Identification of Ligand-Selective Peptide Antagonists of the Mineralocorticoid Receptor Using Phage Display

Abstract

The mineralocorticoid receptor (MR) is a member of the nuclear receptor superfamily. Pathological activation of the MR causes cardiac fibrosis and heart failure, but clinical use of MR antagonists is limited by the renal side effect of hyperkalemia. The glucocorticoid cortisol binds the MR with equivalent affinity to that of the mineralocorticoids aldosterone and deoxycorticosterone. In nonepithelial tissues, including the myocardium, which do not express the cortisol-inactivating enzyme 11β hydroxysteroid dehydrogenase 2, cortisol has been implicated in the activation of MR. The mechanisms for ligand- and tissue-specific actions of the MR are undefined. Over the past decade, it has become clear that coregulator proteins are critical for nuclear receptor-mediated gene expression. A subset of these coregulators may confer specificity to MR-mediated responses. To evaluate whether different physiological ligands can induce distinct MR conformations that underlie differential coregulator recruitment and ligand-specific gene regulation, we utilized phage display technology to screen 10⁸ 19mer peptides for their interaction with the MR in the presence of agonist ligands. We identified ligand-selective MR-interacting peptides that acted as potent antagonists of MR-mediated transactivation. This represents a novel mechanism of MR antagonism that may be manipulated in the rational design of a ligand- or tissue-selective MR modulator to treat diseases like heart failure without side effects such as hyperkalemia.

Peptide phage display was used to identify novel MR-selective interacting peptides that act as antagonists in mammalian cells which may facilitate the rational design MR modulators to treat diseases like heart failure.

The mineralocorticoid receptor (MR) is a member of the nuclear receptor (NR) superfamily which also includes the glucocorticoid receptor (GR), androgen receptor (AR), estrogen receptor (ER), and progesterone receptor (PR) (1). It plays a crucial role in sodium and potassium transport in epithelial tissues, such as the kidney and colon (2), and is also present in nonepithelial tissues, including the blood vessel wall, cardiac myocytes, hippocampus, and adipose tissue (3, 4, 5, 6). Activation of the MR in the cardiovascular system promotes cardiac fibrosis, inflammation, and failure (7, 8, 9). Moreover, clinical use of MR antagonists can significantly reduce morbidity and mortality from heart failure as demonstrated by the Randomized Aldactone Evaluation Study (10) and the Eplerenone Post-Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (11). However, the widespread use of MR antagonists is currently limited by their adverse renal effects, specifically hyperkalemia (12). Thus, a selective MR modulator that acts as an antagonist in the heart but an agonist in the kidneys would be of significant therapeutic benefit.

The human MR is made up of 984 amino acids separated into three principal domains that are characteristic of the NR superfamily: the N-terminal domain (NTD), the DNA-binding domain (DBD), and the C-terminal ligand-binding domain (LBD) (13). Ligand binding to the LBD induces a conformational change in the MR, allowing it to dissociate from cytoplasmic chaperone proteins, homodimerize, and translocate to the nucleus. There the MR binds to DNA response elements and recruits coregulators to direct target gene transcription (14, 15). Coregulators consist of coactivators, which enhance gene transcription, and corepressors, which attenuate gene transcription. Many coactivators interact with the activation function (AF)-2 domain of NRs through a canonical LxxLL motif, whereas corepressors bind via an I/L-xx-I/V-I motif (16, 17, 18).

Aldosterone is considered the primary physiological ligand for the MR in humans. Deoxycorticosterone (DOC) is another endogenous mineralocorticoid also capable of activating the MR. However, the glucocorticoid cortisol which circulates at concentrations 100- to 1000-fold higher than those of aldosterone can bind the MR with equal affinity and promote gene transcription (19, 20). In epithelial tissues, prereceptor metabolism of cortisol by 11β-hydroxysteroid dehydrogenase type 2 (11βHSD2) protects the MR from inappropriate activation by cortisol (21, 22). However, 11βHSD2 expression is minimal or absent in other MR-expressing tissues, such as the myocardium and hippocampus, where cortisol has been implicated in the activation of MR (23, 24). Recent studies have shown specific effects of cortisol that are distinct from aldosterone after activation of the cardiac MR, suggesting that additional mechanisms must exist to confer ligand specificity at the MR in tissues where 11βHSD2 is absent (25, 26, 27). To this end, it has been observed that aldosterone dissociates more slowly from the MR and induces a greater transactivation response than cortisol at any concentration (28, 29). Moreover, the interaction between the MR-NTD and the LBD (the N/C interaction) is much stronger in the presence of aldosterone than for cortisol and may serve to stabilize ligand binding (30, 31). At the postreceptor level, an expanding library of over 300 NR coregulators has been identified and is postulated to play a role in determining tissue- and ligand-specificity due to its structural and functional diversity (32, 33).

Many insights into the role of coregulators in NR selectivity initially came from studies of selective ER modulators that demonstrated ligand-selective gene modulation at the ER by way of ligand-specific receptor conformation and therefore recruitment of specific coactivators (34, 35, 36, 37). More recently, similar coregulator-dependent regulation of ligand-selective responses has been demonstrated for the PR and AR (38, 39, 40, 41).

The MR is the least studied member of the steroid hormone receptors. Limited studies have identified only a handful of MR coregulators (reviewed in Ref. 42). Two previous studies have explored ligand selectivity at the MR from the perspective of differential coregulator recruitment (43, 44). Hultman et al. (43) used a mammalian two-hybrid system to analyze 50 coregulator-derived peptides, containing either the LxxLL or I/L-xx-I/V-I motifs, for their interactions with the MR-LBD, whereas Li et al. (44) screened 38 peptides in an alpha screen for competitive binding to ligand-bound MR against an established LxxLL motif. Neither study found a difference in the peptide binding profiles for aldosterone and cortisol (or corticosterone in rodents), suggesting that MR bound to these ligands adopted essentially identical conformations, although Hultman et al. (43) found some differences in binding when mutant MR-LBDs were used. In interpreting these results, one needs to be aware that both of these studies used the MR-LBD as bait rather than full-length MR, which may not reflect the physiologic MR conformation upon ligand binding. This has been shown to be the case for the ER, where the full-length receptor adopted a different conformation upon estrogen binding compared with the LBD alone (45). The use of LBD alone can be misleading specially for NR that posses N-terminal C interaction, given that binding sites on the LBD may be masked by the N terminus as the case for AR. Moreover, these studies only screened peptides containing the common motifs, LxxLL or I/L-xx-I/V-I, which are known to interact with the LBD. A small number of MR-interacting coregulators that do not contain the LxxLL motif have been identified by other investigators to interact with the MR-NTD (46, 47, 48, 49, 50). One of these coregulators, RNA helicase A, was demonstrated to interact with the AF-1 domain of MR in an aldosterone-selective manner (46). It remains to be seen whether other non-LxxLL-containing proteins may be important for ligand-specific interactions with the MR.

Combinatorial peptide phage display is a high throughput method of studying protein-protein interactions using phage particles expressing short peptide fragments and has been used to explore ligand-induced changes in the structure of NRs (51, 52). Previous studies on the ER and AR have confirmed that these short peptides are able to reveal ligand-induced differences in receptor structure and antagonize the interaction between the receptor and its coregulators (52, 53, 54).

We hypothesize that the MR adopts distinct conformations in the presence of different ligands, resulting in differential coregulator recruitment, and applied phage display technology to test this hypothesis. The objectives of our study were to: 1) identify selective MR-interacting peptides that demonstrate ligand specificity, 2) explore the impact of these peptides on MR-mediated transactivation, and 3) characterize these peptides for features that may differentiate ligand-specific binding for the MR. Phage libraries expressing both random and LxxLL-constrained 19mer peptides were used to probe the full-length human MR in the presence of three different agonists, aldosterone, cortisol, and DOC. We identified peptides that interact with the MR in a ligand-selective manner and act as potent antagonists of MR-mediated transactivation. Analysis of the inducible MR-interacting peptides revealed a consensus-binding motif of MPxLxxLL, which may confer preference for MR binding. Our results provide proof of concept that the MR does adopt distinct conformations upon binding by different ligands and demonstrate a valid approach that can be used in further characterization of ligand-specific MR-interacting coregulators.

Results

Selection of high-affinity MR-binding peptides using combinatorial peptide phage display

Combinatorial peptide phage display was used to probe the surface of full-length human MR in the presence of aldosterone, cortisol, or DOC to identify ligan-specific high-affinity MR-binding peptides. Two phage libraries expressing 19mer peptides were used in the screening process. One library was designed to include a central LxxLL motif flanked by seven random amino acids on either side (X₇-LxxLL-X₇) to target the peptide to the AF-2 region of the MR, whereas the other was an entirely random phage peptide library to enable the identification of additional peptides that may bind elsewhere on the MR.

Highly purified full-length biotinylated human MR was produced by baculoviral overexpression in Spodoptera frugiperda (Sf9) cells and was demonstrated to possess similar ligand-binding and coactivator recruitment activities as the wild-type MR (55). Five rounds of panning were performed with each phage library in the presence of aldosterone, cortisol, or DOC. The phage particles that bound to ligand-activated MR were eluted, and their DNA was extracted, amplified by PCR, and sequenced.

A total of 912 peptides was sequenced, from which 82 nonredundant peptides from the LxxLL-constrained phage library and 83 from the random library were identified (Table 1).

Table 1.

Screening and classification of M13 phage peptides

Ligand	LxxLL-constrained peptides					Unconstrained peptides
		Aldosterone	Cortisol	DOC	Aldosterone	Cortisol	DOC
Number of colonies picked for each ligand	193	127	138	185	130	139
Number of distinct peptides within each ligand category	48	27	26	29	25	34
Number of peptides specific to each ligand	42	24	16	27	23	33

Interaction of peptides with the MR in mammalian cells demonstrate ligand selectivity

To confirm a direct interaction between the peptides and the MR, cell-based mammalian two-hybrid assays were performed using CV-1 cells (African green monkey kidney fibroblast cells). The peptide cDNAs were amplified from the phage and shotgun cloned into a pM3.1 vector that enabled each peptide to be expressed as a fusion protein in frame with the Gal4 DBD (pM3.1-peptide). Full-length MR was subcloned as a fusion protein with the VP16 acidic activation domain (pVP16-MR). Interactions between pVP16-MR and pM3.1-peptide were assessed using a luciferase reporter gene containing a Gal4 response element (pG5Luc). MR-VP-16-induced luciferase activity was equivalent to that for pVP-16 and was not increased by any of the hormone treatments (data not shown). Ligand-dependent interaction of the two proteins was arbitrarily defined as a 2-fold or greater increase in luciferase activity after the addition of ligand.

A total of 165 peptides was analyzed by mammalian two-hybrid assays; 66 peptides interacted with the MR in a nonligand-inducible manner. Two representative peptides of this class of peptides, F3 and AL7, as shown in Fig. 1A, were able to bind the MR in the absence of a ligand, and their interactions were not enhanced by the addition of ligands. Eighteen peptides were found to interact with the MR in a ligand-inducible fashion, including 17 from the LxxLL-constrained library and one from the unconstrained library. For all 18 peptides, the interaction with MR was enhanced by the addition of each of the three ligands. The interaction of representative peptides (AL15, AL31, AL33, and AL34) in mammalian two-hybrid assays is shown in Fig. 1B. Their interactions with the MR were uniformly increased by the addition of aldosterone, cortisol, or DOC. Six of these 18 peptides displayed ligand-selective interactions with the MR, including five that were LxxLL constrained and one that did not contain an LxxLL. The five LxxLL-constrained peptides interacted significantly more with the MR in the presence of aldosterone than cortisol or DOC, as demonstrated by the representative peptides (AL27, AL39, and AL45) (Fig. 1C). In contrast, the unconstrained peptide, A25, displayed greater interaction with the MR in the presence of cortisol when compared with aldosterone or DOC (Fig. 1C). The demonstration that different groups of peptides interacted differentially with the MR in the presence of different ligands is consistent with the notion that ligand-induced changes in MR conformation may be sufficient to recruit unique coregulator proteins and enable ligand-specific transcriptional effects.

Fig. 1.

Peptides interact with the MR in CV-1 cells. Mammalian two-hybrid assays were performed to evaluate the ligand-dependent recruitment of M13 phage peptides by the MR. CV-1 cells were transfected with pVP16-MR, and each of the 165 peptides expressed as a fusion protein with the Gal4 DBD (pM3.1-peptide) along with the Gal4-responsive luciferase reporter gene (pG5Luc). The cells were treated with vehicle (white bars) or 10 nm of each ligand: aldosterone (light gray bars), cortisol (dark gray bars), or DOC (black bars). Results are expressed as mean luciferase activity in RLU ± sem (n = 3). A, The empty vectors pM3.1 and pVP16 were used as negative controls and demonstrated that the interactions were peptide and MR dependent. Two of the 66 peptides, F3 and AL7, that interacted with the MR in a ligand-independent manner are shown. B, Four of the 18 peptides that interacted with the MR in a ligand-inducible manner are shown, including AL15, AL31, AL33, and AL34. *, P < 0.05, compared with vehicle-treated samples. C, Six of the 18 ligand-dependent MR-interacting peptides displayed ligand selectivity, with four representatives, AL27, AL39, AL45, and A25, shown here. *, P < 0.05, compared with aldosterone-treated samples; #, P < 0.05, compared with cortisol-treated samples.

Identification of an extended consensus binding motif for the MR

Of the 17 LxxLL-constrained peptides that interacted with the MR in a ligand-dependent fashion, approximately 50% contained a unique motif, MPxLxxLL (Table 2). This motif was not present in peptides that interact with the MR in a ligand-independent manner (data not shown). Studies with other steroid receptors have shown that residues flanking the central LxxLL motif may confer receptor specificity and modulate the receptor’s interaction with various coregulators (53, 56, 57). The motif, MPxLxxLL, has not to the best of our knowledge been previously described and may represent a MR-preferred peptide binding sequence.

Table 2.

Identification of an extended consensus binding motif for the MR

MP

AL2 FPYPMPTLRALLESDALSL

AL27 DFGPMPLLRSLLEENIGTF

AL4 PEDAMPLLAMLLSDAGAGT

AL8 AEEAMPLLRKLLIEETSGW

AL36 TEPTMPLLRMLLMAPMEDH

AL34 LEERMPLLSGLLTGTYLTG

AL39 GEMRMPILTGLLTSHPYQE

AL29 ECTNMPRLCKLLGGDEMEM

Consensus … . MPlL..LL… . . . .

All

AL15 VPEPMSMLRALLSNDDFSG

AL2 FPYPMPTLRALLESDALSL

AL43 VPADGSMLRYLLSEPEAAM

AL33 PPPEQSILHRLLTADVSDL

AL27 DFGPMPLLRSLLEENIGTF

AL16 YDGPLPILARLLRDAPMEI

AL45 LDHQFPLLTQLLRSYDAGL

AL31 LSETHPLLWTLLSSEGDSM

AL4 PEDAMPLLAMLLSDAGAGT

AL8 AEEAMPLLRKLLIEETSGW

AL36 TEPTMPLLRMLLMAPMEDH

AL21 LHDRPSILSALLTDPPRTE

AL34 LEERMPLLSGLLTGTYLTG

AL39 GEMRMPILTGLLTSHPYQE

AL14 QEVHSPILRGLLLDMQYEW

AL29 ECTNMPRLCKLLGGDEMEM

AL19 QEFSAGMLEWLLTHDEPPV

A25 SCDNSYCNIRSWFSDRVIS

Consensus … . mp.L..LL… . . . .

Bold Face, are amino acid abbreviations: L, Leucine; M, methionine; P, proline.

MR-interacting peptides are potent antagonists of MR-mediated transactivation

The interaction between the MR and the peptides is presumed to resemble the interaction between the MR and endogenous coregulators. Therefore, the coexpression of the peptides in a transactivation assay should competitively block coregulator binding and suppress MR-mediated transcriptional activity. To determine whether the 18 ligand-inducible, MR-interacting peptides could antagonize MR-mediated transcriptional activity, they were introduced into a MR transactivation interference assay together with a MR expression plasmid (pRShMR) and a MR-responsive mouse mammary tumor virus (MMTV) promoter fused to a luciferase reporter gene [MMTV luciferase (MMTV-luc)]. All of the LxxLL-containing peptides were capable of inhibiting MR-mediated transactivation in a dose-dependent manner (Fig. 2, A and B, and data not shown). Some peptides, such as AL4, were very potent in their repression of MR-mediated activity. In contrast, the only non-LxxLL-containing MR-interacting peptide, A25, achieved minimal inhibition of MR transactivation when compared with the empty vector pM3.1 (Fig. 2, C and D). We confirmed that peptides that did not bind the MR, or that bound the MR in a ligand-independent manner, did not exhibit antagonist activity against the MR (Supplemental Figs. 1 and 2, published on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). The ability of a peptide to bind the MR in a ligand-selective manner did not translate into a ligand-specific inhibition of MR-mediated transactivation, because all of the LxxLL-containing peptides repressed MR activity equally in the presence of the three ligands (Fig. 2, A and B, and data not shown). This was independent of the concentration of hormone used in the assay (Supplemental Fig. 3). Nevertheless, these peptides represent novel antagonists of the MR that function by competing with coactivator binding, thereby inhibiting transcription of MR-responsive reporter genes.

Fig. 2.

LxxLL-containing MR-interacting peptides inhibit MR-mediated transactivation via the MMTV promoter. CV-1 cells were transfected with the pRShMR and MMTV-luc along with all 18 MR-interacting peptides (pM3.1-peptide) to examine the effect of the peptides on MR-mediated transactivation. The cells were treated with 10 nm of aldosterone (light gray bars), cortisol (dark gray bars), or DOC (black bars). Results are expressed as mean luciferase activity in RLU ± sem (n = 3), with increasing amounts of the peptides or empty vector (pM3.1) indicated on the x-axis. A and B, Two representative LxxLL-containing peptides, AL4 and AL45, both disrupted MR-mediated transactivation in a dose-dependent manner in the presence of all three ligands. **, P < 0.01; ***, P < 0.0001, compared with MR-mediated transactivation response in the absence of the peptides. C and D, Peptide A25, which does not contain the LxxLL motif, did not produce a significant effect on MR-mediated transactivation when compared with increasing doses of the empty vector pM3.1, irrespective of the ligand present.

Specificity of the peptides for the MR

To evaluate the specificity of these peptides for the MR vs. other NRs, we tested their interactions with the GR, ERα, ERβ, AR, PR-A, and PR-B in the presence of their respective agonist ligands; 14 of the peptides interacted with a range of receptors, in a ligand-dependent manner, with one representative peptide, AL8, as shown in Fig. 3A. Two of the peptides interacted only with ligand-bound GR and MR, as illustrated by the representative peptide, AL4, in Fig. 3B. Two other peptides, AL39 and A25, displayed high specificity for the MR in the presence of aldosterone and cortisol, respectively (Fig. 3, C and D).

Fig. 3.

Receptor specificity of MR-interacting peptides. The interactions between the ligand-dependent MR-interacting peptides and other NRs were studied using mammalian two-hybrid assays. CV-1 cells were transfected with one of the 18 peptides and pVP16-MR, GR, ERα, ERβ, AR, PR-A, and PR-B and treated with vehicle (white bars) or ligand (shaded bars). The ligands used were: aldosterone 10 nm, cortisol 100 nm, estradiol 100 nm, testosterone 100 nm, or progesterone 100 nm according to the type of receptor. Results are expressed as mean luciferase activity in RLU ± sem (n = 3). A, Fourteen of the 18 peptides were nonspecific binders and interacted with a wide range of NRs upon ligand binding. One representative peptide, AL8, is shown here. B, Two of the 18 peptides interacted most strongly with the MR and GR, with one representative, AL4, shown here. C, Peptide AL39 interacts almost exclusively with the MR in a ligand-dependent manner. D, Peptide A25 also interacts almost exclusively with the MR in the presence of aldosterone (10 nm) and cortisol (10 nm). Only data for cortisol-treated MR are shown here.

Further characterization of MR-interacting peptides

In view of the binding specificity of AL39 and A25 for the MR, their interactions with the MR were further characterized in mammalian two-hybrid assays. Both peptides were transfected into CV-1 cells, together with the MR, and treated with either MR agonists (aldosterone, cortisol, or DOC) or antagonists (eplerenone, progesterone, or spironolactone). AL39 interacted with the MR even in the presence of antagonists (Fig. 4A), whereas A25 did not interact at all with antagonist-bound MR (Fig. 4B). Of note, the antagonist-induced interaction between AL39 and MR was abolished when the isolated MR-LBD was used (Fig. 4C). This emphasizes the importance of using full-length MR in screening for peptide interactions, because the tertiary structure presumably contributes to both receptor and ligand-binding stability.

Fig. 4.

MR-specific peptides differ in their interaction with the MR in the presence of agonists and antagonists. Two MR-specific peptides, AL39 and A25, were further characterized in mammalian two-hybrid assays for their interactions with the full-length MR (A and B, respectively) in the presence of vehicle (white bars), agonists, or antagonists. AL39 was also tested for an interaction with the MR-LBD (C). The agonist ligands used were aldosterone (light gray bars), cortisol (dark gray bars), or DOC (black bars), 10 nm each. The antagonist ligands used were eplerenone 1000 nm (horizontally striped bars), progesterone 100 nm (vertically striped bars), or spironolactone 100 nm (diagonally striped bars). Bars represent mean luciferase activity in RLU ± sem (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.0001; compared with samples treated with vehicle.

Peptide A25 is interesting, because it does not contain an LxxLL motif, which typically binds to the AF-2 of NRs, and is the only peptide that binds most avidly to the MR in the presence of cortisol. It was tested in mammalian two-hybrid assays against a range of MR fragments to ascertain the location of its interaction. As before, its interaction with the full-length MR was most significant in the presence of cortisol (Fig. 5A). Surprisingly, it interacted only weakly with the MR-LBD and did not interact with the MR-NTD (Fig. 5A). Given its lack of the LxxLL motif, we anticipated that it would bind to the NTD. The weak interaction between A25 and the MR-LBD was abolished when the wild-type LBD was replaced by a mutant (E962A) with a defective AF-2 domain (Fig. 5B). A25 was also tested for interaction with the MR-DBD and MR-DBD+LBD. It interacted weakly with the MR-DBD+LBD but not with the MR-DBD alone (Fig. 5C). The results confirm the importance of using full-length MR in screening for peptide interactions. Peptide A25 interacted more strongly with the full-length MR than the NTD, DBD, or LBD alone, with a response that far exceeds an additive effect.

Fig. 5.

Peptide A25 requires the full-length MR for maximal interaction. Mammalian two-hybrid analysis was performed to evaluate the interaction between the non-LxxLL containing peptide, A25, and the full-length MR (A and C), MR-LBD (A and B), MR-NTD (A), MR-LBD mutant E962A (B), MR-DBD+LBD (C), and MR-DBD (C), in the presence of vehicle (white bars) or 10 nm of aldosterone (light gray bars), cortisol (dark gray bars), or DOC (black bars). Bars represent mean luciferase activity in RLU ± sem (n = 3). *, P < 0.05; **, P < 0.01; ***, P < 0.0001; compared with samples treated with vehicle.

Discussion

It is now widely accepted that the overall conformation of many NRs is determined by the type of ligand and their subsequent interaction with distinct subsets of coregulators (58). This concept has not been clearly established for the MR despite an urgent need to understand the molecular signaling pathways of the MR in different organ systems in response to different ligands to achieve tissue- and ligand-specific MR modulation that may offer superior heart failure treatment with reduced adverse effects. Several studies have suggested that differential agonist-dependent coregulator recruitment may underlie agonist-selective gene regulation by the MR (43, 46). However, definitive evidence to support this hypothesis is lacking. In the current study, we have provided proof of principle that the MR adopts distinct conformations in the presence of different agonist ligands by isolating a number of MR-interacting peptides, which showed selectivity for different ligands.

We used phage display technology to study the ligand-dependent structural changes in the MR and identified a group of ligand-inducible MR-interacting peptides. The demonstration that several LxxLL-containing peptides bind the MR preferentially in the presence of aldosterone, and one non-LxxLL containing peptide, A25, binds MR preferentially in the presence of cortisol, suggests that aldosterone and cortisol induce unique conformational ting as agonists. Hall et al. (59) have also previously demonstrated hormone-specific patterns of peptide binding efficacy with ERβ that was supported by crystallography studies showing ligand-specific ERβ conformations. Based on transcriptional interference assays, we have also shown that all of the LxxLL-containing ligand-inducible peptides were able to inhibit MR-mediated transactivation in a dose-dependent manner. This is presumably due to their ability to compete with the binding of endogenous coactivators that contain the LxxLL motif. Indeed, other studies have also shown the utility of such short peptides as NR antagonists by targeting the coregulator binding interface, with one study showing the ability of such a peptide to inhibit the transcription of endogenous peroxisome proliferator activated receptor target genes (56). However, we found discordance between the ligand-selective interactions of these peptides with the MR in mammalian two-hybrid assays and their ability to block MR-mediated transactivation indiscriminate of ligand in transcriptional interference assays. It is possible that MR-mediated transactivation requires complex protein-protein interaction surfaces that are not fully characterized by these short peptide probes, such that subtle differences in ligand-induced conformations are missed. In particular, these peptides would only be able to compete with coregulator binding in the AF-2 region of the MR where the LxxLL motif would normally bind. Because many different coregulators share the same LxxLL motif, our short peptides would not be expected to discriminate between them despite their other regional differences. Therefore, our results do not exclude the possibility that the differential recruitment of LxxLL-containing coregulators by the MR occurs in the presence of different agonist ligands. The unconstrained peptide, A25, which interacted strongly with the MR after the addition of cortisol, was not able to significantly inhibit MR-mediated transactivation even in the presence of cortisol. This observation suggests alternative binding surfaces on the MR that may be important for protein-protein interaction, but not necessarily transactivation. The discordance between ligand specificity in mammalian two-hybrid assays and transactivation interference assays has been noted previously in a study of the ER (35). In this study, one peptide was able to interact with both estradiol- and tamoxifen-bound receptors but only able to inhibit the transcriptional activity of tamoxifen, whereas another peptide interacted with PR-B when bound by RU486 but was unable to block the partial agonist activity mediated by PR-B/RU486. These results mirror our findings and suggest that receptor binding does not necessarily correlate with the ability to block receptor-mediated transcription. These peptides do, however, still represent novel antagonists of the MR, because they are able to competitively inhibit coactivator binding and suppress MR-mediated transcription of reporter genes.

We found that the MR has a preference for peptides with a consensus motif MPxLxxLL. Other papers have shown the importance of amino acids in the −2 and −3 positions with respect to the LxxLL motif in receptor selectivity. The presence of a proline at the −2 position has been described for a few classes of LxxLL peptides (53). For example, PPARγ1 prefers HPLLxxLL, whereas liver receptor homologue-1 prefers PILxxLL (56, 57). MPxLxxLL has not been described before for other NRs and is not found in the sequences of naturally occurring coregulators known to interact with the MR. This motif was used to search a nonredundant protein database in the Basic Local Alignment Search Tool and extracted a large number of proteins with the potential to be coregulators, which warrant exploration in future studies. One particular candidate, gem (nuclear organelle)-associated protein 4, was selected for characterization, because it contained three LxxLL motifs, including one MPxLxxLL (MPLLAMLL), and is known to participate in pre-mRNA splicing in the nucleus as part of a complex with gem (nuclear organelle)-associated protein 3 (60). We were not able to demonstrate a direct interaction between Gemin4 and the MR in mammalian two-hybrid assays or transactivation assays (data not shown), possibly due to the absence of other components of the gemin complex. Further studies are required to characterize the biological importance of this motif.

Short LxxLL-containing peptides are often able to bind to a range of NRs, because LxxLL is the preferred motif for binding to the AF-2 region. However, we were able to identify one peptide, AL39, which was selective for the MR over other steroid receptors. It again highlights the importance of flanking residues of the central LxxLL motif in determining receptor selectivity. Interestingly, peptide A25, which does not contain an LxxLL, also displayed MR selectivity. In characterizing these two unique peptides, we found that they interacted with full-length MR differentially in the presence of MR antagonists, including spironolactone, eplerenone, and progesterone. AL39 was able to bind MR, although to a lesser extent, in the presence of antagonists, whereas A25 could not. The differential interaction of AL39 and A25 with the MR in the presence of antagonists suggest MR conformational differences that may contribute to the agonist selectivity of these two peptides. The ability of an LxxLL-containing peptide to interact with the MR in the presence of an antagonist is quite remarkable, because it is normally only recruited to an agonist-activated receptor. Chang et al. (53) found that none of their ER antagonists or selective ER modulators tested was able to facilitate full-length ERα-LxxLL interactions, except for one novel antiestrogen (GW5638) that functioned as an ER agonist in bone and in the cardiovascular system (61). Similarly, Hall et al. (59) did not find an interaction between their LxxLL peptides and ERβ in the presence of antagonists. Our data suggest that antagonists may also be capable of acting as agonists and recruit coactivators depending on the cell type and the presence of particular coregulators.

We were surprised that screening the unconstrained phage peptide library did not isolate more ligand-dependent MR-interacting non-LxxLL-containing peptides. This potentially reflects the various affinities of peptides binding to the MR and is consistent with phage display being biased toward selecting high-affinity interactions that presumably involves the LxxLL motif and the AF-2 region. One peptide, A25, did not possess the LxxLL motif, which strongly argues that its interaction is not with the AF-2 domain. We sought an interaction with the MR-NTD, LBD, DBD, and DBD+LBD. No interaction was observed with the DBD or NTD, and only a weak interaction was observed with the LBD and DBD+LBD. The E962A mutation in helix 12 of the MR also abolished the interaction between A25 and the MR-LBD, which is intriguing, because this mutation is known to prevent the binding of LxxLL-containing coactivators to AF-2 (31). This data demonstrates that the full-length receptor and its tertiary structure with an intact AF-2 are pivotal to the binding of peptides, even for non-LxxLL-containing peptides like A25. As for the LxxLL-containing peptides, such as AL39, the full-length receptor rather than the LBD alone is also crucial to the study of its binding characteristics, because the ability of AL39 to bind in the presence of antagonists was abolished when the MR-LBD was substituted for full-length MR. There are probably critical interdomain interactions, such as the N/C interaction, which has been described as being important for receptor stability (30, 31), to explain the dramatic differences between the interactions of these peptides and various fragments of the MR vs. full-length MR.

Overall, our study has demonstrated that by screening vast phage peptide libraries using the full-length MR, it was possible to identify interacting peptides with ligand selectivity, in contrast to previous studies screening known LxxLL motifs with the MR-LBD. Furthermore, although an LxxLL motif appeared necessary for competitive inhibition of MR-mediated transactivation, it was not obligatory for interaction with the MR. The finding of a relatively cortisol-selective peptide that does not contain an LxxLL motif is novel and may be useful in studying glucocorticoid-mediated signaling at the MR. The finding of a unique consensus binding motif MPxLxxLL is also novel and may be of use in the search for more MR-interacting coregulator proteins. At present, the MR antagonists in clinical use, namely spironolactone and eplerenone, function by competing with aldosterone for binding to the MR ligand binding pocket and destabilizing the active conformation of the receptor (62, 63). They are neither tissue nor ligand specific, and spironolactone is not even receptor specific, resulting in undesirable side effects. It would be highly desirable for a drug to directly target the coactivator binding pocket of the MR and alter the recruitment of tissue- and ligand-specific coregulators. We have identified potent peptide antagonists of the MR via interference with coactivator binding, although more work is required to identify peptides that are tissue and ligand specific as the first step in the rational development of a selective MR modulator for the treatment of diseases like heart failure.

These peptides could also be used to develop a screen for ligands that dissociates the interactions with MR AF2 function -LXXLL peptides motifs or to identified ligands that could abolish the interaction with those peptides that interact with unliganded receptor.

Materials and Methods

Plasmids

For mammalian two-hybrid assays, pVP16-MR and pM3.1-peptide constructs were used. pVP16-MR was constructed using a two-step approach. The first 1139 residues of the MR were PCR amplified from an expression vector containing full-length MR cDNA (PRshMR, kindly provided by Ron Evans; Salk Institute, La Jolla, CA) using custom designed primers MR forward 1EcoRI (GATCGAATTCATGGAGACCAAAGGCTAC; Sigma, St. Louis, MO) and MRR1BamHI (GATCGGATCCAGCAGAGGTGCCAGAA; Sigma). The PCR product, MR1139, was excised with BamHI and EcoRI and subcloned into pVP16 (Clontech, Mountain View, CA) to produce pVP16-MR1139. The remaining 1913 residues of the MR were PCR amplified using another set of custom designed primers MRF2BamHI (GATCGGATCCAGTACATTGCGGGA; Sigma) and MRR2HindIII (GATCAAGCTTTCACTTCCGGTGGAAGTAGAG; Sigma). The second PCR product was then excised using BamHI and HindIII and subcloned into pVP16-MR1139 digested with the same restriction enzymes to produce pVP16-MR. All other pVP16-NR constructs, including pVP16-GR, pVP16-ERα and pVP16-ERβ, pVP16-AR, and pVP16-PRA and pVP16-PRB, have been described previously (64). The construction of pM3.1-peptide fusions has also been described previously (53). Among the various MR fragments, pVP16-MR N-terminal domain and pVP16-MR ligand binding domain, as well as pM-MRLBD-E962A, were obtained as described (31). pVP16-MRDBD was constructed by PCR amplification of a 520-base pair region surrounding the DBD from PRshMR using custom designed primers MRDBDF2_Mlu1 (GATCACGCGTTATTGTTGGGGTG; Sigma) and MRDBDR2_Hind3 (GATCAAGCTTGAAGGTGTGAGCG; Sigma). The PCR product was excised with MluI and HindIII and subcloned into pVP16. pVP16-MRDBD+LBD was constructed by PCR amplification of the DBD and the entire LBD, spanning 1220 base pairs, from PRshMR using the primers MRDBDF2_Mlu1 and MRR2HindIII. The PCR product was excised with MluI and HindIII and subcloned into pVP16. pVP16-MRLBD-E962A was constructed by excising the mutant MRLBD-E962A from its pM vector with SalI and HindIII and subcloning the fragment into pVP16 digested with the same restriction enzymes. All PCR products were sequenced to ensure the fidelity of the resulting products. For MR transactivation assays, full length MR in an expression vector (pRShMR) and MMTV-luc were used.

Production and purification of recombinant full-length human MR and M13 phage panning

Full-length human MR was prepared as previously described (55). The M13 phage display protocol has been described in detail previously (51). Some modifications were introduced for studying the MR; 2 μg of purified baculovirus-expressed, biotinylated MR were added to each well of a 96-well plate in the presence of 100 mm NaHCO₃ to a total volume of 100 μl; 1.0 μm aldosterone, cortisol, or DOC was added to the wells containing MR purified in the presence of each respective ligand. The MR was allowed to bind to plastic for 2 h at room temperature, after which the protein-coated wells were blocked with 2% milk/PBS for 1 h at room temperature. The wells were washed five times with PBS + 0.1% Tween 20 (PBST). Approximately 10⁸ plaque-forming units of phage libraries, either LxxLL constrained or unconstrained, were precleared in 2% milk in PBST, then added to each well and incubated with the MR for 3 h at room temperature. The wells were washed five times with PBST and bound phage was eluted with 100 μl of 0.1 m HCl. The eluate was neutralized with 50 μl of 1 m Tris-HCl (pH 7.4), and the eluted phage was amplified in DH5αF′ cells (Invitrogen, Carlsbad, CA) for 5 h in a shaking incubator at 37 C. Amplified phage were recovered from the bacterial supernatant and used for the subsequent round of panning. Enrichment for MR-binding phage over five successive rounds of panning was demonstrated using a phage ELISA as described previously (51).

Amplification and sequencing of phage isolated from phage display

PCR using mBAX primers (mBAX forward, ATTCACCTCGAAAGCAAGCTG; and mBAX reverse, ACCCTCATAGTTAGCGTAACG) and phage stock from panning rounds 3–5, which showed significant enrichment of MR-binding phage, allowed amplification of the peptide inserts isolated from those rounds. The PCR products were digested with XhoI and XbaI and run on a nondenaturing polyacrylamide gel to allow separation and retrieval of the peptide fragments before subcloning into the expression vector pM3.1 for use in mammalian two-hybrid analysis. Templiphi 100 Amplification kit (GE Healthcare, Princeton, NJ) was used according to manufacturer’s instructions to amplify and sequence the peptide fragments within the pM3.1 vector.

Mammalian cell culture, transfection, and reporter assays

CV-1 cells (African green monkey kidney fibroblast cells) were maintained in DMEM (Invitrogen) supplemented with 10% fetal bovine serum, 1 mm nonessential amino acid, 1 mm l-glutamate, and 1 mm penicillin (10 U/liter) in a humidified 37 C incubator with 5 CO₂. The cells were trypsinized and seeded at a density of 5 × 10⁴ cells per well in 24-well plates 1 d before transfection. Transfections were performed using FuGENE 6 (Roche Molecular Biochemicals, Indianapolis, IN) in triplicate for each datapoint following manufacturer’s protocol, and each experiment was repeated two to three times. Media were changed to DMEM with 10% charcoal-stripped fetal bovine serum. After 24 h, media were refreshed again using DMEM with 10% charcoal-stripped fetal bovine serum, and appropriate hormones were added. Cells were harvested 24 h after hormone addition, and luciferase assays were performed. Results are expressed as luciferase activity in relative light units (RLU) ± sem per triplicate sample of cells.

For mammalian two-hybrid assays, the following plasmids (and quantities per well) were transfected: pVP16-NR constructs (100 ng), pM3.1-peptide (100 ng), and pG5-luciferase (300 ng). For transactivation assays, the following plasmids (and quantities per well) were used: MMTV-luc (500 ng), PRshMR (200 ng), pM3.1-peptide (0–400 ng), and pBlueScript (to adjust for varying amounts of peptide).

Statistical analysis

Significance was calculated with one-way ANOVA followed by Tukey’s multiple comparison test with GraphPad Prism version 5.02 for Windows (GraphPad Software, San Diego, CA).