unusual although not novel because a similar mechanism was reported for the interaction of ERR␣ with an LLXXL motif in the coactivator PGC-1␣ (45).
A characteristic feature of many NR box-AF2 interactions is the use of a charge clamp, consisting of a conserved glutamic acid residue on H12 and a conserved arginine/lysine on H3; these residues typically form hydrogen bonds with backbone atoms in the core NR box ϩ1 and ϩ5 residues, stabilizing the NR box in the AF2 groove (20). FTZ-F1 utilizes these charge clamp residues (Arg-625 on H3 and Glu-795 on H12) to form hydrogen bonds with the LXXLL NR box of FTZ (21). However, we did not observe any charge clamp interactions in the lowest energy structures for the FTZ-F1⅐MET and FTZ-F1⅐GCE interactions (Fig. 6, B and C). Moreover, a comprehensive analysis of interactions from the top 10 models (Table 2) found no hydrogen bond formation between FTZ-F1 and the core LIXXL residues. The docking experiments therefore predict that the NR box of MET and GCE requires extensive hydrophobic contacts to interact with FTZ-F1 but does not require the canonical charge clamp residues.
To test the mechanism of interaction identified from docking experiments, point mutations were introduced in FTZ-F1 and were tested in two-hybrid assays. We chose hydrophobic residues from each helix of the AF2 that are predicted to interact with the core NR box motif as well as the charge clamp residues predicted not to play a role in the interactions with MET and GCE. When individual hydrophobic residues from H3 (Val-621), H4 (Val-635, Met-639, and Leu-642), or H12 (Leu-792 and Met-796) of FTZ-F1 are mutated to a charged aspartic acid, the interaction of FTZ-F1 with MET is significantly (p Ͻ 0.05) reduced (Fig. 7A). Mutation of Leu-642 in particular severely reduced the interaction and was equivalent to the FTZ-F1 ⌬H12 deficiency. Although mutation of Leu-792 produced no significant effect on the interaction of FTZ-F1 with GCE, mutating each of the other hydrophobic residues did significantly reduce this interaction (Fig. 7B). The double mutation L792D/M796D severely reduced the interaction with both MET and GCE and was also equivalent to the FTZ-F1 ⌬H12 deficiency. Whereas the hydro-  phobic residues play a similar role in binding both paralogs, mutation of residues in H4 or the C-terminal hydrophobic residue in H12 (Met-796) has a greater impact on the interaction with GCE compared with MET, consistent with the positioning of the GCE peptide closer to this side of the AF2 groove (Fig. 6A). By sharp contrast, when either of the charge clamp residues Arg-625 on H3 or Glu-795 on H12 are mutated to alanine FTZ-F1 retains the full interaction with both MET and GCE, and in fact the E795A mutation enhances the interaction with GCE (Fig. 7, A and B). The results from the two-hybrid assays are thus consistent with the predictions made by docking experiments, indicating that FTZ-F1 utilizes hydrophobic residues of its AF2 and not the canonical charge clamp to interact with the novel NR box in MET and GCE.

FTZ-F1 Mediates JH Signaling through Mechanism Common
to Both Isoforms-FTZ-F1 is a critical transcriptional regulator in the embryonic, larval, and pupal stages of development whose expression is strictly isoform-specific; ␣FTZ-F1 is only present in early embryos, whereas ␤FTZ-F1 is expressed in sharp peaks during postembryonic development (40,41). We found that only ␣FTZ-F1 is present in S2 cells (Fig. 1B), thus implicating ␣FTZ-F1 as the isoform responsible for mediating JH action in this cell line. However, JH induction can be mediated by either natural isoform or by FTZ-F1 lacking the isoform-specific N terminus (Fig. 2), suggesting that an important component of JH signaling resides in the common domains of FTZ-F1. Based on our observations that FTZ-F1 is bound in vivo to multiple enhancers upstream of E75A (12) and interacts with the candidate JH receptors MET and GCE (Fig. 3), we hypothesized that a critical interaction site common to both isoforms may enable FTZ-F1 to recruit MET or GCE to the target promoter. We focused on the FTZ-F1 LBD because this domain often plays a critical role in transcriptional regulation by nuclear receptors. We found that disruption of AF2 by truncation or point mutations prevents FTZ-F1 from interacting with either MET or GCE (Figs. 3 and 7). The AF2 is a characteristic feature of nuclear receptor LBDs, and its role in recruiting transcriptional coregulators is well documented (20). So far, however, the only known function of this domain in FTZ-F1 occurs during embryonic development, when it interacts with the NR box of the homeodomain protein FTZ (21,30,31). We have now identified an additional function for the FTZ-F1 AF2: mediating JH-dependent transcriptional activation. Because the AF2 is present in both isoforms, the findings described here suggest that JH activation through FTZ-F1 is an isoform-independent process that could occur during embryogenesis through ␣FTZ-F1 or during postembryonic development through ␤FTZ-F1.
Candidate JH Receptors MET and GCE Interact JH-dependently with FTZ-F1-Although a clear role for MET in JH regulation is apparent in several insects (3,5,6,8,12,15), there is still significant debate over the molecular basis for JH action. In Drosophila, both MET and GCE mediate sensitivity to exogenous JH and mutation of either protein disrupts JH function in vivo (11), suggesting that both paralogs may be involved in JH signaling. In light of our recent discovery that FTZ-F1 and GCE are required for JH-dependent activation of E75A (12), we were interested to determine the role of JH in the interaction between FTZ-F1 and the paralogous JH receptor candidates. By examining the FTZ-F1⅐MET and FTZ-F1⅐GCE interactions using a quantitative two-hybrid assay, we found that both paralogs need JH to form heterodimers with FTZ-F1 (Figs. 3, 5, and 7). Our results thus show that both MET and GCE can mediate JH-dependent protein-protein interactions, providing additional evidence that both paralogs are involved in JH signaling.
Novel NR Box Enables JH-dependent Interaction of MET and GCE with FTZ-F1-We previously found that MET and GCE form functional heterodimers with FTZ-F1 that can activate transcription through the FTZ-F1 response element (12). Because our findings may have important implications in understanding the molecular basis for JH signaling, we were interested in characterizing the mechanisms driving the FTZ-F1⅐MET and FTZ-F1⅐GCE interactions. To identify potential nuclear receptor interacting motifs, we used homology modeling and conservation analysis to characterize the secondary and tertiary structure of MET and GCE. One result of this analysis was the identification of the complete PAS domains for these proteins. PAS domains are known to function as LBDs with ligand typically bound in a pocket formed by the ␤-sheet and ␣-helices (14). MET and GCE each possess two tandem PAS domains, and we found that they both form canonical folds (Fig.  4A). PAS Fold 2 is composed of the previously identified PAS B and adjacent PAC motifs, an arrangement that is typical of many PAS domains (18). PAS Fold 1 has a less conventional structure because its C-terminal half is made from discrete blocks of conserved sequences interrupted by flexible loops and does not form a characteristic motif. Loops are common in PAS domains; they bind cofactors and mediate intramolecular interactions and are thought to provide functional diversity to PAScontaining proteins (18,51,52). Interestingly, although sequences from structural elements of the PAS folds are conserved (Table 1), the sequence and length of the PAS loops are highly variable, suggesting that the MET and GCE loops may confer functional diversity on homologs in different insects. Overall, characterization of the PAS folds should aid future structure-function studies of MET and GCE in their capacity to bind ligand and form heterodimers.
A second result of the in silico studies was the identification of two ␣-helices in MET and GCE located C-terminal to PAS Fold 2. One of these possesses a sequence resembling an NR box, LIXXL, that is critical to the interactions with FTZ-F1 (Fig.  5). There is a growing list of functional NR boxes with noncanonical sequences similar to the LXXLL motif, having hydrophobic residues arranged in ϩ1,ϩ2,ϩ5 or ϩ1,ϩ4,ϩ5 orientation and an ␣-helical structure (43)(44)(45). Indeed, docking simulations and two-hybrid assays suggest that the conserved LIXXL motif of MET and GCE is a non-conventional NR box (Figs. 6 and 7). Similar to canonical NR boxes like the one found in FTZ, the MET/GCE NR box interacts with hydrophobic residues from the AF2; however, it possesses two unusual features. First, the LIXXL motif uses an isoleucine residue at the ϩ2 position to interact with residues on H4 of the AF2 (Fig. 6); by contrast, LXXLL motifs use a ϩ4 leucine to interact with resi-dues on H3 (20). Second, the charged residues Glu-795 and Arg-625 of FTZ-F1 that form hydrogen bonds with the peptide backbone of canonical NR boxes, such as the one in FTZ (21), are dispensable to the interactions with MET and GCE (Figs. 6 and 7), suggesting these interactions are formed primarily through hydrophobic residues. These two findings indicate that MET and GCE utilize a non-canonical NR box whose mechanistic basis of interaction differs from the traditional LXXLLtype NR box. Our finding is consistent with emerging evidence that diverse modes of interaction exist for NR box-AF2 complexes. An interesting example of this comes from the coactivator PGC1-␣, which possesses two functional NR boxes. A canonical LXXLL NR box interacts with TR␤1 and PPAR␥ through a mechanism that does not require charged residues (53, 54); a second, non-canonical LLXXL NR box interacts with ERR␣ through a charge clamp-dependent mechanism (45). To our knowledge, the MET/GCE motif represents the first example of an NR box that is both non-canonical in sequence and utilizes a charge clamp-independent mode of interaction.
Functional Divergence Is Evident in C Terminus of MET and GCE-Gene duplication is an important mechanism allowing gene products to assume new functional roles over time (55). The Met gene likely arose by duplication of ancestral gce prior to the origination of the Drosophila genus (56), and consistent with an evolutionarily recent duplication, MET and GCE share substantial sequence similarity in their bHLH and PAS domains and are largely redundant in vivo (8 -10). In our study, we found that in addition to the NR box, there is a Q R region in the C terminus that has paralog-specific function and is used only by MET as a secondary nuclear receptor interaction site (Fig. 5). Inspection of the Q R region in drosophilids reveals that in addition to distinct positioning of the Q R within the proteins (Fig. 4B) there are clear differences in structure and conservation; the Q R of MET is significantly shorter (ϳ30 residues) and contains polyglutamine tracts similar to those found in some SRC family members (26 -29), whereas the Q R of GCE is over 70 residues in length and is homologous to a Q R region found in mosquito MET (supplemental Fig. S3). In addition, the MET and GCE possess paralog-specific acidic and P/S regions in the C terminus ( Fig. 4B and supplemental Fig. S3). Because the in vivo functions of MET and GCE are not entirely redundant (9), we suggest that functional divergence may be attributed in part to the paralog-specific Q R regions and other domains located in the C termini of MET and GCE.
In summary, we have examined the molecular basis of JH-dependent interaction between the nuclear receptor FTZ-F1 and the bHLH-PAS proteins MET and GCE. A novel NR box enables both MET and GCE to interact JH-dependently with the AF2 of FTZ-F1. We suggest that this mechanism functions in the JH transcription activation of E75A and may play a significant role in mediating JH signaling in vivo.