DnaK, hsp73, and Their Molten Globules TWO DIFFERENT WAYS HEAT SHOCK PROTEINS RESPOND TO HEAT*

The thermal stability of bovine brain hsp73, Esche- richia coli DnaK, and its mutant T199A was studied by a combination of spectroscopic and chromatographic methods. DnaK undergoes a temperature-in- duced conformational change that leads to the formation of a molten globule at physiologically relevant temperatures (midpoint of the transition, t,, 41 “C). Native DnaK binds to a denatured form of a-lactalbu- min in a temperature-dependent manner with maximum rate at about 40 OC. The molten globule of DnaK is unable to bind denatured a-lactalbumin but recovers native structure and activity upon cooling. The half-life of the refolding process is 10 min at 35 O C . Mg/ ATP and Mg/ADP increase the thermal stability of DnaK; in the presence of these nucleotides the t , is shifted to 59 OC. Binding of Mg/ATP (but not Mg/ADP or Mg/adenosine S’-[r-thio]triphosphate) causes a conformational change in DnaK as determined by the emission fluorescence spectrum. The DnaK mutant T199A which lacks the threonine residue that is essen- tial for ATP hydrolysis and autophosphorylation activity (McCarty, J. S., and Walker, G. C. (1991) Proc. Natl. Acad. Sci.

The thermal stability of bovine brain hsp73, Escherichia coli DnaK, and its mutant T199A was studied by a combination of spectroscopic and chromatographic methods. DnaK undergoes a temperature-induced conformational change that leads to the formation of a molten globule at physiologically relevant temperatures (midpoint of the transition, t,, 41 "C). Native DnaK binds to a denatured form of a-lactalbumin in a temperature-dependent manner with maximum rate at about 40 OC. The molten globule of DnaK is unable to bind denatured a-lactalbumin but recovers native structure and activity upon cooling. The halflife of the refolding process is 10 min at 35 O C . Mg/ ATP and Mg/ADP increase the thermal stability of DnaK; in the presence of these nucleotides the t , is shifted to 59 OC. Binding of Mg/ATP (but not Mg/ADP or Mg/adenosine S'-[r-thio]triphosphate) causes a conformational change in DnaK as determined by the emission fluorescence spectrum. The DnaK mutant T199A which lacks the threonine residue that is essential for ATP hydrolysis and autophosphorylation activity (McCarty, J. S., and Walker, G. C. (1991) Proc. Natl. Acad. Sci. U. S. A. 88,9513-9517) shows nearly identical properties to the wild type in the presence or absence of nucleotides. Hsp73 undergoes similar temperature-induced transitions as determined by spectroscopic methods (Palleros, D. R., Welch, W. J., and Fink, A. L. (1991) Proc. Natl. Acad. Sci. U. S. A. 88,5719-5723); however, contrary to DnaK, the molten globule of hsp73 irreversibly aggregates at temperatures higher than its t , (42 "C).
Members of the 70-kDa family of heat shock proteins (hsp7O) encompass a group of highly conservedproteins whose presence is ubiquitous in eukaryotic and prokaryotic cells (reviewed by Georgopoulos et al., 1990;Pelham, 1990;Lindquist and Craig, 1988). hsp70 family members have been implicated in various aspects of protein maturation events such as binding to nascent polypeptide chains (Beckmann et al., 1990) and translocation of proteins into organelles (Chirico et al., 1988;Deshaies et al., 1988;Hass and Wabl 1983;Bole et al., 1986;Gething et al., 1986). They have also been proposed to function in the recognition and binding to mature intracellular proteins which become unfolded in the * This work was supported by National Institutes of Health Grant GM45316. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
! I To whom correspondence should be addressed. cell experiencing heat shock and other types of metabolic insults (Pelham, 1986).
The cytosolic forms of mammalian hsp7O are present in cells as two different gene products (Lindquist, 1986): a constitutive member, hsp73, also known as hsc70, and an inducible form, hsp72. In contrast, only one hsp7O gene product has been identified in prokaryotic cells: the DnaK protein.
Due to the high sequence identity (50%) between the cytosolic forms of hsp70 and DnaK (Bardwell and Craig, 1984;O'Malley et al., 1985), it is assumed that these proteins have similar functions. This belief has been supported, in part, by experimental evidence. For example: (i) mammalian hsp70, as well as DnaK, are nucleotide binding proteins (Welch and Feramisco, 1985) and have weak endogenous ATPase activity (Zylicz and Georgopoulos, 1983;Rothman and Schmid, 1986;Chappell et al., 1986). (ii) The cytosolic forms of mammalian hsp70, like their prokaryotic counterpart, have been shown to interact with aberrant polypeptide chains such as mutants of the cellular oncoprotein p53 (Pinhasi-Kimhi et al., 1986;Hinds et al., 1987;Sturzbecher et al., 1987;Clarke et al., 1988) and to form stable complexes with a variety of unfolded protein targets but not with their properly folded counterparts (Palleros etal., 1991;Liberek et al., 1991). (iii) DnaK possesses autophosphorylating capabilities (Zylicz et al., 1983;McCarty and Walker, 1991). In uitro experiments showed that the threonine residue at position 199 is phosphorylated upon incubation with ATP in a process that is highly temperature dependent (McCarty and Walker, 1991). Although other members of the mammalian hsp70 family have been shown to undergo autophosphorylation (Leustek et al., 1989), reports on the autophosphorylation of hsp73 are contradictory (McCarty and Walker, 1991;Flaherty et al., 1990). (iv) It has been shown in in vitro experiments that DnaK protects RNA polymerase from heat inactivation in an ATP-dependent manner. Also heat-inactivated and aggregated RNA polymerase is reactivated by incubation with DnaK in the presence of ATP (Skowyra et al., 1990). Furthermore, it has been reported that DnaK in conjunction with DnaJ and GrpE proteins assists the renaturation of denatured XcI857 mutant protein (Gaitanaris et al., 1990). hsp73 has been shown to be capable of dissociation of protein structures such as clathrin (Rothman and Schmidt, 1986;Greene and Eisenberg, 1990) and heat-treated ribosomes in the nucleus (Pelham, 1984).
Despite all these similarities, mammalian hsp73 expressed in Escherichia coli does not complement DnaK mutations'; homologous DnaK genes from other bacterial species also failed to complement E. coli DnaK mutations (Sussman and Setlow, 1987).
hsp7O-like proteins consist of two functional domains S. Sadis and L. Hightower, personal communication. (Chappell et al., 1987;Flaherty et al., 1990). The nucleotide binding domain, a 44-kDa N-terminal segment that has been highly conserved throughout evolution and a more variable C-terminal domain that has been proposed to be responsible for binding to target proteins (Chappell et al., 1987). The three-dimensional structure of the 44-kDa N-terminal domain, at 2.2 A resolution, has been published recently (Flaherty et al., 1990). Since hsp73 and DnaK have only a 25% sequence identity in the C-terminal domain (Bardwell and Craig, 1984;O'Malley et. al., 1985), differences in biochemical and physiological properties are likely to arise. Unfortunately, little is known about the biophysical properties of these proteins, particularly, those in connection with their response to heat. In a previous study we reported that bovine hsp73 undergoes a conformational transition that leads to oligomerization in the vicinity of temperatures used to induce the heat shock response. The oligomerization process was shown to be irreversible, although it can be shifted to higher temperatures if nucleotides such as Mg/ATP or Mg/ADP are present (Palleros et al., 1991).

5279
In the present study, we investigated the thermal stability of E. coli DnaK in the temperature range 20-90 "C by a combination of spectroscopic and chromatographic methods and the effects of nucleotides on DnaK conformation and stability. We also examined the mutant DnaK T199A. Threonine 199 is the residue which is autophosphorylated in vitro (McCarty and Walker, 1991) and is anticipated to be in close proximity to the y-phosphate residue of bound ATP (Flaherty et al., 1990).

EXPERIMENTAL PROCEDURES
Materials-Bovine brain hsp73 was isolated and purified as described (Welch and Feramisco, 1985). E. coli DnaK and its mutant T199A were isolated and purified by essentially the same procedure as previously described (McCarty and Walker, 1991) using a plasmid from which the DnaJ gene was deleted. Reduced carboxymethylated a-lactalbumin (RCMLA)' was obtained from Sigma. ATP (disodium salt) was from Pharmacia LKB Biotechnology Inc. ADP (monosodium salt) and adenosine 5'-[ythio]triphosphate (ATP+) (tetralithium salt) were from Calbiochem. Guanidine hydrochloride (GdnHC1) was from ICN Biochemicals. All other reagents were analytical grade. Buffer A was 20 mM sodium phosphate, 150 mM sodium chloride, and 10 mM potassium chloride, pH 7.0. Buffer B was 20 mM Tris-HC1, pH 7.5. Buffer C was 20 mM sodium phosphate and 0.20 M potassium chloride, pH 6.5. hsp73 stock solutions were in buffer A or buffer B plus 2 mM fl-mercaptoethanol; concentrations were in the range 10-20 p~, as determined by absorbance = 47800 cm" M-'; Palleros et al., 1991). DnaK and DnaK T199A stock solutions (60-300 pM) were made in 25 mM Hepes-KOH, 0.4 M potassium chloride, 1 mM EDTA, 5 mM P-mercaptoethanol, and 10% glycerol, pH 7.6. Protein concentration was determined as described (McCarty and Walker, 1991). hsp73 and DnaK stock solutions were diluted to their final concentrations (about 6 p~ for CD and gelfiltration experiments and 1.3 p M for fluorescence studies) using buffer A or B. RCMLA stock solution (180 p~) was in 0.01 M HCl. RCMLA concentration was determined by absorbance (ezm = 27200 M" cm"; Ikeguchi and Sugai, 1989). GdnHCl stock solution was 8.0 M in buffer A; the concentration was determined by refractive index (Nozaki, 1972). ATP stock solutions were: 9.6 mM in buffer B plus 10 mM MgClz, pH 6.7; 10 mM in buffer A, pH 6.7, or 5 mM in buffer A plus 5 mM MgC12, pH 6.7. ADP stock solution (10 mM) was in buffer C plus 10 mM MgC12, pH 6.8; ATPrS stock solution (10 mM) was in buffer A plus 10 mM MgC12, pH 6.2.
Methods-Far UV-circular dichroism (CD) spectra were recorded on an Aviv Associates (Lakewood, NJ) model 60DS instrument using a 1-mm path cell. Mean residue ellipticity (OMRW) was calculated as The abbreviations used are: RCMLA, reduced carboxymethylated a-lactalbumin; GdnHC1, guanidine hydrochloride; Hepes, 4-(2-hy-droxyethy1)-1-piperazineethanesulfonic acid; FPLC, fast protein liquid chromatography; HPLC, high performance liquid chromatography; ATPyS, adenosine 5'-[y-thio]triphosphate. described elsewhere (Schmid, 1989). Data were collected with a time constant of 5 s, every 1 nm and a 1.5-mm constant spectral bandwidth. The cell block was thermostated using a Neslab bath equipped with temperature programmer. For the temperature-melt experiments the temperature was increased at a constant rate of 0.33 "C/min. Fluorescence experiments were performed on a Perkin-Elmer Cetus spectrophotometer model MPF-4; the cell block was thermostated as described above.
Gel-filtration-HPLC gel-filtration was performed with a Beckman instrument using a Bio-Si1 SEC-250 silica column (600 X 7.8 mm; Bio-Rad) at 22 "C; buffer C was used as the mobile phase at a flow rate of 1 ml/min; detection was by absorbance at 215 nm. The column was calibrated using a low molecular weight gel-filtration calibration kit (Pharmacia) and sodium azide to determine the total volume. FPLC gel-filtration was carried out on a Pharmacia system using a Superose 6HR 10/30 column (Pharmacia) and buffer C as mobile phase. The flow rate was 0.5 ml/min; detection was by absorbance at 214 nm. The column was calibrated using a high molecular weight gel-filtration calibration kit from Pharmacia and sodium azide.
Refolding Kinetics-DnaK or DnaK T199A stock solutions were diluted with the desired buffer (with or without the addition of nucleotide), the solution was filtered using an HV Millipore membrane (0.45 pm) and collected directly in a 1 ml-fluorescence cuvette, equipped with stirrer and temperature probe. The cuvette was heated on a thermostated block at the desired temperature to unfold the protein (52-65 "C) for 15 min and then placed in an ice water bath for a few seconds until the temperature reached the desired final temperature (20 or 35 "C) and immediately transferred to the fluorescence cell compartment thermostated at the final temperature. The increase in intensity at 330 nm was followed with excitation at 280 nm, 6-nm excitation slit, and 10-nm emission slit; to avoid lightinduced protein oxidation, data were collected at fixed intervals, and the shutter was kept closed the remainder of the time. Typical recovery of fluorescence intensity was 90-95%. When the refolding kinetics were followed by far UV-CD, a similar technique was used; since a shorter path length cell was used (1 mm) the ice water cooling step was unnecessary. The ellipticity at 222 nm was followed at 1min intervals with a time constant of 5 s.

RESULTS
The thermal stability of DnaK and its mutant T199A was examined by far UV-CD, fluorescence, HPLC gel-filtration, and its ability to bind to unfolded proteins. Since DnaK is a nucleotide binding protein, the effects of ATP and ADP on its conformation and thermal stability were also studied. In addition, we have extended our previous studies on the thermal stability of bovine brain hsp73. Previously we reported that in the absence of nucleotides at approximately 40 "C, hsp73 undergoes a cooperative transition that leads to oligomerization (Palleros et al., 1991). In the present investigation the thermal unfolding of hsp73 was studied in a wider temperature range in order to compare with DnaK. Also, the oligomerization of hsp73 was further investigated by FPLC gel-filtration. Thermal Stability Monitored by Far UV-CD- Fig. 1 shows the far UV-CD spectra of DnaK at 20.2, 52.7, and 86.7 "C in buffer A and in 7.2 M guanidine hydrochloride at 20.4 and 90.9 "C. The spectrum for DnaK at 86.7 "C still shows the presence of substantial secondary structure. When the ellipticity at 222 nm was monitored as a function of temperature in the range 20-90 "C, it was observed that the protein undergoes a cooperative conformational change in the vicinity of 40 "C. This transition is accompanied by a relatively small change in ellipticity (a decrease of about 25%); a second and much less cooperative transition takes place in the temperature range 60-90 "C ( Fig. 2). In the presence of Mg/ATP or Mg/ADP, the stability of DnaK is substantially increased. Although the protein undergoes similar transitions, the midpoint of the first transition is shifted to 59 "C ( Fig. 2). In the presence of ATP without Mg2+, the midpoint of the first transition is at 52 "C. Similar behavior to DnaK was observed with the mutant T199A in the presence or absence of nucle-  FIG. 2. Thermal stability of E. coli DnaK followed by CD: mean residue weight ellipticity at 222 nm as a function of temperature. A, in buffer A without nucleotide; B, in buffer A with 1.0 mM Mg/ATP. Temperature was increased from 20 to 90 "C at a rate of 0.33 "C/min; data were collected every 30 s with a time constant of 5 s. Protein concentration was 5.7 p~. The increased noise in the presence of nucleotide is due to the sample's higher absorbance.
otides. Table I summarizes

ATP-induced Conformational
Change-The packing of the aromatic side groups in the protein core can be studied using tryptophan fluorescence as a probe. DnaK has a single Trp residue in position 102 (Bardwell and Craig, 1984;Lindquist and Craig, 1988). Fig. 3 shows the fluorescence emission spectra of DnaK at 21.2 and 47.0 "C and in the presence of an excess of Mg/ATP and Mg/ADP at 21.2 "C. Mg/ATP causes a blue-shift in Amax from 333 to 327 nm as well as a decrease in intensity (Fig. 3C); no change in the Amax was observed in the presence of an excess of Mg/ADP (Fig. 3B).
A similar spectrum to that observed with Mg/ADP was obtained when Mg/ATPyS (a slowly hydrolyzable ATP analogue) was used instead (not shown). The fluorescence spectrum of DnaK at 47.0 "C ( Fig. 3E) resembles that of the unfolded protein in 6.3 M GdnHCl at 48.2 "C ( Fig. 30). Similar results were obtained with the mutant DnaK T199A (not shown). ' ND, not determined.
Buffer B.

Loss of Tertiary Structure during Thermal Unfolding-
When the thermal stability of DnaK was followed by fluorescence intensity at 330 nm ( Fig. 4) a cooperative transition with nearly identical t, (42 "C, Table I) to that found by CD was observed. Further increase of temperature did not result in a second transition but rather a steady decrease in fluorescence intensity due to the increasing temperature. The addition of Mg/ATP also resulted in a shift in the t, (to 60 "C) similar to that observed by CD (see Table I and Fig. 4). hsp73, which has 2 tryptophans residues in positions 90 and 580 (O'Malley et al., 1985), showed similar thermal unfolding behavior to DnaK. Thermal Unfolding Monitored by Gel-filtration-In order to further characterize the thermally unfolded state (after the first transition) of DnaK and hsp73, we used a combination of HPLC and FPLC gel-filtration techniques. When bovine hsp73 (5.3 ~L M in buffer A) was incubated at 37 "C for 10 min and injected into the HPLC, peaks at 17.0 ml (monomer) and 14.4 ml (dimer) were detected as already reported (Palleros et al., 1991). When the incubation temperature was increased to 60 "C, only one peak at the void volume of the column (10.5 ml; molecular mass cut-off 300 kDa) was detected, indicating the formation of soluble oligomers. The nature of these aggregates was explored by FPLC using a Superose 6 column (molecular mass cut-off 5 X lo6 Da). hsp73 (5.3 PM) was incubated in buffer A at 45 "C for different periods of time (5-60 min) and injected into the FPLC. At 45 "C the formation of oligomers is a relatively slow process; after 20 min three peaks were observed monomeric hsp73 (35%), a broad distribution of oligomers with mean molecular mass -3.5 X IO6 Da (55%), and a peak at the void volume (lo%), see Fig. 5B. At longer incubation times the amount of monomeric hsp73 diminished and the molecular mass distribution of oligomers shifted toward larger molecular mass; after a 60min incubation the distribution was: 25% monomeric hsp73, 50% oligomers in the range 7 X lo5 to 5 X lo6 Da, 25% oligomers larger than 5 X lo6 Da (Fig. 5C). When the same process was studied at 60 "C more than 95% of the protein was converted into oligomers larger than 5 X lo6 Da within 20 min. Similar results were observed in buffer B.
When DnaK was incubated at 37 "C (5.8 ~L M , in buffer A or B) and injected into the HPLC two peaks ( V = 16.3 ml (monomer) and 14.0 ml (dimer)) were observed (Fig. 6A). In contrast to hsp73, when DnaK was incubated at temperatures above its first thermal transition (50-65 "C) and immediately analyzed by HPLC, no soluble aggregate peak was detected; instead a peak at V = 15.4 ml eluted along with the peak for the native protein (Fig. 6B). When the same experiment was repeated but this time the preincubated DnaK solution was allowed to equilibrate at 20 "C for different periods of time (0.5-4 h) and then injected into the HPLC, it was found that the peak at 15.4 ml decreased in intensity; this decrease was accompanied by a corresponding increase in the amount of native protein (Fig. 6, C-E). After 20 h of incubation at 20 "C, only the peak corresponding to the native protein was detected (not shown).
Kinetics of DnaK Refolding-Our HPLC studies indicate that contrary to hsp73, DnaK thermally unfolded in the vicinity of 50 "C recovers its native-like structure upon lowering the temperature. The renaturation process requires several hours to go to completion at 20 "C. Since the recovery of DnaK activity may play a vital role in the survival of cells after heat shock, we investigated the effect of the final temperature on the kinetics of refolding. The effect of Mg/ATP on such processes was also studied for DnaK and its mutant T199A. The results are shown in Table 11.
The rates of refolding are the same for the wild type and the mutant T199A. A 10-fold increase in the rate constants for the wild type and the mutant is observed when the refolding is carried out at 35 "C instead of 20 "C; Mg/ATP has no noticeable effect at either temperature. DnaK Activity Assays-DnaK activity was tested by its ability to bind unfolded proteins. RCMLA, a stable unfolded form of a-lactalbumin, has been shown previously to bind to hsp73 in a temperature-dependent manner (Palleros et al., 1991). The formation of complex between DnaK and RCMLA was monitored by gel-filtration HPLC. The complex was   detected as a well resolved peak with elution volume V = 14.6 ml (Fig. 6F). Complex formation was studied in the temperature range 20-50 "C in buffer B. Control experiments showed that neither DnaK nor RCMLA gave peaks at 14.6 ml when incubated separately under the same conditions. The complex was dissociated with Mg/ATP (data not shown). Fig. 7 shows that DnaK binding of RCMLA reaches a maximum activity at around 40 "C followed by a sharp decrease as the temperature increases to 50 "C. The amount of complex formed during a 15-min incubation was used to estimate the second order rate constant as a function of temperature in the range 20-40 "C. The Arrhenius plot yielded a linear correlation ( r = 0.997), and the calculated activation energy is 12.4 kcal/ mol. The activity of DnaK (in buffer B without nucleotide) after being preincubated at 50 "C for 15 min and allowed to recover for 20 h at 20 "C was compared with that of the untreated protein by binding to RCMLA. It was found that both protein preparations reacted to RCMLA in identical fashion as determined by amount of complex formation, thus indicating that renatured DnaK was fully active.

DISCUSSION
Our far UV-CD studies (Fig. 1) indicate that thermally unfolded DnaK at 90 "C has a significant amount of secondary Thermal Stability 5283 structure in contrast with the fully unordered state found in 7.2 M GdnHCl (the increase in negative ellipticity in GdnHCl at 90.9 "C ( Fig. ID) as compared with 20.4 "C ( Fig. 1E) is due to the temperature dependence of the ellipticity of peptide bonds (Privalov et al., 1989)). Similar results were obtained with hsp73 (data not shown). That denatured proteins would retain some features of the native state is not uncommon when denaturants such as temperature, pH, and salts are used. However, the most striking feature of our CD experiments is the fact that in the vicinity of physiologically relevant temperatures, DnaK undergoes a cooperative transition (t, = 41 "C) in the absence of nucleotides (Fig. 2 and Table I). Similar behavior was observed with hsp73 (Palleros et al.,  1991). The change in ellipticity at 222 nm associated with this transition of DnaK corresponds to only 25% of the signal of the native protein. Although the ellipticity at 222 nm is often taken to be solely a measure of secondary structure, it is well known that the packing of aromatic side chains, and therefore the tertiary structure of the protein, also contribute to the CD spectra in the far UV region (Sears and Beychok, 1973). It is clear that the first thermal transition of DnaK involves a loss of tertiary structure as shown by the fluorescence experiments at 330 nm (Fig. 4). Thus, it is not apparent whether the change in ellipticity in the far UV-CD is due to a change of secondary and/or tertiary structure. There is a very good agreement in the t , determined by both fluorescence and far UV-CD (Table I). Further increase in temperature resulted in a second and less cooperative transition as determined by far UV-CD. This transition was not detectable by fluorescence. This implies that all the tertiary structure is lost in the first transition as confirmed by the similarities in the fluorescence spectra of DnaK at temperatures above 47 "C with that in 6.3 M GdnHCl (Fig. 3).
The lack of tertiary structure, as shown by our fluorescence experiments, and the high content of secondary structure as reflected by far UV-CD, suggests that the thermally denatured state of DnaK at 50 "C possesses the properties of a molten globule (Ohgushi and Wada, 1983;Ptitsyn, 1987;Kuwajima, 1989; Christensen and Pain, 1991), i.e. a compact conformation with unordered tertiary structure. Although there have been only a few scattered reports on the formation of molten globule states by thermal denaturation Ptitsyn, 1987;Dolgikh et al., 1985;Kuwajima et al., 1986), it has been shown that these temperature-induced molten globule states are thermodynamically equivalent to those formed by acids and other chemiotropic agents. Also, regardless of the conditions in which they are formed, most molten globule states share a common characteristic feature, i.e. contrary to native proteins, they do not show cooperative thermal transitions (Ptitsyn, 1987). The CD data presented in Fig. 2 are in agreement with this behavior. It can be observed that for the second temperature-induced transition of DnaK the loss in secondary structure spans a temperature range of 30 "C, indicating lack of cooperativity.
Molten globule states are usually slightly expanded relative to the native state; the Stokes radius increasing by 10-20% over that of the native state (Gast et al., 1986;Goto et al., 1990). Our HPLC data confirmed this for the thermally unfolded state of DnaK at 50 "C. The calculation of the Stokes radius based on elution volumes according to the method developed by Roche (Corbett and Roche, 1984) gives 43 A for the molten globule state of DnaK and 38 A for native DnaK.
This represents a 13% increase, confirming that this denatured state is compact. The calculated Stokes radius for the unfolded protein in GdnHCl is 80 A.
Since most physical methods for studying the conformation of unfolded states of proteins give information at low levels of resolution, the molecular details of the architecture of the molten globule state have been the subject of speculation (Ptitsyn, 1987;Kuwajima, 1989;Christensen and Pain, 1991). However, it has been shown by NMR and other spectroscopic techniques that in these compact unfolded states the intramolecular interactions between side groups are considerably weakened although not completely eliminated (Ptitsyn, 1987). This is consistent with the idea of a fluctuating tertiary structure which allows the motion of side groups and the penetration of solvent molecules into the compact core (Dolgikh et al., 1981(Dolgikh et al., , 1985. This lack of unique tight packing of side chains usually leads to a loss of biological activity (Dolgikh et al., 1981;Mitchinson and Pain, 1985) as discussed below. Also, as a result of solvent exposure of hydrophobic residues, molten globule states have a tendency to aggregate (Goto and Fink, 1989;Cleland and Wang, 1990).
We tested DnaK "activity" by its ability to bind to a denatured form of a-lactalbumin (RCMLA). DnaK has been shown to bind to aberrant polypeptide chains such as eukaryotic p53 (Clarke et al., 1988) and carboxymethylated bovine pancreatic trypsin inhibitor (Liberek et al., 1991). RCMLA has been shown previously to form a stable 1:l complex with hsp73 which can be isolated by gel-filtration HPLC (Palleros et al., 1991). Complex formation between DnaK and RCMLA was studied in the temperature range 20-50 "C. The results presented in Fig. 7 clearly indicate that DnaK's ability to bind RCMLA mirrors its thermal stability; at 50 "C, where only the molten globule is populated, no complex was detected.
Maximum activity was found near its t, (40 "C) (where the populations of native and molten globule state should be very similar), rather than at lower temperatures (which favor the native state), due to a relatively high activation energy for complex formation. Consistent with the refolding of the molten globule, when thermally unfolded DnaK at 50 "C was allowed to "recover" at 20 "C for various periods of time and then reacted with RCMLA, an increase in its binding capability with time was observed (data not shown). After 20 h at 20 "C, renatured DnaK was as active as untreated DnaK, indicating that the formation of the molten globule state is a reversible process from which DnaK can fully recover. No interaction between molten globule and native DnaK was detected by HPLC (which should be seen as a peak at an elution volume smaller than that for the DnaK dimer), indicating that the molten globule of DnaK lacks the motif that elicits recognition by the native protein.
An alternative interpretation of the thermal unfolding data could be that of independent domain folding. The biphasic transition observed by CD and the single transition seen by fluorescence could be explained as follows. The single tryptophan of DnaK is found in the N-terminal domain, and its fluorescence thus is a probe of that domain. The transition observed around 40 "C would therefore correspond to the unfolding of the N-terminal domain. The second transition around 70 "C observed only by CD would correspond to the C-terminal domain unfolding and would be silent by fluorescence, since there is no tryptophan residue in this domain. However, this interpretation can be eliminated on the following grounds: 1) proteolysis data indicate that the C-terminal domain is the less stable (Chappell et al., 1987), but this unfolding model would require that the N-terminal domain be less stable; 2) hsp73 shows also one thermal unfolding transition by fluorescence (t, = 42 "C, Table I) yet has 2 tryptophans, one in each domain (O'Malley et al., 1985); and 3) the Stokes radius data are inconsistent with this alternative model.
The results presented here indicate that both DnaK and bovine hsp73 have similar thermal stabilities as judged by their t, under several conditions. This is not surprising considering that both proteins have an overall identity of about 50%. However, their ultimate response to heat is very different. Although hsp73 irreversibly aggregates at temperatures above 40 "C, DnaK forms a molten globule from which it can refold on cooling. In the light of DnaK behavior, the oligomerization of hsp73 can now be interpreted as a result of the aggregation of a molten globule state. Which intrinsic properties of these proteins lead to oligomerization in the case of hsp73 and the formation of a monomeric molten globule state in the case of DnaK is still an open question. However, since both proteins are only 25% identical in the 30-kDa C-terminal domain, uersus an overall identity of 50% (Bardwell and Craig, 1984;O'Malley et. al., 1985), it is likely that differences in this portion of the protein account for their disparate behaviors.
The aggregation of hsp73 does not lead to oligomers of specific size but rather to a broad distribution of molecular mass, as our FPLC data indicate (Fig. 5). The oligomerization is irreversible; furthermore, the oligomers could not be dissociated by exogenous hsp73 in the presence of Mg/ATP nor by Mg/ATP alone (Palleros et al., 1991). Conversely, the refolding kinetics of DnaK presented in Table I1 show that the protein regains its native-like structure with T~/~ =lo0 min at 20 "C. A 10-fold increase in the kinetics of refolding is observed as the temperature is raised to 35 "C (~~1~ =lo min), indicating that the recovery process is fast at physiologically relevant temperatures.
DnaK is known to bind Mg/ATP and to have ATPase and autophosphorylating activity (Zylicz et al., 1983;Zylicz and Georgopoulos, 1984). Fig. 3 shows that Mg/ATP induces a conformational change in DnaK at 20 "C. The fact that there is no significant change in the far UV-CD spectrum of DnaK at the same temperature when Mg/ATP is added (not shown) indicates that this conformational change brought about by Mg/ATP is manifested predominantly in tertiary structure changes. In the presence of Mg/ADP or Mg/ATPrS no change in the X, , , was detected. Based on proteolysis experiments, it has been reported recently (Liberek et al., 1990) that DnaK changes conformation upon ATP hydrolysis. However, the conformational change observed in our fluorescence experiments is likely to be a result of specific Mg/ATP binding rather than ATP hydrolysis, since the mutant DnaK T199A, which lacks the threonine residue that is essential for ATP hydrolysis (McCarty and Walker, 1991), was affected by Mg/ ATP in the same way as the wild type. Our results do not exclude, however, that a second conformational change takes place upon ATP hydrolysis.
The results presented in Table I1 show that the rate of DnaK refolding is not affected by Mg/ATP in the temperature range 20-35 "C, indicating that autophosphorylation and ATP hydrolysis do not play an important role in the refolding process. This is further confirmed by the fact that the mutant T199A (which lacks autophosphorylation activity; McCarty and Walker, 1991) refolds with nearly identical rate constants to the wild type in the presence or absence of Mg/ATP.
In the presence of Mg/ATP or Mg/ADP, DnaK, as well as hsp73 (Palleros et al., 1991), undergoes its first thermal transition at a 20 "C higher temperature; Mg2+ contributes, but it is not essential, to this increased stability. That the thermal stability of DnaK is not affected by autophosphorylation and ATP hydrolysis is made evident by the facts that: (i) the thermal stability of the mutant T199A is affected by Mg/ATP in the same way as the wild type and (ii) Mg/ADP is as effective as Mg/ATP in increasing DnaK thermal stability (Table I). This added stability due to the nucleotide may have important physiological consequences. At temperatures normally used to induce stress response in the cell (40-45 "C), provided the cellular levels of ATP and ADP are high enough, neither DnaK nor hsp73 would experience any major structural change. However, if the nucleotides are not available for binding due to a decrease in their total (or local) concentrations as a result of heat shock, hsp73 would irreversibly aggregate while DnaK would go to its inactive molten globule state. In this scenario, DnaK and hsp73 could be sensors that trigger the heat shock response. Unfortunately, a detailed study on nucleotide metabolism and localization during stress is lacking at the present time. Early reports showed a sharp decrease in cellular ATP levels during heat shock (Findly et al., 1983).
It has been proposed that autophosphorylation and ATPase activity, which are stimulated by temperature in the range 30-50 "C, may serve as chemical thermometers that sense the environmental temperature and would trigger the heat shock response (McCarty and Walker, 1991). It has also been suggested that DnaK senses the consequences of an increase in temperature by binding to proteins that become unfolded as a result of heat shock. The binding to substrate proteins would deplete the pool of free DnaK and would induce the heat shock response (Craig and Gross, 1991). The results presented in this paper show that DnaK and hsp73 are indeed direct sensors of the environmental temperature at low nucleotide levels. Heat shock could drastically reduce the concentration of active protein by inducing a conformational change that results in the formation of an inactive molten globule. The irreversibility of aggregate formation in the case of hsp73, in contrast to the recovery of DnaKs activity upon decreasing the temperature, raises the question whether some other member(s) of the more diverse eukaryotic hsp70 family (perhaps hsp72 or mitochondrial hsp70) fulfill the role left vacant by the absence of DnaK.