Molecular Basis of Coiled Coil Coactivator Recruitment by the Aryl Hydrocarbon Receptor Nuclear Translocator (ARNT)*

The aryl hydrocarbon receptor nuclear translocator (ARNT) serves as the obligate heterodimeric partner for bHLH-PAS proteins involved in sensing and coordinating transcriptional responses to xenobiotics, hypoxia, and developmental pathways. Although its C-terminal transactivation domain is dispensable for transcriptional activation in vivo, ARNT has recently been shown to use its N-terminal bHLH and/or PAS domains to interact with several transcriptional coactivators that are required for transcriptional initiation after xenobiotic or hypoxic cues. Here we show that ARNT uses a single PAS domain to interact with two coiled coil coactivators, TRIP230 and CoCoA. Both coactivators interact with the same interface on the ARNT PAS-B domain, located on the opposite side of the domain used to associate with the analogous PAS domain on its heterodimeric bHLH-PAS partner HIF-2α. Using NMR and biochemical studies, we identified the ARNT-interacting motif of one coactivator, TRIP230 as an LXXLL-like nuclear receptor box. Mutation of this motif and proximal sequences disrupts the interaction with ARNT PAS-B. Identification of this ARNT-coactivator interface illustrates how ARNT PAS-B is used to form critical interactions with both bHLH-PAS partners and coactivators that are required for transcriptional responses.

PAS (Per-ARNT-Sim) 3 domains are small, modular domains that mediate interactions with proteins (and often small molecule ligands) to coordinate cellular responses to diverse environmental stimuli (1). The largest class of eukaryotic PAS-containing proteins are transcription factors containing a basic helix-loop-helix (bHLH) DNA-binding domain followed by two PAS domains, known as the bHLH-PAS family (2). As functional heterodimers, bHLH-PAS complexes bind DNA at specific promoter elements and recruit transcriptional coactiva-tors via their C-terminal transactivation domains (TADs) to regulate transcriptional responses to diverse stimuli. In particular, the ubiquitously expressed protein ARNT is of central importance within the bHLH-PAS family, acting as the obligate heterodimeric partner for the aryl hydrocarbon receptor (AHR), hypoxia-inducible factor-␣ (HIF-␣), and single-minded (SIM) to regulate transcriptional responses to xenobiotics, hypoxia, and neurogenesis, respectively (3). ARNT heterodimeric complexes are involved in the etiology and progression of many forms of human cancer through the metabolic activation of dietary and environmental carcinogens (AHR⅐ARNT) and the adaptation of solid tumors to chronic hypoxia (HIF-␣⅐ARNT, also known as HIF) (4,5).
Heterodimerization of bHLH-PAS proteins is mediated by contacts between both the bHLH and the tandem PAS domains, with inter-PAS domain contacts playing a critical role in the specificity and stability of heterodimer formation (Fig.  1A) (6 -9). Structures of the HIF-2␣⅐ARNT PAS-B heterodimer demonstrate that the two PAS-B domains associate via their ␤-sheets (Refs. 10, 11 and Fig. 1B) and disruption of this interface by point mutations significantly attenuates the transcriptional response to hypoxia by destabilizing the fulllength heterodimer complex (8,9). In addition to forming stable complexes on specific DNA promoters, bHLH-PAS heterodimers must also recruit transcriptional coactivators, a diverse pool of proteins that are required for histone modification and/or recruitment of general transcriptional machinery. The dynamic association of coactivators with sequence-specific transcription factors allows the temporal-and tissue-specific modulation of transcription, as well as influencing the specificity of target gene induction (reviewed in Ref. 12).
The use of glutamine-rich C-terminal TADs by HIF-␣ and AHR to recruit coactivators is well documented (13)(14)(15), though the ARNT C-terminal glutamine-rich domain is apparently dispensable for coactivator recruitment in vivo (16,17). The recent discovery of coiled coil coactivators that target domains other than the C-terminal TADs establishes that multiple structural motifs are used by bHLH-PAS proteins to recruit coactivators (18 -21). Domain-swapping studies have shown that PAS domains within the bHLH-PAS family contribute to the specificity of target gene induction, suggesting that PAS domains themselves participate in the recruitment of specific coactivators (22). Two candidates for such PAS-directed coactivators are CoCoA (coiled coil coactivator) (20) and TRIP230 (thyroid hormone receptor/retinoblastoma-interacting protein 230, also known as TRIP11) (18), which appear to interact with the N-terminal bHLH and/or PAS domains of ARNT. TRIP230 is recruited to endogenous promoters after hypoxia or xenobiotic stress and is required for the function of ARNT heterodimers in vivo (18), highlighting the previously unappreciated role of the N-terminal domains of ARNT in transcriptional regulation.
In this study, we investigate the molecular basis for the interaction of ARNT with coiled coil coactivators. We demonstrate that the coiled coil coactivators TRIP230 and CoCoA bind a similar interface on ARNT that maps to the ␣-helical face of PAS-B. Our data suggest an unrecognized commonality in PAS domain interactions, because an analogous interface has been described for the interaction the PAS-B domain of the SRC-1 coactivator with several helical targets, mediated by an LXXLL motif (L is Leu; X is any amino acid) (23,24). Notably, detailed characterization of the ARNT-interacting sequence on TRIP230 revealed that an LXXLL-like motif is critical for the interaction, demonstrating functional similarities in recruitment of helical targets by PAS domains. Collectively, these data point to a model of simultaneous engagement of bHLH-PAS heterodimeric partners and coactivators by ARNT PAS-B.
Proteins were expressed in Escherichia coli strain BL21(DE3) (Novagen). Isotopically labeled proteins were prepared using M9 minimal medium containing 15 NH 4 Cl and [ 13 C 6 ]glucose as the sole nitrogen and carbon sources, respectively. ARNT PAS domains were expressed as His 6 or His 6 G␤1 fusions and purified from bacterial lysate by Ni 2ϩ -Sepharose (GE Healthcare) affinity chromatography. The His 6 tag was cleaved from U-15 N His 6 -ARNT PAS-B for NMR spectroscopy by incubating overnight with His 6 -TEV (Tobacco Etch Virus) protease (27), followed by Ni 2ϩ -Sepharose and gel filtration (Superdex 75) chromatography, resulting in purified U-15 N ARNT PAS-B with 4 vector-derived N-terminal residues (GAMD). All coactivator fragments were expressed as GST fusions and purified from bacterial lysate by glutathione-Sepharose (GE Healthcare) affinity chromatography. Proteins were cleaved overnight with His 6 -TEV, followed by Ni 2ϩ -Sepharose and either gel filtration (Superdex 75) or cation exchange (Mono S) chromatographies to remove the protease and GST, respectively. Purified coactivator fragments contain a 17-residue vector-derived N-terminal sequence (GAMDPEFKGLRRRAQLV). All protein concentrations were determined by absorbance at 280 nm using predicted extinction coefficients (in M Ϫ1 cm Ϫ1 ) of: His 6 GB1-ARNT PAS-A, 33 Pull-down Assays-Soluble E. coli lysate expressing GSTtagged coactivator fragments was generated by lysing cells in phosphate-buffered saline (10 mM sodium/potassium phosphate, pH 7.4, 137 mM NaCl, 2.7 mM KCl) by high-pressure extrusion, followed by centrifugation at 20,000 ϫ g for 30 min at 4°C. Glycerol was added to a final concentration of 8% (v/v) to the supernatant, and aliquots were quick frozen in liquid nitrogen for storage at Ϫ80°C. Overexpression of GST-tagged fragments within soluble extracts was analyzed by SDS-PAGE to estimate equal input for pull-down assays. Equal amounts of GST-tagged coactivator fragments were incubated with 12 l of Ni 2ϩ -Sepharose beads (GE Healthcare) in binding buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 20 mM imidazole, 5 mM ␤-mercaptoethanol) in the presence or absence of 2 M His 6 -tagged ARNT PAS domain constructs (0.03-0.07 mg/ml) in a total volume of 400 l for 4 -16 h at 4°C. Beads were washed twice with 400 l of binding buffer and eluted with 8 l of 4ϫ SDS Laemmli buffer. Samples were resolved on a 10% Bis-Tris NuPAGE gel run with 1ϫ MES resolving buffer (Invitrogen) and Coomassie Blue-stained for detection of bound proteins.
Circular Dichroism Spectroscopy-Protein samples were buffer-exchanged into 10 mM sodium phosphate, pH 6.5, 17 mM NaCl. All spectra were collected with 15 M protein concentrations (ranging from 0.09 to 0.2 mg/ml) on an Aviv 62D Spectropolarimeter at 25°C using a 0.1-cm path length quartz cuvette, recording every 1.0 nm between 190 and 260 nm. The reference signal from buffer was subtracted from all spectra, which represent the mean of three independent scans. Molar residue ellipticities were calculated from raw ellipticities (mdeg) using Equation 1.
NMR Spectroscopy-All NMR experiments were performed at 25°C on Varian INOVA 600 or 800 MHz spectrometers equipped with 1 H, 13 C, 15 N indirect detection triple resonance probes and a Z-axis pulsed field gradient. 15 N/ 1 H HSQC titration experiments of 200 M U-15 N ARNT PAS-B were carried out by the stepwise addition of natural abundance Trip230 1583-1688 or CoCoA 407-535 (50 -300 M over 4 steps); samples were buffer-exchanged to 50 mM Tris, pH 7.5, 17 mM NaCl, 5 mM dithiothreitol, concentrated to 500 l using an Amicon stirred cell concentrator with a 10 kDa MWCO filter, and D 2 O was added to a final concentration of 10% (v/v). Differential broadening analysis was performed in NMRView (29) by obtaining the ratio of intensities of well-resolved ARNT PAS-B peaks in the presence and absence of coactivator fragments; residues broadened Ͼ 1 beyond the mean were mapped on the ARNT PAS-B structure (PDB 1X0O) (10). 15 N/ 1 H HSQC titration experiments of 100 M U-15 N TRIP230 1583-1620 were carried out by stepwise addition of ARNT or HIF-2␣ PAS-B (50 -900 M in 6 steps) in 50 mM MES pH 6.5, 17 mM NaCl, 5 mM dithiothreitol, and a 3-kDa MWCO filter was used during concentration. Significantly perturbed residues for which the ⌬␦ TOT Ͼ 0.05 ppm were obtained using Equation 2,

Interaction of ARNT PAS-B with Coiled Coil Coactivators
normalized for proton with the scale factor ϭ 0.17, established from estimates of atom-specific chemical shift ranges in a protein environment (30).
AlphaScreen Assay-Binding affinities for GST coactivators and ARNT PAS-B-FLAG were determined by AlphaScreen assay using glutathione donor beads and FLAG acceptor beads according to the manufacturer's instructions (PerkinElmer Life Sciences). Briefly, purified GST coactivators were titrated from either 12.

ARNT PAS-B Is the Primary Site of Coiled Coil Coactivator
Binding-Prior biochemical studies on the ARNT binding abilities of several coiled coil coactivators identified large fragments required for interaction. An ARNT-interacting region of 133 residues was identified for TRIP230 spanning residues 1583-1716 in the conserved C-terminal domain, near its thyroid hormone receptor-binding LXXLL motif (18,35). The ARNT interaction site was not defined on the 631-residue CoCoA protein (20); however, a central coiled coil region critical for interaction with the bHLH-PAS domains of the GRIP1 coactivator is predicted to contain two distinct coiled coil regions (36). Based on these predictions, we subcloned both coiled coil regions to determine their interaction with ARNT by pull-down assay and identified residues 407-514 as the ARNTinteracting motif in CoCoA (data not shown).
To complement these data with higher resolution identification of the coiled coil coactivator-interacting regions of ARNT, we purified both His 6 -tagged ARNT PAS domains alone and in tandem for use in pull-down assays with GST-tagged coactivators (Fig. 1C). ARNT PAS-A is considerably larger than PAS-B because of the presence of several long loops compared with the canonical PAS domain, and care was taken during subcloning to maintain the required secondary structure elements of the PAS fold. An examination of all purified PAS domains by 15 N/ 1 H HSQC spectra demonstrated excellent 1 H chemical shift dispersion consistent with well-folded PAS domains (data not shown). Purified His 6 -tagged PAS domains were incubated with E. coli lysates containing overexpressed GST-tagged coactivator fragments or GST alone, and binding was assessed by pull-down assay using Ni 2ϩ -Sepharose. ARNT PAS-B, but not the isolated PAS-A domain, interacted directly with fragments of either TRIP230 or CoCoA (Fig. 1D). Moreover, the tandem PAS-AB fragment bound equivalent amounts of both coiled coil coactivators as the PAS-B domain alone, suggesting that it is unlikely that the PAS-A domain has a high affinity secondary binding site on the PAS-A domain that cooperatively recruits coiled coil coactivators. Therefore, ARNT PAS-B domain is the primary site of interaction for both TRIP230 and CoCoA.
Mapping  (Fig. 2A). Significantly broadened peaks were found throughout the primary sequence and plotted onto the structure of ARNT PAS-B (Fig. 2B), mapping to the ␣-helical face of the domain. The NMR titration experiments were repeated with CoCoA 407-535 and a similar interface was identified (Fig.  2C). In contrast, the interaction of ARNT PAS-B with HIF-2␣ PAS-B selectively perturbs residues located on the opposite side of the protein on the ␤-sheet (Fig. 2D and Ref. 10), demonstrating that interaction with coactivators and HIF-2␣ occur on opposite sides of the ARNT PAS-B domain.
Mutation of ARNT ␣-Helical Residues Disrupts Interaction with TRIP230 but Not HIF-2␣ PAS-B-We then mutated a subset of residues on the ARNT ␣-helical face that were significantly broadened by interaction with both coactivators (Fig. 2E) to disrupt the interaction. Purified His 6 -tagged WT or mutant PAS-B domains were incubated with E. coli lysates containing overexpressed GST-TRIP 1583-1688 and binding was monitored by pull-down assay (Fig. 2F). Mutations for I369A, V397A, and K419A, either singly or in combination, partially disrupted the interaction with GST-TRIP 1583-1688 . We then tested whether the ARNT PAS-B I369A/V397A double mutant could still interact with HIF-2␣ PAS-B, which utilizes the opposing ␤-sheet interface (10). The 15 N/ 1 H HSQC spectrum of I369A/ V397A displayed minimal chemical shift perturbations that were localized to the ␣-helical surface (supplemental Fig. S1).
As shown in Fig. 2G, both WT and I369A/V397A ARNT PAS-B domains exhibited similar HIF-2␣-induced chemical shift perturbations consistent with binding of the HIF-2␣ PAS-B on the intermediate-fast exchange timescale as previously described (10). Therefore, mutation of the ARNT PAS-B ␣-helical face selectively disrupts interaction with coactivators.
Identifying a Minimal Interacting Region of TRIP230-We subsequently chose to focus on TRIP230 to characterize the ARNT-interacting motif because it is required for transcriptional responses by both AHR⅐ARNT and HIF-␣⅐ARNT complexes and interacts exclusively with ARNT within the bHLH-PAS family (18). Our initial mapping studies showed that a very C-terminal fragment of the TRIP230 ARNT-interacting motif spanning residues 1663-1716 (18) was not required for interaction with ARNT PAS-B (supplemental Fig. S2). Furthermore, other constructs containing N-terminal extensions beyond residue 1583 had no additional effect on ARNT PAS-B binding (data not shown). We then made a series of truncations of the remaining fragment based on secondary structure and coiled coil predictions (Fig. 3A) and performed pull-down assays with His 6 -ARNT PAS-B to qualitatively determine which fragments retained binding to the PAS-B domain. The N-terminal segment containing residues 1583-1620 retained its interaction with PAS-B, while the C-terminal fragment did not (Fig. 3B). Surprisingly, a truncation eliminating 10 residues with no predicted secondary structure (residues 1583-1593) significantly decreased binding to ARNT PAS-B, demonstrating the importance of these residues for interaction with ARNT.
Structural Characteristics of the Minimal TRIP230 Peptide-We obtained circular dichroism spectra to determine the secondary structure of the minimal TRIP230 peptide comprising residues 1583-1620. The peptide was not a stable coiled coil, but instead displayed random coil secondary structure with some ␣-helical content as measured by negative ellipticity at 200 nm with a shoulder at 222 nm, respectively (Fig. 3C). In contrast, the longer fragment of TRIP230 comprising residues 1583-1688 had a circular dichroism spectrum consistent with coiled coil structure, indicated by equal minima at 208 and 222 nm ( 222 / 208 ϭ 1.005) (37). Truncation of the C-terminal coiled coil element from residues 1663-1688 resulted in loss of coiled coil structure in all TRIP230 peptides (supplemental Fig. S3), demonstrating that this region is important for coiled coil formation or stability. However, peptides of the length required to maintain coiled coil formation (residues 1663-1688) and the ARNT interaction motif (residues 1583-1620) were not optimal for NMR, displaying poor chemical shift dispersion and peak intensities (data not shown). We therefore focused on characterizing the TRIP230 1583-1620 peptide and obtained full backbone chemical shift assignments complemented by partial side-chain assignments. Analysis of 13 C␣ chemical shifts (⌬␦ ϭ ⌬␦ obs Ϫ ⌬␦ random coil ) (38) in TRIP230 1583-1620 show a trend toward ⌬␦ Ͼ 0 for residues 1594 -1610, suggesting ␣-helix (Fig. 3D). This is in agreement with more comprehensive TALOS analysis, which uses 15 N, 13 CO, 13 C␤, and 1 H␣ chemical shifts to predict secondary structure (bar over plot in Fig. 3D) (39).
Mapping ARNT PAS-B Binding on TRIP230 Minimal Fragment by NMR-We then used NMR titration experiments to define the ARNT-interacting region on TRIP230 1583-1620 . 100 M U-15 N/ 13 C-labeled TRIP230 1583-1620 was incubated with ARNT PAS-B (from 50 -900 M, in 6 steps) and binding was monitored by 15 N/ 1 H HSQC spectra. Selected amide peaks in the N terminus of the peptide exhibited both broadening and chemical shift changes in the 15 N/ 1 H HSQC (Fig. 4A), while a similar titration experiment performed with the HIF-2␣ PAS-B showed no significant chemical shift or peak intensity changes (supplemental Fig. S4). A highlighted region of the 15 N/ 1 H HSQC spectra containing residues Leu-1588, Leu-1589, and the last residue of vector-derived sequence (a Val at position 17, V*17) (Fig. 4B) showed chemical shift perturbations and broadening while residues at the C-terminal end of the peptide were essentially unaffected. These are consistent with our pull-down data, which established that the first 10 residues of this peptide contain the important determinants of ARNT binding. We also obtained 13 C/ 1 H HSQC spectra at each of the titration points to monitor changes in TRIP230 side chains affected by ARNT PAS-B. Consistent with the 15 N/ 1 H HSQC data, well resolved 13 C␣/ 1 H␣ peaks of residues at the N terminus of the peptide (V*17C␣, S1594C␣, and T1596C␣) were significantly broadened while others in the C terminus (V1613C␣, T1614C␣) or vector-derived extreme N terminus (P*5C␣) of the peptide were unaffected (Fig. 4C).

Identification of Conserved LXXXLL Motif as the ARNT PAS-B Interaction Motif-ARNT-induced backbone chemical
shift changes were mapped by residue onto the TRIP230 1583-1620 peptide sequence (Fig. 5A). We noted a distinct periodicity of chemical shift perturbation in the N terminus of the peptide occurring every 3-4 residues. Although this region is disordered in the apo protein, the periodicity of perturbed residues suggested formation of ␣-helix upon binding ARNT PAS-B. A helical wheel projection of the first 18 residues of the peptide is consistent with formation of an amphipathic ␣-helix (Fig. 5B), with residues showing the greatest ARNT-induced perturbations mapping to the largely hydrophobic face of the helix. Notably, this region contains an LXXLL-like nuclear receptor (NR) box (LRNHLL) located within the N terminus and on the hydrophobic face of the peptide. LXXXLL motifs function similarly to NR boxes in mediating transcription factor/coactivator interactions (40). Alignment of TRIP230 sequences within this region shows that residues within the ARNT interaction motif we identified are highly conserved across insect and vertebrate animal species (underline, Fig. 5C).

Mutation of ARNT PAS-B Interaction Motif on TRIP230
Attenuates Binding-To assess the functional importance of the LXXXLL motif, we serially mutated tandem residue pairs within this motif to alanine (Fig. 5D) and monitored their effect on ARNT PAS-B binding by pull-down assay (Fig. 5E). Mutation of Site 1 to create an LXXXAA motif significantly decreased binding to ARNT PAS-B, consistent with previous loss-of-function mutations of LL sequences in NR boxes (41). Residues adjacent to NR boxes are often critical for determining the specificity and affinity of binding (42,43). We mutated conserved residues on the largely hydrophobic face of the amphipathic helix containing the LXXXLL motif. While the backbone amides of some of these residues did not exhibit significant backbone chemical shift changes in the 15 N/ 1 H HSQCmonitored titration with ARNT PAS-B, significant broadening was observed for side chain atoms of residues Asp-1593 and Thr-1596 in 13 C/ 1 H HSQC spectra (supplemental Fig. S5). Mutation of Site 2 (E1592A/D1593A) decreased PAS-B binding, demonstrating that the interruption of hydrophobicity within the amphipathic helix by these residues may be a functionally important feature. Mutation of Site 3 (Y1595A/ T1596A) also decreased binding, consistent with an important role for residues within the preformed ␣-helix in maintaining the interaction with ARNT PAS-B upon loss of the first ten residues (Fig. 3B and the ⌬N10 fragment in Fig. 5E). Furthermore, these data are consistent with quantitative measurements of the affinity between ARNT PAS-B and TRIP230 fragments as measured using AlphaScreen, a bead-based luminescence proximity assay we engineered to report on the  (48) and COILS coiled coil (49) predictions indicated above in black or gray bars, respectively. B, pull-down assay of His 6 -tagged ARNT PAS-B with GST-tagged TRIP230 fragments. C, circular dichroism spectra of TRIP230 peptides. Far-UV CD spectra of 15 M TRIP230 1583-1620 , in gray, and TRIP230 1583-1688 , in black, were recorded at pH 6.5 (25°C). D, secondary structure determination from 13 C␣ chemical shift comparison. TRIP230 1583-162013 C␣ shifts were compared against the Chemical Shift Index (38), with significant deviation from random coil set to Ϯ 0.7 ppm. TALOS prediction of secondary structure from TRIP230 1583-1620 chemical shifts is indicated above the plot in schematic form (39).
interaction of a FLAG-tagged ARNT PAS-B and GST-tagged TRIP230 proteins. Titrating GST-TRIP230 fragments into samples of ARNT PAS-B-FLAG, we obtained a K d of ϳ65 nM for the ARNT⅐TRIP230 1583-1688 complex. Mutation at Site 1 (K d ϳ 970 nM) or deletion of the N-terminal 10 amino acids (K d ϳ 3900 nM) significantly decreased the strength of this interaction by ϳ1.7-2.5 kcal/mol (supplemental Fig. S6). Collectively, these data support a role for an LXXXLL motif and adjacent residues within the coiled coil coactivator TRIP230 in interacting with the ␣-helical face of ARNT PAS-B.

DISCUSSION
Simultaneous recruitment of multiple coactivators by different domains within transcription factor complexes is important for transcriptional activation, whereby interactions of varying stability are important for determining specificity of gene induction and cross-talk with other pathways (44,45). Within the bHLH-PAS family, the use of C-terminal activation domains to recruit coactivators is well documented (13-15), but PAS domains have also been shown to contribute to the specificity of gene induction (22), suggesting that they too interact with coactivators. Recent reports of coiled coil coactivators that target the N-terminal bHLH and/or PAS domains of ARNT (18,20,21) support a role for ARNT in transcriptional activation in the absence of a functional C-TAD (16,17). Here we show that ARNT interacts with the coiled coil coactivators TRIP230 and CoCoA through a single PAS domain, utilizing the same ␣-helical face to interact with both coactivators.
The ARNT PAS-B domain makes critical interactions with its heterodimeric partner HIF-2␣ that are required for the stability and activity of the HIF-2 complex in vivo (8,9), and we demonstrated here that ARNT PAS-B is the primary site of interaction for coiled coil coactivators. Although the ARNT PAS-A domain did not appear to contain a high affinity coactivator binding site, nor did it cooperatively enhance coactivator binding by ARNT PAS-B within a tandem PAS-AB construct, we cannot rule out the possibility that coactivators make weak or transient interactions with ARNT PAS-A. However, our data suggest a model in which the ARNT PAS-B domain is critically important for hypoxia signaling, simultaneously mediating heterodimer formation with HIF-2␣ on its ␤-sheet and using its opposing ␣-helical face to recruit coactivators such as TRIP230 that are required for transcriptional responses (18,20). We tested for formation of the HIF-2␣ PAS-B⅐ARNT PAS-B⅐TRIP230 heterotrimer in our pull-down assays but were unable to detect the complex in vitro, most likely due to the relatively weak affinity of the two isolated PAS-B domains for one another (8,11).
The use of LXXLL-like motifs to mediate transcription factor-coactivator interactions is widespread; these short, linear motifs bind their cognate receptors with low to moderate affinity, ideal for transient signaling interactions (46). Furthermore, disorder to order transitions in LXXLL-like motifs upon binding appears to be another way to achieve specificity with lower affinity (reviewed in Ref. 47). Our TALOS secondary structure analysis indicates that the TRIP230 1583-1620 LXXXLL motif is disordered in the apo state, while the periodicity of chemical shift perturbations suggests that it forms a helix upon interaction with ARNT, supporting the idea that binding to ARNT induces helical structure in the TRIP230 LXXXLL motif. Furthermore, residues proximal to LXXLL-like motifs are often critical for determining the affinity of binding (42,43). Based on our mutation and quantitative binding data, residues C-terminal to the TRIP230 1583-1620 LXXXLL motif are also important for binding to ARNT PAS-B. While the specific ARNT interaction motif of CoCoA was not identified in the present study, it is interesting to speculate whether a motif with similar helical and charge propensities will mediate that interaction as well.
We note remarkable similarities of the ARNT PAS-B⅐ TRIP230 interaction with complex formation of the SRC-1 PAS-B with different LXXLL motifs that use the same PAS ␣-helical interface (23,24). Of particular interest is the unique structure of the SRC-1 PAS-B domain in its co-crystal structure with a STAT-6 peptide (24), an inverse of typical signaling logic where the PAS-B domain of the SRC-1 coactivator interacts with the LXXLL motif of the STAT-6 transcription factor. The ␣-helical face of the SRC-1 PAS-B domain has undergone dra- Residues with weak or overlapping peaks (F) were not included in the analysis. B, helical wheel projection of TRIP230 residues 1583-1600. Residues are notated as the single amino acid code and peptide numbering (beginning with residue 1583), and those with significant ⌬␦ TOT or chemical shift broadening are outlined in black. Coloring scheme: polar/charged, green or purple; polar/uncharged, blue; nonpolar, beige. C, alignment of TRIP230 protein sequences containing the ARNT-interaction motif (underlined). Residues conserved 100% from insect to humans are colored yellow, while highly conserved/similar residues are colored light blue. D, mutation of ARNT-interacting motif in TRIP230. Sequences of site 1-3 mutants and the ⌬N10 fragment (GST-TRIP230 1593-1688 ) are schematically shown. E, pull-down assay of His 6 -tagged ARNT PAS-B with GST-tagged TRIP230 1583-1688 WT and mutant or truncated peptides. matic rearrangement from the three short ␣-helices seen in a prototypical PAS domain into a single long ␣-helix that contacts the LXXLL motif of STAT-6, and is the only known PAS domain structure to have this unique ␣-helical arrangement (24). We currently have no evidence for specific structural rearrangements on ARNT PAS-B as a result of coactivator binding, precluded to a large degree by broadening of PAS-B peaks upon interaction with TRIP230 and CoCoA. Further structural studies will possibly resolve whether coactivator binding by ARNT is coupled to a similar structural rearrangement of its ␣-helical interface.
In conclusion, this study highlights the versatility of small modular domains that mediate protein-protein interactions within larger complexes. ARNT uses its small, 14-kDa PAS-B domain to simultaneously coordinate two sets of protein interactions that are critical for transcriptional responses to hypoxia by using interfaces on opposing sides of the domain. The recruitment of different coactivators by the TADs of HIF-␣ and the PAS-B domain of ARNT likely contributes to the combinatorial nature of transcriptional activation leading to a diverse array of transcriptional responses to a given stimulus. Further regulation of this response by environmental factors could provide additional temporal or tissue-specific regulation of transcriptional responses.