Contributions of the LPPVK Motif of the Iron-Sulfur Template Protein IscU to Interactions with the Hsc66-Hsc20 Chaperone System*

Hsc66 (HscA) and Hsc20 (HscB) from Escherichia coli comprise a specialized chaperone system that selectively binds the iron-sulfur cluster template protein IscU. Hsc66 interacts with peptides corresponding to a discrete region of IscU including residues 99–103 (LPPVK), and a peptide containing residues 98–106 stimulates Hsc66 ATPase activity in a manner similar to IscU. To determine the relative contributions of individual residues in the LPPVK motif to Hsc66 binding and regulation, we have carried out an alanine mutagenesis scan of this motif in the Glu98–Cys106 peptide and the IscU protein. Alanine substitutions in the Glu98–Cys106 peptide resulted in decreased ATPase stimulation (2–10-fold) because of reduced binding affinity, with peptide(P101A) eliciting <10% of the parent peptide stimulation. Alanine substitutions in the IscU protein also revealed lower activities resulting from decreased apparent binding affinity, with the greatest changes in Km observed for the Pro101 (77-fold), Val102 (4-fold), and Lys103 (15-fold) mutants. Calorimetric studies of the binding of IscU mutants to the Hsc66·ADP complex showed that the P101A and K103A mutants also exhibit decreased binding affinity for the ADP-bound state. When ATPase stimulatory activity was assayed in the presence of the co-chaperone Hsc20, each of the mutants displayed enhanced binding affinity, but the P101A and V102A mutants exhibited decreased ability to maximally simulate Hsc66 ATPase. A charge mutant containing the motif sequence of NifU, IscU(V102E), did not bind the ATP or ADP states of Hsc66 but did bind Hsc20 and weakly stimulated Hsc66 ATPase in the presence of the co-chaperone. These results indicate that residues in the LPPVK motif are important for IscU interactions with Hsc66 but not for the ability of Hsc20 to target IscU to Hsc66. The results are discussed in the context of a structural model based on the crystallographic structure of the DnaK peptide-binding domain.

and bacteria (8,9) indicate that Hsc66 and its eukaryotic homolog assist in the biogenesis of iron-sulfur proteins. Although the exact role of Hsc66 in this process is unknown, the importance of Hsc66 in iron-sulfur cluster assembly may arise from its interactions with the IscU protein (10). IscU is a small, dimeric protein that is capable of forming labile iron-sulfur clusters in vitro and is proposed to serve as a template for iron-sulfur cluster formation in vivo (10 -12). Recent studies have provided evidence that IscU behaves as a substrate for Hsc66 (13), and Hsc66 binding of IscU may serve to regulate the type or stability of iron-sulfur clusters formed on IscU.
Recently, we found that Hsc66 interacts with cellulose-bound peptides containing a conserved motif of IscU (residues 99 -103, LPPVK), suggesting that Hsc66 binds a single discrete region of the protein (14). In addition, a synthetic peptide corresponding to IscU residues 98 -106 (ELPPVKIHC) was capable of eliciting IscU-like effects on Hsc66. The Glu 98 -Cys 106 peptide stimulates Hsc66 ATPase activity with affinity similar to that of full-length IscU (14), whereas denatured proteins that bind Hsc66 but lack the LPPVK motif fail to stimulate Hsc66 ATPase activity (3). These results indicate that residues in the LPPVK motif play a key role in regulation of Hsc66 by IscU. Studies with a synthetic peptide selected using Hsc66 in affinity panning experiments, SLWPPVSGG, also suggest that the LPPVK motif is important for binding to and regulation of Hsc66 (14). This peptide contains three residues, Pro-Pro-Val, found in the LPPVK motif but exhibits reduced affinity for stimulation of Hsc66 ATPase activity and a decreased ability to affect Hsc66 chaperone activity (14). These results suggest that the LPPVK motif is an important determinant for Hsc66-IscU interactions, but the relative contributions of individual residues in the LPPVK motif to binding Hsc66 and regulation of Hsc66 ATPase activity are unknown.
The interaction of IscU with Hsc66 is enhanced by the J domain co-chaperone Hsc20, which directly binds to both IscU and Hsc66 (2,10). Together Hsc20 and IscU synergistically stimulate the ATPase activity of Hsc66 greater than 400-fold (10). In contrast to the IscU protein, however, the Glu 98 -Cys 106 peptide does not function synergistically with Hsc20, and Hsc20 does not interact with IscU peptides arrayed on a cellulose blot (14). These findings suggest that Hsc20 may bind a region of IscU not present in small, unstructured peptides and that this interaction is necessary for synergistic stimulation of Hsc66 ATPase activity. The contributions of individual residues of the LPPVK motif to direct interactions between Hsc20 and IscU or the ability of Hsc20 and IscU to synergistically stimulate Hsc66 ATPase activity remain unclear.
To better understand IscU interactions with Hsc66, we have investigated the effect of alanine substitutions in the LPPVK motif of the IscU protein and of the Glu 98 -Cys 106 peptide on interactions with Hsc66. In addition, the ability of IscU alanine mutants to bind Hsc20 and to be targeted to Hsc66 by Hsc20 was examined. Our findings indicate that three residues in the LPPVK motif, Pro 101 , Val 102 , and Lys 103 , play key roles in interactions with Hsc66. Targeting of IscU to Hsc66 by Hsc20 can override deficiencies in binding affinity caused by these changes, but substitution of Pro 101 or Val 102 decreases the maximal stimulation elicited by IscU in the presence of Hsc20. The importance of residues Pro 101 , Val 102 , and Lys 103 to IscU interactions with the substrate-binding domain of Hsc66 is discussed in the context of a structural model based on the crystal structure of the substrate-binding domain of DnaK.

EXPERIMENTAL PROCEDURES
Materials-E. coli DH5␣F'IQ cells were from Invitrogen, and BL-21(DE3)pLysS cells were from Novagen. Materials for peptide synthesis were from NovaBiochem. Enzymes for DNA manipulations were from Roche Applied Science or U. S. Biochemical Corp. Synthetic nucleotides were obtained from Sigma-Genosys. Bacterial growth media components were from Difco, and other reagents were from Sigma.
Mutants of IscU were expressed in an iscU strain of E. coli M6 1655, in which the IscU gene was disrupted by the cat gene (designated strain OD110; James Imlay, University of Illinois, Champagne-Urbana). The cells were grown under selection with chloramphenicol and ampicillin during expression of IscU. All of the mutants were purified as for wild-type IscU and displayed similar chromatographic properties. The protein concentrations were determined using molar extinction coefficients of 11,200 (M⅐cm) Ϫ1 for IscU and IscU mutants, 19,600 (M⅐cm) Ϫ1 for Hsc66, and 16,800 (M⅐cm) Ϫ1 for Hsc20 based on the average molar absorption of tryptophan and tyrosine residues (15)(16)(17).
Peptide Synthesis-Modified forms of the Glu 98 -Cys 106 peptide were synthesized using standard N-(9-fluorenyl)methoxycarbonyl chemistry (18). The peptides were purified by reverse phase high pressure liquid chromatography, and their compositions were confirmed by matrixassisted laser desorption ionization time-of-flight mass spectrometry. The peptide concentrations were determined by reaction of the Cterminal cysteine residue using Ellman's reagent (19).
ATPase Assays-Steady-state ATPase rates were determined at 23°C in HKM buffer (50 mM Hepes, pH 7.3, 150 mM KCl, 10 mM MgCl 2 ) containing 1 mM DTT and 0.5 mM ATP by measuring phosphate released using the EnzCheck coupled enzyme phosphate assay kit (20) (Molecular Probes) as previously reported (2,3,10,13,14,21). Under these conditions Hsc66 has a basal turnover number of Х0.10 min Ϫ1 . The error bars for all of the figures represent individual values observed in two separate experiments and are shown when they fall outside the symbols used.
Surface Plasmon Resonance Analysis-Surface plasmon resonance studies were carried out at 25°C using a Biacore 2000 instrument as previously described (10,13,14). Hsc66 in the presence of 1 mM ATP and 10 mM magnesium chloride was randomly cross-linked to the surface of the sensor chip by amine coupling. The experiments were conducted in HKM buffer containing 5 mM DTT, and maximal signals were measured during 1-2-min injections. Binding of IscU appeared to be specific because no interaction was observed using sensor chips prepared without Hsc66. Binding of IscU proteins was measured sequentially at each concentration to ensure that degradation of the sensor surface during the experiment did not contribute to differences observed. The curves shown represent a least squares fit of the data to a hyperbolic saturation function. The amount of immobilized Hsc66 capable of binding IscU was estimated by comparison of the response signals obtained following immobilization of Hsc66 (Х6000 relative response units), and those resulting from IscU binding corrected for the relative masses of the proteins, assuming that IscU binds as a dimer (10).
Isothermal Titration Calorimetry-A Microcal Omega microcalorimeter (Amherst, MA) was used to investigate the binding of IscU proteins to Hsc66 and Hsc20. Measurements were carried out at 25°C as previously described (10,13). For Hsc66 a series of 7-l injections of 1.5 mM IscU were made into a cell containing 1.348 ml of 0.1 mM Hsc66 in HKM buffer containing 3 mM DTT and 1 mM ADP. For Hsc20, measurements were made using 3 mM IscU and 0.2 mM Hsc20 in HKM buffer containing 5 mM DTT. The data were corrected for heat produced by the addition of buffer alone (Ͻ0.5 kcal/mol) and integrated heat caused by binding (Q inj ) are plotted versus the molar ratio of IscU to Hsc66 or Hsc20 in the titration cell. Injection of IscU proteins into buffer gave a small amount of heat of dilution (Ͻ0.5 kcal/mol); the data presented are not corrected for these effects.
Hsc66 Substrate-binding Domain Structural Model-A structural model of the Hsc66 substrate-binding domain was constructed based on the crystallographic structure of the substrate-binding domain of DnaK (Protein Data Bank code 1DKX) using the Swiss model program (22). The peptides were modeled into the substrate-binding domain and energy-minimized via simulated annealing using CNS (23). 2

Stimulation of Hsc66
ATPase by Alanine-substituted Peptides-In a previous study, we demonstrated that Hsc66 selectively interacts with cellulose-bound peptides containing residues 99 -103 (LPPVK) of IscU (14). In addition, a synthetic peptide corresponding to this region, Glu 98 -Cys 106 , was found to be capable of stimulating the ATPase activity of Hsc66 with an affinity similar to full-length IscU (14). To determine the contribution of individual residues in the LPPVK motif to interactions with Hsc66, peptides in which each residue of the motif was individually substituted with alanine were synthesized and assayed for their ability to stimulate the ATPase activity of Hsc66. Affinity panning experiments demonstrate that Hsc66 prefers peptides with aliphatic residues (14), and alanine was used to reduce side chain size while retaining the aliphatic character of Hsc66 substrates. Fig. 1 shows the stimulation of Hsc66 ATPase activity observed using 10, 40, and 200 M concentrations of each peptide. All of the alanine-substituted peptides were able to stimulate Hsc66 ATPase activity, but stimulatory activity for the alanine 1 The abbreviations used are: DTT, dithiothreitol; e.u., entropy units. 2 Atomic coordinate files for the Hsc66 structural models are available from the authors. peptides relative to the parent peptide was significantly reduced at the lower concentrations. At the highest concentration tested, however, all of the peptides except peptide (P101A), displayed stimulatory activity within 2-fold of the wild-type Glu 98 -Cys 106 peptide. These results suggest that whereas Leu 99 , Pro 100 , Val 102 , and Lys 103 are important for high affinity binding, the nature of the side chain is not critical for signaling between the substrate-binding domain and the ATPase domain. Peptide(P101A), in contrast, elicited very little stimulation at all of the concentrations tested, suggesting that the nature of the side chain at position 101 is critical for interactions with Hsc66.
Alanine Scanning Mutagenesis of the LPPVK Region of IscU-To ascertain the effects of substitutions in the LPPVK motif in the context of the folded protein IscU mutants L99A, P100A, P101A, V102A, and K103A were constructed using site-directed mutagenesis. All five mutants displayed chromatographic properties, solution molecular masses (Х26 kDa, corresponding to the IscU dimer), and circular dichroism spectra similar to wild-type IscU (data not shown). In addition, each of the mutants was able to form iron-sulfur clusters similar to wild-type IscU (10,24), suggesting that the overall structure of IscU was not affected by the substitutions and that none of these residues are essential for cluster formation (data not shown).
IscU Mutant Interaction with Hsc66⅐ATP-To investigate which residues of the LPPVK motif are important for IscU binding to the ATP-bound T state of Hsc66, we first investigated binding directly using surface plasmon resonance. Fig. 2 shows the results of titrations in which Hsc66 was immobilized on a sensor chip and exposed to different concentrations of the IscU mutants. Wild-type IscU bound with an apparent affinity of 23 M with the maximal signal observed corresponding to Х16% of the immobilized Hsc66. Each of the mutants displayed reduced binding affinity, and a decrease in the maximal signal was observed relative to wild-type IscU. The largest changes in apparent binding affinity were for IscU(K103A) (95 M) and IscU(P101A) (102 M). Because of the differences in maximal binding, however, it is difficult to interpret the results quantitatively. IscU behaves as a dimer in solution, and multidentate binding as well as surface immobilization effects may affect the kinetics of the IscU-Hsc66 interaction (10).
As a separate measure of binding to the ATP state of Hsc66, we examined the effects of IscU alanine mutants on the stimulation of Hsc66 ATPase activity. As shown in Fig. 3 all five mutants were found to stimulate Hsc66 ATPase activity in a concentration-dependent manner and elicit a maximal stimulation of Hsc66 ATPase activity equal or greater than that caused by wild-type IscU (ϳ7-fold). These findings suggest that substitutions in the LPPVK motif do not affect communication between the substrate-binding domain and the ATPase domain of Hsc66. However, each of these mutants displayed reduced apparent affinity compared with wild-type IscU. Alanine substitution at positions 99, 100, and 102 resulted in small (2-4fold) increases in the concentration necessary for half-maximal stimulation of Hsc66 ATPase, but substitution at positions 101 and 103 increased the concentration necessary by ϳ70and 15-fold, respectively. The ability of the IscU(P101A) mutant to stimulate Hsc66 at high concentrations suggests that the low level of stimulation observed using peptide(P101A) may reflect a decreased affinity for Hsc66.
IscU Mutant Binding to Hsc66⅐ADP-ATP hydrolysis by Hsc66 results in a conformational change that converts Hsc66 from the low affinity T state to the high affinity R state (3,13,21). To investigate whether substitution of Pro 101 , Val 102 , and Lys 103 affects binding to the high affinity R state of Hsc66, we measured binding affinities of IscU and IscU alanine mutants directly using isothermal titration calorimetry (Fig. 4). Under  with alanine decreased the affinity of IscU for Hsc66⅐ADP ϳ7and 3-fold, respectively, compared with wild-type IscU. These findings indicate that residues Pro 101 and Lys 103 , found to be important for IscU interactions with the ATP complex of Hsc66, are also important for binding to the ADP complex.
Hsc20 Effects on IscU Mutant Stimulation of Hsc66 ATPase-J domain co-chaperones behave as specificity factors that guide substrates to their cognate chaperone (25)(26)(27), and the interaction of Hsc20 with IscU is consistent with this behavior (10). Hsc20 enhances both the binding of IscU to Hsc66 (10) and the degree of stimulation of Hsc66 ATPase activity (3). To determine whether changes in the LPPVK sequence affect the ability of Hsc20 to target IscU to Hsc66, we investigated Hsc66 ATPase activity in the presence of Hsc20 over a range of IscU concentrations. A concentration of Hsc20 (50 M) was used that had previously been shown to elicit maximal synergistic stimulation (10). Fig. 5 shows that IscU and Hsc20 synergistically stimulate Hsc66 ATPase activity ϳ450-fold with halfmaximal stimulation occurring at Х5 M IscU. Hsc20 also enhanced the binding of each of the alanine mutants to Hsc66 with the concentration of each mutant necessary for half-maximal stimulation ranging from 4 to 8 M. These results indicate that substitutions in the LPPVK motif do not significantly affect the ability of Hsc20 to enhance the affinity of IscU for Hsc66. The striking enhancement in binding affinity for the P101A and K103A mutants, which exhibit very low affinity in the absence of Hsc20 (K m Ն 300 M), establishes that Hsc20 binding to IscU can override the effects of changes in the LPPVK motif.
Whereas all of the mutants displayed binding affinity similar to wild-type IscU, IscU(P101A) and IscU(V102A) were impaired in their ability to elicit maximal stimulation of Hsc66 ATPase activity. The maximal stimulation with IscU(P101A) in the presence of Hsc20 was only 24% of that of wild-type IscU, and the maximal stimulation with IscU(V102A) was 58%.
These findings indicate that the nature of the amino acid side  IscU Mutant Interactions with Hsc20 -The enhanced binding affinity of Hsc66 for IscU mutants in the presence of Hsc20 could arise from an increase in the affinity of these mutants for Hsc20. To investigate this possibility, the binding affinities of Hsc20 for the two IscU mutants that exhibited the greatest enhancement in affinity for Hsc66 in the presence of Hsc20, IscU(P101A) and IscU(K103A), were examined using isothermal titration calorimetry (Fig. 6). Neither mutant displayed enhanced affinity compared with wild-type IscU (K d ϭ 28 M). IscU(P101A) bound Hsc20 with an affinity of Х34 M, and IscU(K103A) exhibited an affinity of Х76 M. These results indicate that enhanced Hsc66 affinity for IscU(P101A) and IscU(K103A) is not a result of increased affinity between the mutants and Hsc20. The enhanced binding affinity elicited by Hsc20 may reflect the ability of Hsc20 to act as a scaffold that positions IscU in the substrate-binding domain of Hsc66.
Effect of the V102E Substitution on IscU Interactions with Hsc66 and Hsc20 -In addition to IscU, diazotrophpic organisms contain a second iron-sulfur template protein, NifU, that functions in the assembly of the nitrogenase protein (28,29). The N-terminal regions of NifU proteins display sequence homology to IscU proteins, but there are no known chaperones associated with iron-sulfur cluster assembly specific to the nitrogen fixation machinery (5,12). Comparison of IscU proteins with NifU proteins reveals that the valine residue in the LPPVK motif is replaced by glutamic acid (LPPEK) in NifU sequences. To investigate the effect of this substitution on interactions with Hsc66, we prepared the IscU(V102E) mutant. The V102E protein exhibited similar general properties to the wild-type protein (chromatographic behavior, dimerization, and iron-sulfur cluster formation; data not shown), indicating that this substitution does not have major effects on IscU structure.
In contrast to the IscU(V102A) mutant, IscU(V102E) failed to stimulate the ATPase activity of Hsc66 even at concentrations as high as 800 M (Fig. 7). This suggests that introduction of a negative charge at this position significantly impairs the ability of IscU to interact with the T state of Hsc66. To deter-mine whether this change affects the ability of IscU to bind to ADP complex of Hsc66, we carried out isothermal titration calorimetry experiments. As shown in the inset to Fig. 7, no significant enthalpic changes were observed upon titration into Hsc66⅐ADP, suggesting that the mutant protein may not bind the R state of Hsc66. We also investigated IscU(V102E)-Hsc66 interactions using surface plasmon resonance, and these experiments also failed to reveal binding to either the ATP state or the ADP state (data not shown). Together these results indicate that replacement of Val 102 with glutamic acid greatly weakens the ability of IscU to interact with both conformational states of Hsc66. This finding is consistent with earlier affinity panning experiments indicating that acidic residues are disfavored in Hsc66 substrates (14).
To determine whether Hsc20 could overcome the effects of the V102E substitution, we carried out Hsc66 ATPase assays in the presence of saturating levels of Hsc20 (Fig. 6). Under these conditions IscU(V102E) was capable of stimulating Hsc66 ATPase activity, and the concentration necessary for half-maximal stimulation (K m ϭ 2 M) is similar to that seen for wildtype IscU in the presence of Hsc20 (Fig. 5). However, the maximum stimulation observed (20-fold) is much lower than that elicited by wild-type IscU (455-fold) or IscU(V102A) (262fold). To determine whether the decreased stimulation might reflect an altered interaction with the co-chaperone, we examined the binding of IscU(V102E) to Hsc20 using isothermal titration calorimetry. As shown in the inset to Fig. 7, the stoichiometry and binding affinity observed are similar to those found for wild-type IscU (Fig. 6), indicating that the reduced activity observed likely results from altered interactions with Hsc66. Thus, although Hsc20 targeting can overcome the effect of the V102E substitution on the affinity of Hsc66 for IscU, introduction of a negative charge in the LPPVK motif has dramatic effects on communication between the substratebinding domain and the ATPase domain. binding to Hsc66 (q) and to Hsc20 (E). The solid line represents the best fit curve to the data for IscU(V102E) binding Hsc20 assuming 1.05 binding sites, K d Х 29.8 M, ⌬H Х Ϫ2.6 kcal/mol, and ⌬S Х 12.1 e.u.

DISCUSSION
The results described herein provide new evidence that residues in the conserved LPPVK motif corresponding to residues 99 -103 of IscU are important for interactions with Hsc66. Studies with synthetic peptides revealed that substitution of alanine for any of the residues in this motif decreased the ability to stimulate the ATPase activity of Hsc66 at low concentrations. The largest effect on peptide-induced stimulation of Hsc66 ATPase activity was observed when the central proline of the LPPVK motif (corresponding to Pro 101 of IscU) was replaced with alanine, suggesting that this residue is a key determinant for peptide stimulation of Hsc66. Studies of alanine mutants of IscU provided additional information about the roles of individual residues in the context of the natural protein substrate. Substitution of each of the residues of the recognition motif with alanine reduced the affinity of IscU for Hsc66⅐ATP as evidenced by surface plasmon resonance binding studies and the increased concentrations necessary for halfmaximal stimulation of Hsc66 ATPase. The largest changes were observed with IscU(P101A), IscU(V102A), and IscU(K103A). The P101A and K103A mutants also exhibited decreased affinity for the R state of Hsc66. The decrease in binding affinity for Hsc66⅐ADP displayed by these mutants was not as pronounced as for the ATP-bound states but establishes that Pro 101 and Lys 103 are important for IscU binding to both of the conformational states of Hsc66. Because initial binding of IscU to Hsc66 in vivo is most likely to occur with the ATPbound state, the larger effects observed in the ATPase assay are more relevant to the physiological interaction between the proteins.
Although substitutions in the LPPVK motif decrease binding affinity of IscU for Hsc66, the addition of Hsc20 can override this deficiency. In the presence of Hsc20, all of the mutants displayed concentrations for half-maximal stimulation (4 -8 M) of Hsc66 ATPase similar to wild-type IscU. Hsc20 targeting of the IscU(V102E) mutant also occurred with affinity similar to wild-type IscU, even though no binding of this mutant could be detected in the absence of Hsc20. These findings indicate that the exact sequence of the LPPVK motif does not affect the affinity enhancement caused by Hsc20 and suggest that Hsc20 binding to IscU will be a key determinant of Hsc66 interactions with IscU. Hsc20 may function as a scaffold to aid in positioning IscU on the substrate-binding domain of Hsc66. A role for Hsc20 in specific targeting of IscU to Hsc66 (10, 13) is consistent with the notion that the functional specificity of Hsp70-type chaperones is determined to a large degree by interactions of their ATPase domain with other components of the cellular machinery (30).
Even in the presence of Hsc20, however, mutations at two positions, Pro 101 and Val 102 , caused a decrease in the maximal stimulation of Hsc66 ATPase elicited by IscU. Replacement of Pro 101 with alanine resulted in a 76% decrease in maximal stimulation, and alanine substitution at Val 102 led to a 42% decrease. Introduction of a negatively charged residue had an even greater effect. The IscU(V102E) mutant gave Ͻ5% of the maximal stimulation of wild-type IscU in the presence of Hsc20. These findings indicate that although residues at position 101 and 102 do not significantly alter IscU binding affinity for Hsc66 in the presence of co-chaperone, they are important for communication between the ATPase domain and the substrate-binding domain of Hsc66. Specific interactions between the side chains of Pro 101 and Val 102 of IscU and the substratebinding domain of Hsc66 appear to be necessary for the allosteric regulation of Hsc66 ATPase activity.
A Model for the Interaction of IscU with Hsc66 -The specificity of Hsc66 for the LPPVK motif and the effects of mutations in this region raise questions regarding how the substratebinding domain of Hsc66 recognizes this amino acid sequence. Crystallographic studies of E. coli DnaK (31) and NMR studies of DnaK (32)(33)(34) and rat Hsc70 (35) have revealed that bound peptide exists in an extended conformation and lies in a groove in the ␤-sandwich subdomain of the substrate-binding domain. Interactions occur between both backbone and side chain atoms of the bound peptide and the chaperone with the most extensive contacts occurring with five central residues at peptide positions Ϫ2, Ϫ1, 0, 1, and 2. The side chain of the central residue 0 projects into a hydrophobic pocket, and the peptide backbone at this position is covered by an arch or bridge formed by residues from ␤-strand loops of the chaperone. Structural features that contribute to peptide binding specificity are not well understood, but comparisons of sequences among different Hsp70 species (36) and mutational studies on DnaK (37,38) suggest that the identity of residues forming the hydrophobic pocket and the arch contribute to selectivity.
The substrate-binding domain of Hsc66 exhibits 43% sequence identity and 60% similarity with that of DnaK, suggesting that Hsc66 is likely to interact with peptides in a manner similar to DnaK (1). To identify structural differences between Hsc66 and DnaK that might contribute to their different peptide recognition specificities, we used the Swiss model program (22) 2 to construct a model of Hsc66 based on the crystal structure of DnaK (31). Fig. 8 shows the structure of the substratebinding domain of DnaK complexed with the synthetic heptapeptide NRLLLTG. This view is as in Fig. 2 of Ref. 31 and shows the peptide oriented with its N terminus facing out of the plane on the "front side" of the ␤-sandwich subdomain and the C terminus projecting into the plane on the "back side" (hereafter designated "N 3 C orientation"). The central leucine residue of the peptide projects downward into a hydrophobic pocket at position 0. Residues Met 404 (from ␤-loop L 1,2 ) and Ala 429 (from ␤-loop L 3,4 ) form an arch over the peptide backbone and make contact with the side chains of leucine residues at positions Ϫ1 and 1. A model for the substrate-binding domain of Hsc66 generated in the absence of a peptide substrate is shown in the lower panel of Fig. 8. The general structure of the Hsc66 model is similar to that of the DnaK template, but there are a number of dissimilarities that may contribute to differences in the peptide binding specificity of the two chaperones. A phenylalanine residue (Phe 426 ) occurs in the arch of Hsc66 in place of Ala 429 in DnaK, and the phenyl side chain projects into the substrate binding cleft. Hsc66 also has a methionine (Met 433 ) at the base of the pocket forming position 0, whereas the corresponding residue in DnaK is a valine (Val 436 ). The bulky side chains of Phe 426 and Met 433 partially block the peptide binding site in the Hsc66 model, reducing its size compared with that observed in the DnaK-peptide complex. There are also differences in surface charge distribution between Hsc66 and DnaK. The front side of Hsc66 contains acidic residues (Asp 422 and Glu 406 ), whereas this side of DnaK contains mostly nonpolar side chains near the region the peptide exits the binding groove (Fig. 8). The back side of Hsc66 has basic residues (Arg 453 and Arg 457 ), whereas this side has neutral or acidic residues (Gln 456 and Asp 460 ) in DnaK (31). Differences in surface charge distribution combined with differences in residues comprising the peptide pocket may contribute to the differences in substrate specificity observed for the two chaperones (3,39) and to the failure of Hsc66 to complement the phenotype of dnaK Ϫ cells (39,40).
To better understand the interaction of Hsc66 with IscU, we modeled a peptide corresponding to IscU residues 98 -104 (ELPPVKI) into the Hsc66 substrate-binding domain. The conformation of this region of the IscU protein is not known, but it is unlikely to be highly structured based on the finding that the Glu 98 -Cys 106 peptide is capable of stimulating Hsc66 ATPase with similar affinity to that of full-length IscU (14). The ELP-PVKI peptide was modeled in an extended conformation similar to that observed for bound peptide in the crystal structure of DnaK. Because Pro 101 is the central residue of the peptide and because of its importance for binding, the equivalent peptide residue was placed in central position 0. The peptide was initially aligned in an orientation similar to that of the NR-LLLTG peptide bound to DnaK (N 3 C; Fig. 8), and the model was subjected to energy minimization using simulated annealing (23). Fig. 9 (A and B) shows two views of the calculated model of the complex. No large structural changes occur in the peptide, which remains in an extended conformation, or in the backbone structure of Hsc66. However, a number of small conformational rearrangements occur in Hsc66 that allow the peptide to be accommodated in the substrate groove. The major change involves rotation of the side chains of Phe 426 and Met 401 out of the cleft to form an arch directly above the backbone of the central peptide residue equivalent to Pro 101 at position 0. The most extensive interactions of Hsc66 occur with this proline and the two adjacent residues corresponding to IscU Pro 100 and Val 102 . The side chain of the central proline lies in a hydrophobic pocket, whereas the side chains of the adjacent proline (position Ϫ1) and valine (position 1) residues are turned upward and make van der Waals' contacts with the arch residue hydrophobic side chains. These interactions are consistent with the observed preference of Hsc66 for nonpolar residues at these positions. The steric restrictions imposed by the arch and by the hydrophobic pocket act to fix this part of the peptide in an extended conformation and force the rotation such that the side chain of proline 0 is directed downward and the side chains of proline Ϫ1 and valine 1 project upward. The extended transconformation favored by proline may contribute to the observed preference of Hsc66 for peptides containing proline in central positions (14). Interactions with residues in the Ϫ2 and 2  positions are less extensive, and the side chains of these residues are more exposed to solvent. The side chains of the two charged residues of the peptide, glutamate at position Ϫ3 and lysine at position 2, are located where the peptide exits the substrate binding groove and lie at the protein surface. As modeled, the negatively charged side chain of glutamate is directed away from acidic residues (Asp 422 and Glu 406 ) on the front side, and the positively charged side chain of lysine 2 points away from basic residues (Arg 453 and Arg 457 ) on the back side of the substrate-binding domain.
In addition to modeling peptide in the N 3 C orientation observed for the DnaK-peptide complex, we also used a similar procedure to model the peptide in the reverse, C 3 N orientation. There are a number of small differences in the structure of the Hsc66-peptide complex in this orientation, but the resulting model (Fig. 9C) has many of the same general features of the N 3 C model, including its extended conformation, the register of the residues within the binding groove, and the orientation of the side chains of the central residues. In the C 3 N orientation, the side chains of proline Ϫ1 and valine 1 are positioned on opposite side of the Hsc66 arch and make different contacts with residues Phe 426 and Met 401 compared with the N 3 C model. In addition, in the C 3 N orientation glutamate Ϫ3 and lysine 2 exhibit favorable electrostatic interactions at the surface of the Hsc66 substrate-binding domain in the C 3 N model. Glutamate Ϫ3, equivalent to Glu 98 of IscU, is able to form a hydrogen bond with the side chain of Arg 453 on the back side, and lysine 2, equivalent to Lys 103 of IscU, is able to interact with the side chain of Glu 406 on the front side of the substrate-binding domain. This can be contrasted with the N 3 C model in which Glu 98 was located on the acidic front side and Lys 103 was located on the basic back side. Because of the complexity of the model structures, however, it is not possible to quantitatively distinguish between the predicted stability of the C 3 N and N 3 C peptide orientations based on energetic calculations, and it is possible that both types of complexes can occur. It is also possible that interactions of other parts of the IscU protein with Hsc66 and/or Hsc20 may play a key role in dictating which orientation is preferred. The two models, however, predict different behaviors in response to alterations of residues of either IscU or Hsc66, and it may be possible to distinguish between them by carrying out mutagenesis experiments. Our finding that the IscU(K103A) mutant exhibits reduced affinity for Hsc66 suggests a specific role for Lys 103 in stabilizing the complex and appears to favor the reverse C 3 N orientation, but additional studies will be required to define the exact nature of the interactions involved. It will also be interesting to investigate the effects of changes in specific residues of the Hsc66 substrate-binding domain on peptide binding affinity and ascertain which residues determine the ability of the chaperone to discriminate between different substrate proteins.