The Small Heat-shock Protein IbpB from Escherichia coli Stabilizes Stress-denatured Proteins for Subsequent Refolding by a Multichaperone Network*

The role of small heat-shock proteins inEscherichia coli is still enigmatic. We show here that the small heat-shock protein IbpB is a molecular chaperone that assists the refolding of denatured proteins in the presence of other chaperones. IbpB oligomers bind and stabilize heat-denatured malate dehydrogenase (MDH) and urea-denatured lactate dehydrogenase and thus prevent the irreversible aggregation of these proteins during stress. While IbpB-stabilized proteins alone do not refold spontaneously, they are specifically delivered to the DnaK/DnaJ/GrpE (KJE) chaperone system where they refold in a strict ATPase-dependent manner. Although GroEL/GroES (LS) chaperonins do not interact directly with IbpB-released proteins, LS accelerate the rate of KJE-mediated refolding of IbpB-released MDH, and to a lesser extent lactate dehydrogenase, by rapidly processing KJE-released early intermediates. Kinetic and gel-filtration analysis showed that denatured MDH preferentially transfers from IbpB to KJE, then from KJE to LS, and then forms a active enzyme. IbpB thus stabilizes aggregation-prone folding intermediates during stress and, as an integral part of a cooperative multichaperone network, is involved in the active refolding of stress-denatured proteins.

The small heat shock proteins IbpA and IbpB from Escherichia coli are two sequence-related 14-and 16-kDa proteins, respectively, co-transcribed during stress by the bacterial heatshock transcription factor 32 (14). IbpA and IbpB share a low sequence homology in their C-terminal region with sHSPs from yeast, plants, and mammals, including ␣B-crystallines (14 -19, 2). Furthermore, they seem to be distantly related to other bacterial chaperones such as PapD and Caf1M. Based on the resolved x-ray structure of the PapD chaperone and on sequence homology, a model has been proposed where the threedimensional structure of IbpB resembles that of immunoglobulins (20).
In E. coli, IbpA and IbpB are found associated with endogenous proteins that aggregate intracellularly during heat-shock (21) and with non-native recombinant proteins in inclusion bodies (14). They are implicated in the solubilization of protein aggregates after stress (21). In vitro, IbpB is a large globular complex, whose oligomeric state may vary at different temperatures and ionic concentrations. 2 Similarly to other small HSPs, light-scattering measurements indicate that IbpB can specifically bind thermally or chemically denatured proteins such as MDH, LDH, and citrate synthase and thus prevent their irreversible aggregation (not shown).
We show here that IbpB can bind and stabilize denatured proteins, and furthermore, deliver them to the DnaK/DnaJ/ GrpE (KJE) chaperone system for subsequent active ATP-dependent refolding. KJE-mediated refolding of IbpB-stabilized proteins can be further activated by LS chaperonins, demonstrating cooperation between the different chaperone systems. IbpB can therefore function as a primordial protein-binding element of a multichaperone network, involved in stabilizing and active refolding of stress-denatured proteins.

EXPERIMENTAL PROCEDURES
Phage Complementation-Wild type and E. coli B mutant T 850 (22) were used for in vivo complementation assays of T4 morphogenesis as described in Ref. 23. Plasmids pBF3-IbpA and pBF3-IbpB were from Dr. C. Hergersberg, Boehringer Mannheim GmbH, and pSE420 was from Invitrogen Inc.
Proteins-IbpB was purified as follows: late log phase E. coli cells, containing plasmid pBF3-IbpB in LB medium and chloramphenicol (50 g/ml), were incubated for an additional 12 h at 37°C in the presence of isopropyl-1-thio-␤-D-galactopyranoside (100 g/ml), collected, and resuspended in 50 mM Tris-HCl, pH 7.5, 150 mM KCl, 20 mM MgAc 2 (Buffer A), containing 5 g/ml leupeptin. Cells were disrupted five times at 25°C in a French Press at 900 p.s.i. Soluble proteins in the 30,000 ϫ g supernatant were incubated with 6% polyethylene glycol 6000 (Merck) for 30 min at 4°C. The protein pellet from 30,000 ϫ g was resuspended in buffer A, filtered through a 0.45-m filter (Schleicher & Schuell), and separated by gel filtration (0.5 ml/min) on a semi-preparative Superose 6B column (Pharmacia) in buffer A. The fractions collected between 6.0 and 7.5 ml (above 2.10 6 daltons), which were highly enriched with soluble oligomeric IbpB, were applied to a resource-Q column (Pharmacia), equilibrated with buffer A. Elution was carried out with a linear gradient from 150 to 500 mM KCl in Buffer A. The fractions eluting at 230 Ϯ 30 mM were collected and frozen as 34 -45 M (protomer) 95% pure IbpB stock solutions. GroEL 14 and GroES 7 were purified as described (24). Pig heart mitochondrial MDH and hog muscle LDH were from Boehringer Mannheim; hexokinase and pyruvate kinase from Sigma. Plasmid-encoded DnaK, DnaJ, and GrpE were overexpressed in E. coli and purified as described (25,26). Protein concentrations were determined by the Bradford protein assay (Bio-Rad). In this study, all chaperone concentrations were expressed in terms of the individual protomers.
3 H Labeling-Native MDH was labeled with NaB( 3 H) 4 as described in Ref. 11. After purification by gel filtration (Superose 6B column, Pharmacia), 3 H-labeled MDH was found to be as active as unlabeled MDH. During heat-denaturation and chaperone-assisted refolding, 3 H-labeled MDH behaved indistinguishably from unlabeled MDH (not shown).
Enzymatic Assays-The activity of MDH was measured as described in Ref. 27, at 25°C in 150 mM potassium phosphate buffer, pH 7.5, 10 mM dithiothreitol, 0.5 mM oxaloacetate, and 0.28 mM NADH (Sigma). The activity of LDH was measured as described in Ref. 28 in 100 mM Tris, pH 7.5, 0.5 mM pyruvate, and 0.28 mM NADH. The time-dependent oxidation of NADH by MDH or LDH was monitored at 340 nm. Native MDH dimers remain stable and active in solutions above a concentration of 6 nM (protomers) (29). The apparent rates of protein refolding were calculated from the time-dependent changes in the enzymatic activity of chaperonin solutions containing more than 20 nM enzyme.

IbpA and IbpB Can Suppress a Phage
Growth groEL Mutation in E. coli-Wild type E. coli B cells are highly sensitive to T4 phages (23). In contrast, the mutant strain T850, which contains a chromosomal groEL with a R268C mutation, 3 do not support T4 growth (Fig. 1). We found that phage growth was specifically restored in T850 cells containing plasmid-encoded ibpA or ibpB genes ( Fig. 1), however, only in the presence of a higher phage titer (10 8 particles/ml) than in wild type cells.
IbpB-stabilized MDH Can be Specifically Refolded by KJE ϩ ATP-When exposed to 47°C, the thermolabile enzyme MDH is inactivated at an apparent rate of 0.13 min Ϫ1 (11) and forms large insoluble protein aggregates that do not refold spontaneously after the heat shock (not shown). In contrast, when exposed to heat stress in the presence of a 4-fold molar excess of thermostable IbpB, MDH was inactivated at a slower rate (0.10 min Ϫ1 ) and formed a stable soluble IbpB⅐MDH complex. 2 We found that the IbpB⅐MDH did not significantly reactivate, even after a 5-h incubation at room temperature ( Fig. 2A). However, when IbpB⅐MDH was supplemented with KJE chaperones and ATP, a significant amount (10%) of native MDH was slowly recovered, at a slow rate of ϳ0.7 nM min Ϫ1 . This KJE-mediated refolding of IbpB⅐MDH was a specific reaction, as it did not occur when IbpB was absent during the heat denaturation (Fig.  2, A and B), or when ATP or GrpE were absent during renaturation (not shown).
KJE-mediated Refolding of IbpB⅐MDH Is Activated by LS-When IbpB⅐MDH was supplemented with LS chaperonins and ATP, only insignificant amounts of MDH were reactivated (1.5%), in addition to the small amounts generated by spontaneous refolding (2.5% within 4 -5 h, Fig. 2A). Remarkably, when KJE and LS chaperones and ATP were concomitantly added to IbpB⅐MDH after the heat shock, a major fraction (40%) of the MDH was actively recovered at an apparent rate 5.5 times higher (3.7 nM min Ϫ1 ), than without LS ( Fig. 2A). Interestingly, a significant amount (10%) of MDH that was heat-inactivated without IbpB was also recovered at a slow rate of 0.8 nM min Ϫ1 , but only in the presence of the two chaperone systems KJE and LS (Fig. 2B). The strong dependence of the refolding rates and yields of MDH, on the presence of IbpB during denaturation, could be used as a first functional assay 3 C. Weiss, S. Gross, and P. Goloubinoff, personal communication. for a small heat-shock protein under stringent conditions.
The yields of the reaction depended on the concentration of IbpB present during the heat-denaturation of MDH (Fig. 3A). When denatured in the presence of a 20-fold molar excess of IbpB, up to 83% of the denatured MDH was recovered by KJE ϩ LS after the stress. Half of the MDH molecules were recovered when denatured in the presence of a 5.7 molar excess of IbpB (Fig. 3A). The refolding rate of IbpB⅐MDH was halfmaximal when the LS concentration (fixed 1:1 ratio between L and S) was one-fifth that of DnaK (4, 4, and 0.5 M of K, J, and E, respectively), or when the KJE concentration (with a fixed molar ratio of 4:4:0.5) was half that of the LS (4:4 M each) (Fig.  3B). This suggests that at near equimolar DnaK and GroEL concentrations, as in the cell (30), the KJE chaperones are more rate-limiting for this reaction than the LS chaperonins.
The apparent refolding rate of heat-preformed KJE⅐MDH complexes was 8.6 times slower than the refolding rate of heat-preformed LS⅐MDH complexes (Table I). This difference was exploited to address, by kinetic measurements, the order of events during multichaperone IbpB/KJE/LS-mediated refolding reactions. The refolding rates of heat-preformed IbpB⅐MDH complexes (Fig. 4A) was compared with heat-preformed KJ⅐MDH and GroEL⅐MDH complexes (Fig. 4B), in the presence of various co-chaperones added either immediately or 53 min after the heat shock.
Significant refolding of IbpB⅐MDH was observed only when supplemented after the heat shock with the three chaperones KJE and ATP, however, at a rate which was about half (  While LS and ATP alone did not directly assist the refolding of IbpB⅐MDH ( Fig. 2A; Table I Remarkably, delaying the addition of E ϩ LS to IbpB⅐MDH which was incubated 53 min with ATP and KJ, increased the rate of refolding from 1.96 to 2.99 nM min Ϫ1 (Fig. 4A; Table I). Likewise, delaying the addition of GroES to IbpB⅐MDH which was incubated 53 min with ATP and KJEL, more than doubled the refolding rates from 1.96 to 4.25 nM min Ϫ1 (Fig. 4A; Table  I). Thus, the rate-limiting dissociation of MDH from IbpB can be overcome by a delay during which it is allowed to transfer on KJ and then on GroEL. Whereas GrpE-mediated refolding of KJ-MDH was inhibited 3-fold in the presence of added free (equimolar) IbpB (Table I, (Table I,  7.2 3 7.64 nM min Ϫ1 ). This indicates that, at this late stage of refolding, GroEL has a much higher affinity to the protein than KJE and IbpB.
The refolding of urea-denatured LDH displayed a similar dependence on the various components of the IbpB/KJE/LS chaperone network. LDH is a thermostable enzyme, which therefore requires first to be denatured in 5 M urea (11) before being diluted into IbpB, LS, or KJ solutions and further incubated as MDH at 47°C. Like MDH, LDH alone did not reactivate from the urea-heat denaturation, and no spontaneous refolding was observed even upon addition of KJE or ATP after the stress (not shown). Nor did urea-heat denatured LDH in the presence of IbpB (IbpB-LDH) refold after the stress in the presence of ATP alone, or with supplemented KJ or LS chaperones alone (not shown). Only in the presence of KJE or  Fig. 2A, and 4 h of refolding at 25°C in the presence of ATP, and both KJE and LS chaperones as in Fig. 2A. B, relative apparent rates of IbpB-bound MDH refolding as in Fig. 2A (Fig. 5). While native complexes of IbpB exceed 2 ϫ 10 6 daltons 2 and consequently eluted near the void volume of the column (6 -7.5 ml), individual native complexes of GroEL, KJ, or MDH, independently resolved at 11-13, 13.5-15.5, and 15.5-18 ml, respectively (arrows shown instead of profiles, Fig. 5). The elution profile of [ 3 H]MDH, which had been heat-denatured in the presence of IbpB alone and then incubated 1 h with ATP alone prior to injection, distributed in two peaks; one together with IbpB, and the other in the low molecular weight region corresponding to free MDH (Fig. 5,  profile 1). The elution profile of heat-denatured IbpB-bound MDH, incubated 1 h with KJ and ATP but without GrpE, showed 3 H label associated to the broad KJ fraction, at the expense of the IbpB containing and the free MDH fractions (Fig. 5, profile 2). Thus, a significant amount of the non-native  Table I), and further incubated at 47°C for 30 min (heat shock). The rate of subsequent LDH renaturation at 25°C was followed in the presence of 2 mM ATP, with or without KJ, E, L, and S chaperones. The apparent rates of LDH refolding were expressed in nanomolar of reactivated LDH, per minute.  MDH was transferred after the heat shock from IbpB and from the free state to the KJ chaperones. In contrast, when IbpB-MDH was incubated with ATP, KJ, and in addition with E ϩ L (without S), the majority of the 3 H label eluted associated to the GroEL fraction, at the expense of the IbpB-, KJ-, and free MDH-containing fractions (Fig. 5, profile 3). Thus, a significant amount of IbpB bound and free MDH, that was initially transferred to KJ, was further transferred to the GroEL.
When KJ chaperones were omitted and IbpB-MDH was incubated directly with GroEL and ATP, less 3 H label co-eluted with the GroEL fraction, and correspondingly, more MDH eluted with the IbpB-containing and free MDH fractions (Fig.  5, profile 4). When S and ATP were subsequently added, MDH activity was fully recovered from the GroEL fractions from profile 3 (ϩKJE ϩ L) but not from profile 4 (ϩL) (data not shown), showing that direct binding and IbpB-released MDH to GroEL is non-productive.
When, in a control experiment, [ 3 H]MDH (0.35 M) was heat-denatured in the presence of an equimolar excess all three types of chaperones together (IbpB, KJ, and L, 10 M each), gel filtration showed that over 90% of the 3 H label co-eluted with GroEL (not shown). When GroEL was omitted during the denaturation, most of the 3 H label co-eluted with KJ. Only when GroEL and KJ were both omitted during denaturation was a majority (60%) of the 3 H label co-eluted with the IbpB (Fig. 5,  track 1). This suggests that changes of affinity can define the sequence of transfer of folding intermediates between the various components of a chaperone network.

DISCUSSION
Whereas small Hsps share with other molecular chaperones the ability to recognize, bind, and prevent the aggregation of non-native proteins under stress, the classification of small Hsps as true chaperones remains an open question because of their poor performance at specifically promoting the correct refolding of the sHsps-stabilized proteins. A first indication that sHsps may collaborate with other chaperones in the refolding of bound denatured proteins came from the observation that the refolding of heat-denatured citrate synthase bound to mammalian Hsp25, can be activated by Hsp70, even without ATP (4).
We observed here that similarly to other small Hsps from eukaryotes (3-5, 10, 16), IbpB from E. coli can stabilize denatured proteins such as citrate synthase, ␣-glucosidase, and MDH in a soluble non-aggregated form. 2 Remarkably, whereas IbpB alone did not promote the active refolding of bound proteins after the stress, IbpB-bound MDH or IbpB-LDH were specifically reactivated by KJE chaperones and ATP, but not directly by LS chaperonins. Hence, the apparent rate of refolding of IbpB-MDH was 50 times higher in the presence of KJE ϩ LS chaperones and ATP, then without KJE ϩ LS and the yield of the reaction was about one recovered MDH for 10 IbpB subunits present during denaturation. Similarly, ATP-and KJE-assisted refolding of IbpB-LDH was 70 -90 times faster than without ATP or KJE.
The kinetic and gel filtration data strongly suggested that MDH, but not necessarily LDH or glucose-6-phosphate dehydrogenase, was refolded in a sequential, multichaperone reaction. When native MDH was heat denatured in the presence of IbpB alone, an IbpB⅐MDH complex was formed which was in equilibrium after the heat shock with a free inactive MDH species. The IbpB-released MDH tended to convert into a kinetically trapped inactive species, unless allowed to bind KJE, but not LS chaperones. The slow dissociation of MDH from the IbpB⅐MDH complex was the rate-limiting step. Delays in the addition of the various chaperone components showed that MDH was first transferred to KJE chaperones, where it accu-mulated, unless allowed to partially fold and then interact with LS chaperonins. In contrast with the IbpB-released species (MDH1), which was characterized by a high affinity for the KJE chaperones and a low (non-productive) affinity for LS chaperonins, the KJE-released species (MDH2) had a low affinity for IbpB and a high affinity for LS chaperonins. The increased affinity of IbpB-released MDH1 for KJE, and of KJEreleased MDH2 for LS provides a preferential pathway for the sequential MDH refolding in the multichaperone network, as follows: Binding to IbpB during the heat shock was also a prerequisite for a strict KJE-and ATP-dependent refolding of urea/ heat-denatured LDH. However, KJE-released LDH species proceeded to the native state almost as rapidly without as with LS chaperonins, suggesting that LS was optional. Interestingly, glucose-6-phosphate dehydrogenase displayed a yet different behavior. Fluorescence and light scattering indicated that heat-inactivated glucose-6-phosphate dehydrogenase was neither aggregated nor bound to IbpB or LS. Whereas denatured glucose-6-phosphate dehydrogenase did not refold spontaneously, it was specifically refolded by KJE ϩ ATP, but not by IbpB and/or by LS ϩ ATP (data not shown). Hence, IbpB can serve as a specific high capacity scavenger for aggregationprone folding intermediates, but it does not necessarily interact with all types of denatured proteins.
It should be noted that the sequential IbpB 3 KJE 3 LS mediated refolding of MDH was observed only when MDH was heat denatured in the presence of IbpB alone. When saturative amounts of LS and KJE were also present during the denaturation, IbpB displayed the lowest affinity for denatured MDH. This implies that in the cell, various protein intermediates can bind and distribute according to the relative amounts, binding capacities of each of the chaperones present, and according to the specific affinity of the folding intermediates for the chaperones present during the stress.
We propose a model that describes the various protein-folding pathways that may exist in the multichaperon network (Fig. 6). Upon denaturation, native proteins form an unstable folding intermediate that can either aggregate irreversibly, refold spontaneously (pathway 1), or distribute between several components of the chaperone network. Whereas acid-denatured barnase refolds spontaneously without chaperones (31), heat-denatured MDH can bind all three components, de-  (4), or several chaperones together. Chaperone-bound proteins can then proceed directly to the native state or, depending on the protein, transfer from IbpB to KJE (5), but not directly to LS. Transfer between KJE and LS can be unidirectional in the case of MDH (6), optional in the case of LDH and in the reverse direction (7) in the case of luciferase.
Plasmid-encoded ibpA or ibpB partially suppressed a chromosomal groEL mutation leading to phage-growth deficiency. Because LS chaperonins are involved in protein-folding, prevention of protein aggregation during stress, and correct refolding after heat shock (11), this suppression is evidence that either IbpA or IbpB can replace, yet with low efficiency, some of the cellular functions carried on by the LS chaperonin.
In eukaryotes, Hsp27, Hsp25, and ␣B-crystallines are found in protein granules of neurodegenerative diseases and after heat shock (33)(34)(35). In Drosophila, heat-induced granules dissolve after stress (36). Similarly, IbpA/B proteins were found in E. coli associated with inclusion bodies and with heat-induced intracellular proteins aggregates. IbpA/B were found to be involved in the re-solubilization of protein aggregates after stress (21). We demonstrated here in vitro that IbpB-bound proteins are stabilized in a conformation that can be subsequently released and specifically refolded by the KJE chaperones and ATP. It is tempting to speculate that stress-denatured proteins that accumulate together with IbpA/B proteins in the E. coli cell, can be solubilized and refolded, or degraded after stress, in processes involving the KJE chaperones, ATP hydrolysis, and possibly other chaperones. Moreover, in eukaryotes where heat-shock granules and amyloid plaques often contain sHSPs (37,35), a similar solubilization mechanism of sHSP-bound protein aggregates, involving Hsp70 and other ATP-hydrolyzing chaperones such as Hsp104, is therefore also conceivable as a defense mechanism against neurodegenerative and prion diseases (38).
There are major differences in the expression level of the various families of Hsps between organisms. While heat shock induces a number of distinct Hsp families in E. coli and yeast (39), Hsp70 is by far the major heat-induced protein in Drosophila (38), and small heat-shock proteins Hsp18.1 and Hsp17.7 are the major heat-induced proteins in higher plants (3,5). We find that small heat-shock proteins, while possibly binding non-native proteins with relatively low affinity, may yet serve as a primary protein-binding matrix for subsequent refolding by a highly cooperative multichaperone machinery. Despite their high binding efficiency, LS chaperonins may not be present in sufficient amounts to bind all non-native protein during cellular stress (39). Hence, auxiliary chaperone systems such as sHsps may become essential in vivo for the stress response.