Structural Plasticity and Noncovalent Substrate Binding in the GroEL Apical Domain

Advances in understanding how GroEL binds to non-native proteins are reported. Conformational flexibility in the GroEL apical domain, which could account for the variety of substrates that GroEL binds, is illustrated by comparison of several independent crystallographic structures of apical domain constructs that show conformational plasticity in helices H and I. Additionally, ESI-MS indicates that apical domain constructs have co-populated conformations at neutral pH. To assess the ability of different apical domain conformers to bind co-chaperone and substrate, model peptides corresponding to the mobile loop of GroES and to helix D from rhodanese were studied. Analysis of apical domain-peptide complexes by ESI-MS indicates that only the folded or partially folded apical domain conformations form complexes that survive gas phase conditions. Fluorescence binding studies show that the apical domain can fully bind both peptides independently. No competition for binding was observed, suggesting the peptides have distinct apical domain-binding sites. Blocking the GroES-apical domain-binding site in GroEL rendered the chaperonin inactive in binding GroES and in assisting the folding of denatured rhodanese, but still capable of binding non-native proteins, supporting the conclusion that GroES and substrate proteins have, at least partially, distinct binding sites even in the intact GroEL tetradecamer.

Several classes of molecular chaperones assist in the folding of newly synthesized polypeptides by preventing off-pathway reactions that lead to aggregation (1)(2)(3). The so-called chaperonins (4) are large cylindrical structures that transiently enclose a partially folded polypeptide and allow it to continue folding in a sequestered environment, blocking intermolecular associations between chains during folding. The chaperonin of Escherichia coli, GroEL, belongs to the group I class of chaperonins and has been studied in great detail (e.g. Refs. 5 and 6). GroEL interacts with an estimated 10 -15% of newly synthesized polypeptides in the bacterial cytosol (7,8).
GroEL is a homotetradecamer of ϳ57-kDa subunits that are arranged as two heptameric rings stacked back-to-back (9,10). GroEL requires the presence of a co-chaperone, GroES, together with a controlled cycle of nucleotide binding and hydrolysis, to complete its functional cycle (1,11,12). GroES is a single heptameric ring of identical ϳ10-kDa subunits that binds in a nucleotide-dependent manner to one end of the GroEL cylinder. A substrate protein binds with highest affinity to the nucleotide-free state of GroEL (reviewed in Ref. 1). On subsequent binding of 7 ATP and GroES, the substrate protein is jettisoned into a cis-folding cage where it is free to fold to its final native form. Whereas the nucleotide-coupled interplay between GroEL, GroES, and protein substrate is understood in great detail, precisely how GroEL binds a wide array of nonnative proteins and assists their folding remains elusive.
Each GroEL subunit consists of an equatorial, an apical, and an intermediate domain (9,10). The equatorial domains of the subunits contain the ATP-binding site and both the N and C termini; they also mediate most inter-subunit contacts within and between rings. The apical domains form the entrance to the GroEL cavity and contain the residues involved in protein and GroES binding (13). Each intermediate domain forms a hinge-like connector between the equatorial and apical domains of that subunit. Studies of polypeptide recognition by GroEL using various techniques (reviewed in Ref. 14) suggest that the binding site involves the apical domains; more recent crystallographic studies suggest that helices H and I within the apical domain form the substrate-binding site (15)(16)(17)(18)(19)(20). Interestingly, these helices are the least well defined regions in the GroEL crystal structure and have relatively high thermal parameters (B-factors) even in the isolated apical domain (21,22), raising the possibility that the ability of GroEL to bind a diverse range of polypeptides may involve this conformational plasticity. Structural mobility and exposed hydrophobic sur-faces are features common to both the apical domains of GroEL and non-native polypeptides.
Here we investigate the properties of the apical domain from GroEL and its peptide binding characteristics using a range of analytical techniques. Specifically, the ability of electrospray mass spectrometry (ESI-MS) to simultaneously observe different protein conformations in solution was exploited to study the conformational dynamics of the GroEL apical domain and the ability of different conformers to bind peptide mimics of the co-chaperone GroES and the substrate protein, rhodanese. Parallel fluorescence emission studies were used to study these complexes and compare the binding properties of the apical domain constructs for each peptide with those of intact GroEL. The data suggest that the apical domain possesses different binding sites for the peptides in that no competition for binding individual peptides to the apical domain was observed. Finally, using surface plasmon resonance and fluorescence studies on a GroEL variant in which the GroES-binding groove is blocked by a covalently bound peptide, we show that the substrate and the GroEL mobile loop-binding sites are at least partially distinct even in the intact GroEL tetradecamer.

EXPERIMENTAL PROCEDURES
GroEL and Apical Domain Constructs of GroEL-The apical domain constructs: ApEL 1 (GroEL-(191-376)), N-His ApEL  containing an N-terminal hexahistidine tag), C-His ApEL (GroEL-(188 -381) containing a C-terminal hexahistidine tag), and ApTrap (ApEL G337S/I349E) were cloned, expressed, and purified as described previously (23,24). The concentration of all apical domain constructs was determined using an extinction coefficient of 4260 M Ϫ1 cm Ϫ1 at 280 nm, calculated by the method of Pace et al. (25). Wild-type GroEL and EL N229C (intact GroEL containing the mutation N229C and in which all the endogenous cysteine residues have been exchanged for alanine) were purified as described previously (26).
Crystallization-All crystallization screens were performed by vapor diffusion using the hanging drop method at 18°C. Sparse matrix screens were used to perform initial crystallization trials in 24-well plates with a well volume of 500 l (27). Lyophilized C-His ApEL was dissolved in, and extensively dialyzed against, buffer (20 mM Tris-HCl, 1 mM DTT, pH 7.2) prior to screening. The final protein solution was 10 mg ml Ϫ1 . Hanging drops contained equal volumes (2 l each) of the protein and well solutions. Crystals of C-His ApEL were obtained after 1-2 weeks with a well solution of 0.6 M NaCl, 14% (w/v) PEG 6000, 100 mM Tris-HCl, pH 8.5.
X-ray Data Collection and Structure Determination-Crystals of C-His ApEL were transferred to a solution of mother liquor containing 15% (v/v) glycerol as a cryoprotectant and rapidly frozen in liquid nitrogen prior to data collection. X-ray data were collected using a 30-cm MAR Research image plate detector at the Synchrotron Radiation Source Station 9.6 (Daresbury, UK) ( ϭ 0.87 Å). Indexing and extraction of the raw x-ray intensities were performed using the program MOSFLM (28). Intensities were merged and amplitudes were calculated using programs from the CCP4 program suite (29). Molecular replacement was carried out with the program AMORE (30) using an initial search model consisting of residues 188 -381 from the refined GroEL tetradecamer (21). The high resolution structure of an apical domain fragment closely resembling our construct, containing GroEL residues 191-376 and an N-terminal histidine tag sequence (22), was used subsequently as a model for further modeling and refinement. Initial model building was carried out with the program FRODO (31,32) and the structure was refined first at medium resolution using X-PLOR (33). Inspection of the resulting model using the program O (34) and full refinement at high resolution including all data was performed using the programs CNS (35) and REFMAC (36).
Helix D and Mobile Loop Synthesis-The peptides helix D (HD) of rhodanese (amino acid residues 248 -267) (37) (HD: Ac-RKGVTACHIA-LAAYLCGKPD-NH 2 ) and mobile loop (ML) of GroES (amino acid residues 13-32) (ML: Ac-KRKEVETKSAGGIVLTGSAA-NH 2 ) were chemically synthesized with a C-terminal amide on an automated peptide synthesizer (Applied Biosystems, Foster City, CA, model 430A) using standard Fmoc methodology and purified by reverse phase high performance liquid chromatography. The N terminus was modified with either an acetyl or dansyl group, the latter to provide a sensitive fluorescent probe for monitoring binding to the apical domain. The purity and molecular masses of the peptides were verified by analysis of a 10 ng l Ϫ1 solution in 1:1 (v/v) acetonitrile, 0.05% aqueous formic acid using positive ionization ESI-MS. The concentrations of dansylated peptides were determined using the extinction coefficient for a single dansyl group of 4500 M Ϫ1 cm Ϫ1 at 330 nm (38). The concentrations of non-dansylated peptides were estimated by dry weight.
Electrospray Ionization Mass Spectrometry-The apical domain constructs were analyzed at a concentration of 20 M by continuous infusion at a flow rate of 5 l min Ϫ1 into the electrospray ionization source of a Platform II (Micromass UK Ltd., Manchester, United Kingdom) single quadrupole mass spectrometer using a syringe pump (model 22, Harvard Apparatus, Holliston, MA). Where "native" electrospray conditions were used, the samples were dissolved in ammonium acetate (50 mM) at pH 7.5. Where "denaturing" electrospray conditions were used, the samples were analyzed in 1:1 (v/v) 0.1% aqueous formic acid, acetonitrile. In both cases the ionization source was maintained at 30°C and nitrogen was employed as both the nebulizing and drying gases at flow rates of 20 and 200 l h Ϫ1 , respectively. Positive ionization electrospray was used with a capillary voltage of 2.3 kV, the counter electrode set at 0 kV, and the sampling cone at 30 V. Data were acquired over the range m/z 500 -3000 at a scan speed of 10 s and processed using MassLynx software (Micromass UK Ltd.). An external calibration using horse heart myoglobin (molecular mass 16,951.5 Da; Sigma) was applied to ensure mass accuracy.
The m/z spectra displayed were smoothed mildly using a Sovitzy-Golay algorithm. The molecular mass (zero charge state) profiles illustrated were generated from the m/z spectra using the Maximum Entropy (39) software supplied with MassLynx.
Thermal Stability of the Apical Domain Constructs Monitored by ESI-MS-To determine the thermal stability of the apical domain constructs, the ionization source temperature was raised in a stepwise fashion from 30 -190°C. Apical Domain and Peptide Complex Formation Monitored by ESI-MS-The apical domain was mixed with the peptides HD or ML in ammonium acetate (50 mM) at pH 7.5 and the mixtures were incubated for 1 h at ambient temperature prior to ESI-MS analysis. The concentration of the apical domain was maintained at a constant 20 M although the concentration of ML was varied from 1 to 200 M and the concentration of HD was maintained at 50 M to maximize the proportion of the complex without incurring problems with aggregation. The ionization source temperature was maintained at 30°C and the sampling cone voltage at 30 V; these conditions were chosen to preserve protein-peptide complexation.
ilized to yield 12 mg of 98% pure material as verified by high performance liquid chromatography and mass spectrometric analysis.
Modification of GroEL N229C-A ϳ3-fold molar excess of SBP-Mal over EL N229C cysteines was applied to achieve quantitative modification. SBP-Mal (10 mM in Me 2 SO) was diluted 20-fold with H 2 O. 200 l of EL N229C (to give a final concentration 9 M) was added to 1 ml of 0.5 mM peptide solution and incubated for 1 h at 25°C. The reaction was stopped by the addition of ␤-mercaptoethanol (50 mM) and excess reagent was removed by gel filtration chromatography. Less than 10% free thiols were detected in EL N229C:SBP-Mal on treatment with maleimidosalicylic acid.
Prevention of Rhodanese Aggregation-Denatured rhodanese (100 M in 6 M guanidinium chloride) was diluted 200-fold into buffer (20 mM MOPS, pH 7.4, 100 mM KCl, 5 mM MgCl 2 ) in the absence or presence of 1.0 M EL N229C or EL N229C:SBP-Mal. Rhodanese aggregation was followed by turbidity measurements at 320 nm. Steady State Fluorescence-Steady state fluorescence emission was measured at 20°C on a spectrofluorimeter (model LS50B, PerkinElmer Life Sciences) using a 1-cm path length. Protein-peptide complexes were equilibrated for 5 min prior to acquiring fluorescence emission spectra. Fluorescence emission scans of dansylated peptides were recorded between 450 and 600 nm, using an excitation wavelength of 350 nm and a scan speed of 60 nm min Ϫ1 . Excitation and emission slit widths were typically between 5 and 10 nm. Peptide binding titrations were performed with similar parameters but in the time drive mode, measuring emission at a single wavelength. The emission wavelength used was 500 nm for HD and 535 nm for ML binding to ApEL, respectively. All titrations were performed by adding small volumes of the dansyl-ML or ApEL to a cuvette containing 2-3 ml of either ApEL (25 mM) or dansyl-HD (3 mM), respectively, in buffer (50 mM Hepes, pH 7.5, 2 mM DTT). After each addition, the solution was mixed thoroughly and allowed to equilibrate thermally for ϳ5 min. Titration of ApEL with the dansyl-ML required ϳ25-min equilibration times. The total volume of ligand added never exceeded 10% of the total volume. For each assay, a control experiment was performed by adding the ligand to a solution of buffer alone and was subtracted from the corresponding reading acquired in the presence of protein or peptide.
Fitting of Ligand Binding Data-All peptide binding curves were fitted using the program Grafit TM (Erithacus Software Ltd.). Single transition ligand-binding profiles were fitted with either a weak (Equation 1) or tight (Equation 2) ligand binding equation as appropriate (41). Typically, the weak binding equation was used under conditions where the dissociation constant was greater than the concentration of binding sites and the tight ligand binding equation was used under conditions where the concentration of binding sites was much greater than the dissociation constant.
Where F is the signal, F max is the maximum signal, K d is the dissociation constant, and L is the concentration of ligand.
Where F is the signal, F max and F min are the maximum and minimum signals, respectively, E 0 is the protein ligand-binding site concentration, K d is the dissociation constant, and L is the total ligand concentration.
Surface Plasmon Resonance-Mutant GroES 98C (42) (mutant cochaperone GroES with an additional cysteine at the C-terminal) was purified as described previously (26) and immobilized (300 resonance units) via a thioether linkage on a CM5 biosensor chip (Biacore 2000 SPR instrument, Biacore AB, Uppsala, Sweden) and the analysis was performed as described previously (43). Binding was followed using buffer (20 mM MOPS, pH 7.4, 100 mM KCl, 5 nM MgCl 2 , 2 mM ATP) at a flow rate of 20 l min Ϫ1 at 25°C. The concentration of GroEL was 250 nM.

RESULTS
Various GroEL apical domain fragments, mainly spanning amino acids 191-376 (with or without a hexahistidine tag) were expressed and purified for these studies (Fig. 1A). ApTrap contains two point mutations in GroEL-(191-376): G337S and I349E, which in intact wild-type (WT) GroEL result in loss of ability of GroEL to release bound polypeptide (13,44). Additionally, C-His ApEL is a slightly larger apical domain construct, encompassing residues GroEL-(188 -381), containing a C-terminal hexahistidine tag.
Crystal Structure Analysis of GroEL Apical Domains-Several crystal structures of GroEL apical domains have been obtained to date (9,19,22,45,46). To compare the structure of these domains with that of the larger construct C-His ApEL constructed here, crystals of C-His ApEL were grown as described under "Experimental Procedures." Crystals appeared within 1-2 weeks and belong to space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 48.6 Å, b ϭ 61.9 Å, and c ϭ 75.2 Å. A 98.8% complete data set was collected at cryogenic temperatures from a single crystal to 2.06-Å resolution. The molecular replacement procedure revealed a clear solution in both the rotation and translation searches. The correlation coefficient corresponding to the highest peak in the translation function was 0.600 and the initial R-factor after rigid-body optimization was 0.387 (47).
The crystal structure of the apical domain fragment, C-His ApEL (residues 188 -381), is well defined. The free R of the final refined model is 0.257 including 95% of all data between 20 and 2.06 Å. In the current model the deviations in ideal bond lengths and bond angles were 0.011 Å and 2.4°, respectively, and the average real-space correlation coefficient for all amino acid residues was 0.936. The backbone torsion angles of all residues are within the allowed regions of the Ramachandran plot (data not shown). The overall data collection and model refinement statistics are collated in Table I and the coordinates have been deposited in the Protein Data Bank as 1LA1.
C-His ApEL shares a virtually identical fold to its counterparts in both the intact GroEL tetradecamer as well as in the shorter domains 191-336, 191-345, 193-345, and 191-376 (9,16,19,22,46). A comparison of five independent models of the apical domain of GroEL, namely C-His ApEL (determined here), N-His ApEL (22), ApEL-(191-336)⅐peptide complex (19), intact GroEL (9), and a GroEL⅐ADP complex (45) is shown in Fig. 1B. The C-His ApEL model can be superimposed onto equivalent ␣-carbon atoms in the intact tetradecamer, the high-resolution N-His ApEL model, the ApEL-(191-336)⅐ peptide complex, and the GroEL⅐ADP complex with overall r.m.s. deviations of 0.72, 0.53, 0.56, and 1.57 Å, respectively. The ␣-carbon atoms in the ␤-sandwich core of the ApEL domain are virtually unchanged in the different models. The r.m.s. deviation of the core structure in the C-His ApEL and N-His ApEL models was only 0.32 Å. Similarly, the overlap between the core of the ApEL domains and the equivalent region in the intact EL tetradecamer was 0.65 Å. However, the most striking feature of the structural comparisons was that the local changes were significantly higher and map predominantly to helices H and I. Displacements of up to 2.0 Å can be found in this part of the ApEL domain (Fig. 1C). Consistent with previous observations (19,22,23) these data suggest that helices H and I, which have been implicated in substrate binding (15)(16)(17)(18)(19)(20), are locally dynamic.
ESI-MS of GroEL Apical Domain Constructs-We first investigated the characteristics of ApEL, the wild-type apical domain construct of GroEL (Fig. 1A) under varying conditions by ESI-MS analysis. Under denaturing conditions at pH 2 in a 1:1 (v/v) 0.1% aqueous formic acid, acetonitrile, the m/z spectrum of ApEL ( Fig. 2A) shows a single, wide distribution of charge states (n) ranging from n ϭ ϩ10 to ϩ29 consistent with the population of a single conformational species of a pure and highly denatured protein of measured molecular mass 20,239.5 Da (calculated mass 20,239.4 Da).
Under native electrospray ionization conditions (50 mM aqueous NH 4 OAc, pH 7.5) the m/z spectrum of the ApEL construct changes significantly (Fig. 2B). The range of charge states is wider (n ϭ ϩ7 to ϩ29) and more complex than anticipated for a folded protein of this molecular mass. Significantly, the number and range of charge state distributions observed is consistent with the protein existing in a number of distinct conformations at this pH. The first distribution, centered on charge state n ϭ ϩ9, represents ϳ30% of molecules, and suggests that these molecules are tightly folded. A second distri-bution of charge states is centered on approximately n ϭ ϩ21. These states represent ϳ60% of the protein molecules, indicating that a substantial proportion of apical domain molecules were expanded relative to the native state, even under neutral pH conditions. Finally, a third charge state distribution centered on n ϭ ϩ13 is observed (ϳ10% of the total protein), corresponding to a third, distinct, population of partially folded molecules. Similar multiple charge state distributions were observed under native conditions for the other apical domain constructs N-His ApEL, C-His ApEL, and ApTrap (data not shown). These data indicate that the apical domain was conformationally dynamic, co-populating multiple species with distinct properties at neutral pH.
To determine whether the different charge state distributions reflect molecules in rapid equilibrium or distinct conformational states of native and more highly denatured species, the effect of varying the ESI-MS conditions on the charge state distribution of ApEL was investigated. The effect of the sampling cone voltage on the charge state distributions for ApEL at pH 7.5 is shown in Fig. 3A. When the cone voltage is increased, the charge states corresponding to the most highly unfolded molecules centered on n ϭ ϩ21 shift to higher m/z and hence lower charge states, as might be expected for the "charge stripping" phenomenon of proteins observed with increasing sampling cone voltage (48). When the cone voltage is raised above 70 V, the protein signal becomes weaker and the signal to noise ratio decreases, most probably because of sample decomposition and possibly fragmentation under the harsh conditions. In contrast, the charge state distribution corresponding to more folded molecules centered on the n ϭ ϩ9 charge state changes little with increasing cone voltage until 190 V, and only then does this particular charge state distribution widen, consistent with partial unfolding at the higher cone voltage. This indicates a remarkable stability of this subset of molecules to cone voltage compared with other native proteins (49). The resistance of this distribution to increases in sampling cone voltage supports the notion that these molecules are natively folded in a stable structure.
In a second series of experiments the stability of ApEL with respect to ionization source temperature was studied (Fig. 3B). As described above, at pH 7.5, with an ionization source temperature of 30°C, multiple charge state series were observed. As the ionization source temperature is increased above 75°C, however, the proportion of unfolded protein molecules increases although the folded protein population decreases concomitantly. At an ionization source temperature of 150°C, virtually all of the protein molecules were unfolded. Similar behavior was exhibited by the other apical domain constructs shown in Fig. 1A (data not shown). These findings are consistent with the notion that the more highly folded populations of the domain unfold at increased electrospray solution temperature suggesting that the different charge state populations observed at lower temperatures represent distinct conformational states of apical domains that are co-populated and in equilibrium with each other at pH 7.5.
Polypeptide Binding to GroEL Apical Domains-ESI-MS offers a unique opportunity to investigate simultaneously the ability of different conformational species of the GroEL apical domain to bind cofactors and substrates. Two model peptides were selected for such binding studies. The first peptide is equivalent to the mobile loop sequence of GroES (residues 13-32) (50, 51), a region that has been shown to mediate the interaction between GroES and GroEL by binding to the GroEL apical domain between helices H and I in the intact GroEL-GroES structure. This region of GroES has been shown to be   conformationally dynamic using both NMR (50) and x-ray crystallography (52). Additionally, previous NMR studies have shown that the isolated ML peptide bound to GroEL adopts the same ␤-hairpin structure as in intact GroES (51).
The second peptide represents helix D of rhodanese, a stringent GroEL substrate. This helical region has been shown to bind to the surfaces of the apical domain in GroEL using cross-linking and proteolysis experiments (53). The HD peptide used in our studies is equivalent to amino acid residues 248 -267 of the sequence of intact rhodanese. This region encompasses the native ␣-helix, together with three flanking residues at the N and C termini to stabilize the helix and to produce a predominantly hydrophobic peptide with a net positive charge to facilitate its solubility. Accordingly, HD was found to be soluble at a concentration of 50 M at pH 7.5 and ϳ30% helical as judged by far UV CD (data not shown).
The ability of ApEL to bind ML or HD was first investigated using fluorescence studies. In these experiments, dansylated peptides (see "Experimental Procedures") were added separately to ApEL and the fluorescence emission spectra of the peptides were examined. The data indicate that the fluorescence emission intensities of the dansylated peptides increase (Fig. 4, A and B) and the max shifts to shorter wavelength (data not shown) upon binding of both peptides to the apical domain construct, suggesting that both peptides bind to the apical domains such that the dansyl group is located in a more hydrophobic environment. Titration data show that both peptides bind to ApEL with similar affinity, K D (ML) ϭ 2.7 Ϯ 0.7 M and K D (HD) ϭ 2.2 Ϯ 0.1 M; their affinity for the apical domain thus lies within the range commonly observed for peptides binding to GroEL (38,54,55). Strikingly, both peptides bind with a 1:1 stoichiometry, demonstrating the integrity of the apical domain protein and indicating that the isolated apical domain retains much of the peptide binding ability of intact GroEL. Difference spectroscopy using far UV CD showed that no significant change occurs upon binding HD to ApEL, demonstrating that this peptide, similarly to helix A from the same protein (55,56), binds to ApEL in a helical conformation (data not shown). Efficient binding of ML to ApEL was found to be slow and to require extended equilibration times in the range of 30 min. This relatively slow binding reaction could result from a slow association rate between ML and the apical domain or from a slow interconversion of apical domain conformations with differing affinities for ML.
In a second series of experiments, peptide binding to ApEL was also investigated using ESI-MS. Surprisingly, given the hydrophobic nature of substrate recognition by GroEL (reviewed in Refs. 11 and 14), measurable quantities of apical domain-peptide complexes could be detected by mass spectrometry under native conditions. Using ApEL at a concentration of 20 M and ML at a concentration of Ն40 M, the 1:1 complex between ML and the apical domain was partially preserved in the gas phase (Fig. 5A). However, the predominant mass species in the spectrum corresponded to uncomplexed protein and peptide. Similar observations were made for the complex between ApEL (20 M) and HD (50 M) under ESI-MS conditions (Fig. 5B). The dissociation of the complex in the gas phase accords with the hydrophobic nature of the noncovalent protein-ligand interactions, the strength of which is critically dependent on the presence of solvent. As hydrophobic interactions are predominantly solution dependent, there is little to sustain these interactions during the ESI process (57), where the removal of water is liable to cause the dissociation of apical domain-peptide complexes in the gas phase. Conversely, complexes that are stabilized by forces such as hydrogen bonds or electrostatic charge interactions are generally much better pre-served in the gas phase (58,59).
An interesting feature of this experiment is that the ESI-MS data show that only ApEL molecules in a folded or partially folded conformation are capable of gas-phase retention of bound peptides (Fig. 5, A and B). To estimate the percentage of the different apical domain conformations bound to ML and HD in the gas phase, the ESI-MS spectra were transposed onto a molecular mass profile (Fig. 6). To accomplish this, the m/z spectra were split into three regions encompassing the charge states associated with the folded (n ϭ ϩ8 to ϩ11), partially folded (n ϭ ϩ13 to ϩ17), and more highly unfolded (n ϭ ϩ20 to ϩ28) populations identified above. Each region was then subjected individually to Maximum Entropy (39) processing, thus producing three mass profiles (Fig. 6). By comparison of the relative ion currents we estimate that ϳ26% (based on relative peak areas) of the most highly folded apical domain species, ϳ14% of the partially folded apical domain species, and 0% of the more highly unfolded apical domain species are detected bound to the ML peptide in the gas phase. There is also evidence for a trace (Ͻ2%) of a ternary complex between the highly folded apical domain and two molecules of ML. Very similar results were obtained also for the HD peptide (data not shown).

Simultaneous Binding of Mobile Loop and Helix D Peptides
to GroEL Apical Domains-Although the binding site on the GroEL apical domain for ML and other peptides in an extended conformation has been mapped in detail by x-ray crystallography (19,22,51) there are currently no high resolution data highlighting the nature of the binding sites for helical substrates within the apical domain binding surface. NMR studies have suggested, however, that helices H and I are involved in the binding of helix A from rhodanese (56, 60) as well as intact proteins. To evaluate the extent of overlap of the binding sites for GroES and substrates in the apical domain of GroEL, a series of biochemical investigations was undertaken.
First, having established by fluorescence titrations that both ML and HD bind to ApEL (Fig. 4) with similar affinity, competition studies were undertaken using equilibrium fluorescence measurements. These were carried out by challenging complexes of ApEL or GroEL with dansylated peptides, with an excess of unlabeled competitor peptide. A decrease in the fluorescence intensity of the protein-dansylated peptide complexes would then be indicative of direct competition for peptide binding, whereas the fluorescence signal should be unperturbed in the absence of competition. The fluorescence emission spectrum of a preformed complex between ApEL (5 M) and dansylated HD (5 M) in the presence or absence of unlabeled ML (100 M) is shown in Fig. 7A. No decrease in signal was observed when excess unlabeled ML was added to the ApEL⅐HD complex, suggesting that ML and HD are binding to distinct sites on the apical domain surface. The inability of ML to displace ApEL-bound dansylated HD is an important result as it is uncomplicated by the possibility of domain movements that can occur upon ligand binding to intact GroEL (10,51). The preformed complex between intact GroEL (0.4 M) and dansylated HD (3 M) was also challenged with excess unlabeled ML (100 M) (Fig. 7B). Again the results show that ML does not displace the bound dansylated HD. The protein-dansylated HD complexes show a slight increase in fluorescence emission in the presence of ML, which may suggest that the latter binds to a site close to dansylated HD, causing minor changes in the environment of the dansyl fluorophore.
The reverse competition experiment of challenging preformed protein-dansylated ML complexes with excess HD could only be performed reliably with intact GroEL, because the signal change accompanying the binding of dansylated ML to the apical domain is relatively small in magnitude (Fig. 4, A  and B). The fluorescence emission spectrum of a preformed complex between GroEL (0.5 M) and dansylated ML (20 M) in the presence or absence of unlabeled HD (100 M) is shown in Fig. 7C. Again the data reveal that ML and HD do not compete for binding to the GroEL surface. The instability of the peptide complexes in the gas phase precluded similar experiments using ESI-MS.
In a second series of experiments, a mutant of intact GroEL (EL N229C) was prepared in which every endogenous cysteine residue was exchanged for an alanine residue and a new cysteine was introduced at Asn 229 (26). Cys 229 is strategically well positioned at one end of the GroES binding groove formed by helices H and I in the apical domain of GroEL. A C-terminal maleimide-derivatized variant of the "strong binding peptide" (SBP-mal) that was shown elsewhere to bind into the same binding groove as the GroES mobile loop (19) was covalently FIG. 6. ESI-MS molecular mass profiles of noncovalently bound ApEL⅐peptide complexes. The molecular mass (or zero charge) profiles generated by maximum entropy (39) processing techniques from the ESI-MS m/z spectrum of the mixture of ApEL and ML. A, the molecular mass profile generated from the m/z region covering the strongly folded apical domain (charge states n ϭ ϩ8 to ϩ11). B, the molecular mass profile generated from the m/z region covering the partially folded apical domain (charge states n ϭ ϩ13 to ϩ17). C, the molecular mass profile generated from the m/z region covering the more highly unfolded population (charge states ϩ20 to ϩ28). ApEL is observed in all populations, although the ApEL⅐ML complex is detected only in conjunction with the folded or partially folded states. coupled to Cys 229 to form EL N229C-SBP (see "Experimental Procedures"). Using a surface plasmon resonance-based binding assay to determine the GroES binding capacities of unmodified EL N229C and the peptide-bound mutant EL N229C-SBP, we found that unmodified EL N229C bound immobilized GroES 98C with an affinity similar to that of WT GroEL (26), whereas by contrast, the modified variant EL N229C-SBP was essentially incapable of cofactor binding (Fig. 8A). From these data we conclude that the covalently bound SBP peptide competes efficiently with the GroES mobile loop structures, and therefore that SBP immobilized at position Cys 229 assumes the same binding mode relative to the intact GroEL 14-mer as the co-crystallized SBP peptide relative to the isolated GroEL apical domain (19). In agreement with these GroES binding data we find that EL N229C-SBP does not refold chemically denatured rhodanese, a stringently GroEL/GroES-dependent model substrate (data not shown). Unmodified EL N229C on the contrary refolds this model substrate with the same efficiency as WT GroEL (26).
Finally we investigated whether blocking the GroES-binding site by SBP in the apical domain of EL N229C impaired sub-strate binding. Comparison of the activity of EL N229C and EL N229C-SBP to prevent rhodanese aggregation showed that EL N229C-SBP still maintained a substantial (ϳ70%) activity in substrate binding compared with the unmodified EL N229C (Fig. 8B). These data indicate that the binding sites of SBP and non-native proteins on the apical domain of GroEL overlap only partially. Rhodanese seems to interact with a more extensive region on the apical domain surface than that masked by SBP binding. This hypothesis is consistent with the fluorescence data shown above, suggesting that the peptides ML and HD, mimicking GroES and substrate, respectively, do not compete directly for the same binding site on the apical domain. In accord with this, further titration experiments showed that dansylated HD binds to EL N229C, but covalently bound SBP does not prevent dansylated HD binding to EL-N229C-SBP (Fig. 8C).   (23,46). The ESI-MS data presented here suggest that this behavior results from the denaturation of different conformational states within the ensemble as the temperature is raised, although noncooperative denaturation within a population cannot be ruled out. Consistent with this, thermal denaturation of the shorter apical domain construct 191-345 (46), in which there is no exchange of conformers, shows a single species at pH 7.0 and only a single set of resonances visualized by NMR (56).
The conformational dynamics observed in ApEL raise the intriguing possibility that substrate binding could occur by a mechanism akin to induced fit. In accord with this, only the partially folded and more highly folded populations bind the GroES and rhodanese substrate analogue peptides tightly, as judged by the survival of these complexes under ESI-MS conditions, yet stoichiometric binding is observed at a high concentration of peptides in solution. Furthermore, superposition of the x-ray structures of intact GroEL and of the various apical domain constructs with structures determined to date show that helices H and I, which are known to be involved in binding GroES and possibly also in the binding of some purported substrate analogues (15)(16)(17)(18)(19)(20)56), show distinct conformational variability. Such a model could account for the different substrates that bind to GroEL, including all ␣, all ␤, as well as mixed ␣/␤ proteins (13), and also a range of peptides with different secondary structural preferences (14,38,62,63).
The ability of the apical domain constructs to bind two very different peptides, one a substrate mimic and one a GroES mimic, albeit with similar binding affinity, accords with similar affinity of the intact protein for its co-chaperone and substrate proteins (reviewed in Ref. 14). Moreover the change in fluorescence observed and the ability of both peptide complexes to survive, at least partially, in the gas phase, suggest that the binding sites for the two peptides are similar in nature. In accord with this, reduced ␣-lactalbumin, denatured pepsin, and the helical peptide A from rhodanese have all been shown to bind close to helices H and I of GroEL, in line with this substrate sharing a similar binding site to that of GroES (60). Hydrophobic interactions have long been known to play an important role in the binding of substrates to intact GroEL (reviewed in Refs. 11 and 14); consistent with this, only a minority of peptide molecules remains bound to ApEL in the gas phase. Nonetheless, a measurable and reproducible proportion of bound peptides is detected in the gas phase using ESI-MS, in agreement with the observation that electrostatic and hydrogen bonding also play some role in binding, as observed in the x-ray structures of the N-terminal protease cleavage tag (22) and the SBP peptide (19) to ApEL.
One of the most interesting results to emerge from this study is the observation that HD and ML do not compete for the same binding site in ApEL, despite having similar binding affinities for the protein. Similarly, no competition was observed by fluorescence emission data when the peptides were added to intact GroEL. These data, which are not influenced by the dynamic movements and cooperativity known to complicate binding experiments in intact GroEL, indicate that the binding sites on the apical domain surface, at least for these small peptides, are distinct. These binding sites are also at least partially distinct in intact GroEL, because HD is able to bind to EL N229C-SBP, albeit with slightly reduced affinity and, despite being unable to bind GroES, EL N229-SBP is able to bind rhodanese. Taken together, therefore, the data suggest that displacement of polypeptide from the apical domain surface by GroES can proceed via an indirect removal of the polypeptidebinding surfaces through the large conformational changes induced in the GroEL apical domains upon GroES and nucle-otide binding, as well as by more direct ejection of the substrate protein into the central cavity upon GroES binding. The dynamic surface of the apical domain and of its partially unfolded substrates suggests that different proteins could bind to different sites on the apical domain surface and the manner by which the substrate is ejected into the central cavity could involve one or both of these mechanisms. Nonetheless, the conformational flexibility shown here directly for the apical domain surfaces using ESI-MS demonstrates that GroEL is exquisitely adapted to bind and fold its broad range of protein substrates.
The question remains, where might substrate proteins bind to the GroEL apical domain? Interestingly, although previous crystallographic studies of complexes with the apical domain have implicated helices H and I as the critical substrate-binding site, the peptide fragments bound in these cases show remarkable sequence similarity to the mobile loop of GroES (14). This raises the intriguing possibility that these peptides mimic GroES binding rather than substrate binding (14). Whereas there could be several binding sites for multivalent substrates within the dynamic apical domain surface, it is clear that significant conformational rearrangements would be required for a folded helix to bind between helices H and I, in a manner identical to that of the mobile loop of GroES. However, the data presented here suggest that HD of rhodanese binds in a helical conformation to a site distinct from that of the GroES mobile loop, in support of the view that intact substrate proteins could, most probably, bind to GroEL at multiple sites. Further experiments will be required to map the location of the binding site for HD and other substrates on the apical domain surface and to test further the relationship between GroES binding and substrate release during the GroEL functional cycle. Unfortunately the limited solubility of the HD sequence precluded studies of this kind, for example, using co-crystallization or NMR analysis. Nonetheless, the data presented here show clearly that substrate and GroES binding to GroEL could be distinct events, at least for some proteins, and highlight that there is still much to learn about the manner in which GroEL binds its substrate proteins and releases them in a manner commensurate for productive protein folding.