Monitoring ligand-mediated nuclear receptor–coregulator interactions by noncovalent mass spectrometry Sarah Sanglier 1 , William Bourguet 2, *, Pierre Germain 2 , Virginie Chavant 2 , Dino Moras 2 , Hinrich Gronemeyer 2 , Noelle Potier 1 and Alain Van Dorsselaer 1 1 Laboratoire de Spectrome ´ trie de Masse Bio-Organique, CNRS UMR 7509, ECPM, Strasbourg, France; 2 Institut de Ge ´ ne ´ tique et de Biologie Mole ´ culaire et Cellulaire, CNRS INSERM ULP Colle ` ge de France, Illkirch, France Retinoid receptors are ligand-dependent transcription factors belonging to the nuclear receptor s uperfamily. Retinoic acid (RARa, b, c) and retinoid X (RXRa, b, c) receptors mediate the retinoid/rexinoid signal to the transcriptional machineries by interacting at the first level with coactivators or c orepressors, which leads to the recruitment of enzymatically active noncovalent c om- plexes at target gene promoters. It has been shown that the interaction of corepressors with nuclear receptors in- volves conserved LXXI/HIXXXI/L consensus sequences termed corepressor nuclear receptor (CoRNR) boxes. Here we describe the use of nondenaturing electrospray ionization mass spectrometry (ESI-MS) to determine the characteristics o f CoRNR box peptide binding to the ligand binding domains of the R ARa–RXRa hetero- dimer. The stability of the RARa–RXRa–CoRNR 1 tern- ary complexes was monitored in the presence of different types of a gonists or a ntagonists for the two receptors, including inverse agonists. These r esults show unambig- uously the differential impact of distinct retinoids on corepressor binding. W e show that ESI-MS is a p owerful technique that complements classical methods and allows one to: (a) obtain d irect evidence for the formation o f noncovalent NR complexes; (b) determine ligand binding stoichiometries and (c) monitor ligand effects on these complexes. Keywords: ESI-MS; noncovalent mass spectrometry; nuc- lear receptors; protein–protein interactions; p rotein–ligand interactions. Retinoid receptors are ligand-dependent transcription f ac- tors belonging to the nuclear receptor (NR) superfamily [1]. They are involved in vertebrate development, cell differen- tiation and proliferation o r apoptosis [2], and display significant promise for cancer therapy and prevention [3,4]. Retinoid X (R XRa, b, c) and retinoic acid (RARa, b, c) receptors form heterodimers, but RXRs can also hetero- dimerize with a great number of other nuclear receptors [5]. NRs display a modular structure with distinct functional and structural domains [1], such as the DNA-binding domain [6] or the ligand binding domain (LBD) [7,8], which corresponds also to the region that interacts with coactiva- tors (CoA) or corepressors (CoR). Although the molecular mechanisms by which NRs control target gene expression are in principle understood, many aspects remain to be solved. This concerns both the details of the ligand-induced early events [9], and the assembly and epigenetic action of multiprotein machineries on target gene promoters [10]. As NRs r espond to small ligand molecules and correspond to potent regulators of cell function [11,12], life and death [13,14], they are particularly attractive targets for the design of novel therapeutic agents. A c ritical step in the drug design process is t he elucidation of the ligand function, as i t c an be deduced from the characterization of the molecular mechanism(s) that are modulated by ligand binding, such as the interaction with coactivator and corepressor complexes that mediate the transcriptional activity o f N Rs. In the case of retinoid receptors, th e unliganded RAR–RXR heterodimers (HD) have been shown to act as repressors of transcription through recruitment of complexes containing corepressors, such as nuclear receptor corepressor [ 15] or silencing mediator for RARs a nd th yroid hormone receptor (TRs) 2 [16]. Upon ligand binding, NRs undergo conformational changes inducing the dissociation of repressor complexes and the sub sequent association w ith coac tivator molecules thereby promoting transcriptional activation. CoAs are like CoRs; large multidomain proteins, which bind t hrough specific motifs (referred to as NR box or LXXLL motif) to Correspondence to N. Potier, Laboratoire de Spectrome ´ trie de Masse Bio-Organique, UMR CNRS 7509, ECPM, Universite ´ Louis Pasteur de Strasbourg, 25 Rue Becquerel, 67082 Strasbourg Cedex 2, F rance. Tel.: +33 3 90 24 26 79, Fax: +33 3 90 24 27 81, E-mail: npotier@chimie.u-strasbg.fr Abbreviations: ESI-MS, electrospray ionization mass spectrometry; RAR, human retinoic acid receptor a; RXR, human retinoid X receptor a; CoR, corepressor; Co A , coactivator; NR, nuclear receptor; NcoR, nuclear corepressor p rotein; CoRNR, c orepressor nuclear receptor; LBD, ligand binding domain; LBP, ligand binding pocket; Vc, accelerating voltage; HD, RAR–RXR heterodimer; TR, thyroid hormone receptor. *Present address: Centre de Biochimie Structurale , CN RS UM R 5 048, UM1 INSERM UMR 554, Faculte ´ de Pharmacie, 15 Avenue Flahault, 34060 Montpellier, France. (Received 17 August 2004, revised 12 October 2004, accepted 27 October 2004) Eur. J. Biochem. 271, 4958–4967 (2004) Ó FEBS 2004 doi:10.1111/j.1432-1033.2004.04466.x agonist-bound NRs. CoRs interact with NRs th rough nuclear receptor interacting domains that contain two helical motifs referred to as corepressor nuclear receptor boxes (CoRNR1 and CoRNR2, comprising LXXI/ HIXXXI/L sequences) [17–19]. Since 1 991 [13], electrospray ionization mass spectro- metry (ESI-MS) has been regularly used to study noncova- lent complexes, and an increasing number of studies of protein–protein, protein–ligand and protein–DNA com- plexes are now reported (reviewed in [20–22]). However, such studies are not yet straightforward because some technical difficulties, such as the presence of nonvolatile buffers or the fragility of the complexes, are encountered. Nevertheless, ESI-MS appears to be an attractive approach and offers new possibilities for the study of such complexes, providing direct e vidence for their f ormation and an accurate determination of their binding stoichiometry. For instance, in the field of NRs, ESI-MS was previously used by Witkowska et al. [23,24] to detect the estrogen rec eptor ligand binding domain in its dimeric noncovalently bound form. Craig et al. [25] analyzed by microESI-MS the binding of the human vitamin D receptor and the retinoid X r eceptor-a to the osteopontin vitamin D response element, and assessed the influence of their endogenous ligands 1a,25-dihydroxyvitamin D3 and 9- cis retinoic acid. The importance of s tructural z inc ions for the interaction between the DNA binding domain of the human vitamin D receptor with a double-stranded DNA containing the vitamin D response element from the m ouse osteopontin gene was determined using ESI-MS [25,26]. Bourguet et al. [27] used ESI-MS to reveal the presence of an unexpected ligandinthemRXRaF318A complex that was apparent at the early stages of the structure refinement and Germain et al. [28] demonstrated by ESI-MS that the RXR subunit of the HD heterodimer 3 can bind its cognate ligand even when RAR is ligand-free. Potier et al . [29] and other groups [30–32] described the strength o f nondenaturing mass spectrometry to identify unexpected ligands coming from the expression host that bind to orphan receptors. More recently, the use of ESI-MS has also been reported for the characterization of protein–ligand complexes involving other nuclear receptors [ 4 33–38]. In the study reported here, we used ESI-MS to charac- terize the binding of CoRNRs to the LBDs of the HD heterodimer. In particular, the influence of ligand binding to RAR o r RXR on the stability of HD–CoRNR complexes was investigated by nondenaturing ESI-MS. We show that supramolecular mass spectrometry is a powerful tool to rapidly and unambiguously determine if a particular pep tide sequence can interact with LBDs of the HD heterodimer and to directly visualize the influence of various ligands on the stability of the corresponding ternary complexes. These results could only be obtained after precisely adjusting of the mass spectrometer interface p arameters, which a re des- cribed here in detail. Materials and methods Materials and chemicals Expression and purification of the HD heterodimer was performed as described by Bourguet et al. [39]. Peptides containing CoRNR1 and CoRNR2 sequences were syn- thesized and used as model peptides for the corepressor proteins. CoRNR1 is a 24 amino acid peptide (THRLIT LADHICQIITQDFARNQV; molecular mass 2806.2 Da) containing the consensus s equence that specifically interacts with the RAR subunit. CoRNR2 contains 14 amino acids (NLGLEDIIRKALMG; molecular mass 1542.8 Da) and interacts specifically with the RXR subunit. Ligands were chosen to interact specifically with either RAR (AM80, BMS493 and BMS614) or RXR (HX531) [28,40,41]. PSG5-based Gal-RARa and VP16-RARa expression vectors, and the (17m)5x-bG-Luc reporter gene have been described [42]. Gal-NCoR corresponds to a fusion between residues 1–147 of Gal4 and residues 1629–2453 of NCoR. Cell culture and transient transfections COS cells, c ultured in Dulbecco’s modified Eagle’s medium/ 5% (v/v) fetal bovine serum, were transiently transfected using the standard calcium phosphate method. Results were normalized to coexpressed b-galactosidase. Electrospray ionization mass spectrometry analysis Prior to any ESI-MS analysis, samples were desalted on Centricon PM30 microconcentrators (Amicon, Millipore, Molsheim, France) 5 in 100 m M ammonium acetate (pH 6.5). Ammonium acetate p resents the advantage not only o f preserving the ternary and quaternary structures of p roteins in solution but also of being compatible with ESI-MS experiments. ESI-MS measurements were performed on an electro- spray time-of-flight mass spectrometer (LCT, Waters, Manchester, UK) 6 . Purity and homogeneity of the retinoid receptors were verified by mass spectrometry analysis in denaturing conditions: p roteins were diluted to 5 pmolÆlL )1 in a water/acetonitrile mixture (1 : 1, v/v) acidified with 1% (v/v) formic acid. Mass spectra were recorded in the positive ion mode on the mass range 500–2500 m/z, after calibration with horse heart myoglobin diluted to 2 pmolÆlL )1 in a water/acetonitrile mixture ( 1 : 1, v/v) acidified with 1% (v/v) formic acid. The f ollowing molecular masses were measured: 29 930 ± 1.8 Da for RARa;26735±2.3Da for RXRa with deletion of the N-terminal methionine and 26 867 ± 2.1 Da corresponding to RXRa. These results were in agreement with the molecular masses calculated from the known amino acid sequences. The mass measurements of the noncovalent complexes were performed in ammonium acetate (50 m M ,pH6.5). Samples were diluted to 10 p molÆlL )1 in the previous buffer and continuously infused into the ESI ion source at a flow rate of 6 lLÆmin )1 through a Harvard syringe pump. Great care was exercised so that non covalent interactions survive the ionization/desorption process. Particularly the a cceler- ating v oltage (Vc), which controls the k inetic energy communicated to the ions in the interface region of the mass spectrometer, was optimized to 50 V i n o rder to prevent ligand dissociation in the gas phase. ESI-MS data were acquired in the positive ion mode on the mass range 1000–5000 m/z. Calibration of the instrument was per- formed using th e multiply charged ions produced by a Ó FEBS 2004 ESI-MS to investigate noncovalent complexes (Eur. J. Biochem. 271) 4959 separate injection of horse heart myoglobin diluted to 2pmolÆlL )1 in a water/acetonitrile mixture (1 : 1, v/v) acidified with 1% (v/v) formic acid. The relative abundance of the different species present o n ESI mass spectra were measured from their r espective peak i ntensities, assuming that relative intensities displayed by the different species on the ESI mass spectrum reflect the actual distribution of these species in solution. The reproducibility of the determination of the relative p roportions of the different species was estimatedtobe±2–3%. Results Optimization of the ESI-MS operating conditions to detect heterodimer–ligand–corepressor complexes To obtain optimal sensitivity without affecting the stability of noncovalent N R complexes, we first optimized the operating co nditions. I ndeed, i t i s c ritical t o p revent dissociation of w eak interactions in the mass spectrometer when transferring ions from the condensed phase to the gas phase. It has been shown by several groups that the gas phase stability of noncovalent complexes is greatly depend- ant on the nature of the interactions involved in the formation of the complexes [37,43–47]. It seems that the gas phase stability strongly increases with an increased contri- bution of electrostatic interactions within a noncovalent complex. Therefore , it is im portant that the Vc, which controls the kinetic energy communicated to the ions in the interface region of the mass spectrometer, is carefully set [37,47]. Operating co nditions were op timized on the HD– ligand–corepressor c omplex with CoRNR1 and BMS493, an inverse pan-RAR agonist [28,48] that does not interact with RXRs. Mass spectra for the ligand-bound HD–BMS493– CoRNR1 complex at different Vc voltages were generated from samples in which a threefold molar excess of CoRNR1 and a twofold molar excess of BM S493 were added to the HD heterodimer preparation (Fig. 1). At low accelerating voltages ( i.e. for Vc £ 30 V), the HD–BMS493–CoRNR1 noncovalent complex is detected as a major component (Fig. 1A). However, the observed peak shapes are broad, which probably r esults from incomplete desolvation. A direct consequence o f this i s a significant loss in mass accuracy. Nevertheless, significant amounts of the HD– BMS493 complex can be detected. Increasing the Vc to 50 V significantly improves the signal to noise ratio and still allows the detection of the HD–BMS493–CoRNR1 com- plex as a main component with a molecular mass of 59 884.1 ± 1.4 Da, while the HD–BMS493 complex with a molecular mass of 57 080.2 ± 0.9 Da is only a minor component (Fig. 1B). Under these conditions an additional ion series displaying a molecular mass of 5 9 474.2 ± 0.1 Da emerges a nd can be attributed to the unliganded HD– CoRNR1 complex. Thus, increasing Vc from 30 V to 50 V improves considerably the desolvation and the accuracy of 17 + V 18 + 16 + 17 + V 16 + 18 + 16 + 15 + 3 m/z 200 3300 3400 3500 3600 3700 3800 3900 17 + V 15 + 18 + 17 + 16 + 15 + Partial peptide dissociation Ligand dissociation DM = 408 Da (100%) Slight ligand dissociation (less than 5%) A B C Fig. 1. Optimization of the operating conditions: ESI-MS mass spectra of the HD–BMS493–CoRNR1 complex at different accelerating voltages (Vc). Positive ESI mass spectra of the HD–BMS493–CoRNR1 diluted to 10 pmolÆlL )1 in AcONH 4 (50 m M ,pH6.5)at(A)Vc¼ 30 V: the HD–BMS493–CoRNR1 noncovalent co mplex is detected without any dissociation; (B) Vc ¼ 50 V: slight dissocitation of the ligand (less than 5%) from th e HD–BMS493–CoRNR1 complex; (C) Vc ¼ 80 V: 100% dissociation of the ligand from the HD–CoRNR1 and partial dissociation of the peptide from the HD–CoRNR1 complex. Peaks labeled with * correspond to the different species with the additional N-terminal methionine. 4960 S. Sanglier et al.(Eur. J. Biochem. 271) Ó FEBS 2004 the mass m easurement but induces concomitantly some dissociation of the ligand-bound HD–BMS493–CoRNR1 complex into the unliganded HD–CoRNR1 complex. A further increase of the accelerating voltage to 80 V (Fig. 1C) provokes dissociation of t he quaternary complex to HD– CoRNR1 (major component) and HD (minor component) complexes. No ions corresponding to a ligand -bound species are detected under these conditions. These data indicate that the interactions involving the ligand are less stable in the gas phase than those involving the corepressor pep tide. Whether HD originates from the gas phase dissociation of the HD– BMS493 complex or from the HD–CoRNR1 complex cannot be distinguished 7 . Increasing the accelerating voltage to 120–150 V (data not shown) does not lead to any major change in the dissociation pattern regarding the complexes, which suggests that the HD–CoRNR1 complex is quite stable in the gas phase. The above data reveal that the chosen operating condi- tions (buffer and control of the Vcvoltage)allowthe transfer of intact noncovalent receptor–ligand–corepressor peptide complexes from solution to the gas phase. It should be noted that such experiments were performed for each HD–ligand–CoRNR complex described in this paper. In all cases, at low accelerating voltages (30–50 V), HD–ligand– CoRNR complexes remained intact after passage into the gas phase. Increasing the Vc v oltage (up to 80–150 V) resulted in a stepwise removal of the ligand first, and subsequent dissociation of the CoR peptide. Consequently, as a compromise between sufficient desolvation, optimal transmission of the ions and nondestructive gas-phase collisions was found at 50 V and we retained such experimental conditions in our further experiments. Interestingly, while various biochemical studies including circular dichroism, proteolytic digestions or gel mobility suggested that NRs undergo conformational changes upon ligand binding, no c harge state distribution changes were detected upon ligand binding or CoR peptide binding (+15 t o +17) 8 by MS. Indeed, it has been proposed by Witkowska et al. [23], that ESI-MS might also be useful for revealing structural changes in protein conformations from subtle changes in the charge state d istribution observed on ESI mass spectra [20,49–52]. So, if conformational changes occur, these do not sufficiently affect the surface of the protein exposed to solvent to be detected by mass spectro- metry. This latter assumption is supported by X -ray crystallographic studies showing no obvious evidence of new exposure of charged amino acids. The mainly local conformational changes appear not to affect the protein surface in a way sufficient to be revealed by ESI-MS. RAR–RXR heterodimer forms a 1 : 1 noncovalent complex with corepressor peptides The mass spectrum obtained at a threefold molar excess of CoRNR1 over HD displays two main ion series at mass to charge ratios between m/z 3000 and m/z 4000 (Fig. 2A). The first series (r) corresponds to the +15 to +17 charge states of the HD, with a molecular mass of 56 669.4 ± 2.5 Da. The second series (d) gives rise to a molecular mass of 59 478.3 ± 0.5 Da and can be attributed to the formation of a 1 : 1 complex between HD and CoRNR1 c orepressor peptide, with a charge state distribution varying from +16 to +18. Using identical experimental conditions for HD–CoRNR2 complexes, two series o f multiply charged ions were detected corresponding to the HD (56 664.8 ± 0.3 Da) and t he HD–CoRNR2 complex ( 58 210.2 ± 1.8 Da) (Fig. 3A). All data were recorded at a cone voltage of 50 V (see above). As ESI-MS a nalyses faithfully reflect solution equilibrium in the previously mentioned experi- mental conditions, we reproducibly observed that only  50% of the detected species correspond to both HD–CoRNR complexes in the absence of ligand. Titration experiments, in which increasing amounts of CoRNR peptide (up to fourfold molar excess, data not shown) were added to the HD, reveal no further saturation of H D in the absence of ligand. This suggests that a twofold molar excess of CoRNR peptide is sufficient to reach the equilibrium state under our experimental conditions. T his ratio (50%) was taken as a reference point in the present study to monitor the effect of various classes of ligands on this equilibrium. Ligand modulation of HD–CoRNR ternary complexes In order to monitor t he effect of different types of ligands on the recruitment of t wo CoRNR peptides, various ligands were added to each HD–CoRNR complex and the stability of each ternary complex was recorded by ESI-MS. The antagonist activity of the inverse agonist BMS493, the RARa-selective a gonist BMS614 and the RXR antagonist HX531 were confirmed by competition o f AM80-induced transactivation using the chimeric Gal-RARa as ligand- dependent activator of a cognate 17mer-luciferase reporter gene (Fig. 4). These data reveal the antagonistic potential of BMS493 and BMS614 and demonstrate also that HX531 has a weak RARa antagonist activity. Effects of RAR ligand binding on the stability of the HD–CoRNR1 complex. Mass spectra obtained for the HD in presence of CoRNR1 with a twofold molar excess of either the inverse pan-RAR antagonist BMS493 (Fig. 2B) or the RARa-selective agonist AM80 (Fig. 2 C), revealed that all ligands bind to the HD and to the HD–CoRNR complexes. More importantly, they have a dramatic effect on the initial HD to HD–CoRN R1 ratio (Fig. 2A–C). In the presence of a twofold molar excess of BMS493, two ion series are observed. The most abundant series (d)hasa molecular mass of 59 884.1 ± 1.4 Da corresponding to the binding of one BMS493 molecule [mass difference (DM) 9 ¼ 408 Da] to the HD–CoRNR1 complex (Fig. 2B). The minor one (r) can be attributed to the fixation of one BMS493 molecule to the free HD. No significant signal corresponding to unliganded species is detected. Even i n the presence of a twofold molar excess of BMS493, no protein/ ligand stoichiometry higher than 1 : 1 is observed, indicating that the specific binding of BMS493 to its cognate receptor is faithfully re vealed by MS and that no Ônonsite-specificÕ ligand addition resulting from any artefactual aggregation during the electrospray ionization process occurs. Similar results were obtained with BMS614, a RARa-selective antagonist, which induced a mass increase of 448 Da for the binding of one BMS614 molecule to the HD and HD– CoRNR complexes (data not shown). As i nterface condi- tions are optimized so that neither peptides nor ligands dissociate during transfer to the gas phase, we con clude that Ó FEBS 2004 ESI-MS to investigate noncovalent complexes (Eur. J. Biochem. 271) 4961 the HD–BMS493 complex is present in solution at equilibrium and does not result from the dissociation of the HD–BMS493–CoRNR1 complex in the interface of the mass spectrometer. For c omparison of t he relative proportions of the different species present in solution, we made the assump- tion that ionization efficiencies a nd response factors of all species are similar, and thus that relative intensities of the main charge states can serve to estimate the relative abundances in solution. This assumption is supported by several points: (a) strictly identical experimental conditions are used for the a ddition of each ligand (molar excess, buffer, pH, Vc, etc.); (b) the mass contribution of the ligand is very small compared to the mass of the HD–CoRNR complex, thus similar response factors are expected; (c) all species exhibit nearly the same charge states and are detected with the same m/z ratios, suggesting that no strong discrimination effects o f the ion species upon focalization in the interface of the mass spectrometer should occur. Table 1 summarizes the relative abundances of all species in the absence or presence of the corepressor peptide. The fact that the relative abundance of the HD–CoRNR1 complex significantly increases in the presence of RAR-antagonist or inverse agonist ligands (75% and 56% of HD–CoRNR1 in the presence of, respectively, BMS493 or BMS614 instead of 43% in the absence of ligand) indicates that these RAR- ligands stabilize the HD–CoRNR1 ternary complex. Above 60 V (data not shown), t he ligand is dissociated fro m all species in the gas phase while the r atio between HD and HD–CoRNR1 stays unchanged. To compare these data with an analysis of the ligand- induced complex formation in living cells in vivo,we performed m ammalian cell two-hybrid e xperiments in which w e monitore d t he interaction of a VP16 activation domain-linked R ARa with a Gal-NCoR hybrid t hat w as capable of activating a 17mer-based luciferase reporter 17 + 16 + • • • No ligand 18 + 16 + 15 + 16 + 351 Da 15 + 17 + 3200 3250 3300 3350 3400 3450 3500 3550 3600 3650 3700 3750 3800 3850 3900 m/z • • • • • • • • +2 +2 +1 +2 +1 +2 16 + 17 + +2 +1 +1 Reference 17 + DM= DM= 408 Da • • • 15 + 16 + 18 + 16 + Stabilizing Effect RAR-antagonist BMS493 RAR-agonist AM80 RXR-antagonist HX531 Destabilizing Effect No apparent Effect * * * * * * * * * * * * * * * * * * * * * * A B C D Fig. 2. Effect of ligand additions on the H D–CoRNR1 ternary complex. Positive ESI mass spectra of HD–CoRNR1. The mass spectra were acquired at Vc ¼ 50 V and at a pressure in the interface of 2.5 mbar. (A) Before any ligand addition. After addition of t wofold molar excess 16 of (B) BM S493: the HD–CoRNR1 complex is stabilized upon BMS493 binding; (C) AM80: the HD–CoRNR1 complex is completely destabilized upon AM80 binding; (D) HX531: no ap parent effect on the stability of the HD–CoRNR1 comple x. 4962 S. Sanglier et al.(Eur. J. Biochem. 271) Ó FEBS 2004 (Fig. 5). Transcription activation indicated N CoR–RARa interaction strongly increased upon exposing the cells to BMS493, while only a weak effect was seen in presence of BMS614, and little, if any, in the presence of HX531 (Fig. 5). We conclude that the MS analysis faithfully monitors the modulation of receptor–corepressor inter- action, as it occurs in living cells. After the addition of a twofold molar e xcess o f t he RARa-specific agonist AM80 in the presence of CoRNR1 (Fig. 2C), one single ion series (r) displaying a mass increase of about 351 Da above the molecular mass of the HD is then observed. No ions corresponding to the ligand- bound HD–AM80–CoRNR1 or to nonliganded HD are detected. Therefore, MS analysis faithfully reveals the previously observed effects of RAR agonists, antagonists and inverse agonists on corepressor–HD interaction . Effects of RAR ligand binding on the stability of the HD– CoRNR2 complex. ESI mass spectra obtained after the addition of a twofold molar excess of d ifferent ligands and CoRNR2 are presented in Fig. 3 for the inverse pan-RAR agonist BMS493 (Fig. 3B) an d t he RAR a agonist AM80 (Fig. 3C). In the absence of ligand a mixture of HD (47%) and HD–CoRNR2 (53%) was observed (Fig. 3A). As for CoRNR1, the addition of RAR-specific ligands induces a dramatic change in the relative amount o f the various complexes. While the agonist AM80 and the antagonist BMS614 induce a complete destabilization of the ternary HD–CoRNR2 complex, no similar effect is exerted b y the inverse agonist BMS493. Indeed, 22% of all observed species correspond to the HD–BMS493–CoRNR2 com- plex. In all cases, a single ligand molecule is qu antitatively bound to the HD complex. 3200 3250 3300 3350 3400 3450 3500 3550 3600 3650 3700 3750 3800 3850 3900 m/z 16 + 17 + + DM= DM= 408 Da 16 + 16 + • • • 17 + 15 + 17 + 17 + 15 + • • 16+ 16 + 351 Da 15 + 17 + • • • • +2 +2 16 + 16 + +2 +2 +1 +1 +2+1 * * * * * * * * * * * A B C D Reference Destabilizing Effect Destabilizing Effect No apparent Effect No ligand RAR-antagonist BMS493 RAR-agonist AM80 RXR-antagonist HX531 Fig. 3. Effect of ligand additions on the HD–CoRNR2 ternary complex. Positive ESI mass spect ra of HD –CoRNR2. The m ass spectra were acquired at Vc ¼ 50 V and at a pressure in the interface of 2.5 mbar. (A) Before any ligand addition. After addition of twofold molar excess 17 of (B) BMS493: the HD–CoRNR2 complex is d estabilized upon BMS493 binding; (C) AM80: the H D–CoRNR2 complex i s completely destabilized upon AM80 binding. (D) HX531: no apparent effect on the stability of the HD–CoRNR2 complex. Ó FEBS 2004 ESI-MS to investigate noncovalent complexes (Eur. J. Biochem. 271) 4963 Effects of RXR ligands on the stability of HD–CoRNR complexes. WhenatwofoldmolarexcessoftheRXR antagonist HX531 is added (Figs 2D and 3D), the inter- pretation of the mass spectra becomes significantly more difficult because multiple species are detected. HD a nd ternary HD–CoRNR1 complexes were observed both as liganded and unliganded species, with the liganded species involving one or even two HX531 molecules. In the presence of this RXR ligand, no major difference concerning the initial HD to HD–CoRNR1 ratio is observed in the mass spectra. The addition of a twofold molar excess of a RXR-specific ligand induces some stabilization of the HD–CoRNR1 complex (51% in the presence of HX531 vs. 43% in the absence of any ligand; Table 1). Because of the presence of a second ligand molecule bound to each detected complex, no clear conclusions regarding t he influence of a RXR-specific ligand on t he stability o f the HD–CoRNR1 complex can be drawn. Discussion ESI-MS allows monitoring of ligand modulations on HD–CoRNR complexes Due to its ability to detect each species present in solution, ESI-MS appears to be a powerful tool to follow coregulator association a nd dissociation to NR upon binding of different classes of ligands. In addition to previous studies [28,37], we now show that nondenaturaing mass spectro- metry can be used, not only to monitor binding of coregulator peptides but also to assess the effect of ligand binding on the stability o f the corresponding NR–coregu- lator complexes 10 . The interpretation of the ESI mass spectra provides several facts 11 : (a) both corepressor peptides (CoRNR1 or CoRNR2) bind to the HD in the absence of any ligand, and (b) ESI-MS data allow one to directly monitor the effect of ligand binding on the association/ dissociation pattern of the heterodimer and the corepressor. RAR-specific ligands display a s trong influence o n the stability of heterodimer–CoRNR complexes, comparing well with previous reports showing that induction of corepressor release is part of the mechanism of action of RAR agonists [1,53]. Conversely, the stabilization of the corepressor complex by RAR inverse agonist enhances transcriptional silencing [28,48]. These MS data are fully in keeping with the corresponding results obtained by transient transactivation studies. Such conclusions are not yet routine and it is o f most i mportance to c ontrol that the solution phase image is not distorted during the ESI-MS analysis and that observed peaks on ESI mass spectra in vacuo can be related to species effectively present in solution. transactivation relative to Am80 10 –8 M (100%) 100 125 75 50 25 0 0–5–8 –7 –6 Ligands [logM] Gal - RARα + Am80 10 –8 M 5 x Gal4 Gal4 LucβG RARα BMS493 BMS614 HX531 Fig. 4. RARa antagonist potential of synthetic retinoids. Transient transactivation assays were performed to assess the antagonist activity of BMS493 (j), BMS614 (n), and HX531 (m). COS cells were co- transfected with reporter (17m)5x-bG-Luc and Gal-RARa.The reporter was activated with 10 n M Am80 (100 %) and increasing con- centrations of the re spective retinoid were added, as indica ted. Table 1. Influence of the addition of ligands on the stability of the noncovalent HD–CoRNR complexes. The relative abundance of the species are obtained by summing the peak intensities of the three predominant charge states of the complexes. Reproducibility of the values is within 2–3%. No ligand Relative amounts of complexes detected by ESI-MS (%) In presence of CoRNR1 In presence of CoRNR2 RAR–RXR + CoRNR1 RAR–RXR RAR–RXR + CoRNR2 RAR–RXR 43 57 47 53 Ligand RAR–RXR/L + CoRNR1 RAR–RXR/L RAR–RXR/L + CoRNR2 RAR–RXR/L AM80 RAR-agonist 0 100 0 100 BMS493 RAR-inverse agonist 75 25 22 78 BMS614 RAR-antagonist 56 44 6 94 HX531 RXR-antagonist 51 49 50 50 4964 S. Sanglier et al.(Eur. J. Biochem. 271) Ó FEBS 2004 If the observed complex results from specific interactions in solution, it must be sensitive to modifications of the experimental conditions affecting its stability. The fact that different ligands induce (a) a substantial change in the mass spectrum and (b) distinct effects on the s tability of HD–corepressor complexes, provides a good support for a Ôstructurally-specificÕ interaction. To establish that t he interactions detected by ESI-MS in presence of RXR ligands are really the result of i n-solution interactions rather than artefactual nonspecific associations occurring during the ESI process [ 54], rigorous control e xperiments were performed. In particular, separate titration experiments performed with the RXR ligand and the monomeric RXRa subunit revealed that, even in the presence of an e xcess of ligand (a threefold molar excess), no b inding of a second HX531 molecule is observed on RXRa (data not shown). At this level, two h ypotheses may be advanced to explain the binding of a second ligand molecule to the h eterodimer: either the second ligand molecule is anchored directly in the free RAR-binding site [this option is strongly supported by the antagonistic action of HX531 on AM80-induced Gal- RARa transactivation ( Fig. 4)], or the second ligand molecule is nonspecifically interacting at the protein surface. Because the single RAR subunit tends to precipitate under ESI-MS compatible conditions, experiments with mono- meric RAR could not be achieved. Relationship between the type of interactions involved in the formation of HD–ligand–CoRNR complexes and their gas phase stabilities The ESI-MS detection of complexes involving few electro- static contacts between the LBP and t he ligands is not straightforward. As described in the literature [37,43–46], the stability of noncovalent complexes in the ESI-MS process depends strongly on the type of interaction (ele c- trostatic contacts, hydrogen bonds, Van der Waals interac- tions) involved in the formation of the complex. During ion transfer from the solution to the gas phase both electrostatic interactions an d hydrogen bonds are emphasized, w hile complexes whose formations in solution are mainly driven by hydrophobic effects appear to be weakened [55,56]. The present study shows that HD–CoRNR and HD–ligand noncovalent complexes have different gas phase stabilities. The Vc, which controls the energy communicated to the ions in the interface of the mass spectrometer, needs to be set to quite low values (30–50 V) to detect the interaction between NRs and ligands, whereas higher Vc values can be used to observe HD–CoRNR complexes, suggesting that these latter complexes are more s table in the gas phase. Such behavior is in complete agreement with crystallographic data showing that interactions between NRs and ligands are mainly hydrophobic, whereas NR– CoRNR 12 interactions involve more polar contacts. Until now, i t h as been 13 difficult to link the number of polar contacts involved in a complex formation and its subse- quent MS detection. For the receptors of retinoids, the carboxylic group of the ligand is the only moiety engaged in polar interactions with the LBP. Two hypotheses can be made to explain the ESI-MS observations: (a) few electro- static contacts are sufficient to maintain the NR–ligand interactions in the gas phase, and (b) the LBP conformation that to tally surrounds the ligand prevents its ejection from the pocket. Whether dissociation experiments might be used to directly describe the type of interactions in volved in the formation of a given noncovalent complex constitutes a new challenge. Conclusion In this study, we demonstrate that noncovalent mass spectrometry is a powerful technique to obtain direct evidence of the formation of NR complexes and to monitor the effect of ligand binding on the stability of the resulting complexes. Thanks to its unique advantage of providing direct insights into all s pecies pr esent in s olution, even moderate changes induced by the addition of a given substrate (ligand, peptides, metallic ions, small organic molecules, etc.) in the association/dissociation pattern can be directly revealed on the E SI mass spectra. Therefore, this nondenaturing ESI-MS approach will efficiently comple- ment other structural methods such as NMR and crystal- lography. ESI-MS is a technique that can be extended to a wide range of noncovalent protein–protein or protein–ligand complexes. Combining rapidity, sensitivity and accuracy, it will play an increasingly important role not only for the analysis of nuclear receptor action and (orphan) receptor ligand detection, but also for transcription regulation in general, which is triggered by noncovalent interactions. 10000 15000 5000 0 relative transactivation etOH Am80 BMS493 BMS614 HX531 Gal-NCoR + VP16 - RARα 5xGal4 Gal4 LucβG NCoR RAR VP16 Fig. 5. Enhanced interaction between RARa and NCoR induced by synthetic retinoids. Mammaliantwo-hybridassaywithGal-NCoRas bait and VP16-RARa as prey were performed to assess the influence of indicated syn thetic retinoids on interaction between RARa and NCoR in a cellular context. Ó FEBS 2004 ESI-MS to investigate noncovalent complexes (Eur. J. Biochem. 271) 4965 Acknowledgements We are grateful to Pascal Eberling fo r the synthesis of the CoRNR1 corepression peptid e, to Dr Lazar for the kind gift o f t he CoRNR2 peptide a nd to Drs Chris Zusi and Koichi Shudo for ligands. S.S. acknowledges the CNRS and Lilly for financial support. Work in the laboratory of H.G. i s supported b y g rants from t he European Commission (QLK3-CT2002-02029, HPRN-CT2002-00268), the Fon- dation de France, the Association for International Cancer Research, the Association pour la Recherche sur le Cancer, the ULP, the INSERM, the CNRS, and Bristol-Myers S quibb. References 1. Laudet,V.&Gronemeyer,H.(2002)The Nuclear Receptor Facts Book. Academic Press, San Diego. 2. Kastner, P., Mark, M. & Chambon, P. (1995) Nonsteroid nuclear receptors: what are g enetic studies telling us about their role in r eal life? Cell 83, 859–869. 3. Altucci, L. & Gronemeyer, H. (2001) The p romise of retinoids to fight against cance r. Nat. Rev. Cancer 1, 181–193. 4. Jimenez-Lara, A.M., Clarke, N., Altucci, L. & Gronemeyer, H. (2004) Retinoic acid-induced apoptosis in leukemia cells and the implications of recent findings for pathogenesis and therapy. Trends Mol. Med. 10, 508–515. 5. Mangelsdorf, D.J. & Evans, R.M. (1995) The RXR heterodimers andorphanreceptors.Cell 83, 841–850. 6. Rastinejad, F. (2001) Retinoid X receptor and its partners in the nuclear rec eptor family. Curr.Opin.Struct.Biol.11, 33–38. 7. Moras, D. & G ronemeyer, H. (1998) The nuclear receptor ligand- binding domain: structure and function. Curr. Opin. Cell Biol. 10, 384–391. 8. Bourguet, W., Germain, P. & Gronemeyer, H. (2000) Nuclear receptor ligand-binding domains: three-dimensional structures, molecular interactions and pharmacological implications. Trends Pharmacol. Sci. 21, 381–388. 9. Greschik, H. & Moras, D. (2003) Structure–activity relationship of nuclear rec eptor–lig and interactions. Curr. Top. Med. Chem. 3, 1573–1599. 10. Metivier, R., Penot , G., Hubner, M.R., Reid, G., B rand, H., Kos, M. & G annon, F. (2003) Estrogen receptor-alpha directs ordered, cyclical, and combinatorial recruitm ent of cofactors on a natural target promoter. Cell 115, 751–763. 11. Evans, R.M., Barish, G.D. & Wang, Y.X. (2004) PPARs and the complex jo urney to obesity. Nat. Med. 10, 355–361. 12. Chawla, A., Repa, J.J., Evans, R.M. & Mangelsdorf, D.J. (2001) Nuclear receptors a nd lipid physiology: ope ning the X-files. Science 294, 1866–1870. 13. Altucci, L . & Gronemeyer, H. (2001) Nuclear receptors in cell life and death. Trends Endocrinol. Metab. 12, 460–468. 14. Clarke, N., Jimenez-Lara, A.M., Voltz, E. & Groneme yer, H. (2004) Tumor suppressor IRF-1 mediates retinoid and inter- feron anticancer signaling to death ligand TRAIL. EMBO J. 23, 3051–3060. 15. Horlein, A.J., Naar, A.M., Heinzel, T., Torchia, J., Gloss, B., Kurokawa, R., Ryan, A., Kamei, Y., Soderstrom, M., Glass, C.K. et al. (1995) Ligand-indep endent repression by the t hyroid h or- mone receptor mediated by a nuclear receptor co-repressor. Nature 377, 397–404. 16. Chen, J.D. & Evans, R.M. (1995) A transcriptional co-repressor that interacts with nuclear hormone receptors. Nature 377, 454– 457. 17. Nagy, L., Kao, H.Y., Love, J.D., Li, C., Banayo, E., G ooch, J.T., Krishna,V.,Chatterjee,K.,Evans,R.M.&Schwabe,J.W.(1999) Mechanism of corepressor binding and release from nuclear hor- mone receptors. Genes Dev. 13 , 3209–3216. 18. Perissi, V., Staszewski, L.M., McInerney, E.M., Kurokawa, R., Krones, A., Rose, D.W., Lambert, M.H., Milburn, M.V., Glass, C.K. & Rosenfeld, M.G. (1999) Molecular d etermina nts of nu clear receptor–corepressor interac tion. Genes Dev. 13, 3198–3208. 19. Hu, X. & Lazar, M.A. (1999) The CoRNR motif controls the recruitment of corepressors by nuclear hormone receptors. Nature 402, 93–96. 20. Loo, J.A. (2000) Electrospray ionization m ass spectrometry: a technology for studying noncovalent macromolecular complexes. Int. J. Mass Spectrom. 200, 175–186. 21. Last, A.M. & Robinson, C.V. (1999) Protein folding an d inter- actions revealed by mass spectrometry. Curr. Opin. Chem. Biol. 3, 564–570. 22. Heck, A.J.R. & Van den Heuvel, R.R. ( 2004) Investigation of intact protein complexes by mass spectrometry. Mass Spectrom. Rev. 23, 368–389. 23. Witkowska, H.E., Green, B.N., Carlquist, M. & Shackleton, C.H. (1996) Intact noncovale nt dimer of estrogen receptor ligand- binding domain can be detected by electrospray ionization mass spectrometry. Steroids 61, 433–438. 24. Witkowska, H.E., Carlquist, M., Engstrom, O., Carlsson, B., Bonn, T., Gustafsson, J.A. & Shackleton, C.H. (1997) Char- acterization of bacterially expressed rat estrogen receptor beta ligand binding domain by mass spectrometry: structural com- parisonwithestrogenreceptoralpha.Steroids 62, 621–631. 25. Craig, T.A., Benson, L.M., Naylor, S. & Kumar, R. (2001) Modulation effects of zinc on the formation of vitamin D receptor and retinoid X receptor alpha-DNA transcriptio n complexes: analysis by microelectrospray mass spectrometry. Rapid Commun. Mass Spectrom. 15, 1011–1016. 26. Veenstra, T.D., Benson, L.M., Craig, T.A., Tomlinson, A.J., Ku- mar, R. & Naylor, S. (1998) M etal mediated sterol receptor–DNA complex association and dissociation determined by electrospray ionization mass spectrometry. Nat. Biotechnol. 16, 262–266. 27. Bourguet, W., Vivat, V., Wurtz, J.M., Chambon, P., Gronemeyer, H. & Moras, D. (2000) Crystal structure of a heterodimeric complex of RAR and RXR ligand-binding domains. Mol. Cell 5, 289–298. 28. Germain, P., Iyer, J., Zechel, C. & Gronemeyer, H. (2002) Cor- egulator recruitm ent and th e mechan ism of retinoic acid receptor synergy. Nature 415, 187–192. 29. Potier, N., Billas, I.M., Steinmetz, A., Schaeffer, C., van Dors- selaer, A., Moras, D. & Renaud, J.P. (2003) Using nondenaturing mass spectrometry to detect fortuitous ligands in orphan nuclear receptors. Protein Sci. 12, 725–733. 30. Dhe-Paganon , S., Duda, K., Iwamo to, M., Chi, Y.I. & Shoelson, S.E. (2002) Crystal structure of the HNF4 alpha ligand binding domain in complex w ith endogenous fatty acid ligand. J. Biol. Chem. 277, 37973–37976. 31. Wisely, G.B., Miller, A.B., Davis, R.G., Thornquest, A.D. Jr, Johnson, R., Spitzer, T., Sefler, A., Shearer, B., Moore, J.T., Willson, T.M. & Williams, S.P. (2002) Hepatocyte nu clear factor 4 is a transcription factor that constitutively binds fatty acids. Structure (Camb.) 10, 1225–1234. 32. Kallen, J.A., Schlaeppi, J.M., Bitsch, F., Geisse, S., Geiser, M., Delhon, I. & Fournier, B. (2002) X-ray structure of t he hROR- alpha LBD at 1.63 A ˚ : structural and functional data that cho les- terol or a cholesterol derivative is t he natural ligand of R ORalpha. Structure (Camb.) 10, 1697–1707. 33. Potier, N., Lamour, V., Poterszman, A ., Thierry, J.C., Moras, D. & Van Dorsselaer, A. (2000) Characterization of crystal content by ESI-MS and MALDI-MS. Acta Crystallogr. D56, 1583–1590. 34. Rocchi,S.,Picard,F.,Vamecq,J., Gelman, L., Potier, N., Zeyer, D.,Dubuquoy,L.,Bac,P.,Champy,M.F.,Plunket,K.D., Leesnitzer, L.M., Blanch ard, S.G., Desreumaux, P., Moras, D., Renaud, J.P. & Auwerx, J. (2001) A unique PPARgamma ligand 4966 S. Sanglier et al.(Eur. J. Biochem. 271) Ó FEBS 2004 with potent insulin-sensitizing yet weak adipogenic activity. Mol. Cell 8, 737–747. 35. Greschik, H., Wurtz, J.M., Sanglier, S., Bourguet, W., Van Dorsselaer, A., Moras, D. & Renaud, J.P. (2002) Structural and functional evidence fo r ligand-independen t tran scriptional acti- vation by the estrogen-related r eceptor 3. Mol. Cell 9, 303–313. 36. Stehlin-Gaon, C., Willmann, D., Zeyer, D., Sanglier, S., Van Dorsselaer, A., Renaud, J.P., Moras, D. & Schule, R. (2003) All- trans r etinoic acid is a ligand for the orphan nuclear receptor ROR beta. Nat. Struct. Biol. 10, 820–825. 37. Bitsch, F., Aichholz, R ., Kallen,J.,Geisse,S.,Fournier,B.& Schlaeppi, J.M. (2003) Identification of natural ligands of retinoic acid receptor-related orphan receptor alph a ligand-binding domain ex pressed in Sf9 cells – a m ass spectrometry approach. Anal. Biochem. 32 3, 139–149. 38. Lengqvist , J., Mata De Urquiza, A., B ergman, A.C., Willson, T.M., Sjovall, J., Perlmann, T. & Griffiths, W.J. (2004) Poly- unsaturated fatty acids i nclu ding do cosahexaeno ic and arac hi- donic acid b ind to t he retinoid X recep tor alpha ligand-binding domain. Mol. Cell. Pro t eo mi cs 3, 692–703. 39. Bourguet,W.,Andry,V.,Iltis,C.,Klaholz,B.,Potier,N.,Van Dorsselaer, A., Chambon, P., Gronemeyer, H. & Moras, D. (2000) Heterodimeric Complex of RAR and RXR Nuclear Receptor Ligand-Binding Domains: Purification, Crystallization, and P reliminary X-Ray Diffraction Analysis. Protein Expr. Purif. 19, 284–288. 40. Gehin,M.,Vivat,V.,Wurtz,J.M.,Losson,R.,Chambon,P., Moras, D. & Gronemeyer, H. (1999) Structural basis for engineering of retinoic acid receptor isotype-selective agonists and antagonists. Chem. Biol. 6, 519–529. 41. Vivat, V., Zechel, C., Wurtz, J.M., Bourguet, W., Kagechika, H., Umemiya, H., Shudo, K., Moras, D., Gronemeyer, H. & Cham- bon, P. (1997) A mutation mimicking ligand-induced conforma- tional change yields a constitutive RXR that senses allosteric effects in heterodimers. EMBO J. 16, 5697–5709. 42. Nagpal, S., Friant, S., Nakshatri, H . & Chambon, P. (1993) RARs and RXRs: evidence fo r two autonomous transactivation func- tions (AF-1 and AF-2) and h eterodimerization in vivo. EMBO J. 12, 2349–2360. 43. Smith, R.D. & Light-Wahl, K.J. (1993) The observation of non- covalent interactions in solution b y electrospray ionization mass spectrometry: promise, pitfalls and prognosis. Biol. Mass Spec- trom. 22, 493–501. 44. Robin son, C.V ., Chu ng, E.W., Kragelund, B .B., Kn udsen, J., Aplin, R.T., Poulsen, F.M. & Dobson, C.M. (1996) Probing the nature of noncovalent interactions by mass spectrometry. A study of protein-CoA ligand binding and assembly. J. Am. Chem. Soc. 118, 8646–8653. 45. Rogniaux, H ., Va n D orsselaer, A ., Barth, P., Biellmann, J.F., Barbanton, J., van Zandt, M., Chevrier, B., Howard, E., Mits- chler, A., Potier, N., Urzhumtseva, L., Moras, D. & Podjarny, A. (1999) Binding o f aldose reductase inhibitors: correlation o f crystallographic and mass spectrometric studies. J. Am. Soc. Mass Spectrom. 10, 635–647. 46. Hernandez, H., Hewitson , K.S ., Roach, P., Shaw, N.M., Baldwin, J.E. & Robinson, C.V. (2001) Observation of the iron-sulfur cluster in Escherichia coli biotin synthase by nanoflow electrospray mass spectrometry. Anal. Chem. 73, 4154–4161. 47. Pramanik, B.N., Bartner, P.L., Mirza, U.A., Liu, Y.H. & Ganguly, A.K. (1998) Electrospray ionization mass spectrometry for the study of non-covalent complexes: an emerging technology. J. Mass Spectrom. 33, 911–920. 48. Klein, E.S., Pino, M.E., Johnson, A.T., Davies, P.J., Nagpal, S., Thacher, S.M., K rasinski, G. & Chandraratna, R.A. (1996) Identification and functional separation of retinoic acid receptor neutral antagonists and inverse agonists. J. Biol. Chem. 271, 22692–22696. 49. Katta, V. & Chait, B.T. (1991) Observation of the Heme-Globin complex in native myoglobin by electrospray-ionization mass spectrometry. J. Am. Chem. Soc. 113, 8534–8535. 50. Chowdhury, S.K., Katta, V. & Chait, B.T. (1990) Probing con- formational changes in proteins by mass spectrometry. J. Am. Chem. Soc. 112, 9012–9013. 51. Loo, J.A., Oo gorzalek-L oo, R.R., Udseth, H.R., Edmo nds, C.G. & Smith, R.D. (1991) Solvent-induced conformational changes of polypeptides probed by electrospray ionisation mass spec tro- metry. Rapid Commun. Mass Spectrom. 5, 101–105. 52. Le Blanc, J.C.Y., Beuchemin, D., Siu, K.W.M., Guevremont, R. & Berman, S.S. (1991) Thermal denaturation of some proteins and itseffectontheirelectrospraymassspectra.Org. Mass Spectrom. 26, 831–883. 53. Hu, X. & Lazar, M.A. (2000) Transcriptional repression by nuclear hormone receptors. Trends Endocrinol. Metab. 11, 6–10. 54. Ding, J. & Anderegg, R.J. (1995) Specific and nonspecific dimer formation in the electrospray ionization mass spectrometry of oligonucleotides. J. Am. Soc. Mass Spectrom. 6, 159–164. 55. Li, Y.T., Hsieh, Y.L., Henion, J.D., Senko, M.W., McLafferty, F.W. & Ganem, B. (1993) Mass spectrometric studies on non- covalent dimers of leucine zipper peptides. J. Am. Chem. Soc. 115, 8409–8413. 56. Li, Y.T., Hsieh, Y.L., Henion, J.D., Ocain, T.D., Schiehser, G.A. & Ganem, B. (1994) Analysis of the energetics of gas-phase immunophilin-ligand comp lexes by ion s pray mass spectrometry. J. Am. Chem. Soc. 116, 7487–7493. Ó FEBS 2004 ESI-MS to investigate noncovalent complexes (Eur. J. Biochem. 271) 4967 . Monitoring ligand-mediated nuclear receptor–coregulator interactions by noncovalent mass spectrometry Sarah Sanglier 1 ,. Poulsen, F.M. & Dobson, C.M. (1996) Probing the nature of noncovalent interactions by mass spectrometry. A study of protein-CoA ligand binding and assembly.