Core L XX LL Motif Sequences in CREB-binding Protein, SRC1, and RIP140 Define Affinity and Selectivity for Steroid and Retinoid Receptors*

An a -helical motif containing the sequence L XX LL is required for the ligand-dependent binding of transcriptional co-activators to nuclear receptors. By using a peptide inhibition assay, we have defined the minimal “core” L XX LL motif as an 8-amino acid sequence spanning positions 2 2 to 1 6 relative to the primary conserved leucine residue. In yeast two-hybrid assays, core L XX LL motif sequences derived from steroid receptor co-activator (SRC1), the 140-kDa receptor interacting protein (RIP140), and CREB-binding protein (CBP) displayed differences in selectivity and affinity for nuclear receptor ligand binding domains. Although core L XX LL motifs from SRC1 and RIP140 mediated strong interactions with steroid and retinoid receptors, three L XX LL motifs present in the global co-activator CBP were found to have very weak affinity for these proteins. Core motifs with high affinity for steroid and retinoid receptors were generally found to contain a hydrophobic residue at position 2 1 relative to the first conserved leucine and a nonhydrophobic residue at position 1 2. Our results indicate that variant residues in L XX LL core motifs influence the affinity and selectivity of co-activa-tors for nuclear receptors.

hydrophobic pocket accommodates the cognate ligand, which upon binding induces a conformational change in the LBD, exposing a co-activator docking site on the LBD surface. Sequence conservation, mutational analyses, and crystal structures have indicated that the co-activator docking site is made up of residues from helices 3, 5, and 12, which form a hydrophobic channel conserved among the NR family members. The integrity of this surface is essential for co-activator binding and as a consequence the transactivation function (AF-2) of the LBD (reviewed in Ref. 2).
We have previously shown that a short ␣-helical sequence, the LXXLL motif, is necessary and sufficient for mediating the interaction of the co-activators RIP140, SRC1 and CBP/p300 with NRs (3). LXXLL motifs have also been shown to mediate binding of other co-activators including TIF1 (4), TIF2/GRIP1 (5,6), p300/CBP-interacting protein/ACTR (7,8), TRAP220/ DRIP205 (9 -11), PPAR␥ co-activator-1 (12), and ASC-2/ RAP250/NRC (13)(14)(15) to NRs. The number and sequence of LXXLL motifs varies considerably among the co-activators and is likely to account for observed differences in binding of coactivators to selected NRs or classes of NRs. The 160-kDa co-activators represented by SRC1, TIF2, and p/CIP each have a nuclear receptor interaction domain (NID) containing three LXXLL motifs. These sequences and the spacing between them are highly conserved, and we and others (3,5,7,16,17) have shown that they mediate high affinity binding to NRs. In addition, the SRC1a isoform has an additional LXXLL motif at its C terminus (16). The 140-kDa protein RIP140 contains nine functional LXXLL motifs that show variable affinity for the LBD of ER␣ (3). We have also shown that CBP (and its paralog p300) contains two conserved LXXLL motifs near the N terminus of the protein, which account for its weak interaction with ER and RAR (3). Some nuclear receptor-binding proteins such as NR-binding SET domain-containing protein contain variant motifs such as FXXLL (18). In addition, a leucine-rich helix (LXX(I/H)IXXX(I/L)) termed the co-repressor of nuclear receptor box mediates the binding of the silencing mediator of the co-repressors for retinoid and thyroid hormone receptors (SMRT) and nuclear receptor co-repressor (NCOR) proteins to the unliganded forms of NRs (19 -21).
The crystal structure of the LBD of PPAR␥ in complex with a portion of the NID of SRC1 revealed that LXXLL motifs adopt an ␣-helical conformation that binds the AF-2 surface of the liganded LBD (22). A similar arrangement was observed in the crystal structures of the liganded TR␤ and ER␣ LBDs complexed with a peptide corresponding to motif 2 of GRIP1 (6,23). The three conserved leucines were found to align on the face of the ␣-helix which packs against the hydrophobic channel on the LBD surface. Two charged residues that are highly conserved among LBDs (a glutamic acid in helix 12 and a lysine residue * This work was supported by the Wellcome Trust, the Imperial Cancer Research Fund, and a Marie Curie fellowship (to D. M. H.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed. Tel.: 44-116-252-3474; Fax: 44-116-252-3369; E-mail dh37@le.ac.uk. 1 The abbreviations used are: LBD, ligand binding domain; SRC1, steroid receptor co-activator; RIP140, receptor interacting protein; CBP, CREB-binding protein; NR, nuclear receptor; AF, activation function; TIF, transcriptional intermediary factor; GRIP1, glucocorticoid receptor interacting protein; ACTR, nuclear receptor co-activator; TRAP220, thyroid receptor-associated protein 220 kDa; DRIP205, vitamin D receptor interacting protein 205 kDa; ASC-2, activating signal co-integrator; RAP250, nuclear receptor activating protein 250 kDa; NRC, nuclear receptor co-activator; NID, nuclear receptor interaction domain; LXM, LXXLL motif; ER, estrogen receptor; RAR, retinoic acid receptor; RXR, retinoid X receptor; TR, thyroid hormone receptor, AR, androgen receptor; PR, progesterone receptor; PPAR, peroxisome proliferator activated receptor; VDR, vitamin D receptor; AAD, acidic activation domain; DBD, DNA binding domain; PCR, polymerase chain reaction; GST, glutathione S-transferase; E2, 17-␤-estradiol; aa, amino acids. in helix 3) and that are especially critical for AF-2 function appear to act as a charge clamp to hold the LXXLL ␣-helix in position (22). In all three structures, both AF-2 surfaces in the LBD homodimer were occupied with one LXXLL ␣-helix. The PPAR␥/SRC1 NID structure indicates that, as a consequence of having multiple LXXLL motifs, a single p160 co-activator protein can contact both AF-2 surfaces in NR dimers. Consistent with this, we and others (5,16,24,25) have found a requirement for at least two functional motifs to support high affinity binding of SRC1 and TIF2/GRIP1 NIDs or full-length proteins to steroid receptors. Similarly, the co-activator TRAP220/ DRIP205, which was isolated on the basis of interaction with class II receptors TR and VDR, contains two functional LXXLL motifs (NR-1 and NR-2), which have been shown to make selective contacts with VDR and RXR in the VDR/RXR heterodimer (10,11). A strong interaction was identified between NR-2 and VDR, whereas a weaker interaction was observed between the NR-1 motif and RXR. In contrast, other putative co-activators appear to contain a single functional LXXLL motif, e.g. TIF1 (4) and ASC-2/RAP250/NRC (13)(14)(15). Thus, the number and sequence of LXXLL core motifs, and possibly flanking sequences, will influence the selectivity and affinity of co-activators for different NRs.
In this study we have defined the minimal or core LXXLL sequence capable of binding to NRs. By comparing the interactions of LXXLL core motifs derived from three different coactivators with a panel of NR LBDs, we have identified elements of the LXXLL core sequence that influence the affinity and selectivity of co-activators for NRs.
Two-hybrid Interaction Assays-The yeast reporter strain used for all two-hybrid assays was W303-1B (HML␣ MAT␣ HMR␣ his3-11, 15 trp1-1 ade2-1 can1-100 leu2-3, 11 ura3) carrying the reporter plasmid pRL⌬21-U3ERE, which contains a lacZ gene driven by a modified URA3 promoter containing three estrogen response elements (28). Transformants containing the desired plasmids were obtained by selection for the appropriate plasmid markers and grown to late log phase in 15 ml of selective medium (yeast nitrogen base containing 1% glucose and appropriate supplements) in the presence or absence of up to 1 M of the appropriate ligand (all-trans-retinoic acid; 9-cis-retinoic acid; 17-␤-estradiol; R5020; mibolerone) or an equivalent volume of vehicle. The preparation of cell-free extracts by the glass bead method and the measurement of ␤-galactosidase activity in the extracts were performed as described previously (28). Two-hybrid experiments were performed three or more times, and reporter activities are expressed as nmol/ min/g protein.
Immunodetection of Proteins-The expression of DBD fusion proteins in yeast cell-free extracts was monitored by immunodetection using a monoclonal antibody recognizing the human ER (a gift from P. Chambon, Strasbourg). The antibody recognizes the "F" region of the LBD in the human ER and also the F region tag at the N termini of the DBD fusion proteins (26). Equal amounts of protein were separated by SDSpolyacrylamide gel electrophoresis and transferred to nitrocellulose for Western blotting. A peroxidase-linked goat anti-mouse IgG was used to detect the primary antibody.

GST Pulldown and Peptide Inhibition Assays-GST fusion proteins
were expressed in Escherichia coli DH5␣ using isopropyl-␤-D-thiogalactopyranoside induction as described previously (3,16). GST-Sepharose beads were loaded with GST alone or GST fusion proteins prepared from bacterial cell-free extracts. 35 S-Labeled mouse ER␣ and SRC1e proteins were generated by coupled in vitro transcription/translation from the plasmids pSP6-MOR (a gift from Roger White) and pSG5 hSRC1e (16) and tested for interaction with GST proteins as described previously (29). Binding was carried out overnight at 4°C with gentle mixing in NETN buffer (50 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 20 mM Tris-HCl, pH 8.0) containing protease inhibitors in the presence or absence of 1 M estradiol (E2) as required in a final volume of 1 ml. Peptides were dissolved in sterile water at a concentration of 4 mg/ml and added to GST-binding reactions at a final concentration of 50 M, immediately before the addition of ligand.

Definition of a Minimal Functional LXXLL Sequence-We
have previously shown that a 14-residue synthetic peptide corresponding to the C-terminal LXXLL motif (motif 4) of SRC1e efficiently competed the ligand-dependent binding of in vitro translated SRC1 proteins to the LBD of ER␣ in GST pulldown experiments (3). This ability to compete the ER␣/ SRC1 interaction was shown to be dependent on the integrity of the LXXLL motif, as a peptide in which two conserved leucines (residues ϩ4 and ϩ5) were replaced with two alanines failed to interfere with SRC1 binding to ER␣. In this study, we have used the peptide competition assay to define the minimal LXXLL sequence required to bind the LBD of ER␣. A series of peptides spanning the high affinity binding motif 4 sequence were tested for their ability to compete the interaction of SRC1e with GST-ER␣ LBD. As shown in Fig. 1, the ligand-dependent recruitment of SRC1e to GST-ER␣ LBD was strongly inhibited by the peptides SLLQQLLTE and SLLQQLLT, respectively. However, omission of the serine (LLQQLLTE) or threonine (SLLQQLL) residues at positions Ϫ2 and ϩ6 (relative to the first conserved leucine residue, designated ϩ1) resulted in a strong reduction in the ability of the resulting peptides to compete SRC1e binding to ER␣ LBD. As expected, peptides in which the LXXLL motif is disrupted or truncated showed no competitive properties in these experiments. Simi- lar results were obtained using the SRC1a isoform (data not shown). These results are consistent with our previous observation that as little as 8 amino acids comprising the LXXLL motif is sufficient to bind to the LBD of ER␣ in yeast twohybrid assays (3) and define the core NR-binding sequence as Ϫ2 to ϩ6.
LXXLL Motifs in SRC1 Display Differential Affinities for the LBDs of Nuclear Receptors-The human SRC1 protein is expressed in a variety of cell lines as two major isoforms (SRC1a and SRC1e) which differ at their C termini (16). Both isoforms contain the nuclear receptor interaction domain (NID) consisting of three LXXLL motifs (motifs 1-3) that are conserved in other p160 co-activators. The SRC1a isoform contains a 4th motif at the extreme C terminus of the protein that is responsible for the ability of this region of SRC1a to interact with nuclear receptors (Figs. 1 and 2 and Refs. 3, 16, and 30).
Our previous results showed that upon addition of ligand, motifs 1-4 of SRC1 bind strongly to the LBD of ER␣ with similar apparent affinities, as determined by yeast two-hybrid experiments (3). In this study, we compared the interactions of SRC1 LXXLL motifs with the LBDs of hER␣, hRAR␣, mRXR␣, hAR, and hPR receptors. As shown in Fig. 2, motif 1 interacted strongly with the LBD of ER␣ as expected but displayed a significantly reduced ability to bind to the LBDs of RAR␣ and AR. In addition, the reporter activation due to interaction of motif 1 with the LBDs of RXR␣ (Fig. 2) and PR (Fig. 5) was consistently reduced in comparison to the other motifs. The motif 2 sequence displayed significant ligand-dependent interaction with all LBDs tested, although a lower affinity for the AR LBD was observed in comparison to other NR LBDs tested (Fig. 2). Motifs 3 and 4 exhibited strong interaction with the panel of LBDs tested in the presence of ligand. However, motif 3 and to a lesser extent motif 4 displayed significant ligandindependent interaction with the LBD of RXR␣. The significance of the ligand-independent interactions with RXR␣ LBD is unclear, as this was only observed in yeast two-hybrid experiments. No significant ligand-independent interaction was observed between GAL-RXR␣ and full-length SRC1, the intact NID or C terminus of SRC1a in transiently transfected mammalian cells (51).
High and Low Affinity LXXLL Motifs in RIP140 -We com-pared the interaction of RIP140 motifs L1 to L9 with the LBDs of ER␣, RAR␣, and RXR␣. As shown in Fig. 3, the L1-L9 core motifs displayed differences in the strength of reporter activation due to binding to ER␣ LBD. In general, similar patterns of interaction were observed for each core motif with the LBDs of ER␣, RAR␣, and RXR␣, with some exceptions as discussed below. By comparison of the reporter activities induced in the presence of ligand, we sought to determine how the sequences of individual core motifs influenced the apparent strength of interactions with NR LBDs in the two-hybrid system. Motifs showing strong NR binding in this assay generally contained a hydrophobic residue at position Ϫ1, usually leucine, valine, or tyrosine. However, the presence of a hydrophobic residue at position ϩ2 (as in motifs L3 and L5) appeared to reduce affinity for NR LBDs. The motif L3 displayed weak interaction with the LBDs of ER␣ and RXR␣ but negligible binding to the RAR␣ LBD. The motif L5, on the other hand, displayed weak interaction with the LBDs of ER␣ and RAR␣ but no detectable binding to RXR␣. Similar selectivity was observed using larger fragments of RIP140 spanning the L3 or L5 motifs, in yeast two-hybrid experiments (data not shown). Low Affinity LXXLL Motifs Mediate the Interaction of CBP with NRs-Several groups have observed ligand-dependent binding of N-terminal fragments of CBP to NRs (31)(32)(33)(34). We compared the ability of in vitro translated SRC1 or CBP fulllength proteins to bind to the LBD of ER␣ in GST pulldown experiments. SRC1e exhibited strong binding to the LBD of ER␣ in the presence of 17-␤-estradiol (Fig. 1). In contrast, full-length CBP showed only a very weak ability to bind the ER LBD in similar experiments (51).
Weak interaction of the N terminus of CBP with RAR␣ has been localized to amino acids 1-101 (31,32). In addition, a second region comprising amino acids 356 -495 has been reported to bind weakly to RXR␣, although this interaction was apparently independent of ligand (32). Analysis of the sequence of CBP revealed it to contain three potential LXXLL motifs (amino acids 68 -78, 355-365, and 2067-2077), which are conserved in p300. Indeed, we have previously reported that the LXXLL motif located between amino acids 68 and 75 displays weak ligand-dependent binding to the LBD of ER␣ in twohybrid assays. However, the strength of binding of this motif to ER␣ (as indicated by reporter activation) was between 10-and 50-fold lower than that achieved by motifs derived from SRC1 or RIP140, despite being expressed at comparable levels (3).
To examine interactions of CBP with NRs in more detail, PCR products spanning the regions 1-101, 345-381, and 1981-2165 (shown schematically in Fig. 4A) were fused in frame with a heterologous DBD. These constructs were used in yeast twohybrid experiments to test for interactions with the LBDs of ER␣, RAR␣, or RXR␣ (Fig. 4, B and D). The CBP N-terminal 1-101 construct displayed potent transcriptional activity in yeast, consistent with reports that the N terminus of CBP functions as a transactivation domain in mammalian cells (35)(36)(37). Reporter activation was slightly enhanced (ϳ1.5-1.8-fold) upon addition of ligand (Fig. 4B). To determine whether the LXXLL motif mediates this weak ligand-induced interaction with NRs, the conserved leucine at residue 72 was mutated to alanine. This resulted in a loss of the weak ligand-induced response (Fig. 4B), supporting the hypothesis that this sequence is required for ligand-induced NR binding. Western blots confirmed that the wild type and mutant fusion proteins were expressed at similar levels (data not shown). The data shown in Fig. 4B were pooled with data from replicated experiments to conduct an analysis of variance test using the SPSS statistics package. The increased reporter activity due to interaction of DBD CBP-(1-101) with VP16-NRs in the presence of ligand was found to be significant (F(1, 11) ϭ 13.007, p Ͻ 0.01). In addition the inability of the DBD CBP-(1-101) (L72A) to increase reporter activity due to interaction with VP16-NRs in response to ligand was also found to be significant (F(1, 11) ϭ 37.309, p Ͻ 0.001).
To confirm this weak interaction in vitro, the CBP-(1-101) sequence was fused in frame with GST and expressed in E. coli for use in pulldown experiments. As shown in Fig. 4C, CBP-(1-101) bound to full-length ER␣ in a ligand-dependent manner. In contrast, the CBP-(1-101) (L72A) mutant failed to bind to ER␣ LBD to a level above the controls. However, it should be noted that the efficiency of binding in the GST pulldown experiments was significantly lower than that achieved with GST-NID of SRC1 (data not shown), indicating that ER␣ binds SRC1 with a higher affinity than to CBP.
The second N-terminal fragment CBP-(345-381) also displayed weak ligand-inducible interaction with the LBDs of ER␣, RAR␣, and RXR␣ in yeast two-hybrid experiments (Fig.  4D). In contrast to CBP-(1-101), no intrinsic reporter activation was observed with this construct. The C-terminal fragment CBP-(1891-2165) displayed a weak interaction with the LBDs of ER␣ and RXR␣ but not RAR␣ in these experiments, enhancing the activity of the reporter 2-3-fold in the presence of ligand (Fig. 4D). GST-pulldown experiments confirmed the weak ligand-dependent interaction of CBP sequences 345-381 and 1891-2165 with full-length ER␣ (Fig. 4E). Note that the amount of ER␣ protein retained on the beads was less than 10% of the input.
We next determined whether the CBP LXXLL core motifs (designated CBP LXM-1-3) could bind the LBDs of steroid and retinoid receptors. We had previously demonstrated that CBP LXM-1 interacted weakly with the LBD of ER␣. As shown in Fig. 5A, CBP LXM-1 and CBP LXM-3 displayed ligand-induced binding to RAR␣ and PR. However, no significant interaction with the LBD of the androgen receptor (AR) was detected (Fig.  5A). Despite several attempts, we were unable to achieve stable expression of the construct DBD-CBP-(356 -366) as determined by Western blots (data not shown). Thus, we were unable to determine whether LXM-2 is sufficient to mediate binding to NRs in the two-hybrid assay. However, the interaction of CBP-(345-381) fragment with the LBDs of ER␣, RAR␣, and RXR␣ in this system (Fig. 4D) and in vitro (Fig. 4E) suggests that LXM-2 also functions as a weak NR-binding motif.
The fold activation of reporter activities achieved with the LXM-1 and LXM-3 core motifs (Fig. 5A) was slightly higher than those obtained due to interaction of LBDs with DBD-CBP-(1-101) (Fig. 4B) and DBD-CBP-(1891-2165) (Fig. 4D). As it has been shown that flanking sequences can influence the affinity of LXXLL motifs for NRs (47,50), this may indicate that core sequences from CBP have a slightly increased affinity for steroid and retinoid receptors compared with larger CBP fragments. Nonetheless, as shown in Fig. 5B, the reporter activity due to interaction of LXXLL motifs derived from SRC1 with the LBD of the PR was 50-100-fold higher than that achieved with CBP motifs. This was not due to lower expression of the CBP fusion proteins, as shown by Western blots (Fig. 4C). Taken together, our results indicate that LXXLL motifs at the N and C termini of CBP and p300 account for the low affinity binding of these proteins to steroid and retinoid receptors.  Fig. 2. DBD-L1 through DBD-L9 represent bait constructs in which the DBD is fused in frame to RIP140 sequences, as indicated in parentheses. Shaded and black columns represent reporter activity in the absence and presence of 500 nM cognate ligand, respectively.

DISCUSSION
In this study we have defined the core LXXLL motif as eight amino acids, requiring two residues N-terminal to the first conserved leucine (ϩ1) and at least one residue C-terminal to the 3rd conserved leucine (ϩ5) (Fig. 6). Consistent with this, it has been previously observed that peptides of 13 and 14 amino acids in length encompassing motif 2 of GRIP1 strongly compete the interaction of the GRIP1 NID with the TR␤ LBD, whereas the hexapeptide LLRYLL did not (6). The crystal structure of the LBD of TR␤ in complex with a synthetic peptide corresponding to motif 2 of GRIP1 showed that the peptide adopts an amphipathic ␣-helix conformation of nearly 3 turns from Lys-688 to Gln-695. Residues outside this sequence were unstructured or in an extended coil conformation. In contrast, the peptide in solution was found to be in a random coil formation, thus ␣-helicity appears to be induced by the interaction with the TR␤ LBD. The ␣-helical structure in motif 2 induced by binding the TR␤ LBD involves amino acids KILHRLLQ and correlates exactly with the boundaries Ϫ2 to ϩ6 that we have defined as the minimum length of a LXXLL motif sufficient for binding to AF-2. In addition, in a phage display screen for peptides that bind to liganded ER␣ LBD, more than 30% of the isolated LXXLL sequences commenced at position Ϫ2. The shortest C-terminal boundary isolated in the same screen ter-minated at ϩ7 (38). Thus, we conclude that the minimal NR binding module is an 8-amino acid ␣-helix.
A number of studies have indicated that co-activators show differential but overlapping binding preferences for NRs. Although there are reports of interactions of fragments of CBP with steroid and retinoid receptors, we and others (51) have found these interactions to be far weaker than the interaction of p160 proteins with steroid or retinoid receptors ( Fig. 1 and  Fig. 5D). Fluorescence resonance energy transfer experiments failed to show association of ER␣ and CBP in vivo, under conditions where the ER␣/SRC1 and PPAR␥/CBP association was readily observed (39). Similarly, others (17, 40 -42) have reported only very weak interactions between steroid, thyroid, and retinoid receptors with CBP. In contrast interactions of CBP and PPARs seem to be more robust, although still considerably weaker than PPAR␥/p160 interactions (39,(43)(44)(45).
We have demonstrated that LXXLL core motifs derived from SRC1a, RIP140, and CBP show distinct preferences for steroid and retinoid receptors. By analysis of the core motif sequences, it is apparent that positions Ϫ1 and ϩ2 (and perhaps ϩ3) have a strong impact on the ability of core motifs to bind to different NR LBDs. We noted that the LXXLL motifs showing the strongest interactions with the panel of LBDs selected tended to have a hydrophobic residue at position Ϫ1 (Fig. 6). For example, motifs 2 and 3 of SRC1/TIF2/ACTR, and motif 4 of SRC1a, all of which display high affinity for ER␣ and other NRs, contain a leucine or isoleucine residue at the Ϫ1 position. In addition, in an affinity selection of randomized peptides that bind the LBD of ER␣, ϳ30% of LXXLL sequences selected contained a hydrophobic residue at Ϫ1 (38). In another study using a "focused" peptide library, 12/17 sequences affinity selected with ER␣ LBD contained a leucine or isoleucine residue at the Ϫ1 position (46). The SRC1 motif 1 has a lysine residue at position Ϫ1. Although this does not appear to affect its interaction with ER␣ in our experiments, motif 1 displayed significantly reduced binding to RAR␣, RXR␣, AR, and PR (Fig.  2). Consistent with this, it has been shown that a peptide based on motif 1 sequence was inferior to other SRC1 LXXLL motif peptides in competition of ER␣/SRC1 interaction in vitro (47).
CBP has been reported to interact weakly with RAR and RXR via the N-terminal 400 amino acids ( Fig. 5; see Refs. 31 and 32). However, it has been shown that deletion of the first 400 residues of CBP did not affect its ability to bind RAR (31,41). This suggested that additional binding site(s) distal to the N terminus are present in CBP or that CBP interaction with NRs can be indirect. CBP contains three LXXLL motifs, and we have demonstrated here that these mediate weak ligand-dependent interactions of the N-and C-terminal fragments of CBP with steroid and retinoid receptors. None of the LXXLL sequences in CBP contain a hydrophobic residue at position Ϫ1. In a recent study of interaction of these sequences with the LBDs of PPAR␥ and RXR␣, the CBP LXM-1 motif was shown to bind to PPAR␥ but not RXR␣ (45). Furthermore, the same authors reported that the replacement of the glutamine at Ϫ1 with leucine significantly increased the ability of CBP LXM-1 to bind RXR␣. This is consistent with our finding that coactivators with high affinity for steroid and retinoid receptors generally contain LXXLL motifs with a hydrophobic residue at position Ϫ1. Thus, while we and others (40,48,51) have shown CBP is required for the function of steroid and retinoid receptors, we propose that at least some NRs recruit CBP indirectly via the p160s.
Although 8 of the 9 nine LXXLL motifs present RIP140 have a hydrophobic residue at position Ϫ1 (the exception being L4), not all demonstrate high affinity binding to the panel of LBDs tested. For example, we noted that motif L5 of RIP140 has very low affinity for NRs, despite satisfying the requirement in having a leucine at position Ϫ1 (Fig. 3). This may be due to the presence of a leucine residue at position ϩ2, which is rare among the LXXLL motifs identified to date. The presence of a bulky hydrophobic side chain at this position may be detrimental to the formation of an amphipathic ␣-helix. However, there are examples of core motifs that tolerate a hydrophobic residue at ϩ2, notably motif 1 of SRC1 which shows reduced affinity for NRs, and the functional LXXLL motif of human ASC-2/NRC/ RAP250, which has a valine at position ϩ2 (13)(14)(15). Further studies will be required to determine the repertoire of amino acids tolerated within core motif sequences.
Structural studies have shown that lysine and glutamic acid residues in the LBD that are almost universally conserved among the NR family may act to clamp to the LXXLL ␣-helix at the AF-2 surface (6,22,23). This involves contacts with residues outside the core motif. In addition, it has been shown that CBP can acetylate a lysine residue just upstream of motif 1 in ACTR and that this modification appears to reduce the binding of ACTR to NRs (49). In addition, reports by several groups (47,50) indicate that sequences flanking the core motif affect the affinity and selectivity of co-activators for NRs. Thus, sequence flanking the core motif may also influence the interaction of co-activators with NRs. Understanding the determinants of LXXLL motif/LBD interactions will facilitate the design of peptides and other antagonists capable of disrupting specific NR/ co-activator interactions.